All tested compounds were obtained from Tocris (Bristol, UK) and their chemical structures are showed in supplementary ( Figure S1 ). Chemicals were dissolved and stored as 10^-2 M stock solutions in dimethyl sulfoxide (DMSO), while the final DMSO concentration never exceeded 0.1% (v/v) of the culture medium during testing.

The plasmid (GAL4RE)5-βGlob-Luc-SVNeo (Chen et al., 1995) contains luciferase gene driven by a pentamer of yeast activator GAL4 binding sites in front of β-globin promoter and a neomycin phosphotransferase gene under the control of SV40 promoter. To construct GAL4-NR chimeras, individual LBDs of human FXR (R193-Q486), LXRα (A187-E447), LXRβ (G199-E461), CAR (M76-S348), and RORγ (S97-K518) were amplified by PCR with primers containing XhoI and SacI restriction sites (FXR), XhoI and BamHI (LXRs, CAR, RORγ) and inserted in pSG5-GAL4(DBD)-puromycin plasmid. To construct GAL4-CAR (+ APYLT) (M76-S353), the five amino acids APYLT were inserted by overlap extension PCR between amino acids P270 and D271 of CAR.

HG5LN and HG5LN PXR cells have already been described (Delfosse et al., 2015). To characterize the specificity of chemicals for human FXR, LXRα, LXRβ, CAR and RORγ, we established in a similar manner HG5LN FXR, LXRα, LXRβ, CAR WT, CAR (APYLT) and RORγ reporter cell lines. HG5LN are HeLa cells stably transfected by the (GAL4RE)5-βGlob-Luc-SVNeo plasmid (Seimandi et al., 2005). These cells were stably transfected by pSG5-GAL4 (DBD)-NRs (LBD)-puromycin plasmid and selected in presence of 0.5 µg/ml puromycin. Three weeks after the initiation of the puromycin selection, maximal luciferase expression of resistant clones was measured in presence of 1 µM GW4064 (HG5LN FXR), 1 µM T0901317 (HG5LN LXRs), DMSO (HG5LN RORγ and CAR), 1 µM CITCO (HG5LN CAR + APYLT) and 0.3 µM luciferin for 24h. Two days later, minimal luminescence expression from the same individual clones was measured in presence of 1 µM DY268 (HG5LN FXR), 1 µM SR9238 (HG5LN LXRs), 10 µM SR2211 (HG5LN RORγ), 1 µM PK1195 (HG5LN CAR and CAR + APYLT), and 0.3 µM luciferin. From each HG5LN NR cells, 5 to 10 clones were chosen for their ligand inducibility of luciferase expression. We noted that the basal activity in absence of ligand strongly differed among NRs. FXR, LXRs, and CAR APYLT had the lowest basal activities, whereas CAR and RORγ had the highest. The most inducible clones were expanded and rechecked for inducibility and aliquots were frozen at different passages. One clone per receptor was maintained in culture and used for ligand screening. HG5LN cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM): Nutrient Mixture F-12 (DMEM/F-12) containing phenol red and 1 g/l glucose and supplemented with 5% fetal bovine serum, 100 units/ml of penicillin, 100 µg/ml of streptomycin and 1 mg/ml geneticin, in 5% CO₂ humidified atmosphere at 37°C. Stably transfected cell lines derived from HG5LN cells were cultured in the same culture medium supplemented with 0.5 µg/ml puromycin.

Individual reporter cell lines were seeded at a density of 40,000 cells per well in 96-well white opaque tissue culture plates (Greiner CellStar) in Dulbecco’s Modified Eagle Medium: Nutrient Mixture F-12 (DMEM/F-12) without phenol red and 1 g/l glucose and supplemented with 5% stripped fetal bovine serum, 100 units/ml of penicillin, 100 µg/ml of streptomycin (test medium). Chemicals to be tested were added 24h later, and cells were incubated at 37°C for 16h. Experiments were performed in quadruplicates. At the end of the incubation period, culture medium was replaced with test medium containing 0.3 mM luciferin solution. Luciferase activity was measured for 2s in intact living cells after 10 min stabilization using a Micro Beta Wallac luminometer (PerkinElmer). EC₅₀ values were calculated using GraphPad Prism (GraphPad Software Inc).

