Ordinal association between cytochrome P450 inhibition and hepatotoxicity severity in rats

Abstract

Cytochrome P450s (P450s) are essential for xenobiotic metabolism, and their inhibition is associated with chemical-induced liver toxicity. While qualitative associations between P450 inhibition and hepatotoxicity have been reported, the quantitative relationship between the degree of inhibition and the severity of hepatotoxicity remains unclear. In this study, we explored the quantitative association between P450 inhibition and hepatotoxicity using lowest observed effect levels (LOELs) from rat repeated-dose toxicity (RDT) studies on 326 chemicals. Inhibitory activities against seven rat P450 isoforms were compared between compounds positive and negative for six liver-related group endpoints (gEPs). The results revealed that inhibitory activity against CYP1A1, CYP2B1, CYP2C6, and CYP3A2 was significantly higher in compounds positive for hepatocellular hypertrophy or dyslipidemia than in negative compounds. Although regression analyses did not show clear linear relationships between P450 inhibition and LOELs, nonparametric trend tests revealed modest monotonic associations, with increased P450 inhibition corresponding to lower LOELs. To identify structural factors influencing inhibition among highly toxic compounds, we compared molecular descriptors between those exhibiting strong or weak P450 inhibition. Descriptors related to aqueous solubility, Verhaar baseline toxicity, and structural complexity were consistently higher in the weak-inhibition group across multiple P450–gEP combinations. These findings suggest that inhibition of CYP1A1, CYP2B1, CYP2C6, and CYP3A2 partly contributes to the severity of hepatocellular hypertrophy and dyslipidemia, whereas highly toxic compounds with low P450 inhibition may exert their toxicity through P450-independent mechanisms.

INTRODUCTION

Cytochrome P450s (P450s) are highly expressed in the liver and are involved in the metabolism of drugs and other chemical substances (Guengerich, 2015). While P450s play key roles in detoxifying xenobiotics, they can also produce reactive intermediates during metabolism (Rendic and Guengerich, 2024). These intermediates can covalently bind to biomolecules such as proteins and DNA, leading to oxidative stress and cellular damage, which may ultimately result in cell necrosis. The P450-mediated formation of reactive metabolites is one of the mechanisms underlying drug-induced liver injury (DILI) (Allison et al., 2023; Feng and He, 2013, Guengerich, 2015; Yu et al., 2014). Accordingly, during drug development, the formation of reactive metabolites, especially those mediated by P450s, is routinely evaluated for drug candidates (Food and Drug Administration, 2020a, 2020b).

In our previous studies using pharmaceuticals, we have demonstrated that inhibition of human CYP1A1 and CYP1B1 is associated with DILI (Kaito et al., 2024; Shimizu et al., 2021), and that most drugs that strongly inhibit CYP1A1 or CYP1B1 cause DILI (Shimizu et al., 2021). In addition, in a recent analysis of rat repeated-dose toxicity (RDT) data for pesticides, we showed that inhibition of rat CYP1A1 and CYP2C6 is related to liver hypertrophy and dyslipidemia in rats (Shibata et al., 2025). Furthermore, analysis of rat RDT data for industrial chemicals combined with P450 inhibition data revealed that inhibition of rat CYP2B1 and CYP2C6 is associated with liver hypertrophy (Watanabe et al., 2020). Regarding CYP1A1, it has also been demonstrated that aryl hydrocarbon receptor (AHR) activation triggered by CYP1A1 inhibition contributes to hepatotoxicity (Yoda et al., 2022). Collectively, these findings indicate that P450 inhibition is associated with chemical-induced hepatotoxicity and suggest that this association is mechanism-based.

However, these findings have so far only demonstrated qualitative associations between P450 inhibition and hepatotoxicity, and the quantitative relationship between the extent of P450 inhibition and the severity of hepatotoxicity remains unknown. Understanding this quantitative relationship would help develop models to predict hepatotoxic risk based on P450 inhibition data. Therefore, in this study, we analyzed the quantitative relationship between P450 inhibition data and the lowest observed effect levels (LOELs) for hepatotoxicity in RDT studies of chemical compounds.

MATERIALS AND METHODS

Dataset

We used the same dataset of 326 compounds as in our previous study (Ambe et al., 2024). Compound IDs (SXXX) were assigned to these 326 compounds. More detailed information on these compounds is provided in Supplementary Table S1.

