Chem. Res. Toxicol. 2009, 22, 267–279
267
ReViews Metabolites in Safety Testing (MIST): Considerations of Mechanisms of Toxicity with Dose, Abundance, and Duration of Treatment Dennis A. Smith*,† and R. Scott Obach‡ Pharmacokinetics, Dynamics, and Metabolism, Pfizer Inc., Sandwich, Kent, U.K., and Pharmacokinetics, Dynamics, and Metabolism, Pfizer Inc., Groton, Connecticut ReceiVed NoVember 4, 2008
In previous papers, we have offered a strategic framework regarding metabolites of drugs in humans and the need to assess these in laboratory animal species (also termed Metabolites in Safety Testing or MIST; Smith and Obach, Chem. Res. Toxicol. (2006) 19, 1570-1579). Three main tenets of this framework were founded in (i) comparisons of absolute exposures (as circulating concentrations or total body burden), (ii) the nature of the toxicity mechanism (i.e., reversible interaction at specific targets versus covalent binding to multiple macromolecules), and (iii) the biological matrix in which the metabolite was observed (circulatory vs excretory). In the present review, this framework is expanded to include a fourth tenet: considerations for the duration of exposure. Basic concepts of pharmacology are utilized to rationalize the relationship between exposure (to parent drug or metabolite) and various effects ranging from desired therapeutic effects through to severe toxicities. Practical considerations of human ADME (absorption-distribution-metabolism-excretion) data, to determine which metabolites should be further evaluated for safety, are discussed. An analysis of recently published human ADME studies shows that the number of drug metabolites considered to be important for MIST can be excessively high if a simple percentageof-parent-drug criterion is used without consideration of the aforementioned four tenets. Concern over unique human metabolites has diminished over the years as experience has shown that metabolites of drugs in humans will almost always be observed in laboratory animals, although the proportions may vary. Even if a metabolite represents a high proportion of the dose in humans and a low proportion in animals, absolute abundances in animals frequently exceed that in humans because the doses used in animal toxicology studies are much greater than therapeutic doses in humans. The review also updates the enzymatic basis for the differences between species and how these relate to MIST considerations. Contents Introduction Dose and Plasma Concentration Response Metabolites in Types A1 and A2 Toxicity Metabolites in Types B and C Toxicity Implications for MIST Strategy Practical Considerations in Identifying and Characterizing Human Circulating Metabolites 7. Enzymatic Basis for Interspecies Differences in Circulating Metabolite Profiles 7.1. Cytochrome P450 Enzymes 7.2. UGT1A4 and Tertiary N-Glucuronidation Reactions 7.3. Aldehyde Oxidase 8. Summary and Conclusions 1. 2. 3. 4. 5. 6.
267 269 271 271 272 273 275 275 276 276 277
1. Introduction In developing a strategy for metabolites in safety testing (MIST1), the types of mechanisms of toxicity that could be * To whom correspondence
[email protected]. † Pfizer Inc., Sandwich, Kent.
should
be
addressed.
E-mail:
caused by metabolites and the chemical structure of the metabolite, relative to the parent drug, have been considered. For the purposes of MIST, we have found that the categorization of toxicity into four overall types (A, B, C, and D) as described by Park (1) is extremely helpful (2-4). Some simpler versions categorizing toxicology into on- and off-target probably represent the future when toxicity is fully understood in terms of drug-protein or other macromolecule interactions. We would like to extend the categorization used earlier, hoping that advances to be made in our understanding of the precise mechanisms underlying toxicity (e.g., identification of specific proteins involved) will permit categories B, C, and D to be rendered obsolete. In type A, the toxic mechanism is a normally reversible interaction with a defined protein or macromolecule (henceforth termed the receptor) leading to a predictable pharmacodynamic outcome. Previously, we termed this as a pharmacological based ‡
Pfizer Inc., Groton, Connecticut. Abbreviations: ADME, absorption, distribution, metabolism, and excretion; ALT, alanine aminotransferase; DILI, drug-induced liver injury; HPLC, high pressure liquid chromatography; MIST, metabolites in safety testing; NAPQI, N-acetyl paraquinoneimine; SAR, structure-activity relationship; ULN, upper limit of normal. 1
10.1021/tx800415j CCC: $40.75 2009 American Chemical Society Published on Web 01/23/2009
268
Chem. Res. Toxicol., Vol. 22, No. 2, 2009
mechanism of toxicity. Two subtypes of type A were considered: one based on the desired target pharmacology (A1) and the other based on other nontarget pharmacologies (A2). For any given drug, type A2 toxicities are influenced by the selectivity of the drug and the concentration at the receptor(s). For type A1 toxicity, the parent drug or an active metabolite of a pro-drug is the most common culprit. This arises as a side effect of the drug in some or all patients and is usually due to the value for receptor occupancy or enzyme inhibition being too high, occupancy or inhibition occurring for too extended a period (supra-pharmacological effects), or affecting a receptor in a tissue that is not the tissue targeted for therapy. Factors that can cause this are poor dose selection and individual variability including drug metabolism enzyme or transporter polymorphism. Examples include GI bleeding due to cyclooxygenase-1 inhibition by NSAIDs, GI motility decreases due to µ-opioid agonism, constipation by 5HT3 antagonists, myopathy hypothesized to be due to alterations in lipid metabolism in muscle by HMGCoA reductase inhibitors, and extrapyramidal effects of dopamine antagonists. Metabolites in which structural modifications are minor and occur on substituents not critical for target receptor activity, and do not substantially change the physicochemical properties of the parent drug can contribute to pharmacological activity, and hence any supra-pharmacological toxic effects. In some cases, such modifications increase affinity for the receptor, and the metabolite is the sole cause of the suprapharmacological toxic effects. Type A2 toxicity is that elicited by binding to and altering the activity of a specific receptor that is not the primary desired pharmacological target. With these cases, the drugs or metabolites do not have sufficient selectivity over the target receptor such that sufficient occupancy of the nontarget receptor mediates an undesirable effect. An example of this is binding to the IKr channel. This causes QT interval prolongation and can result in fatal cardiac arrythmias. It is interesting to note that side effects for one drug may be the therapeutic benefit of another. For example, the opioid agonist loperamide used to alleviate diarrhea by decreasing GI motility represents the same effect that is considered a negative side effect of centrally acting opioids used for pain. Thus one man’s poison is another man’s cure. Types B, C, and D toxicities have much less well defined targets and pathways. However, all toxicity mechanisms ultimately rely upon the toxic molecule interacting with specific receptors, because, whether it is a specific protein adducted by the toxic molecule (among many to which adducts may be formed) or a region of DNA critical to a cellular process. Types B, C, and D are distinguished from type A2 simply by the fact that we lack knowledge of the identities of specific macromolecule targets. Our hope is that classifications B, C, and D will eventually become obsolete, as knowledge of specific macromolecule targets becomes available for individual toxins, perhaps through the application of proteomic and genomic approaches. At that point, all toxicities would be considered either on- or off-target. Type B refers to idiosyncratic toxicities, such as drug-induced allergy, that do not exhibit classic dose-response relationships and are observed in very low numbers of patients. Type B toxicity requires repeat administration of the drug to achieve a toxic response. Mechanisms of type B toxicity are not wellestablished, but the first pivotal event is considered to be activation of the drug to a reactive metabolite that nonselectively covalently bonds to proteins. The haptenized proteins can trigger an immune response that could either target only haptenized
