Changes in gene expression induced by aromatic amine drugs: Testing the danger hypothesis

Abstract

Virtually all drugs that contain a primary aromatic amine are associated with a high incidence of idiosyncratic drug reactions (IDRs), suggesting that this functional group has biological effects that may be used as biomarkers to predict IDR risk. Most IDRs exhibit evidence of immune involvement and the ability of aromatic amines to form reactive metabolites and redox cycle may be responsible for initiation of an immune response through induction of cell stress, as postulated by the Danger Hypothesis. If true, danger signals could be biomarkers of IDR risk. A previous attempt to test the Danger Hypothesis found that sulfamethoxazole (SMX), the only aromatic amine tested, was also the only drug not associated with an increase of cell stress genes in mice. To ensure that these observations were not species-specific, and to determine biomarkers of IDR risk common to aromatic amines, rats were treated with SMX and two other aromatic amine drugs, dapsone (DDS) and aminoglutethimide (AMG), and hepatic gene expression was determined using microarrays. As in mice, SMX induced minimal gene changes in the rat, and none indicated cell stress, whereas DDS and AMG induced several changes including up-regulation of enzymes such as aldo-keto reductase, glutathione-S-transferase, and aldehyde dehydrogenase, which may represent danger signals. Early insulin-induced hepatic gene (Eiih) was up-regulated by all three drugs. Some mRNA changes were observed in the Keap-1-Nrf2-ARE pathway; however, the pattern was significantly different for each drug. Overall, the most salient finding was that the changes in the liver were minimal, even though aromatic amines cause a high incidence of IDRs. The liver generates a large number of reactive species; however, the ability of aromatic amines to be bioactivated by cells of the immune system may be why they cause a high incidence of IDRs.

Keywords::

• Aromatic amines
• idiosyncratic drug reactions
• Danger Hypothesis
• microarray
• sulfamethoxazole
• dapsone
• aminoglutethimide
• biomarkers
• early insulin-induced hepatic gene
• Keap1-Nrf2-ARE
• thioredoxin reductase
• immune tolerance

Abbreviations
IDR,	=	Idiosyncratic drug reaction;
SMX,	=	Sulfamethoxazole;
DDS,	=	Dapsone;
AMG,	=	Aminoglutethimide;
CYP450,	=	Cytochrome P450;
FDR,	=	False discovery rate;
Eiih,	=	early insulin-induced hepatic gene;
β2M,	=	β2-microglobulin;
Dusp1,	=	Dual specificity phosphatase 1;
Sgk,	=	Serum/glucocorticoid-regulated kinase 1;
Txnrd,	=	Thioredoxin reductase;
GSH-S-Tr,	=	Glutethione-S-Transferase

Introduction

Drugs that contain the primary aromatic amine functional group are almost always associated with a high incidence of idiosyncratic drug reactions (IDRs), and this is considered a structural alert for drug development (Uetrecht, 2002). The inability to predict which drug candidates will cause such reactions, and the fact that they are usually not detected until very late in development or after the drug has been marketed, markedly increases the risks associated with the development of new/novel drugs.

The exact mechanisms of IDRs remain unclear; however, many lines of evidence suggest that most IDRs are mediated by the adaptive immune system (Uetrecht, 2009). The Danger Hypothesis has been proposed as a mechanism for the initiation of immune-mediated IDRs (Li and Uetrecht, 2010). This hypothesis posits that the tissues are responsible for initiating an immune response through the release of danger signals in response to cell stress or damage (Matzinger, 1994). This hypothesis compliments other hypotheses such as the hapten hypothesis because it may result in activation of antigen-presenting cells leading to costimulatory signals, referred to as Signal 2, that are required for activation of T-helper cells (Figure 1). In the context of IDRs, a drug, or more likely its reactive metabolite, could induce cell stress or damage, and this might contribute to the ability of a drug to cause IDRs.

Figure 1.  The Danger Hypothesis applied to IDRs. The drug or its reactive metabolite can form antigenic adducts with endogenous molecules that are presented to T-cells on major histocompatibility complexes of antigen-presenting cells (Signal 1). Additionally, the drug or reactive metabolite may induce cell stress or damage through processes such as redox cycling and oxidative damage. This may lead to the release of danger signals to up-regulate co-stimulatory molecules such as B7 and CD40 on antigen-presenting cells to activate T-cells (Signal 2). Signals 1 and 2 are required concurrently to initiate an adaptive immune response and certain drugs have the ability to induce both signals.

To date, many potential danger signals have been identified such as heat shock proteins, high-mobility group protein B1, ATP, cytokines, and nuclear DNA (Gallucci and Matzinger, 2001). However, it is difficult to rigorously test whether the induction of danger signals is an important biomarker of IDR risk. Danger signals vary widely in their characteristics; the only consistent feature being that they are always endogenous molecules, either actively secreted or released due to destruction of cell membrane integrity in the event of cell stress or damage. If the release of danger signals could be used as a biomarker to predict the risk that a drug candidate would cause IDRs, it would have a profound effect on the process of drug development.

The IDRs associated with aromatic amine drugs are thought to be due to the ease with which they form reactive metabolites (Figure 2). Aromatic amines can be oxidized by cytochrome P_450s (CYP450s) and also myeloperoxidase to hydroxylamine metabolites, which are not very reactive, but can auto-oxidize to electrophilic nitroso metabolites that can covalently bind endogenous proteins. The nitroso metabolite can be easily reduced back to the parent amine with antioxidants, and this redox cycling could lead to oxidative stress. Thus, aromatic amine drugs have the potential to form antigenic substances and to cause cell stress and damage, which could produce danger signals and initiate immune-mediated IDRs.

Figure 2.  Metabolic scheme for the formation of reactive metabolites of aromatic amine drugs. Oxidation of aromatic amines to hydroxylamines can occur through enzymes such as CYP450s or myeloperoxidase (MPO); this is prevented by N-acetylation. Hydroxylamines are not very reactive but can auto-oxidize to electrophilic nitroso metabolites. The nitrosoamine can react with thiol-containing nucleophiles such as glutathione or endogenous proteins to form sulfinamide adducts. Additionally, the nitrosoamine can be reduced back to the parent drug, leading to redox cycling and oxidative stress.

Sulfamethoxazole (SMX) is an aromatic amine antimicrobial that is widely used in combination with trimethoprim for many indications including prophylaxis against and to treat Pnuemocystis carinii pneumonia in HIV-infected individuals. SMX can cause a generalized hypersensitivity reaction and is also associated with among the highest incidences of Stevens-Johnson syndrome and toxic epidermal necrolysis, which can be very serious and even fatal (Chan et al., 1990; Mockenhaupt, 2011).

