Association between in vitro nuclear receptor-activating profiles of chemical compounds and their in vivo hepatotoxicity in rats

Abstract

The liver plays critical roles to maintain homeostasis of living organisms and is also a major target organ of chemical toxicity. Meanwhile, nuclear receptors (NRs) are known to regulate major liver functions and also as a critical target for hepatotoxic compounds. In this study, we established mammalian one-hybrid assay systems for five rat-derived NRs, namely PXR, PPARα, LXRα, FXR and RXRα, and evaluated a total of 326 compounds for their NR-activating profiles. Then, we assessed the association between their NR-activating profile and hepatotoxic endpoints in repeated-dose toxicity data of male rats from Hazard Evaluation Support System. In the in vitro cell-based assays, 68, 38, 20, 17 and 17 compounds were identified as positives for PXR, PPARα, LXRα, FXR and RXRα, respectively. The association analyses demonstrated that the PXR-positive compounds showed high frequency of endpoints related to liver hypertrophy, such as centrilobular hepatocellular hypertrophy, suggesting that PXR activation is involved in chemical-induced liver hypertrophy in rats. It is intriguing to note that the PXR-positive compounds also showed statistically significant associations with both prolonged activated partial thromboplastin time and prolonged prothrombin time, suggesting a possible involvement of PXR in the regulation of blood clotting factors. Collectively, our approach may be useful for discovering new functions of NRs as well as understanding the complex mechanism for hepatotoxicity caused by chemical compounds.

INTRODUCTION

The safety of chemical compounds has traditionally been evaluated using animal testing. Especially, since the repeated-dose toxicity (RDT) tests are designed to generate data for systemic toxicity in whole body, the mechanism of toxicity observed is extremely complicated to be fully understood. Moreover, many compounds, as it stands, have not been studied for safety because it is time-consuming and costly to assess individually. Therefore, in order to improve human safety and promote animal welfare, it is desired to develop efficient and reliable alternatives to animal testing based on the mechanism of action of compounds. In recent years, for animal-free hazard and risk assessment of chemical compounds, the Organization for Economic Co-operation and Development (OECD) has been actively working on the development of adverse outcome pathways (AOPs) (Villeneuve et al., 2014). The AOP consists of three major components: molecular initiating event (MIE) in which a chemical compound acts on target molecule, following key events (KEs) that occur across different levels of biological organization and an adverse outcome (AO), and is a conceptual pathway that organizes the causal relationships between a MIE, the subsequent series of KEs and the final AO, based on the information available at present (Ankley et al., 2010).

Nuclear receptors (NRs), a family of ligand-dependent transcription factors, regulate the expression of target gene networks pivotal for various biological processes such as metabolism, reproduction and development, in response to exogenous and endogenous small lipophilic ligands (Evans and Mangelsdorf, 2014; Francis et al., 2003; Moore et al., 2006; Woods et al., 2007). In other words, xenobiotics that have ligand activity for NRs can alter their functions, and as a result, may provoke undesirable effects on human health. Therefore, the activation of NRs is expected to be a useful MIE in the AOP concept.

The liver is the primary organ for maintaining homeostasis of living organisms, playing crucial roles in metabolism, detoxification, immunity and so on, but due to its unique functional characteristics, it is also a major target organ for toxicity caused by chemical compounds (Gu and Manautou, 2012). Like others, in the liver, NRs are responsible for major liver functions, by up- and down-regulating the gene expression (Francis et al., 2003; Woods et al., 2007). Thus, NRs have also been known as critical targets for hepatotoxic compounds.

Among NRs in the liver, pregnane X receptor (PXR) has a distinctly large ligand-binding pocket, so that it enables to accept wide-ranged compounds as ligand (Kliewer et al., 2002). Along with constitutive androstane receptor (CAR), PXR plays pivotal roles in drug metabolism and protection of the body from environmental compounds by up-regulating drug-metabolizing enzymes and drug transporters. Occasionally, their up-regulation by PXR triggers drug-drug interactions. Moreover, since some enzymes also inactivate endobiotics such as steroids, PXR activation by xenobiotics may interfere with the endocrine system. Furthermore, it is well known that PXR activation induces hypertrophic phenotypes in the liver of rodents, but the precise mechanism is still unknown (Yoshinari, 2019). Recent studies have revealed that the diverse roles of PXR in lipid and carbohydrate metabolism, cell proliferation, inflammatory, immune responses and so on (Mackowiak et al., 2018; Yoshinari, 2019). In particular, chemical activation of PXR has been reported to increase lipogenesis, resulting in a fatty liver phenotype in mice.

Peroxisome proliferator-activated receptor (PPAR) α, along with two other isoforms δ and γ, acts as a fatty acid (FA) sensor to regulate energy combustion, wherein it plays pivotal roles in the regulation of lipid and carbohydrate metabolism (Pawlak et al., 2015). In particular, PPARα is a primary regulator for FA oxidation and ketogenesis in the liver. Additionally, it is found to be involved in the regulation of inflammatory responses. PPARα also has a large ligand binding pocket and accommodates a variety of endogenous and exogenous lipophilic compounds. Due to these characteristics, application of PPARα agonists to pharmaceuticals has been expected (Kersten and Stienstra, 2017). For example, fibrates such as bezafibrate (BZA) and fenofibrate are used for the treatment of hyperlipidemia. On the other hand, like PXR, PPARα activation is known to induce liver hypertrophy in rodents (Corton et al., 2018). Moreover, especially, chronic administration of PPARα agonists is known to induce liver cancer in rodents but not in humans.

Liver X receptor (LXR) α and farnesoid X receptor (FXR) work closely together to control important liver functions. LXRα acts as a cholesterol sensor that regulates the expression of transcription factors, metabolic enzymes and transporters involved in cholesterol/bile acid (BA) metabolism to promote cholesterol catabolism and BA synthesis, and oxysterol has been identified as its endogenous ligand (Wang and Tontonoz, 2018). LXRα is also responsible for FA synthesis, wherein the receptor upregulates the sterol regulatory element-binding protein (SREBP)-1c pathway, a master regulator of lipogenic gene expression. Meanwhile, FXR acts as a BA sensor and controls entero-hepatic circulation (Stofan and Guo, 2020). Importantly, as opposed to LXRα, FXR represses the rate-limiting enzyme in the pathway to convert cholesterol to BAs. In addition, FXR interferes with LXRα-mediated activation of the SREBP-1c pathway, resulting in decreased FA synthesis. Moreover, both LXRα and FXR are known to be involved in carbohydrate metabolism (Commerford et al., 2007; Ma et al., 2006; Ploton et al., 2018). So far, it has been found that especially in liver steatosis, FXR activation has beneficial effects, whereas LXRα activation has adverse ones (Tanaka et al., 2017).

