2018 Volume 43 Issue 2 Pages 75-87
Several drugs have been withdrawn from the market or restricted to avoid unexpected adverse outcomes. Drug-induced liver injury (DILI) is a serious issue for drug development. Among DILIs, idiosyncratic DILIs have been a serious problem in drug development and clinical uses. Idiosyncratic DILI is most often unrelated to pharmacological effects or the dosing amount of a drug. The number of drugs that cause idiosyncratic DILI continue to grow in part because no practical preclinical tests have emerged that can identify drug candidates with the potential for developing idiosyncratic DILIs. Nevertheless, the implications of drug metabolism-related factors and immune-related factors on idiosyncratic DILIs has not been fully clarified because this toxicity can not be reproduced in animals. Therefore, accumulated evidence for the mechanisms of the idiosyncratic toxicity has been limited to only in vitro studies. This review describes current knowledge of the effects of cytochrome P450 (CYP)-mediated metabolism and its detoxification abilities based on studies of idiosyncratic DILI animal models developed recently. This review also focused on antiepileptic drugs, phenytoin (diphenyl hydantoin, DPH) and carbamazepine (CBZ), which have rarely caused severe adverse reactions, such as fulminant hepatitis, and have been recognized as sources of idiosyncratic DILI. The studies of animal models of idiosyncratic DILIs have produced new knowledge of chronic administration, CYP inductions/inhibitions, glutathione contents, and immune-related factors for the initiation of idiosyncratic DILIs. Considering changes in the drug metabolic profile and detoxification abilities, idiosyncratic DILIs caused by antiepileptic drugs will lead to understanding the mechanisms of these DILIs.
Adverse drug reactions can be defined as any undesirable response from therapeutic use. Adverse outcomes can be divided into those that are dose-dependent and largely predictable from the known pharmacological properties of the compound and those that are dependent on characteristics that are unique to susceptible individuals or that are idiosyncratic in nature (Park et al., 1992). In the latter category, the term idiosyncratic in the context of an adverse drug event refers to the determinants of susceptibility that are unique to an individual and therefore do not imply any particular mechanism (Glauser, 2000; Ju and Uetrecht, 2002; Knowles et al., 2000). The frequency of occurrence of idiosyncratic adverse drug reactions is very low (Ju and Uetrecht, 2002), and these reactions are often not detected until the drug has exposure in a large patient population. In addition, these reactions have generally never been predicted in standard regulatory animal toxicity studies.
Idiosyncratic adverse drug reactions can affect any organ, including the liver (Srivastava et al., 2010), skin (Wolf et al., 2005; Sevketoglu et al., 2009; Svensson, 2009), and bone marrow (Handoko et al., 2006; Andrès et al., 2008), and these organs have been considered as the most common targets. Among these, drug-induced liver injury (DILI) is the most frequent reason for drug withdrawal and is also a major cause of attrition in drug development (Kaplowitz, 2004). Idiosyncratic DILI is most often unrelated to a drug’s pharmacological effects and dosing amount. The number of drugs that cause idiosyncratic DILI continues to grow in part because there are no practical preclinical tests that can identify drug candidates with the potential for developing idiosyncratic DILI (Kaplowitz, 2005). This problem is considered to be based on a lack of understanding of the mechanisms for these reactions because, importantly, most idiosyncratic DILIs cannot be reproduced in animal models. This failure would contribute to hampering the understanding of the mechanism of idiosyncratic DILI development. Although several hypotheses to explain them have emerged over the years, the reactions have remained poorly understood. One possibility for a crucial factor of idiosyncratic DILI is drug metabolism-related factors (Amacher, 2006; Attia, 2010).
The liver is exposed to relatively high concentrations of drugs and their metabolites after oral administration. The basic principle of drug metabolism is to convert a lipophilic drug or xenobiotic to a hydrophilic metabolite to accelerate excretion from the body. However, some of the drugs or xenobiotics may be activated to chemically electrophilic reactive metabolites. This bioactivation is known to be the initial event in many drug-induced toxicities, including DILI (Amacher, 2006; Attia, 2010). However, the relationship between the development of idiosyncratic DILI, drug metabolism-related factors (including reactive metabolite formation) has not been fully clarified. One of the causes is considered to be the lack of an animal model that shows a phenotype for idiosyncratic toxicity.
