Association of CYP1A1 and CYP1B1 inhibition in in vitro assays with drug-induced liver injury

Abstract

Drug-induced liver injury (DILI) is one of the major causes for the discontinuation of drug development and withdrawal of drugs from the market. Since it is known that reactive metabolite formation and being substrates or inhibitors of cytochrome P450s (P450s) are associated with DILI, we systematically investigated the association between human P450 inhibition and DILI. The inhibitory activity of 266 DILI-positive drugs (DILI drugs) and 92 DILI-negative drugs (no-DILI drugs), which were selected from Liver Toxicity Knowledge Base (US Food and Drug Administration), against 8 human P450 forms was assessed using recombinant enzymes and luminescent substrates, and the threshold values showing the highest balanced accuracy for DILI discrimination were determined for each P450 enzyme using receiver operating characteristic analyses. The results showed that among the P450s tested, CYP1A1 and CYP1B1 were inhibited by DILI drugs more than no-DILI drugs with a statistical significance. We found that 91% of drugs that showed inhibitory activity greater than the threshold values against CYP1A1 or CYP1B1 were DILI drugs. The results of internal 5-fold cross-validation confirmed the usefulness of CYP1A1 and CYP1B1 inhibition data for the threshold-based discrimination of DILI drugs. Although the contribution of these P450s to drug metabolism in the liver is considered minimal, our present findings suggest that the assessment of CYP1A1 and CYP1B1 inhibition is useful for screening DILI risk of drug candidates at the early stage of drug development.

INTRODUCTION

Drug-induced liver injury (DILI) is a major cause for the withdrawal of approved drugs from the market, restriction of their use, and discontinuation of drug development (Kullak-Ublick et al., 2017; Stevens and Baker, 2009; U.S. Food and Drug Administration, 2009; Watkins, 2005). It is known to be very difficult to reproduce DILI in animal models and its incidence in humans is extremely low. Moreover, the pathogenesis of DILI is very complex and the mechanism has not yet been fully understood. To date, the production of reactive metabolites and protein binding, mitochondrial dysfunction, triglyceride and phospholipid accumulation, and cell membrane damage have been shown to be involved in the development of DILI (Chen et al., 2016a; Cuykx et al., 2018; Goda et al., 2017; Russmann et al., 2010).

It has been reported that cytochrome P450 (P450)-generated reactive metabolites covalently bind to DNA and proteins and/or induce oxidative stress, leading to DILI development via mitochondrial dysfunction, cellular stress, immune and inflammatory responses (Chalasani and Björnsson, 2010; Oda and Yokoi, 2015). For example, it has been reported that diclofenac causes DILI through the formation of quinone imine intermediates by CYP2C9 and CYP3A4, causing changes in mitochondrial functions and other adverse effects (Feng and He, 2013). Mitochondrion is thus one of the main targets of liver injury-inducing drugs, and it has been reported that approximately 50% of the drugs that received a black box warning for hepatotoxicity cause mitochondrial dysfunction (Dragovic et al., 2016; Han et al., 2013; O’Brien et al., 2006). In addition, persistent oxidative stress conditions caused by reactive oxygen species can lead to oxidative damage to proteins, lipids and nucleic acids in living organisms and disrupt their functions, and this oxidative stress is known to cause chronic inflammation (Barnham et al., 2004; Cichoż-Lach and Michalak, 2014; Khansari et al., 2009). Based on these facts, the risk of developing DILI has been assessed by in vivo and in vitro studies related to these phenomena (Albrecht et al., 2019; Xu et al., 2008). However, their predictive accuracy is not always high and these experimental systems are complicated and cumbersome to operate, making it extremely difficult to identify compounds with a high risk of developing DILI in the early stages of drug development.

Although the mechanism of DILI pathogenesis has not been elucidated, some factors are suggested to be associated with DILI inducibility, which include a daily dose of 100 mg or more, a logP value of 3 or more, and being a substrate or inhibitor of P450s (Yu et al., 2014). P450s consist of gene superfamily containing multiple isoforms, and each form has different substrate specificity. They play important roles in the detoxification of xenobiotics including pharmaceutical drugs, while they are also involved in the development of toxicity through metabolic activation (Feng and He, 2013; Lammert et al., 2010). Since their inhibition may cause drug-drug interactions, inhibitory effects of drug candidates have been evaluated relatively early in drug development. Furthermore, we recently reported the association of the inhibition of some P450 forms with certain endpoints of rat repeated-dose toxicity tests (Watanabe et al., 2020). Based on these backgrounds, we have hypothesized that P450 inhibition data are useful for discriminating drugs with DILI risk during drug development processes. In this study, we tested this hypothesis.

