Rapid assessment of drug metabolism in the drug discovery process

European Journal of Pharmaceutical Sciences 11 Suppl. 2 (2000) S61–S72www.elsevier.nl / locate /ejps

Rapid assessment of drug metabolism in the drug discovery process*Marc Bertrand, Peter Jackson, Bernard Walther

` ´ ´Technologie Servier, 25 –27 Rue Eugene Vignat, BP 1749, F-45007 Orleans Cedex 1, France

Received 15 June 2000; received in revised form 13 July 2000; accepted 19 July 2000

Abstract

For a few years, in vitro models have been used as part of high-throughput screening (HTS) programs to characterize metabolicstability, drug permeability and drug solubility. This has allowed the rapid selection of lead candidates based not only on pharmacologicalendpoints but also on biopharmaceutical specifications. What has now become clear is that the huge amount of data produced to sort seriesof compounds has a limited predictive value when used to predict human pharmacokinetic parameters. More complex in vitro teststogether with some simple in vivo tests used as validation steps have been developed in order to provide absolute data that may be used asa complement to lead selection providing reliable predictions not only of human pharmacokinetics but also of potential drug–druginteractions. These models may be used as part of selective drug screening (SDS) programs. Further advances in analytical and in vitrotechniques will see some of these models shifting from SDS to HTS programs putting the emphasis on the use of expert systems andphysiologically based pharmacokinetic models (PBPK) to provide meaningful endpoint data. 2000 Elsevier Science B.V. All rightsreserved.

Keywords: Screening; HTS; Metabolism; Caco-2; Modeling; PBPK

1. Introduction of the major biopharmaceutical characteristics of newdrugs is possible. Integration of these tools into the drug

The aim of the pharmaceutical discovery and develop- discovery process has been made possible because of thement process is to identify new pharmacologically active recent advancement of powerful analytical techniques suchchemical entities that, once marketed, can be simply and as LC–MS–MS together with the increased use of labora-safely prescribed by physicians. tory automation.

In the past few decades, new chemical entities were only Today, the emphasis is more on the appropriate use ofscreened for pharmacological activities, after which pre- these tools to solve the right problem at the right moment,clinical and clinical development was performed. Today, in other words to master the balance between theirthe increasing costs of pre-clinical and especially clinical systematic and selective use. Now that data can bedevelopment, together with the impossibility to increase in generated with great rapidity, additional issues such asparallel the marketed price, have pushed companies to validation and data analysis have arisen.select new chemical entities to be developed in a different This paper focuses, using specific examples, on themanner. This has been achieved by combining chemistry predictive value of certain metabolic tools at the variousand pharmacology together with biopharmacy during stages of the drug discovery process.earlier discovery stages in order to improve candidateselection for pre-clinical and clinical development.

In vitro tools, which have been introduced years ago into 2. Biopharmaceutical screeningthe drug metabolism field, are now well-mastered. Whenpharmacokinetic parameters are combined with physico- A critical step in any quantitative drug design approachchemical descriptors in the discovery process, a prediction has always been to describe the molecules with appropriate

physicochemical parameters. This also applies to the drugmetabolism field. Because the biological processes in-*Corresponding author. Tel.: 133-2-3881-6121; fax: 133-2-3881-volved in drug metabolism, as well as the level of6177.

E-mail address: [email protected] (B. Walther). interaction with biological environments, are of a multi-

0928-0987/00/$ – see front matter 2000 Elsevier Science B.V. All rights reserved.PI I : S0928-0987( 00 )00165-2

S62 M. Bertrand et al. / European Journal of Pharmaceutical Sciences 11 Suppl. 2 (2000) S61 –S72

dimensional nature (Testa et al., 1997) it is not possible to results (or even HTS results) need to be confirmedexplore all these processes at a screening stage. The tools (validated) for compounds that have been selected forused in drug metabolism screening must therefore corre- pre-clinical development using, for example, animal n-in-spond to each important individual limiting factor for the one dosing. This serves to increase the confidence in adevelopment of a new chemical entity, i.e., absorption particular model used in the screening stages. When morecharacteristics including dissolution and permeability, rate informative parameters are obtained during the SDSand extent of metabolism, enzyme inhibition and induction stages, accurate predictions of drug plasma levels, vari-potential, as well as the identity of the liver enzymes ability in clearance and risk of drug–drug interactions canimplicated in primary metabolic pathways (Fig. 1). be made (Fig. 2). However, the screening process does not