Chemicals tested for agonism on CAR WT and RORγ were incubated in presence of 1 µM PK11195 and 1 µM SR2211. Chemicals tested for antagonism on FXR, LXRs, and PXR were incubated in presence of 100 nM GW4064 (HG5LN FXR), 20 nM T0901317 (HG5LN LXRα), 100 nM T0901317 (HG5LN LXRβ) 200 nM SR12813 (HG5LN PXR). Chemicals tested for agonism on FXR, LXRs, CAR (+ APYLT), PXR, and antagonism on CAR and RORγ were tested alone, without addition of any other ligand.

In the transactivation assay, each compound was tested at various concentrations in three independent experiments at least. For each experiment, tests were performed in quadruplicates for each concentration, and data are expressed by means values with standard deviations. Individual agonist dose–response curves, in the absence and presence of antagonist, were fitted using the sigmoid dose-response function of a graphics and statistics software program (GraphPad Prism, version 5.0). EC₅₀ (effective concentration for half-maximal luciferase activity) and IC₅₀ (half-maximal inhibitory concentration) values were calculated from equations used to fit the data in this graphic software. Transactivation data are presented as EC₅₀ and IC₅₀ values for each compound tested. To analyze significances, we compared individual compound treatments with controls using one-way analysis of variance (ANOVA) with the help of GraphPad Prism software.

To evaluate the FXR, LXRα, LXRβ, PXR, CAR, and RORγ specificity of pharmaceutical chemicals, we established HG5LN reporter cell lines expressing human FXR, LXRα, LXRβ, CAR, CAR (+ APYLT), and RORγ. Then, we tested the agonistic/antagonistic potential of 22 commercially available pharmaceutical compounds ( Figure S1 ) on these cells and in the previously established PXR cell line (Delfosse et al., 2015). The 22 chemicals were first tested for non-specific modulation of luciferase expression on the HG5LN parental cell line, which contains the same reporter gene as HG5LN-GAL4-NRs cells, but lacks of Gal4-NRs. We found XY018, SPA70 and clotrimazole, as the most toxic representatives, while GW4064, DY268, GSK2033, WAY252623, and CITCO revealed vice versa, significant nonspecific activation ( Figure S2 ). The basal (in absence of ligand) and the maximal (in presence of NR-specific full agonist) activities of HG5LN and the different HG5LN GAL4-NRs were measured. The maximal activity of FXR, LXRs, and PXR was obtained in the presence of 1 µM GW4064 (FXR), 1 µM T0901317 (LXRs), 3 µM SR12813 (PXR), respectively. The expression of FXR induced the slight repression of basal transcription, while both LXRs and PXR had no effect, or just slightly activated the basal transcription ( Figure 1 ). Consequently, the baseline activities were normalized as percentage of maximal activity mediated by full agonists, as follows: 6% for FXR, 8% for LXRα, 9% for LXRβ, 19% for PXR ( Table 1 ). On the contrary, the basal activity was increased by the expression of CAR and RORγ. Due to the inherent constitutive activity, we determined this activity equal to 100% as the specific agonists were not able to reveal their transactivation potential alone, but they were able to reverse the inhibition caused by antagonists/reverse agonists. To characterize easily CAR agonists, we used cell line expressing a variant of CAR (CAR + APYLT) with reduced basal (Auerbach et al., 2003). In this cell line, the baseline activity (11%) was normalized as percentage of maximal activity mediated by full agonist CITCO ( Table 1 ) (Maglich et al., 2003). The difference in baseline activities of different NRs potentially results from receptor-specific recruitment of endogenous corepressors and/or coactivators presented in their apo forms. Subsequently, we tested all selected pharmaceutical compounds on our stable HG5LN GAL4-NRs reporter cell lines.