P450 inhibition assay

We used the P450 inhibition data obtained in our previous study (Ambe et al., 2024). Briefly, inhibitory activity (%) against rat CYP1A1, CYP1A2, CYP2B1, CYP2C6, CYP2D1, CYP2E1, and CYP3A2 was determined using luminescent substrates (Promega, Madison, WI, USA) and recombinant proteins (Supersomes; Corning, Corning, NY, USA) as enzyme sources.

Rat RDT study data

Rat RDT data for the test compounds were obtained from the HESS database (National Institute of Technology and Evaluation, Tokyo, Japan) (Sakuratani et al., 2013; Takeshita et al., 2024). The LOELs for hepatic findings were extracted from the database. Similar findings were grouped, and six liver-related group endpoints (gEPs) were defined (Supplementary Table S2). For each gEP, the lowest LOEL among its findings within that gEP was assigned as the LOEL for that gEP. Details of gEP LOELs for each compound are provided in Supplementary Table S3. When any LOEL was reported for any individual findings within a given gEP, the compound was considered positive for that gEP; otherwise, it was considered negative. For some analyses, LOEL values were classified into three toxicity classes: Class I (<300 mg/kg/day), Class II (300–1000 mg/kg/day), and Class III (>1000 mg/kg/day or no reported LOEL).

Molecular descriptors

Two-dimensional (2D) molecular descriptors were calculated using alvaDesc (ver. 2.0.1.2, Alvascience, Lecco, Italy) based on the SMILES information obtained in our previous study (Ambe et al., 2024). In this study, we focused on four descriptor groups: constitutional indices, molecular properties, ring descriptors, and functional group counts, using a total of 266 descriptors. Details of these descriptors are provided in Supplementary Table S4. Descriptors exhibiting constant values across test compounds or containing missing values were excluded, resulting in 203 descriptors.

Statistical analyses

All statistical analyses were performed using Python (version 3.12.3) in Jupyter Notebook. The libraries used were pandas (2.2.3), numpy (2.2.3), scipy (1.15.2), matplotlib (3.10.1), seaborn (0.13.2), statsmodels (0.14.4), and scikit-learn (1.6.1).

Generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used Grammarly (version 1.134.1.0) in order to check and improve English grammar. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the publication's content.

RESULTS

Results of P450 inhibition assay

The inhibitory activities of 326 test compounds against seven rat P450 isoforms were measured at three concentrations. For each compound, the highest inhibition value (%) across the concentrations was defined as the maximum inhibitory activity (MAX). In addition, the sum of the inhibition values at all three concentrations was also calculated and defined as SUM. All MAX and SUM data are provided in Supplementary Table S5. As shown in Fig. 1, inhibitory activities against CYP1A1, CYP1A2, CYP2B1, and CYP2C6 tended to be higher than those against the other three isoforms, with this trend being more pronounced for SUM than for MAX.

Fig. 1

Box plots of P450 inhibition data of the test compounds. P450-inhibitory activities (MAX and SUM) of the 326 compounds at three concentrations are presented using box plots. These plots show the median (lines within boxes), interquartile range (IQR; boxes), whiskers (1.5 × IQR), and outliers (dots).

Association between P450 inhibition and hepatotoxicity

To investigate differences in P450-inhibitory activities between compounds with and without hepatotoxicity (positive vs. negative for each gEP), the Wilcoxon rank-sum test was performed (Table 1). Significant differences (p < 0.05) in inhibitory activities against CYP1A1, CYP2B1, CYP2C6, and CYP3A2 were observed for both LV05 and LV06. When comparing MAX and SUM as indices of P450 inhibition, MAX consistently showed smaller p-values than SUM for all P450 isoforms, indicating greater sensitivity in detecting differences. Therefore, MAX was used as the indicator of P450 inhibition data in the subsequent analyses.

Table 1. Associations between P450 inhibition and gEPs.

Quantitative relationships between gEP LOEL values and P450 inhibition data (MAX) were then evaluated via regression analysis for the pairs with significant differences in the Wilcoxon rank-sum test (Table 1). The coefficients of determination (R²) were below 0.02 for all combinations of endpoints and P450 isoforms (Fig. 2). However, excluding compounds with LOEL values of 1000 mg/kg/d, most compounds were located in the lower-left triangular regions of the scatter plots, with few compounds showing both high LOEL values and high inhibitory activities; this pattern was more pronounced for LV05. Although regression analysis did not reveal statistically significant quantitative relationships, the scatter plot pattern suggested a potential ordinal association.