Retinoid X receptor (RXR) α is a NR that responds to retinoic acid derivatives as ligands. RXRα, along with other two isoforms β and γ, regulates transcription of the target genes by acting either as a heterodimer with its partner NRs or as its own homodimer (Dawson and Xia, 2012). In fact, most NR members classified as non-steroid receptors form heterodimer with RXRα, and some of the heterodimers, including PPARα/RXRα, LXRα/RXRα and FXR/RXRα, are also activated by the RXRα agonists alone. So far, these NRs have been shown to act as permissive RXRα partners. Therefore, RXRα plays critical roles in regulating hepatic functions such as cholesterol, lipid and glucose metabolism, detoxification and is involved in hepatic homeostasis (Evans and Mangelsdorf, 2014). Although RXRα agonists have been expected as therapeutic agents for metabolic syndrome for many years, they have been shown to have unwanted side effects, such as dyslipidemia (de Vries-van der Weij et al., 2009; Pinaire and Reifel-Miller, 2007; Standeven et al., 2001).

As mentioned above, NRs play pivotal roles in the regulation of hepatic functions in response to both endogenous and exogenous ligands (Francis et al., 2003; Woods et al., 2007). NRs have thus become attractive targets for drug discovery (Gronemeyer et al., 2004), whereas they can inadvertently be critical targets for the toxicity caused by industrial chemicals and their unintended by-products (Diamanti-Kandarakis et al., 2009). For examples, in rodent models for toxicity test, phthalates used as plasticizers have been shown to induce hepatotoxicity by activating PPARα signals (Feige et al., 2010; Lapinskas et al., 2005). Bisphenol A and its analogs widely used as plasticizers, stabilizers and antioxidants for plastics are concerned about endocrine disrupting effects via activating estrogen receptor α/β (Delfosse et al., 2012). Recent studies have also revealed that they interact with both human PXR and CAR (Kojima et al., 2019; Sui et al., 2012). Hence, assessing the NR-chemical interactions would provide important information for elucidating the mechanisms of their adverse actions in the living organisms.

Here, we assessed the association between the rat-derived NR-activating profiles of compounds and their hepatotoxic endpoints observed in in vivo RDT tests of male rats in order to obtain useful information that led to the discovery of new functions of NRs as well as the understanding of the complex mechanism for hepatotoxic endpoints caused by chemical compounds.

MATERIALS AND METHODS

Chemicals

Dimethyl sulfoxide (DMSO), deionized distilled water (DDW), ethanol (EtOH) and the PPARα agonist BZA was purchased from FUJIFILM Wako Pure Chemical (Osaka, Japan). PXR agonist pregnenolone 16α-carbonitrile (PCN) was purchased from Sigma-Aldrich (St. Louis, MO, USA). The LXRα agonist GW3965 and RXRα agonist bexarotene (BEX) were obtained from Cayman Chemical (Ann Arbor, MI, USA). The FXR agonist GW4064 was purchased from ChemScene (Monmouth Junction, NJ, USA). All five NR agonists were dissolved in DMSO at appropriate concentrations. Test compounds (Table 1) were purchased from Angene Chemical (Namjing, China), Combi-Blocks (San Diego, CA, USA), Fluorochem (Derbyshire, UK), FUJIFILM Wako Pure Chemical, MP Biomedicals (Santa Ana, CA, USA), Nacalai Tesque (Kyoto, Japan), Sigma-Aldrich or Tokyo Chemical Industry (Tokyo, Japan). The test compounds were dissolved in DMSO at 100 mM except the followings: S152 and S204 were dissolved in DMSO at 50 mM. Due to poor solubility in DMSO, some test chemicals were dissolved in EtOH at 100 mM (S011, S012, S019, S038, S112, S120, S125, S130, S145, S151, S158, S159, S254 and S276) or DDW at 10 mM (S055, S134, S195). All the test compounds were successively diluted to the desired concentrations with culture medium immediately before use.

Table 1. List of test compounds and their NR-activating profiles.

Plasmids

pFN26A (Promega, Madison, WI, USA) was used to express the following yeast GAL4 DNA-binding domain (DBD) (aa 1-146) fusion proteins: rat PXR ligand-binding domain (LBD) (aa 136-431), rat PPARα LBD (aa 195-468), rat LXRα LBD (aa 201-445), rat FXR LBD (aa 238-469) and rat RXRα LBD (aa 230-467). Each expression plasmid was constructed by inserting a cDNA fragment corresponding to each NR LBD into the SgfI and PmeI sites of pFN26A. For GAL4-rat RXRα, the cDNA encoding the fusion protein was transferred from pFN26A into pcDNA3.1-VH (Thermo Fisher, Waltham, MA, USA). pTargeT, pGL4.35 and pGL4.74 were purchased from Promega.

Cell culture

Human hepatocarcinoma HepG2 cells were obtained from RIKEN BioResource Center (Ibaraki, Japan). The cells were maintained in minimum essential medium (MEM; Sigma-Aldrich) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Cytiva, Marlborough, MA, USA), minimum nonessential amino acids (FUJIFILM Wako Pure Chemical) and antibiotics-antimycotic (Thermo Fisher Scientific) in a humidified atmosphere of 5% CO₂ at 37°C.

Reporter assay

Trypsinized cells were co-transfected with the appropriate combination and amount of plasmids using Lipofectamine 3000 (Life Technologies, Carlsbad, CA, USA), and then were seeded in 96-well plates (3.2 x 10⁴ cells/well). For GAL4-rat PXR, GAL4-rat PPARα, GAL4-rat LXRα, GAL4-rat FXR and GAL4-rat RXRα, transfection mixture per well contained 20 ng of pGL4.35, 20 ng of pFN26A and 60 ng of pTargeT. For GAL4-rat RXRα, transfection mixture per well contained 20 ng of pGL4.31, 20 ng of pcDNA3.1-VH and 40 ng of pTargeT. In addition, 20 ng of pGL4.74 for Renilla luciferase was included in the transfection for GAL4-rat RXRα as a control. Twenty-four hours after transfection, the culture medium was replaced with FBS-free-MEM containing various chemicals, and then the cells were cultured for another 24 hr. All chemicals tested were dissolved in DMSO or EtOH. Normally, the final concentration of vehicle in the medium was 0.1% (v/v) for DMSO. For test compounds dissolved at 50 mM in DMSO, the medium contained a final concertation of 0.2% (v/v) DMSO. On the other hand, in the case of test compounds prepared with EtOH, the medium contained EtOH at final concentration of 0.1% (v/v) was used. Since one of the main objectives of this study was to evaluate whether test compounds have NR agonist activity, their treatment concentrations were set at a minimum value of 10 µM, which is often employed, and a maximum value of 100 μM, which was set based on the cytotoxicity and solubility limitation. In each assay, all the test compounds were evaluated in triplicate at each concentration (10, 30 and 100 μM). Each of assay plates also included both vehicle and positive controls. Luciferase reporter activities were determined using the Dual-Luciferase Reporter Assay System (Promega) with GlowMax Navigator (Promega). Firefly luciferase activity was normalized to Renilla luciferase activity. Relative luciferase activity was calculated by taking the activity of the cells treated with vehicle as one.