Advances in understanding of the mechanisms of idiosyncratic DILI have been hampered by the lack of animal models. Acetaminophen-induced liver injury in mice is the only widely studied animal model of DILI and is probably the only reliable model used thus far for investigating the mechanisms of DILI. Acetaminophen-induced liver injury is classified as a type of “intrinsic DILIs” which are dose-related; that is, the toxic response becomes more severe with increasing exposure (Roth and Ganey, 2010). Intrinsic DILI is readily reproducible in animals and occurs dose-dependently at sublethal doses. However, this widely studied model of intrinsic DILI alone does not encompass the all mechanistic features of DILI development in patients, especially idiosyncratic DILI. Idiosyncratic DILIs present themselves very differently than intrinsic ones; they happen in a minority of patients, with variable time of onset and no obvious relationship to drug dose, and they are not reproducible in usual animal tests. Therefore, understanding of mechanisms of idiosyncratic DILI had been limited by in part the result of a lack of animal models. Thus, it is critical to have several idiosyncratic DILI animal models.
From the late 2000s, several animal models of idiosyncratic DILI have been developed as typified by diclofenac (Yano et al., 2012), flucloxacillin (Takai et al., 2015), dicloxacillin (Higuchi et al., 2011), carbamazepine (CBZ, Higuchi et al., 2012) and phenytoin (diphenyl hydantoin (DPH), Sasaki et al., 2013). Mostly, these models have been established by using wild type mice and rats with a non-lethal dose of drug. These animal models contribute more directly to elucidating the production of reactive metabolites and immune-related factors, and their contribution goes toward development of idiosyncratic DILI than past events. In fact, according to the above reports, the drug metabolism-related factors (reactive metabolite formation) and inflammation-related as well as immune-related factors have been raised as onset and exacerbation factors, respectively, for developing idiosyncratic DILIs.
For the purpose of this review, we clarify the relationship between drug-metabolism, reactive metabolite formation and risk of antiepileptic drug-induced liver injury, which is considered to be idiosyncratic toxicity with reference to the research results of accumulated reports. Furthermore, the importance of considering changes in drug metabolism profiles and reactive metabolites formation is discussed. From these findings, we provide consideration for individual differences in the susceptibility for developing antiepileptic drug-induced liver injury development from the viewpoint of drug metabolisms.
The antiepileptic drug hypersensitivity reaction is an adverse drug reaction associated with the aromatic antiepileptic drugs phenytoin (DPH), carbamazepine (CBZ), phenobarbital (PB) and primidone (Shear and Spielberg, 1988; Handfield-Jones et al., 1993). A hypersensitivity reaction associated with DPH therapy was reported in 1950 (Chaeiken et al., 1950); however, another 20 years passed before a consensus was reached that DPH could produce characteristic hypersensitivity reactions, such as fever, peripheral leukocytosis and sometimes life-threatening hepatic necrosis (Licata and Louis, 1996). Other aromatic antiepileptic drugs, such as CBZ and PB, could also be associated with hypersensitivity and were termed as antiepileptic drugs that induce the hypersensitivity reaction in 1988 (Shear and Spielberg, 1988; Schlienger and Shear, 1998).
Antiepileptic drug-associated hypersensitivity appears within 2-8 weeks after initiation of pharmacotherapy. In the early stage, fever, followed by cutaneous reaction, lymphadenopathy, and pharyngitis may develop within 1-2 days. These events, followed by various internal organ failures, most commonly the liver, may occur (Shear and Spielberg, 1988; Vittorio and Muglia, 1995). The most prominent events include hepatitis, eosinophilia, blood dyscrasias, and nephritis (Shear and Spielberg, 1988; Flowers et al., 1987).
The incidence of a hypersensitivity reaction associated with DPH and CBZ was estimated to be 2.3-4.5 per 10,000 and 1.0-4.1 per 10,000, respectively (Tennis and Stern, 1997). Among them, the rate of liver failure involvement in patients has been reported to range from 34% (Haruda, 1979) to 94% (Powers and Carson, 1987). Thus, DPH- and CBZ-induced severe liver injuries have been considered idiosyncratic DILI. In addition, mortality of patients who developed DPH- or CBZ-induced liver injury is up to about 20%, and liver injury cannot be predicted before and under the medication until onset of the liver injury (Schlienger and Shear, 1998). Therefore DPH- and CBZ-induced liver injuries have been a serious issue in clinical settings over a long time. The hypersensitivity reaction accompanied by a serious liver injury can be found in all of the older aromatic antiepileptic drugs. Among them, DPH and CBZ have had the most cases reported (Schlienger and Shear, 1998). This review focuses on the cases of DPH- and CBZ-induced idiosyncratic liver injury focusing on the role of cytochrome P450-mediated metabolism and involvement of reactive metabolite formation on antiepileptic drug-induced liver injuries.
Although there has been a lack of direct evidence for the involvement of reactive metabolites in the development of DPH-induced liver injury, the capabilities of cytochrome P450 (CYP)2C9 and CYP3A4 producing electrophilic metabolites were investigated in in vitro studies (Munns et al., 1997; Roy and Snodgrass, 1988, 1990).