The different types of screening tools to be used in a stop there and has been stretched out to the first humanscreening process may depend on many factors. These may administration for compounds having a good in vitrodiffer depending on corporate strategies to integrate bio- profile (e.g., good metabolic stability) but for which verypharmaceutics into the screening process, but the screening few in vivo animal data are available.tools must always be adapted to the number of compounds The major question to be addressed is whether we areto be tested. Rodrigues (1997) has emphasized the need for really able to maintain in vivo predictability when in vitroa rational approach in drug screening to apply the right methods are simplified to be used in HTS, or whether wemodel to the right problem or to the right number of are just generating simple metabolic parameters with nocompounds. real predictive use when dealing with large series of

If the objective is to sort a large number of molecules in compounds.high throughput screening (HTS) processes in order todiscover a lead, there is no doubt that the only way is toobtain the simplest parameters by the simplest methods 3. Evolution in assay technologies(e.g., a single data point for the evaluation of the rate ofbiotransformation). Rapid assessment of drug metabolism parameters has

In contrast, when working with a smaller chemical been possible with the combined use of:series in a more selective drug screening (SDS) process,one can generate more informative parameters (e.g., Km

and V values in metabolic stability tests) which can be • robotics, using 96-well (or more) plates, for manipula-max

used to directly predict the corresponding in vivo parame- tions including preparation of solutions, in vitro incuba-ters (Fig. 2). This stage can follow a preceding HTS stage, tions, the preparation of the samples before analysiswith the aim of transforming leads into candidates of (Simpson et al., 1998), or the maintenance of cells usedpotential clinical use, or it can correspond to a backup for in these in vitro screens. These robotic systems increasedrugs that have failed during the development process. In the throughput of all simple in vitro tools, allowingthe latter case, some in vivo data may be available to aid scientists to focus more on the interpretation of datathe choice of the screening tool to be used. rather than on the routine aspects of sample manipula-

In many of the screening strategies described, SDS tion

Fig. 1. Objectives of a rapid assessment of drug metabolism parameters.

M. Bertrand et al. / European Journal of Pharmaceutical Sciences 11 Suppl. 2 (2000) S61 –S72 S63

Fig. 2. Schematic presentation of screening tools in metabolism.

• LC/MS/MS as an analytical tool, the sole able to lipophilicity and when dividing the bioavailability of drugsrapidly achieve the sensitivity and specificity required into absorption and metabolism processes, one can see aswhen analysing chemicals in complex mixtures (Ber- schematized in Fig. 3 that lipophilicity is important both inman et al., 1997). Development of these multi-com- terms of metabolism (increased metabolic clearance withpound quantitative analytical methods has been shor- lipophilicity) and absorption (limited absorption for verytened with the help of powerful software such as polar drugs together with protein efflux or solubilityQuanlynx (Micromass) guiding the analytical chemist limitation when too lipophilic).not only in the critical MS–MS optimization phase but However, this is probably too much of a simplification,also in the design of acquisition and quantification and classical lipophilicity parameters are not fully satisfac-methods as well as in automating the reporting phase. tory for drug metabolism prediction at the screening stage

and are applicable only within structurally related series.Their real value lies in the fact that they can be estimated

4. Screening parameters in drug metabolism very early by simple calculations, and when used as alertsby synthetic chemists, can provide lipophilicity ranges

One of the simplest physicochemical parameters one compatible with acceptable pharmacokinetic profiles.may use to explain the relationship between chemicalstructure and some drug metabolism parameters is the 4.1. Drug absorptionlipid /water partition coefficient. Many pharmacokineticparameters are in one way or another correlated with Oral absorption depends on the pharmaceutical phase

Fig. 3. Schematic relation between lipophilicity and drug absorption and metabolism.


