Inhibition and induction of human cytochrome P450 enzymes: current status

Abstract

Variability of drug metabolism, especially that of the most important phase I enzymes or cytochrome P450 (CYP) enzymes, is an important complicating factor in many areas of pharmacology and toxicology, in drug development, preclinical toxicity studies, clinical trials, drug therapy, environmental exposures and risk assessment. These frequently enormous consequences in mind, predictive and pre-emptying measures have been a top priority in both pharmacology and toxicology. This means the development of predictive in vitro approaches. The sound prediction is always based on the firm background of basic research on the phenomena of inhibition and induction and their underlying mechanisms; consequently the description of these aspects is the purpose of this review. We cover both inhibition and induction of CYP enzymes, always keeping in mind the basic mechanisms on which to build predictive and preventive in vitro approaches. Just because validation is an essential part of any in vitro–in vivo extrapolation scenario, we cover also necessary in vivo research and findings in order to provide a proper view to justify in vitro approaches and observations.

Introduction

Activities of cytochrome P450 (CYP) enzymes are affected by numerous genetic, endogenous host, and environmental factors, making drug metabolism exceedingly variable and even individualistic. This variability has important repercussions to drug development, clinical drug therapy and in general to sensitivity to chemicals foreign to the body, i.e. xenobiotics. Among environmental factors, compounds causing inhibition and induction are amongst the most important ones, or at least the most researched.

Inhibition of CYP enzymes by other drugs or chemicals has received considerable attention since cimetidine was shown to affect drug metabolism in both animals and humans (Puurunen and Pelkonen 1979; Rendic et al. 1979). Mechanistic insights, predictive in vitro assays and modeling and validation of in vitro–in vivo extrapolation have revolutionized this part of drug development. However, inhibitory interactions in the metabolism of compounds other than pharmaceuticals have been less well developed and the significance of chemical–drug or chemical–chemical interactions is still not very well understood.

Although the role of Ah receptor in the induction of aryl hydrocarbon hydroxylase was elucidated already more than 20 years ago, it was only after the unraveling of the role of pregnane X receptor (PXR) and constitutive androstane receptor (CAR) in more general P450 induction phenomena that the research field experienced almost logarithmic growth. With the advent of several other nuclear receptors, their crosstalk in regulating CYP induction and pleiotypic responses after the exposure of the organism to nuclear receptor ligands, we have started to understand the complex interwoven regulatory networks, which link drug metabolism with the regulation of many facets of intermediary metabolism, such as bile acids, lipids, glucose and so on.

In this review we present an updated view on the scope and significance of both inhibition and induction of CYP enzymes, especially in humans, but taking examples also from animal studies. We have tried to cover pertinent literature until March 2008, but the earlier literature, which was covered in our earlier review article (Pelkonen et al. 1998), will be mentioned only sporadically. It is impossible to cover all the literature on CYP inhibition and induction published during the last 10 years, so we have to apologize for any omission of significant references and authors.

CYP enzymes in human tissues

In February 2008, the CYP superfamily consisted of more than 7,000 named sequences in animals, plants, bacteria and fungi (http://drnelson.utmem.edu/CytochromeP450.html). The human genome has 57 CYP genes, and the function for most of the corresponding enzymes is known at least to some degree. Fifteen individual CYP enzymes in families 1, 2 and 3 metabolize xenobiotics, including the majority of small molecule drugs currently in use. A typical feature of these CYPs is broad and overlapping substrate specificity (Guengerich et al. 2005). Other CYPs with much narrower substrate specificity are devoted mainly to the metabolism of endogenous substrates, such as sterols, fatty acids, eicosanoids, and vitamins. It has become evident that expression patterns of many individual CYPs in different tissues and cell types of an organ have important physiological roles (Seliskar and Rozman 2007).

Cytochrome P450 enzymes are found in practically all tissues, with highest abundance and largest number of individual CYP forms present in the liver. CYPs reside also in the intestine, lung, kidney, brain, adrenal gland, gonads, heart, nasal and tracheal mucosa, and skin. In human liver CYP enzymes comprise approximately 2% of total microsomal protein (0.3–0.6 nmol of total CYP per mg of microsomal protein). The content of drug-metabolizing CYPs is much lower in other tissues (Table 1). While extrahepatic metabolism may have clinically significant local effects, systemic metabolic clearance of drugs occurs in the liver with a significant contribution by the gut wall in special cases.

Table 1 Total CYP content in selected human tissues (Hrycay and Bandiera 2008)

Metabolism is the main route of clearance for approximately 70% of currently used drugs. Ten individual CYP forms in the adult human liver carry out virtually the whole CYP-mediated metabolism. CYP3A4 is the highest abundance form and it metabolizes the greatest number of drugs and a very large number of other xenobiotics. A minority of Caucasian people have relatively high amount of CYP3A5 in the liver, and CYP3A7 is a fetal enzyme. Also CYP2D6, although of much lower abundance, mediates the metabolism of numerous drugs. Together CYP2B6, CYP2C9, CYP2C19, CYP2D6 and CYP3A4 are responsible for more than 90% of known oxidative drug metabolism reactions (Wienkers and Heath 2005; Guengerich 2008). Figure 1 illustrates the relative abundance of individual CYP forms in the liver, and lists some examples of substrates, inhibitors and inducers. The CYP enzymes are well known for their capacity to metabolize a vast number of structurally diverse xenobiotics. Several reviews (e.g. Rendic 2002; Guengerich et al. 2005; Liu et al. 2007; Brown et al. 2008; Hrycay and Bandiera 2008) and Internet sites (e.g. http://medicine.iupui.edu/flockhart) contain exhaustive lists of xenobiotics that are CYP substrates.

Fig. 1

Relative abundance of individual CYP forms in the liver and some examples of substrates, inhibitors and inducers. CYPs with clinically significant genetic polymorphism in gray. Phenob. phenobarbital, TCDD 2,3,7,8-tetrachlorodibenzo-p-dioxin, nk not known

The genes encoding CYP enzymes are highly polymorphic (http://www.cypalleles.ki.se). Numerous studies have established that several variant alleles of individual CYP genes encode functionally deficient enzymes, the prime example being CYP2D6. When challenged with a CYP2D6 substrate drug, e.g. dextromethorphan, individuals with a deficient enzyme phenotype [poor metabolizers (PMs)] may experience adverse effects due to excessive serum concentrations of the drug. On the other hand, individuals with multiple copies of the CYP2D6 gene (ultrarapid metabolizers) will have insufficient clinical response since the drug is eliminated during first-pass metabolism (Kirchheiner et al. 2005). Thus, genotyping patients for CYP2D6 and other drug-metabolizing genes before implementing drug therapy would be advantageous. Nevertheless, implementing this type of genetic information into practice is a daunting task with several obstacles to overcome before individualized drug therapy is a reality (Nebert et al. 2008).

Most drugs cleared by the CYP system are metabolized through several CYP forms. As a general rule, drugs that are metabolized by a single CYP form are more susceptible to drug interactions than drugs metabolized by multiple forms. For investigational purposes, an ideal marker (probe) drug should be metabolized by a single CYP form. The FDA has issued a draft guidance on drug interaction studies (Huang et al. 2007). In this guidance, examples are given on substrates, inhibitors and inducers that can be used in clinical drug interaction studies (Table 2).

Table 2 Examples of in vivo substrate, inhibitor, and inducer for CYP enzymes recommended for study (oral administration) (Huang et al. 2007)

Inhibition of CYP enzymes

Inhibition of CYP enzymes is the most common cause of harmful drug–drug interactions and has led to the removal of several drugs from the market during the past years (Friedman et al. 1999; Lasser et al. 2002). Inhibition can lead to increased bioavailability of the parent compound normally subject to extensive first-pass elimination or to decreased elimination of compounds dependent on metabolism for systemic clearance. If a drug is metabolized mainly via single pathway, inhibition may result in increased steady-state concentration and accumulation ratio and non-linear kinetics as a consequence of the saturation of enzymatic processes. Especially with prodrugs, inhibition may result in a decrease in the amount of the active drug form. Thus, inhibition of CYPs may lead to the toxicity or lack of efficacy of a drug.

The type of CYP inhibition can be either irreversible (mechanism-based inhibition) or reversible. The distinction is relative and can be hard to determine if the inhibitor binds tightly to the enzyme and is released slowly (Wienkers and Heath 2005). Irreversible inhibition requires biotransformation of the inhibitor, while reversible inhibition can take place directly, without metabolism. Reversible inhibition is the most common type of enzyme inhibition and can be further divided into competitive, non-competitive, uncompetitive, and mixed-type inhibition (Lin and Lu 1998; Hollenberg 2002; Madan et al. 2002).

Mechanism-based inhibition

Mechanism-based inhibition can occur via the formation of metabolite intermediate complexes or via the strong, covalent binding of reactive intermediates to the protein or heme of the CYP. The most important phenomenon of mechanism-based inhibition is the time-, concentration-, and NADPH-dependent enzyme inactivation (Halpert 1995; Lin and Lu 1998). Mechanism-based inhibition is terminated by enzyme resynthesis and is therefore usually long-lasting (Halpert 1995; Ito et al. 1998; Kent et al. 2001). In some cases, the metabolic product inactivates the enzyme completely (suicide inhibition). Classical mechanism-based inhibitors include the CYP1A2 inhibitor furafylline (Sesardic et al. 1990; Kunze and Trager 1993) and the CYP3A4-inhibitor gestodene (Guengerich 1990; Back et al. 1991).

Reversible inhibition

Reversible inhibition occurs as a result of competition at the active site of the enzyme and probably involves only the first step of the P450 catalytic cycle. Binding to the enzyme takes place usually with weak bonds, which are both formed and broken down easily. Consequently, reversible inhibitors act rapidly, but do not permanently destroy the enzyme (Lin and Lu 1998; Hollenberg 2002; Madan et al. 2002).

In competitive inhibition, competition between the substrate and inhibitor to bind to the same position on the active site of the enzyme takes place. In the noncompetitive mode of inhibition, the active binding site of the substrate and inhibitor is different from each other. In the case of uncompetitive inhibition, the inhibitor binds to the enzyme–substrate complex, but not to the free enzyme entity. In practice, mixed-type inhibition displaying elements of both competitive and noncompetitive inhibition are frequently observed (Madan et al. 2002).

Inhibition of individual CYP enzymes: examples of substrates and inhibitors

Many individual members of CYP families exhibit distinct, but often overlapping, selectivity towards certain substrates and inhibitors. The most commonly used probe substrates and diagnostic inhibitors for each CYP form are collected in respective tables and discussed briefly below. The K _m and K _i values are collected from the appropriate in vitro studies.

CYP1 family

CYP1A2 is the only hepatic member of the CYP1 family. CYP1A1 and CYP1B1 are the other enzymes in this family, of which CYP1A1 is the major human extrahepatic CYP form (Ding and Kaminsky 2003). The hepatic expression of CYP1B1 is insubstantial, but otherwise it is known to be expressed in almost every other tissue (Sutter et al. 1994; Shimada et al. 1996a, b). All members of this family are regulated by the AhR-receptor (see “Induction of CYP enzymes: mechanisms”). Besides detoxification, the CYP1 family members are often responsible for metabolic activation of polycyclic aromatic hydrocarbons (PAHs) and aromatic amines and thus they have been linked to chemical carcinogenesis (Boobis et al. 1994; Ioannides and Lewis 2004).