Major metabolic pathways of DPH in humans have been reported (Fig. 1) (Maguire, 1988; Szabo et al., 1990; Komatsu et al., 2000). The studies using cDNA-expressed P450 enzymes have revealed that 4’-hydroxy DPH (4’-HPPH), which is the major metabolite of DPH identified in humans and rats, was formed by CYP2C9- and CYP2C11-mediated metabolism in humans (Komatsu et al., 2000) and rats (Yamazaki et al., 2001), respectively. 4’-HPPH is formed via an arene oxide metabolite; however, this intermediate is chemically unstable and is speculated to mediate protein adducts by forming covalent bonds in rat and mouse liver microsomes (Roy and Snodgrass, 1988, 1990).
Proposed major metabolic pathways of DPH in humans. The structure within the brackets is inferred from the products. CYP, cytochrome P450; Cys, cysteine; DPH, phenytoin; GGT, γ-glutamyl transferase; Glu, glucuronide; GSH, glutathione; mEH, microsomal epoxide hydrolase; NAC, N-acetylcysteine; UGT, UDP-glucuronosyltransferase.
The arene oxide intermediate is transferred into a stable dihydrodiol metabolite that would be catalyzed by microsomal epoxide hydrolase (mEH). The study using human liver microsomes has indicated that 4’-HPPH can be further metabolized into a catechol metabolite and that this catechol is spontaneously oxidized into an unstable electrophilic intermediate, o-quinone metabolite (Munns et al., 1997; Leeder et al., 1992; Cuttle et al., 2000). o-Quinone metabolite is also considered to be capable of covalently binding to protein. The formation of catechol followed by o-quinone is catalyzed by CYP-mediated metabolism in humans and rats and covalent bonding to human liver microsomal protein (Munns et al., 1997; Roy and Snodgrass, 1988, 1990).
DPH and 4’-HPPH can also be converted into glucuronidation catalyzed by UDP-glucuronosyltransferase (UGT), N-glucuronide (DPH-Glu, Smith et al., 1977) and 4’-HPPH O-glucuronide (Maynert, 1960), and these were detected in human urine samples. However, these glucuronides are chemically stable and have not been reported as being associated with toxicity.
Taken together, putative reactive metabolites of DPH are considered to be the same arene oxide and the o-quinone metabolites, and both are formed by CYP-mediated metabolism. However, most of the accumulated evidence was based on in vitro studies and implications for idiosyncratic DILI had not been demonstrated because there were no suitable animal models to demonstrate the involvement of reactive metabolites on the development of DPH-induced liver injury directly.
The novel animal model of DPH-induced liver injury was developed using mice that received repeated administration of DPH for 5 days (Sasaki et al., 2013). As we refer to in the report, 5 days of repeated administration of DPH was essential to developing liver injury, and the injuries were exacerbated by co-treatment with the γ-glutamylcysteine synthetase (γ-GCS) inhibitor, l-buthionine S,R-sulfoximine (BSO), which reversibly depletes organ glutathione (GSH) contents (Watanabe et al., 2003). This model has slight elevation of plasma alanine aminotransferase (ALT) levels at 4 days of continuous DPH administration (approximate ALT: 200 U/L) that is remarkably increased at the 5th day of administration (approximate ALT: 700-1,000 U/L). The liver injury was histopathologically confirmed and typified by degradation, ballooning and apoptosis of hepatocytes. Interestingly, the elevation of ALT and histopathological features of liver injury are attenuated by co-treatment with the Cyps inhibitor, 1-aminobenzotriazole (ABT, approximate ALT: 400-600 U/L). Taken together, DPH-induced liver injury likely occurs by CYP-mediated metabolism and GSH is critical for detoxification of the reactive metabolite(s).
Numerous reports indicate that CYP-dependent covalent bond formation of DPH is inhibited in the presence of GSH or N-acetylcysteine (NAC) in vitro and in vivo (Munns et al., 1997; Roy and Snodgrass, 1988, 1990). These previous observations correspond to the study of the DPH-induced liver injury mouse model (Sasaki et al., 2013). Although the presence of thiol conjugates of DPH had been suggested from previous studies (Munns et al., 1997; Roy and Snodgrass, 1988, 1990), the existence of thiol-conjugated metabolites has not been proved.
Our previous report first identified GSH-, NAC-, and cysteine (Cys)-conjugated metabolites from the plasma and bile of the DPH-induced liver injury model mouse (Sasaki et al., 2015). Changes in these plasma thiol-conjugate concentrations are correlated with those of plasma ALT levels. In other words, these metabolites appear only after the onset of liver injury. This observation strongly suggests that the reactive metabolite, which can be converted into a thiol-conjugate, is involved in the development of liver injury.