corresponding to the in vivo dissolution of the drug, to involvement of the P-gp efflux protein are available. Theyphysiological conditions in the GI tract such as gastric are membrane-based, quantifying ATPase activityemptying and presence of bile acids, and with the passage (Druekes et al., 1995) or competition with a referenceof the drug across the intestinal membrane. substrate, or they are cell-based and measure the intracellu-

Drug dissolution and intestinal permeability are there- lar accumulation of a fluorescent substrate (Sharom et al.,fore the two major factors to be evaluated in this biological 1999).process (Lipinski et al., 1997). The less a drug is absorbed, While allowing permeability comparisons within athe more pharmacokinetic variation will occur from one chemical series (SDS), with use of calibration curvessubject to another, and the more it may be influenced by based on compounds with a known in vivo absorption, theexogenous factors such as diet and dosing regimen, leading Caco-2 model can also be used to predict in vivo absorp-to a more difficult clinical development. tion for large numbers of compounds. The combined use of

n-in-one incubations with LC/MS/MS analyses can in-4.1.1. Drug dissolution crease throughput to such an extent that its use in HTS

Solubility under physiological conditions is probably the may be envisaged.first physicochemical parameter to be evaluated before One of the drawbacks of the Caco-2 cell line is themeasuring in vitro drug metabolism parameters (per- difference in metabolic competence of these cells com-meability, metabolic stability, etc.). This probably should pared to normal intestinal enterocytes with an over-expres-also apply to in vitro pharmacological testing, where sion of CYP1A1 and a down regulation of the constitutivesolubility can often be a problem. But in order to fully CYP3A4. This can be overcome by restoring CYP3A4understand the limiting factors in the absorption process, (Fig. 4) with a simple modification to the culture mediumdissolution profiles over time or as a function of the pH are (method derived from Schmiedlin-Ren et al., 1997), or bymore appropriate measurements than just intrinsic solu- using CYP3A4 expressing Caco-2 cell lines (Crespi et al.,bility values used mainly for ranking purposes. 1996). When validated against in vivo intestinal metabo-

Many attempts have been made to miniaturize classical lism, the intestinal passage, with both its permeability anddissolution tests used for pharmaceutical formulations and metabolism components, could then be assessed using theto make them applicable to the drug discovery process. same in vitro model.Ideally, these tests should consume a few milligrams ofpowder and should allow the evaluation of the influence of 4.2. Metabolismproteins, bile acids, etc., in the media.

A small scale shake flash has been described (Johansson In addition to the important connection with oralet al., 2000) which allows intrinsic drug solubility to be bioavailability, metabolic parameters are also important inassessed in 96-well plates. Others authors have combined pharmaceutical development as they may explain inter-dissolution and permeation tests, mimicking almost the subject variability, drug–drug interactions, non-linear phar-entire absorption process. The solubility versus pH profile macokinetics and toxic effects. If we define metabolism asof drugs and their passage through artificial membranes the chemistry of enzymatic and non-enzymatic processes(Avdeef, 2000) or Caco-2 cultures (Kataoka, 2000) have (Mayer and van de Waterbeemd, 1985), this covers the ratebeen used in discovery programs. (metabolic stability) and extent (metabolic routes) of

metabolism, enzyme inhibition and induction potential of4.1.2. Drug permeation the drug concerned as well as the identity of the liver

The intestinal passage is a function of intestinal per- enzymes implicated in its primary metabolic pathways. Inmeability and pre-systemic metabolism. Different in situ vitro models for the evaluation of drug metabolic stabilityanimal models have been described to estimate these two have already been extensively used in the pharmaceuticalparameters (Schurgers et al., 1986), but they are generally industry and have easily been adapted to screening stages.more adapted to the study of complex drug absorption Other tools in drug metabolism (i.e., those used to studyproblems than to screening. Several cellular models may metabolic routes, inhibition or induction potential, etc.)be adapted to screening, but the Caco-2 model is currently have also been integrated into the drug screening pro-the standard method. The success of this human cell line is cesses.mainly due to its simplicity since it is able to spontaneous-ly differentiate as a monolayer of enterocytes with well- 4.2.1. Metabolic stabilitydeveloped tight junctions after only several days in culture. The hepatic microsomal model, when compared to moreThis allows measurement of both transcellular passive complete cellular models such as liver slices (whichtransport through the membrane and paracellular passive express all liver enzymes together with intact cell–celltransport through the tight junctions. Active transport communications) or hepatocyte cultures (expressing all theincluding P-gp efflux may also be studied relatively simply metabolic liver enzymes), is the simplest and the bestusing the Caco-2 model. adapted tool for early drug screening stages.