MATERIALS AND METHODS

Reagents

Test drugs were purchased from Combi-Blocks (San Diego, CA, USA), FLUKA (Charlotte, NC, USA), Kyowa Medex (Tokyo, Japan), LKT Laboratories (St. Paul, MN, USA), MedChemExpress (Princeton, NJ, USA), MP Biomedicals (Santa Ana, CA, USA), Selleck Chemicals (Houston, TX, USA), Sigma-Aldrich (St. Louis, MO, USA), Tokyo Chemical Industry (Tokyo, Japan), Toronto Research Chemicals (North York, ON, Canada) or FUJIFILM Wako Pure Chemical (Osaka, Japan). Quinidine was purchased from Tokyo Chemical Industry. Sulfaphenazole, α-naphthoflavone, and flutamide were purchased from Sigma-Aldrich. Ketoconazole was purchased from LKT Laboratories. Ticlopidine was purchased from FUJIFILM Wako Pure Chemical. β-Nicotinamide-adenine dinucleotide phosphate (NADP+), D-glucose-6-phosphate, and glucose 6-phosphate dehydrogenase were purchased from Oriental Yeast (Tokyo, Japan).

Test drugs and their liver injury inducibility

Test drugs were selected from the list of DILIrank drugs (Chen et al., 2016b), excluding Ambiguous-DILI-concern drugs. ^vMost-DILI-concern and ^vLess-DILI-concern drugs were defined as DILI-positive drugs (DILI drugs) and ^vNo-DILI-concern drugs were as DILI-negative drugs (no-DILI drugs) in this study. A total of 358 test drugs (266 DILI drugs and 92 no-DILI drugs), which were commercially available and could be dissolved in dimethyl sulfoxide (DMSO) at a 100 mM, were used (Table S1).

Cytochrome P450 inhibition assays

Inhibitory effects of the test drugs at 10 µM against 8 human P450s (CYP1A1, CYP1A2, CYP1B1, CYP2B6, CYP2C9, CYP2C19, CYP2D6 and CYP3A4) were determined using the P450-Glo CYP1A1 Assay, P450-Glo CYP1A2 Induction/Inhibition assay, P450-Glo CYP1B1 Assay, P450-Glo CYP2B6 Assay, P450-Glo CYP2C9 Assay, P450-Glo CYP2C19 Assay, P450-Glo CYP2D6 Assay, and P450-Glo CYP3A4 Assay with Luciferin-IPA (Promega, Madison, WI, USA), as described previously (Watanabe et al., 2020), with minor modifications. As enzyme sources, supersomes (Corning, Corning, NY, USA) were used and a brief summary of the reactions is presented in Table S2. The final DMSO concentration in the reaction mixture was < 0.1%, and equivalent amounts of DMSO were added to the vehicle control reactions. Typical P450 inhibitors (Table S2) were used to confirm assay conditions (Watanabe et al., 2020). The test drugs were used at a constant concentration to compare their inhibitory activity under the same conditions, and 10 µM was selected because it was the highest concentration to be prepared without influence of DMSO on the enzymatic activity.

Chemical descriptors

The SMILES format data of all the compounds used in this study were obtained from PubChem database. The SMILES were converted to MOL format using OpenBabelGUI (O’Boyle et al., 2011). The chemical descriptors contained in 6 Blocks (Constitutional indices, Ring descriptors, Functional group counts, Atom-centred fragments, Molecular properties and Drug-like indices) were calculated with Dragon 7 software (Talete, Milano, Italy) with MOL format data. The descriptors that were incalculable and those that were constant among all the test drugs were excluded, and the remaining 306 descriptors were used for analyses.

Data analyses

The inhibitory activity (%) of each test drugs against each P450 was calculated by subtracting the residual activity (%) against vehicle control, which was set to 100%, from 100. The values less than 0 were rounded up to 0 and the values above 100 were rounded down to 100.