CYP1A2

Initially the expression of CYP1A2 was thought to be limited only to the liver, but recent studies have shown that it is expressed along with CYP1A1 in the lung (Wei et al. 2002; Liu et al. 2003). Over 20 single nucleotide polymorphisms (SNPs) within CYP1A2 have been reported, though most of them have been found to be very rare (Nakajima et al. 1994; Sachse et al. 1999). Despite extensive interindividual variation in CYP1A2 activity and systematic sequencing efforts, no predictive CYP1A2 polymorphisms have been reported (Ingelman-Sundberg et al. 2007). However, recent twin studies have suggested a strong role of genetic factors in CYP1A2 function (Rasmussen et al. 2002).

Substrates and inhibitors of CYP1A2

CYP1A2 is a major enzyme in the metabolism of a number of important chemicals, which typically belong structurally to the group of planar polyaromatic amides and amines (Lewis 2004). Ethoxyresorufin, caffeine, phenacetin, theophylline, clozapine, melatonin, and tizanidine are biotransformed predominantly by this CYP form (Table 3). Caffeine clearance has been regarded as ‘the golden standard’ for in vivo phenotyping purposes due to the predominating role of CYP1A2 in the overall metabolism of caffeine and the excellent tolerability of this probe substrate (Faber et al. 2005). Recently, also oral melatonin has been suggested as a suitable marker for CYP1A2 phenotyping (Härtter et al. 2001; Faber et al. 2005). For in vitro purposes, especially phenacetin, but also ethoxyresorufin are recommended, whereas the xanthines caffeine and theophylline are not so favored due to the low turnover of these compound in vitro (Tucker et al. 2001; Bjornsson et al. 2003).

Table 3 Substrates and inhibitors of CYP1A2 enzyme

Potent inhibitors of CYP1A2 include furafylline, fluvoxamine, ciprofloxacin, and rofecoxib. Also oral hormone replacement therapy and oral contraceptives have been shown to significantly inhibit CYP1A2-mediated metabolism (Laine et al. 1999; Pollock et al. 1999). Furafylline, a methylxanthine analog, is a selective and potent mechanism-based inhibitor of several CYP1A2-mediated reactions and is widely employed in in vitro studies. However, it is not available for in vivo use, since it has severe interactions with caffeine (Tarrus et al. 1987). For in vivo study purposes, selective serotonin reuptake inhibitor fluvoxamine and fluoroquinolone antibiotic ciprofloxacin are usually applied.

CYP2 family

The human CYP2 family is very diverse and comprises a number of important drug-metabolizing CYPs. Members of this family do not share any common regulation patterns and their substrate specificities and tissue expression vary substantially. CYP2B6, CYP2D6, and CYP2E1 are the only functional enzymes in their subfamilies, whereas CYP2A contains two, and CYP2C four functional members. The clinically most important CYP polymorphisms are found within the CYP2 family (i.e. CYP2C9, CYP2C19, and CYP2D6).

CYP2A6

At the quantitative level, CYP2A6 is a minor component among hepatic CYPs (Rostami-Hodjegan and Tucker 2007). Several variant CYP2A6 alleles with distinct frequencies between ethnic populations have been characterized. Some of these alleles have been associated with altered nicotine pharmacokinetics and furthermore to differing smoking habits in variant genotype populations (London et al. 1999; Raunio et al. 2001). Relatively large variability in the enzyme activity between individuals has been described, with a fair proportion of Japanese known to lack the functional protein completely (Shimada et al. 1996b; Pelkonen et al. 2000).

Substrates and inhibitors of CYP2A6

Substrates of CYP2A6 are usually structurally small and planar molecules (Lewis 2004). CYP2A6 has a predominant role in the overall metabolism of nicotine and its metabolite cotinine (Hukkanen et al. 2005). The 7-hydroxylation of coumarin is known to be solely catalyzed by CYP2A6 and therefore coumarin has been traditionally employed as the prototypical model substrate for this enzyme (Pelkonen et al. 2000) (Table 4). In addition to pharmaceuticals, bioactivation of some toxicologically significant substances such as aflatoxin B₁ and nitrosoamines are known to be mediated at least to some extent via CYP2A6 (Pelkonen et al. 2000; Raunio et al. 2001).

Table 4 Substrates and inhibitors of CYP2A6 enzyme

A number of potent inhibitors with variable selectivity against CYP2A6 have been characterized. The most used in vitro inhibitors include tranylcypromine and methoxsalen (8-methoxypsoralen).

CYP2B6

Initially CYP2B6 was regarded as a minor hepatic CYP in humans, essentially expressed only in a few livers and with minor significance in the overall xenobiotic metabolism. This view has changed over the very last few years and currently CYP2B6 belongs to a set of important hepatic drug-metabolizing CYPs (Turpeinen et al. 2006). An extensive interindividual variability in the expression of CYP2B6 has been reported, mainly due to genetic factors, and CYP2B6 has been estimated to represent approximately 1–10% of the total hepatic CYP pool (Rostami-Hodjegan and Tucker 2007; Zanger et al. 2007).

Substrates and inhibitors of CYP2B6

The list of CYP2B6 substrates has increased drastically in the past few years (Turpeinen et al. 2006). CYP2B6 usually metabolizes non-planar, neutral, or weakly basic molecules with one or two hydrogen bond acceptors (HBAs) (Lewis 2004). For the metabolic pathway and kinetics of bupropion, cyclophosphamide, ifosfamide, efavirenz, ketamine, and propofol, CYP2B6 is of considerable importance (Table 5). In addition to pharmaceuticals, CYP2B6 appears to both detoxify and bioactivate a number of procarcinogens (Code et al. 1997; Smith et al. 2003a). Bupropion has been suggested as a good model substrate both for in vitro and in vivo studies. It is extensively metabolized in human liver, resulting in low levels of the parent compound in plasma. The major metabolite, pharmacologically active hydroxybupropion, is formed selectively by CYP2B6 (Turpeinen et al. 2006).

Table 5 Substrates and inhibitors of CYP2B6 enzyme

Potent inhibitors of CYP2B6 include thienopyridine derivatives clopidogrel and ticlopidine and the anticancer agent thioTEPA, which is a very selective inhibitor of this particular CYP.

CYP2C8

Like in the case of CYP2B6, the importance of CYP2C8 for drug metabolism has been elucidated quite recently (Totah and Rettie 2005). Also a number of functional CYP2C8 polymorphisms have been published during recent years (Dai et al. 2001; Niemi et al. 2003a). Some SNPs or their combinations in the CYP2C8 gene have been associated with certain disease states or adverse drug reactions, but more studies about the importance of CYP2C8 polymorphisms and also the general role of this enzyme in drug metabolism are still needed.

Substrates and inhibitors of CYP2C8

Drugs metabolized by CYP2C8 do not share any common structure or chemical pattern. There seems to be some overlapping especially with CYP2C9 and CYP3A4 substrates. Drugs with major importance of CYP2C8 include amodiaquine, paclitaxel, cerivastatin, and several oral antidiabetics such as repaglinide, pioglitazone, and rosiglitazone (Table 6). Paclitaxel 6α-hydroxylation has been regarded as the typical index of CYP2C8 activity, but partly due to the high costs of authentic chemical standards and unsuitability for in vivo use, other probe substrates have also gained interest. Recently, the N-deethylation of the antimalarial amodiaquine was demonstrated as a good model substrate for CYP2C8 with high affinity and turnover rate. So far the applicability of glitazones as model substrates has been restricted by difficulties in obtaining metabolite standards.

Table 6 Substrates and inhibitors of CYP2C8 enzyme

Known CYP2C8 inhibitors include quercetin, which has been used for several years for in vitro purposes, and leukotriene receptor antagonists montelukast and zafirlukast. Although montelukast and zafirlukast are potent inhibitors of CYP2C8 in vitro, they both are highly bound to plasma proteins (>99%) resulting in very low free fraction in humans. Thus these two drugs are not suitable for in vivo inhibition purposes (Jaakkola et al. 2006a; Kim et al. 2007). The lipid-lowering drug gemfibrozil inhibits CYP2C8 potently in vivo via its phase II metabolite, gemfibrozil 1-O-β-glucuronide. Besides CYP2C8, gemfibrozil inhibits the organic anion-transporting polypeptide-2 (OATP2)-transporter, which should be noted when evaluating its in vivo effects (Shitara et al. 2004; Schneck et al. 2004).

CYP2C9

CYP2C9 is the predominant CYP2C form with high abundance among hepatic CYPs (Rostami-Hodjegan and Tucker 2007). It is polymorphically expressed, and the importance of the SNPs within CYP2C9 gene is emphasized especially with S-warfarin, which uses CYP2C9 as a major metabolic pathway and possesses a narrow therapeutic window with a potentially fatal side-effect profile (Aithal et al. 1999; Daly and King 2003; Kirchheiner and Brockmöller 2005).

Substrates and inhibitors of CYP2C9

Besides S-warfarin, CYP2C9 catalyses the metabolism of a number of other clinically relevant drugs such as fluoxetine, fluvastatin, losartan, and several non-steroidal anti-inflammatory agents, as well as the classical probe substrate tolbutamide (Table 7). Tolbutamide methylhydroxylation and diclofenac 4′-hydroxylation are both validated for CYP2C9 marker reactions both in vitro and in vivo.

Table 7 Substrates and inhibitors of CYP2C9 enzyme

Among the recognized inhibitors of CYP2C9 are amiodarone, and fluconazole, which both are suitable for in vivo use, but are relatively unselective. For in vitro purposes, sulphaphenazole is traditionally employed, because of very high potency and selectivity.

CYP2C19

Drugs metabolized via CYP2C19 are usually amides or weak bases with two HBAs (Lewis 2004). Compared to CYP2D6 (see later), polymorphisms of the CYP2C19 gene represent a smaller proportion and perhaps have less clinical significance in Caucasians, but in Orientals the frequency of CYP2C19 PMs has been characterized to be up to 20% of the population (Bertilsson 1995; Ingelman-Sundberg et al. 2007).

Substrates and inhibitors of CYP2C19

CYP2C19 participates in the metabolism of many commonly used pharmaceuticals, e.g. diazepam, citalopram, amitriptyline, mephenytoin, proguanil, and phenytoin. The metabolism of most of the proton pump inhibitors (omeprazole, esomeprazole, lansoprazole and pantoprazole) is mediated mainly by CYP2C19. Classical marker reactions for this enzyme include S-mephenytoin 4′-hydroxylation and omeprazole 5-hydroxylation (Table 8).

Table 8 Substrates and inhibitors of CYP2C19 enzyme

No selective drug inhibitors for CYP2C19 have been found yet, but at least omeprazole, ticlopidine, nootkatone, and fluconazole—all with some affinity towards other CYPs, too—have been employed for this purpose.

CYP2D6

Since the characterization of the interindividual differences in the oxidation of debrisoquine and sparteine in the late 1970s, CYP2D6 has become the most studied CYP with respect to pharmacogenetics. The genetic polymorphism within the CYP2D6 gene causes wide and clinically important variability in CYP2D6 enzyme activity (Eichelbaum et al. 2006; Ingelman-Sundberg et al. 2007). Ultimate examples of the polymorphism of the CYP2D6 gene include the PMs lacking the functional enzyme, and the ultra-rapid metabolizers (UMs) having duplications or multiplications of the gene. Approximately 7 and 5.5% of Caucasians have been genotyped for CYP2D6 PMs and UMs, respectively (Zanger et al. 2004; Ingelman-Sundberg et al. 2007). Wide variability between ethnic groups with respect to CYP2D6 phenotype exists; for instance, the PM phenotype is practically absent in Oriental populations and the UM phenotype is very frequent in certain Arabian and Eastern African populations (Ingelman-Sundberg et al. 2007). CYP2D6 belongs to the set of most relevant target genes where genotype/phenotype testing has been suggested as a useful tool in dosing and monitoring in clinical practice (Dahl 2002; Kirchheiner et al. 2005; Eichelbaum et al. 2006).