A number of simpler tests assessing the potential Hepatic microsomal fractions are available for all animal


Fig. 4. Activity of CYP3A4 in either vitamin D3-induced or CYP3A4-transfected Caco-2 cells (method derived from Schmiedlin-Ren et al., 1997).

species including humans, and express all the CYP super- scaled up to the in vivo situation to obtain an estimate offamily of enzymes together with other oxido-reductases the metabolic oral bioavailability. Simulations may beand certain conjugation enzymes. This model may be used performed for different oral dose levels in order to modelto obtain simple or more complex metabolic parameters saturation of first-pass metabolism and therefore non-linearwhich may be used to predict the in vivo situation in pharmacokinetic effects (Fig. 6).humans. It is also easily used in an automated system and This approach, which has been developed primarily tois therefore totally adapted to HTS. predict a metabolic clearance in humans before the first

Because of the simplicity of use of the microsomal administration of the drug, is also well adapted to SDSmodel, one tends to forget the importance of experimental when automated and used with cassette LC–MS–MSconditions in in vitro incubations. Many of the enzymes analyses. Its throughput may be increased by using theimplicated in the metabolism of xenobiotic compounds are same hepatic microsomal model but with only one veryhighly saturable and therefore methods of calculating rates low substrate concentration in order to minimize any riskof metabolism based on a single substrate concentration of metabolic saturation, allowing a determination of intrin-may, if this concentration is not low enough, produce sic clearance (Cl ). As shown in Fig. 7, this method hasint

results that drastically underestimate the in vivo situation. yielded a good in vitro–in vivo correlation for 37 structur-For this reason, we use a simple method that provides ally diverse compounds in the rat. In humans, goodapparent K and V determinations for the overall metabo- correlations with fewer compounds have also been ob-m m

lism of a drug, based on computer fitting of in vitro served. Further simplification (e.g., one substrate con-disappearance curves using the Michaelis–Menten inte- centration and one time point) allows an estimate ofgrated equation (Fig. 5). When used with certain physio- metabolic rates that is extremely well adapted to HTS andlogical parameters (hepatic blood flow, liver weight, can rank compounds according to in vitro rates of metabo-microsomal yields, etc.), these simple in vitro data can be lism. However, a higher risk of failure exists when


Fig. 5. Non-linear regression fitting of the Michaelis–Menten equation to calculate apparent V and K metabolic parameters (A), and their use form m

metabolic bioavailability prediction for a simulated administered dose (B).

Fig. 6. Prediction of metabolic bioavailability and its evolution with dose compared with actual bioavailability in vivo in the rat, dog and human.


Fig. 7. Correlation between metabolic bioavailability predicted from K /V values and measured in vivo.m m

predicting in vivo clearances using these simplified meth- may be toxic, thus decreasing the therapeutic margin. In aods. few cases, metabolites may even have a reverse pharmaco-

Assessing K and V in parallel for certain compounds logical activity (agonist versus antagonist) resulting in am max

at an SDS level with animal and human material, together compound that is very difficult to develop. Even whenwith in vivo animal data on selected compounds, re- some metabolites have pharmacological activities similarinforces in vivo human prediction if the in vitro / in vivo to that of the parent drug, development is, in general, morecorrelation in rat is good within a series. This may render expensive and hence such problems should be identified asmore reliable future predictions with new compounds and early as possible. This is sometimes possible in a screeningcan be seen as a useful validation step. context by combining a pharmacological test and an in