In vivo hepatotoxicity of test compounds

Test compounds were selected from Hazard Evaluation Support System (HESS) (Sakuratani et al., 2013) and their results of around 4- to 6-week RDT tests of male rats were extracted from HESS. The endpoints used in this study is listed in supplementary Table S2. Each endpoint of each test compound was binarized as “positive” and “negative”. If the lowest-observed-effect level (LOEL) of a test compound was determined for a certain endpoint, the compound was judged as positive for the endpoint, and if not, the combination of the compound and endpoint was considered “negative”.

Data analyses

For assay performance evaluation of in vitro cell-based reporter assays, the values of signal/background ratio (S/B), coefficient of variation (CV) and Z’ factor were calculated for each plate using the following equations (1) to (3):

In the equations, PC, NC and SD represent positive control, negative control (vehicle) and standard deviation, respectively.

The values of threshold for each NR in the reporter assays were determined by adding three times the maximal SD value of the vehicle controls of the corresponding NR assays to the value of vehicle control, 1, and rounding up to the nearest whole number.

The frequency of “positives” in each phenotype was compared between all the test compounds and the positive compounds of each NR or among each of clusters. Statistical analyses were performed using R version 4.0.5. Fisher’s exact tests for Tables 2 and 4 were performed using R function “fisher.test()”. For clustering, each test compound was given a score of 1-5 based on the fold-induction values for each NR (Table 3A), and then subjected to non-hierarchical cluster analysis based on those values. K-means clustering was performed using R function “Hartigan-Wong” as the default algorithm.

Table 2. The association between in vitro rat-derived NR-activating profiles of test compounds and their hepatotoxic endpoints in RDT tests of male rats.

Table 3. Clustering of all the test compounds based on their potential to activate each of the five NRs.

Table 4. The analysis using the clustering of test compounds based on their NR-activating profiles.

RESULTS

Establishment of mammalian one-hybrid assay systems for rat-derived NRs

We newly constructed the mammalian one-hybrid NR reporter assay systems in human hepatocarcinoma HepG2 cells designed to detect the agonist activity of the test compounds on rat-derived NRs, including PXR, PPARα, LXRα, FXR and RXRα. The one-hybrid NR reporter assay systems simply consist of the chimeric receptor in which the LBD of each NR is fused to the C-terminus of yeast GAL4 DBD and the luciferase reporter gene carrying the GAL4 response sequence. Upon ligand binding, the chimeric NR binds to and activates the GAL4-targeted luciferase reporter gene. The assay system with chimeric NRs were termed as the M1H-NR system.

Using these systems, we first evaluated the five M1H-NR systems for their responses to typical agonists: PCN for PXR, BZA for PPARα, GW3965 for LXRα, GW4064 for FXR, BEX for RXRα (supplementary Fig. S1). Each of chimeric receptors responded to its own typical agonist in a dose-dependent manner with very low background activity and high inducible activity. The agonist activity of PCN, BZA, GW3965 and GW4064 on their targeting NRs exhibited >100-fold over the vehicle control at the concentrations of 10, 100, 3 and 2 μM, respectively. BEX produced around 70-fold activation of RXRα over the vehicle control at 10 nM. Based on these results, we decided to use these agonists as positive controls and their fold-induction values as evaluation indexes for each NR.

Evaluation of agonist activity of 326 test compounds toward the rat-derived NRs

We evaluated the agonist activity of the 326 test compounds toward the five NRs at fixed concentrations (10, 30 and 100 μM), by using the M1H-NR systems. As described in Materials and Methods, each of assay plates included both vehicle and positive controls. For each receptor, the assay performance statistics in the evaluation of the test compounds are shown in supplementary Table S1. The average of Z’ factor was around 0.8 for all of the five M1H-NR systems. These results indicate that all assays were considered to have been performed with sufficient precision.

Table 1 summarizes the agonist activity of the 326 test compounds toward the five NRs. Of the 326 test compounds, 317 compounds were appropriately evaluated, but 9 compounds were excluded from the results since their Renilla luciferase/firefly luciferase activity ratio was less than 0.4 to the vehicle control in either analysis, which was an index of cytotoxicity determined in this study. To evaluate and classify the NR agonist activity of the test compounds, we adopted the fold-induction values over vehicle control. Therefore, as described in Materials and Methods, the threshold values were set based on the SD values of the corresponding vehicle controls for each NR, and they are shown in supplementary Table S1.

As expected, the number of test compounds that activated PXR-mediated transcriptional activity was largest among the five NRs (Table 1). A total of 68 test compounds met the threshold of 3-fold induction or more for PXR-positive, among which the fold-induction of 13 compounds exceeded 100, and S040 and S320 exhibited >200-fold induction. However, none of the test compounds activated PXR more strongly than the positive control PCN.

Regarding PPARα, a total of 38 compounds were found to meet the threshold of 3-fold induction or more and judged as positive (Table 1). The PPARα-positives included four known PPARα agonists: S298, S301, S305 and S325, which, as expected, activated PPARα moderately or potently (Kersten and Stienstra, 2017). They also included four non-steroidal anti-inflammatory drugs, of which S300 and S309 exhibited moderate agonist activities. Moreover, S204 activated PPARα at 10 μM up to about the half of activation by BZA. However, S204 showed clear cytotoxicity against HepG2 cells at concentrations above 30 μM. Furthermore, two environmental compounds S039 and S267, which have been known as PPARα agonists, were confirmed to activate PPARα weakly but clearly (Nakane et al., 2012; Wolf et al., 2008). Notably, S232 was found to strongly activate PPARα about 120-fold at 10 μM, although it had cytotoxic effects on HepG2 cells at the concentrations above 30 μM.

Contrary to expectations, none of the test compounds exhibited strong agonist activity on either LXRα or FXR. A total of 20 test compounds activated LXRα more than 4-fold to be judged as positive and the maximum induction was about 8-fold, whereas a total of 17 test compounds activated FXR more than 3-fold to be judged as positive, of which two compounds, S049 and S244, exhibited approximately 20-fold induction (Table 1). As for S244, its maximal activity was about 14% relative to the endogenous FXR agonist chenodeoxycholic acid at 100 μM (data not shown) (Wang et al., 1999). Thus, S244 was presumed to have a weak but significant agonist activity.

