Safety Assessment of Drug Metabolites: Implications of Regulatory

Jan 26, 2009 - Safety Assessment of Drug Metabolites: Implications of Regulatory Guidance and Potential Application of Genetically Engineered Mouse Mo...
0 downloads 9 Views 69KB Size
Chem. Res. Toxicol. 2009, 22, 257–262

257

Safety Assessment of Drug Metabolites: Implications of Regulatory Guidance and Potential Application of Genetically Engineered Mouse Models that Express Human P450s Mark W. Powley,* Clay B. Frederick, Frank D. Sistare, and Joseph J. DeGeorge Department of Safety Assessment, Merck Research Laboratories, West Point, PennsylVania ReceiVed NoVember 17, 2008

Species differences in drug metabolism present two challenges that may confound the nonclinical safety assessment of candidate drugs. The first challenge is encountered when metabolites are formed uniquely or disproportionately in humans. Another challenge is understanding the human relevance of toxicities associated with metabolites formed uniquely or disproportionately in a nonclinical species. One potential approach to minimize the impact of metabolite related challenges is to consider genetically engineered mouse models that express human P450 enzymes. Human P450 expressing mouse models may have the ability to generate major human metabolites and eliminate or reduce the formation of mouse specific metabolites. Prior to determining the utility of any particular model, it is important to qualify by characterizing protein expression, establishing whether the model generates an in vivo metabolite profile more closely related to that of humans than the wild-type mouse, verifying genetic stability, and evaluating animal health. When compared to the current strategy for handling metabolite challenges (i.e., direct administration of metabolite), identifying an appropriate human P450 expressing model could provide a number of benefits. Such benefits include improved scientific relevance of the evaluation, decreased resource needs, and a possible reduction in the number of animals used. These benefits may ultimately improve the quality and speed by which promising new drug candidates are developed and delivered to patients. Contents 1. Background 1.1. Historical Perspective 1.2. Implications for Nonclinical Safety Assessment 2. Examples 2.1. Major Human Metabolites 2.2. Metabolites Associated with Species Specific Toxicity 3. Potential Application of Genetically Engineered Mouse Models 3.1. Proof of Concept 3.2. Qualification Strategy 4. Summary

257 257 258 259 259 259 260 260 261 261

1. Background Nonclinical safety evaluations are based on the premise that animal models will be useful for identifying potential human toxicities associated with candidate drugs. The complexity of such evaluations is compounded by well-known species differences in drug metabolizing enzymes and drug transporters. These differences can involve substrate specificity, enzyme induction, and tissue distribution leading to two major challenges in the drug development process: (1) assessing the safety of metabolites formed uniquely or disproportionately in humans, and (2) determining whether toxicities observed in nonclinical species are due to metabolites formed uniquely or disproportionately in animals and may, therefore, be irrelevant to humans. To minimize these uncertainties, it would be desirable to have * To whom correspondence should be addressed. E-mail: mark_powley@ merck.com.

access to animal models capable of generating a metabolite and/ or induction profile more similar to that of the human than the standard nonclinical test species. The question we are currently seeking to answer is whether genetically engineered mouse models (GEMMs1) that express human P450 enzymes will be useful in addressing challenges resulting from species differences in metabolism. These models may have the ability to generate important circulating metabolites identified in humans as well as eliminate or reduce the formation of mouse specific metabolites leading to toxicities of unknown relevance to humans. This manuscript details limitations of the current safety assessment paradigm for dealing with major human metabolites as well as potential applications of GEMMs in evaluating drugs with the metabolite challenges described above. While species differences in phase II enzymes and drug transporters are known to exist and could affect the concentrations of circulating metabolites, the current discussion focuses on P450s principally responsible for forming the metabolites of concern. 1.1. Historical Perspective. The original Metabolites In Safety Testing (MIST) position paper represents an industry wide effort to respond to regulatory concerns associated with potential metabolite toxicity (1). Since that time, a number of papers dealing with various aspects of drug metabolite safety have been published (2-9). Particularly interesting is an analysis of drugs withdrawn from the market due to safety concerns (8). Of the list detailing 24 withdrawn molecules, 14 had toxicities possibly involving metabolites. The authors indicated that a majority of these metabolites reached exposures in humans that 1 Abbreviations: BAC, bacterial artificial chromosome; CROs, contract research organizations; GEMMs, genetically engineered mouse models; NOEL, no observable effect level; PhIP, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine.

10.1021/tx8004373 CCC: $40.75  2009 American Chemical Society Published on Web 01/26/2009

258

Chem. Res. Toxicol., Vol. 22, No. 2, 2009

Powley et al.