Non-CYP dependent pathways can be assessed either vitro metabolism model (e.g., hepatic microsomal incuba-with hepatic microsomes or with hepatocytes preparations tions together with receptor binding studies with the(immobilized or on suspension). For peptides or esters, incubates).blood plasma or cell culture medium incubations can be High-throughput methods for in vitro formation andcarried out for large series of compounds. mass spectrometric characterization of microsomal drug

metabolites have been developed (van Breemen et al.,4.2.2. Metabolic routes 1998). However, it is doubtful that the determination of the

Metabolism can be regarded as a biological process extent of metabolism at the HTS and SDS stages has anywhich, when extensive, may make a chemically pure other interest than to fill databases, even if this approachcompound (e.g., 100%) become highly impure (e.g., 1%) may be valuable if performed for a few selected com-within a biological system. The impurities (metabolites) are pounds which can help the chemist apply metabolicgenerally pharmacologically inactive but in certain cases stabilizing modifications. Due to metabolic switching,


where a different route of metabolism can be favoured by • the number and nature of CYP enzymes involved (oneblocking a previous route, the real impact of any chemical major CYP implicated producing a risk of interactionmodification on metabolic stability must be verified by with all drugs that can inhibit this given CYP);further in vitro experiments. • the inhibition potential based on the nature of the CYP

More interesting is the possible use of structural tools inhibited and on the in vivo extrapolation of thesuch as LC–MS–MS at the HTS or SDS stages to evaluate inhibition constant (K ) (a K close to the in vivo hepatici i

the possible formation of reactive metabolite intermediates, concentrations estimated from maximal plasma levelswhich may cause problems in terms of binding to macro- producing a risk of interaction with other co-adminis-molecules and provide a risk of hepatic toxicity or tered drugs metabolized principally by the inhibitedgenotoxicity. Recent in-house experiments have provided a CYP);sensitive method for the detection of reactive metabolites • the induction potential based also on the dose requiredby using a combination of hepatic microsomal incubations to produce induction compared to the in vivo dosein the presence of GSH (used to trap reactive inter- regimen (an induction potential at a low dose being amediates) together with LC/MS/MS detection of GSH risk of interaction of the drug on all other co-adminis-adducts using the scan of the possible parent ions of a tered drugs metabolized principally by the CYP in-GSH-specific fragment. As shown in Fig. 8, this sensitive duced).method may be performed with non-radiolabelled com-pounds and allows both detection and structural elucidation This strategy, which has been adapted from the FDAof the reactive intermediates. and European Agency guidelines, gives a better under-

standing of how to use some in vitro data (sometimes4.3. CYP enzymes: identity, inhibition and induction generated in the earlier stages of development). But wepotential may wonder whether there is a need to integrate these

parameters into the screening process?CYP enzymes are involved in many of the primary

metabolic routes of xenobiotics, and as such, they de-termine the metabolic clearance of many drugs. CYP 4.3.1. Inhibition of drug-metabolizing enzymesenzymes may also be inhibited or induced by drugs. When The tools available for inhibition studies are the same ascombined, these three characteristics give a good under- previously described for metabolic stability determination.standing of the risk of drug–drug interactions. Hepatic microsomes are used in combination with specific

These parameters are generally assessed during pre- CYP substrates. They may clarify the inhibition potentialclinical stages, and we have developed an in vitro strategy of new entities and the nature of the CYP enzymesfor the rational assessment of risks of interaction, mainly to inhibited.rationalize the use of such preclinical data into a clinical The inhibition potential of a given candidate can becontext (Fig. 9). This strategy, which defines the type of assessed for large series of compounds and is a valuablepotential drug–drug interaction and the clinical studies parameter to predict drug–drug interactions, the mostneeded, is based on three types of in vitro data: hazardous being irreversible inhibition (e.g., suicide sub-

Fig. 8. Reactive intermediate detection by parent scan analysis and structural elucidation.