Substrates and inhibitors of CYP2D6

CYP2D6 contributes to the metabolism of dextromethorphan, debrisoquine, and bufuralol, which all have been used as model substrates for this enzyme. Since debrisoquine is no longer available for in vivo studies, also newer substances like atomoxetine have been introduced for phenotyping purposes. The metabolism of several β-adrenoceptor antagonists like metoprolol and propranolol, several important antidepressants such as fluoxetine and paroxetine, and atypical antipsychotics like risperidone and aripiprazole is mediated predominantly via CYP2D6 (Table 9). On structural basis, the common characteristic for CYP2D6 substrates seems to be, that they are mostly basic molecules with protonatable nitrogen atom 4–7 Å from the site of metabolism (Lewis 2004).

Table 9 Substrates and inhibitors of CYP2D6 enzyme

CYP2D6 is inhibited potently by a variety of different drugs, of which a large proportion belongs also to the list of CYP2D6 substrates. Traditionally an antiarrhythmic drug quinidine has been utilized as a highly selective and very efficient CYP2D6 inhibitor for metabolism studies. Incidentally, quinidine is not a substrate of CYP2D6.

CYP2E1

Although CYP2E1 is one of the most abundant hepatic CYPs, only a few pharmaceuticals are metabolized via this enzyme. However, from the toxicological perspective, the role of CYP2E1 is without dispute. CYP2E1 has been studied extensively due to its central role in the metabolism of ethanol (Kessova and Cederbaum 2003; Lieber 2004), in the bioactivation of several industrial solvents (Raucy et al. 1993), as an activator of chemical carcinogenesis, and as a producer of free radicals causing tissue injury (Caro and Cederbaum 2004; Gonzalez 2005). CYP2E1 is also linked to acetaminophen-related hepatotoxicity (Rumack 2004). Although a number of SNPs within the CYP2E1 gene have been described, no polymorphisms leading to a loss of function have been reported (Gonzalez 2005).

Substrates and inhibitors of CYP2E1

The substrates of CYP2E1 usually consist of hydrophobic and low molecular weight compounds (Lewis 2004). For modeling purposes, chlorzoxazone is probably the most widely used, but also the metabolism of acetaminophen, enflurane, and halothane seems to be mediated to some extent by CYP2E1 (Table 10). It is noteworthy that several substrates of CYP2E1 (e.g. ethanol, acetone and pyrazole) act as inducing agents of this enzyme.

Table 10 Substrates and inhibitors of CYP2E1 enzyme

Inhibitors of CYP2E1 include pyridine and disulfiram, the latter being utilized in clinical practice as a treatment of alcohol dependence.

CYP3 family

The human CYP3 family represents about 30% of the total hepatic P450 content and is considered to be the most important CYP subfamily in the biotransformation of drugs. This family contains one subfamily including three functional proteins: CYP3A4, CYP3A5, and CYP3A7, and one pseudoprotein, CYP3A34 (Ingelman-Sundberg 2005). These enzymes have overlapping catalytic specificities and their tissue expression patterns differ.

CYP3A5 is a minor polymorphic CYP form in human liver (Westlind-Johnsson et al. 2003), but in extrahepatic tissues it is consistently expressed in kidney, lung, colon, and esophagus (Ding and Kaminsky 2003; Burk and Wojnowski 2004). Despite a few exceptions, the substrate and inhibitor specificity of CYP3A5 seems to be highly similar to CYP3A4, albeit the catalytic capability might be somewhat lower (Wrighton et al. 1990; Williams et al. 2002).

CYP3A7 is mainly expressed in embryonic, fetal, and newborn livers, where it is the predominant CYP form (Kitada and Kamataki 1994; Hakkola et al. 2001), whereas in the adult liver, CYP3A7 seems to be a minor form (Schuetz et al. 1994). CYP3A7 has an important role during the fetal period in the hydroxylation of several endogenous substances like retinoic acid and steroid hormones, and therefore it has relevance to normal embryonal development (de Wildt et al. 1999; Hines and McCarver 2002). In drug metabolism, the role of CYP3A7 is not yet clear.

CYP3A4

CYP3A4 is the sixth most abundant enzyme in human liver at the mRNA level and constitutes the major CYP form in the liver and the small intestine (Kivistö et al. 1996; von Richter et al. 2004; Paine et al. 2006; Rostami-Hodjegan and Tucker 2007). CYP3A4 has a pivotal role in xenobiotic metabolism, and it has been estimated to be involved in the metabolism of approximately 50% of the drugs in clinical use. The active site of CYP3A4 is very large and flexible allowing multiple small molecules to be present simultaneously in the active site. The substrate binding is principally based on hydrophobicity with some steric interactions. A concept where multiple conformations of the enzyme can exist both in the presence and absence of substrate has been proposed (Ekins et al. 2003; Scott and Halpert 2005). The kinetic interaction between CYP3A4 and its substrates is often atypical, making the prediction and modeling of CYP3A4-mediated drug–drug interactions troublesome (Ekins et al. 2003; Houston and Galetin 2005).

Substrates and inhibitors of CYP3A4

The known substrates of CYP3A4 vary widely in size and structure. Among the substrates of CYP3A4 are several clinically important drugs, e.g. cyclosporine A, erythromycin, nifedipine, felodipine, midazolam, triazolam, simvastatin, atorvastatin, and quinidine (Table 11), as well as several endogenous agents including testosterone, progesterone, androstenedione, and bile acid (Waxman et al. 1991; Patki et al. 2003). Consequently, altered CYP3A4 activity can lead to notable drug–drug interactions and adverse effects. Bioactivation of some procarcinogens such as aflatoxin B₁ (Aoyama et al. 1990) and PAHs (Hecht 1999) are also mediated partially via CYP3A4.

Table 11 Substrates and inhibitors of CYP3A4/5 enzyme

A relatively low degree of substrate selectivity makes CYP3A4 susceptible to inhibition by different chemicals. Inhibitors of CYP3A4 cover a broad variety of structurally unrelated substances. The most well established and clinically the most relevant inhibitors include certain azole antifungal agents (ketoconazole and itraconazole), antimicrobials (clarithromycin, erythromycin and ritonavir), antihypertensives (verapamil and diltiazem) and several herbal and food constituents, e.g. grapefruit juice and bergamottin (He and Edeki 2004; Fujita 2004). It is noteworthy that IC₅₀ values for CYP3A4 inhibitors are highly substrate-dependent, and the use of multiple probe substrates for inhibition studies is thus recommended (Kenworthy et al. 1999; Wang et al. 2000b; Galetin et al. 2003). The detailed characteristics of several CYP3A4 substrates and inhibitors are summarized recently in a review by Liu et al. (2007).

Inhibition: in vitro–in vivo extrapolation

Inhibitory potency in vitro-inhibition of clearance

Based on the requirements of the authorities, new drugs need to be tested with respect to their potential to cause drug–drug interactions (EMEA 1997, U.S. FDA 1997). These estimations rely primarily on projected in vivo concentrations of compounds and on estimates of their inhibitory constants obtained from in vitro studies.

The degree of inhibition depends also on the inhibition pattern when the substrate concentration is high. However, when the substrate concentration [S] ≪ K _m—which is the most usual case in clinical use—the degree of inhibition (R) can be expressed by the following equation independent of the inhibition pattern, except in the case of the uncompetitive inhibition (Tucker 1992; Ito et al. 1998):

$$ R = \frac{1}{{1 + [I]/K_{\rm i}}} $$

When [S] ≥ K _m, the degree of the inhibition can be estimated from the following assuming competitive inhibition (Tucker 1992; Ito et al. 1998):

$$ R = \frac{{(K_{{\text{m}}} + [S])}}{{[K_{{\text{m}}} (1 + [I]/K_{{\text{i}}}) + S]}} $$

In the case of competitive inhibition, it is possible to calculate the inhibition constants on the basis of experimentally determined IC₅₀ values using the Tornheim equation (Tornheim 1994):

$$ K_{{\text{i}}} = \frac{{[I]}}{{(V_{0}/V - 1)(1 + [S]/K_{{\text{m}}})}} $$

A recent study by Obach et al. (2006) involving a variety of drugs attempted to estimate the utility of in vitro data for prediction of drug–drug interactions in clinical situations. They concluded that in vitro inhibition data could be reliably used for predictions for at least CYP1A2, CYP2C9, CYP2C19, and CYP2D6, while for CYP3A4, the effects of both hepatic and intestinal metabolism should be considered. Other factors affecting in vivo–in vitro extrapolation will be discussed below.

Factors affecting in vitro–in vivo extrapolation

As presented above, affinity and CYP specificity for an inhibitor can be studied in vitro and further, the potential of a compound to cause interactions can be revealed. However, this does not necessarily mean that the compound would cause clinically significant drug–drug interactions. For such interactions to take place in vivo, two prerequisites have to be fulfilled: first, the concentration of the substrate in the in vivo situation should be high enough for the inhibition to occur in clinical situation, and second, the therapeutic index of the drug should be narrow, so that the change caused by the inhibitor would be manifested in adverse effects (Pelkonen et al. 1998, 2005). Semiquantitative classifications for the extrapolation purposes have been constructed such as that of Bjornsson et al. (2003) based to the ratio of C _max over K _i predicting the clinical relevance of the interaction in the case of competitive inhibition (Bjornsson et al. 2003).

However, free inhibitor concentrations at the site of action (adjacent to the enzyme) are in most cases unknown in in vivo situations. Based on the hypothesis that only an unbound fraction of a drug is capable of diffusing into hepatic tissue, predictions are made assuming that unbound inhibitor concentrations in plasma and hepatic tissue are equal (Ito et al. 1998; von Moltke et al. 1998b). For very lipophilic compounds, this assumption is known to be false; despite an extensive binding to plasma proteins, their hepatic concentrations are multiple to their plasma values (Chou et al. 1993; von Moltke et al. 1998b; Schmider et al. 1999; Cook et al. 2004). A recent analysis by Ito et al. (2004) suggested that total inhibitor concentrations with in vitro K _i values would probably be the most useful approach for the categorization of CYP inhibitors.

Although the expression of CYPs is centered in the liver, several other barrier tissues such as intestinal mucosa, lungs, and skin contain metabolic enzymes on a smaller scale and contribute to xenobiotic biotransformation (Kapitulnik and Strobel 1999; Ding and Kaminsky 2003). Knowing that the systemic bioavailability depends both on the dose absorbed and the fraction surviving from hepatic and extrahepatic metabolism, it should be noted that especially intestinal metabolism may affect in vitro–in vivo extrapolations (Wu et al. 1995; Hall et al. 1999; Kivistö et al. 2004). However, often the amount of extrahepatic metabolism is not known and consequently, estimations concerning the in vivo situation may be misleading.

The role of transporters, especially P-glycoprotein (P-gp) and human organic anion-transporting polypeptides (OATPs), has been recognized as a major contributor to drug–drug interactions. P-gp is known to possess a significant substrate overlap with the CYP3A family: drug substrates for both CYP3A4 and P-gp include cyclosporin A, verapamil, quinidine, erythromycin, and HIV-1 protease inhibitors (Wacher et al. 1995; Kim et al. 1999). Among drug substrates of OATPs are, e.g fexofenadine (Cvetkovic et al. 1999; Dresser et al. 2002) and pravastatin (Hsiang et al. 1999; Nishizato et al. 2003). Nevertheless, estimations concerning the net effect of transporters on interactions are very uncertain and have been so far poorly taken into account in in vitro–in vivo extrapolation calculations.