Table 1. Summary of Strategies for Evaluating the Safety of Major Human Metabolites approach to evaluating metabolite toxicity

pros

cons

Use a standard nonclinical safety assessment model capable of generating adequate plasma concentrations of major human metabolite(s) following administration of the parent molecule

form metabolite(s) in vivo (i.e., generate metabolite profile closely related to human) satisfy regulatory requirements outlined in guidance without conducting additional studies in certain cases, there will be no delays or additional costs

may require testing in multiple species or species that are not optimal for development (e.g., selecting primate as nonrodent species solely for metabolite coverage)

Use a standard nonclinical safety assessment model capable of generating adequate plasma concentrations following direct administration of the metabolite(s)

robust historical database for traditional toxicology end points is available satisfy regulatory requirements outlined in guidance

data of questionable relevance significant number of animals may be needed

costs associated with: -synthesizing and characterizing kg amounts of metabolite -optimizing formulation -verifying metabolite TK -conducting studies development delays Use GEMMs capable of generating adequate plasma concentrations following administration of the parent molecule

form metabolite in vivo (i.e., generate metabolite profile closely related to human) improved scientific relevance compared to direct administration of metabolite fewer animals are needed compared to direct administration of the metabolite; this will be especially evident if multiple major human metabolites are identified minimize costs and delays compared to those associated with direct administration of metabolite satisfy regulatory requirements outlined in guidance

were adequately covered in nonclinical species, suggesting that further nonclinical safety evaluation would not have predicted the clinical toxicity observed. 1.2. Implications for Nonclinical Safety Assessment. The recent FDA guidance describing safety testing of drug metabolites (10) has substantial implications for nonclinical safety assessment. When a major human metabolite, described as one that accounts for >10% of parent exposure, is identified during the development process, there are three general strategies for assessing safety. In addition to being described below, the strategies are also summarized in Table 1. (1) In many cases, at least one nonclinical safety species is capable of generating sufficient plasma concentrations of major circulating human metabolites following administration of the parent compound. When this is true, further nonclinical safety evaluation of the metabolite is ordinarily not required (10). Using a nonclinical model that forms major human metabolites in vivo and satisfies regulatory safety testing requirements without additional studies is desirable. However, in some cases it may require testing in multiple species or those that are not optimal for development (e.g., selecting a primate as nonrodent species solely for metabolite coverage). (2) For some molecules, there may not be an animal model capable of generating adequate concentrations of a major circulating human metabolite. Under these circumstances, the FDA metabolite guidance suggests that the major human

qualification of model needed. lack of historical background data for traditional toxicology end points costs involved in developing or obtaining model screening costs

metabolite(s) be synthesized and tested in appropriate nonclinical safety assessment studies. Results of these studies must be submitted prior to phase 3 trials. Encountering major human metabolites not covered by a nonclinical model can significantly hinder drug development, is resource intensive, and essentially requires a separate development program. A number of liabilities are inherent to this strategy. Prueksaritanont et al. (11) as well as Smith and Obach (8) describe potential complications resulting from pharmacokinetic differences (e.g., absorption and clearance) between direct administration of synthesized metabolite and metabolite formed in vivo following the administration of the parent compound. Because of these well documented differences, data collected following the direct administration of metabolite to a nonclinical species may not accurately represent toxicity resulting from a systemically generated metabolite. In addition to concerns about the data generated, there are costs and delays associated with this strategy that are not encountered when the metabolite is formed in vivo. Among these are the costs of synthesizing/characterizing the metabolite, optimizing a formulation for administration, conducting toxicokinetic studies to verify that adequate plasma concentrations can be achieved, and performing nonclinical safety studies described in the FDA guidance. A significant commitment of both time and resources is likely to be associated with each of these steps. Most notable are the resources required for synthesis

PerspectiVe

Chem. Res. Toxicol., Vol. 22, No. 2, 2009 259

Table 2. Costs Associated with Running GLP Studies at CRO required evaluations

Table 3. Steady State Exposures Following Oral Administration of Candidate Drug Moleculea

costa

Ames assay chromosomal aberrations assay 1 month toxicity study in rat 6 month toxicity study in rat embryofetal toxicity study in rat carcinogenicity study in ratb carcinogenicity study in transgenic mouseb total of safety studies

$6000 $20,000 $180,000 $420,000 $100,000 $480,000 $200,000 $926,000-$1,206,000c

synthesis of metabolite (kg quantities)

widely variable, possibly >$100,000

a

Approximate average of costs provided by multiple CROs. b Anticipated cost of adding 1 metabolite dose level to the carcinogenicity evaluation with the parent molecule. c Range dependent on carcinogenicity evaluation in transgenic mouse or rat.