FXR (NR1H4) was characterized as a nuclear receptor weakly activated by farnesol metabolites and by primary bile acids (Forman et al., 1995; Makishima et al., 1999). FXR also responds to lipophilic signaling molecules, including endocrine hormones, vitamins, xenobiotics, and dietary lipids. Three FXR ligands, two agonists (GW4064, feraxamine), and one antagonist (DY268) were tested, in HG5LN GAL4-FXR cells ( Figures 2A, B ). The full agonist GW4064 (Merk et al., 2019) acted as the most potent and efficacious FXR agonist with half-maximal effective concentration (EC₅₀) equal to 41 nM ( Table 1 ), which was in accordance with Zhang and co-workers (2015) obtained on HEK293T cells. Contrary to GW4064, feraxamine recently described as partial agonist (Merk et al., 2019) did not reach maximal FXR agonistic potential with an EC₅₀ equal to 612 nM ( Figure 2A , Table 1 ). Consistent with its partial agonist profile, feraxamine inhibited agonist-induced reporter activity of FXR chimeric receptor with corresponding concentration required to produce 50% of inhibition (IC₅₀) calculated to 687 nM ( Figure 2B , Table 2 ). As some full FXR agonists decreased the level of high-density lipoprotein (HDL), while increasing the level of low-density lipoprotein (LDL), leading finally to cardiovascular-related adverse effects, a partial agonists/antagonists could be promisingly less toxic than full agonists.

Finally, DY268, trisubstituted-pyrazol carboxamide (IC₅₀ of 468.5 nM in FXR cell-based antagonistic assays published by Yu et al., 2014), behaved as a full FXR antagonist, able to inhibit FXR activity with an IC₅₀ of 148 nM ( Figure 2B , Table 2 ). Interestingly, all these compounds showed full FXR selectivity, as they did not activated/inhibited the other tested NRs ( Tables 1 and 2 ). FXR ligand specificity is the important observation, considering the beneficial effects of FXR ligands in cholestasis and hypercholesterolemia (Jonker et al., 2012; Amano et al., 2018; Keitel et al., 2019).

LXRα and LXRβ (also known as NR1H3 and NR1H2) originally considered to be orphan nuclear receptors, were “adopted” following the discovery that metabolites of cholesterol bind to and activate these receptors at physiological concentrations and induce the expression of genes involved in lipid metabolism (Willy, 1995; Gill et al., 2008). Six LXR ligands, three agonists (T0901317, GW3965 and WAY252623) and three antagonists (SR9238, SR9243 and GSK2033) were tested in HG5LN GAL4-LXRs cells ( Figures 3 and 4 ). T0901317 was settled as reference 100% value of activity (1 µM) for both, LXRα and LXRβ. T0901317 revealed mild subtype specificity favoring α subtype with EC₅₀ values corresponding to 9 nM (LXRα) and 34 nM (LXRβ) ( Figures 3A and 4A , Table 1 ). Indeed, T0901317 is the agonist for multiple targets, which possesses EC₅₀ values of 20 nM for LXRα and lesser, micromolar extent for LXRβ, respectively, in HEK293 cells (Schultz et al., 2000; Kanno et al., 2013). GW3965 was also a potent full agonist that revealed 12-fold LXRβ specificity ( Figures 3A and 4A , Table 1 ). EC₅₀ values corresponded to 176 nM (LXRα) and 15 nM (LXRβ). WAY252623 activated LXRs with lower potency and almost no subtype specificity (EC₅₀ values of 595 and 321 nM for LXRα and LXRβ ( Figures 3A and 4A , Table 1 ). In accordance with our results, tertiary amine GW3965, activated a luciferase reporter gene containing LXR response elements in HEK293 cells expressing LXRα or LXRβ, while revealed nanomolar LXRβ agonistic potency (Peng et al., 2011).