Finally, when estimating clinical significance of the interaction, one should take into account that results obtained from in vitro test systems are highly dependent on several technical aspects including the microsomal protein amount, incubation time, and initial velocity conditions used in the test system. The best predictive value is usually obtained when the substrate concentration used is within the linear part of the time and protein concentration dependence curve for the metabolite formation (Mäenpää et al. 1998; Lin and Lu 1998; Yuan et al. 1999). It is supposed, with some restrictions, that the values obtained from inappropriate experimental settings would lead to the greatest extrapolation error with compounds of intermediate inhibitory potency (Ghanbari et al. 2006).

Induction of CYP enzymes: mechanisms

Transcriptional regulation by ligand-activated transcription factors

Induction of CYP enzymes by exogenous compounds is mediated to major extent by group of ligand-activated transcription factors. These intracellular receptors involve aryl hydrocarbon receptor (AhR) that belongs structurally to the class of basic-helix-loop-helix (bHLH)-Per-Arnt-Sim (PAS) proteins and nuclear receptors pregnane X receptor (PXR, NR1I2) and constitutive androstane receptor (CAR, NR1I3). Together these receptors are able to sense a great variety of xenobiotics and consequently regulate numerous phase I and phase II drug-metabolizing enzymes and drug transporters in order to adjust the organism to the requirements of the chemical environment. In addition to these well-established xenosensors some other nuclear receptors such as estrogen receptor (ER) α and glucocorticoid receptor (GR) may be involved in some induction phenomena (Higashi et al. 2007a, b; Hukkanen et al. 2003).

AhR

AhR is expressed widely in human tissues with highest expression in placenta, lung, heart, pancreas and liver (Dolwick et al. 1993). AhR typically accepts hydrophobic, planar compounds such as classical AhR activator 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) as ligands. However, numerous molecular structures and both xenobiotics and endogenous compounds have been described to be ligands for AhR (recently reviewed by Nguyen and Bradfield 2008). The major classes of xenobiotic ligands include halogenated dioxins, polychlorinated biphenyls and PAHs. Furthermore, some dietary compounds can activate AhR. Indole-3-carbinol present in cruciferous plants appears to, perhaps through active derivates produced in gastrointestinal track, be able to activate AhR function (Bjeldanes et al. 1991). Also several endogenous compounds are AhR agonists; however, it remains to be shown if any of these thus far identified compounds represent true endogenous ligands (Nguyen and Bradfield 2008).

In the absence of a ligand, AhR is located in cytosolic compartment in complex with chaperone proteins Hsp90, ARA9 (also known as XAP2 or AIP) and p23 (Carlson and Perdew 2002). Binding of an agonist launches a conformational change that allows nuclear localization signal to interact with nuclear import machinery. In the nucleus AhR dimerizes with another bHLH-PAS protein AhR nuclear translocator (ARNT) thus forming the actual DNA-binding complex. The AhR/ARNT heterodimer interacts with XRE-binding elements with consensus core sequence 5′-TNGCGTG-3′. The DNA bound AhR/ARNT complex then activates transcription through recruitment of complex of multiple coactivators including CBP/p300, SRC-1, NCOA-2 and Mediator (Kawajiri and Fujii-Kuriyama 2007). The coactivators in turn modulate the chromatin structure and interact with the general transcription factors to allow activation of transcription.

AhR primary regulates expression of genes in CYP family 1, i.e. CYP1A1, CYP1A2 and CYP1B1. However, also some CYP2 family members are AhR target genes (Rivera et al. 2002; Arpiainen et al. 2005) (see Table 12). Our understanding on molecular details of AhR function is mainly based on extensive study of CYP1A1 induction mechanism (for a review see Ma 2001). CYP1A1 is induced extremely powerfully by AhR ligands while effect on most other AhR target genes is less pronounced. This may be due to multiple XRE sites in the CYP1A1 5′ flanking region (Kress et al. 1998). Furthermore, the low constitutive expression of CYP1A1 may emphasize the magnitude of induction. Ligand-activated AhR induces expression of AhR repressor (AhRR), which is able to dimerize with ARNT and bind XRE, but in contrast to AhR, represses transcription (Mimura et al. 1999). This may represent a negative feedback loop.

Table 12 Comparison of methods to assess receptor-mediated induction

PXR

Discovery of mouse and subsequently human PXR at the end of 1990s represented a major breakthrough in understanding the molecular mechanisms of many clinically significant induction phenomena (Kliewer et al. 1998; Lehmann et al. 1998; Blumberg et al. 1998; Bertilsson et al. 1998). Human PXR ligands (recently reviewed by Chang and Waxman 2006) include many therapeutic drugs, such as rifampicin, known to induce drug metabolism. PXR has large and flexible ligand-binding pocket which enables binding of numerous compounds of varying size and structure (Watkins et al. 2001). There are, however, major differences in PXR ligand preferences between species because of poorly conserved ligand-binding domain (Lehmann et al. 1998; Blumberg et al. 1998). Thus rifampicin is a good agonist for human PXR, but poorly activates mouse PXR. The opposite is true for mouse PXR agonist pregnenolone-16α-carbonitrile. These sharp differences in PXR ligand preferences affect and limit extrapolation of results from experimental animals to humans. PXR ligands include also endogenous compounds such as some bile acids (Staudinger et al. 2001; Xie et al. 2001).

Tissue distribution of PXR is quite narrow. In human, the main sites of PXR expression are liver and small intestine while limited expression can be detected in kidney and lung (Miki et al. 2005). This expression profile is in good agreement with the putative role as an environmental xenosensor and coincides with that of a major target gene CYP3A4. PXR protein level is regulated by microRNA, which in turn affects CYP3A4 expression level (Takagi et al. 2008). There has been some controversy about subcellular localization of unliganded PXR. While Koyano et al. (2004) reported constant nuclear localization of PXR regardless of the presence or absence of agonist some other recent studies have suggested cytoplasmic localization and nuclear transport after ligand binding (Kawana et al. 2003; Squires et al. 2004). Ligand bound PXR forms heterodimer with another nuclear receptor RXR. This heterodimer is able to bind several distinct DNA elements including both direct and everted repeats of sequence AGGTCA and its’ variants. A number of different coregulators including SRC-1, p300 and PGC-1 have been reported to interact with PXR (Orans et al. 2005). Interestingly, in addition to ligand, also tissue and target gene promoter appear to affect PXR coactivator interactions (Masuyama et al. 2005).

PXR target genes include members in subfamilies CYP2A, CYP2B, CYP2C and CYP3A (see Table 13). Especially regulation of CYP3A4 by PXR has been studied extensively [for reviews see Burk and Wojnowski (2004) and Plant (2007)]. Initially PXR was found to interact with ER6 motif present in the proximal promoter of the CYP3A4 gene (Lehmann et al. 1998). Subsequently Goodwin et al. 1999 identified a so-called XREM (xenobiotic-response enhancer module) in the distal CYP3A4 promoter -7784/-7672 that was found to play a major role in CYP3A4 induction by PXR ligands. The XREM contains several nuclear receptor-binding elements of which a DR3 element is of major importance. Both the DR3 element in the XREM and the ER6 element in the proximal promoter are needed for the maximal induction by PXR. Recently an additional PXR-binding element was identified in the far upstream region -11400/-10500 (Liu et al. 2008). Similar to XREM also this far module seem to act in collaboration with the proximal promoter. Hepatocyte nuclear factor 4 α (HNF4α) has been shown to augment PXR mediated induction of CYP3A4 (Tirona et al. 2003). However, there is some controversy if HNF4α DNA binding is necessary. Tirona et al. (2003) reported that HNF4-binding element in the XREM is needed for enhancement of PXR function by HNF4α. On the other hand, Li and Chiang (2006) suggested that rifampicin-activated PXR interacts with HNF4α through protein–protein interaction independently from HNF4α DNA binding. Regulation of CYP3A4 by PXR has been schematically presented in Fig. 2. Small heterodimeric partner (SHP) is able to interact with PXR and repress its transcriptional activity (Ourlin et al. 2003). Rifampicin-activated PXR downregulates SHP transcription (Li and Chiang 2006), which may enable maximal induction of PXR target genes such as CYP3A4.

Table 13 Induction of CYP enzymes in humans in vivo by xenobiotics

Fig. 2

Schematic model of CYP3A4 regulation by ligand activated PXR. Both ER6 sequence at the proximal promoter -170/-153 (P-ER6) and XREM sequence at the distal 5′ flanking region -7784/-7672 bind PXR and are necessary for maximal induction. Recently a novel ER-6 type PXR binding element was identified in the far module (F-MOD) -11400/-10500 that also appears to coordinate CYP3A4 induction together with the proximal ER-6 site. HNF4α interacts with PXR and augments PXR mediated induction. PXR recruits a number of coactivators which consequently modify chromatin structure and engage the transcription initiation complex

Constitutive androstane receptor

Constitutive androstane receptor is the closest relative of PXR and is present only in mammals suggesting that CAR arose from a duplication of an ancestral PXR gene (Reschly and Krasowski 2006). CAR expression is limited to human liver and kidney and very low in other tissues (Nishimura et al. 2004). The special feature of CAR is constitutive activity, i.e. transactivation in the absence of a ligand. This is because of the unusual structure of the CAR ligand-binding domain in which the AF-2 (helix 12) is stabilized to the active confirmation (Xu et al. 2004; Suino et al. 2004; Shan et al. 2004). Therefore activators of CAR involve both direct ligand-dependent and ligand-independent mechanisms. Similar to PXR also CAR ligands display species specificity (for review see Chang and Waxman 2006). Well-established agonists for mouse and human CAR are 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene (TCPOBOP) and 6-(4-chlorophenyl)imidazo[2,1-b][1,3]thiazole-5-carbalehyde O-[3,4dichlorobenzyl)oxime (CITCO), respectively (Chang and Waxman 2006). CAR constitutive activity can be inhibited with inverse agonist such as androstanol for mouse CAR (Forman et al. 1998).

In the absence of an activator CAR is retained in the cytosol in a complex with cytoplasmic CAR retention protein (CCRP) and Hsp90 (Kobayashi et al. 2003). Phenobarbital is a classical inducer of drug-metabolizing enzymes and CAR was shown to mediate phenobarbital induction of mouse Cyp2b10 gene (Honkakoski et al. 1998a). However, phenobarbital is not a direct ligand for CAR (Moore et al. 2000). Instead phenobarbital induces nuclear translocation of CAR by a mechanism involving 30 amino acid leucine-rich region in the C-terminus of CAR (Zelko et al. 2001) and subsequently ligand independent transactivation. The detailed mechanisms of phenobarbital induction and CAR translocation are still unclear. Phosphorylation events appear to be important. Upon phenobarbital induction cytosolic CAR complex recruits protein phosphatase 2A (Yoshinari et al. 2003). Furthermore, extracellular signal-regulated kinase affects CAR subcellular location (Koike et al. 2007). Phenobarbital also activates AMP-activated protein kinase, which has been suggested to be necessary for phenobarbital induction (Rencurel et al. 2006). In nucleus CAR heterodimerizes with RXR and binds to DNA-binding elements, of which many are shared with PXR. Several coactivators including SRC-1, PGC-1 and GRIP1 have already been shown to interact with CAR (Timsit and Negishi 2007).