Fig. 2

Correlation analyses of P450 inhibition data with LOEL values. Scatter plots of LOEL values and P450 inhibition (MAX) for compounds positive (LOEL ≤ 1000 mg/kg/d) for LV05 or LV06 are shown. Red dashed lines represent regression lines, and R² indicates the coefficient of determination. The triangular distribution (few compounds with both high LOEL and high inhibition) suggests that strong P450 inhibition is more frequently observed among more toxic compounds (lower LOEL), whereas weak inhibitors can also be highly toxic, indicating that P450 inhibition is not the sole driver of toxicity.

Next, to examine the ordinal relationship between LOEL and P450 inhibition, LOEL values were categorized into three classes: Class I (most toxic), Class II, and Class III (least toxic). For LV05, 95, 49, and 182 compounds were assigned to Classes I, II, and III, respectively; for LV06, the numbers were 129, 50, and 147. Inhibitory activities (MAX) were then compared among these classes (Fig. 3). The Jonckheere–Terpstra test showed statistically significant monotonic trends (p < 0.05) in all comparisons for both LV05 and LV06. Kendall’s rank correlation coefficients (τ) ranged from −0.12 to −0.21, indicating weak but consistent negative associations between toxicity class and inhibitory activity, meaning that higher inhibition tended to be observed in more toxic classes. The strongest trend was observed for LV05 and CYP2B1 (τ = −0.21). However, for LV05 with CYP1A1, CYP2B1, or CYP2C6, and for LV06 with CYP1A1, Class I compounds exhibited a wide range of inhibitory activities (0–100%), suggesting the presence of highly toxic compounds with both minimal and strong P450 inhibition.

Fig. 3

Box plots of P450 inhibition data across toxicity classes. Box plots of P450 inhibition across toxicity classes (based on LOEL values) for LV05 and LV06. Box plots show the median (line), interquartile range (box), whiskers (1.5 × IQR), and outliers (dots). P-values from the Jonckheere–Terpstra (JT) test indicate the statistical significance of monotonic trends across classes. Kendall’s rank correlation coefficient (τ) denotes the strength and direction of the monotonic association between toxicity class order and inhibitory activity.

Differences in physicochemical properties of weak and strong P450 inhibitors among highly toxic compounds

Finally, we investigated structural features associated with variation in P450 inhibition among highly toxic compounds in toxicity Class I. Class I compounds were ranked by inhibitory activity, and the top 25% (high-inhibition group) and bottom 25% (low-inhibition group) were selected. Molecular descriptor distributions were then compared between the two groups using the Wilcoxon rank-sum test with Bonferroni correction. The resulting p-values indicate the statistical significance of differences between groups, while Cliff’s δ (ranging from −1 to 1) serves as a nonparametric effect size, reflecting the magnitude and direction of these differences: negative values indicate higher values in the low-inhibition group, whereas positive values indicate higher values in the high-inhibition group. All statistical results are presented in Supplementary Table S6.

Overall, the descriptor analysis indicates that, within toxicity Class I, strong P450 inhibitors tend to be larger and more lipophilic (with higher ring/topological and polarizability-related features), whereas weak inhibitors tend to be more water-soluble and structurally more complex.

In detail, across multiple P450–gEP combinations, several molecular descriptors consistently differed between the low- and high-inhibition groups. The low-inhibition group showed significantly higher median values for ESOL (estimated aqueous solubility; logS), BLTF96, BLTD48, and BLTA96 (Verhaar baseline toxicity descriptors derived from MLOGP for fish, Daphnia, and algae, respectively), as well as for GD (graph density, reflecting structural complexity). These descriptors were predominantly associated with negative Cliff’s δ values, indicating a tendency toward higher values in the low-inhibition group. In contrast, the high-inhibition group consistently exhibited significantly higher median values for MW (molecular weight), Sp (sum of atomic polarizabilities), ring-related descriptors (Rperim, NRS, and D/Dtr06), and multiple lipophilicity- and refractivity-related descriptors (i.e., several alternative logP estimates and molar refractivity indices: ALOGP, ALOGP2, LOGP99, LOGPcons, MLOGP, MLOGP2, AMR, MR99, and MRcons), yielding positive Cliff’s δ values. Collectively, these results indicate that, within toxicity Class I, weak and strong P450 inhibitors are differentiated by consistent shifts in physicochemical and structural descriptor profiles across multiple P450–gEP combinations.