Lastly, for RXRα, a total of 17 compounds exhibited more than 3-fold activation and were judged as positive (Table 1). Among the RXRα-positives, S175, S254 and S325 exhibited stronger agonist activities than the others. Among them, S325 was recently reported as an agonist of human RXRα (Pollinger et al., 2019). Notably, the RXRα activation levels of S175 and S254 at 100 μM were almost comparable to BEX (10 nM).

The association between in vitro rat NR-activating profiles and their hepatotoxicity in vivo in rats

As the NR-activating profiles of the 317 compounds were obtained by the M1H-NR systems, we assessed the association between the NR activation and the in vivo endpoints known to be associated with hepatotoxicity observed in RDT tests of male rats. Particularly, we focused on the endpoints related to liver hypertrophy and changes in hepatic lipid and carbohydrate metabolism. The endpoints assessed in this study are listed in supplementary Table S2.

In this study, all the endpoints were binarized as positive and negative based on the LOEL availability as described in Materials and Methods. Then, the frequency of “positive” in each endpoint was compared between the frequency of “positive” in each NR assay for all the test compounds.

First, we took up four representative endpoints related to liver hypertrophy, increased absolute liver weight, increased relative liver weight, enlarged liver and centrilobular hepatocellular hypertrophy (Table 2). The frequencies of the liver hypertrophy-related endpoints, except for enlarged liver, were significantly higher in the PXR-positive groups than in all the test compounds. In addition, those in both the PPARα- and RXRα-positive groups tended to be higher. On the other hand, the tendency was not clear in either LXRα- or FXR-positive groups.

Next, we focused on the results of blood biochemical tests that are highly relevant to liver functions. Interestingly, the frequency of prolonged activated partial thromboplastin time (APTT) was significantly higher in the PXR-positive group than in all the test compounds (Table 2). In addition, for prolonged prothrombin time (PT), which is also related to blood coagulation disorders, the frequency in the PXR-positive group was higher than in all the test compounds even though it was without statistical significance. Of the PXR-positive compounds, 12 were positive for both prolonged APTT and prolonged PT. Moreover, among them, S039 was also positive for hepatocyte necrosis in centrilobular, while S082 and S229 were positive for elevated alanine aminotransferase (ALT) levels. In the LXRα- and RXRα-positive groups, the frequency of prolonged APTT and/or prolonged PT tended to be high. Meanwhile, there was a significant difference in the frequency of elevated alkaline phosphatase (ALP) levels both in the PPARα- and LXRα-positive groups compared to all the test compounds. In addition, a significant difference was also observed in the decreased total protein levels in the LXRα-positive group.

Besides, from a histopathological point of view, the endpoints of both bile duct proliferation and hepatocyte mitosis were significantly more frequent in the LXRα-positive group than the others (Table 2). When looking at other in vivo endpoints, three out of five of the relevant LXRα-positives, S188, S189 and S318, were positive for both the endpoints, and elevated levels of ALT, total bilirubin and γ-GTP were also observed in all of these three.

The association analysis using the clustering of test compounds based on their NR-activating profiles

The test compounds were found to have agonist activity toward multiple NRs evaluated. Therefore, we classified them into seven clusters based on their potential to activate the five NRs, and compared the frequency of hepatotoxic endpoints between each cluster. The cluster classification of each test compound is listed in Table 1. As shown in Table 3A, each test compound was given a score of 1-5 based on the fold-induction values for each NR (i.e. “NR activating score”; 1, negative; 2-5, positive), and then the compounds were subjected to the K-means clustering. Briefly, for the NR activating score, the threshold values of each NR were used to discriminate between negative and positive, and the fold-induction values of the test compounds were used as an index for scoring in order to take into account their biological activity relative to those of positive control compounds. Table 3B shows the results of clustering, which includes the number of test compounds in each cluster and the average of NR activating scores on each NR of all the test compounds in each cluster. For example, the test compounds included in the cluster 1 showed no agonist activity for almost all the NRs. Meanwhile, for the clusters 2 and 3, the average of NR activating scores of the test compounds on PXR was 2.3 and 4.7, respectively, indicating that the cluster 3 included the compounds with stronger agonist activity on PXR compared to the cluster 2. Thus, the frequency of hepatotoxic endpoints in each cluster was compared with that of the cluster 1 (Table 4).

Regarding the four endpoints related to liver hypertrophy mentioned above, in the clusters 2, 3 and 5, which included the test compounds with high agonist activity for PXR or PPARα, the frequency of at least one of the four endpoints was significantly higher than that in the cluster 1. Notably, the frequencies of all the four endpoints in the cluster 3 were significantly higher than those in the cluster 1, and even higher than those in the cluster 2.

For the results of blood biochemical tests, the frequency of prolonged APTT was significantly higher in the clusters 2 and 3 than in the cluster 1, and the frequency was significantly higher in the cluster 3 (52.6%) than in the cluster 2 (20%). Moreover, for the cluster 3, the frequency of prolonged PT was also significantly higher than that of the cluster 1. On the other hand, in the cluster 5 with high NR activating scores for PPARα, the frequency of reduced triglyceride (TG) levels was significantly higher than that in the cluster 1. In the cluster 5, the frequency of elevated ALP levels was also higher than that in the cluster 1 although no statistical significance was observed. Moreover, in the cluster 4, a significantly high frequency was observed for reduced total cholesterol levels even though the number of the test compounds included was limited. Incidentally, as compared to the cluster 1, the frequency of bile duct proliferation was higher in the cluster 6, in which two of the relevant compounds, S205 and S308, were also positive for toxic endpoints, such as elevated levels of ALT and γ-GTP and reduced levels of albumin.

DISCUSSION

The liver plays critical roles to maintain homeostasis of living organisms and is also a major target for chemical toxicity (Gu and Manautou, 2012). Therefore, understanding of the mechanisms of their action is quite important for the development of new toxicity tests. In the liver, NRs have pivotal roles in the regulation of major liver functions (Francis et al., 2003; Woods et al., 2007). Given that NRs function in response to xenobiotics as well as endobiotics, NRs have attracted attention as critical targets for hepatotoxic compounds. Hence, in this study, in order to elucidate the complex toxicological mechanisms, we focused on the five NRs mainly expressed in the liver, PXR, PPARα, LXRα, FXR and RXRα, and investigated the association between the in vitro NR-activating profiles of test compounds and their in vivo hepatotoxic endpoints in RDT data of male rats. Our association analyses clearly show that among the five NRs, PXR can be more closely involved in the hypertrophic changes in rat liver caused by xenobiotic exposure. Moreover, we first report that compounds that strongly activate PXR often have anticoagulant properties, providing new insights into a role of PXR in the regulation of blood clotting factors. In recent years, as the OECD promotes the development of AOPs (Villeneuve et al., 2014), the activation of NRs by chemical compounds have been attracting attention as a useful MIE in the AOP concept (Ankley et al., 2010). Thus, our findings can provide useful information for the development of AOPs related to chemical-induced hepatotoxicity.