Statistical analyses were performed using Microsoft Excel, JMP Pro 14 (SAS Institute, Cary, NC, USA) and R version 3.x. The heatmap was drawn with the R function “heatmap.2()”. The calculation of Pearson’s correlation coefficient and the drawing of histograms and scatter plots were performed using R function “pairs.panels()”. Fisher’s exact tests for 2 x 10 contingency tables were performed using R function “fisher.test()” with Monte Carlo simulation. Brunner-Munzel tests (Brunner and Munzel, 2000) were performed using R function “brunner.munzel.test()” in R package “brunnermunzel”.

Receiver operating characteristic (ROC) analyses were performed using the R package ‘ROCR’. The cutoff values, showing the highest balanced accuracy (BA) of each P450 assay, were determined with the package.

The 358 test drugs were divided into the five folds, keeping the ratio of the numbers of DILI drugs and no-DILI drugs, to confirm of the dependency to the learning data of the threshold values. Thresholds for each set of the four folds (training data) were determined using ROC analyses and the rest (test data) was used to check discrimination performance. This trial was repeated five times on each training data set.

As the indices of discrimination performance, accuracy, sensitivity, specificity and BA were calculated using the following equations (1) to (4):

In the equations, TP, TN, FP, and FN represent the number of true positives, the number of true negatives, the number of false positives, and the number of false negatives, respectively.

Drug information

The drugs’ indications and dosages were obtained from SuperDRUG2 (http://cheminfo.charite.de/superdrug2/index.html).

RESULTS

Assessment of inhibitory activity of test drugs against P450s

We investigated the inhibitory effects of the 358 test drugs against 8 human P450s at a fixed concentration (10 µM), and the results are shown as a heatmap (Fig. 1). CYP1A2 and CYP2B6 were less sensitive than other P450s. Since P450 forms have diverse but overlapping substrate specificity, Pearson’s correlation analyses were performed to check their substrate specificity (Fig. 2). There were no significant associations between all the pairs of P450s except CYP1A1 vs. CYP1B1 (r = 0.82). These results suggest that the values of inhibitory activity against each P450 are available as independent parameters.

Fig. 1

The heatmap of the inhibitory activity of the test drugs. Inhibitory activity of 358 test drugs against 8 P450 forms was determined as described in Materials and Methods.

Fig. 2

Correlations among the inhibitory activity against the P450s tested. The histograms of inhibitory activity against each P450 are shown on the diagonal. Pearson correction coefficients (r) between the inhibitory activity against given 2 P450 forms were determined as described in Materials and Methods and are shown above the diagonal. The bivariate scatter plots of inhibitory activity are shown below the diagonal. The red lines represent the regression lines.

Comparison of the P450 inhibitory activity between DILI drugs and no-DILI drugs

To investigate the association between DILI and inhibition of any P450, we compared the inhibitory activity against each P450 of the test drugs between DILI drugs and no-DILI drugs using statistical methods.

First, comparison of the mean values between DILI drugs and no-DILI drugs was performed using Welch’s t-test (Fig. 3). Statistically significant differences were observed between DILI drugs and no-DILI drugs for CYP1A1 (p < 0.01), CYP1B1 (p < 0.001), CYP2B6 (p < 0.05) and CYP2C9 (p < 0.05).

Fig. 3

Comparison of the inhibitory activity of DILI drugs and no-DILI drugs. Inhibitory activity of DILI drugs and no-DILI drugs for each P450 were compared by Welch’s two sample two-sided t-test. The asterisks indicate statistical differences (*, p < 0.05; **, p < 0.01; ***, p < 0.001).

Next, inhibitory activity against each P450 was divided into 10 levels with a 10% range, and the distribution of inhibitory activity and DILI inducibility (DILI drugs vs. no-DILI drugs) are summarized as a 2 x 10 contingency table (Table 1). To investigate the relationships between DILI and the distribution of inhibitory activity, tests of independence on the contingency table were performed using Fisher’s exact test. The p-values obtained are shown in Table 1. Statistically significant associations with DILI were observed for CYP1A1 and CYP1B1 inhibition.

Table 1. The contingency table between DILI/no-DILI and the inhibitory activity against P450s.