The detected thiol-conjugates are thought to be metabolites in which the thiol is directly bonded to the benzene ring of DPH (Sasaki et al., 2015). In the case where allene oxide, presumed as a reactive metabolite of DPH, is conjugated with GSH, GSH binds to the 3rd or 4th position of the benzene ring as well as to the metabolite in which GSH or the hydroxyl group is bound to the 3rd or 4th, respectively, and will be converted. After that, it was inferred that the hydroxyl group of the benzene ring was rapidly removed by a dehydration reaction, resulting in GSH bonding to the benzene ring of DPH. Benzene and thienylic acid produce a GSH conjugate with a similar mechanism of epoxy degree, including allene oxide, that has been reported (Ross, 1996; Nishiya et al., 2008). When GSH is added to the o-quinone form, it is presumed that GSH binds to the 2nd or 5th position of the benzene ring of DPH, followed by the quinone being reduced. As a result, it will form a stable catechol metabolite; however, no mass spectra suggesting the catechol body were observed in tandem mass spectrometry analyses (Sasaki et al., 2015). This result suggests that a metabolic pathway prior to 4’-HPPH formation rather than after 4’-HPPH oxidation would be critical for reactive metabolite formation. Therefore, it is considered that CYP-mediated metabolism via formation of the allene oxide metabolite is involved in the onset of DPH-induced liver injury.
CBZ has been widely used for the treatment of partial seizures. It is reported that patients who developed CBZ-induced serious liver injury were under drug treatment for an average of 30 weeks (Björnsson, 2008), which suggests that long-term CBZ treatment is a risk factor for liver injury development. Similarly, with DPH, liver injury caused by CBZ has been rarely observed and is often fatal (Schlienger and Shear, 1998). Therefore CBZ-induced liver injury has been classified into idiosyncratic DILI. It is also known that some drugs can change drug metabolism profiling by inducing or inhibiting drug-metabolizing enzyme activities. In the case of CBZ, CYP1A, CYP2A, CYP2B, CYP2C, and CYP3A can be induced by CBZ treatment in both humans and rats (Tateishi et al., 1999; Oscarson et al., 2006). Thus, involvement of the drug metabolism in the development of liver injury has been suggested. The association of electrophilic metabolite(s) on developing CBZ-induced liver injury has been suggested. For example, GSH or NAC can reduce the irreversible binding of [C14]-CBZ to human and mouse microsomes (Pirmohamed et al., 1992; Lillibridge et al., 1996). Several reactive metabolites of CBZ have been speculated. The proposed metabolic pathways of CBZ in humans are described in Fig. 2.
Proposed major metabolic pathways of CBZ in humans. The structure within the brackets is inferred from the products. CBZ, carbamazepine; CYP, cytochrome P450; GSH, glutathione.
CBZ is mainly metabolized to CBZ 10,11-epoxide, which is a pharmacologically active metabolite, and it is catalyzed by CYP3A4 and CYP2C8 (Kerr et al., 1994). CBZ 10,11-epoxide is further metabolized to trans-10,11-dihydroxy CBZ by mEH (Mather and Levy, 2000). CBZ can also be metabolized to 2-hydroxy CBZ and 3-hydroxy CBZ by several CYPs in an in vitro study (Pearce et al., 2002). These hydroxyl metabolites have been considered to be candidates for the precursors of reactive metabolites. In particular, the in vitro study indicates that 2-hydroxy-CBZ and/or 3-hydroxy-CBZ can be further metabolized to 2,3-dihydroxy CBZ (Lertratanangkoon and Horning, 1982), which has been proposed to form an o-quinone metabolite (Pearce et al., 2002). Also, 2-hydroxy CBZ has been shown to be converted into 2-hydroxyiminostilbene (Pearce et al., 2005), which can be readily oxidized to an iminoquinone species (Ju and Uetrecht, 1999). As described in the above in vitro studies, although there are many findings about the relationship between CBZ-induced toxicity and drug metabolism, the relationship between bioactivation and the development of CBZ-induced liver injury has not been fully investigated. One of the reasons is that the appropriate animal model for CBZ-induced liver injury did not exist. Therefore, it was not possible to analyze the relationship between liver injury and CBZ metabolism.