and characterization of large (e.g., kg) quantities of metabolite and conducting additional nonclinical safety studies. It is possible that synthesis will entail nothing more than simple modification of the parent compound, thereby requiring minimal resources. However, this will not always be true especially if dealing with molecules with complex stereochemistry. The time needed to complete all of these activities can be significant and may ultimately delay NDA submission. The battery of nonclinical safety studies needed to qualify a metabolite will vary considerably. In order to demonstrate the impact on costs, consider a theoretical drug candidate intended for chronic administration to treat a nonlife threatening indication in a patient population that includes women of child bearing potential. The additional cost to a development program designed to meet minimal criteria for evaluating a single major human metabolite is provided in Table 2. The total cost of conducting the suggested safety studies, determined using average estimates obtained from multiple contract research organizations (CROs), is >$900,000. Note that there may be additional costs required to conduct supporting studies (e.g., range finding, toxicokinetic studies, etc.). It is also important to point out the increase in the number of animals used for this option. (3) Another potential option is to use GEMMs that express human P450 enzymes for conducting safety evaluations of both parent and metabolite simultaneously following parent compound administration. This could either be for all nonclinical safety studies conducted in the rodent or for a limited number of bridging studies. One obvious prerequisite for applying this strategy to an evaluation of both metabolite and parent is that all human metabolites of interest be identified early in development. Another obvious requirement is that an appropriate GEMM is identified. Identification and qualification of appropriate GEMMs that express human P450 enzymes is a major focus of this manuscript and is subsequently discussed. Major advantages of this option are avoiding time delays inherent in conducting studies with metabolite as well as providing a more scientifically justified assessment of metabolite safety. However, the cost of selecting an appropriate P450 expressing model must be considered. At a minimum, a shortterm toxicokinetic study could be used to determine whether the model can produce the metabolite of interest. In addition, the initial costs of developing and qualifying GEMMs can be significant and should not be overlooked.

2. Examples As described above, species differences in drug metabolism result in two distinct types of metabolite challenges that can be

parent

metabolite

AUC0-24 h (µM·h)

AUC0-24 h (µM·h)

0.665

8.64

humanb week 4 ratc week 13 week 26

11 (female) 14 (female)

7.43 (male) 7.67 (male)

0 (female) 0.018 (female)

0.327 (male) 0.290 (male)

dogc week 5 week 13

2.86 2.82

0 0

mousec week 4

4.18

0.312

a

Data are presented as the mean for sexes-combined unless otherwise noted. b At anticipated clinical dose. c At NOEL for toxicologically significant changes.

encountered during nonclinical safety assessment. The first challenge is presented by metabolites that are uniquely or disproportionately formed in humans. The likelihood that a truly unique human metabolite is identified is less than the probability of identifying a metabolite formed at disproportionately higher levels in human versus nonclinical species. While the incidence of either scenario has been minimal under current practice, the impact of identifying a major human metabolite can be substantial. Furthermore, the probability of encountering a major human metabolite in the future has increased because of the new need to screen candidate drugs for human metabolites at an earlier stage when a greater number of molecules are still in development. The second type of challenge is determining whether a toxicity accompanying the formation of a metabolite uniquely or disproportionately in a nonclinical species is relevant to humans. There currently is little understanding of how often this occurs; however, this may result in discontinuing the development of a compound because of a nonmonitorable and/ or severe toxicity that may actually be of questionable relevance in humans. 2.1. Major Human Metabolites. The following is an example of a metabolite formed at disproportionately high levels in humans when compared in nonclinical safety species (Table 3). In this case, a P450-mediated metabolite was present in humans at exposures approximately 13-fold greater than the parent molecule. Each nonclinical animal model evaluated provided sufficient coverage of the parent molecule at the no observable effect level (NOEL) for toxicologically significant changes. However, none of these models generated sufficient plasma concentrations of the major human metabolite in question. In addition to the species shown in Table 3, metabolite formation was assessed in other nonclinical models, including multiple strains of rodents. Because adequate plasma concentrations could not be achieved following oral administration of the parent compound, a separate development program was initiated to evaluate the metabolite. 2.2. Metabolites Associated with Species Specific Toxicity. Table 4 presents data from an example where a metabolite was formed disproportionately in a single animal model when compared to other animal models and, most importantly, the human. Following administration of the parent molecule to nonclinical animal models, a large difference in plasma concentrations of a P450-mediated metabolite was observed between mice, rats, and dogs. Accompanying these high plasma concentrations in mice was a tissue specific lesion not observed

260

Chem. Res. Toxicol., Vol. 22, No. 2, 2009

Powley et al.