Previously we tried to test the Danger Hypothesis with a study of changes in gene expression induced by tienilic acid, which causes idiosyncratic liver injury. Its reactive metabolite binds principally to CYP450 that is unlikely to cause much cell stress. However, it binds to several other proteins and it induced several changes in gene expression in the liver that likely represented danger signals (Pacitto et al., 2007). In contrast, treatment of mice with SMX, which was meant as a positive control because it can covalently bind to a wider variety of proteins and also induce oxidative stress, did not lead to up-regulation of genes that were likely to represent danger signals, and the few changes that were induced in the liver were mostly down-regulation of gene expression (Pacitto et al., 2007). It is possible that down-regulation of mRNAs represents a danger signal. However, metabolism of SMX in mice is limited (Farrell et al., 2003), and the metabolism of SMX in rats more closely resembles its metabolism in humans (Gill et al., 1997). Thus, a global screen of hepatic gene expression was performed to determine the effects of SMX in the rat to investigate whether it can induce cell stress leading to the release of danger signals. In addition, two other aromatic amine drugs associated with a high incidence of IDRs, dapsone (DDS) and aminoglutethimide (AMG; Figure 3), were tested to determine whether similar expression profiles exist between aromatic amine drugs that could potentially indicate a similar mechanism of action and be useful as biomarkers for drugs with the aromatic amine functional group. As the major site of drug metabolism and target for IDRs, the liver (or hepatocytes) are extensively used for reactive metabolite and toxicity studies during drug development. Therefore, we chose the liver in which to study changes in gene expression induced by aromatic amines; possible downstream effects of these gene changes were also tested.

Figure 3.  Structures of the aromatic amine drugs used in this study and some associated IDRs.

Materials and methods

Reagents

Sulfamethoxazole (SMX) and dapsone (DDS) were purchased from Sigma (Oakville, Ontario, Canada). AMG was purchased from Toronto Research Chemical (North York, Ontario, Canada). Nevirapine was obtained from Boehringer Ingelheim (Ridgefield, CT). Clozapine was provided by Novartis (Dorval, Quebec, Canada). RNeasy Mini kits were obtained from Qiagen (Missis-sauga, Ontario, Canada). Gene chips were purchased from Affymetrix (Santa Clara, CA). Light Cycler FastStart DNA Master SYBR Green I was purchased from Roche Applied Science (Laval, Quebec, Canada). Primers were obtained from Integrated DNA Technologies (Coralville, IA).

Animals

Male Brown Norway rats (≈ 8-weeks-old, 200–250 g) were purchased from Charles River (Montreal, Quebec, Canada). Rats were housed under standard conditions (doubly housed in plastic cages, standard rat chow, automatic watering, 12:12 h light:dark cycle, and 22°C). Food and water were provided ad libitum. Rats were acclimatized for 1 week before the start of experiments. In other studies performed with penicillamine and nevirapine in our laboratory, the Brown Norway rat was more susceptible to IDRs induced by these drugs (i.e., autoimmunity and skin rash, respectively) than other strains; for this reason, this strain was selected for use in the current study. The study protocol was approved by and performed in accordance with the University of Toronto Faculties of Medicine and Pharmacy Animal Care Committee.

Treatments

The treatment protocol was similar to that of Pacitto et al. (2007), with some modifications. Each treatment had four rats per group unless otherwise stated. The drugs were administered by oral gavage at doses of 150 mg/kg (SMX), 20 mg/kg (DDS), and 80 mg/kg (AMG). These dosages were chosen from previous rat studies (Schwartz and Rieder, 1970; Nicholls et al. 1984; Gill et al., 1997; Helton et al., 2000) to mimic peak drug concentrations in the blood of patients taking these drugs for therapy (≈37 µg/ml for 1000 mg SMX (Goodman et al., 2006), 1.7 µg/ml for 100 mg DDS (Goodman et al., 2006), and 3 µg/ml for 250 mg AMG (Lonning et al., 1985)).

Drugs were suspended in 0.5% methylcellulose due to the poor solubility of DDS and AMG. For the microarray study, rats were euthanized by CO₂ asphyxiation after 12 h drug treatment. Treatments for follow-up studies were performed as described previously using several different timepoints, with the exception that for timepoints > 24 h, rats were given an additional dose of drug every 24 h. At necropsy, the liver was extracted using aseptic techniques, immersed in RNAlater RNA Stabilization Reagent (Qiagen), and stored according to manufacturer′s instructions at −80°C until RNA extraction. Liver tissue was also immersed in phosphate buffered saline (PBS, pH 7.4) with protease inhibitor (Sigma) for activity assays. PBS was used instead of lysis buffer to minimize possible interferences by the solvent with assay matrices.

RNA extraction

Liver tissue (≈20–30 mg) stored in RNAlater RNA Stabilization Reagent was homogenized with a rotor stator homogenizer (IKA Ultra-Turrax T25 S1, Janke & Kunel, Staufen, Germany). Total RNA was extracted using an RNeasy Mini Kit (Qiagen) as per manufacturer′s protocol. Extracted RNA was eluted in 40 µl of RNase-free water and stored at −80°C. The quality and purity of the extracted RNA was initially measured using UV spectrophotometry based on the A₂₆₀/A₂₈₀ ratio before a further quality check through capillary electrophoresis using the Agilent Bioanalyzer (Affymetrix).

Microarray analysis

Affymetrix RatGene 1.0 ST chips were used for detection of changes in gene expression. Microarray processing of the liver samples was performed at the microarray facility in The Centre for Applied Genomics (Hospital for Sick Children, Toronto, Ontario, Canada) based on Affymetrix protocols. For microarray data, Expression Console software was used for quality control analysis before fold-changes were determined using Partek software. The Robust Multi-array Analysis method was used for background correction, which included Quantile Normalization and Median Polish summarization. Data was log-base 2 transformed and fold-change values were obtained for treatments as compared to the control group. One-way ANOVA analysis was performed to determine the statistical significance based on differences between the treatment and control mean intensities for each gene. Attempts were made to obtain more accurate p-values by performing multiple test corrections using the false discovery rate (FDR).

