Review For reprint orders, please contact
[email protected]
From known knowns to known unknowns: predicting in vivo drug metabolites ‘It is better to be useful than perfect’. This review attempts to critically cover and assess the currently available approaches and tools to answer the crucial question: Is it possible (and if it is, to what extent is it possible) to predict in vivo metabolites and their abundances on the basis of in vitro and preclinical animal studies? In preclinical drug development, it is possible to produce metabolite patterns from a candidate drug by virtual means (i.e., in silico models), but these are not yet validated. However, they may be useful to cover the potential range of metabolites. In vitro metabolite patterns and apparent relative abundances are produced by various in vitro systems employing tissue preparations (mainly liver) and in most cases using liquid chromatography–mass spectrometry analytical techniques for tentative identification. The pattern of the metabolites produced depends on the enzyme source; the most comprehensive source of drug-metabolizing enzymes is cultured human hepatocytes, followed by liver homogenate fortified with appropriate cofactors. For specific purposes, such as the identification of metabolizing enzyme(s), recombinant enzymes can be used. Metabolite data from animal in vitro and in vivo experiments, despite known species differences, may help pinpoint metabolites that are not apparently produced in in vitro human systems, or suggest alternative experimental approaches. The range of metabolites detected provides clues regarding the enzymes attacking the molecule under study. We also discuss established approaches to identify the major enzymes. The last question, regarding reliability and robustness of metabolite extrapolations from in vitro to in vivo, both qualitatively and quantitatively, Olavi Pelkonen1†, Ari cannot be easily answered. There are a number of examples in the literature suggesting Tolonen2 , Timo Korjamo2 , Miia Turpeinen1,2,3 & that extrapolations are generally useful, but there are only a few systematic and Hannu Raunio4 comprehensive studies to validate in vitro–in vivo extrapolations. In conclusion, † Author for correspondence: extrapolation from preclinical metabolite data to the in vivo situation is certainly useful, 1 Department of Pharmacology but it is not known to what extent. Background Xenobiotics are eliminated from the human (or animal) body principally as metabolites, products of (predominantly) enzymatic biotransformation reactions. The metabolites formed are usually inactive (i.e., real detoxication products), but sometimes they are pharmacologically or toxicologically active, contributing to therapeutic effects or adverse reactions and tissue injuries [1,2] . Elucidation of the metabolic properties of candidate molecules during drug discovery and development is therefore crucial. Clinically relevant information on metabolism and metabolites is obtained during clinical trials. Typically, in an in vivo pharmacokinetic study, the parent drug and its metabolites are measured in easily accessible body fluids, such as blood and urine, as a function of time after the administration, and their metabolic (or formation) clearances are calculated. Major and minor metabolites are
identified and the kinetics, interactions and further metabolism of major metabolites are studied (if of interest). Furthermore, their pharmacological and toxicological potential can be elucidated if deemed necessary. Certain metabolites, such as glutathione (GSH) conjugates and more distal breakdown products, may provide some indication of the potential production of reactive metabolites. However, this clinically relevant in vivo information is available only after in-depth pharmacok inetic studies in humans. These clinical studies can answer the most important questions concerning clinical relevance of metabolites: what are the major metabolites requiring further studies? How significant are these metabolites regarding pharmacokinetics, interactions, therapeutics or toxic activity? The crucial question posed in this review is: is it possible (and if it is, to what extent) to predict in vivo metabolites and their abundances on the basis of in silico, in vitro and preclinical animal studies.