Liver hypertrophy characterized by an increase in liver weight associated with hepatocellular hypertrophy is the most observed change in animal testing, being considered as an adaptive change as long as liver homeostasis is maintained (Hall et al., 2012; Maronpot et al., 2010). Although the detailed mechanism remains unclear, it is known that induction of hepatic drug-metabolizing enzymes (DMEs) is often observed prior to liver hypertrophy. Their induction mainly occurs via activation of NRs, such as CAR, PXR and PPARα (Waxman, 1999). In this study, the frequency of the endpoints related to liver hypertrophy, namely increased absolute liver weight, increased relative liver weight, enlarged liver and centrilobular hepatocellular hypertrophy, was significantly higher in the PXR-positive compounds than in all the test compounds. Moreover, it was found that in all the four endpoints evaluated, their frequency increased as the extent of PXR activation increased. On the other hand, for PPARα, which is also involved in DME induction, a significant difference was observed in a limited endpoint, specifically enlarged liver, only in the cluster 5 that contained the strong PPARα-positive compounds. These results suggest that among the DMEs, the group of enzymes up-regulated mainly by PXR might be more associated with the development of liver hypertrophy than those induced mainly by PPARα. For future studies, it will be of interest to assess the association between the DME-inducing profiles of these compounds and in vivo endpoints related to liver hypertrophy.

Intriguingly, compared to all test compounds, the frequencies of both prolonged APTT and prolonged PT were higher in the PXR-positive group, in which 12 compounds were positive for both the phenotypes related to blood coagulation disorders. Of these, 9 compounds were negative for endpoints related to hepatopathy. Moreover, in the cluster 3, which had the highest average NR activating score against PXR, the frequencies of both prolonged APTT and prolonged PT were significantly higher than the cluster 1. These results suggest that PXR-activating compounds affect the two cascades of blood coagulation, both the intrinsic pathway and the extrinsic pathway (Smith et al., 2015). As a possible interaction between the PXR signaling and the regulation of blood coagulation, PXR activation was suggested to be involved in the regulation of blood clotting factors, such as prothrombin, fibrinogen, factors V and X, which are involved in the common pathway downstream of the two pathways. Incidentally, blood coagulation has been known to be closely linked to inflammation (Esmon, 2005). During inflammation, blood coagulation system is activated and anticoagulation system is suppressed, while coagulation factors also function as inflammatory mediators. As PXR is known to have anti-inflammatory function (Okamura et al., 2020), it may be associated with unknown anticoagulant effects of PXR-activating compounds.

Meanwhile, the frequencies of bile duct proliferation and hepatocyte mitosis were significantly higher in the LXRα-positive compounds than in all the test compounds. The three relevant compounds, N-nitrosodiethylamine (S188), N-nitrosodimethylamine (S189) and methapyrilene hydrochloride (S318) exhibited agonist activity only against LXRα among the NRs evaluated in this study. However, given that these three compounds are known to induce liver injury, the observed bile duct proliferation and hepatocyte mitosis might have occurred to repair from the damage and have no relevance to LXR activation (George et al., 2019; Uehara et al., 2008).

PPARα has been an attractive target for drug discovery because of its functions in hepatic energy metabolism, but it has also been reported to be responsive to various environmental compounds such as diheptyl phthalate (S039 in this study) and perfluorooctanoic acid (Nakane et al., 2012; Wolf et al., 2008). In this study, 2,4,6-tribromophenol (S232) and 2,4-diamino-6-phenyl-1,3,5-triazine (S233) were newly identified as PPARα agonists. In particular, the relative agonist activity of S232 (10 μM) reached approximately 80% of that of BZA (100 μM). Recently, S232 was reported to directly bind to human PPARγ LBD (Fang et al., 2015). The endpoints observed in rats, such as hepatic hypertrophy, suppression of fatty change and lowered blood TG levels caused by them, are presumed to be due to PPARγ activation in addition to PPARα activation.

To our best knowledge, there are few reports that comprehensively evaluated the agonist activity of industrial compounds on RXRα. In this study, 17 compounds were found to be rat RXRα agonists. Among them, 2,6-di-tert-butyl-4-ethylphenol (S132), 4-sec-butyl-2,6-di-tert-butylphenol (S175) and 1,3,5-tri-tert-butylbenzene (S254) are structurally different from typical RXRα agonists such as BEX and 9-cis retinoic acid (de Almeida and Conda-Sheridan, 2019), have a low molecular weight and contain two or more alkyl groups on a benzene ring. Both S175 and S254 activated RXRα at high concentrations to a same extent as BEX. According to rat RDT data from HESS, all these three were reported to induce centrilobular hepatocyte hypertrophy. Moreover, phenolic antioxidants S132 and S175 were reported to elevate blood total cholesterol levels, while S254 was reported to reduce both blood glucose and TG levels. They might alter the function of permissive RXRα partners, e.g., PPARs, LXRα and FXR, resulting in changes in lipid and glucose metabolism (Boergesen et al., 2012; Lalloyer et al., 2009; Mukherjee et al., 1997; Repa et al., 2000). Therefore, the activation of RXRα by these compounds is of interest from a toxicological point of view.

As tools for drug development, screening systems with high specificity for substances with high affinity are desired. On the other hand, fairly sensitive assay systems capable of detecting weakly active compounds are more reliable for human and/or environmental risk assessment purposes. Our M1H-NR systems showed much higher inducibility (i.e. fold-induction by agonists) than conventional systems in which NRs act as a heterodimer with RXRα. For example, in the case of PPARα, the fold-induction values jumped from about 3-fold in the conventional types to about 170-fold in our system (data not shown) (Abe et al., 2017). One of the reasons why the M1H-NR systems showed much higher fold-induction values is that the GAL4-targeted Luc2P reporter protein that was used in this study has short half-life of approximately one hour (Paguio et al., 2010). Hence, the M1H-NR systems are considered favorable to identify weak agonist compounds. Furthermore, since all the elements other than the NR LBD are common to every system, our M1H-NR systems enable us to comprehensively compare the agonist activity of compounds among multiple NRs.

In this study, we processed data of in vitro NR-activating profiles of chemical compounds in two classes; (A) classifying test compounds as positive or negative for each NR with the threshold values based on fold-induction and (B) clustering test compounds into seven groups by scoring based on the intensity of agonist activity for each NR. And then we performed association analyses with in vivo data using both the classification. Notably, in the latter classification, mutual comparison between clusters allowed us to assess the effects of chemical compounds on living organisms depending on the agonist activity of a particular NR. For example, clusters containing test compounds with stronger PXR agonist activity were found to have a higher frequency of endpoints related to both liver hypertrophy and blood coagulation disorders.