Real-time (RT)-PCR of early insulin-induced hepatic gene (Eiih)

Extracted RNA samples were converted to cDNA using the Omniscript RT Kit (Qiagen), with oligo(dT15) primers (Roche) and RNase inhibitor (Roche), as per the manufacturer′s protocol. Quality was checked using the spectrophotometry method as previously described. For each sample, 2 µg of RNA was converted to cDNA in a total reaction volume of 20 µl. Using Light Cycler FastStart DNA Master SYBR Green 1 (Roche), PCR was performed using a Light Cycler instrument (Roche) as per the following conditions: pre-incubation at 95°C for 10 min and amplification for 45 cycles (denaturation at 95°C for 15 s, annealing at 60°C for 5 s, and elongation at 72°C for 10 s). Primers (Integrated DNA Technologies) for Eiih were as follows: forward primer 5′-AGCTCTCCAGCTCTGGATTCTT-3′, reverse primer 5′-CACACCCAGAACAGAGTCTTAC-3′. β₂-Microglobulin (β₂M) was used as the housekeeping gene and the primers (Integrated DNA Technologies) were as follows: forward primer 5′-TCAGTTCCACCCACCTCAGATAGA-3′, reverse primer 5′-TGTGAGCCAGGATGTAGAAAGAC-3′. Data was analyzed using the RelQuant software and normalized to a calibrator control.

An additional two drugs, nevirapine and clozapine, were both tested in the rat for Eiih expression using RT-PCR. The treatment protocol, tissue extraction, and PCR analysis were similar to previously stated. Nevirapine and clozapine were administered at 150 mg/kg and 50 mg/kg, respectively, through oral gavage, and samples were taken 6 and 12 h after treatment.

Activity assays

Thioredoxin reductase (Txnrd) and glutathione-S-transferase (GSH-S-Tr) activity in the liver was measured using a Txnrd assay kit from Cayman Chemical (Ann Arbour, MI) and a GSH-S-Tr activity colorimetric assay kit from Abcam (Cambridge, MA), respectively, as per manufacturer′s instructions. Assayed samples contained a total protein concentration of 4 mg/ml in PBS with proteinase inhibitor. The levels of sensitivity of the Txnrd and GSH-S-Tr kits/assays were 0.015 and 0.025 µmol/min/ml, respectively.

Statistical analysis

Data obtained from PCR and activity assays were analyzed by ANOVA using GraphPad Prism 5. Bonferroni post-tests were used to determine statistical significance between treatment and control groups.

Results

Microarray quality

The extracted rat liver RNA was of good quality as demonstrated via UV spectrophotometry and the Agilent Bioanalyzer methods. The A₂₆₀/A₂₈₀ ratios of the samples were in the range of 1.96–2.29, indicating good RNA quality. Only one sample in the SMX treatment had a RNA Integrity Number < 7, but was still sufficient for microarray processing. Expression Console software revealed no outlier arrays and the Spearman Rank Correlation (r²) was > 0.982, indicating a good correlation between signal values of different gene chips. Thus, all arrays were used for microarray data analysis.

SMX-induced hepatic gene expression

SMX induced very few gene changes >2-fold expression in the rat 12 h after treatment (see Appendix A). When FDR was applied for multiple test corrections, none of the changes were significant (p < 0.05); of the gene changes that occurred, none seemed to explicitly indicate cell stress or initiation of an immune response. Interestingly, genes involved with the acute cell response were down-regulated, including dual specificity phosphatase 1 (Dusp1) and serum/glucocorticoid-regulated kinase (Sgk).

DDS-induced hepatic gene expression

In contrast to SMX, a greater number of genes were up-regulated > 2-fold with DDS treatment in the rat (see Appendix B). However, only changes in six genes were statistically significant (p < 0.05), with the FDR applied: aldo-keto reductase 7A3 (Akr7a3, 5.38-fold), glutathione-S-transferase Yc2 subunit (Yc2, 3.18-fold), aldehyde oxidase 1 (Aox1, 1.63-fold), ferritin light polypeptide (Ftl1, 1.44-fold), UDP glucosyltransferase 1A3 (Ugt1a3, 1.39-fold), and glutathione reductase (Gsr, 1.38-fold). Up-regulation of genes such as Akr7a3, Yc2, Ugt1a3, and Gsr could indicate cell stress due to their function in drug metabolism and cytoprotection against oxidative stress. Again, Dusp1 and Sgk were among the most down-regulated genes, similar to what was observed with SMX.

AMG-induced hepatic gene expression

AMG induced the most gene changes of the aromatic amine drugs tested (see Appendix C). Among these changes, 552 genes were significantly different using the FDR. Of these genes, quite a few were involved in antioxidant and drug metabolizing pathways, including CYP450 oxidoreductase (Por), aldehyde dehydrogenase 1A1 (Aldh1a1), UDP glucuronosyltransferase 2B1 (Udpgtr2), epoxide hydrolase 1(Ephx1), aldo-keto reductase 7A3 (Akr7a3), glutathione-S-transferase Yc2 sub-unit (Yc2), thioredoxin reductase 1 (Txnrd1), glutathione reductase (Gsr), and UDP glycosyltransferase 1A3 (Ugt1a3), which could indicate that cell stress was induced. Additionally, several heat shock proteins were up-regulated such as Hspca, Hsph1, Hspa1b, and Hspb8 that could be potential danger signals to initiate an immune response. The pattern of decreased gene expression was also similar to the other aromatic amine drugs, as Dusp1 and Sgk were among the most down-regulated.

Aromatic amine gene expression profile

Although the gene changes were different for each aromatic amine drug, in general, aromatic amines induced greater down-regulated than up-regulated gene changes (Table 1). Among these, hepatic protein EIIH (Eiih; aka early insulin-induced hepatic gene) was the most highly up-regulated gene in all three drugs tested. Increased Eiih gene expression was confirmed through RT-PCR for both DDS and AMG treatment, but not SMX (Figure 4). Further time-course investigation using RT-PCR found that Eiih was expressed only acutely, with the highest expression at 8 and 12 h after AMG treatment (Figure 5). Other drugs known to cause IDRs such as nevirapine and clozapine also increased expression of Eiih at early time points in treated rats (Figure 6).

Table 1.  Similar genes differentially regulated at least 1.4-fold change in the same direction in all aromatic amine drugs tested 12 h after treatment.