Smith and Obach
proteins (resulting in toxicity only when the drug is administered) or could begin to also recognize native proteins (resulting in autoimmune toxicity that does not require continued drug administration). Type B toxicity can occur in a variety of tissues and can even occur in different tissues in different patients for the same drug. Normally, though, the three prime sites of toxicity are the liver, blood cells, and skin. Many of the drugs causing type B toxicity exhibit effects on all three. Type C toxicity is distinguished from type B toxicity by a rapid ensuing response and can occur after a single high dose. The effect of the drug is due to a chemical reaction between drug or metabolite and tissue macromolecules, normally caused by bioactivation of the drug to a reactive metabolite (although some alkylating agents used in cancer chemotherapy can directly cause these effects). The effects have close to a classical dose-response relationship, although nonlinearities can occur due to the depletion of nucleophiles such as glutathione. For example, liver toxicity elicited by high doses of acetaminophen via generation of the N-acetyl paraquinoneimine metabolite arises by depletion of hepatic GSH stores. Type D toxicity is toxicity in which the period of dosing is critical and exemplified by carcinogenicity and teratology. Carcinogenesis can be caused by genotoxins which, apart from alkylating agents used in cancer chemotherapy, are generally bioactivated to reactive metabolites. Alternatively, carcinogenicity can be caused by disruption of the endocrinological processes. Thus, it possesses mechanisms overlapping with types A, B, and C toxicities. Even though somewhat more is known about the macromolecular interactions and the pathway leading to the effects, carcinogenicity needs a discrete classification as it has a dose-response relationship that differs from A, B, or C, in that duration of exposure is more important than concentration or dose size. Like carcinogenicity, teratology has a number of possible mechanisms similar to types A, B, and C including receptor mediated and oxidative stress. However, administration of the toxic agent has to occur during a critical time period coinciding with a particular stage of fetal development to elicit the response. Oxidative stress has been suggested (5) as a common mechanism due to the possibility that teratogens (or their metabolites) may cause oxidative stress and influence redox-sensitive signaling pathways. Receptor mediated teratogenesis has been demonstrated in an unequivocal nature with a number of drug targets including blocking of the IKr potassium channel (6) and ETA and ETB receptor antagonism (7). Overall, most metabolites implicated in type A (and D: endocrinological carcinogenicity and receptor mediated teratogenicity) toxicities will be circulating, and the circulating concentrations (particularly unbound concentrations) will reflect the concentrations at the receptor. Excreted metabolites will not contribute per se to the risk/hazard identification for these types of toxicity. For types B and C (and D: genotoxic carcinogenicity) toxicity, the actual metabolites that could be responsible for toxicity are extremely unlikely to be detected per se, and evidence for their presence will be gained by observation of downstream adducts and metabolites arising from rearrangements or conjugation with nucleophiles. The most likely matrix in which these downstream products will be identified is the excreta (due to their now highly polar nature and rapid clearance). With some compounds, covalent binding of metabolites can be observed by radioactivity persistent in the circulation due to irreversible binding to serum albumin. However, actual identification of the pathway to the reactive intermediates and later products is normally largely accom-
ReView
Chem. Res. Toxicol., Vol. 22, No. 2, 2009 269 Table 1. Toxicities Categorized against Time and Concentration/Dose Response for Each Toxicity Type
type
drug
metabolite implicated
A1 A2 A2
tramadol dolasetron fenfluramine
O-desmethyl tramadol hydrodolasetron norfenfluramine
B
troglitazone
C
acetaminophen
oxidation at the chromane ring, oxidative cleavage of the thiazolidinedione ring, NAPQI
D
cigarette smoke
various reactive metabolites of carcinogens postulated
time, dose concentration response
toxicity
directly related to concentration directly related to concentration onset 1 or >months of dosing with dose/ concentration response onset 0.5 or >months, no defined dose size response
ataxia ECG interval change valvulopathy
24-36 h, direct concentration/dose response dose2 × duration4.5 ) risk
hepatotoxicity
hepatotoxicity
lung cancer
Table 2. Drug Withdrawals from 1980 to Date Grouped into Categories of Reason for Withdrawala type A1
a
type A2
type B/C
generic name
daily dose (mg)
generic name
daily dose (mg)
generic name
daily dose (mg)
Alosetron Cerivastatin Encainide Flosequinan Rofecoxib
1 0.3 150 100 25
Astemizole Cisapride Dexfenfluramine Fenfluramine Grepafloxacin Mibefradil Rapacuronium Terfenadine
10 40 15 15 400 100 100 120
Benoxaprofen Bromfenac Nomifensine Remoxipride Suprofen Temafloxacin Ticrynafen Tolcapone Troglitazone Trovafloxacin Zomepirac
600 100 125 300 800 600 400 300 400 200 400
Compounds shown in bold were withdrawn due to hepatotoxicity. The table emphasizes the different ranges in daily dose size in each of the groups.
Table 3. Drugs with Active Metabolites More Potent than the Parent in Keeping with SAR Requirements receptor
characteristics of ligand/ receptor
drug
active metabolite
histamine
charged primary amine
loratidine
desacetyl loratidine
5HT
charged primary amine
zolmitriptan
desmethyl zolmitriptan
5HT
charged primary amine
fenfluramine
norfenfluramine
5HT
charged primary amine
dolasetron
hydrodolasetron
noradrenaline/dopamine transporter
charged primary amines
sibutramine
desmethyl and didesmethyl sibutramine
Mu opioid
phenolic hydroxyl essential for activity 17-hydroxyl group of testosterone essential for receptor binding
tramadol
desmethyl tramadol
flutamide
hydroxyflutamide
androgen
sar introduction of basic charged center tertiary to secondary amine, closer to primary amine secondary to primary amine, mimicking the natural ligand introduction of basic charged centera tertiary to secondary to primary amine, mimicking the natural ligands introduction of a phenolic hydroxyl introduction of H-bonding hydroxyl group to mimic testosterone
a In the case of dolasetron, the introduction of a basic center also gives ion channel properties in keeping with ion channel SAR and is responsible for type A2 toxicities.
plished from a study of excreted metabolites, although hydrolysis of the alkylated protein has also identified possible pathways (8).