10.4155/BIO.09.32 © 2009 Future Science Ltd
Bioanalysis (2009) 1(2), 393–414
and Toxicology, PO Box 5000, FIN-90014, University of Oulu, Finland Tel.: +358 8537 5230 Fax: +358 8537 5247
[email protected] 2 Novamass Ltd, Medipolis Center, Kiviharjuntie 11, FIN-90220, Oulu, Finland 3 Institute of Clinical Pharmacology, Stuttgart, Germany and Department of Clinical Pharmacology, University of Tübingen, Auerbachstrasse 112, DE-70376 Stuttgart, Germany 4 Department of Pharmacology and Toxicology, PO Box 1627, FIN-70211 University of Kuopio, Finland
ISSN 1757-6180
393
Review | Pelkonen, Tolonen, Korjamo, Turpeinen & Raunio for the in vitro–in vivo extrapolation of metabolite profile & abundance Profile and abundance of metabolites produced in in vitro experimental systems are naturally dependent on the biological preparation used. Ideally, the biological preparation should contain all the potential enzymes in relative proportions resembling those in the intact organ, which are known to catalyze biotransformations of drug molecules. It is worth stressing that chemicals are individuals and even structurally closely related chemicals may have different major and minor metabolites. This is the principal rationale for comprehensiveness of the biological preparation used in in vitro systems. It is also possible to produce metabolites virtually, by in silico model systems. There are a large number of models available either publicly or commercially; some are used extensively to help with experimental work. However, there is no consensus of how useful or robust the outcomes from these computational approaches are, or what their place in the elucidation of metabolite profiles may be. Currently, there is strong pressure, both ethical and commercial, to move from live animals to in vitro studies and further on technologies to in silico approaches, thus it Prerequisites
Metabolite identification Tentative or absolute chemical structure of a given metabolite; the use of various analytical tools to attain identification
is useful to consider the current state-of-the-art regarding these in silico approaches. Proper elucidation of metabolism is based on the availability and employment of analytical techniques to measure both the parent molecule and pharmacokinetically, pharmacologically and/or toxicologically significant metabolites. In the past and sometimes even today, various chromatographic techniques for the separation and quantitation of the parent drug and its metabolites, with detection by UV or fluorescence, were the mainstream analytical means. However, metabolite identification was often a very difficult task (and in many cases still is), which required synthesis of standard compounds and often months of laborious work. Currently, the advances in liquid chromatography (LC) separation and mass spectrometry (MS)-based detection and identification have enabled fast, reliable and robust analytical means to be used in various in vitro and in vivo situations. Framework
for the in vitro–in vivo extrapolation An idealized workflow or framework for the in vitro–in vivo extrapolation of preclinical metabolite studies is presented in Figure 1. This framework serves as a structure for the following
In silico screening of potential metabolites
In vitro incubations with human (liver) preparations (preferably hepatocytes or homogenates)
Extrahepatic preparations; specific enzymes
Metabolite profile and apparent abundances
Animal in vivo (and in vitro) data (if available)
Metabolite data for the preparation of the ‘first-into-man’ clinical study Validation database for in vitro–in vivo extrapolation and prediction Pharmacokinetic and metabolite profile study in humans
Figure 1. Workflow chart for in vitro–in vivo extrapolation.
394
Bioanalysis (2009) 1(2)
future science group
Predicting in vivo drug metabolites sections, starting with in silico approaches and ending with the most important conclusions. This framework is ‘idealized’, because several important areas have not yet been developed sufficiently for clear conclusions. For example, although primary in vitro metabolite profiles in human hepatocytes most probably represent the situation in the liver in situ, at least in terms of primary and some secondary and more distal metabolites, the in vitro system is still mostly static and hepatocytes do not represent the whole population of cells in the intact liver. Based on the framework described above, we try to address, at least partially, the following questions: n What kind of in silico approaches are available and how useful are they for the prediction of metabolites from a specific compound? Which analytical techniques are currently available and how suitable are they for identifying and quantitating produced metabolites?
n
Should we always use liver as a source of drug-metabolizing enzymes? When are there reasons to employ other tissues?
n
How comprehensive should we be regarding the incorporation of enzymes into in vitro systems?
n
Is it of significance to identify the most important enzymes metabolizing a particular chemical?
n
When should we use animals or animal tissues for metabolism studies? What is the added value of this? How significant are interspecies differences?
n
Which are the important in vivo metabolites that should be detected in in vitro studies?
n
How reliable and robust (both qualitatively and quantitatively) are extrapolations from in vitro to in vivo regarding metabolites? Are data on in vivo metabolites reliable enough for establishing the validity of correlations?
n
| Review
(ADMET) properties early in the drug development process. Another driver is the desire to predict ADMET (especially toxicity) features of the tens of thousands of chemicals we and the environment are exposed to. In this review, we will focus on cytochrome P450 (CYP) enzymes and in silico methods to predict metabolite formation parameters, especially the site of metabolism (SOM) and metabolites generated. The aspects of metabolic behavior that would be useful to predict in silico are listed in Box 1. These aspects are also major topics in in vitro and experimental animal studies during preclinical drug development (see later sections). In silico methods to predict metabolite formation can be divided into two main classes; comprehensive (global) and specific (local). Comprehensive methods (also called expert systems, knowledge-based systems and rule-based systems) mimic human reasoning and attempt to formalize existing knowledge on metabolism pathways. Predictive expert systems portray metabolites of a compound based on knowledge rules. These systems recognize target groups and their metabolic reactions, and attempt to prioritize such reactions [3] . Specific systems generally apply to specific enzymes or metabolic reactions. Specific methods can be grouped into ligand- and targetbased techniques. Ligand-based approaches analyze biological activities of ligands and derive interaction models without need of prior knowledge of target protein structure. Ligandbased methods use statistical tools to explore the relationships between specified structural descriptors and observed parameters of a particular metabolic property. The most commonly Box 1. Aspects of metabolic behavior to be predicted. Biotransformation
Chemical structure of single metabolites Metabolic tree Most-probable metabolic route Warnings for possible toxic intermediates
Binding to enzymes
In silico systems to suggest in vitro/in vivo metabolites During the last few years, substantial advances have been made in developing modeling (in silico) techniques to predict metabolism pathways of compounds. One of the main drivers has been the need to analyze absorption, distribution, excretion, metabolism and toxicity future science group
Identification of concerned enzymes Inhibition of these enzymes Induction of these enzymes
Catalytic reaction
Rate of metabolism Extent of metabolism Mechanism of reaction
Data from [141].