Gene	Function	Fold-change
		SMX	DDS	AMG
Eiih (hepatic protein EIIH)	Unknown function; induced early in rats upon insulin treatment; may play role in carbohydrate metabolism through association with glucose transporter 2 (Coffy et al., 2005).	2.86	4.23	12.94*
Socs2 (suppressor of cytokine signaling 2)	Inhibits growth hormone activity; role in mediating the anti-inflammatory effects of lipoxins (Machado et al., 2006).	1.48	2.08	1.41
Dusp1 (dual specificity phosphatase 1)	Involved with dephosphorylation of tyrosine and serine/threonine residues; role in signalling pathways due to interaction with p38, JNK, ERK; inducible early response to LPS, hypoxia, heat shock, oxidative stress (Patterson et al., 2009).	−3.58	−3.01	−3.61
Sgk (serum/glucocorticoid regulated kinase)	Serine/threonine protein kinase involved with cellular signaling pathways of metabolism, proliferation, and differentiation in response to glucocorticoids and serum; induced transiently during early response to stress (Webster et al., 1993).	−2.69	−3.72	−5.37*
G0s2 (G0/G1 switch gene 2)	Involved with mitochondrial induction of apoptosis and growth arrest in human fibroblasts; target of retinoic acid and peroxisome proliferator-activated receptor agonists; may play role in adipogenesis (Zandbergen et al., 2005; Welch et al., 2009).	−2.11	−2.03	−1.59
Jun (Jun proto-oncogene)	Activated by JNK phosphorylation; interacts with c-Fos to form AP-1 transcription factor that regulates cell proliferation and apoptosis pathways in response to stress (Leppa and Bohmann, 1999).	−1.97	−2.49	−1.51
Cyr61 (cysteine rich protein 61)	Heparin binding angiogenesis inducer, involved with endothelial adhesion and migration and tumor growth (Babic et al., 1998).	−1.87	−2.5	−1.94
Slc25a25 (solute carrier family 25, member 25)	Regulates ATP homeostasis as a shuttle across inner mitochondrial membrane; involved with maintaining metabolic efficiency (Anunciado-Koza et al., 2011).	−1.86	−2.67	−4.04*
Acat2 (acetyl-Coenzyme A acetyltransferase 2)	Cytosolic acetyl-Coenzyme A acetyltransferase activity; involved with synthesis of cholesteryl esters and cholesterol absorption; mainly localized to liver and intestines; may play a role in atherosclerosis (Lada et al., 2003).	−1.63	−1.63	−2.28
Slc34a2 (solute carrier family 34, member 2)	Sodium-phosphate transporter found in intestines, kidney, and lung; may be involved with regulating phosphate levels in intestinal epithelium (Xu et al., 1999).	−1.59	−1.82	−3.02*
Sqle (squalene epoxidase)	Involved in rate-limiting step of cholesterol biosynthesis; oxidizes squalene to squalene epoxide; negatively regulated by cholesterol (Chugh et al., 2003).	−1.51	−1.42	−1.72
Inhbe (inhibin beta E)	Member of TGF-β family; may induce apoptosis in hepatoma cells and down-regulated in hepatocarcinogenesis (Grusch et al., 2006).	−1.44	−1.51	−2.01*

Data expressed as fold-change between treated and controls. * p < 0.05 with FDR applied (n = 4).

Figure 4.  Eiih expression induced by aromatic amine drugs 12 h after treatment. Among the treatment groups (solid bars), AMG induced a significant increase in Eiih expression as compared to control (open bars). DDS also induced an increase in Eiih; however, no change was observed for SMX. Rats were given 150 mg SMX/kg, 20 mg DDS/kg, or 80 mg AMG/kg, and liver samples were taken for RT-PCR analysis. Eiih expression was normalized to β₂-microglobulin (β₂M) as a housekeeping gene and a calibrator control. RT-PCR was run in triplicate (n = 4; *** p < 0.001).

Figure 5.  Time-course of hepatic Eiih expression induced by AMG. Eiih was expressed the greatest 8 and 12 h after AMG treatment (solid bars) as compared to control (open bars). Rats were treated with 80 mg AMG/kg, and liver samples were taken for RT-PCR analysis. Eiih expression was normalized to β₂-microglobulin (β₂M) as a housekeeping gene and a calibrator control. RT-PCR was run in triplicate (n = 2).

Figure 6.  Hepatic gene expression of Eiih induced by several other drugs associated with IDRs. Clozapine (CLZ), an atypical anti-psychotic drug, induced an increase in Eiih 12 h after treatment of rats, and nevirapine (NVP), an anti-retroviral drug, increased Eiih expression at both 6 and 12 h after treatment of rats. Data expressed as fold-change between treated and controls (n = 4; * p < 0.05, ** p < 0.01).

A greater number of similarities were also observed between DDS and AMG in terms of genes regulated by the Keap1-Nrf2-ARE pathway (Table 2). Further testing of downstream effects of these gene changes found that enzyme activity did not always correlate with changes in mRNA expression. Greater Txnrd activity was found after SMX treatment when Txnrd1 gene expression was low compared to that induced by DDS or AMG, whereas gene expression of Txnrd1 was higher with the other two drugs, when Txnrd activity was low (Figure 7a). Nevertheless, a time-course study with AMG found that Txnrd activity was only increased at 48 h after treatment (Figure 7b). GSH-S-Tr activity was also tested, but no significant changes were observed except for a possible increase in activity 48 h after AMG treatment (Figure 8).

Table 2.  Keap1-Nrf2-ARE-regulated genes that were changed ≥1.4 fold by at least one of the aromatic amine drugs (SMX, DDS, or AMG) tested 12 h after treatment.

Keap-1-Nrf2-ARE-regulated genes	Fold-change
	SMX	DDS	AMG
Xenobiotic metabolism
Glucuronidation:
Udpgtr2 (UDP glucuronosyltransferase 2 family, polypeptide B1)	1.12	1.28	3.59*
Ugt1a3 (UDP glycosyltransferase 1 family, polypeptide A3)	1.09	1.39*	2.06*
Ugt2b4 (UDP glycosyltransferase 2 family, polypeptide B4)	−1.09	1.10	1.45
Glutathione conjugation:
Yc2 (glutathione S-transferase Yc2 subunit)	1.01	3.19*	2.23*
Gstm5 (glutathione S-transferase, mu 5)	1.08	1.53	1.14
Gsta2 (glutathione-S-transferase, alpha type2)	−1.02	1.31	1.49*
Other:
Aldh1a1 (aldehyde dehydrogenase family 1, member A1)	−1.23	2.33	4.15*
Ephx1 (epoxide hydrolase 1, microsomal)	1.02	1.71	2.33*
Nqo1 (NAD(P)H dehydrogenase, quinone 1)	1.13	1.57	1.53
Regulation of cellular redox/antioxidant status
Txnrd1 (thioredoxin reductase 1)	1.16	1.40	2.23*
Gsr (glutathione reductase)	1.08	1.38*	2.06*
Energy production
Me1 (malic enzyme 1)	1.09	1.42	2.76*
Heme catabolism
Hmox1 (heme oxygenase (decycling) 1)	1.09	1.06	1.72

Data expressed as fold-change between treated and controls. *p < 0.05 with FDR applied (n = 4).