Table 4. Steady State Exposures Following Oral Administration of Candidate Drug Moleculea parent species/dose

study week AUC0-24 h (µM·h)

human maximum doseb mouse dose level 1 dose level 2c dose level 3c dose level 4c

week 5

rat dose level 1 dose level 2 dose level 3

week 13

dog dose level 1

week 13

metabolite AUC0-24 h (µM·h)

45

177

51.3 111 141 187

523 2050 4130 5040

55.5 120 298

74.6 281 484

33.5

dose level 2

203

dose level 3

476

74.7 (female) 328 (female) 1100 (female)

22.7 (male) 203 (male) 239 (male)

a Data are presented as the mean for sexes-combined unless otherwise noted. b Estimated. c Doses at which species specific toxicity was observed following 14 weeks of dosing.

in the other animal models. The lowest metabolite systemic exposure where toxicity was observed in mice was at least 2-fold higher than that of the other species evaluated. Because of this discrepancy, further investigative studies were conducted, in part, to understand whether the parent or metabolite was responsible for the toxicity in mice.

3. Potential Application of Genetically Engineered Mouse Models The two previous examples highlight the need for animal models that more closely mimic human drug metabolism. One potential solution to these metabolite challenges are GEMMs that express human P450 enzymes. It is important to note distinctions between models that express human P450 enzymes on a wild-type mouse background (i.e., in most cases transgenic models) and those that express human P450 enzymes in the absence of the mouse P450 orthologue(s) (i.e., humanized models). Techniques used to incorporate the human gene of interest into the mouse genome range from random insertion of a transgene by pronuclear injection through precise placement of the transgene utilizing site-specific recombinases (e.g., CreLox, etc.) or homologous recombination with human genes, or cDNAs flanked by mouse sequences. These models hold promise for several reasons, the most significant being the ability to generate a metabolite profile more closely associated with humans than wild-type mice. In doing so, the models may potentially form human P450-mediated metabolites and eliminate or reduce the production of P450mediated mouse metabolites. In addition, GEMMs expressing metabolic sensors such as PXR or CAR have been developed (12-14). If similar models are crossed with GEMMs expressing human P450 enzymes, the result would be a mouse humanized for not only the P450 enzyme of interest but also the gene responsible for controlling the expression of the enzyme. An example of a model expressing human CYP3A4 and human PXR was recently described by Ma et al. (15). It is estimated that CYP3A4 is involved in the metabolism of ∼50% of marketed drugs, while CYP2D6 and CYP2C

enzymes (i.e., 2C9 and 2C19) are involved in ∼30% and ∼10%, respectively (16). Because of their importance in drug metabolism, models expressing these enzymes would be of greater interest. Also, there are a number of clinically important P450 variants in existence (17) that may be of interest as well. There are several benefits that may result from identifying an appropriate humanized model for evaluating metabolite safety. First and most important, the data generated from a metabolite formed in vivo is likely to have greater scientific relevance than data generated following the direct administration of a synthesized metabolite. Second, there would be a decrease in the time delays and costs associated with initiating an independent development program to evaluate major human metabolite safety or determining the relevance of species specific metabolites. Another important benefit will be the possible reduction in animal usage, especially in cases where multiple major human metabolites are identified. All of these benefits ultimately serve to improve the quality and speed by which promising new therapeutics can be delivered to patients with serious medical needs. 3.1. Proof of Concept. There are a number of human P450 expressing GEMMs described in the literature. These include models expressing CYP1A1/1A2 (18-21), CYP2D6 (22), CYP2E1 (23), and CYP3A4 (24-27) as well as a mouse model expressing both CYP3A4 and CYP2D6 (28). Excellent detailed overviews of this class of models have been published in recent reviews (29-32). The information below is not intended to provide another review but to demonstrate proof of concept by describing metabolism and pharmacokinetic data for the relevant models. Both the functionality of human protein as well as the ability to generate a qualitative/quantitative in vivo metabolite profile more closely associated with the human than the wildtype mouse are important in this respect. Two related GEMMs have been developed that express CYP1A1/CYP1A2 in the absence of mouse Cyp1a2. Cheung et al. (18) describe a humanized model made by crossing a Cyp1a2 knockout mouse with a CYP1A1/1A2 expressing strain. The P450 expressing model was generated by pronuclear injection of bacterial artificial chromosome (BAC) containing the genes for both human P450 enzymes. In evaluating the resulting humanized model, the authors examined urinary concentrations of 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) metabolites. Urinary concentrations of potentially mutagenic metabolites formed in the human metabolic pathway (i.e., N2-hydroxyPhIP and related glucuronides) were significantly increased in the humanized model compared to control mice. Conversely, urinary concentrations of a metabolite associated with PhIP detoxification in rodents (i.e., free and deconjugated 4-hydroxyPhIP) were lower in humanized mice. Derkenne et al. (19) evaluated a related model by monitoring the metabolism of theophyline to 3-methylxanthine, a prominent human metabolite. The humanized model formed substantial amounts of 3-methylxanthine in vitro, while levels of this metabolite in wild-type mice were below the limits of detection. A transgenic model expressing CYP2D6, another important human P450, was developed through pronuclear injection of the entire CYP2D6 gene (22). Systemic exposure (AUC0-24 h) to 4-hydroxydebrisoquinine, a prominent human debrisoquinone metabolite, was ∼9-fold higher in the transgenic model compared to wild-type controls. Several CYP3A4 models have been described in the literature. Using pronuclear injection of a CYP3A4 containing BAC, Granvil et al. (24) generated a transgenic model that expressed CYP3A4 in the small intestine. Following oral administration