2. Dose and Plasma Concentration Response In Table 1, the duration of dosing, concentration, and dose response is explored for each of the classes of toxicity in which the causal agent is likely or proven to be a metabolite. A fuller discussion of these and other examples has been recently published (4). As indicated in Table 1, types A and C toxicities traditionally have a near classic dose-response relationship. Type B, the type traditionally referred to as idiosyncratic drug-induced injury (e.g., hepatotoxicity), is often described as not dose-related. This position has been increasingly questioned (9-11) since most medicines withdrawn from marketing or receiving a black-box warning because of hepatotoxicity were or are prescribed at daily doses greater than 100 mg/day (Table 2).
The concept of daily dose size being a major factor in druginduced liver injury has been further validated using databases in the U.S. and Sweden (12). Medications were categorized into e10 mg/day, 11-49 mg/day, and g50 mg/day groups. Among US prescription medicines, a statistically significant relationship was observed between the daily dose of oral medicines and reports of liver failure (P ) 0.009), liver transplantation (P < 0.001), and death caused by DILI (P ) 0.004). In Sweden, 9% of idiosyncratic drug-induced liver injury occurred in the e10 mg/day group and 14.2% in the 11-49 mg/day group, and 77% of cases were caused by medications given at a dose of g50 mg/day. A statistically significant relationship was noted between daily dose and poor outcome (death or liver transplantation) of Swedish DILI cases (2%, 9.4%, and 13.2% in e10, 11-49, and g50 mg/day groups, respectively; P ) 0.03). Factors compounding the unraveling of complex dose-response relationships are not only the relatively rare occurrence of the event and unknown individual variability elements among human beings that make some susceptible to certain toxins and
270
Chem. Res. Toxicol., Vol. 22, No. 2, 2009
Smith and Obach
Figure 1. Schematic representation of the pharmacological basis of target and toxic effects (see text). The concept is that all responses are due to ligand-receptor interaction theory and therefore obey a standard sigmoidal concentration-effect curve. The figure replaces the range of responses in an individual or group normally shown with the proportion of the population experiencing a response.
others not susceptible but also the actual administration of the medicine (dose and time). Moreover, the classification used here cannot be seen as exclusive. For instance, doubt still exists around the role of reactive metabolites in troglitazone hepatotoxicity (13). The most studied hepatotoxin is acetaminophen. This has a generally regarded safe dose of 4 g/day when administered to adults. A recent placebo controlled clinical study demonstrated that at this dose 39% of acetaminophen treated patients had ALT measurements 3× the ULN, 25% 5× the ULN, and 8% > 8× the ULN (14). It is not understood how these markers of hepatic function link directly to pathological change, but the definition of a safe dose becomes much harder. Furthermore, a few individuals do show pronounced hepatotoxicity at lower doses than 4 g/day: risk factors include enzyme induction by other drugs, fasting, and alcohol abuse (15), but not all cases are explained. The pathway to acetaminophen toxicity is beginning to be unraveled and consists of a metabolic phase, an oxidative phase, the involvement of several cytokines of the innate immune system with stress kinases regulating the cytokine responses (16). That the toxicity has an immune response component and can occur at therapeutic doses in a very small proportion of individuals begins to blur the type B/type C definitions. Some of the drugs with higher incidence of hepatotoxicity, particularly that defined by serum transaminase levels (a biomarker for potential hepatotoxicity) rather than overt organ disease, also blur the distinction between types B and C. Isoniazid hepatotoxicty is classed as idiosyncratic but occurs in between 2 and 28% of treated patients (17), the variation partly depending on the population studied and the definitions of hepatotoxicity. Evidence of a conventional dose-reponse relationship for hepatotoxicity can be gleaned from clinical
reports for both isoniazid and another antituberculosis drug rifampin. Isoniazid demonstrated toxicity at 1000 mg, but not at 400 mg in the same patients (18). In a clinical trial comparing 450 mg with 600 mg rifampin mild (grade 1 or 2), hepatotoxicity was more common in the higher-dose group (46 vs 20%), although severe toxicity was not observed (19). Evidence of a dose-response relationship has been observed for bosentan (20). Lammert et al. (12) anecdotally cites many cases where an individual receives an increased dose of a medication and develops hepatotoxicty. Specific reference is made to a doubling of the duloxetine dose to 60 mg/day causing pronounced hepatotoxicity in an individual patient. Many cases of idiosyncratic severe hepatotoxicity occur against a background of raised transaminase enzymes in a much higher proportion of individuals, the severity of the increase inversely correlating to its occurrence. For instance, bromfenac shows this pyramid of response: when prescribed for over one month, 2.8% of patients have ALT > 3× ULN and 0.5% of patients >8× ULN. Full hepatotoxicity (acute liver failure) with the drug occurred in 0.006% of patients (21). Likewise, for troglitazone 1.3% developed ALT > 3× ULN and 1% of patients >8× ULN (22) with full hepatotoxicty seen in 0.03% (23) These increases commonly show a dose-response relationship; examples include trovafloxacin (22) and talcapone (24), which shows 3× ULN in 1-3% of patients receiving 100 mg of TID and 3.7% of patients receiving 200 mg of TID. The overwhelming evidence points to considering types A and B toxicities from a classical concentration-response standpoint and to the fact that type C is more dose- or concentration-size-related than sometimes stated. Fundamental questions that remain to be answered unequivocally are as follows: (i) For low incidence idiosyncratic reactions, would the incidence increase if higher doses were administered? (ii)
ReView
Are the low incidence of effects seen because we are at the bottom of the dose-response curve? (iii) Is the apparent insensitivity to dose seen because the dose size variation is too small (classical dose-response curves generally spanning 2 log units of dose or concentration from minimum to maximum effect)? To further explore this, we suggest that all pharmacological responses that are triggered by a single ligand-receptor interaction follow a hyperbolic concentration-response curve (with some variations due to special binding mechanisms such as cooperativity). Dose-response curves can be more complex if there are multiple receptors/enzymes involved in the response that can result in curves that deviate from this simple hyperbola (like U-shaped dose-responses), although these are extremely rare. The scheme in Figure 1 depicts five concentration-response curves plotted for different effects of the same molecule. As the concentrations increase, both the severity of effect and incidence rate increase. In this example, the dashed line at 100 indicates a beneficial response rate of 90%, a desired outcome for most drugs. At that level, there is also a readily measured response rate for side effects at 25%, which is not uncommon. (These could be nuisance effects such as transient headache, somnolence, or mild nausea, which are typically tolerated in order to gain the therapeutic benefit.) At the standard dose (100), the incidence of toxicity biomarkers (e.g., phenomena such as increases in circulating liver transaminases or QTc prolongation, etc.) is at 1%, which is not an uncommonly observed incidence when surveying such information in drug package inserts. What poses even greater challenges are the incidences of severe toxicity, which on this plot are at 0.01 and 0.001% of subjects (i.e., 1 in 10000 and 1 in 100000) when the 90% efficacy rate is attained. These incidences are not even visible in plot A and require log transformation of the y-axis to observe them. In clinical trials, toxicity phenomena such as these are not observable due to the number of patients it is reasonable to test in phase 3 trials. Yet, pharmacological concentration-response curve theory mandates that these patients must exist and that their frequency (and hence ability to be observed) is a function of the differential between the desired effect potency and the undesired effect potency (severe toxicity). Thus, in this example, even when there is a split of 5 orders of magnitude between the pharmacological potency for the desired effect and the potency for the worst of all effects (severe toxicity), it could be expected that 1 patient in 100000 would suffer severe toxicity for drug-related reasons. However, such incidences are so low as to be detected only sporadically, and cause-effect can rarely be ascertained in any given case. This extremely low incidence rate can mislead us into thinking that such effects require the achievement of a threshold rather than be part of a continuous dose or concentration response curve. Why the question is important is that the designs of many in vitro and even in vivo experiments to investigate the toxicity of these drugs resort to high concentrations or doses to get a reproducible effect. The scientific plausibility of these tests has to relate to some form of conventional dose-response curve and an increasing incidence of toxicity with dose. Moreover, there has to be some common mechanism either in total or part and the belief that predictable and unpredictable (idiosyncratic) toxicity are part of a continuum. Greater study of the pyramid of response outlined above and the frequency of classical dose response at the base of the pyramid may help to answer these questions.