DISCUSSION

We have previously demonstrated qualitative associations between P450 inhibition and hepatotoxicity (Shibata et al., 2025; Shimizu et al., 2021; Watanabe et al., 2020). However, these studies did not address whether P450 inhibition is quantitatively related to the severity of hepatotoxicity. Therefore, in this study, we aimed to clarify the quantitative and ordinal relationships between P450 inhibition and LOELs for hepatotoxicity. Our analyses showed that inhibitory activities against CYP1A1, CYP2B1, CYP2C6, and CYP3A2 are associated not only with the presence or absence of liver hypertrophy (LV05) and dyslipidemia (LV06), but also display ordinal associations with the severity of these effects when LOELs are considered as ordered categories.

Using the Wilcoxon rank-sum test, we first examined the associations between the inhibitory activities against seven P450 isoforms and six gEPs, finding significant differences (p < 0.05) in the inhibitory activities against CYP1A1, CYP2B1, CYP2C6, and CYP3A2 between compounds positive and negative for LV05 and LV06. We then focused on these isoform–endpoint combinations to evaluate the quantitative relationship between P450 inhibition and LOEL values. Simple linear regression analyses yielded very low coefficients of determination (R² < 0.02) for all combinations, indicating a lack of clear linear relationships. However, when treating LOEL values as ordered categories and analyzing them with trend-based nonparametric tests, P450 inhibition consistently showed a significant negative correlation with LOEL (Jonckheere–Terpstra test p-values of < 0.05; Kendall’s τ values of < −0.1). These findings indicate that chemicals with lower LOEL values (stronger hepatotoxicity) tend to exhibit higher inhibitory activities against CYP1A1, CYP2B1, CYP2C6, and CYP3A2, suggesting that P450 inhibition is ordinally associated with liver hypertrophy and dyslipidemia.

In the analyses above, inhibitory activity was summarized using the MAX value across the three tested concentrations. We also evaluated SUM (the simple sum across the three concentrations) to account for concentration-dependent inhibition patterns. As MAX reflects only the highest inhibitory activity among the three concentrations, it cannot distinguish compounds that show inhibition at low concentrations from those that show inhibition only at high concentrations. Therefore, we also evaluated SUM to capture whether P450 inhibition is observed at low concentrations or becomes apparent only at high concentrations. However, regarding the presence or absence of hepatotoxicity (or LOEL-based severity), our results suggest that the magnitude of inhibitory activity under toxicity-relevant conditions is more important than whether inhibition is detectable at low concentrations. In addition, because SUM is an additive metric across the three concentrations, it tends to yield larger values for compounds exhibiting moderate inhibition at multiple concentrations and may overestimate the overall degree of inhibition, potentially diluting between-group differences. Taken together, SUM may not adequately represent inhibitory activity at the time of toxicity manifestation, and in our dataset, MAX appears to be a more appropriate representative inhibition index.

Our laboratory previously reported that hepatotoxic drugs often exhibit strong CYP1A1 inhibition (Shimizu et al., 2021). This aligns with our present observation that CYP1A1 inhibition correlates with liver hypertrophy and dyslipidemia. In a pesticide dataset, inhibition of CYP1A1 and CYP2C6 was significantly associated with hepatomegaly-related findings, such as increased liver weight and centrilobular hepatocellular hypertrophy, and dyslipidemia characterized by elevated blood cholesterol levels (Shibata et al., 2025). Another HESS-based study by our group further confirmed significant associations between inhibition of CYP2B1 and CYP2C6 and hepatomegaly-related endpoints corresponding to LV05 (Watanabe et al., 2020). Although the specific combinations of P450 isoforms showing statistical significance vary across studies, independent analyses of pharmaceuticals, pesticides, and industrial chemicals consistently indicate that inhibition of certain P450 isoforms, particularly CYP1A1, CYP2B1, and CYP2C6, is associated with liver hypertrophy and dyslipidemia. The present study extends these findings by demonstrating that P450 inhibition is not only qualitatively associated with hepatotoxicity but also quantitatively related to the severity of LV05 and LV06, as reflected in their LOEL values.