Figure 7.  Hepatic Txnrd activity after treatment with aromatic amine drugs. Thioredoxin reductase activity was increased at both 12 and 24 h after SMX treatment (a) but not with the other two aromatic amine drugs (n = 4; *** p < 0.001). However, a time-course study (b) found that AMG appeared to increase thioredoxin reductase activity 48 h after treatment (n = 2). Rats were treated with 150 mg SMX/kg, 20 mg DDS/kg, or 80 mg AMG/kg before liver samples were taken for activity assay. For timepoints > 24 h, rats were given an additional dose of drug every 24 h.

Figure 8.  Time-course of hepatic GSH-S-Tr activity in AMG-treated rats. No significant change was observed in GST activity after treatment with 80 mg AMG/kg (solid squares) in the liver as compared to controls (open circles). For timepoints > 24 h, rats were given an additional dose of drug every 24 h (n = 2).

Discussion

If biomarkers that predict the risk that a drug candidate will cause an unacceptable risk of IDRs could be found it would have a profound effect on the process of drug development. The liver is a logical place to look because it is the site of most reactive metabolite formation, and it is the target of IDRs that are most likely to lead to drug withdrawal. Therefore, when drug candidates are screened for reactive metabolite formation, this is done almost exclusively in the liver or with hepatocytes. Accordingly, much of the screening for toxicities of drug candidates is focused on the liver.

Drugs that are primary aromatic amines represent a good starting point because this functional group is notorious for being associated with IDRs. Microarray technology represents the easiest way to examine a wide variety of possible biomarkers. The initial goal of these experiments was to verify whether the changes in hepatic gene expression induced by SMX were similar to our previous study performed in mice and whether there were changes that could be danger signals, biomarkers, or both. We chose here to focus on early time-points because subtle signs of cell stress, through changes in mRNA expression, would likely happen rapidly upon drug treatment and thus be good predictors. In comparison, effects on host immune responses would not be expected to occur—or to be detected—as early. Our findings indicate that, as in mice, SMX induced minimal hepatic gene changes in the rat. Further, of these changes, none met the strict criteria set here for denoting statistical significance.

It was surprising that there was a significant difference between the gene expression profile of SMX and DDS because they are structurally very similar with similar pharmacology, although DDS causes hemolytic anemia and methemoglobinemia and SMX does not (Uetrecht, 1989); this presumably reflects differences in the amount of redox cycling and distribution to red cells (Naisbitt et al., 1999). AMG is a first generation aromatase inhibitor used to treat estrogen-responsive breast and prostate cancer. AMG is structurally different from the other two drugs; in particular, there is no electron-withdrawing group on the aromatic ring to decrease the electron density of the aromatic amine, and this should change its redox potential. AMG was included to determine the extent of the involvement of the aromatic amine group in the induction of IDRs to determine whether the mere presence of the moiety is enough to induce similar changes. Interestingly, AMG induced the most significant number and magnitude of gene changes. This is not surprising because AMG is associated with a variety of adverse reactions in ≈ 50% of patients that take this drug, although only 1% of patients experience idiosyncratic hematologic toxicities including thrombocytopenia, leucopenia, pancytopenia, and agranulocytosis (Young et al. 1984). To some degree the similarities between DDS and AMG correlate with the fact that they are more prone to induce blood dyscrasias and bone marrow toxicity than SMX (Young et al., 1984; Coleman, 2001). Covalent adducts of SMX hydroxylamine have been found to localize on the surface of normal human epidermal keratinocytes in vitro, whereas covalent adducts of DDS hydroxylamine were found to localize intracellularly (Roychowdhury et al., 2005). These distinct covalent binding patterns may play a role in the observed differences and types of IDRs induced because intracellular adducts may have a greater ability to induce a strong immune response.

DDS and AMG induced some genes that could be considered danger signals. Our previous microarray studies on drugs including tienilic acid, carbamazepine, phenytoin, and D-penicillamine have all shown induction of cell stress through the up-regulation of genes involved with the Keap1-Nrf2-ARE pathway (Seguin et al., 2005; Pacitto et al., 2007; Lu et al., 2008), and this may potentially be a sign of danger to the immune system. The Keap1-Nrf2-ARE pathway is responsible for regulating the transcription of a variety of genes involved with detoxification and cytoprotection. The up-regulation of these genes suggests that cell stress occurred and there is an increased need to modulate potential damage that could eventually lead to an immune response.

In the present study, the same general pattern of increased expression of Keap1-Nrf2-ARE-regulated genes was observed, although mainly for DDS and AMG. Nevertheless, although the changes induced by SMX were not statistically significant, the same Keap1-Nrf2-ARE-regulated genes as DDS and AMG appeared to be slightly elevated. Txnrd is an enzyme analogous to glutathione reductase that functions to provide reducing equivalents to thioredoxin to act as an antioxidant, and it is crucial in maintaining cell redox status (Mustacich and Powis, 2000). Surprisingly, it was SMX that induced the greatest increase in Txnrd activity. Furthermore, there was a lag period between gene induction and Txnrd activity during AMG treatment. This implies that gene expression does not necessarily correlate with functional changes and that gene expression was likely induced before protein expression, which may explain the inconsistent relationship between gene and protein activity. This also raises an important issue as to whether gene changes are meaningful on their own. The activities of certain proteins are governed by post-translational modifications or by cytoplasmic ‘regulators′. In the canonical pathway, Nrf2 is kept inactivated by Keap1 in the cytoplasm, and only when Keap1 undergoes proteolytic cleavage is Nrf2 released to migrate into the nucleus where it then binds to the antioxidant response element and initiates transcription (Nguyen et al., 2009). Thus, changes in gene expression of Nrf2 may not be useful because activation is regulated post-translationally.

The acute cell response genes Sgk and Dusp1 were consistently the most down-regulated with all aromatic amine drugs tested. Sgk, a serine/threonine protein kinase involved in regulating cellular metabolism, proliferation, and differentiation, can be induced transiently by gluco-corticoids and serum (Webster et al., 1993). Dusp1, a nuclear phosphatase involved in regulating cellular signalling pathways, has been reported as inducible in early response to LPS, heat shock, and oxidative stress (Patterson et al., 2009). Down-regulation of these genes opposes what would be expected upon drug treatment; however, decreased Sgk expression upon tienilic acid treatment was attributed to the short half-life of Sgk mRNA (Webster et al., 1993; Pacitto et al., 2007). It is also plausible that a lack of signal could alert the cell to stress and initiate an immune response, although the exact mechanism is unclear.