www.future-science.com
395
Review | Pelkonen, Tolonen, Korjamo, Turpeinen & Raunio Box 2. Examples of in silico systems used to predict metabolic pathways and toxicity of compounds. Metabolic pathways MetabolExpertTM
[206]
Expert system with open architecture. Predicts possible metabolic pathways in mammals and plants
METATM
[207]
Predicts metabolic reactions based on the dictionary chosen. Contains >1400 transformation reactions produced by >20 different types of enzymes
METEORTM
[208]
Uses a knowledge base of structure–metabolism rules. A metabolic tree output
MetaSite
[202]
Predicts structure of metabolites formed with a ranking derived from the site of metabolism predictions. Restricted to CYP-mediated carbon oxidation
SimcypTM
[204]
A platform that performs automated in vitro extrapolation to predict in vivo outcomes, supporting the assessment of large numbers of compounds metabolized by multiple enzymes
StarDropTM
[209]
The CYP models in this platform identify a molecule’s most reactive sites towards metabolism for human CYP3A4, CYP2D6 and CYP2C9
Toxicity MetaDrugTM
[210]
Generates networks around proteins and genes. Evaluation of biological effects of small‑molecule compounds on the human body
DEREKTM
[208]
Expert system. Predicts whether a chemical is toxic in mammals and bacteria. Provides information about potential toxicity when no experimental data are available
TOPKAT TM
[203]
Uses crossvalidated quantitative structure–activity relationship models. Assessment of various toxicity end points from a chemical’s molecular structure
used ligand-based methods are quantitative structure–activity relationships (QSAR) and pharmacophore modeling. QSAR tools are pattern recognition methods that describe any mathematical or statistical method that might be used to detect or reveal patterns in data [4,5] . Pharmacophore technology defines the minimum functionality of a molecule to exhibit activity [6] . Target-based methods calculate atomic interactions between ligands and their target macromolecules involved in metabolic processes. They require 3D structures of ligands and macromolecules and thus need more computational power. Algorithms from target- and ligand-based approaches are integrated to gain additional structural insights and to further validate the individual models. Today, ligandand target-based methods are often combined (so-called mixed or ligand plus target approaches) 7,8] . Many homology models of human CYP proteins have been published. These models were first based on bacterial CYPs 396
Bioanalysis (2009) 1(2)
and later on the rabbit CYP2C5 structures [9,10] . A recent review by Lewis and Ito demonstrates that it is possible to generate realistic models for the major human CYPs using homology modeling and docking tools [11] . Box 2 lists examples of comprehensive and specific in silico systems that are currently available for predicting compound metabolism. Also included are some systems aimed at predicting toxicity of compounds, as these systems are often linked. Prediction of organ toxicity by these mostly expert systems is primarily based on their ability to predict putative reactive metabolites. Accurate prediction of reactive or otherwise unwanted metabolites would be of paramount importance in the early drug development process. The features as well as strengths and limitations of these systems have been recently reviewed by Kulkarni et al. [12] , Crivori and Poggesi [8] and Muster et al. [13] . In addition to these systems, there are several more general software that have been applied in SOM prediction. Examples include future science group
Predicting in vivo drug metabolites SYBYL (by Tripos [201] ), GRID and Almond (Molecular Discovery [202] ) and Catalyst TM (Accelrys [203]). Although physiologically based pharmacokinetic models such as Simcyp TM are not designed for metabolite prediction, they nevertheless allow the incorporation of modules for the simulation and prediction of metabolite kinetics [204] . In the early phases of drug discovery, data are derived from in vitro assays using, for example, animal and human liver homogenates, microsomes, cell cultures, expressed enzymes. Results of in silico screening based on data from such systems are by no means directly applicable to the whole organism. Thus, to predict in vivo metabolism pathways, it is necessary to apply several other methods. Later in the drug developm ent pipeline, when experimental data are more accurate, in silico models can be developed on focused series of compounds. Screening & identification of in vitro metabolites As stated earlier, in the screening phase the most important objective for metabolite profiling studies is comprehensiveness (i.e., no important metabolite should be missed and at least tentative identification of the metabolite structures and semiquantitative information regarding their abundance should be obtained). This information is needed to provide a basis for further studies for the assignment of metabolizing enzyme(s) and also for comparison of metabolic profiles between different species, which in turn is needed to select the most suitable species for in vivo toxicity studies. The metabolite structures are also of very high importance to evaluate their possible toxicity (reactivity) and to point the metabolic soft spots in the parent drug structure, so that its metabolic stability be correctly modified by synthetic approaches, if needed. These requirements severely restrict the usefulness of available analytical approaches. Consequently, the metabolites are screened and identified usually by LC–MS techniques. Several recent reviews describe typical procedures and chromatographic and MS instrumentation used in metabolite identification [14–16] . LC–MS methods utilized in these types of study are not usually optimized, but fast generic chromatographic methods (e.g., gradient runs with 5–90% acetonitrile or methanol vs aqueous phase in a few minutes), are used and the data are acquired using MS. future science group