Although the precise mechanisms underlying the association between P450 inhibition and liver toxicity remain unclear, we propose that altered levels of endogenous signaling molecules caused by P450 inhibition may contribute to the hepatotoxicity investigated in this study. For example, CYP1A1 metabolizes the endogenous aryl hydrocarbon receptor (AHR) agonist 6-formylindolo[3,2-b]carbazole (FICZ) (Wincent et al., 2009). Thus, long-term inhibition of CYP1A1 could increase FICZ levels, prolong AHR signaling, and potentially induce AHR-mediated hepatotoxicity, including hepatocellular hypertrophy and lipid accumulation (Yoda et al., 2022). In addition, CYP3A enzymes are involved in bile acid metabolism; therefore, sustained reductions in CYP3A activity may alter both bile acid levels and composition (Klyushova et al, 2022). As some bile acids act as agonists of nuclear receptors such as PXR and FXR (Kliewer and Willson, 2002; Jonker et al., 2012), these changes may disrupt receptor signaling and contribute to hepatocellular hypertrophy and disturbances in lipid metabolism. Moreover, CYP2C enzymes convert arachidonic acid into signaling lipids such as epoxyeicosatrienoic acids (EETs); thus, inhibition of CYP2C enzymes may alter these lipid mediators, potentially affecting lipid homeostasis and inflammation (Zeldin, 2001; Shibata et al., 2025). Taken together, P450 inhibition may alter the levels of various endogenous signaling metabolites and thereby contribute to the onset of hepatotoxicity classified as LV05 and LV06. This hypothesis should be tested in future studies using targeted metabolomics and analyses of receptor-related gene expression.

In our dataset, CYP1A1 inhibition showed a clear association with hepatotoxicity, whereas CYP1A2 inhibition did not, suggesting that CYP1A1 and CYP1A2 play distinct roles in maintaining the homeostasis of endogenous signaling molecules. These differences likely reflect their overlapping but distinct substrate specificities (Floreani et al., 2012; Yamazoe et al., 2016; Yamazoe and Yoshinari, 2017; Yamazoe and Yoshinari, 2020). For example, in the metabolism of the AHR agonist FICZ, both CYP1A1 and CYP1A2 catalyze its hydroxylation, but CYP1A2 displays slower kinetics than CYP1A1 (Bergander et al., 2004). Although genetic knockout is not identical to chemical inhibition, studies using knockout rats suggest that the biological consequences differ between loss of CYP1A2 alone and simultaneous loss of CYP1A1 and CYP1A2. CYP1A2 knockout rats exhibit elevated serum cholesterol and free testosterone, along with mild liver injury and lipid accumulation (Sun et al., 2021). In contrast, CYP1A1/CYP1A2 double-knockout rats show more pronounced lipid and cholesterol abnormalities, including increased total cholesterol in blood and liver, accompanied by alterations in lipid regulatory pathways, such as activation of the LXRα–SREBP1–SCD1 axis (Lu et al., 2023). Moreover, in rats, the extent of in vivo CYP1A1 and CYP1A2 inhibition may differ from that observed in vitro, depending on hepatic exposure and the mode of inhibition. Although further studies are needed, these considerations may help explain why CYP1A1, but not CYP1A2, inhibition showed a clear association with hepatotoxicity in the present study.

Among highly toxic chemicals, we observed compounds exhibiting both strong and weak P450 inhibition (Fig. 1). To further characterize highly toxic compounds with weak P450 inhibition, we compared molecular descriptor distributions between the high- and low-inhibition groups within toxicity Class I using the Wilcoxon rank-sum test with Bonferroni correction. Descriptors such as ESOL, an indicator of aqueous solubility, and GD, a measure of structural complexity, showed significantly higher values in the low-inhibition group across multiple P450–gEP pairs (Supplementary Table S6). These results suggest that highly toxic compounds with low P450 inhibition tend to be more water-soluble and structurally complex. In addition, several descriptors were consistently higher in the high-inhibition group across all tested P450–gEP pairs (p < 0.05 with positive Cliff’s δ), including MW and Sp, ring- and topology-related descriptors (Rperim, NRS, and D/Dtr06), and multiple lipophilicity- and refractivity-related indices (ALOGP, ALOGP2, LOGP99, LOGPcons, MLOGP, MLOGP2, AMR, MR99, and MRcons). Collectively, these descriptive patterns indicate that, within Class I, strong inhibitors tend to be larger and more lipophilic and to exhibit more pronounced ring/topological features and higher polarizability-related properties than weak inhibitors. Considering these trends alongside the higher solubility observed for weak inhibitors, lipophilicity-related properties may represent a key factor underlying differences in P450 inhibition among highly toxic compounds. Previous studies have shown that P450 substrate-binding pockets are predominantly hydrophobic (Guengerich, 2008; Guengerich, 2015), and QSAR analyses have identified lipophilicity as an important determinant of inhibitory activity for several P450 isoforms (Lewis et al., 2007). Consistent with these findings, our results support the hypothesis that highly toxic compounds with weak P450 inhibition may bind less efficiently to hydrophobic P450 active sites and may instead elicit hepatotoxicity through mechanisms that do not primarily involve P450 inhibition.