Possibly the most interesting change was the increase in Eiih mRNA that was observed with all three drugs, although we failed to confirm increases in Eiih expression with RT-PCR for SMX. Additional testing of two other drugs associated with IDRs: clozapine and nevirapine, found a rapid increase in Eiih expression. These other drugs are not primary aromatic amines, and therefore it is possible that this is a more general biomarker of IDR risk. However, without testing a variety of drugs including negative controls that are not associated with a significant risk of IDRs it would be premature to conclude that this represents a useful biomarker of IDR risk. Unfortunately, little is known about the function of this gene. To date, there has only been one published study on Eiih that found it was induced early and acutely upon insulin treatment in rats with expression mainly localized to the liver, intestine, and islets of Langerhan, and there is some indication for its involvement in carbohydrate metabolism (Coffy et al., 2005). Its predicted protein arrangement suggests that it has the potential to be secreted or membrane bound, which implies that Eiih could possibly be a danger signal if secreted or act as a receptor-signalling complex on cellular membranes.

Although there were changes in gene expression induced by these primary aromatic amines, especially AMG, which could represent a danger signal, the most salient finding of this study is that primary aromatic amines do not appear to cause many changes in gene expression in the liver. Given that the primary aromatic amine is such a notorious structural alert and is readily oxidized to reactive metabolites in the liver and can also redox cycle, this is surprising. SMX and AMG frequently cause liver injury: the incidence of elevated γ-glutamyltransferase is greater than 60% in AMG-treated patients (Nagel et al., 1982). However, SMX-induced liver injury is usually part of a more generalized hypersensitivity IDR, and the more severe AMG-induced liver injury is more commonly cholestatic rather than hepatocellular. The most likely explanation for these results is that, although this functional group is readily oxidized to reactive metabolites in the liver, the liver is well equipped to deal with reactive metabolites. For example, in the case of an animal model of nevirapine-induced skin rash, there was more covalent binding in the liver than in the skin, but there were many more significant changes in gene expression in the skin (> 400, unpublished observation, manuscript in preparation). It is likely that what makes the aromatic amine functional group a structural alert is that it can also be oxidized by myeloperoxidase, which is present in antigen presenting cells. The formation of reactive metabolites by antigen presenting cells has been shown to lead to their activation (Elsheikh et al., 2010), and this is likely to be a strong stimulus for the induction of an immune response. Specifically, dendritic cells treated with SMX and nitroso metabolite were found to increase expression of the co-stimulatory molecule CD40 (Sanderson et al., 2007). Similarly, keratinocytes were found to up-regulate their expression of heat shock protein 70 (which could be a potential danger signal) upon treatment with the hydroxylamine metabolite of SMX (Khan et al., 2007).

Even in the case of immune-mediated liver injury, it appears that activation of the immune system outside of the liver is necessary in order to cause an immune response in the liver that results in injury (Bowen et al., 2005). It is notable that all three of these drugs are associated with a relatively high incidence of idiosyncratic agranulocytosis; this presumably involves oxidation of the drug to reactive metabolites by myeloperoxidase, the major oxidizing enzyme in neutrophils. This clinical picture is most consistent with activation of the immune system outside of the liver with the immune response sometimes extending to the liver. Therefore, the present focus on the liver by the pharmaceutical industry may miss signals that predict IDR risk, especially for drugs that also cause IDRs outside of the liver. In particular, bioactivation of easily-oxidized drugs, such as those containing aromatic amines, by myeloperoxidase in neutrophils and antigen-presenting cells, may play an important role in the induction of many types of IDRs.

Appendices

Appendix A. SMX-induced hepatic gene changes at 12 h (fold-change between treated and controls).

RefSeq	Gene symbol	Gene name	Fold change	p value
Up-regulation
NM_001009965	Eiih	hepatic protein EIIH	2.86	0.0227
J03863	Sds	serine dehydratase	2.15	0.0237
NM_021660	Ihpk2	inositol hexaphosphate kinase 2	1.91	0.0206
NM_030845	Cxcl1	chemokine (C-X-C motif) ligand 1	1.67	0.0476
NM_198776	MGC72973	hemoglobin, beta, adult major chain	1.67	0.1509
NM_022671	Onecut1	one cut homeobox 1	1.61	0.3665
NM_033234	Hbb	hemoglobin, beta	1.54	0.1647
NM_001034125	Per1	period homolog 1 (Drosophila)	1.53	0.1517
NM_001013853	LOC287167	globin, alpha	1.52	0.1826
NM_001010946	LOC288741	huntingtin interactor protein E	1.51	0.0480
NM_012571	Got1	glutamate oxaloacetate transaminase 1	1.51	0.0035
NM_001007146	Tob2	transducer of ERBB2, 2	1.50	0.0456
Down-regulation
NM_017136	Sqle	squalene epoxidase	−1.51	0.0336
NM_017268	Hmgcs1	3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1	−1.52	0.0098
NM_080886	Sc4mol	sterol-C4-methyl oxidase-like	−1.55	0.0196
NM_057107	Acsl3	acyl-CoA synthetase long-chain family member 3	−1.57	0.0374
NM_013060	Id2	inhibitor of DNA binding 2	−1.58	0.0221
NM_053380	Slc34a2	solute carrier family 34 (sodium phosphate), member 2	−1.59	0.0076
NM_001012072	Ppp1r3c	protein phosphatase 1, regulatory sub-unit 3C	−1.61	0.0383
NM_001008352	Pmvk	phosphomevalonate kinase	−1.62	0.0108
NM_001006995	Acat2	acetyl-Coenzyme A acetyltransferase 2	−1.63	0.0259
NM_001106882	Unc5cl	unc-5 homolog C (C. elegans)-like	−1.66	0.0357
NM_019237	Pcolce	procollagen C-endopeptidase enhancer protein	−1.67	0.0075
NM_198738	Psat1	phosphoserine aminotransferase 1	−1.67	0.0772
NM_032063	Dll1	delta-like 1 (Drosophila)	−1.68	0.0029
NM_212505	Ier3	immediate early response 3	−1.69	0.0022
NM_053328	Bhlhb2	basic helix-loop-helix domain containing, class B2	−1.72	0.0969
NM_001013071	Tm7sf2	transmembrane 7 superfamily member 2	−1.76	0.0139
NM_206950	Mig12	MID1 interacting G12-like protein	−1.76	0.0010
NM_145677	Slc25a25	solute carrier family 25 (mitochondrial, phosphate), member 25	−1.86	0.0162
NM_001033066	Ddhd1	DDHD domain containing 1	−1.87	0.0323
NM_031327	Cyr61	cysteine rich protein 61	−1.87	0.0025
NM_021835	Jun	jun proto-oncogene	−1.97	0.0019
NM_001009632	G0s2	G0/G1 switch gene 2	−2.11	0.0201
NM_019232	Sgk	serum/glucocorticoid regulated kinase 1	−2.70	0.0023
NM_053769	Dusp1	dual specificity phosphatase 1	−3.58	0.0020

Data shown are gene changes ≥ 1.50-fold expression. p-values are unadjusted (n = 4).