Classical CAR targets include the CYP2B family members, in humans CYP2B6 (Honkakoski and Negishi 1997; Sueyoshi et al. 1999). CYP2B genes contain 51 bp phenobarbital responsive enhancer module (PBREM) in their regulatory regions that constitutes of two DR4 type nuclear receptor-binding sites binding CAR/RXR heterodimer and nuclear factor I (NFI) site in between (Honkakoski et al. 1998b). NFI site is needed for the full activity of the PBREM (Kim et al. 2001). Mouse PBREM favors CAR over PXR while human PBREM is less selective (Mäkinen et al. 2002). This may explain induction of human CYP2B6 by typical PXR ligands such as rifampicin (Goodwin et al. 2001). In addition to CYP2B subfamily, CAR regulates members in the CYP2C and CYP3A families (see Table 13). Also CYP1A2 and CYP2A6 are modestly induced by phenobarbital in human hepatocytes but direct involvement of CAR is still to be shown (Donato et al. 2000; Madan et al. 2003).

Crosstalk of xenosensing receptors

Detailed investigation of signaling mechanism involving xenosensing receptors has revealed extensive crosstalk with each other as well as with number of other factors (for recent review see Pascussi et al. 2008). The levels of crosstalk involve sharing of ligands, sharing of DNA-binding elements, receptor–receptor interactions, interaction with common coactivators and secondary regulation of the regulators. For example CAR and PXR share some ligands and DNA-binding elements. Furthermore, increasing evidence shows that all AhR, CAR and PXR participate in regulation of also many other important cellular functions besides regulation of metabolism and transport of xenobiotics. For example, PXR has been found to interact with several important transcription factors inducing HNF4α, FOXA2, FOXO1, CREB and to modify via them glucose and lipid metabolism (Li and Chiang 2005; Kodama et al. 2004, 2007; Nakamura et al. 2007). This emerging area of research should help us in the future to understand complex responses to xenobiotics.

Post-transcriptional induction

Significant post-transcriptional regulation has been shown only for few CYPs. CYP2E1 appears to be the only CYP regulated mainly at the post-translational level by xenobiotics such as ethanol, acetone, pyrazole and isoniazid (Song et al. 1989; Carroccio et al. 1994). CYP2E1 protein has short half life which is significantly increased by CYP2E1 inducing compounds. In rat liver CYP2E1 protein is degraded in two phases with half-lives of 7 and 37 h. However, after 10-day acetone treatment the fast degradation phase was abolished (Song et al. 1989). This stabilization may involve inhibition of proteosomal degradation pathway (Cederbaum 2006).

Regulation of mRNA stability has been found to mediate induction by xenobiotics in a few cases. Most extensively has been studied regulation of mouse CYP2A5 by pyrazole which involves binding of heterogeneous nuclear ribonucleoprotein A1 to the 3′-untranslated region of the CYP2A5 mRNA (Raffalli-Mathieu et al. 2002). Similar mechanism was reported to regulate also human CYP2A6 (Christian et al. 2004).

Experimental tools to study induction

Here, we present the current status of methods that have been developed and used to predict and detect induction in humans or in human-derived preparations. The methods involve either assaying various outcomes of agonist binding to the receptors that govern induction of CYP enzymes, or direct detection of induced CYP mRNAs and/or activities. The relative benefits and drawbacks of these methods are compared in Table 12.

In silico methods

The ability of a chemical to bind the cognate receptor responsible for CYP induction can be computationally estimated, and the evolving status of in silico models for AhR, CAR and PXR has been reviewed (Lewis et al. 2002; Ekins et al. 2002; Jacobs 2004; Poso and Honkakoski 2006; Schuster et al. 2006; Vedani et al. 2006). In silico screening can be performed in two fundamental ways. Briefly, in ligand-based methods (e.g. QSAR and pharmacophore models), molecular descriptors extracted from a set of known receptor ligands will provide rules that will classify other chemicals as potential CYP inducers via receptor-mediated mechanisms. The number and breadth of chemical structures used for model building and the use of separate validation sets are critical for the applicability and predictivity of the models. In the protein-based approach, candidate ligands are docked into the 3D structure or a homology model of the receptor, and evaluation of the binding fitness by various scoring functions will identify the chemicals with highest potential for receptor binding and CYP induction. Combinations of protein- and ligand-based methods have been also reported (e.g. Schuster and Langer 2005; Lemaire et al. 2007). In homology models, inappropriate template structures will lead to erroneous binding cavities for the modeled receptor (e.g. Xiao et al. 2002; Jacobs 2004), inability to distinguish between agonistic and antagonistic binding (Schuster and Langer 2005), and naturally, problems in prediction as noted elsewhere (Schuster et al. 2006; Windshügel et al. 2007). The dynamic flexibility of the receptor protein has been mostly omitted in the dockings although it is important for nuclear receptor function. Only recently, modeling employing induced fit (Sherman et al. 2006; Nabuurs et al. 2007; Repo et al., submitted) or with molecular dynamics simulations (Poso and Honkakoski 2006) has been done with nuclear receptors.

For AhR, comprehensive QSAR models covering several classes of polyhalogenated and polycyclic hydrocarbons have been constructed (e.g. Waller and McKinney 1992; Mekenyan et al. 1996). A later homology model yielded a quite robust model with cross-validation coefficient (q ²) of 0.58 or greater (Lo Piparo et al. 2006). In these studies, π–π stacking interactions, planarity of the compounds, and electronegative features of the ligand seem to determine high AhR-binding affinity. Quantum mechanical calculations that incorporate dispersion interactions and electrophilicity have successfully explained AhR binding and/or EROD inducibility to a level of 70% or above (Arulmozhiraja and Morita 2004; Gu et al. 2007). Due to the lack of 3D AhR structure, direct docking studies have not been done.

Only few purely ligand-based PXR studies have been conducted (Ekins and Erickson 2002; Jacobs 2004), in which PXR ligands seem to have one HBA and four hydrophobic regions (HPR) and models can classify ligands as either potent or weak PXR agonists. A machine-learning system based on structures of 128 PXR activators and 77 non-activators could correctly predict approximately four out of five PXR agonists (Ung et al. 2007). Combined approaches utilizing key residues within the five 3D PXR LBD structures (Poso and Honkakoski 2006; Xue et al. 2007) suggest the presence of one or two HBAs and three to five HPRs in PXR activators (Schuster and Langer 2005; Ekins et al. 2007; Lemaire et al. 2007). These ligand characteristics roughly correspond with the key features of the PXR LBD with Gln285 and His407 acting as hydrogen bond donors and several residues contributing to hydrophobic interactions.

There is only one pharmacophore model and one QSAR model on inhibition for the mouse CAR (Ekins et al. 2002; Jyrkkärinne et al. 2003). The studies published on human CAR deal with detailed homology modeling and molecular dynamics simulations (Jyrkkärinne et al. 2005; Windshügel et al. 2007). Hydrophobic features of the ligand and interactions with the key LBD residues including Phe161 appear to be critical for CAR activation. On-going studies in our laboratory have identified a novel 3D-QSAR model and shown that virtual screening can be used to identify novel CAR agonists (Jyrkkärinne et al., submitted; Küblbeck et al., submitted).

Cell-free assays

These in vitro assays fall into two different categories. The first category includes assays where the suspected inducer competes with a high-affinity radiolabeled ligand (such as [³H]TCDD for AhR, [³H]clotrimazole for CAR, and [³H]SR12183 for PXR) for receptor binding in an extract or a purified receptor preparation, and the amount of bound label is quantified by standard separation techniques (Roberts et al. 1990) or by scintillation proximity assays which have a higher throughput (Moore et al. 2000; Jones et al. 2000). However, it seems that ligand binding to PXR is not a very good measure of PXR activation or CYP3A4 induction (Zhu et al. 2004), and due to the high basal activity of CAR, inverse agonists cannot be distinguished from agonists with the radiolabel-binding assay (Poso and Honkakoski 2006).

The assays in the second category utilize a part of the signaling pathway to detect the activated, agonist-bound receptor. For example, ligand-elicited formation of AhR/ARNT complexes on dioxin-responsive DNA elements has been detected initially by gel electrophoresis (Seidel et al. 2000) and later by more rapid PCR, fluorescence or ELISA techniques (Sun et al. 2004; Fukuda et al. 2004; You et al. 2006). However, the DNA-binding assay is reported to give more false positive results than the cell-based AhR reporter assay (Seidel et al. 2000), and the gel-based assay for ligand-dependent DNA/receptor complexes for CAR and PXR has a narrow linear range (Mäkinen et al. 2002). A better choice is to measure ligand-dependent association of CAR and PXR with NR-interacting peptides from a co-activator such SRC-1 by gel electrophoresis (Frank et al. 2004) or more easily with fluorescent resonance energy transfer techniques (Moore et al. 2000; Maglich et al. 2003). In the latter case, other detection systems such as fluorescently labeled microbeads or chemiluminiscence could be adapted to increase assay sensitivity or throughput (Rouleau et al. 2003; Iannone et al. 2001).

Cell-based reporter assays

The chemically activated luciferase expression (CALUX) assay has been used to detect AhR-activating chemicals (Murk et al. 1996; Whyte et al. 2004). The CALUX is a very sensitive method, surpassing the sensitivity of the CYP1A-mediated EROD assay. The CALUX system is based on the stable transfection of a luciferase reporter, driven by TCDD-responsive gene promoters or DNA elements, into hepatoma cells that express a functional AhR such as rat H4IIE (Long et al. 2003), mouse H1L1.1c2 (Seidel et al. 2000) or human HepG2 cells (Yueh et al. 2005) Therefore, cell line- and species-dependent factors can affect the results as noted by Long et al. (2003). To streamline the detection system, green fluorescent protein (GFP) has been utilized as a simple and inexpensive reporter for AhR (Nagy et al. 2002).

Because PXR and CAR are not expressed at significant levels in most continuous cell lines, most activation assays have relied on transient transfection of CAR/PXR cDNA and an appropriate responsive reporter plasmid into the cells, followed by chemical treatment and reporter assays. In many cases, full-length CAR and PXR receptors control the expression of natural CYP3A4 or CYP2B6 gene promoters or PXR/CAR-responsive derivatives thereof (Ogg et al. 1999; Goodwin et al. 1999; Sueyoshi et al. 1999; El-Sankary et al. 2001; Moore et al. 2002; Mäkinen et al. 2002). To avoid competition by endogenous NRs for the response elements and to utilize one common reporter construct, the mammalian 1-hybrid system has been employed. Here, the yeast GAL4 DBD/NR LBD construct drives the ligand-dependent expression of a GAL4-responsive luciferase reporter (Mäkinen et al. 2002; Vignati et al. 2004; Jyrkkärinne et al. 2005). Such systems have been useful in delineating, e.g. species differences in CAR and PXR ligand responses (Moore et al. 2002; Mäkinen et al. 2002; Vignati et al. 2004). PXR- and CAR-responsive double-stable cell lines have been developed in order to streamline the procedure and to reduce variability from the transient transfection step (Trubetskoy et al. 2005; Lemaire et al. 2007).