In our dataset, the LOELs for LV05 (liver hypertrophy) and LV06 (dyslipidemia) overlapped, although their dose ranges were not identical. For compounds with LOEL values for both endpoints, the LOELs for LV05 were generally lower than or comparable to those for LV06 (Supplementary Fig. S1). This suggests that liver hypertrophy is often detectable at doses where dyslipidemia has not yet occurred, though both effects can sometimes appear at similar dose levels. However, since these observations are based on LOELs from RDT studies, they do not provide direct temporal information on when the lesions arise and therefore cannot confirm a general sequential progression from LV05 to LV06. Mechanistic studies have shown that activation of the nuclear receptors PXR and CAR can induce both hepatocellular hypertrophy and lipid metabolism alterations (Cai et al., 2021; Sonkar et al., 2024), suggesting that different pathways may produce LV05 and LV06 with distinct dose–response characteristics. Taken together, our findings suggest that LV05 is often detectable at lower or equal doses compared with LV06 in this dataset, but do not establish a universal order of occurrence. Determining the temporal and causal relationships between liver hypertrophy and dyslipidemia will require dedicated mechanistic and time-resolved in vivo or in vitro studies.

This study used the LOEL as a surrogate measure of hepatotoxicity severity. However, because LOEL values were derived from heterogeneous RDT studies, they inevitably reflect differences in study design (e.g., dose spacing and dosing duration). For example, widely spaced dose levels can cause the observed LOEL to be driven more by the selected dose interval than by the underlying toxicity threshold. In addition, the LOEL does not capture time-to-onset and may fail to detect toxicities that depend on exposure duration. Consequently, LOEL-based severity has inherent limitations in terms of both comparability across studies and temporal resolution. The monotonic associations observed in this study should therefore be interpreted in light of these sources of heterogeneity.

A further limitation is that the inhibition assays used in this study cannot distinguish between competitive (reversible) inhibition and mechanism-based inhibition (MBI), even though P450 enzymes can be inhibited via multiple mechanisms. Since MBI can cause a more sustained loss of P450 activity in vivo, it may result in more severe hepatotoxicity than competitive inhibition. Future studies incorporating time-dependent inhibition data may help to refine the relationship between P450 inhibition and the severity of LOEL-based hepatotoxicity.

In conclusion, the results of this study suggest that inhibition of P450, particularly CYP1A1, CYP2B1, CYP2C6, and CYP3A2, is at least in part quantitatively associated with chemical-induced hepatotoxicity, including liver hypertrophy and dyslipidemia. In addition, the presence of highly toxic compounds with low P450 inhibition in this dataset suggests mechanisms of hepatotoxicity that are independent of P450 inhibition. The quantitative relationships between P450 inhibition and hepatotoxicity, along with the structural features of these highly toxic compounds, may provide useful insights for developing quantitative prediction methods and for differentiating P450 inhibition-dependent from independent mechanisms of liver injury.

ACKNOWLEDGMENTS

We thank Dr. Takuomi Hosaka, Dr. Takamitsu Sasaki, Ms. Chie Nakayama, and other laboratory members for their contributions to data collection and curation, as well as suggestions for this study at University of Shizuoka.

Funding

The study was supported in part by a grant from the Japan Chemical Industry Association (JCIA) Long-range Research Initiative (LRI; #22-6-02).

Conflict of interest

The authors declare no competing interests.

Data availability

The data in this study are included in the article/supplementary materials. Contact the corresponding author(s) directly to request the underlying data.