Recently, we have reported a similar approach using the results of inhibition assays of cytochrome P450s, which are major DMEs and play critical roles in chemical toxicity (Shimizu et al., 2021; Watanabe et al., 2020). In these studies, we found specific associations between the cytochrome P450 inhibition profiles of chemical compounds and toxicity endpoints, as demonstrated in the present study. Taken together, the classification of chemical compounds based on the results of in vitro assays (i.e. biological activity) might be useful to identify a role of target proteins in the chemical toxicity and to construct in vitro toxicity evaluation systems.

Some effects of chemical compounds on living organisms appear immediately after or within a few days of exposure, while others appear only after long-term exposure. Moreover, living organisms gradually adapt to exposure to chemical compounds, and as a result, their effects may disappear. In this study, we utilized only the results of rat RDT tests of around 4- to 6-week period. Besides, some chemical compounds are rapidly converted in the liver into metabolites that are easily eliminated from the body, while others have a long half-life and are highly accumulated. In this study, we utilized a hepatocarcinoma-derived cell line, which has low metabolic activity, as a host for the M1H-NR systems. Therefore, our present results may have some limitations. First, different findings might be obtained when conducting association analysis with in vivo dataset of short-term chemical exposure for several days to a week or with those of much longer exposure tests, such as 90-day and 2-year tests. Second, for some test compounds, the effect of metabolism on NR activation may not have been properly assessed. Particularly for those that are easily metabolized, the NR-activating profiles of their metabolites need to be evaluated. Finally, the number of positives for each NR was limited. In the future, it will be necessary to assess more test compounds and their possible in vivo metabolites to accumulate data as well as to perform association analyses with different in vivo dataset for improving the reliability of the information obtained from our approach. Nonetheless, assessing the association between the in vitro NR-activating profiles of chemical compounds and their in vivo rat hepatotoxicity can be useful for discovering new functions of NRs as well as understanding the mechanisms for complex toxicity such as RDT.

ACKNOWLEDGMENTS

This study was supported in part by grants from Artificial Intelligence-based Substance Hazard Integrated Prediction System project (AI-SHIPS) of the Ministry of Economy, Trade and Industry of Japan. We thank all the project members for suggestions and discussion on the results obtained.

Conflict of interest

The authors declare that there is no conflict of interest.