Chem. Res. Toxicol., Vol. 22, No. 2, 2009 271
3. Metabolites in Types A1 and A2 Toxicity Analysis of a wide variety of drugs indicate that active metabolites tend to mirror the pharmacology of the parent rather than introduce novel de novo pharmacology (2-4). This applies both to primary and secondary pharmacology. Because drugs are carefully designed and synthesized to have high affinity and selectivity for a target, most metabolites will have lower affinity than the parent. Moreover, the narrow nature of structure-activity relationships (SAR) mean that only metabolites with a structure similar to that of the parent will have any affinity. Such metabolites will be the result probably of a single oxidation or reduction step. In many drugs, the presence of a tertiary amine function leads to metabolism along a sequential pathway of N-dealkylation reactions through the secondary amine to the primary amine. Because of the increased metabolic stability of the metabolites, it is often found that the metabolites have greater duration, in the circulation, than the parent and thus exert a longer lasting effect if active (e.g., fluoxetine vs norfluoxetine). Moreover, more accumulation of the metabolite will occur leading to a greater influence of activity. The few occasions where metabolites are more active can readily be explained and predicted by structure-activity relationships (4, 25) as demonstrated by the examples in Table 3 and exemplified by flutamide in Figure 2. Since circulating metabolites are in the vast majority of cases equally active or less active than the parent, unbound concentrations of circulatory material are a major guide. All compounds lose selectivity at high concentrations, or conversely, most compounds will have some limited amount of affinity for any given receptor, even though the affinity is very weak. Analysis of high throughput screening programs and the theory of ligand binding would suggest that such statements are true at concentrations above 10 or 100 µM. At these concentrations, less than 10% of the molecular structure of an average drug (mw 350) is efficiently interacting with the protein (4). This breakdown in structure-activity relationships needs to be considered with metabolites since very different circulating concentrations may occur due to changes in physicochemistry. A conservative concentration based on the above is 1 µM. The free concentration is particularly important: acidic metabolites are often generated from neutral and basic drugs (oxidation of primary alcohols formed by oxidation of terminal methyl groups or some N-dealkylation reactions). These compounds will often have much higher plasma protein binding than the parent (acidic drugs tending to have high affinity for albumin) and be present in high total drug concentrations in the circulation.
4. Metabolites in Types B and C Toxicity Most cases of types B and C toxicity implicate irreversible binding of a reactive metabolite to a macromolecule or oxidative stress induced by a reactive metabolite via redox recycling. These types of interactions can lead to immunoallergic toxicity or more direct damage to the cell or subcellular units such as mitochondria. With few exceptions (such as acyl glucuronides), chemically reactive metabolites are not readily detected in circulation or excreta due to their transient nature. More commonplace identification occurs in the excreta where downstream metabolites (such as mercapturic acid conjugates) may be observed. Quantification of the metabolites of toxic drugs by analysis of downsteam metabolites (2, 3) provides a guide to the no effect types of limits referred to above in the section on dose response. The mass of NAPQI formed at toxic doses of acetaminophen represents over 1 g total body burden per day in a human
272
Chem. Res. Toxicol., Vol. 22, No. 2, 2009
Smith and Obach
Table 4. Characterization of Metabolites in Terms of Class of Toxicity, Matrix, and Structure type A Structure: close-in analogues of parent or concentrations above 1 µM Location: circulation Amount: for type A1 (target activity), consideration is based on contribution to receptor activity (circulating free concentration × intrinsic potency). Metabolites that exceed 25% of parent in terms of receptor activity should be considered. For type A2 (secondary pharmacology), a similar formula can be used but all metabolites that exceed a circulating concentration of 1 µM free should also be considered.
overdose situation and is around 400 mg at the maximum clinical dose of 4 g/day. The reactive metabolite of felbamate (2-phenypropenal) is implicated in the aplastic anemia and hepatotoxicity observed in some patients receiving felbamate. From a standard daily dose of 2.4 g/day, it can be estimated that about 150 mg is formed each day.
5. Implications for MIST Strategy The above considerations have been woven into a strategy to place metabolites into the context of their structure, the matrix in which they are observed, and the toxicity that they could exert. This strategy is shown in Table 4. The unique nature of carcinogenicity in MIST is that its onset is delayed, and the initiation requires prolonged dosing. In Figure 3, we show average initiation times for the various classes of toxicity. This requirement for prolonged dosing changes the safety considerations (4) for many drugs as outlined. The use of a single drug for extended periods is largely confined to ages above 50. It should be noted that while pharmacotherapy for a given chronic indication for those below age 50 also exist (e.g., asthma, psychosis, etc.) it is very common that a specific medicine will be taken for a less prolonged period. Very few drugs are taken for periods exceeding 5 years (4). If we assume a time of first exposure to a carcinogen, chronic duration requirement, and latency periods to produce an increase in carcinoma of 30 years, there would be minimal risk to
Figure 2. Structures of testosterone and hydroxyflutamide showing a key interacting functionality at the receptor. The metabolically introduced hydroxyl group greatly increases receptor affinity in accord with expected the SAR. For further structural considerations of Table 3, see ref 4.
