Advances in the generation of mouse models to elucidate the pathways of drug metabolism in rodents and man

Animal models remain a cornerstone of drug development; however, such studies have a major limitation in that pathways of drug metabolism can vary markedly from those in humans, resulting in profound species differences in drug efficacy, toxicity and pharmacokinetics. Advances in transgenic technology have now allowed the generation of animal models, in which endogenous enzymes are replaced with their human counterparts, thereby creating a system that more closely reflects the human situation. In this article, we will describe advances in this research area, focusing on the strengths and limitations of recent mouse models that will undoubtedly assist in overcoming some of the issues associated with species-specific differences in drug-metabolism studies.

An adaptive response system has evolved in mammalian species, which protects them against a chemically challenging environment. This detoxification system consists of a complex network of xenoreceptors, drug-metabolizing enzymes (DMEs) and transporters, and it does not differentiate between synthetic chemicals, drugs or environmental toxins. Although transporters undoubtedly play an important role in drug disposition, they will not be discussed further in this article and the reader is referred to recent publications [1,2]. Xenoreceptors, for example, the nuclear receptors constitutive androstane receptor (CAR) and pregnane X receptor (PXR), are activated by a range of endogenous molecules or xenobiotics and regulate the expression of DMEs. DMEs may be categorized into Phase I and II enzymes; the former largely relies on the biotransformation of xenobiotics by enzymes encoded by the cytochrome P450 (CYP) superfamily. Parent compounds and their Phase I metabolites can either be eliminated directly from the body (via transporters) or may be subject to further metabolism by Phase II enzymes prior to elimination. Phase II enzymes are encoded by a diverse group of gene families but share the common feature of adding a chemical moiety to parent compounds or Phase I metabolites, such as glucuronide (UDP-glucuronosyl transferases [UGTs]), glutathione (glutathione transferases [GSTs]) or sulphate (sulfotransferases [SULTs]). The overall outcome of the actions of Phase I and Phase II enzymes is to render molecules more hydrophilic and, thus, more easily excretable from the body.

Owing to the availability of murine embryonic stem cells and the feasibility of targeted manipulation of genes of interest, mice have become a favored model organism to elucidate the function of genes in drug metabolism. A particularly useful approach in order to study the contribution of a nuclear receptor or DME to the metabolism of a specific compound in vivo is the targeted disruption or ‘knockout’ of the corresponding gene. Many DME-knockout mice have become available over the past decade, and have been used to discover or confirm the role of individual DMEs in the metabolism of specific compounds (Box 1). For example, mice deprived of the Cyp2e1 gene are resistant to the hepatotoxic effects of acetaminophen, chloroform and carbon tetrachloride [3–5], and CAR-knockout mice exhibit significantly increased sensitivity to zoxazolamine, failing to recover from the paralysis induced by this drug, and also display increased cocaine hepatotoxicity as a consequence of their inability to upregulate expression of the CYP enzymes that would normally metabolically inactivate these drugs [6]. Recently, gene-targeting techniques have been refined such that it is possible to delete or alter genes in a conditional manner, for example, spatio–temporal or tissue specific [7,8]. This is exemplified by deletion of NADPH-CYP oxidoreductase (POR), the unique electron donor to microsomal CYPs; whereas a global knockout of POR is embryonically lethal [9,10], conditional deletion in the liver using Cre/loxP technology allowed the generation of a hepatic reductase-null viable mouse line [11,12], which has been used to investigate the relationship between drug metabolism and disposition and toxicokinetics, using the anticancer prodrug cyclophosphamide as an example [13,14]. A further example of conditional gene knockout in the area of drug metabolism is the hepatic deletion of microsomal cytochrome b5, a hemoprotein that transfers electrons to the CYP system and has been the subject of vigorous debate over several decades as to its role in CYP function. Determination of the pharmacokinetics of several probe drugs in hepatic b5-null (HBN) mice demonstrated unequivocally that cytochrome b5 exercised a stimulatory effect on CYP activity in vivo[15].

As the technology to carry out defined genetic alterations in rats, a favored model of the pharmaceutical industry, has been described recently, it appears conceivable that comparable xenoreceptor- or DME-knockout rats will become available in the near future [16].

Although the combined use of knockout animals can provide important insights into the function(s) of proteins involved in drug metabolism, there are limitations to this approach caused mainly by interspecies differences in the interaction of DMEs and nuclear receptors with drugs and chemicals. Thus, compounds may be metabolized variably between species, due to differential binding of ligands to xenoreceptors and/or to differences in the catalytic activity or multiplicity of DMEs themselves, therefore affecting the bioavailability, toxicity and efficacy of a drug and making extrapolation of data from animals to man complex and fraught with difficulties.