P450	DILI	The number of drugs										p-value
		0≤x<10	10≤x<20	20≤x<30	30≤x<40	40≤x<50	50≤x<60	60≤x<70	70≤x<80	80≤x<90	90≤x≤100
CYP1A1	DILI	117	34	13	15	11	9	9	8	16	34	0.0010
	no-DILI	47	9	4	11	3	8	6	1	0	3
CYP1A2	DILI	211	18	9	6	5	3	4	5	2	3	0.3468
	no-DILI	84	1	2	1	0	0	1	1	0	2
CYP1B1	DILI	131	31	14	8	15	14	12	13	8	20	0.0030
	no-DILI	64	10	8	4	2	1	0	1	1	1
CYP2B6	DILI	208	21	12	4	6	7	1	1	2	4	0.8921
	no-DILI	80	5	3	1	2	0	0	0	0	1
CYP2C9	DILI	123	49	19	11	12	13	8	9	8	14	0.6232
	no-DILI	55	14	7	2	2	3	4	1	1	3
CYP2C19	DILI	123	37	20	16	7	7	8	17	10	21	0.5437
	no-DILI	42	20	6	5	3	0	2	2	5	7
CYP2D6	DILI	139	26	19	13	6	5	12	9	12	25	0.7426
	no-DILI	40	12	5	5	2	1	3	4	8	12
CYP3A4	DILI	113	60	23	17	9	10	9	8	10	7	0.8421
	no-DILI	47	15	10	3	4	3	3	2	2	3

x, inhibitory activity.

Finally, ROC analyses were performed to investigate whether each P450 inhibitory activity could discriminate between DILI drugs and no-DILI drugs and to determine the threshold values of each P450 assay for DILI discrimination (Table S3). The cutoff values showing the highest BA for each P450 inhibition assay were used as the thresholds. The thresholds for CYP1A1 and CYP1B1 inhibition were 64% and 40%, respectively, and discrimination with these values demonstrated extremely high specificity (> 90%) although sensitivity was low (25% and 31%, respectively) and BA values were moderate (60% and 62%). The thresholds for other P450s did not seem applicable in terms of experimental variations.

Taken together, these results suggest that among the P450s tested the inhibitory activity data on CYP1A1 and CYP1B1 might be useful for DILI risk evaluation.

Discrimination of DILI drugs and no-DILI drugs using inhibitory activity against CYP1A1 and CYP1B1

Discrimination between DILI drugs and no-DILI drugs of the 358 test drugs was performed using the thresholds for CYP1A1 and CYP1B1 inhibition determined above. A two-dimensional scatter plot of CYP1A1 and CYP1B1 inhibition was illustrated and the drugs were separated by the thresholds into four areas, termed A, B, C and D (Fig. 4). The numbers of drugs contained in each area are summarized in Table 2. We found that almost all the drugs in the areas B-D were DILI drugs (91%, 88 of 97) in the scatter plot. Among the DILI drugs used, 33% (88 of 266) were in the areas B-D while the ratio of no-DILI drugs was 9.8% (9 of 92). Interestingly, the ratio of ^vMost-DILI-concern drugs (40%, 48 of 119) was higher than that of ^vLess-DILI-concern drugs (27%, 40 of 147). The 9 no-DILI drugs in the areas B-D, numbered in the scatter plot, are listed in supplementary Table S4 with label compound names from DILIrank, and their indications and dosages obtained from SuperDRUG2.

Fig. 4

The scatter plot of CYP1A1 and CYP1B1 inhibition of test drugs. The inhibitory activity of 266 DILI drugs (red circles) and 92 no-DILI drugs (blue triangles) were plotted for CYP1A1 inhibition on the X-axis and CYP1B1 inhibition on the Y-axis. The area was partitioned by two lines of the threshold values for CYP1A1 inhibition (64%) and CYP1B1 inhibition (40%) determined by ROC analyses and the 4 partitions are termed A, B, C and D as indicated. The false positive no-DILI drugs are labeled with numbers 1 to 9 and their detailed information is shown in supplementary Table S3.

Table 2. The contingency table between DILI/no-DILI and area distribution.