PerspectiVe

of midazolam, a model CYP3A4 substrate, plasma concentrations of 1′-hydroxymidazolam were monitored. Systemic exposure (AUC0-3 h) to 1′-hydroxymidazolam was ∼3-fold higher for the transgenic model when compared to the wild-type mouse. Liver-specific expression of CYP3A has also been achieved in both transgenic and humanized mice (25, 26). In the first example, a transgenic model was generated through pronuclear injection of an Apo-E promoter driven expression cassette containing CYP3A4 cDNA (25). By monitoring in vivo midazolam metabolism, the authors determined that 1′-hydroxymidazolam exposure (AUC0-3 h) was ∼1.5-fold greater in the transgenic model versus the wild-type mouse. This same laboratory generated a Cyp3a cluster knockout model, which was subsequently crossed with the transgenic described above as well as a small intestine specific CYP3A4 model created by pronuclear injection of a villin promoter driven expression cassette containing CYP3A4 cDNA (26). The authors provide data showing that docetaxol metabolites were just above the detectable levels in the liver of the cluster knockout model, while the wild-type mouse and the humanized CYP3A4 liver specific model had comparable levels of metabolite in liver. Similarly, the humanized CYP3A4 small intestine specific model had tissue concentrations of metabolites at higher levels than the knockout mouse. While the data does not provide evidence that the model generates a human-like metabolite profile, the data does show that the CYP3A4 protein is functional. The data described above demonstrate the promising nature of these models; however, there are additional factors that must be considered when determining the utility of a particular GEMM. 3.2. Qualification Strategy. Prior to regarding a humanized P450 mouse as fit for use in nonclinical safety testing of metabolites, the model must be qualified. At a minimum, this process will involve characterizing protein expression, establishing whether the model generates an in vivo metabolite profile more closely related to humans than the wild-type mouse, verifying genetic stability, and determining the impact of the genetic modification on animal health. Key aspects of protein characterization include evaluating the expression of mouse P450s as well as understanding both where and to what extent the human P450 protein is expressed. As mentioned previously, a truly humanized model will express the human P450 enzyme in the absence of equivalent mouse orthologues. Regardless of what technique is used to accomplish this humanization (e.g., breeding a knockout mouse with a transgenic mouse expressing the human P450 protein or by knocking-in the human gene), it is important to demonstrate the absence of the mouse P450 protein and activity. In addition, because there may be a compensatory response to the genetic modification, looking for changes in the expression pattern of other P450 enzymes is also prudent. Such a compensatory response was demonstrated by van Waterschoot et al. (33), who discovered that Cyp2c enzymes, most significantly Cyp2c55, were upregulated in Cyp3a knockout mice. The upregulation of Cyp2c resulted in significant midazolam metabolism. Ideally, human P450s will be expressed with levels and tissue distribution mimicking that observed in humans. Although current models do not possess optimal expression patterns and, in some cases, are not truly humanized (i.e., still express mouse P450s), they may be sufficient for metabolite safety assessment. However, it is crucial to uncover and define their capabilities and limitations. With experience, more refined GEMMs may be developed.