of various LC–MS techniques for metabolite studies In drug metabolism studies, where practically all analytes have ionizable functional groups, such as hydroxyl, carbonyl, amine, amide, sulfone and so on, electrospray ionization (ESI) is clearly the most used interface (ion source) between LC and the mass spectrometer [17–20] . As the ESI is strongly affected by ionization of the functional groups in the liquid phase (HPLC eluents), it provides excellent cover for almost all kinds of drug compounds and their metabolites. Another ionization source used in LC–MS studies is atmospheric pressure chemical ionization, which provides good ionization for more nonpolar compounds, such as steroids [21–23] , and also the use of the quite recently introduced atmospheric pressure photoionization is increasing [23–26] . Recently, Cai et al. described that the atmospheric pressure photoionization clearly provided better ionization efficiency for a number of neutral drug-like compounds over ESI or atmospheric pressure chemical ionization [26] . Even though various approaches for each analytical task can be utilized, the basic rule of thumb has been that LC with time-of-flight (TOF) or quadrupole (Q)TOF mass spectrometers are the instruments of choice for screening (metabolite screening and tentative identification), whereas the triple quadrupole mass spectrometers are superior in high sensitivity quantitative work of known target compounds (Table 1) . In addition, recent developments in ion-trap technology, especially the introduction of linear ion-trap mass spectrometers and their combination with triple quadrupole mass spectrometer (i.e., triple quadrupole-linear ion trap; Q-Trap) have opened more possibilities regarding also their use in metabolite screening [27,28] . During the last few years, a number of other new types of mass spectrometers or new hybrid instruments have become commercially available and the analytical possibilities of these instruments, such as ion-trap TOF [29] or orbitrap mass spectrometers [30,31] , can provide some new insights into the analyses. Furthermore, the developments in separation techniques have leapt forward during the last decade.
| Review
Suitability
Drug metabolism Also known as biotransformation, it is an enzymatically catalyzed biochemical modification or degradation of a drug
Metabolite screening Use of various analytical techniques, by which production of drug metabolites in a given experimental system is assessed
Tentative
identification of metabolites Tentative identification of biotransformations, that is, metabolites produced, are identified by mass spectrometric data. With soft atmospheric pressure ion sources, such as electrospray, www.future-science.com
397
Review | Pelkonen, Tolonen, Korjamo, Turpeinen & Raunio Table 1. Comparison of features of time-of-flight and triple quadrupole mass spectrometers in in vitro drug metabolism studies. Feature Full scan detection sensitivity (for metabolite screening) Mass accuracy (for exact mass measurement) Data acquisition rate (for fast chromatography)
MS/MS capability (for elucidation of biotransformation sites) Linearity of response (for quantitative studies) Precursor ion or neutral loss scanning Detection sensitivity for known compounds Price Optimal use
Time-of-flight
Triple quadrupole
[37–39,43–44]
[33–36]
Very high
Low
Very good Very high
Poor Low–very high (depending on scan mode) Yes
No (yes in QTOF) Good No Very high High Metabolite screening; identification of biotransformations
Very good Yes Extremely high Moderate–high Metabolic stability; interaction studies
MS: Mass spectrometry; QTOF: Quadrupole time-of-flight.
it is mainly the intense molecular ion that is detected in the spectrum, enabling the elucidation of the molecular weight change during the biotransformation reaction. Some common changes in molecular weights caused by metabolic reactions are given in Table 2 . Depending on the type of MS, fragment ion MS data can also be obtained, thereby enabling the elucidation of the biotransformation site. The most commonly used fragment ion data providing tandem MS (MS/MS) instruments are triple quadrupole and ion-trap mass spectrometers [32–37] . Some other instrument types, such as TOF [38–40] , orbitrap or ion cyclotron resonance (ICR) Fourier transform (FT) mass spectrometers [30,31,41] , in turn have high mass resolution (capability to separate between ions with mass-to-charge ratio [m/z] values close to each other) and very good mass accuracy (ability to produce accurate mass data with