Accordingly, much effort has been made in recent years to generate humanized mouse models for drug-metabolism studies, encompassing both xenoreceptors and DMEs [17]. In addition to older humanized models for PXR and CAR, in which the orthologous mouse receptor was deleted and replaced with a random transgene expressing human PXR or CAR cDNAs under the control of the liver-specific albumin promoter, second-generation models have been described recently. In one such approach, a PXR-humanized mouse model was created by random integration of the entire human PXR gene and its promoter on a Pxr^-/- background [18]. Pretreatment of these mice with the human-specific PXR agonist rifampicin resulted in faster metabolism of midazolam, while no change in midazolam metabolism was observed in wild-type mice. The PXR-humanized mice, therefore, accurately reflected the drug interaction found in humans when drugs are coadministered with rifampicin. In another approach, PXR- and CAR-humanized mice were generated by a knockin strategy to express human PXR and CAR under the control of their corresponding mouse promoters, accompanied by a deletion of the endogenous genes [19]. In this way, a panel of models was established containing all possible combinations of single or double humanized and knockout PXR and CAR mice. These models were used to show that, in contrast to previous in vitro results, the in vivo induction of Cyp3a11 and Cyp2b10 by phenobarbital was mediated exclusively by CAR and, by using an in vivo pharmacokinetics approach, a difference in the metabolism of Cyp3a11 and Cyp2b10 probe substrates was demonstrated between wild-type and CAR-humanized mice after treatment with a human-specific CAR agonist. PPARα-humanized mice have been generated, in which human PPARα was expressed either with a tetracycline responsive system [20] or under the control of the human PPARα promoter [21]. PPARα target genes are induced by PPARα ligands in these humanized models, but in contrast to wild-type mice, there is no hepatocellular proliferative response following PPARα activation, therefore reflecting the human situation. Thus, PPARα-humanized mice can help to assess the true risk of PPARα ligands to induce nongenotoxic liver tumor promotion. Aryl hydrocarbon receptor (AHR) humanized mice, generated by knockin of the human AHR cDNA into the mouse Ahr gene locus and under the control of the mouse promoter [22], show induction of typical AHR target genes following treatment with AHR ligands, but they are functionally less responsive to the environmental toxicant 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) than wild-type mice, again reflecting the human situation.

As with the xenoreceptors, species-specific differences in the functionality and multiplicity of DMEs are now well known. A striking example is the CYP2D family, with only one functional gene, CYP2D6, in humans compared with nine genes in the mouse. Similar differences also exist for other CYP genes, such as the CYP3A family. Therefore, CYP-humanized mice can be valuable tools for drug-metabolism studies. A variety of random transgenic mice expressing human CYPs have been described in recent years, including mice expressing human CYP2D6, CYP3A4, CYP2E1, CYP1A1, CYP1A2, CYP1B1 and others [23]. Debrisoquine, used as a probe substrate of CYP2D6 activity in humans, is metabolized efficiently in mice expressing human CYP2D6 [24] and the rate of midazolam metabolism and clearance after oral administration was increased in CYP3A4-transgenic mice [25]. A major drawback of most of the current transgenic mice expressing human CYPs is the presence of the homologous mouse genes in these models. As a consequence, both mouse and human enzymes could potentially contribute to the metabolism of a compound, resulting in a mixed profile of drug metabolism. This can be overcome by expressing the human DME gene on a null background for the corresponding mouse gene. For example, human CYP1A1/1A2 has been expressed in a mouse lacking Cyp1a1/1a2[26], or human CYP3A4 has functionally replaced eight mouse Cyp3a genes in either the liver or the intestine [27]. In the latter model, absorption of docetaxel into the bloodstream is decreased by CYP3A4 expression in the intestine, while hepatic expression increased the systemic docetaxel clearance, showing that these mice could be used to analyze the relative contribution of liver and intestine to the CYP3A4-mediated metabolism of a drug. However, further studies on the underlying Cyp3a-knockout mouse showed that the metabolism of the human CYP3A4 probe substrate midazolam was essentially unchanged when compared with wild-type mice [28], an effect explained by a compensatory upregulation of Cyp2c enzymes, which are upregulated in the Cyp3a-knockout mouse. In order to achieve a useful functional humanization of CYP-mediated metabolism of a compound, it may therefore be necessary to delete such compensatory enzyme systems.

Relatively little information on species differences related to Phase II metabolism is available; accordingly, reports of Phase II humanized mouse models are rare in the literature. However, one such model was generated by random integration of the human UGT1 locus into the mouse genome [29]. The utility of this model to elucidate human UGT1 metabolism of a drug is yet to be fully verified.