In the first half of 2010, animal models of CBZ-induced liver injury were reported in mice (Higuchi et al., 2012) and rats (Iida et al., 2015). The involvement of drug metabolism in the development of CBZ-induced liver injury has been indicated. In the rat model, repeated administration of CBZ caused severe liver injury, yet single administration failed to produce liver injury even under GSH-depleted conditions (Iida et al., 2015). Similar to the rat model, repeated administration is required for developing liver injury in a mouse model (Higuchi et al., 2012). The hepatic CYP3A contents and its enzyme activities were increased in parallel with the administration period. In addition, CYP3A inhibitor-treatment strongly suppressed the development of liver injury in rats (Iida et al., 2015), whereas in the mouse model, Cyp3a inhibitor co-treatment exacerbated liver injury accompanied by acceleration of the metabolic pathway toward 2-hydroxy and 3-hydroxy CBZ formations (Higuchi et al., 2012). Thus, CYP3A-mediated metabolism and the metabolic pathways toward 2-hydroxy and 3-hydroxy CBZ formations are critical factors for liver injury development.
In reports from both rat and mouse models, electrophilic metabolite(s) formation is considered to be a crucial step in the development of liver injury because hepatic GSH is reduced by repeated administration of CBZ and treatment with the GSH depletor BSO exacerbates the CBZ-induced liver injury in the rat model (Iida et al., 2015). In the rat model, reducing the 2-hydroxy CBZ concentration in plasma is associated with the degree of liver injury. This phenomenon can be explained by further metabolism of 2-hydroxy CBZ forming reactive metabolite(s). In particular, in vitro studies revealed that 2-hydroxy CBZ can be further metabolized to 2-hydroxyiminostilbene (Pearce et al., 2005) followed by oxidized to an iminoquinone species (Ju and Uetrecht, 1999). In addition, 2-hydroxy CBZ can be a precursor of CBZ-o-quinone, which has an electrophilic property (Pearce et al., 2002). The iminoquinone can react with GSH or NAC to form stable thiol-conjugates, which was detected in the urine of patients who were administered CBZ (Ju and Uetrecht, 1999; Pearce et al., 2005). Therefore, enhancement of the metabolic pathway to 2-hydroxy CBZ followed by the iminoquinone species or o-quinone metabolite formation is a key factor for liver injury development. The mouse model also provides supporting evidence for CBZ o-quinone formation via enhancement of the metabolic pathway to 3-hydroxy CBZ, which is involved in the development of liver injury (Higuchi et al., 2012).
In vitro cytotoxic responses of reactive metabolite(s) have been demonstrated in both DPH and CBZ, and the results suggest that decreased detoxification capacity is a possible determinant of susceptibility to DPH- and CBZ-induced severe adverse reactions. Deficiency of mEH activity increased cytotoxicity in an in vitro assay for DPH and CBZ (Pirmohamed et al., 1991, 1992; Spielberg et al., 1981), which suggests that mEH enzyme activity can be associated with susceptibility to developing DPH- and CBZ-induced liver injury. However, no genetic variation that alters the structure or enzyme activity of mEH has been identified since 1994 (Gaedigk et al., 1994). From this fact, decreases in mEH activity would be an unlikely direct cause for DPH- or CBZ-induced liver injury.
GSH has been shown to be a crucial factor implicated in the detoxification of electrophilic reactive metabolites as typified by acetaminophen (Jollow et al., 1973; Mitchell et al., 1973; Dahlin et al., 1984), ticlopidine (Shimizu et al., 2009, 2011), tienilic acid (Nishiya et al., 2008), and DPH (Sasaki et al., 2015), and CBZ (Ju and Uetrecht, 1999; Pearce et al., 2005; Iida et al., 2015). Most soft electrophilic metabolites can react with GSH or NAC by forming stable GSH- or NAC-conjugates. In the cases of DPH and CBZ, protecting the role of hepatic GSH in the development of DPH- or CBZ-induced liver injury is indicated by using GSH-depleted animal models (Sasaki et al., 2013; Iida et al., 2015). Taking together these results, the GSH-mediated detoxification pathway for the electrophilic metabolite(s) is considered to be a key factor in understanding the development of liver injury. Putative reactive metabolite(s) have emerged for DPH and CBZ, as described in sections 3 and 4, respectively.
In humans, hepatic GSH contents are decreased by alcohol intake (Lauterburg and Velez, 1988), aging (Hernanz et al., 2000) and virus infection (Staal et al., 1992). These factors can decrease hepatic GSH contents, which would lead to a decrease in detoxification ability for electrophilic reactive metabolites. In addition, it is speculated that chronic administration of drug, which can be metabolized to electrophilic metabolite(s), leads to sustainable decreases in hepatic GSH contents that can be caused by GSH consumption through electrophilic metabolites. In fact, repeated administration of DPH and CBZ decreases hepatic GSH contents in mice (Higuchi et al., 2012; Sasaki et al., 2013, 2015) and rats (Iida et al., 2015).