LXRs antagonists were tested in presence of T0901317 (20 and 100 nM for LXRα and LXRβ respectively). The IC₅₀ of SR9238 was 26 nM for LXRα and 5 nM for LXRβ showing high LXR potency and 5-fold LXR β-subtype specificity to the detriment of α subtype ( Table 2 ). SR9243 antagonized T0901317-induced luciferase expression in both, HG5LN GAL4-LXRα and LXRβ cell lines with higher potency for LXRβ (IC₅₀ of 560 and 59 nM for LXRα and LXRβ, respectively) ( Figures 3B and 4B , Table 2 ). GSK2033 resulted as well, in high potent inhibition of LXRs with appropriate IC₅₀ values corresponding to 22 nM for LXRα and 10 nM for β subtype ( Figures 3B and 4B , Table 2 ). Remarkably, these three chemicals repressed the basal activities of LXRα and LXRβ that clearly correspond to their inverse agonistic potential, which is in accordance with previous studies (Griffett et al., 2013; Flaveny et al., 2015).

LXRs ligands appeared to be remarkably nonspecific among the tested molecules. Previous studies showed that T0901317 activated/repressed several tested receptors (Houck et al., 2004; Xue et al., 2007; Kanno et al., 2013). Indeed, it appeared to be the least selective compound ( Figure 5A ), since apart from its agonistic activity on LXRs was able to dose-dependently activate PXR (EC₅₀ of 128 nM), inhibit agonist-induced reporter activity of FXR (IC₅₀ of 5300 nM) and even repress constitutive activities of CAR (IC₅₀ of 4641 nM) and RORγ (IC₅₀ of 1418 nM) ( Figure 5A , Tables 1 and 2 ). With lower extent, GW3965 showed additional agonistic properties against PXR and even antagonized FXR ( Tables 1 and 2 ). LXR antagonists, SR9238, SR9243 and GSK2033 behaved in the similar manner, as they did not only efficiently inhibited T0901317-induced LXR activity under receptor basal level, but all of them, with SR9238 on the forefront, activated PXR and slightly repressed constitutive activity of CAR WT ( Figure 5B , Tables 1 and 2 ). We also observed substantial receptor non-specificity of GSK2033, activating PXR with EC₅₀ equal to 505 nM on the one hand ( Table 1 ), and repressing constitutive activity of CAR with IC₅₀ of 556 nM on the other ( Table 2 ). In accordance with our observations, recently published hypothesis believes in promiscuity of GSK2033, as it potentially targets several receptors (Griffett and Burris, 2016).

RORγ (NR1F3) binds and is preferentially modulated by oxysterol derivatives and not by retinoic acid (Fauber and Magnuson, 2014). SR1078 and SR0987 were identified as chemicals that induced the expression of the ROR target genes, where SR1078 showed concentration dependent induction of reporter gene expression with EC₅₀ of 800 nM, while its synthetic analogue SR0987 demonstrated better efficacy in full-length RORγt transfected cells (The correct citation is: Rene et al., 2014; Chang et al., 2016). By applying referred agonists, SR1078 and SR0987 in combination with inverse agonist SR2211 (1 μM) on our chimeric reporter cell line containing just LBD of RORγ, both SR1078 and SR0987 revealed agonistic properties with corresponding EC₅₀ of 2,091 and 3,186 nM ( Figure 6A , Table 1 ).