		Area					Total
		A	B	C	D	B - D
no-DILI	Count	83	4	2	3	9	92
	% within area	31.8%	14.8%	3.3%	30.0%	9.3%	25.7%
DILI	Count	178	23	58	7	88	266
	% within area	68.2%	85.2%	96.7%	70.0%	90.7%	74.3%
Most*	Count	71	11	33	4	48	119
	% within area	27.2%	40.7%	55.0%	40.0%	49.5%	33.2%
Less*	Count	107	12	25	3	40	147
	% within area	41.0%	44.4%	41.7%	30.0%	41.2%	41.1%
Total	Count	261	27	60	10	97	358
	% within area	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

*Most and Less represent ^vMost-DILI-concern and ^vLess-DILI-concern in DILIrank dataset, respectively.

When the drugs in area A were classified as DILI-negative and the drugs in areas B-D as DILI-positive, the sensitivity, specificity, accuracy and BA of this discrimination were 33%, 90%, 48% and 62%, respectively. Since the area A has 83 no-DILI drugs and 178 DILI drugs, we performed ROC analyses to investigate inhibitory activity against other P450s could discriminate between DILI drugs and no-DILI drugs in the area A, but no valid thresholds were obtained (data not shown).

Finally, internal 5-fold cross-validation was performed for the discrimination of the test drugs using CYP1A1 and CYP1B1 inhibition thresholds. The results of five trials and thresholds were shown in the scatter plots of CYP1A1 vs. CYP1B1 for each test data (20% of the original drugs) (Supplementary Fig. S1). The thresholds of CYP1A1 inhibition were almost constant (about 64%) among the datasets while those of CYP1B1 inhibition relatively depended on the dataset. The mean ratios of DILI drugs in the areas B-D for the 5-fold cross-validation were 89% for learning data and 87% for test data, which was comparable to that obtained with the whole dataset.

Differences in the chemical properties between CYP1A1/CYP1B1-inhibiting DILI drugs and non-inhibitors

Since one-third of DILI drugs inhibited CYP1A1 or CYP1B1 by more than threshold values but the others did not, we investigated the differences in the chemical properties associated with these differences. To this end, chemical descriptors were calculated and those of DILI drugs in the area A and those of DILI drugs in the areas B-D were compared. As results, significant differences with p values of < 0.001 using Brunner-Munzel test were recognized between the area A and areas B-D in 32 chemical descriptors (Table S5, Fig. 5). These descriptors include those related to lipophilicity (nRCOOH, nROH, O-057, Hy, MLOGP and MLOGP2) and those associated with molecular saturation degree of carbon (nBM, SCBO, nC, C%, nCsp2, Uc and Ui).

Fig. 5

The box and violin plots of chemical descriptors showing significant differences between DILI drugs in the areas A and B-D. The boxplots are overlayed on the violin plots. The chemical descriptors of DILI drugs in the areas A and B-D were compared using Brunner-Munzel test. Y-axes represent the values of the descriptors indicated and the distributions of the values are shown as violin plots.

DISCUSSION

It has been suggested that reactivity with P450s (being substrates or inhibitors) is associated with DILI risk. Therefore, in this study, we investigated whether the inhibitory activity against 8 human P450s could be useful for evaluating DILI risk. With the 358 test drugs selected from the list of DILIrank drugs, we found that the target P450s and the degree of inhibition differed among the drugs as expected. Less inhibition was observed for CYP1A2 and CYP2B6 than other P450s. Pearson coefficient correlation analyses demonstrated that almost all the pairs of P450 forms had no significant association except the pair of CYP1A1 and CYP1B1, which is consistent with their reported substrate specificity (Guengerich, 2015). Nonetheless, these results suggest that P450 inhibitory activity data can be used as the characteristics of drugs as proposed in our previous study (Watanabe et al., 2020).

To investigate the usefulness of P450 inhibition data for predicting DILI risk, we performed various statistical analyses. Comparison of the mean values of inhibitory activity of DILI drugs and no-DILI drugs using Welch’s t-test demonstrated that the values for the inhibition of CYP1A1, CYP1B1, CYP2B6 and CYP2C9 were significantly different between DILI drugs and no-DILI drugs. With Fisher’s exact probability test for 2 x 10 contingency tables, we found significant associations between the inhibition of CYP1A1 or CYP1B1 and DILI. Based on these results, we preformed ROC analyses and the thresholds with the highest BA for DILI drug discrimination were determined as 64% for CYP1A1 inhibition and 40% for CYP1B1 inhibition. The classification of the 358 test drugs using these threshold values demonstrated that 91% of the test drugs that inhibited either CYP1A1 or CYP1B1 more than the threshold values were DILI drugs. These results are consistent with the previous findings showing the relationship between DILI and the drug property of being substrates and/or inhibitors of P450s (Yu et al., 2014). More importantly, the classification with CYP1A1 and CYP1B1 inhibition data might be useful for detecting potential DILI drugs with a high sensitivity.