Genetic polymorphisms are considered to be a factor for change susceptibility to developing DPH- and CBZ-induced liver injury. γ-Glutamylcysteine synthetase (γ-GCS) is the rate-limiting enzyme for GSH synthesis and is polymorphically expressed in humans (Nakamura et al., 2002). However, genotyping data on the relationship between polymorphisms in γ-GCS and susceptibility to developing DPH- or CBZ-induced liver injury in humans have not yet been reported. Glutathione S-transferase (GST) µ is known to detoxify reactive epoxide and is also polymorphically expressed in humans, where gene deletion causes absence of the gene in 50% of the population (Eaton and Bammler, 1999). However, no data have revealed the relationship between these genetic polymorphisms and the risk of idiosyncratic DPH- and CBZ-induced liver injuries.
Developing CBZ- and DPH-induced liver injury requires repeated administration in both animal models (Higuchi et al., 2012; Sasaki et al., 2013) and humans (Schlienger and Shear, 1998). When focusing on drug metabolism factors at this point, auto-inductions of CYPs should be considered. In fact, both DPH (Chaudhry et al., 2010; Fleishaker et al., 1995; Yamazaki et al., 2001; Hagemeyer et al., 2010) and CBZ (Tateishi et al., 1999; Oscarson et al., 2006) can induce CYPs in humans and rodents. In addition, CYPs, which can be induced by drugs, are also involved in metabolizing the drugs themselves. In clinical uses, dosages of DPH and CBZ can be changed by medication terms because chronic treatment of these drugs leads to requiring higher doses to reach effective therapeutic plasma concentrations (Johannessen and Landmark, 2010) because metabolism of these drugs can be induced by chronic treatment.
The enhancement of metabolism of these drugs can be considered a risk factor for developing liver injury because both DPH and CBZ are partially metabolized to electrophilic reactive metabolites as discussed in sections 3 and 4, respectively. For DPH, the enhancement of metabolism was observed over 5 days of continuous DPH administration compared to 3 days in mice (Sasaki et al., 2015). In addition, the remarkable elevation of plasma ALT levels was observed after 5 days of continuous DPH administration, which suggests that enhancement of DPH metabolism would be a crucial step for liver injury development. For CBZ, although a single administration of CBZ failed to develop liver injury in CYP3A-induced rats, which were created by CYP3A inducer dexamethasone pre-treatment, it led to success for a single CBZ administration-induced liver injury (Iida et al., 2015). This result strongly suggests that enhancement of CBZ metabolism followed by increasing reactive metabolite formation is responsible for the development of liver injuries.
In addition to CYP inductions, we first reported that reduced activities of Cyps were observed in DPH-induced liver injuries in mice, although the expression levels of Cyp3a protein were increased by DPH administration (Sasaki et al., 2015). This inhibition considered that Cyps can be irreversibly modified by reactive metabolites, followed by a loss of catalytic activity known as mechanism-based inhibition (MBI, Leeder, 1998). This phenomenon would contribute to changes in drug metabolism and activation of immune systems by forming neo antigen (Landsteiner and Jacobs, 1935; Park, 1998). Thus, MBI through repeated administration of DPH would be involved in developing liver injury.
The altered DPH or CBZ seems likely to also be involved in clinical cases of DPH- and CBZ-induced liver injuries. Clinically, antiepileptic drugs including DPH, CBZ and PB may often be used in combination (Johannessen and Landmark, 2010). These complicated combinations may also lead to changes in the metabolic profiles of DPH and CBZ. For example, PB treatment induces CYP3A and CYP2B enzyme activities in humans (Waxman and Azaroff, 1992). Indeed, PB affects the metabolizing pathways of DPH, CBZ and clomipramine in clinical use (Patsalos and Perucca, 2003; Perucca, 2006). By accounting for these clinical situations, there is a possibility that the altered metabolism of DPH or CBZ might be involved in the onset of idiosyncratic liver injury not only in animal models but also in human clinical cases.