REFERENCES

•  Abe,  T.,  Takahashi,  M.,  Kano,  M.,  Amaike,  Y.,  Ishii,  C.,  Maeda,  K.,  Kudoh,  Y.,  Morishita,  T.,  Hosaka,  T.,  Sasaki,  T.,  Kodama,  S.,  Matsuzawa,  A.,  Kojima,  H. and  Yoshinari,  K. (2017): Activation of nuclear receptor CAR by an environmental pollutant perfluorooctanoic acid. Arch. Toxicol., 91, 2365-2374.
•  Ankley,  G.T.,  Bennett,  R.S.,  Erickson,  R.J.,  Hoff,  D.J.,  Hornung,  M.W.,  Johnson,  R.D.,  Mount,  D.R.,  Nichols,  J.W.,  Russom,  C.L.,  Schmieder,  P.K.,  Serrrano,  J.A.,  Tietge,  J.E. and  Villeneuve,  D.L. (2010): Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk assessment. Environ. Toxicol. Chem., 29, 730-741.
•  Boergesen,  M.,  Pedersen,  T.A.,  Gross,  B.,  van Heeringen,  S.J.,  Hagenbeek,  D.,  Bindesbøll,  C.,  Caron,  S.,  Lalloyer,  F.,  Steffensen,  K.R.,  Nebb,  H.I.,  Gustafsson,  J.A.,  Stunnenberg,  H.G.,  Staels,  B. and  Mandrup,  S. (2012): Genome-wide profiling of liver X receptor, retinoid X receptor, and peroxisome proliferator-activated receptor α in mouse liver reveals extensive sharing of binding sites. Mol. Cell. Biol., 32, 852-867.
•  Commerford,  S.R.,  Vargas,  L.,  Dorfman,  S.E.,  Mitro,  N.,  Rocheford,  E.C.,  Mak,  P.A.,  Li,  X.,  Kennedy,  P.,  Mullarkey,  T.L. and  Saez,  E. (2007): Dissection of the insulin-sensitizing effect of liver X receptor ligands. Mol. Endocrinol., 21, 3002-3012.
•  Corton,  J.C.,  Peters,  J.M. and  Klaunig,  J.E. (2018): The PPARα-dependent rodent liver tumor response is not relevant to humans: addressing misconceptions. Arch. Toxicol., 92, 83-119.
•  Dawson,  M.I. and  Xia,  Z. (2012): The retinoid X receptors and their ligands. Biochim. Biophys. Acta, 1821, 21-56.
•  de Almeida,  N.R. and  Conda-Sheridan,  M. (2019): A review of the molecular design and biological activities of RXR agonists. Med. Res. Rev., 39, 1372-1397.
•  de Vries-van der Weij,  J.,  de Haan,  W.,  Hu,  L.,  Kuif,  M.,  Oei,  H.L.,  van der Hoorn,  J.W.,  Havekes,  L.M.,  Princen,  H.M.,  Romijn,  J.A.,  Smit,  J.W. and  Rensen,  P.C. (2009): Bexarotene induces dyslipidemia by increased very low-density lipoprotein production and cholesteryl ester transfer protein-mediated reduction of high-density lipoprotein. Endocrinology, 150, 2368-2375.
•  Delfosse,  V.,  Grimaldi,  M.,  Pons,  J.L.,  Boulahtouf,  A.,  le Maire,  A.,  Cavailles,  V.,  Labesse,  G.,  Bourguet,  W. and  Balaguer,  P. (2012): Structural and mechanistic insights into bisphenols action provide guidelines for risk assessment and discovery of bisphenol A substitutes. Proc. Natl. Acad. Sci. USA, 109, 14930-14935.
•  Diamanti-Kandarakis,  E.,  Bourguignon,  J.P.,  Giudice,  L.C.,  Hauser,  R.,  Prins,  G.S.,  Soto,  A.M.,  Zoeller,  R.T. and  Gore,  A.C. (2009): Endocrine-disrupting chemicals: an Endocrine Society scientific statement. Endocr. Rev., 30, 293-342.
•  Esmon,  C.T. (2005): The interactions between inflammation and coagulation. Br. J. Haematol., 131, 417-430.
•  Evans,  R.M. and  Mangelsdorf,  D.J. (2014): Nuclear receptors, RXR, and the big bang. Cell, 157, 255-266.
•  Fang,  M.,  Webster,  T.F.,  Ferguson,  P.L. and  Stapleton,  H.M. (2015): Characterizing the peroxisome proliferator-activated receptor (PPARγ) ligand binding potential of several major flame retardants, their metabolites, and chemical mixtures in house dust. Environ. Health Perspect., 123, 166-172.
•  Feige,  J.N.,  Gerber,  A.,  Casals-Casas,  C.,  Yang,  Q.,  Winkler,  C.,  Bedu,  E.,  Bueno,  M.,  Gelman,  L.,  Auwerx,  J.,  Gonzalez,  F.J. and  Desvergne,  B. (2010): The pollutant diethylhexyl phthalate regulates hepatic energy metabolism via species-specific PPARα-dependent mechanisms. Environ. Health Perspect., 118, 234-241.
•  Francis,  G.A.,  Fayard,  E.,  Picard,  F. and  Auwerx,  J. (2003): Nuclear receptors and the control of metabolism. Annu. Rev. Physiol., 65, 261-311.
•  George,  J.,  Tsuchishima,  M. and  Tsutsumi,  M. (2019): Molecular mechanisms in the pathogenesis of N-nitrosodimethylamine induced hepatic fibrosis. Cell Death Dis., 10, 18.
•  Gronemeyer,  H.,  Gustafsson,  J.Å. and  Laudet,  V. (2004): Principles for modulation of the nuclear receptor superfamily. Nat. Rev. Drug Discov., 3, 950-964.
•  Gu,  X. and  Manautou,  J.E. (2012): Molecular mechanisms underlying chemical liver injury. Expert Rev. Mol. Med., 14, e4.
•  Hall,  A.P.,  Elcombe,  C.R.,  Foster,  J.R.,  Harada,  T.,  Kaufmann,  W.,  Knippel,  A.,  Küttler,  K.,  Malarkey,  D.E.,  Maronpot,  R.R.,  Nishikawa,  A.,  Nolte,  T.,  Schulte,  A.,  Strauss,  V. and  York,  M.J. (2012): Liver hypertrophy: a review of adaptive (adverse and non-adverse) changes--conclusions from the 3rd International ESTP Expert Workshop. Toxicol. Pathol., 40, 971-994.
•  Kersten,  S. and  Stienstra,  R. (2017): The role and regulation of the peroxisome proliferator activated receptor alpha in human liver. Biochimie, 136, 75-84.
•  Kliewer,  S.A.,  Goodwin,  B. and  Willson,  T.M. (2002): The nuclear pregnane X receptor: a key regulator of xenobiotic metabolism. Endocr. Rev., 23, 687-702.
•  Kojima,  H.,  Takeuchi,  S.,  Sanoh,  S.,  Okuda,  K.,  Kitamura,  S.,  Uramaru,  N.,  Sugihara,  K. and  Yoshinari,  K. (2019): Profiling of bisphenol A and eight its analogues on transcriptional activity via human nuclear receptors. Toxicology, 413, 48-55.
•  Lalloyer,  F.,  Pedersen,  T.Å.,  Gross,  B.,  Lestavel,  S.,  Yous,  S.,  Vallez,  E.,  Gustafsson,  J.A.,  Mandrup,  S.,  Fiévet,  C.,  Staels,  B. and  Tailleux,  A. (2009): Rexinoid bexarotene modulates triglyceride but not cholesterol metabolism via gene-specific permissivity of the RXR/LXR heterodimer in the liver. Arterioscler. Thromb. Vasc. Biol., 29, 1488-1495.
•  Lapinskas,  P.J.,  Brown,  S.,  Leesnitzer,  L.M.,  Blanchard,  S.,  Swanson,  C.,  Cattley,  R.C. and  Corton,  J.C. (2005): Role of PPARα in mediating the effects of phthalates and metabolites in the liver. Toxicology, 207, 149-163.
•  Ma,  K.,  Saha,  P.K.,  Chan,  L. and  Moore,  D.D. (2006): Farnesoid X receptor is essential for normal glucose homeostasis. J. Clin. Invest., 116, 1102-1109.
•  Mackowiak,  B.,  Hodge,  J.,  Stern,  S. and  Wang,  H. (2018): The roles of xenobiotic receptors: beyond chemical disposition. Drug Metab. Dispos., 46, 1361-1371.
•  Maronpot,  R.R.,  Yoshizawa,  K.,  Nyska,  A.,  Harada,  T.,  Flake,  G.,  Mueller,  G.,  Singh,  B.,  Ward,  J.M. and  Botts,  S. (2010): Hepatic enzyme induction: histopathology. Toxicol. Pathol., 38, 776-795.