Appendix B. DDS-induced hepatic gene changes at 12 h (fold-change between treated and controls).

RefSeq	Gene symbol	Gene name	Fold-change	p-value
Up-regulation
NM_013215	Akr7a3	aldo-keto reductase family 7, member A3 (aflatoxin aldehyde reductase)	5.38	2.89 × 10^−7
NM_001009965	Eiih	hepatic protein EIIH	4.23	0.0038
NM_001013181	Zbtb16	zinc finger and BTB domain containing 16	3.55	0.0154
NM_001009920	Yc2	glutathione S-transferase,Yc2 sub-unit	3.19	7.52 × 10^−7
NM_001012111	Lpin1	lipin 1	2.59	0.0200
NM_001014162	RGD1309578	acyl-CoA synthetase medium-chain family member 5	2.37	0.0329
NM_022407	Aldh1a1	aldehyde dehydrogenase family 1, member A1	2.33	0.0049
NM_058208	Socs2	suppressor of cytokine signaling 2	2.08	0.1430
NM_012603	Myc	myelocytomatosis oncogene	2.00	0.0238
BC89763	Nr1i3	nuclear receptor subfamily 1, group I, member 3	1.95	0.0052
NM_145775	Nr1d1	nuclear receptor subfamily 1, group D, member 1	1.94	0.0885
NM_001025623	Vnn1	vanin 1	1.92	0.0957
NM_052798	Zfp354a	zinc finger protein 354A	1.85	0.1089
NM_198776	MGC72973	hemoglobin, beta, adult major chain	1.80	0.1046
Down-regulation
NM_001012072	Ppp1r3c	protein phosphatase 1, regulatory sub-unit 3C	−1.81	0.0129
NM_053380	Slc34a2	solute carrier family 34 (sodium phosphate), member 2	−1.82	0.0014
NM_032063	Dll1	delta-like 1 (Drosophila)	−1.85	0.0008
NM_031678	Per2	period homolog 2 (Drosophila)	−1.86	0.0074
NM_012565	Gck	glucokinase	−1.89	0.0450
NM_001013175	Apol3	apolipoprotein L, 3	−1.90	0.0005
NM_053328	Bhlhb2	basic helix-loop-helix domain containing, class B2	−1.93	0.0500
NM_138912	Ppp1r3b	protein phosphatase 1, regulatory sub-unit 3B	−1.96	0.0038
NM_001009632	G0s2	G0/G1 switch gene 2	−2.03	0.0263
NM_019138	Cyp7b1	cytochrome P450, family 7, sub-family b, polypeptide 1	−2.05	0.0286
NM_012703	Thrsp	thyroid hormone responsive protein	−2.08	0.0023
NM_001003706	LOC360228	WDNM1 homolog	−2.17	0.0022
NM_021835	Jun	jun proto-oncogene	−2.49	0.0002
NM_031327	Cyr61	cysteine rich protein 61	−2.50	0.0001
NM_001033066	Ddhd1	DDHD domain containing 1	−2.50	0.0040
NM_206950	Mig12	MID1 interacting protein 1	−2.59	1.06 × 10^−5
NM_145677	Slc25a25	solute carrier family 25 (mitochondrial, phosphate), member 25	−2.67	0.0008
NM_053769	Dusp1	dual specificity phosphatase 1	−3.01	0.0052
NM_019232	Sgk	serum/glucocorticoid regulated kinase 1	−3.72	0.0003

Data shown are gene changes ≥ 1.80-fold expression. p-values are unadjusted (n = 4).

Appendix C. AMG-induced hepatic gene changes at 12 h (fold-change between treated and controls).