However, the use for different promoter and reporter constructs, different cell lines with variable coactivator and corepressor contents, varying culture conditions and the lack of validation gives rise to widely different responses even with the established PXR ligands (Stanley et al. 2006). For examples, the activation by 10 μM rifampicin has been reported to vary from fourfold to more than 50-fold, depending on the assay set-up (e.g. Goodwin et al. 1999; El-Sankary et al. 2001; van Giersbergen et al. 2002a), and contradictory results for CAR modulators such as clotrimazole have been reported (Moore et al. 2000; Jyrkkärinne et al. 2005; Faucette et al. 2007). Nevertheless, due to the high throughput of the assay, transient transfection techniques have already been used to assess human PXR activation for hundreds of chemicals (Luo et al. 2002; Zhu et al. 2004; Persson et al. 2006; Sinz et al. 2006). This method seems acceptable for rapid screening of potential CYP inducers although their actual influence on the metabolic CYP activity cannot be predicted in this way (Luo et al. 2004). The high constitutive activity of CAR gives problems for detection of its ligands. For example, Moore et al. (2002) could not detect any CAR agonists in their assay in CV1 cells, and agonist responses have often been modest (Kobayashi et al. 2005; Stanley et al. 2006). Such problems have been circumvented by addition of CAR inverse agonists in the assay medium (Mäkinen et al. 2003; Jyrkkärinne et al., submitted) or by using a CAR splice variant with attenuated basal activity (Faucette et al. 2007). Both modifications introduce uncertainties about actual EC₅₀ values or ligand specificities, respectively. In our experience, careful selection of the host cell line is essential for a robust CAR activation assay (Jyrkkärinne et al. 2005; Küblbeck et al., submitted).

Continuous, immortalized and stem cell lines

Many human cell lines express AhR and its partner ARNT and therefore, increases in CYP1A1-mediated activities can be detected with the traditional EROD assay, with novel P450-Glo substrates or by measurement of CYP1A1 mRNA with quantitative RT-PCR methods (e.g. Behnisch et al. 2001; Westerink and Schoonen, 2007). Even then, different sources of the same cell line vary in the enzyme profile, requiring careful characterization (Hewitt and Hewitt 2004). It is notable, however, that in these cells CYP1A1 is induced while the preferred form in primary hepatocytes or in liver is CYP1A2 (Zhang et al. 2006). In contrast, the basal levels of other major CYP mRNAs (1A2, 2B6, 2Cs, 2D6, 3A4), CYP-mediated activities and their induction responses are generally very low in HepG2 or other hepatoma cells (Rodríguez-Antona et al. 2002; Donato et al. 2008) although the presence of PXR and the response of CYP3A4 mRNA to rifampicin has often been demonstrated (e.g. Sumida et al. 2000; Gómez-Lechón et al. 2001; Westerink and Schoonen 2007; Martin et al. 2008).

More substantial induction of CYP3A4 mRNA and/or activities and the presence of functional PXR has been reported in other cell lines of hepatic origin such as FLC-5 (Iwahori et al. 2003), Huh7 (Wang et al. 2006) and HC-04 (Lim et al. 2007). The same is true for some intestinal cell lines such as LS180 and LS174 but not for Caco-2 cells commonly used as a permeability model (Pfrunder et al. 2003; Hartley et al. 2006) in which CYP3A4 is up-regulated by the vitamin-D receptor (Schmiedlin-Ren et al. 1997). As far as we know, functional CAR is absent or at very low levels in all these cell lines. However, it seems possible to up regulate CAR, PXR and/or CYP expression to some extent by optimizing culture conditions (Korjamo et al. 2005; Osabe et al. 2008; Martin et al. 2008), by encapsulating cells in alginate (Elkayam et al. 2006) or by selective culture techniques (Rencurel et al. 2005). Most hepatomas that have been immortalized by, e.g. transfection of cell cycle inhibitors or telomerase or by other means appear to express some CYP enzymes. However, their utility for metabolism or induction studies is limited (Vermeir et al. 2005). One exception is the SV40-immortalized Fa2N-4 cell line (Hariparsad et al. 2008) which seems to reproduce PXR-dependent induction rather well. However, Fa2N-4 cells were unresponsive to CAR agonists indicating a similar lack of CAR expression as in other continuous cell lines.

A novel cell line HepaRG, derived spontaneously from a human hepatocellular carcinoma, has recently been introduced (Gripon et al. 2002). At high seeding density and after differentiation with 2% DMSO, HepaRG cells express the major CYPs and their regulators including CAR at or near the levels found in freshly isolated hepatocytes (Aninat et al. 2006). The extent of CYP mRNA induction in HepaRG varies according to the culture conditions, but often reaching values obtained with human hepatocytes (Aninat et al. 2006; Kanebratt and Andersson 2008). The long-term stability of HepaRG cultures as compared to primary hepatocytes makes these cells an attractive alternative for prolonged in vitro toxicity studies (Josse et al. 2008).

Many human stem cell lines display hepatocyte markers upon differentiation in culture and they have been anticipated to provide better in vitro cell models. However, in most cases, these cell lines have very low expression of CYP mRNAs, proteins or activities (Cai et al. 2007; Ek et al. 2007; Agarwal et al. 2008, Campard et al. 2008) with substantial expression of only CYP1A2 and CYP3A4/7 (Ek et al. 2007). Further work in the development, culture and differentiation of stem cells is thus warranted.

Fresh, cryopreserved and fetal primary hepatocytes

The golden standard and requirement by the authorities for induction studies are cultured primary human hepatocytes, which express all the relevant metabolic enzymes, transporters and their regulators. Their properties and difficulties in their procurement, variable quality, differences in genetics and prior exposure to inducers of donors and problems related to the time-dependent decreases in enzyme and transporter activity have been excellently reviewed (Lecluyse 2001; Gómez-Lechón et al. 2003; Parkinson et al. 2004; Hewitt et al. 2007a). The current status of how induction studies are and should be conducted has also been reviewed (Hewitt et al. 2007b, c). Briefly, hepatocytes from several donors are preincubated for 24–48 h in a (sandwich-type) monolayer in the presence of low concentrations of dexamethasone to allow stabilization of CYP expression. Cells are then exposed to increasing concentrations of inducing agents and established inducers for 3–4 days before measurement of CYP marker activities. It should be noted that the induction of CYP mRNA precedes and often exceeds that of enzyme activity, and therefore, mRNA levels are often quantified after 24 h of treatment. The assessment of induction at both mRNA and CYP activity levels will help identify inducing agents (e.g. ritonavir and troleandomycin) that also inhibit CYP activities, and the simultaneous measurement of cytotoxicity only adds to the versatility of the hepatocyte culture system (Kostrubsky et al. 1999; Luo et al. 2002; Madan et al. 2003). Induction of phase II transferases and transporters can also be seen (Soars et al. 2004; Sahi et al. 2003).

Cryopreserved hepatocytes have been plagued by the low and unpredictable extent of cell attachment after seeding and lower CYP activities than in fresh hepatocytes, rendering them questionable for induction studies (Li et al. 1999; Hengstler et al. 2000). Technical improvements have led to the wider acceptance of cryopreserved hepatocytes for CYP induction studies despite the fact that basal levels of some CYP activities are quite low (Garcia et al. 2003; Roymans et al. 2005; Hewitt et al. 2007b).

Fetal hepatocytes are able to proliferate in culture, but this advantage is offset by the fact that the profile and regulation of CYPs and other metabolizing enzymes do not match the adult situation due to, e.g. low expression of several transferases, high levels of CYP3A7 and marginal PXR-dependent induction (McCarver and Hines 2002; Matsunaga et al. 2004; Maruyama et al. 2007).

Liver slices

Because all cellular systems described above lack proper 3D contacts between hepatic cells, precision-cut liver slices have been used to investigate drug metabolism and clearance in different species (Lerche-Langrand and Toutain 2000). Despite the loss of CYP activities in prolonged culture, recent studies have shown that induction of the major CYP mRNAs can be detected in human liver slices (Martin et al. 2003; Persson et al. 2006). These data indicate that both AhR-, PXR- and CAR-dependent induction can be mimicked in slices although the extent of induction is often lower than in primary hepatocytes (Martin et al. 2003; Edwards et al. 2003). This is most likely due to poor permeability of the inducing compounds within the slice fragment. Variability in responses can also complicate the use of liver slices.

Animal models for human CYP induction

Investigations in an in vivo system would provide better estimation of clearance, pharmacokinetic/pharmacodynamic and toxicological consequences of enzyme induction. Due to species differences in the ligand/substrate/product specificities of the receptors and CYP enzymes, in vivo studies in laboratory animals are not predictive for human CYP induction. A classical example is the selectivity difference between humans and rodents for induction of CYP3A enzymes by rifampicin and pregnenolone 16α-carbonitrile, respectively, which is due to variation at key ligand-binding pocket residues in the corresponding PXR forms (Stanley et al. 2006). Genetically modified mice carrying the human AhR (Moriguchi et al. 2003), PXR (Xie et al. 2000; Ma et al. 2007) or CAR (Huang et al. 2003) in lieu of the murine receptor have been created. These humanized mice are useful for studies of in vivo-like responses to human inducers. Their wider use may be compromised by the presence of other mouse proteins that are relevant for compound permeability, metabolism and potential species differences in the receptor’s target gene repertoire. To overcome such problems and to provide a renewable source of human hepatocytes, immunodeficient mice in which transplanted human hepatocytes can colonize the liver have been developed (Tateno et al. 2004). In these chimeric mice, at least the profile and inducibility of CYP3A4 and CYP1A2 seems to be similar to those in human hepatocytes (Nishimura et al. 2005; Emoto et al. 2008; Katoh et al. 2008). A recent study reported a more advanced mouse strain (Azuma et al. 2007) that harbors transplanted human hepatocytes in which a broad array of CYPs, other enzymes and transporters and CAR and PXR are expressed at or near normal levels. More studies on the applicability of these mice for toxicological research are imminent.

In vivo studies

The focus of human in vivo induction studies has been on the CYP3A4 enzyme and its reaction products from either endogenous or exogenous substances. The increases in urinary excretion of 6β-hydroxycortisol, typically less than twofold by PB but up to sevenfold by rifampicin, have been widely used to reflect changes in the CYP3A4 activity (Galteau and Shamsa 2003). Hydroxylated metabolites of bile acids, produced by CYP3A4, can be quantified in the urine of test subjects (Furster and Wikvall 1999; Bodin et al. 2005), and plasma levels of 4β-hydroxycholesterol have been reported to reflect CYP3A4 activity and to be significantly elevated by several CAR/PXR activators (Bodin et al. 2001; Diczfalusy et al. 2008). For other CYPs, similar endogenous biomarkers have not yet been identified. Of the earliest exogenous test substances, the erythromycin breath test measures [¹⁴C]CO₂ in the exhaled air formed by CYP3A4-mediated N-demethylation of intravenously administered radiolabeled erythromycin (Watkins et al. 1989; McCune et al. 2000). The safety issues, together with relative insensitivity and tediousness of the assay, limit its use mainly to clinical studies. The changes in pharmacokinetic parameters or metabolic ratios of specific probe substrates such as midazolam (CYP3A4), caffeine (CYP1A2) and others (see “Induction of CYP enzymes in humans in vivo”) have been detected after pretreatment with inducers, although the extent of induction is highly dependent on the selectivity of metabolism and disposition of the test substrate (Gurley et al. 2002; Niemi et al. 2003a, b; Faber et al. 2005; Hukkanen et al. 2005; Loboz et al., 2006). Identifying better CYP form-specific probe substances and the use of cocktail protocols could provide a wider test battery for human in vivo induction studies.