When the structural characteristics of the DILI drugs classified into the area A and areas B-D were compared, significant differences (p < 0.001) in the median of molecular descriptors used with Brunner-Munzel test were found for those related to lipophilicity and molecular saturation. CYP1A1 is known to metabolize polycyclic aromatic hydrocarbons (PAHs), such as benzo[a]pyrene (Guengerich, 2015; Nebert and Gelboin, 1968; Shou et al., 1996), and the catalytic activity of CYP1B1 is similar to that of CYP1A1 (Guengerich, 2015; Shimada et al., 1998). Thus, the high lipid solubility and low molecular saturation of the DILI drugs in the areas B-D suggest that these DILI drugs have physicochemical properties similar to typical CYP1A1 and CYP1B1 substrates, although most of the drugs do not have PAH-like structures (supplementary Fig. S2-S4).

In the classification with CYP1A1 and CYP1B1 inhibition data, there were 9 false positive no-DILI drugs. According to their information on usage, most of them are administrated with parenteral routes. Pyrantel, an anthelmintic, and primaquine, an antimalarial drug, are often used with a single dose. Norethindrone is a progestin drug and used orally or intramuscularly. Based on these facts, these false positive drugs have much lower exposure to the liver than other drugs that are used orally. In addition, the number of patients treated with these drugs might be relatively low. Although the drugs may have chemical structures that cause hepatotoxicity according to our present results, these facts imply that these false positive drugs are classified as no-DILI drugs in the DILIrank dataset due to their unique clinical use.

CYP1A1 is expressed in fetal livers but its expression in adult livers is relatively low and mainly expressed in extrahepatic tissues in adults (Guengerich, 2015; Omiecinski et al., 1990). CYP1B1 is also known to be barely expressed at the protein level in hepatocytes (Guengerich, 2015; Shimada et al., 1996). Because of these reasons, there is little information on the involvement of CYP1A1 and CYP1B1 in drug metabolism and their inhibition are rarely assessed during drug development processes. Information on the inhibitory activity of CYP1A1 and CYP1B1 by the test drugs obtained in this study would be useful for future research on these P450s.

Since the contribution of CYP1A1 and CYP1B1 to drug metabolism in the liver is likely to be minimal, mechanistic linkage between the inhibition of these P450s and the onset of liver injury remains unclear. At this moment, we have raised the following 3 possibilities. First, inhibition of CYP1A1 and/or CYP1B1 expressed in extrahepatic tissues may indirectly induce hepatotoxicity. These P450s are known to be involved in the metabolism of several endogenous compounds, such as estrogen and tryptophan metabolites (Guengerich, 2015), and thus long-term inhibition of the metabolism of these compounds may indirectly affect liver functions, leading to hepatotoxicity. Second, the chemical structures of the drugs that inhibit CYP1A1 or CYP1B1, rather than their inhibition itself, might be associated with hepatotoxicity. Since P450s recognize substructures of chemical compounds as well as whole molecules, CYP1A1 and CYP1B1 might recognize as yet undefined chemical structures associated with DILI. In this regard, we have recently reported that inhibitions of some specific P450 forms are associated with various types of endpoints in rat repeated-dose toxicity, including hepatocyte hypertrophy/swelling, blood γ-glutamyl transpeptidase increase, kidney basophilic change/regeneration, anemia and coagulation abnormality, and stomach hyperplasia/thickening epithelium, proposing P450 reactivity information as new molecular descriptors to characterize chemical compounds that cannot be recognized by conventional chemical descriptors (Watanabe et al., 2020). Finally, CYP1A1 or CYP1B1 inhibition might be a surrogate of other biological action of the inhibiting compounds. Since target molecules of DILI are largely unknown, it is possible that there are unknown biomolecules related to the hepatotoxicity and their chemical reactivity in vivo is similar to that of CYP1A1 or CYP1B1. Although further investigations are necessary to test these hypotheses, we believe that our present findings provide novel insights into the onset of DILI and the development of a useful preclinical method(s) to screen DILI risk during drug development.