The clinical manifestations of DPH and CBZ hypersensitivity reactions, including patients who develop idiosyncratic liver injury, are consistent with both immune reactions, and hepatic damage is estimated in approximately 50% of patients (Schlienger and Shear, 1998). Indeed, approximately 50% of individuals are believed to have experienced DPH and CBZ hypersensitivity, including patients who develop idiosyncratic liver injury based on a positive in vitro re-challenge test revealing serum anti-microsomal antibodies (Leeder et al., 1992). Subsequent studies revealed that the predominant antigen(s) of detected antibodies in patients who developed hypersensitivity reactions was a member of the CYP3A and CYP2C subfamily (Riley et al., 1993). These enzymes have been known to be involved in the metabolism of CBZ and DPH in in vitro studies, which suggests that the covalent bonds of reactive metabolites to CYPs are formed in humans (Munns et al., 1997; Roy and Snodgrass, 1988, 1990). However, this phenomenon was only demonstrated in in vitro studies. One hypothesis that has been raised is that the protein-drug adduct(s), which is recognized as foreign, can be processed by antigen-presenting cells that, in turn, can trigger B cell- or T cell-mediated responses (Zaccara et al., 2007). However, these hypotheses do not directly verify the association with the onset of liver injury. Therefore, further research is needed.
Until animal models for idiosyncratic DILI are established, the critical factor(s) for activating the immune reaction in idiosyncratic DILI will not be fully understood. However, as multiple in vivo studies suggest, damage-associated molecular patterns (DAMPs), such as high-mobility group box 1 (HMGB1), S100A8, and S100A9 (which is an agonist of toll-like receptor (TLR) 4 (Yao and Brownlee, 2010)), were shown to be involved through in vivo studies (Higuchi et al., 2012; Sasaki et al., 2013). DAMPs are secreted from the damaged cells and stimulate the TLR4 signal, followed by promotion of activation of innate immune-related cells such as hepatic stellate cells, Kupffer cells and dendritic cells (Guo and Friedman, 2010). In fact, the onset of liver injury is strongly suppressed by administering the TLR4 antagonist eritoran in DPH- and CBZ-induced liver injury mouse models (Higuchi et al., 2012; Sasaki et al., 2013). The importance of DAMP release has also been demonstrated in vivo in mice using an anti-HMGB1 antibody (Higuchi et al., 2012; Sasaki et al., 2013). In addition, gene expression levels of S100A8 and S100A9 increased in the livers of mice with DPH- and CBZ-induced liver injuries. The role for DAMP secretion in developing DILI is speculated to be cell damage through reactive oxygen species (ROS) or reactive metabolite formation (Martin-Murphy et al., 2010). Indeed, Cyps inhibitor pre-treatment inhibited the development of liver injury in DPH- and CBZ-induced liver injury in mice and rats (Sasaki et al., 2015; Iida et al., 2015). Therefore, it has been considered that secreted DAMPs have an amplifying role in reactive metabolite(s)-mediated liver injury via TLR4 activation.
The in vitro studies demonstrated that DPH and CBZ are metabolically activated to produce ROS (Lu and Uetrecht, 2008; Pirmohamed et al., 1992). It is known that ROS activates NACHT, LRR and PYD domain-containing protein 3 (NALP3); secretes interleukin (IL)-1β; and causes inflammation (Bryant and Fitzgerald, 2009; Shimada et al., 2012). NALP3 has the extraordinary capacity of sensing DAMPs such as ATP (Mariathasan et al., 2006), which suggests that ROS productions would amplify the DAMP-mediated reactions via an increase in NALP3 activation. In fact, in the case of DPH, gene expression levels of Nalp3 in the liver were elevated in the early stage of liver injury in mice (Sasaki et al., 2013). In addition, elevation of plasma levels of IL-1β protein, which is produced by cleavage of pro-IL-1β via NALP3 activation (Martinon et al., 2002), was observed in the initial stage of DPH (Sasaki et al., 2013) and diclofenac (Yano et al., 2012) induced liver injuries in mice.
ROS can induce the gene expression levels of TLR4 and the receptor for advanced glycation end products (RAGE), which is known to act as a DAMP receptor (Yao and Brownlee, 2010). In fact, both gene expression levels of TLR4 and RAGE were elevated in the early stage of CBZ-induced liver injury in mice (Higuchi et al., 2012). As described above, the onset of liver injury is strongly suppressed by administering the TLR4 antagonist eritoran in DPH- and CBZ-induced liver injuries in mice (Higuchi et al., 2012; Sasaki et al., 2013). Taking these results together, increasing DAMP expression and its receptor expression as well as DAMP release followed by TLR4, NALP3 inflammasome and RAGE signal activation involve amplifying liver injury through activating immune- and inflammation-related factors. The proposed mechanism of reactive metabolite-mediated immune activation in DILIs is indicated in Fig. 3.