A different method of humanizing mice for drug metabolism studies is the generation of mice with a chimeric humanized liver. The rationale of this approach is based on the fact that the liver is the primary organ of drug metabolism and, therefore, mice with a liver entirely or partially derived from human hepatocytes may be useful to predict the metabolism of a drug in humans. The feasibility of this technique was demonstrated in two recent publications. In both cases, human hepatocytes were transplanted into immunodeficient mouse strains with induced liver failure, resulting in highly engrafted mice in which the mouse liver could be replaced by up to 90% with human hepatocytes [30,31]. Furthermore, it could be shown that typical human DMEs, such as CYP3A4 and CYP1A1/2, were expressed in the liver of these mice, and that human-specific metabolites of probe substrates could be detected in the serum of the chimeric mice, thus demonstrating their utility for assessing drug interactions via enzyme induction and inhibition [32]. A clear advantage of these chimeric humanized liver models is that it is possible for all xenoreceptors and DMEs to be expressed by the human hepatocytes and at the single cell level, they potentially represent a complete humanization of the drug-metabolism pathway. However, a complete replacement of mouse liver with human hepatocytes has not yet been achieved, resulting in a mixed profile of mouse and human metabolism at the organ level. Humanization by this approach is currently restricted to the liver and, as the chimeric phenotype is not passed to the next generation, mice humanized by this approach must be continuously remade, resulting in a labor-intensive and expensive procedure.

Reporter mouse models

A promising method of defining the in vivo induction profile of DMEs by a compound is the use of reporter mouse models, as exemplified by Campbell et al., who used the CYP1A1 promoter to drive expression of lacZ and create a model system with which to determine the environmental and hormonal factors regulating expression of CYP1A1 [33,34]. Two further models have been made using the CYP3A4 promoter, one employing the firefly luciferase gene [35] and the other employing lacZ[36]. In the former, induction of reporter gene expression by common CYP3A4 inducers was assessed using a noninvasive in vivo imaging system. As many inducers of CYP3A4 are ligands for the xenoreceptors PXR and CAR, this reporter can be used to uncover the in vivo activation of these receptors by a compound. In order to increase the value of this system, a combination of the CYP3A4-luc reporter with humanizations of PXR and CAR, as described above, would be beneficial. Similar to the CYP3A4-luc reporter, a Cyp1a2 reporter has been generated, in which the firefly luciferase is linked to 8.4 kilobases of the mouse Cyp1a2 promoter [37]. Owing to the induction of Cyp1a2 in response to activation of the aryl hydrocarbon receptor, this mouse line could potentially be used to analyze the interaction of a compound with this receptor. However, as it is known that different mouse strains express different levels of the aryl hydrocarbon receptor protein, the induction of Cyp1a2 upon treatment with aryl hydrocarbon receptor activators can vary considerably among different strains.

The generation of suitable animal models for a better prediction of drug efficacy and toxicity in humans is an active field of research. Although the development of both transgenic and chimeric humanized mouse models for investigating drug metabolism studies is still at an early stage, their potential to better predict the pathways of drug disposition and toxicity in humans is already clear. The utility of such models will increase as they reach a number of key milestones:

• • Representation of the bona fide expression level and tissue distribution of the corresponding proteins in humans;

• • Deletion of the corresponding homologous mouse gene(s) in order to achieve a true humanized profile of drug metabolism;

• • Combination of multiple humanizations of xenoreceptors and DMEs in a single mouse model, covering a larger part of the human pathway of drug metabolism.

Chimeric mice with humanized livers are useful tools for drug-metabolism studies whenever it is acceptable that the humanization is restricted to the liver and that the pharmacokinetics is superposed with a mouse-specific drug-metabolism profile caused by residual mouse liver cells. Mouse reporter models can assist in the prediction of the in vivo induction profile of DMEs by a compound and to help identify the xenoreceptors mediating this effect. However, more work is required in order to define the sensitivity and specificity of these reporter systems, especially with regard to the question as to how far they truly reflect the situation in humans. The value of these models would significantly increase by a combination with humanized xenoreceptors and the incorporation of multiple reporters within a single mouse model.

Box 1. Drug-metabolizing enzyme and xenoreceptor gene knockouts in mice.

Phase I

• • Cyp1a1

• • Cyp1a2

• • Cyp1b1

• • Cyp(1a1 or 1a2)/1b1

• • Cyp1a1/1a2/1b1

• • Cyp2e1

• • Cyp1a2/2e1

• • Cyp2g1

• • Cyp3a family

• • POR: conditional

• • POR: complete

Phase II

• • Gsta4

• • Gsto1

• • Gstp1/p2

• • Gstt1

Xenoreceptor

• • CAR

• • PXR

• • PPARα

• • AhR

Financial & competing interests disclosure

Colin J Henderson is a consultant to CXR Biosciences. C Roland Wolf is a cofounder and Chief Scientific Officer of CXR Biosciences. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

Notes

CAR: Constitutive androstane receptor; PXR: Pregnane X receptor.