Figure 3. Average times to onset of toxicity in humans. Type D refers solely to carcinogenicity. Note the logarithmic axis.
types B and C Structure: indicative of past reactivity Location: excreta (i.e., estimated total body burden) Amount: 10 mg as a threshold above which the role of the reactive metabolite should be considered Cannot measure reactive metabolite directly and therefore must use downstream products (e.g., mercapturic acids, methyl catechols, etc.).
humans, in terms of cancer incidence, of an unqualified metabolite on the basis of these statistics. Changes to a different drug (whether in the same class or not) attenuate the risk. Most patients have regular changes to therapy even in relatively stable chronic conditions (e.g., hypertension where drug therapy regimes will usually change every 3-5 years to stabilize the patient). Unless a drug is to be used widely for long periods (greater than 5 years) in a young patient group, it is suggested that carcinogenic risk is essentially mitigated. It is already established that drugs of short duration use (e.g., antibiotics used for less than 2 weeks) are not routinely assessed in 2-year animal carcinogenicity studies. We would suggest that a human metabolite, which is not present in animal species or is present only in lower amounts in animal species, does not need further investigation for possible carcinogenicity or mutagenicity if (a) the drug is not a lifetime drug; (b) the metabolite either is pharmacologically similar to the parent or is inactive; and (c) does not possess a genotoxic toxicophore not already present in the parent drug. We believe that in silico methods and appropriate pharmacology screening provide effective methods to evaluate the carcinogenic risk of metabolites. These considerations arose from examining carcinogenicity data from cigarette smoking, aflatoxin, 2-napthylamine, cyclophosphamide, benzene, vinyl chloride, asbestos, and dioxin (4). The most extreme example found to date is aristolochic acids, a family of substances found in herbal medicines. Aristolochic acid represents a very potent kidney toxin and carcinogen. The herbal medicines were used as part of a slimming preparation. Progression to nephropathy occurred between 3 and 85 months after taking the herbal medicine containing aristolochic acid (26). The lesion was tubular necrosis leading to interstitial fibrosis. Duration of treatment was normally around 15 months. Further progression to urothelial carcinoma occurred in about half the patients and devoloped 68-169 months after cessation of treatment (27). Many of these patients were on immunosuppressive drugs due to kidney transplantation. Patients were characterized by the presence of DNA adducts formed selectively by aristolochic acids. The 7-(deoxyadenosine-N6-yl)-aristolactam I DNA adduct is associated with mutations to the H ras proto-oncogene in rodents and the p53 gene in humans. Toxicity and carcinogenicity appear to be dose related in terms of the herbal dose, and extrapolation suggests that an accumulative dose of around or greater than 130 mg is required (28). It is not clear if the nephropathy and the carcinoma share a common pathway in reactive intermediates or exactly how the mechanisms relate. Although worrying, because of the high incidence of toxicity and eventual cancer and its relatively rapid progression, the example is unusual and should be considered from the perspective of the nondrug-like structure of the aristolochic acid(s) and their classical toxin features. Aristolochic acids are nitrophenanthrene derivatives, as illustrated in Figure 4. Examination
ReView
Chem. Res. Toxicol., Vol. 22, No. 2, 2009 273 Table 5. Parameters of Human Circulating Metabolites for Typical New Drugsa number of human circulating metabolites at >10% of: number of human metabolites in circulation
number of human circulating metabolites at >10% parent compound that are:
compound
indication
parent compound in human circulation (%)
bicifadine brasofensine carisbamate CP-122721 CP-615003 dasatinib MK-0524 muraglitazar OR-1896 PHA-543613 prasugrel ruboxistaurine sitagliptin traxoprodil varenicline
antinociceptive Parkinson’s disease anticonvulsant psychotherapeutic psychotherapeutic anticancer cardiovascular diabetes cardiovascular psychotherapeutic cardiovascular diabetic complications diabetes stroke addiction
6.1 45 >95 0 to 5.6b 19 26 57 94 36 59 0 50 83 21 to 58b 91
3 6 1 4 7 15c 1 8 2d 3 13 4 5 6 4
3 2 0 3 2 1 1 0 1 1 2 1 0 3 0
3 5 0 4 6 12 1 0 2 1 13 1 0 4 0
0 2 0 0 2 0 0 0 1 0 1 0 0 0 0
0 4 0 2 4 13 0 0 2 0 N/Ae 0 0 0 0
44 45 0 to >95
5.5 4 1 to 15
1.3 1 0 to 3
3.5 2 0 to 13
0.4 0 0 to 2
1.8 0 0 to 13
average median range
total radioactivity
parent
not detected in animal circulation
less than 10% in animal circulation
a References: bicifadine (32, 33); brasofensine (34); carisbamate (35, 36); CP-122721 (37, 38); CP-615003 (39, 40); dasatinib (41, 42); MK-0524 (43, 44); muraglitazar (45); OR-1896 (46); PHA-543613 (47, 48); prasugrel (49, 50); ruboxistaurin (51, 52); sitagliptin (53, 54); traxoprodil (55, 56); varenicline (57). b Describes parent compound in circulation of CYP2D6 extensive and poor metabolizers, respectively. c Some metabolites were not resolved on HPLC. d For OR-1896, one metabolite was determined, and the rest of the drug-related material was described as other. e For prasurgel, the metabolites in animal circulation were described as detected or not detected.
of the structure indicates potential hazards with the compounds being completely planar and containing two structural alerts (methylene dioxy and aromatic nitro). The compounds are identified as potentially carcinogenic by in silico approaches including MDL-QSAR, TOXLITE, Oncologic, and DEREK (29). The reduction of the nitro group of aristolochic acids and cyclization form the ultimate carcinogen aristolactam nitrenium ion (Figure 4), which forms the DNA adducts. A number of enzyme systems including CYP1A2, NADPH/CYP reductase, NAD(P)H/ quinone oxidoreductase, xanthine oxidase, and prostaglandin H synthetase can catalyze this reduction in the liver and kidney (27). Although data is lacking, aristocholic acids are lipophilic and would be expected to be substrates for kidney and liver drug transporters due to their anionic nature. Excretion and subsequent reabsorption of the compounds due to their lipophilicity would give rise to high proximal tubule concentrations. Studies in rats show high and very persistent concentrations of aristolochic acid I in the kidney of rats after radiolabeled compound administration (30). As part of a general drug development program, considerable information is available to characterize metabolites as the program progresses. Ideally, to satisfy all viewpoints, all metabolites would be identified early in in vitro screens, be synthesized, and subjected to broad ligand binding and also be present in animal studies. However, many factors can impinge
Figure 4. Metabolism of aristolochic acid to a carcinogenic delocalized nitrenium ion (R ) H or OCH3).
on this, and different bodies and organizations can have different viewpoints on considering the data. Moreover, the pivotal study used to define the whole of the human metabolic pathway and clearance mechanisms is the human radiolabeled study, which is typically only done after a compound has demonstrated early indications of therapeutic success in humans (i.e., done in late phase 2). No matter how the study results are used and which guidelines are followed, there are technical considerations that influence the conduct and perhaps the success or failure of a study to support MIST. These are considered below.