Author contributions

Conceptualization: Nana Uchida, Kouichi Yoshinari

Formal analysis: Nana Uchida, Kouichi Yoshinari

Funding acquisition: Kouichi Yoshinari

Investigation: Nana Uchida, Minami Shibata, Akira Ooka, Ryota Shizu, Jun-ichi Takeshita, Kouichi Yoshinari

Project administration: Kouichi Yoshinari

Supervision: Kouichi Yoshinari

Visualization: Nana Uchida

Writing – original draft: Nana Uchida, Kouichi Yoshinari

Writing – review & editing: Nana Uchida, Minami Shibata, Akira Ooka, Ryota Shizu, Jun-ichi Takeshita, Kouichi Yoshinari

Ethical approval and consent to participate

Not applicable.

Patient consent for publication

Not applicable.

REFERENCES

•  Allison,  R.,  Guraka,  A.,  Shawa,  I.T.,  Tripathi,  G.,  Moritz,  W. and  Kermanizadeh,  A. (2023): Drug induced liver injury - a 2023 update. J. Toxicol. Environ. Health B Crit. Rev., 26, 442-467.
•  Ambe,  K.,  Nakamori,  M.,  Tohno,  R.,  Suzuki,  K.,  Sasaki,  T.,  Tohkin,  M. and  Yoshinari,  K. (2024): Machine Learning-Based In Silico Prediction of the Inhibitory Activity of Chemical Substances Against Rat and Human Cytochrome P450s. Chem. Res. Toxicol., 37, 1843-1850.
•  Bergander,  L.,  Wincent,  E.,  Rannug,  A.,  Foroozesh,  M.,  Alworth,  W. and  Rannug,  U. (2004): Metabolic fate of the Ah receptor ligand 6-formylindolo[3,2-b]carbazole. Chem. Biol. Interact., 149, 151-164.
•  Cai,  X.,  Young,  G.M. and  Xie,  W. (2021): The xenobiotic receptors PXR and CAR in liver physiology, an update. Biochim. Biophys. Acta Mol. Basis Dis., 1867, 166101.
•  Feng,  S. and  He,  X. (2013): Mechanism-based inhibition of CYP450: an indicator of drug-induced hepatotoxicity. Curr. Drug Metab., 14, 921-945.
• Food and Drug Administration. (2020a): In Vitro Drug Interaction Studies -Cytochrome P450 Enzyme- and Transporter-Mediated Drug Interactions Guidance for Industry. (Services, U.S.D.o.H.a.H., ed.), U.S.A.
• Food and Drug Administration. (2020b): Safety Testing of Drug Metabolites Guidance for Industry. (Services, U.S.D.o.H.a.H., ed.), Silver Spring, MD: Center for Drug Evaluation and Research.
•  Floreani,  M.,  Gabbia,  D.,  Barbierato,  M.D.,  Martin,  S. and  Palatini,  P. (2012): Differential inducing effect of benzo[a]pyrene on gene expression and enzyme activity of cytochromes P450 1A1 and 1A2 in Sprague-Dawley and Wistar rats. Drug Metab. Pharmacokinet., 27, 640-652.
•  Guengerich,  F.P. (2008): Cytochrome p450 and chemical toxicology. Chem. Res. Toxicol., 21, 70-83.
• Guengerich, F.P. (2015): Human Cytochrome P450 Enzymes. In: Cytochrome P450: Structure, Mechanism, and Biochemistry (Ortiz de Montellano, P.R., ed.), pp. 523-785, Springer International Publishing, Cham.
•  Jonker,  J.W.,  Liddle,  C. and  Downes,  M. (2012): FXR and PXR: potential therapeutic targets in cholestasis. J. Steroid Biochem. Mol. Biol., 130, 147-158.
•  Kaito,  S.,  Takeshita,  J.I.,  Iwata,  M.,  Sasaki,  T.,  Hosaka,  T.,  Shizu,  R. and  Yoshinari,  K. (2024): Utility of human cytochrome P450 inhibition data in the assessment of drug-induced liver injury. Xenobiotica, 54, 411-419.
•  Kliewer,  S.A. and  Willson,  T.M. (2002): Regulation of xenobiotic and bile acid metabolism by the nuclear pregnane X receptor. J. Lipid Res., 43, 359-364.
•  Klyushova,  L.S.,  Perepechaeva,  M.L. and  Grishanova,  A.Y. (2022): The Role of CYP3A in Health and Disease. Biomedicines, 10, 2686.
•  Lewis,  D.F.,  Lake,  B.G. and  Dickins,  M. (2007): Quantitative structure-activity relationships (QSARs) in inhibitors of various cytochromes P450: the importance of compound lipophilicity. J. Enzyme Inhib. Med. Chem., 22, 1-6.