Chem. Res. Toxicol., Vol. 22, No. 2, 2009 261

The most crucial piece of data generated during qualification is determining whether or not the model can generate a humanlike metabolite profile. It should be noted that selection of appropriate substrates to demonstrate this is paramount. It is also important to note that the amount of available data for mice is limited, highlighting the need to characterize metabolism catalyzed by mouse P450 enzymes. Ideally, the substrate will have a unique human metabolite formed through the action of a single P450 enzyme. Comparing the in vitro metabolite profiles of these substrates generated by GEMMs, wild-type mouse, and humans provides a resource sparing means of determining the probability of success for a particular model. Analytical techniques such as accurate mass LC-MS/MS methodology that employ data mining tools yield qualitative metabolite data (34) that can be used to compare species. These in vitro data will be helpful in deciding whether or not to proceed to more comprehensive and resource intensive in vivo studies. While qualitative information from in vivo studies will again be useful, a more precise quantitative toxicokinetic evaluation will provide the most informative data. The ultimate answer will come from a comparison of systemic exposure, preferably at steady state, in GEMMs and wild-type mice against clinical data for probe substrates. A model will have obvious utility if it is capable of providing coverage of a major human metabolite not covered by wild-type mice. After characterizing protein expression and establishing that a model can generate a more human metabolite profile, the next step will be determining what impact the genetic modification has on the overall animal health. In order to understand any limitations in evaluating a metabolite, it is important to understand effects on life-span, ante-mortem end points (e.g., body weights, behavior, serum biochemistry, hematology, and urine analysis), and postmortem end points (e.g., organ weights, gross observations, and histopathological alterations) as well as developmental and reproductive toxicology end points (e.g., fertility, maternal toxicity, embryo-fetal development, and preand postnatal development). By evaluating these parameters for appropriate durations, a control database will be generated to increase confidence in one’s ability to detect changes related to administration of a candidate drug. The use of mouse as a routine nonclinical safety model presents a number of challenges in itself since the rat is currently the preferred rodent species; however, the generation of sufficient control data should establish the utility of this alternate nonclinical safety assessment model. Furthermore, from a regulatory perspective, the mouse is an acceptable species for conducting the necessary safety evaluations (35-38).

4. Summary The current nonclinical safety assessment paradigm for handling unique or disproportionate metabolites is to directly administer the metabolite to an animal model. Conducting these studies could result in data of questionable utility and will require significant time and resources. To more appropriately comply with recent FDA guidance addressing the safety testing of drug metabolites, animal models capable of generating a more human-like metabolite and/or induction profile must be identified. One possibility is GEMMs that express human P450 enzymes, preferably in the absence of the mouse orthologue. Examples where this class of models were capable of generating human metabolites and/or reducing the formation of mouse metabolites have been detailed in the literature and summarized above. While further evaluation (e.g., characterizing protein expression, metabolism, genetic stability, and animal health)

262

Chem. Res. Toxicol., Vol. 22, No. 2, 2009

must be completed before declaring these models appropriate for regulatory decision making, they hold potential in improving the quality and speed of nonclinical drug development. Acknowledgment. We acknowledge Dr. Tom Rushmore for his input as well as Dr. Raymond Evers for his input and critical reading of the manuscript.