RefSeq		Gene symbol		Gene name	Fold-change	p-value
Up-regulation
NM_001009965	Eiih		hepatic protein EIIH		12.94	0.0072
NM_145775	Nr1d1		nuclear receptor sub-family 1, group D, member 1		5.77	0.0291
NM_031576	Por		P450 (cytochrome) oxidoreductase		4.78	0.0003
NM_133401	Abcb1a		ATP-binding cassette, sub-family B, member 1A		4.50	0.0001
NM_133586	Ces2		carboxylesterase 2C		4.38	0.0051
NM_022407	Aldh1a1		aldehyde dehydrogenase family 1, member A1		4.15	0.0121
NM_001008321	Gadd45b		growth arrest and DNA-damage-inducible, beta		3.81	0.0009
NM_001008363	Zfand2a		zinc finger, AN1-type domain 2A		3.77	0.0076
NM_022927	Mid1		midline 1		3.71	0.0021
NM_001009474	Pir		pirin (iron-binding nuclear protein)		3.67	2.98 × 10^−5
NM_173295	Udpgtr2		UDP glucuronosyltransferase 2 family, polypeptide B1		3.59	0.0008
NM_144743	Ces6		carboxylesterase 2A		3.48	3.33 × 10^−5
NM_013105	Cyp3a23		cytochrome P450, family 3, subfamily a, polypeptide 23/polypeptide 1		3.37	3.79 × 10^−5
NM_144755	Trib3		tribbles homolog 3 (Drosophila)		2.79	0.0329
NM_012600	Me1		malic enzyme 1, NADP(+)-dependent, cytosolic		2.76	0.0076
NM_033352	Abcd2		ATP-binding cassette, sub-family D (ALD), member 2		2.57	0.0039
NM_001047090	Aph1b		anterior pharynx defective 1b homolog (C. elegans)		2.56	0.0291
NM_175761	Hspca		heat shock protein 90, alpha (cytosolic), class A member 1		2.52	0.0145
NM_013063	Parp1		poly (ADP-ribose) polymerase 1		2.46	0.0002
NM_001001509	Hdmcp		solute carrier family 25, member 47		2.46	0.2365
NM_001007706	RGD1359191		tRNA methyltransferase 61 homolog A (S. cerevisiae)		2.40	0.1436
NM_001025623	Vnn1		vanin 1		2.35	0.3351
NM_173116	Sgpl1		sphingosine-1-phosphate lyase 1		2.34	0.0246
NM_001011901	Hsph1		heat shock 105/110 protein 1		2.34	0.0048
NM_001034090	Ephx1		epoxide hydrolase 1, microsomal		2.33	0.0009
NM_001109022	LOC368066		indolethylamine N-methyltransferase		2.32	0.1376
NM_013215	Akr7a3		aldo-keto reductase family 7, member A3 (aflatoxin aldehyde reductase)		2.23	0.0305
NM_199103	Ikbkg		inhibitor of kappa light polypeptide gene enhancer in B-cells, kinase gamma		2.23	0.0006
NM_001009920	Yc2		glutathione S-transferase Yc2 subunit		2.23	0.0066
NM_031614	Txnrd1		thioredoxin reductase 1		2.23	0.0003
NM_131906	Slco1a4		solute carrier organic anion transporter family, member 1A2		2.22	0.0702
NM_001005384	Osmr		oncostatin M receptor		2.21	0.0501
NM_212504	Hspa1b		heat shock 70kD protein 1B		2.20	0.0130
NM_178091	Insig2		insulin induced gene 2		2.20	0.0018
M58040	Tfrc		transferrin receptor		2.19	0.1851
NM_024127	Gadd45a		growth arrest and DNA-damage-inducible, alpha		2.19	0.1613
NM_144752	Oas1b		2-5 oligoadenylate synthetase 1B		2.14	0.0020
NM_145091	Pdp2		pyruvate dehyrogenase phosphatase catalytic subunit 2		2.09	0.0115
NM_001007687	Cndp1		carnosine dipeptidase 1 (metallopeptidase M20 family)		2.08	0.0635
NM_001010946	LOC288741		huntingtin interactor protein E		2.08	0.0782
NM_053906	Gsr		glutathione reductase		2.06	0.0000
NM_001013181	Zbtb16		zinc finger and BTB domain containing 16		2.06	0.5563
NM_201424	Ugt1a3		UDP glycosyltransferase 1 family polypeptide A3		2.06	2.86 × 10^−5
AF082126	Ahr		aryl hydrocarbon receptor		2.05	0.0733
NM_019283	Slc3a2		solute carrier family 3 (activators of dibasic and neutral amino acid transport), member 2		2.05	0.0485
NM_001008301	Usp14		ubiquitin specific peptidase 14 (tRNA-guanine transglycosylase)		2.04	0.0135
ENSRNOT00000056174	Usp18		ubiquitin specific peptidase 18		2.04	0.0300
NM_153312	Cyp3a2		cytochrome P450, family 3, sub-family a, polypeptide 2		2.03	0.0116
NM_001007690	Fbxo30		F-box protein 30		2.01	0.0246
NM_053612	Hspb8		heat shock protein B8		2.01	0.0036
Down-regulation
NM_031678	Per2		period homolog 2 (Drosophila)		−2.00	0.1059
NM_001003706	LOC360228		WDNM1 homolog		−2.00	0.1222
NM_001108118	Itga5		integrin, alpha 5 (fibronectin receptor, alpha polypeptide)		−2.01	0.0205
NM_031815	Inhbe		inhibin beta E		−2.01	0.0132
NM_001011985	Mknk2		MAP kinase-interacting serine/threonine kinase 2		−2.02	0.0612
NM_080782	Cdkn1a		cyclin-dependent kinase inhibitor 1A		−2.03	0.3319
NM_198780	Pck1		phosphoenolpyruvate carboxykinase 1 (soulable)		−2.05	0.0272
NM_012941	Cyp51		cytochrome P450, subfamily 51		−2.07	0.0346
NM_001109430	LOC680883		leucine-rich repeats and transmembrane domains 2		−2.07	0.0635
NM_012508	Atp2b2		ATPase, Ca++ transporting, plasma membrane 2		−2.08	0.0006
NM_001106873	LOC301113		solute carrier family 25 (mitochondrial; phosphate), member 23		−2.11	0.0257
NM_001034081	Paqr7		progestin and adipoQ receptor family member VII		−2.14	0.0071
NM_178096	Nrep		neuronal regeneration related protein		−2.15	0.0273
ENSRNOT00000027756	Fads2		fatty acid desaturase 2		−2.20	0.0116
NM_001006995	Acat2		acetyl-Coenzyme A acetyltransferase 2		−2.28	0.0526
BC107921	RGD1309879		family with sequence similarity 89, member A		−2.31	0.0121
NM_022392	Insig1		insulin induced gene 1		−2.32	0.0272
NM_012703	Thrsp		thyroid hormone responsive		−2.33	0.0441
NM_198790	Rgsl2h		regulator of G-protein signaling like 2 homolog (mouse)		−2.34	0.0385
NM_001033066	Ddhd1		DDHD domain containing 1		−2.41	0.1299
NM_001004096	Reg4		regenerating islet-derived family, member 4		−2.47	0.0026
NM_022392	Insig1		insulin induced gene 1		−2.50	0.0298
NM_022280	Lrat		lecithin-retinol acyltransferase (phosphatidylcholine-retin)		−2.59	0.0212
U11038	Lox		lysyl oxidase		−2.70	0.3845
NM_001109912	Tsc22d1		TSC22 domain family, member 1		−2.72	0.0080
NM_013144	Igfbp1		insulin-like growth factor binding protein 1		−2.73	0.5886
ENSRNOT00000024658	Slc28a2		solute carrier family 28 (sodium-coupled nucleoside transporter), member 2		−2.80	0.0076
NM_053922	Acacb		acetyl-Coenzyme A carboxylase beta		−2.80	0.0197
NM_001008352	Pmvk		phosphomevalonate kinase		−2.83	0.0063
NM_053380	Slc34a2		solute carrier family 34 (sodium phosphate), member 2		−3.02	0.0025
NM_019138	Cyp7b1		cytochrome P450, family 7, subfamily b, polypeptide 1		−3.04	0.0815
NM_001012621	Slc13a4		solute carrier family 13 (sodium/sulfate symporters), member 4		−3.15	0.0023
NM_139258	Bmf		Bcl2 modifying factor		−3.25	0.0123
NM_053769	Dusp1		dual specificity phosphatase 1		−3.61	0.0712
NM_145677	Slc25a25		solute carrier family 25 (mitochondrial, phosphate), member 25		−4.05	0.0076
NM_019232	Sgk		serum/glucocorticoid regulated kinase 1		−5.37	0.0064

Data shown are gene changes ≥ 2-fold expression. p-values are adjusted using FDR (n = 4).

Acknowledgments

We thank Xiaochu Zhang for providing the nevirapine liver samples. This research was funded by the Canadian Institutes of Health Research. J.P.U. is the Canada Research Chair in Adverse Drug Reactions.

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.