•  Lu,  J.,  Shang,  X.,  Yao,  B.,  Sun,  D.,  Liu,  J.,  Zhang,  Y.,  Wang,  H.,  Shi,  J.,  Chen,  H.,  Shi,  T.,  Liu,  M. and  Wang,  X. (2023): The role of CYP1A1/2 in cholesterol ester accumulation provides a new perspective for the treatment of hypercholesterolemia. Acta Pharm. Sin. B, 13, 648-661.
•  Rendic,  S.P. and  Guengerich,  F.P. (2024): Formation of potentially toxic metabolites of drugs in reactions catalyzed by human drug-metabolizing enzymes. Arch. Toxicol., 98, 1581-1628.
•  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.
•  Shibata,  M.,  Uchida,  N.,  Ooka,  A.,  Shizu,  R. and  Yoshinari,  K. (2025): Cytochrome P450 inhibition correlates with hepatotoxicity of pesticides: analysis using repeated-dose toxicity data in rats. Drug Metab. Pharmacokinet., in press.
•  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.
•  Sonkar,  R.,  Ma,  H. and  Waxman,  D.J. (2024): Steatotic liver disease induced by TCPOBOP-activated hepatic constitutive androstane receptor: primary and secondary gene responses with links to disease progression. Toxicol. Sci., 200, 324-345.
•  Sun,  D.,  Lu,  J.,  Zhang,  Y.,  Liu,  J.,  Liu,  Z.,  Yao,  B.,  Guo,  Y. and  Wang,  X. (2021): Characterization of a Novel CYP1A2 Knockout Rat Model Constructed by CRISPR/Cas9. Drug Metab. Dispos., 49, 638-647.
•  Takeshita,  J.I.,  Goto,  Y.,  Yamamoto,  S.,  Sasaki,  T. and  Yoshinari,  K. (2024): Comprehensive analysis of the toxicity-related findings from repeated-dose subacute toxicity studies of industrial chemicals in male rats. Crit. Rev. Toxicol., 54, 996-1010.
•  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.
•  Wincent,  E.,  Amini,  N.,  Luecke,  S.,  Glatt,  H.,  Bergman,  J.,  Crescenzi,  C.,  Rannug,  A. and  Rannug,  U. (2009): The suggested physiologic aryl hydrocarbon receptor activator and cytochrome P4501 substrate 6-formylindolo[3,2-b]carbazole is present in humans. J. Biol. Chem., 284, 2690-2696.
•  Yamazoe,  Y.,  Ito,  K.,  Yamamura,  Y.,  Iwama,  R. and  Yoshinari,  K. (2016): Prediction of regioselectivity and preferred order of metabolisms on CYP1A2-mediated reactions. Part 1. Focusing on polycyclic arenes and the related chemicals. Drug Metab. Pharmacokinet., 31, 363-384.
•  Yamazoe,  Y. and  Yoshinari,  K. (2017): Prediction of regioselectivity and preferred order of metabolisms on CYP1A2-mediated reactions. Part 2: solving substrate interactions of CYP1A2 with non-PAH substrates on the template system. Drug Metab. Pharmacokinet., 32, 229-247.
•  Yamazoe,  Y. and  Yoshinari,  K. (2020): Prediction of regioselectivity and preferred order of CYP1A1-mediated metabolism: solving the interaction of human and rat CYP1A1 forms with ligands on the template system. Drug Metab. Pharmacokinet., 35, 165-185.
•  Yoda,  T.,  Tochitani,  T.,  Usui,  T.,  Kouchi,  M.,  Inada,  H.,  Hosaka,  T.,  Kanno,  Y.,  Miyawaki,  I. and  Yoshinari,  K. (2022): Involvement of the CYP1A1 inhibition-mediated activation of aryl hydrocarbon receptor in drug-induced hepatotoxicity. J. Toxicol. Sci., 47, 359-373.
•  Yu,  K.,  Geng,  X.,  Chen,  M.,  Zhang,  J.,  Wang,  B.,  Ilic,  K. and  Tong,  W. (2014): High daily dose and being a substrate of cytochrome P450 enzymes are two important predictors of drug-induced liver injury. Drug Metab. Dispos., 42, 744-750.
•  Zeldin,  D.C. (2001): Epoxygenase pathways of arachidonic acid metabolism. J. Biol. Chem., 276, 36059-36062.