Fig. 9. Rational strategy for assessment of potential drug–drug interactions.

strates). Inhibition potential can therefore be considered as animals to humans is unreliable. In addition, as thea useful parameter to be measured at an early screening quantity of compound required for such studies is normallystage. Particular attention must be paid to compounds with a limiting factor in any screening project, the need fora very good metabolic stability to control that it is not reliable, small-scale in vitro tools has become urgent.linked with an inhibition potential (inhibitory metabolites). Today, long-term hepatocyte cultures allowing the in-Should inhibition be observed, the inhibition constant (K ) duction process to be completely established are the besti

¨must be measured in order to scale up the effect of predictive tools (Ferrini et al., 1997; Langouet et al.,inhibition on plasma levels by comparison with predicted 1997). Specific CYP activity measurements complementedor actual in vivo concentrations in humans. This approach with Western blot analyses allow the detection of anyis therefore limited to a small number of chemicals. induction potential directly with human hepatocytes. Using

Inhibition measurements can be simplified by estimating reference inducers, the induction potency of test com-an IC value, making them suitable for SDS. In parallel, pounds can be evaluated. However, these tests are not50

the use of microsomes prepared from cells transvected entirely satisfactory because of the limited availability ofwith one human CYP together with a simple enzyme assay human hepatocytes and the inter-subject variability ob-detection method have been proposed for HTS (Crespi et served in response to inducers. For SDS screens, ratal., 1997). hepatocytes cultures are a possible alternative and can be

used to sort molecules with respect to their induction4.3.2. Induction of drug metabolizing enzymes potential. Once a link between both species has been

Strategies for evaluating the induction potential of drug established.candidates have often been, and still are, based on ex-vivo The use of cell lines with inductive properties (deanimal liver analyses obtained from toxicological studies. Waziers et al., 1992) has not been convincing either. ToolsBecause of the important species differences observed in well adapted to large scale screens will probably beterms of inducibility of liver enzymes, scaling from genetically engineered cells expressing all or part of the


CYP regulation machinery coupled with an easily quantifi- CYPs are variable and there is no reason to point towards aable reporter gene protein (Ogg et al., 1999). These tools particular one.could be envisaged as becoming part of the HTS process. Ideally, a drug should be slowly metabolized but by a

number of CYP’s, thus minimizing the consequences of4.3.3. Identification of liver enzymes drug–drug interactions (induction or inhibition). It now

Identifying the number and nature of enzymes impli- appears justified to weed out during screening programs allcated in the metabolism of drugs allows to predict the compounds metabolized by a single enzyme, and those forpotential variability in in vivo clearance as a result of the which a polymorphic enzyme accounts for much of theinvolvement of drug metabolizing enzymes. Some CYP metabolic clearance.enzymes are known to be particularly variable from onesubject to another (e.g., 1A1, 1A2, 3A4), whereas othersare polymorphically expressed (e.g., 2D6, 2C19). 5. Handling the data

Classically, three approaches are run in parallel for theidentification of the enzymes implicated in the metabolism With the introduction of automated systems at theof a drug: various stages of the screening programme, an incredible

leap in the data acquisition process is seen. Attention isnow focusing on data storage in structured databases, and

• the application of specific inhibitors to a pool of human on data retrieval together with appropriate analysis.microsomes gives information on the nature and rela- One can see in Fig. 10 that data are collected at alltive importance of the enzymes involved; stages of the screening process and that the type of

• the use of heterologous drug metabolizing enzymes information collected can vary according to the modelexpressing a single CYP, allows to assess the nature of used, from primary parameters such as a single data pointthe isozymes involved as well as their relative contribu- allowing the evaluation of the rate of metabolism within ation to the various metabolic routes; series of compounds, to secondary parameters combining a

• the correlation with specific CYP activities obtained number of primary parameters as bioavailability or drug–with a bank of human microsomes previously character- drug interaction prediction, up to collections of in vivoized with specific CYP substrates. data from n-in-one dosing experiments in animals.