Due to the constitutive activity of RORγ, the antagonists SR2211, SR1001 and XY018 inhibited the constitutive activity of RORγ with IC₅₀s of 505, 1,312, and 7,385 nM for SR2211, XY018 and SR1001, respectively ( Figure 6B , Table 2 ). It is noteworthy, that administration of the inverse RORγ agonists has proven pharmaceutical potential, as SR1001 significantly reduced diabetes incidence and insulitis in a type 1 diabetes model mice (Solt et al., 2015) and SR2211 has been reported to inhibit inflammation in a collagen-induced arthritis mouse model (Xue et al., 2016). RORγ ligands appeared relatively specific among tested molecules, as only SR0987 activated PXR and CAR in the submicromolar concentrations ( Tables 1 and 2 ).

CAR (NR1I3) similarly to PXR is a key regulator of metabolism and play a major role in detoxification. The constitutive activity of CAR and its constitutive nuclear localization in cancer cell lines did not allow to test directly chemicals with agonistic activity. To reveal the agonistic activity of 22 chemicals in our HG5LN GAL4-hCAR cell line, we tested their ability to reverse the inhibitory effect of PK11195 (1 μM), one of the most potent CAR antagonist. In these conditions, only the well know CAR agonist CITCO (EC₅₀ of 49 nM in the CAR/SRC-1 FRET assay published by Maglich et al., 2003) showed strong transactivation potential with EC₅₀ of 253 nM (data not shown). As ligand-dependency of CAR is increased in a CAR variant having an insertion of five amino acids (APYLT) residues into the ligand binding domain (Auerbach et al., 2003), we generated a HG5LN cells expressing this CAR variant (HG5LN CAR + APYLT). In accordance with our expectations, CITCO fully activated our chimeric CAR+APYLT cell line with EC₅₀ equal to 380 nM ( Figure 7A , Table 1 ). It is noteworthy, that the insertion of five amino acids with consequent decrease of receptor constitutive activity, facilitates the screening of CAR agonists.

To reveal the antagonistic activity of 22 chemicals, we tested their ability to inhibit effect the basal activity in our HG5LN GAL4-CAR WT cell line. In this regard, the potency of the most efficacious reference compounds S07662, CINPA1 and PK11195 was sought in dose-response. These experiments resulted in significant suppression of receptor constitutive activity with IC₅₀s of 251, 306 and 1056 nM for CINPA1, PK11195 and S07662, respectively ( Figure 7B , Table 2 ). Except S07662, all of the CAR ligands behaved as PXR agonists, while CITCO being the most potent with an EC₅₀ for PXR of 1.3 µM, as it was recently published (Anderson et al., 2011; Lin et al., 2020).

PXR, another xenobiotic sensor among nuclear receptors, is commonly known to be activated by a very structurally diverse array of endogenous and exogenous molecules referring its expansive (~1,300 Å3), hydrophobic, roughly spherical ligand binding pocket with the flexibility to accommodate large molecules. This fact was fully supported by our experiments, where several tested pharmaceutics unveiled agonistic properties for PXR.

SR12813 referred as high affinity ligand for PXR (EC₅₀ of 0.44 μM in hPXR LBD assay performed by Shukla et al., 2011) showed strong PXR selectivity with corresponding EC₅₀ of 153 nM in our experiments. Dabrafenib (Creusot et al., manuscript in preparation) was little more potent with an EC₅₀ of 115 nM, while less potency revealed clotrimazole with EC₅₀ of 1,278 nM ( Figure 8A , Table 1 ). Plenty of drugs bind to and activate PXR while upregulating drug-metabolizing enzymes, decrease drug efficacy and increase resistance. Taking together, PXR antagonists could have therapeutic exploitation. Indeed, SPA70 was reviewed as potent and selective PXR antagonist (IC₅₀ of 510 nM in the cell-based hPXR antagonistic assay), while inhibited PXR activities in human hepatocytes or humanized mouse models and enhanced the chemosensitivity of cancer cells, consistent with the role of PXR in drug resistance (Lin et al., 2017). In our hands, SPA70 was able to inhibit agonist-induced reporter activity with an IC₅₀ of 278 nM and displayed a strong selectivity for PXR ( Figure 8B , Table 2 ).