6. Practical Considerations in Identifying and Characterizing Human Circulating Metabolites The metabolites that demand further consideration or study (major) have been defined in different ways. In the first proposal regarding metabolites in safety testing (MIST), a pragmatic approach was adopted: a metabolite present as >25% of the AUC of total radioactivity in human was described as important (31). It was proposed that assurance be gained in those cases that such human metabolites were also present in abundance in animal species used in safety assessments. Our position, described previously and above (2-4), has been one wherein percentage values are replaced with absolute amounts/concentrations when defining thresholds of concern since toxicity is related to absolute exposure values. This information is combined with knowledge of the chemical structure of the metabolite (i.e., relatedness to the parent drug or indicative of chemical reactivity) and in which matrix the metabolite is found. A regulatory guidance for industry was subsequently issued by the U.S. Food and Drug Administration (in 2008), which defines the threshold for concern as human circulating metabolites present at 10% of the parent drug in circulation. It is recommended that assurance be obtained that such metabolites exceeding this threshold in humans be present in the circulation of animal species used in safety assessments at levels greater
274
Chem. Res. Toxicol., Vol. 22, No. 2, 2009
than in humans (at steady-state). When described solely in this manner, some challenges can arise in the simple application of this approach: when the parent drug is either very potent (i.e., present at low concentrations in circulation) or extensively metabolized, or both. It must be stated that the guidance stresses a case by case basis for study and open dialogue; in this spirit, it should be remembered that guidelines are not obligatory. Rather, they provide a framework for thoughtful consideration and dialogue to be undertaken as the need arises. Irrespective of the different defined thresholds, all approaches rely upon data generated during the standard human radiolabel ADME study. In this type of study, the administered parent compound contains one or more radionuclei (usually carbon14), which permits defining a comprehensive and quantitative picture of the metabolic profile in circulation and excreta after a single dose. It is important to note that the human radiolabel study has great value in defining the clearance pathways for the drug. To understand the complexity different MIST strategies pose, the scientific literature was searched to find examples of quantitative human circulating metabolite profiles to compare with the corresponding profiles in animals (Table 5). In order to ensure that the experimental methods used to define and identify metabolites in the examples are up-to-date, the literature was searched for human radiolabel ADME studies published from 2006 onward. The reports had to include a quantitative assessment of human circulating metabolite profiles using radiometric HPLC. For those cases reported, there also had to be a publication available containing the corresponding data obtained in laboratory animals (at least one species) for comparison. This yielded 15 examples with adequate data sets (Table 5) ranging over a variety of intended clinical indications. (Note that there was a considerably greater number of human radiolabel studies reported, but many did not have the circulating metabolite data reported, or the animal data were not available.) As a percentage of the total, the parent compound can range considerably from undetectable by radiometric HPLC (e.g., prasugrel, CP-122721) to nearly all of the drug-related material (e.g., carisbamate, muraglitazar). The number of metabolites present in human circulation varied considerably among the 15 examples with some compounds having just one metabolite to others having 13 (prasugrel) or 15 (dasatinib). If the criteria for detailed consideration was based solely on 10% of total drugrelated material, then theoretically, a maximum of 10 metabolites could meet the threshold; however if based on 10% of the parent, that number could be almost infinite if the parent itself was present at a low relative percentage. Such an example is exemplified by dasatinib in which the parent drug comprises 26% of total drug-related material, and there are just two metabolites in circulation that are present at >10% of total drugrelated material. However, there are 13 more metabolites present at between 2.6 and 10% of total radioactivity, which places them at >10% of parent drug and hence warranting further consideration. These metabolites were also minor circulating entities in animal species. If the doses used in toxicology studies were substantially greater than doses administered humans, it is likely that absolute exposures to these myriad human metabolites were greater in animals, assuming that none of them accumulated substantially more in human than in animals upon multiple dosing. Overall, a typical drug will have one metabolite present in human circulation at greater than 10% of drug-related material (range of 0 to 3) and two metabolites greater than 10% of parent (range of 0 to 13). Four of the 15 compounds (27%) had one or two metabolites present in human circulation at greater than 10% of the parent that were not reported as detected in animal
Smith and Obach
circulation. In 5 of 14 compounds (36%), metabolites at >10% of the parent compound in human circulation were present at less than 10% of the total drug-related material in animal circulation (i.e., a major human metabolite was minor in animals). Most of the time, this was four or fewer metabolites, but in one case (dasatinib), there are 13 metabolites present at >10% of the parent in human circulation and less than 10% of total radioactivity in animals. It should be noted that the example of prasugrel could not be analyzed in this manner because the animal circulating metabolites were described only as detected or not detected, but it is highly likely that the number would also be high. One approach to responding to a perceived need for further consideration is to confirm that the concentration of a particular metabolite is higher in animals used in preclinical testing. Overall, using solely the criteria described by the USFDA guidance and based on these representative examples gathered from the recent scientific literature, about one in three of the new chemical entities being developed will require, at a minimum, quantitative bioanalysis of at least one metabolite (in most cases more than one) in humans and animals at steady state to provide assurance that the metabolite is present in animals used for safety assessments. It is not known how many of these cases would require direct testing of the metabolite(s) in animal toxicology studies. If a criterion is used based on metabolites as a percentage (10%) of total drug-related material, the number of cases would drop by half, and instances wherein multiple metabolites would require assessment (e.g., dasatinib and prasugrel) would not exist. Irrespective of the criteria used, the principles outlined by Smith and Obach (2-4) can be applied before committing to extensive bioanalytical studies. First, having identified the metabolites, consider their structure and the type of toxicity they could exert. In the case of dasatinib, the parent drug is the major circulating material, being approximately 2-fold higher in concentration than the next most abundant drug-related entity (M20), which is formed by hydroxylation of the chloro-methylphenyl group. The primary metabolites (i.e., those arising from a single metabolic step and therefore closest in structure to the parent, and likely to be contributors to type A1 or A2 toxicity) including M20 are all at least 10-fold weaker as kinase inhibitors than the parent, with the exception of the desethanol metabolite (M4), which is equipotent (41, 42) This metabolite circulates at a concentration approximately 20-fold lower than dasatinib. The maximum total concentration reached by a metabolite (M20) is around 400 nM, well below the 1 µM unbound concentration where nonspecific receptor affinity may be observed. Assuming broadly similar protein binding and conservation of target selectivity in the metabolites, dasatinb is the only component in the circulation likely to exert a pharmacodynamic effect via kinase inhibition. The pharmacological activity of metabolites and the parent has been assessed by in vitro growth of carcinoma cells. Dasatinib is a multiple kinase inhibitor, including Bcr-Abl, Src family, c-Kit, EPHA2, and platelet-derived growth factor β kinases (41, 42), and the actual selectivity of the metabolites has not been published. Assuming selectivity is similar, it is arguable that no further consideration needs to be applied to these circulating metabolites. Prasugrel, essentially being a prodrug, represents a different case. Here, the active principle is formed following hydrolysis to an inactive thiolactone, which is ring opened by a cytochrome P450-dependent ring opening to form the active species. Prasugrel is not observed in plasma, so all metabolites are subject to further consideration if 10% of parent concentration is a threshold. The thiolactone, the thiol-containing active
ReView
species, and its pharmacologically inert metabolites (S-methyl and cysteine conjugate) are the major circulating species and represent 61% of radioactivity at Cmax and around 40% of radioactivity as AUC. Clearly, this pathway to the (re)active species must be recreated in preclinical species and comparison made that the pathway terminates in similar downstream products. Quantitatively, the only comparisons that should be made are for the (re)active metabolite and possibly its precursor. The multiple other inactive products formed are extremely unlikely to contribute to any toxic or pharmacodynamic effects and should not be subject to further examination or consideration. Reinforcing this is the low dose of prasugrel (15 mg/ day) and the low concentration observed of all metabolites (the major circulating S-methyl metabolite has a total Cmax of 400 nM and a 24 h total Cavg of around 100 nM). We believe that a stepwise approach of identifying the circulating metabolites, consistent with practicality followed by a rigorous appraisal of likely risk before further consideration of qualification, is the correct approach.