At a very early step when dealing with large series ofMost of these techniques are potentially applicable to compounds, the integration of warnings is a simple way to

HTS, but even their systematic use to screen for a use the data. This has already been applied to solubilityparticular enzyme at the HTS level would not predict (Lipinski et al., 1997) and can easily be used to sortpharmacokinetic variability. Indeed, most of the major molecules in large series. Another classical way to analyze

Fig. 10. Flow of metabolic data information during drug screening and development.


these data is to use quantitative methods already applied in preparations containing single CYP enzymes. Togetherthe QSAR field in a stepwise multiple quantitative struc- with scaling factors based on the levels of individual CYPture–pharmacokinetic relationships (QSPR) (Mayer and enzymes present in a bank of microsomes, expected ratesvan de Waterbeemd, 1985). These data can then be of metabolism are calculated with Michaelis–Mentencombined with pharmacological activities and the infor- equations and compared with actual rates within the samemation used within and across chemical series in a more microsomal bank in order to validate these parameters.predictive way. After the incorporation of kinetic parameters and scaling

Ideally in drug metabolism, all the individual data factors into the PBPK model, prediction of ranges ofcollected at each stage should enter an appropriate model plasma levels can be made, representing the interindividualhelping to rebuild the biological process and hence validate variability of the bank used. Fig. 11 shows an example ofthe approach by appropriate in vivo experiments performed the predicted range of plasmatic concentrations comparedwith a certain number of compounds. This has already with the actual range found in clinical studies. This modelbeen shown with bioavailability predictions from K and allows the quantitative description of the potential impactm

V parameters. However, we can go further and one way of metabolism-based interindividual variability on in vivom

of illustrating these approaches is the prediction of drug pharmacokinetic parameters. This model can be furtherplasma levels using a physiologically based phar- completed by adding IC or K values, affording an even50 i

macokinetic (PBPK) model containing only in vitro data. better prediction of the risk of drug–drug interactions. ThisSuch models (Nestorov et al., 1997) use actual physiologi- tool can also be simplified by using apparent K and Vm max

cal parameters such as tissue volumes and weights, blood values for the total drug disappearance helping us again toflow rates, as well as drug-specific parameters such as predict drug plasma levels in humans.partition coefficients, clearances, permeabilities of tissues Discovery chemists are asked to focus more and moremembranes and protein binding. on the survival of drug candidates up to the first clinical

The example presented by Bogaards et al. (in press) is trials. Screening packages have therefore been developedbased on the determination of kinetic parameters (K and allowing these physiological models to be applied tom

V ) for the main metabolic pathways using enzyme selected compounds. But the use of these interpretationmax

Fig. 11. Prediction of in vivo drug plasma concentrations using PB/PK modelling of in vitro data.


Druekes, P., Schinzel, R., Palm, D., 1995. Photometric microtiter assay oftechniques does not necessarily fit with the objectives ofinorganic phosphate in the presence of acid-labile organic phosphates.HTS where absolute measurements are generally lessAnal. Biochem. 230, 173–177.

important than relative ones. Ferrini, J.B., Pichard, L., Domergue, J., Maurel, P., 1997. Long-termprimary cultures of adult human hepatocytes. Chem.-Biol. Interact.107, 31–45.

Johansson, C.A.S., Luthman, K., Artursson, P., 2000. A small scale shake6. Conclusionflask method for determination of intrinsic drug solubility. In: TheSecond Lipophilicity Symposium, University of Lausanne, March

Progress has been very rapid in drug metabolism during 2000, Abstract O-C4.the last 5 years, providing a large range of metabolic tools Kataoka, M., 2000. In vitro screening system for oral absorption forwhich are well adapted to screening and allow a rapid poorly water-soluble drugs. In: Millennial World Congress of Pharma-

ceutical Sciences, San Francisco, April 2000, Abstract 1-4034, 108.understanding of the in vivo fate of a potential candidate,¨ ´Langouet, S., Maheo, K., Berthou, F., Morel, F., Lagadic-Gossman, D.,well before clinical studies. In one sense, things have

Glaise, D., Coles, B., Ketterer, B., Guillouzo, A., 1997. Effects ofmoved too fast as most of the automated in vitro tests are administration of the chemoprotective agent oltipraz on CYP1A Z andalready in place but we have not yet fully mastered the CYP2B in rat liver and rat hepatocytes in culture. Carcinogenesis 18,handling of the enormous amounts of data thus produced. 1343–1349.