•  Moore,  J.T.,  Collins,  J.L. and  Pearce,  K.H. (2006): The nuclear receptor superfamily and drug discovery. ChemMedChem, 1, 504-523.
•  Mukherjee,  R.,  Davies,  P.J.,  Crombie,  D.L.,  Bischoff,  E.D.,  Cesario,  R.M.,  Jow,  L.,  Hamann,  L.G.,  Boehm,  M.F.,  Mondon,  C.E.,  Nadzan,  A.M.,  Paterniti,  J.R. Jr. and  Heyman,  R.A. (1997): Sensitization of diabetic and obese mice to insulin by retinoid X receptor agonists. Nature, 386, 407-410.
•  Nakane,  F.,  Kunieda,  M.,  Shimizu,  S.,  Kobayashi,  Y.,  Akane,  H.,  Akie,  Y.,  Saito,  A.,  Noguchi,  M.,  Kadota,  T. and  Mitsumori,  K. (2012): Twenty-six-week oral toxicity of diheptyl phthalate with special emphasis on its induction of liver proliferative lesions in male F344 rats. J. Toxicol. Sci., 37, 527-537.
•  Okamura,  M.,  Shizu,  R.,  Abe,  T.,  Kodama,  S.,  Hosaka,  T.,  Sasaki,  T. and  Yoshinari,  K. (2020): PXR Functionally Interacts with NF-κB and AP-1 to Downregulate the Inflammation-Induced Expression of Chemokine CXCL2 in Mice. Cells, 9, 2296.
•  Paguio,  A.,  Stecha,  P.,  Wood,  K.V. and  Fan,  F. (2010): Improved dual-luciferase reporter assays for nuclear receptors. Curr. Chem. Genomics, 4, 43-49.
•  Pawlak,  M.,  Lefebvre,  P. and  Staels,  B. (2015): Molecular mechanism of PPARα action and its impact on lipid metabolism, inflammation and fibrosis in non-alcoholic fatty liver disease. J. Hepatol., 62, 720-733.
•  Pinaire,  J.A. and  Reifel-Miller,  A. (2007): Therapeutic potential of retinoid x receptor modulators for the treatment of the metabolic syndrome. PPAR Res., 2007, 94156.
•  Ploton,  M.,  Mazuy,  C.,  Gheeraert,  C.,  Dubois,  V.,  Berthier,  A.,  Dubois-Chevalier,  J.,  Maréchal,  X.,  Bantubungi,  K.,  Diemer,  H.,  Cianférani,  S.,  Strub,  J.M.,  Helleboid-Chapman,  A.,  Eeckhoute,  J.,  Staels,  B. and  Lefebvre,  P. (2018): The nuclear bile acid receptor FXR is a PKA- and FOXA2-sensitive activator of fasting hepatic gluconeogenesis. J. Hepatol., 69, 1099-1109.
•  Pollinger,  J.,  Gellrich,  L.,  Schierle,  S.,  Kilu,  W.,  Schmidt,  J.,  Kalinowsky,  L.,  Ohrndorf,  J.,  Kaiser,  A.,  Heering,  J.,  Proschak,  E. and  Merk,  D. (2019): Tuning Nuclear Receptor Selectivity of Wy14,643 towards Selective Retinoid X Receptor Modulation. J. Med. Chem., 62, 2112-2126.
•  Repa,  J.J.,  Turley,  S.D.,  Lobaccaro,  J.A.,  Medina,  J.,  Li,  L.,  Lustig,  K.,  Shan,  B.,  Heyman,  R.A.,  Dietschy,  J.M. and  Mangelsdorf,  D.J. (2000): Regulation of absorption and ABC1-mediated efflux of cholesterol by RXR heterodimers. Science, 289, 1524-1529.
•  Sakuratani,  Y.,  Zhang,  H.Q.,  Nishikawa,  S.,  Yamazaki,  K.,  Yamada,  T.,  Yamada,  J.,  Gerova,  K.,  Chankov,  G.,  Mekenyan,  O. and  Hayashi,  M. (2013): Hazard Evaluation Support System (HESS) for predicting repeated dose toxicity using toxicological categories. SAR QSAR Environ. Res., 24, 351-363.
•  Shah,  I.,  Houck,  K.,  Judson,  R.S.,  Kavlock,  R.J.,  Martin,  M.T.,  Reif,  D.M.,  Wambaugh,  J. and  Dix,  D.J. (2011): Using nuclear receptor activity to stratify hepatocarcinogens. PLoS One, 6, e14584.
•  Shimizu,  Y.,  Sasaki,  T.,  Yonekawa,  E.,  Yamazaki,  H.,  Ogura,  R.,  Watanabe,  M.,  Hosaka,  T.,  Shizu,  R.,  Takeshita,  J.I. and  Yoshinari,  K. (2021): Association of CYP1A1 and CYP1B1 inhibition in in vitro assays with drug-induced liver injury. J. Toxicol. Sci., 46, 167-176.
•  Smith,  S.A.,  Travers,  R.J. and  Morrissey,  J.H. (2015): How it all starts: initiation of the clotting cascade. Crit. Rev. Biochem. Mol. Biol., 50, 326-336.
•  Standeven,  A.M.,  Thacher,  S.M.,  Yuan,  Y.D.,  Escobar,  M.,  Vuligonda,  V.,  Beard,  R.L. and  Chandraratna,  R.A. (2001): Retinoid X receptor agonist elevation of serum triglycerides in rats by potentiation of retinoic acid receptor agonist induction or by action as single agents. Biochem. Pharmacol., 62, 1501-1509.
•  Stofan,  M. and  Guo,  G.L. (2020): Bile Acids and FXR: Novel Targets for Liver Diseases. Front. Med. (Lausanne), 7, 544.
•  Sui,  Y.,  Ai,  N.,  Park,  S.H.,  Rios-Pilier,  J.,  Perkins,  J.T.,  Welsh,  W.J. and  Zhou,  C. (2012): Bisphenol A and its analogues activate human pregnane X receptor. Environ. Health Perspect., 120, 399-405.
•  Tanaka,  N.,  Aoyama,  T.,  Kimura,  S. and  Gonzalez,  F.J. (2017): Targeting nuclear receptors for the treatment of fatty liver disease. Pharmacol. Ther., 179, 142-157.
•  Uehara,  T.,  Kiyosawa,  N.,  Hirode,  M.,  Omura,  K.,  Shimizu,  T.,  Ono,  A.,  Mizukawa,  Y.,  Miyagishima,  T.,  Nagao,  T. and  Urushidani,  T. (2008): Gene expression profiling of methapyrilene-induced hepatotoxicity in rat. J. Toxicol. Sci., 33, 37-50.
•  Villeneuve,  D.L.,  Crump,  D.,  Garcia-Reyero,  N.,  Hecker,  M.,  Hutchinson,  T.H.,  LaLone,  C.A.,  Landesmann,  B.,  Lettieri,  T.,  Munn,  S.,  Nepelska,  M.,  Ottinger,  M.A.,  Vergauwen,  L. and  Whelan,  M. (2014): Adverse outcome pathway (AOP) development I: strategies and principles. Toxicol. Sci., 142, 312-320.
•  Wang,  B. and  Tontonoz,  P. (2018): Liver X receptors in lipid signalling and membrane homeostasis. Nat. Rev. Endocrinol., 14, 452-463.
•  Wang,  H.,  Chen,  J.,  Hollister,  K.,  Sowers,  L.C. and  Forman,  B.M. (1999): Endogenous bile acids are ligands for the nuclear receptor FXR/BAR. Mol. Cell, 3, 543-553.
•  Watanabe,  M.,  Sasaki,  T.,  Takeshita,  J.I.,  Kushida,  M.,  Shimizu,  Y.,  Oki,  H.,  Kitsunai,  Y.,  Nakayama,  H.,  Saruhashi,  H.,  Ogura,  R.,  Shizu,  R.,  Hosaka,  T. and  Yoshinari,  K. (2020): Application of cytochrome P450 reactivity on the characterization of chemical compounds and its association with repeated-dose toxicity. Toxicol. Appl. Pharmacol., 388, 114854.
•  Waxman,  D.J. (1999): P450 gene induction by structurally diverse xenochemicals: central role of nuclear receptors CAR, PXR, and PPAR. Arch. Biochem. Biophys., 369, 11-23.
•  Wolf,  C.J.,  Takacs,  M.L.,  Schmid,  J.E.,  Lau,  C. and  Abbott,  B.D. (2008): Activation of mouse and human peroxisome proliferator-activated receptor alpha by perfluoroalkyl acids of different functional groups and chain lengths. Toxicol. Sci., 106, 162-171.
•  Woods,  C.G.,  Heuvel,  J.P. and  Rusyn,  I. (2007): Genomic profiling in nuclear receptor-mediated toxicity. Toxicol. Pathol., 35, 474-494.
•  Yoshinari,  K. (2019): Role of nuclear receptors PXR and CAR in xenobiotic-induced hepatocyte proliferation and chemical carcinogenesis. Biol. Pharm. Bull., 42, 1243-1252.