7. Enzymatic Basis for Interspecies Differences in Circulating Metabolite Profiles In the final section, recognizing that species differences in metabolism can occur, we have reviewed and updated their physicochemical and enzymological basis. Additionally, we have provided guidance as to the relevance of some of these differences to MIST decisions. Several factors influence the concentration of a metabolite in plasma. These include (a) the rate and extent to which the parent compound is converted to the metabolite; (b) the rate at which the metabolite is subsequently cleared (by sequential metabolic steps or direct excretion); and (c) the relative distribution of the metabolite between plasma and tissues (which is a function of the ability of the metabolite to penetrate tissues and its relative binding to plasma proteins and tissues). Item c will be a contributor to the plasma concentration of metabolite relative to the parent and largely a function of physicochemical properties. For example, metabolism generally increases the hydrophilicity of a molecule, and a more hydrophilic metabolite may not distribute to tissues as well as a more hydrophobic parent drug. This can cause higher metabolite/ parent ratios in circulation. N-Dealkylation of amines and subsequent oxidation of the aldehyde intermediates to carboxylic acids is a common metabolic pathway. Carboxylic acids tend to bind plasma proteins better than amines, and the latter tend to partition into phospholipids in tissues better than the former. This can yield high metabolite/parent drug ratios in plasma, even when the overall extent of metabolism through this pathway is minor, relative to the total metabolism. A good example of this is shown in the metabolism of the experimental compound CP122,721 (37, 38). The parent drug is a lipophilic amine present in circulation at less than 1% of the total radioactivity, while a major circulating metabolite that is a carboxylic acid (5trifluoromethoxy salicylic acid) represented over 50% of circulating drug-related material (Figure 5). However, while predominating in circulation, the salicylic acid metabolite represented a mere 1.4% of the total dose in excreta. Thus, overall the metabolite was a very minor component to total disposition, yet circulating concentrations were high. Unpublished studies in rats showed that the volume of distribution of the metabolite, administered directly, was very low (0.14 L/kg), while that of the parent was high (20 L/kg), and that the plasma protein binding of the salicylic acid metabolite was high relative
Chem. Res. Toxicol., Vol. 22, No. 2, 2009 275
Figure 5. Structures of CP-122721 and its circulating metabolite.
to the parent (fu ) 0.005 and 0.037 for metabolite and parent, respectively; Pfizer, data on file). Items a and b can differ across the species because humans and animals exhibit differences in drug metabolizing enzymes. The study of interspecies differences and similarities in drug metabolism has been an area of research dating back to the 1950s and owes considerable debt to one of its pioneers, Tecwyn Williams. The early work lacked the sophisticated identification and quantitation methods available today; therefore, full pathways were not delineated with only the most abundant metabolites characterized, and almost all the attention focused on metabolites excreted in the urine. Many species differences appeared therefore as absolutes in terms of possessing or lacking a particular pathway. This was supported by the knowledge that certain species lack certain types of enzymes, such as the dog lacking N-acetyltransferase activity, cats lacking the capability to glucuronidate xenobiotics, and different species using different amino acids to generate conjugates of xenobiotic carboxylic acids. The earlier literature emphasized species differences in metabolism; for instance, Caldwell (58) suggested that the rat poorly reflected the human metabolite profile greater than 60% of the time and that other species (except primates) also did not generate good matches to the human-like metabolite pattern. The Rhesus monkey, a species not that commonly used in routine toxicology studies of new drugs, provided the best performance at generating a human-like metabolite profile. The majority of drugs have primary (and indeed secondary and tertiary) metabolites formed by P450 enzymes. Interspecies differences may be considered complex at the enzyme level for reactions catalyzed by the P450 enzyme family since there are orthologous P450 enzymes across species (58), rather than cases of simple presence or absence of an enzyme. Approximately 60% of drugs metabolized by P450s are cleared by CYP3A, an enzyme that is present in all common laboratory species and humans (although there will be some differences in amino acid sequence). This single fact tends to mean that species differences are rarer than was supposed in the early days. Moreover, the promiscuity of the various enzymes means that although proportions of metabolites may vary, it is very rare that there are absolute species differences. Some drug metabolizing enzymes certainly do pose challenges with certain compounds in terms of preclinical data assessment; for example, the CYP2C family of enzymes show considerable differences among the species in terms of their substrate affinities. Unlike P450 enzymes, for certain other enzymes expression can occur in some species and not others. Examples are described that lead to challenges with regard to animal toxicology data and comparative metabolite exposures in animals and humans. These examples are representative, rather than comprehensive. 7.1. Cytochrome P450 Enzymes. Reviews of P450 differences across species have been published, and the reader is referred to these for greater detail (59-62). Among the P450 enzymes, the CYP2 family (especially 2C) shows a great deal
276
Chem. Res. Toxicol., Vol. 22, No. 2, 2009
Table 6. Examples of Demonstrated or Proposed CYP2C9 Substrates in Which the Hydroxyl Metabolite Formed in Humans Is Formed to a Much Smaller Degree in Dogs 2-benzylthio-5-trifuoromethyl benzoic acid bucloxic acid bumetamide 6,7-dichloro-2-methyl-1-oxo-5-indanyloxy acetic acid 6,8-diethyl-5-hydroxy-4-oxo-1-benzopyran-2-carboxylic acid etodolac fenoprofeb flurbiprofen MK-473 phenytoin probenecid proxicromil sultosilic acid tetrahydrocannabidiol tienilic acid torasemide