Proposed mechanism of reactive metabolite-mediated inflammatory reactions in liver for idiosyncratic drug-induced liver injury. The overall flow of developing inflammation reactions is as follows: (a) Reactive metabolite formation occurs in hepatocytes. At the same time, reactive oxygen species (ROS) formation occurs. ROS can activate innate immune system-related factors such as NACHT, LRR and PYD domain-containing protein 3 (NALP3) and S100A8/9. (b) Accumulation of reactive metabolite(s) and ROS leads to necrotic cell death. (c) Necrotic cell death releases damage-associated molecular patterns (DAMPs) that activate Kupffer cells via toll-like receptors (TLRs) and the NALP3 inflammasome. (d) Activated Kupffer cells release cytokines (e.g., tumor necrosis factor (TNF)-α, interleukin (IL)-1β) and chemokines (e.g., macrophage inflammatory protein (MIP)-2), which lead to neutrophil infiltration in the liver. (e) Neutrophil infiltration contributes to exacerbating liver damage as well as causing inflammation-mediated liver injury. (f) This inflammation-mediated liver damage also accelerates inflammation reactions via further DAMP release from damaged hepatocytes. Such a loop of DAMP-mediated reactions would exacerbate liver damage, and reactive metabolites act to first trigger this inflammation reaction loop via primary hepatocyte cell death.
The in vitro assay system for predicting the risk of DILI while considering both drug metabolism and immune activation has been proposed (Yano et al., 2014). Yano et al. (2014) reported the establishment of a human monocytic cell-based in vitro DILI assessment tool. In particular, this in vitro assay system can account for the influence of CYP-mediated bioactivation of drugs by using human liver microsome-mediated metabolic reactions in the presence of nicotinamide adenine dinucleotide phosphate (NADPH). In this in vitro assay tool, the immune- and inflammation-related genes were used as indicators for risk of DILI, and these genes were also clearly elevated in the liver with the developing idiosyncratic DILI mouse models, such as DPH (Sasaki et al., 2013), CBZ (Higuchi et al., 2012) and diclofenac (Yano et al., 2012). In addition, Oda et al. (2016) reported that a conditioned medium from HepaRG cells with drugs can activate human promyelocytic leukemia HL-60 cells. Furthermore, the gene expression levels of immune- and inflammation-related factors in HL-60 cells correlated well with the risk of DILI development for 72 drugs (Oda et al., 2016). This correlation was not achieved when HepG2 cells were used instead of HepaRG cells. The cause of this discrepancy is considered to be due to the relatively low expression levels of CYPs in HepG2 cells compared to HepaRG cells (Hart et al., 2010; Gerets et al., 2012). Taken together, reactive metabolite formation is important for initiation of immune and inflammation reactions in DILIs by leading to weak or moderate hepatocyte damage followed by release of DAMPs. These results support the evidence that metabolic activation would be required for initiating immune activation for idiosyncratic DPH- and CBZ-induced liver injuries.
Taking all these findings together, it is concluded that at least metabolic activation plays an important role in creating a trigger for immune activation via ROS production, DAMP release and TLR4 signal activation, which are phenomena that would lead to immune and inflammation reactions in the liver.
This review focused on antiepileptic drug (especially DPH and CBZ)-induced liver injury and summarized the importance of drug metabolism in the initiation of liver injury. Since the 1980s, DPH and CBZ have been presumed to involve drug metabolism in their toxic reactions; however, it had not been possible to demonstrate the relationships between drug metabolism and the development of liver injury because there were no appropriate animal models for idiosyncratic DILIs. By establishing model animals, we found that changes in drug metabolism activities are one of the determining factors for susceptibility to developing liver injury caused by DPH or CBZ treatment. Unlike intrinsic DILIs, the majority of idiosyncratic DILIs have been caused by continuous administration of drugs. However, it is also clear that changes in drug metabolism profiles through continuous administration of DPH or CBZ contributed to reactive metabolite production. Furthermore, in vivo evidence indicates that reactive metabolite formation would contribute to secreting DAMPs through mild cell damage. This DAMP release is an initial step for immune activation that causes inflammation- and immune-mediated severe liver injury. Overall, by considering both the generation of reactive metabolite(s) as an initial step and the immune activation through DAMP release as an exacerbating step, accurate predictive evaluation of idiosyncratic DILIs could be achieved. Therefore, animal models of idiosyncratic DILIs would contribute to determining the mechanisms and establishing idiosyncratic DILI prediction systems. This accumulated knowledge will provide a change in the concept of “idiosyncratic toxicity”, which has been considered a black box, to be recognized as “mechanism-based toxicity”.
The work summarized in this review was conducted at Drug Metabolism and Toxicology, Faculty of Pharmaceutical Sciences, Kanazawa University (Kakuma-machi, Kanazawa, Japan), and it was supported by funding from the Health and Labor Sciences Research Grants from the Ministry of Health, Labor and Welfare of Japan (H23-BIO-G001).
Conflict of interestThe authors declare that there is no conflict of interest.