References (1) Baillie, T. A., Cayen, M. N., Fouda, H., Gersons, R. J., Green, J. D., Grossman, S. J., Klunk, L. J., LeBlanc, B., Perkins, D. G., and Shipley, L. A. (2002) Drug metabolites in safety testing. Toxicol. Appl. Pharmacol. 182, 188–196. (2) Davis-Bruno, K. L., and Atrakchi, A. (2006) A regulatory perspective on issues and approaches in characterizing human metabolites. Chem. Res. Toxicol. 19, 1561–1563. (3) Guengerich, F. P. (2006) Safety assessment of stable drug metabolites. Chem. Res. Toxicol. 19, 1559–1560. (4) Hastings, K. L., El-Hage, J., Jacobs, A., Leighton, J., Morse, D., and Ostertag, R. E. (2003) Letter to the Editor. Toxicol. Appl. Pharmacol. 190, 91–92. (5) Humphreys, W. G., and Under, S. E. (2006) Safety assessment of drug metabolites: characterization of chemically stable metabolites. Chem. Res. Toxicol. 19, 1564–1569. (6) Luffer-Atlas, D. (2008) Unique/major human metabolites: why, how, and when to test for safety in animals. Drug Metab. Dispos. 40, 447– 463. (7) Smith, D. A., and Obach, R. S. (2005) Seeing through the MIST: abundance versus percentage. Commentary on metabolites in safety testing. Drug Metab. Dispos. 33, 1409–1417. (8) Smith, D. A., and Obach, R. S. (2006) Metabolites and safety: What are the concerns, and how should we address them? Chem. Res. Toxicol. 19, 1570–1579. (9) Smith, D. A., Obach, R. S., Wiliams, D. P., and Park, B. K. (2008) Clearing the MIST (metabolites in safety testing) of time: The impact of duration of administration on drug metabolite toxicity. Chem.-Biol. Interact. [Online early access]. Published Online, September 26, 2008. (10) FDA (2008) Guidance for Industry: Safety Testing of Drug Metabolites. (11) Preuksaritanont, T., Jiunn, H. L., and Baillie, T. A. (2006) Complicating factors in safety testing of drug metabolites: Kinetic differences between generated and preformed metabolites. Toxicol. Appl. Pharmacol. 217, 143–152. (12) Ma, X., Shah, Y., Cheung, C., Guo, G. L., Feigenbaum, L., Krausz, K. W., Idle, J. R., and Gonzalez, F. J. (2007) The pregnane X receptor gene-humanized mouse: a model for investigating drug-drug interactions mediated by cytochromes P450 3A. Drug Metab. Dispos. 35, 194–200. (13) Scheer, N., Ross, J., Rode, A., Zevnik, B., Niehaves, S., Faust, N., and Wolf, C. R. (2008) A novel panel of mouse models to evaluate the role of human pregnane X receptor and constitutive androstane receptor in drug response. J. Clin. InVest. 118, 3228–3229. (14) Xie, W., Barwick, J. L., Downes, M., Blumberg, B., Simon, C. M., Nelson, M. C., Neuschwander-Tetri, B. A., Brunt, E. M., Guzelian, P. S., and Evans, R. M. (2000) Humanized xenobiotic response in mice expressing nuclear receptor SXR. Nature 406, 435–439. (15) Ma, X., Cheung, C., Krausz, K. W., Shah, Y. M., Wang, T., Idle, J. R., and Gonzalez, F. J. (2008) A double transgenic mouse model expressing human pregnane X receptor and cytochrome P450 3A4. Drug Metab. Dispos. [Online early access]. Published Online, September 17, 2008. (16) Zuber, R., Anzenbacherova, E., and Anzenbacher, P. (2002) Cytochromes P450 and experimental models of drug metabolism. J. Cell. Mol. Med. 6, 189–198. (17) Zhou, S.-F., Di, Y. M., Chan, E., Du, Y.-M., Chow, V.D.-W., Xue, C. C., Lai, X., Wang, J.-C., Li, C. G., Tian, M., and Duan, W. (2008) Clinical pharmacogenetics and potential application in personalized medicine. Curr. Drug Metab. 9, 738–784. (18) Cheung, C., Ma, X., Krausz, K. W., Kimura, S., Feigenbaum, L., Dalton, T. P., Nebert, D. W., Idle, J. R., and Gonzalez, F. J. (2005) Differential metabolism of 2-amino-1-methyl-6-phenylimidazo[4,5b]pyridine (PhIP) in mice humanized for CYP1A1 and CYP1A2. Chem. Res. Toxicol. 18, 1471–1478. (19) Derkenne, S. D., Curran, C. P., Shertzer, H. G., Dalton, T. P., Dragin, N., and Nebert, D. W. (2005) Theophylline pharmacokinetics: comparison of Cyp1a1(-/-) and Cyp1a2(-/-) knockout mice, humanized hCYP1A1_1A2 knock-in mice lacking wither the mouse

Powley et al.

(20)

(21) (22)

(23)

(24)

(25)

(26)

(27)

(28)

(29) (30) (31) (32)

(33)

(34)

(35) (36) (37) (38)