of divergence, and all of the xenobiotic metabolizing P450s show considerably greater divergence across species relative to P450s involved in metabolism of endogenous substrates (63). Humans possess CYP2C8, 2C9, and 2C19, dogs CYP2C21 and 2C41, and rats CYP2C6, 2C11, 2C12, and 2C13 (with the latter three being gender specific). Well established human CYP2C9 substrates such as diclofenac and tolbutamide are handled differently in dogs. The major metabolite of tolbutamide in dogs is the cleaved product p-tolylsulfonylurea rather than the 4-hydroxy and 4-carboxytolbutamide metabolites observed in humans and rats (64, 65). For diclofenac, dogs primarily generate the taurine conjugate and very little of the 4′hydroxydiclofenac metabolite that is major in humans (66). Smith (67) provided a list of known and proposed human CYP2C9 substrates, in addition to tolbutamide and diclofenac, which showed much lower rates of oxidation (hydroxylation) in dogs (Table 6) and illustrated that the difference resides at the enzyme level across multiple substrates. The metabolism of rosiglitazone offers an interesting example of cross-species differences among CYPs. Rosiglitazone has been shown to be primarily metabolized via N-demethylation and p-hydroxylation by CYP2C8 and 2C9 (68). These metabolic routes are major in vivo, as demonstrated in a radiolabel human metabolism study (69). Metabolic products are similar between species, but a minor route of metabolism in rats and humans is deamination of the 2-aminomethylpyridine moiety to a carboxylic acid metabolite (SB-271258), which in humans is present at only 3.7% of the dose in urine and was not observable in plasma. However, in dogs, SB-271258 is a more prevalent metabolite at 24% of dose in excreta and up to 8.6% of circulating drug-related material (70). Rosiglitazone in preclinical testing caused cardiac hypertrophy in the dog at high doses. A mechanistic investigation showed that this metabolite could have small to moderate trophic effects on cardiac myocytes possibly suggesting some involvement in the toxicity (71). It is tempting to conclude that species differences in metabolism may lead to species differences in response, and indeed, this has often been the first question scientists will ask. However, studies in knockout mice (cardiomyocyte-restricted knockout of insulin receptors, CIRKO mice) demonstrate that the effects are general for other PPAR-γ agonists and are likely due to the consequences of plasma volume expansion, which these class of compounds also show (72). Thus, rosiglitazone serves as an example of a CYP2C cleared drug that shows a difference between dogs and humans in metabolite and toxicity profiles, but that the findings are probably not causally related. In fact, the effect is most likely due to a suprapharmacological effect of the target pharmacology (type A1) and unique to dogs. This
Smith and Obach
is an example that nicely illustrates that more often than not, metabolites with new off-target activities are not a frequent explanation for the finding of an unexpected toxicity and that the parent compound and its target pharmacology are more involved. In addition to differences in CYP2C family enzymes, other notable species differences that could generate different metabolite profiles between humans and animals include greater activity of CYP2D orthologues in the rat and the importance of CYP2B11 in the dog. The substrate selectivity and level of expression of CYP2D1 in the rat explains some of the classic early species differences cited 30 years ago including the abundant aromatic hydroxylations of amphetamine, lidocaine, and diazepam (73) in the rat compared to other routes in humans. Again, these major differences in primary metabolism have little consequences in understanding the side effects or toxicity of these drugs. 7.2. UGT1A4 and Tertiary N-Glucuronidation Reactions. A striking example of a route close to those described as an absolute species difference is often encountered with tertiary amine drugs. There are several examples of tertiary aliphatic amine drugs that are converted to quaternary N-glucuronides in humans. These include mirtazepine (74), citalopram (75), clozapine (76), olanzapine (77, 78), several antihistamines (79), nicotine (80), tamoxifen (81), afloqualone (82), trifluoperazine (83), and midazolam (84) among others. Among species, humans appear to glucuronidate tertiary amines rapidly, with rabbits and guinea pigs being most similar and rats and dogs having low or no activity (85). Rabbits may be the best model species for tertiary N-glucuronidation reactions (86). A molecular basis for this is the presence of functional UGT1A4 enzymes in humans and rabbits (87) but not rats (88). Furthermore, recent research has revealed the specific features of active site amino acids that confer the capability of UGT1A4 to glucuronidate amines as opposed to other UGT enzymes that glucuronidate phenols (89). As an example, olanzapine glucuronidation at the 10-position yields a circulating metabolite in humans, with concentrations approximately half of those of the parent drug at about 10 ng/ mL (77). Using criteria described above to define major metabolites as 10% or more relative to the parent in circulation, olanzapine 10-N-glucuronide clearly achieves that threshold. Furthermore, the N-glucuronide metabolite was not observed in the plasma of mice, rats, dogs, or monkeys, and only trace quantities were observed in dog urine. The metabolism of aliphatic amines to quaternary N-glucuronides represents a pathway that is frequently more prevalent in humans and can give rise to metabolites of greater relative and absolute abundance than in animal species. These metabolites, to date, are devoid of significant pharmacological activity due to the large structural and physicochemical changes this metabolism step produces. Thus, although significant in terms of compound disposition these metabolites do not appear to be significant in risk assessment and do not warrant further investigation in terms of MIST. 7.3. Aldehyde Oxidase. Aldehyde oxidase, despite its name, catalyzes not only the oxidation of aldehydes to carboxylic acids but also, importantly for drug metabolism, the oxidation of carbon-nitrogen double bonds present in aromatic azaheterocyclic drugs. In addition to the primary metabolic pathways, aldehyde oxidase catalyzes the second step in the common oxidative metabolic conversion of cyclic amines such as pyrrolidines and piperidines to lactams. Investigations of aldehyde oxidase activities across species have shown that dogs have
ReView
Chem. Res. Toxicol., Vol. 22, No. 2, 2009 277
Table 7. Exposures (AUC, in µg-h/mL) of Zaleplon and Aldehyde Oxidase Derived Metabolites in Humans and Preclinical Species (Sonata approval FDA, 1999)a compound
human
mouseb
rat
dose zaleplon 5-oxozaleplon desethyl 5-oxozaleplon
10 mg 0.068 (100) 0.198 (290) 0.02 (29)
50 mg/kg 3.5 (100) 0.76 (22) 0.20 (6)
50 mg/kg 134 (100) 4.8 (4) 0.5 (0.3)
b
rabbitb 50 21 14 no
mg/kg (100) (67) data
dogb 40 mg/kg 208 (100)