Induction of CYP enzymes in humans in vivo

The two preceding sections show that there is considerable understanding of basic mechanisms of induction as well as a variety of tools and experimental setups to study induction in silico, in vitro and in vivo. It is equally evident from the above sections that despite the considerably increased knowledge about induction in human, humanized, or human-derived or mimicking systems, there is a notable gap between experimental studies and clinical observations. Consequently, in the following sections and Table 13 we will present a systematic survey of the inducers of the specific CYP enzymes in vivo in humans. Studies with experimental animals and human cell lines including human primary hepatocytes, which have demonstrated many other in vitro inducers, are not included here. In addition, established in vivo inducers (such as clotrimazole, troglitazone and moricizine), that are no longer in systemic clinical use and are without current toxicological interest, are not reviewed here due to space constraints.

CYP1 family

Although the induction of both CYP1A1 and CYP1B1 enzymes by various inducers are classic examples of CYP induction, human in vivo induction of CYP1A1 and CYP1B1 has been difficult to study due to their very low hepatic levels when compared to CYP1A2 (Chang et al. 2003). Since these three enzymes have overlapping substrate specificities and their induction share regulatory features with each other as discussed previously, the induction seen in phenotyping studies usually reflects the induction of CYP1A2 in liver. Only when gene and enzyme-specific methods are applied at tissue level can it be construed that CYP1A1 and CYP1B1 are induced. With such methods, smoking has been shown to induce CYP1A1 and CYP1B1 in tissues such as lung, liver (only mRNA is detected) and placenta (Chang et al. 2003; Hakkola et al. 1998; Hukkanen et al. 2002; Huuskonen et al. 2008). Also, topical coal tar and ultraviolet-B radiation treatments induce CYP1A1 and CYP1B1 in skin (Katiyar et al. 2000; Smith et al. 2006). Furthermore, CYP1B1 is induced in peripheral leukocytes of waste incinerator workers, coke oven workers and smokers (Hanaoka et al. 2002; Hu et al. 2006; Lampe et al. 2004; van Leeuwen et al. 2007). The induction of CYP1A1 mRNA and CYP1A protein by omeprazole occurs in duodenum (Buchthal et al. 1995; McDonnell et al. 1992). Duodenal CYP1A1 mRNA and protein are also induced by charbroiled meat containing diet (Fontana et al. 1999).

The induction of CYP1A2 in vivo has been widely studied. AhR ligands such as indole-3-carbinole (in cruciferous vegetables), PAHs (in tobacco smoke, charbroiled meat and coffee) and TCDD induce CYP1A2-associated activities in humans (Djordjevic et al. 2008; Faber et al. 2005; Landi et al. 1999; Reed et al. 2005). Topically applied coal tar induced CYP1A2 mRNA in skin (Smith et al. 2006). The activity of CYP1A2 enzyme as assessed by theophylline, clozapine, tizanidine, ropivacaine or caffeine pharmacokinetics is induced by pharmaceuticals such as omeprazole (Ma and Lu 2007), rifampicin (Backman et al. 2006a, b), phenytoin (Miller et al. 1984; Wietholtz et al. 1989), carbamazepine (Parker et al. 1998), phenobarbital (Landay et al. 1978), pentobarbital (Dahlqvist et al. 1989), secobarbital (Paladino et al. 1983), sulfinpyrazole (Birkett et al. 1983) and ritonavir (Hsu et al. 1998; Yeh et al. 2006).

CYP2A6

CYP2A6 has recently been shown to be under the regulation of ERα (Higashi et al. 2007a, b). This is reflected in the increased CYP2A6-mediated nicotine and cotinine metabolism in oral contraceptive users (Benowitz et al. 2006; Berlin et al. 2007). Subjects taking combination oral contraceptives and estrogen-only contraceptives had accelerated nicotine metabolism, whereas progesterone-only contraceptives did not affect nicotine metabolism (Benowitz et al. 2006). The induction of CYP2A6 by estrogens is supported by the findings that female gender and pregnancy induce nicotine metabolism (Benowitz et al. 2006; Berlin et al. 2007; Dempsey et al. 2002). Also, CYP2A6 protein is induced in the glandular cells of the endometrium in the proliferative phase when compared to the secretory phase (Higashi et al. 2007a, b). There is some evidence for the induction of CYP2A6 in vivo by phenobarbital and other anticonvulsant drugs. Coumarin phenotyping shows increased metabolism in epileptic patients treated with carbamazepine, phenobarbital, and/or phenytoin (Sotaniemi et al. 1995). Two-day treatment with phenobarbital prior to a liver biopsy resulted in induction of metabolism of nicotine to cotinine in hepatocytes (Kyerematen et al. 1990). Liver microsomes from phenobarbital-treated patients have higher amounts of CYP2A6 protein than microsomes from untreated patients (Cashman et al. 1992; Yamano et al. 1990). A recent study showed that artemisinin (antimalarial) administration affected significantly the pharmacokinetics of both nicotine and coumarin suggesting induction of CYP2A6 (Asimus et al. 2008).

CYP2B6

Although induction of CYP2B enzymes by phenobarbital is the archetypal example of enzyme induction in experimental animals, the induction of human CYP2B6 in vivo is less well characterized. However, the induction of CYP2B6 has been shown to occur with several drugs as evidenced by the changes in the pharmacokinetics of bupropion by rifampicin, ritonavir and carbamazepine (Ketter et al. 1995; Kharasch et al. 2008; Loboz et al. 2006), efavirenz by rifampicin and carbamazepine (Sustiva (efavirenz); Lopez-Cortes et al. 2002), S-mephenytoin (N-demethylation) by several artemisinin antimalarials (Elsherbiny et al. 2008; Simonsson et al. 2003) and cyclophosphamide by phenytoin and phenobarbital (Jao et al. 1972; Slattery et al. 1996). Carbamazepine use has been associated with high hepatic CYP2B6 protein content and enzymatic activity in two carbamazepine-exposed liver samples when compared to 85 non-exposed liver samples (Desta et al. 2007). Additionally, metamizole was recently shown to induce CYP2B6 protein and activity in human liver in vivo (Saussele et al. 2007).

CYP2C8

Studies with pioglitazone, rosiglitazone and repaglinide as substrates show induction in CYP2C8 activity by rifampicin (Bidstrup et al. 2004; Jaakkola et al. 2006a, b; Niemi et al. 2000, 2004; Park et al. 2004). Rifampicin also induces CYP2C8 protein in jejunal enterocytes in vivo (Glaeser et al. 2005). Paclitaxel has been used as a CYP2C8 probe (Rodriguez-Antona et al. 2007) and paclitaxel metabolism is induced in patients treated with phenytoin, carbamazepine or phenobarbital (Chang et al. 1998; Fetell et al. 1997). However, it seems that these antiepileptics preferentially induce the minor CYP3A4-mediated pathway (Chang et al. 1998; Cresteil et al. 1994).

CYP2C9

The induction of CYP2C9 activity has been shown with rifampicin (Kay et al. 1985; Niemi et al. 2001; O’Reilly 1974; Williamson et al. 1998; Zilly et al. 1975), phenobarbital (Goldberg et al. 1996; Orme and Breckenridge 1976; Udall 1975), pentobarbital (Yoshida et al. 1993), secobarbital (O’Reilly et al. 1980; Udall 1975), carbamazepine (Herman et al. 2006), St. John’s wort (Jiang et al. 2006; Jiang et al. 2004), ritonavir (in combination with lopinavir)(Lim et al. 2004; Yeh et al. 2006), aprepitant (Depre et al. 2005; Shadle et al. 2004) and bosentan (van Giersbergen et al. 2002b; Weber et al. 1999) when studied with warfarin, losartan, phenytoin, tolbutamide or glibenclamide pharmacokinetics (reviewed in (Miners and Birkett 1998). Rifampicin also induces CYP2C9 protein in jejunal enterocytes (Glaeser et al. 2005). There is some evidence concerning the inducing effect of phenytoin on CYP2C9 but good quality studies on the subject are lacking (Dickinson et al. 1985; Levine and Sheppard 1984).

CYP2C19

CYP2C19 activity measured using S-mephenytoin, omeprazole or hexobarbital as probes is induced by rifampicin (reviewed in Desta et al. 2002), phenobarbital (Richter et al. 1980), pentobarbital (Heinemeyer et al. 1987), St. John’s wort (Wang et al. 2004a, b), ritonavir (in combination with lopinavir) (Yeh et al. 2006) and artemisinin antimalarials (artemisinin, artemether, arteether) (Asimus et al. 2007; Elsherbiny et al. 2008; Mihara et al. 1999; Svensson et al. 1998). Phenobarbital treatment induces CYP2C19 protein and activity in liver in vivo (Lecamwasam et al. 1975; Perrot et al. 1989).

CYP2E1

As discussed above, both transcriptional and posttranscriptional mechanisms influence the induction of CYP2E1 with stabilization of mRNA and protein having major significance in contrast to many other CYP forms (reviewed in Lieber 1999). Only a few human in vivo CYP2E1 inducers are known. The characteristic inducer is ethanol as shown with increased CYP2E1 mRNA and protein in liver biopsies and as increased chlorzoxazone hydroxylation after ethanol administration and in alcoholics (Girre et al. 1994; Lucas et al. 1995; Oneta et al. 2002; Perrot et al. 1989; Takahashi et al. 1993; Tsutsumi et al. 1989). In addition, CYP2E1 mRNA and protein are induced in peripheral blood lymphocytes in alcoholics and correlate with chlorzoxazone clearance in vivo (Raucy et al. 1999; Raucy et al. 1997). Also the full-term placentas of heavily drinking mothers express increased levels of CYP2E1 protein (Rasheed et al. 1997). CYP2E1-related activities are induced by isoniazid and smoking (Benowitz et al. 2003; Chien et al. 1997; Mazze et al. 1982; O’Shea et al. 1997; Zand et al. 1993). The induction of CYP2E1 protein in the brain of the smoking alcoholics when compared to nonalcoholic nonsmokers has been proposed (Howard et al. 2003). A relatively long-term administration (28 days) of St. John’s wort induces chlorzoxazone hydroxylation (Gurley et al. 2002, 2005). Several pathologic conditions such as diabetes, nonalcoholic steatohepatitis and obesity have been associated with the increased levels of CYP2E1 (reviewed in Lieber 2004).

CYP2S1

The AhR-regulated CYP2S1 enzyme has been implicated in the chemical carcinogenesis (Saarikoski et al. 2005). CYP2S1 is induced by coal tar, ultraviolet radiation and all-trans retinoid acid in skin (Smith et al. 2003b). Smoking may induce CYP2S1 in bronchoalveolar macrophages but not in pulmonary bronchi or placenta (Huuskonen et al. 2008; Thum et al. 2006).

CYP3A4

A multitude of compounds induce CYP3A4. Since the literature on human CYP3A4 induction is vast and rapidly expanding, we refer the reader to a recent review (Luo et al. 2004) and aim to complement its list of inducers with latest findings and to fill in certain omissions. As reviewed comprehensively by Luo and coauthors, there is convincing evidence for the induction of CYP3A4 activities in vivo by carbamazepine, phenobarbital, phenytoin, rifampicin, ritonavir, St. John’s wort and topiramate, as well as troglitazone (withdrawn due to hepatotoxicity) (Luo et al. 2004).