ACKNOWLEDGMENTS

This work was supported in part by a Grant-in-Aid from the Ministry of Education, Culture, Sports, Sciences and Technology of Japan (grant numbers 17K08419) and a grant from The Mochida Memorial Foundation for Medical and Pharmaceutical Research (FY 2020). This work was also financially supported by Asahi Kasei Pharma Corporation.

Conflict of interest

The authors declare that there is no conflict of interest.

REFERENCES

• Albrecht, W., Kappenberg, F., Brecklinghaus, T., Stoeber, R., Marchan, R., Zhang, M., Ebbert, K., Kirschner, H., Grinberg, M., Leist, M., Moritz, W., Cadenas, C., Ghallab, A., Reinders, J., Vartak, N., van Thriel, C., Golka, K., Tolosa, L., Castell, J.V., Damm, G., Seehofer, D., Lampen, A., Braeuning, A., Buhrke, T., Behr, A.C., Oberemm, A., Gu, X., Kittana, N., van de Water, B., Kreiling, R., Fayyaz, S., van Aerts, L., Smedsrød, B., Ellinger-Ziegelbauer, H., Steger-Hartmann, T., Gundert-Remy, U., Zeigerer, A., Ullrich, A., Runge, D., Lee, S.M., Schiergens, T.S., Kuepfer, L., Aguayo-Orozco, A., Sachinidis, A., Edlund, K., Gardner, I., Rahnenführer, J. and Hengstler, J.G. (2019): Prediction of human drug-induced liver injury (DILI) in relation to oral doses and blood concentrations. Arch. Toxicol., 93, 1609-1637.
• Barnham, K.J., Masters, C.L. and Bush, A.I. (2004): Neurodegenerative diseases and oxidative stress. Nat. Rev. Drug Discov., 3, 205-214.
• Brunner, E. and Munzel, U. (2000): The nonparametric Behrens-Fisher problem: asymptotic theory and a small-sample approximation. Biometrical. J., 42, 17-25.
• Chalasani, N. and Björnsson, E. (2010): Risk factors for idiosyncratic drug-induced liver injury. Gastroenterology, 138, 2246-2259.
• Chen, M., Borlak, J. and Tong, W. (2016a): A Model to predict severity of drug-induced liver injury in humans. Hepatology, 64, 931-940.
• Chen, M., Suzuki, A., Thakkar, S., Yu, K., Hu, C. and Tong, W. (2016b): DILIrank: the largest reference drug list ranked by the risk for developing drug-induced liver injury in humans. Drug Discov. Today, 21, 648-653.
• Cichoż-Lach, H. and Michalak, A. (2014): Oxidative stress as a crucial factor in liver diseases. World J. Gastroenterol., 20, 8082-8091.
• Cuykx, M., Claes, L., Rodrigues, R.M., Vanhaecke, T. and Covaci, A. (2018): Metabolomics profiling of steatosis progression in HepaRG^® cells using sodium valproate. Toxicol. Lett., 286, 22-30.
• Dragovic, S., Vermeulen, N.P., Gerets, H.H., Hewitt, P.G., Ingelman-Sundberg, M., Park, B.K., Juhila, S., Snoeys, J. and Weaver, R.J. (2016): Evidence-based selection of training compounds for use in the mechanism-based integrated prediction of drug-induced liver injury in man. Arch. Toxicol., 90, 2979-3003.
• Feng, S. and He, X. (2013): Mechanism-based inhibition of CYP450: an indicator of drug-induced hepatotoxicity. Curr. Drug Metab., 14, 921-945.
• Goda, K., Kobayashi, A., Takahashi, A., Takahashi, T., Saito, K., Maekawa, K., Saito, Y. and Sugai, S. (2017): Evaluation of the potential risk of drugs to induce hepatotoxicity in human-relationships between hepatic steatosis observed in non-clinical toxicity study and hepatotoxicity in humans. Int. J. Mol. Sci., 18, 810.
• Guengerich, F.P. (2015): Human cytochrom P450 enzymes. In: Cytochrome P450. Structure, Mechanism, and Biochemistry, 4th edn. (Ortiz de Montellano, P. R., ed.), pp 532-785. Springer, Cham, Switzerland.