Cyp1a1 or Cyp1a2 gene, and Cyp1(+/+) wild-type mice. Pharmacogenet. Genomics 15, 503–511. Jiang, Z. W., Dalton, T. P., Jin, L., Wang, B., Tsuneoka, Y., Shertzer, H. G., Deka, R., and Nebert, D. W. (2005) Toward the evaluation of function in genetic variability: characterizing human SNPs and establishing BAC-transgenic mice carrying the CYP1A1_CYP1A2 locus. Hum. Mutat. 25, 196–206. Dragin, N., Uno, S., Wang, B., Dalton, T. P., and Nebert, D. W. (2007) Generation of ‘humanized’ hCYP1A1_1A2_Cyp1a1/1a2(-/-) mouse line. Biochem. Biophys. Res. Commun. 359, 635–642. Corchero, J., Granvil, C. P., Akiyama, T. E., Hayhurst, G. P., Pimprale, S., Feigenbaum, L., Idle, J. R., and Gonzalez, F. J. (2001) The CYP2D6 humanized mouse: effect of the human CYP2D6 transgene and HNFR on the disposition of debrisoquine in the mouse. Mol. Pharmacol. 60, 1260–1267. Cheung, C., Yu, A.-M., Ward, J. M., Krausz, K. W., Akiyama, T. E., Feigenbaum, L., and Gonzalez, F. J. (2005) The CYP2E1-humanized transgenic mouse: role of CYP2E1 in acetaminophen hepatotoxicity. Drug Metab. Dispos. 33, 449–457. Granvil, C. P., Yu, A.-M., Elizondo, G., Akiyama, T. E., Cheung, C., Feigenbaum, L., Krausz, K. W., and Gonzalez, F. J. (2003) Expression of the human CYP3A4 gene in the small intestine of transgenic mice: in vitro metabolism and pharmacokinetics of midazolam. Drug Metab. Dispos. 31, 548–558. van Herwaarden, A. E., Smit, J. W., Sparidans, R. W., Wagenaar, E., van der Kruijssen, C. M. M., Schellens, J. H. M., Beunen, J. H., and Schinkel, A. H. (2005) Midazolam and cyclosporin A metabolism in transgenic mice with liver-specific expression of human CYP3A4. Drug Metab. Dispos. 33, 892–895. van Herwaarden, A. E., Wagenaar, E., van der Kruijssen, C. M. M., van Waterschoot, R. A. B., Smit, J. W., Song, J.-Y., van der Valk, M. A., van Tellingen, O., van der Hoorn, J. W. A., Rosing, H., Beijnen, J. H., and Schinkel, A. H. (2007) Knockout of cytochrome P450 3A yields new mouse models for understanding xenobiotic metabolism. J. Clin. InVest. 117, 3583–3592. Cheung, C., Yu, A.-M., Chen, C.-S., Krausz, K. W., Byrd, L. G., Feigenbaum, L., Edwards, R. J., Waxman, D. J., and Gonzalez, F. J. (2006) Growth hormone determines sexual dimorphism of hepatic cytochrome P450 3A4 expression in transgenic mice. J. Pharmacol. Exp. Ther. 316, 1328–1334. Felmlee, M. A., Lon, H.-K., Gonzalez, F. J., and Yu, A.-M. (2008) Cytochrome P450 expression and regulation in CYP3A4/CYP2D6 double transgenic humanized mice. Drug Metab. Dispos. 36, 435– 441. Gonzalez, F. J., and Yu, A.-M. (2006) Cytochrome P450 and xenobiotic receptor humanized mice. Annu. ReV. Pharmacol. Toxicol. 46, 41–64. Lin, J. H. (2008) Applications and limitations of genetically modified mouse models in drug discovery and development. Current Drug Metab. 9, 419–438. Muruganandan, S., and Sinal, C. J. (2008) Mice as clinically relevant models for the study of cytochrome P450-dependent metabolism. Clin. Pharm. Ther. 83, 818–827. Cheung, C., and Gonzalez, F. J. (2008) Humanized mouse lines and their application for prediction of human drug metabolism and toxicological risk assessment. J. Pharmacol. Exp. Ther. 327, 288– 299. van Waterschoot, R. A. B., van Herwaarden, A. E., lagas, J. S., Sparidans, R. W., Wagenaar, E., van der Kruijssen, C. M. M., Goldstein, J. A., Zeldin, D. C., Beijnen, J. H., and Schinkel, A. H. (2008) Midazolam metabolism in cytochrome P450 3A knockout mice can be attributed to up-regulated CYP2C enzymes. Mol. Pharmacol. 73, 1029–1036. Tiller, P. R., Yu, S., Castro-Perez, J., Fillgrove, K. L., and Baillie, T. A. (2008) High through-put, accurate mass liquid chromotagraphy/ tandem mass spectrometry on a quadrupole time-of-flight system as a “first-line” approach for metabolite identification studies. Rapid Commun. Mass Spectrom. 22, 1053–1061. ICH (1993) ICH harmonised tripartite guideline: Detection of toxicity to reproduction for medicinal products and toxicity to male fertility S5(R2). ICH (1997) ICH Harmonised Tripartite Guideline: Testing for carcinogenicity of pharmaceuticals - S1B. ICH (1998) ICH Harmonised Tripartite Guideline: Duration of chronic toxicity testing in animals (rodent and non rodent toxicity testing) S4. ICH (2000) ICH Harmonised Tripartitite Guideline: Maintenance of the ICH guideline on non clinical safety studies for the conduct of human clinical trials for pharmaceuticals - M3(R1).

TX8004373