In addition to rifampicin, another rifamycin antibiotic rifabutin induces CYP3A4 activities (Barditch-Crovo et al. 1999; Perucca et al. 1988). Other PXR ligands with proven in vivo CYP3A4 inducing properties include bosentan (Dingemanse and van Giersbergen 2004), sulfinpyrazone (Staiger et al. 1983; Wing et al. 1985) and artemisinin antimalarials (Asimus et al. 2007). Aprepitant, which is predicted to be a PXR ligand based on in silico methods (Ekins et al. 2006), induces CYP3A4 slightly as studied with midazolam as a probe (Shadle et al. 2004). Modafinil and its R-enantiomer armodafinil induce CYP3A4 based on their effects on triazolam, midazolam and ethinyl estradiol pharmacokinetics (Darwish et al. 2008; Robertson et al. 2002). Besides ritonavir, antiretrovirals such as efavirenz (Fellay et al. 2005; Mouly et al. 2002), nevirapine (Dailly et al. 2006; Mildvan et al. 2002; Solas et al. 2004) and amprenavir (Justesen et al. 2003; Kashuba et al. 2005) induce CYP3A4-related activies. Metamizole was recently shown to induce CYP3A4 protein and activity in human liver in vivo (Saussele et al. 2007). Several corticosteroids such as dexamethasone (McCune et al. 2000; Watkins et al. 1989), methylprednisolone (Kuypers et al. 2004), prednisolone (van Duijnhoven et al. 2003) and prednisone (Anglicheau et al. 2003) induce CYP3A4 activities. However, smaller doses of corticosteroids do not seem to induce CYP3A4 (Villikka et al. 1998, 2001).

CYP3A5

The study of CYP3A5 induction in vivo has been hampered by the overlapping substrate and inducer specificities with CYP3A4, and lower hepatic expression levels when compared to CYP3A4. Thus, the induction of CYP3A-specific activities is usually construed as a sign of CYP3A4 induction. In analogy to CYP1A1 and CYP1B1 in relation to CYP1A2, only when gene and enzyme-specific methods are applied at tissue level can it be ascertained if CYP3A5 is induced in vivo. There is only limited evidence of the CYP3A5 induction in vivo. Rifampicin administration induced duodenal CYP3A5 mRNA in three of eight subjects with some indication of the effect of CYP3A5 genotype (Burk et al. 2004). Phenobarbital did not induce hepatic CYP3A5 protein in phenobarbital-treated children (Busi and Cresteil 2005). A recent study on the induction of CYP enzymes in skin biopsies showed a significant induction of CYP3A5 mRNA by topical administration of the glucocorticoid clobetasol 17-propionate (Smith et al. 2006). To the best of our knowledge, there is no evidence of the induction of CYP3A7 by xenobiotics in vivo.

Clinical and toxicological consequences of enzyme induction

As discussed above, the scope of compounds capable of inducing CYP enzymes in humans is extensive. The clinical and toxicological significance of CYP induction depends on several factors. These include inducer-specific aspects such as the potency of the inducer, the dose and the concentration of the inducer needed for the induction to occur, the duration of the exposure needed for the induction to happen, the metabolic properties of the inducer, the length of the exposure to the inducer, the duration of the induction once the inducer is withdrawn, the route of the exposure (e.g. orally, topically or by inhalation), and the anatomical location of the CYP enzymes induced (e.g. intestine, skin or lung). Some inducers are also inhibitors of the same CYP enzyme they induce further complicating the situation. Especially many of the antiretrovirals with CYP3A4 inducing properties are potent CYP3A4 inhibitors as well (Antoniou and Tseng 2005). Significant induction of CYP3A4 by a potent inducer such as rifampicin may not have clinical consequences if used short term in the setting where no comedication is metabolized by CYP3A4.

On the other hand, if the duration of rifampicin treatment is prolonged, complications, such as drug-induced osteomalacia via the CYP3A4-mediated catabolism of vitamin-D (1,25-(OH)2-D3) (Zhou et al. 2006), may arise even without co-medications. The properties of the comedications are important factors predicting the consequences of induction; the risk of significant interactions is increased if concurrently administered drugs have low therapeutic indexes or high first-pass metabolism, or the induced CYP is the major pathway of their metabolism (Park et al. 1996). For example, the AUC of oral midazolam (a drug with high first-pass metabolism and CYP3A4-mediated metabolism) is reduced by 96% if given after rifampicin induction when compared to AUC in noninduced state (Backman et al. 1996).

Another multitude of factors affecting the consequences of the inducers are the host-specific aspects such as genetic variations in the transporters of the inducer, in the enzymes metabolizing the inducer and in the receptors mediating the induction and in the CYP genes themselves as well as diseases, nutritional status, gender and age of the host. All these factors lead to marked interindividual variability in the induction of CYP enzymes (Tang et al. 2005). In general, the lower the baseline enzymatic activity, the higher the induction achieved with inducers (McCune et al. 2000; Vesell and Page 1969) unless the low baseline activity is caused by CYP null alleles or other genetic factors.

The most obvious characteristic affecting the importance of a specific inducer is the magnitude of the induction; inducer may cause a detectable induction of a certain CYP enzyme in the controlled setting of a pharmacokinetic study but if the magnitude of the induction is minor, it might not have any discernible ramifications in clinical setting. Nevertheless, the interplay between the patient’s genetic makeup, the dose of the inducer and the comedications often complicates the issue. As an example, omeprazole, an established inducer of CYP1A2, is usually considered not to have any significant inducing effect at the dose of 40 mg per day used clinically. However, the same dose is effective in inducing CYP1A2 in patients who are PMs of omeprazole (mediated by CYP2C19) and furthermore, a higher dose of 120 mg is sufficient to cause induction also in extensive metabolizers (Ma and Lu 2007). Although not yet studied, this type of scenario might also transpire if a potent CYP2C19 inhibitor like fluconazole was administered together with omeprazole. Another example of the interplay between genetic makeup of the host and the induction potential is the induction of CYP2E1 by isoniazid, which is only seen in patients with slow N-acetylation status (a major pathway of isoniazid metabolism) leading to higher isoniazid concentrations (O’Shea et al. 1997).

Generally, for drugs that are active in their parent form, induction may increase the drug’s elimination and decrease its pharmacological effect. A well-established case is the use of CYP3A4 inducers such as St. John’s wort together with cyclosporine in organ transplant patients leading to reduced cyclosporine concentrations and organ rejection (Zhou et al. 2004). Other well-known examples are the increased risk of pregnancy with oral contraceptives when combined with enzyme inducing antiepileptics or rifampicin, and the problems encountered with warfarin anticoagulation when inducers are started (reduction in anticoagulant activity) or when inducers are withdrawn (increased risk of hemorrhages) (Perucca 2006; Zhang et al. 2007). For prodrugs, compounds that require metabolic activation and whose effects are produced by the active metabolites, enhanced pharmacodynamic effects may be expected. Thus, there is some evidence for the enhanced antiplatelet activity of prodrug clopidogrel with CYP3A4 inducers such as rifampicin and St. John’s wort (Lau and Gurbel 2006). In addition, induction may lead to increased toxicity if the increased metabolism of the parent compound is accompanied by the increase in exposure to a toxic metabolite. For example, anecdotal evidence links the CYP2E1 induction by ethanol to the increased risk of carbon tetrachloride toxicity in a setting of accidental occupational exposure to carbon tetrachloride of fire extinguishing liquids (Manno et al. 1996).

As is well known, CYP1 family of enzymes is of importance due to their toxicological significance in the activation of several procarcinogens to more mutagenic forms. This is reflected in the finding that the expression levels of pulmonary CYP1A1 mRNA and protein correlate positively with the aromatic/hydrophobic DNA adduct levels in human lung tissue (Cheng et al. 2000; Mollerup et al. 1999). However, the latest paradigm concerning the CYP1A enzymes based on studies in CYP1 knockout mouse lines emphasizes the beneficial effects of these enzymes in protection against chemical carcinogenesis in intact organisms (Ma and Lu 2007; Nebert and Dalton 2006). This kind of hypothesis may explain the finding that phenobarbital treatment has been associated with decreased amounts of the aromatic amine-hemoglobin adducts in smoking epileptics (Wallin et al. 1995). The detailed picture of the significance of the carcinogen-metabolizing CYP enzymes in general and their induction specifically is yet to emerge.

Research needs and future trends

Inhibition and drug–drug interactions

Despite some deficiencies and uncertainties in in vitro inhibition screens (see “Inhibition: in vitro–in vivo extrapolation”), they are used widely in drug industry (and also increasingly in food and chemical industries) and there seems to be a general consensus that their performance is relatively good. Drug regulatory agencies have provided guidances in the hope of harmonizing the approaches to screen drugs by in vitro and in vivo investigations. Harmonization is intimately linked with the standardization of CYP probe substrates, inhibitors and inducers and with the development of classification systems to improve the risk communication to all concerned stakeholders (Bjornsson et al. 2003).

Despite promising results of in vitro inhibition screens, there are still some areas which need further development. One is the relatively narrow focusing on CYP enzymes. Admittedly they are of primary importance for drug and chemical metabolism, but there are a huge number of other drug-metabolizing enzymes, which may be of importance for individual drugs and chemicals and which are not covered by current screening systems. This area certainly needs much further work in the future. A partial solution to this problem may be a use of substrate loss assays in a cell (hepatocyte)-based assays for inhibition screens, because such a system should be able to take into consideration the totality of enzymes metabolizing a particular drug or a chemical.

In silico tools are also available for predicting substrates and inhibitors of CYP enzymes as well as ligands of AhR, PXR, and CAR, but their prediction power needs to improve before they can be routinely used in drug metabolism or safety evaluation. Another important question concerns extrapolation models; the precision of prediction is dependent on what kind of model is selected. The application of some physiologically based pharmacokinetic models, such as Simcyp or PK-Sim (see Pelkonen et al. 2008), seem capable of providing a fairly reliable projection to the in vivo situation. Modeling and simulation, however, always brings forth the necessity of validation, which is a primary concern from the application point of view of any in silico (or in vitro, for that matter) approach.

Induction

Looking back at the phenomenal development of nuclear receptor field, one cannot help thinking that their role should be more or less completely defined in the near future. On the other hand, increasing number of transcription factors, coactivators, corepressors, their interwoven roles and cross-talk between various regulatory pathways are making the regulatory studies very complex and we expect that new factors and concepts will emerge to better understand the regulation of induction phenomenon. Actually it seems that ‘induction’ as understood until now is only a small part of regulatory devices to keep homeostasis and adjust the organism to a changing chemical and biological internal milieu and environment. However, it may be difficult to predict these developments at least to what extent they affect drug development, toxicological risk assessment and clinical drug therapy.

In our opinion, three major fields will emerge within development of experimental tools. First, receptor-based reporter assays will become a widely used system to predict CYP induction although their validation with primary hepatocytes will be needed. Second, further research on renewable cell lines is warranted to provide an unlimited source of hepatocyte-mimicking cells for metabolism, induction and toxicity studies. Third, improved in silico algorithms will be needed that can help distinguish true receptor agonists from large sets of chemicals. No doubt, these developments, if fulfilled, will help enormously especially drug development, but also chemical risk assessment.

The development of hepatocyte-mimicking cells will also help the targeting of in vivo induction studies. Currently there is a large gap between a profusion of compounds suggested to be inducers on the basis of current in vitro tools and reliable in vivo induction data in humans. Another problem in characterizing induction in vivo (and partially in vitro cellular systems) is the outdated terminology (Smith et al. 2007). The response should be characterized by both potency (e.g. EC50 or ED50) and maximal response in a given system to be able to compare various compounds and their effects at the cellular or organism levels. This would help also judge about whether the induction by a given compound would cause clinically relevant consequences (Smith 2000).