• Han, D., Dara, L., Win, S., Than, T.A., Yuan, L., Abbasi, S.Q., Liu, Z.X. and Kaplowitz, N. (2013): Regulation of drug-induced liver injury by signal transduction pathways: critical role of mitochondria. Trends Pharmacol. Sci., 34, 243-253.
• Khansari, N., Shakiba, Y. and Mahmoudi, M. (2009): Chronic inflammation and oxidative stress as a major cause of age-related diseases and cancer. Recent Pat. Inflamm. Allergy Drug Discov., 3, 73-80.
• Kullak-Ublick, G.A., Andrade, R.J., Merz, M., End, P., Benesic, A., Gerbes, A.L. and Aithal, G.P. (2017): Drug-induced liver injury: recent advances in diagnosis and risk assessment. Gut, 66, 1154-1164.
• Lammert, C., Bjornsson, E., Niklasson, A. and Chalasani, N. (2010): Oral medications with significant hepatic metabolism at higher risk for hepatic adverse events. Hepatology, 51, 615-620.
• Nebert, D.W. and Gelboin, H.V. (1968): Substrate-inducible microsomal aryl hydroxylase in mammalian cell culture. I. Assay and properties of induced enzyme. J. Biol. Chem., 243, 6242-6249.
• O’Brien, P.J., Irwin, W., Diaz, D., Howard-Cofield, E., Krejsa, C.M., Slaughter, M.R., Gao, B., Kaludercic, N., Angeline, A., Bernardi, P., Brain, P. and Hougham, C. (2006): High concordance of drug-induced human hepatotoxicity with in vitro cytotoxicity measured in a novel cell-based model using high content screening. Arch. Toxicol., 80, 580-604.
• O’Boyle, N.M., Banck, M., James, C.A., Morley, C., Vandermeersch, T. and Hutchison, G.R. (2011): Open Babel: an open chemical toolbox. J. Cheminform., 3, 33.
• Oda, S. and Yokoi, T. (2015): Establishment of animal models of drug-induced liver injury and analysis of possible mechanisms. Yakugaku Zasshi, 135, 579-588.
• Omiecinski, C.J., Redlich, C.A. and Costa, P. (1990): Induction and developmental expression of cytochrome P450IA1 messenger RNA in rat and human tissues: detection by the polymerase chain reaction. Cancer Res., 50, 4315-4321.
• Russmann, S., Jetter, A. and Kullak-Ublick, G.A. (2010): Pharmacogenetics of drug-induced liver injury. Hepatology, 52, 748-761.
• Shimada, T., Hayes, C.L., Yamazaki, H., Amin, S., Hecht, S.S., Guengerich, F.P. and Sutter, T.R. (1996): Activation of chemically diverse procarcinogens by human cytochrome P-450 1B1. Cancer Res., 56, 2979-2984.
• Shimada, T., Yamazaki, H., Foroozesh, M., Hopkins, N.E., Alworth, W.L. and Guengerich, F.P. (1998): Selectivity of polycyclic inhibitors for human cytochrome P450s 1A1, 1A2, and 1B1. Chem. Res. Toxicol., 11, 1048-1056.
• Shou, M., Krausz, K.W., Gonzalez, F.J. and Gelboin, H.V. (1996): Metabolic activation of the potent carcinogen dibenzo[a,h]anthracene by cDNA-expressed human cytochromes P450. Arch. Biochem. Biophys., 328, 201-207.
• Stevens, J.L. and Baker, T.K. (2009): The future of drug safety testing: expanding the view and narrowing the focus. Drug Discov. Today, 14, 162-167.
• U.S. Department of Health and Human Services, Food and Drug Administration. (2009): Guidance for Industry Drug-Induced Liver Injury : Premarketing Clinical Evaluation.
• 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.
• Watkins, P.B. (2005): Idiosyncratic liver injury: challenges and approaches. Toxicol. Pathol., 33, 1-5.
• Xu, J.J., Henstock, P.V., Dunn, M.C., Smith, A.R., Chabot, J.R. and de Graaf, D. (2008): Cellular imaging predictions of clinical drug-induced liver injury. Toxicol. Sci., 105, 97-105.
• 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.