Lipinski, C.A., Lombardo, F., Dominy, B.W., Feeney, P.J., 1997. Ex-Rapid metabolism assessment in the discovery phase is aperimental and computational approaches to estimate solubility andcontinuing process. The simpler the metabolic tests, thepermeability in drug discovery and development settings. Adv. Drug

more adapted they are to classifying large series of Deliv. Rev. 23, 3–25.compounds (HTS) and producing warnings. Inversely, the Mayer, J.M., van de Waterbeemd, H., 1985. Development of quantitativemore sophisticated tools and/or models give a better structure-pharmacokinetic relationships. Environ. Health Perspect. 61,

295–306.understanding to how to use more simple HTS data andNestorov, I.A., Aarons, L., Rowland, M., 1997. Physiologically basedallow the rapid development of new tests. This bi-direc-

pharmacokinetic modelling of a homologous series of barbiturates intional continuum is the sole way to combine both valida- the rat: a sensitivity analysis. J. Pharmacokinet. Biopharm. 25, 413–tion and use of new tools whilst the domain is evolving 447.rapidly; and yet we are already facing another possible Ogg, M.S., Williams, J.M., Tarbit, M., Goldfarb, P.S., Gray, T.J.B.,

Gibson, G.G., 1999. A reporter gene assay to assess the molecularleap with the advent of genomics.mechanisms of xenobiotic-dependent induction of the human CYP3A4The difficulty is and will remain to use the right toolsgene in vitro. Xenobiotica 29, 269–279.

for the right problem and the right models to interpret data, Rodrigues, D., 1997. Preclinical drug metabolism in the age of high-getting as close as possible to our target, the human throughput screening: an industrial perspective. Pharm. Res. 14, 1504–patient. 1510.

Schmiedlin-Ren, P., Thummel, K.E., Fisher, J.M., Paine, M.F., Lown,K.S., Watkins, P.B., 1997. Expression of enzymatically active CYP3A4by Caco-2 cells grown on extracellular matrix-coated permeable

References supports in the presence of 1a,25-dihydroxyvitamin D3. Mol. Phar-macol. 51, 741–754.

Avdeef, A., 2000. High-throughput measurement of solubility profiles. In: Schurgers, N., Bijdendijk, J., Tukker, J.J., Crommelin, D.J.A., 1986.The Second Lipophilicity Symposium, University of Lausanne, March Comparison of four experimental techniques for studying drug absorp-2000, Abstract L-12. tion kinetics in the anesthetized rat in situ. J. Pharm. Sci. 75, 117–119.

Berman, J., Halm, K., Adkison, K., Shaffer, J., 1997. Simultaneous Sharom, F.J., Yu, X., Lu, P., Liu, R., Chu, J.W.K., Szabo, K., Muller, M.,pharmacokinetic screening of a mixture of compounds in the dog Hose, C.D., Monks, A., Varadi, A., Seprodi, J., Sarkadi, B., 1999.Using API LC/MS/MS analysis for increased throughput. J. Med. Interaction of the P-glycoprotein multidrug transporter (MDR1) withChem. 40, 827–829. high affinity peptide chemosensitizers in isolated membranes, reconsti-

Bogaards, J.J.P., Hissink, A.M., Briggs, M., Weaver, R., Jochemsen, R., tuted systems, and intact cells. Biochem. Pharmacol. 58, 571–586.Jackson, P., Bertrand, M., van Bladeren, P.J., in press. Prediction of Simpson, H., Barthemy, A., Buhrman, D., Burton, R., Newton, J., Kealy,drug plasma levels in vivo from individual enzyme kinetic data and M., Wu, D., 1998. High throughput liquid chromatography/massphysiologically-based pharmacokinetic (PBPK) modeling. Eur. J. spectrometry bioanalysis using 96-well disk solid phase extractionPharm. Sci. plate for the sample preparation. Rapid Commun. Mass Spectrom. 12,

Crespi, C.L., Penman, B.W., Hu, M., 1996. Development of Caco-2 cells 75–82.expressing high levels of cDNA-derived cytochrome CYP3A4. Pharm. Testa, B., Kier, L.B., Carrupt, P.A., 1997. A systems approach toRes. 13, 1635–1641. molecular structure, intermolecular recognition, and emergence-dis-

Crespi, C.L., Miller, V.P., Penman, B.W., 1997. Microtiter plate assays for solvence in medicinal research. Med. Res. Rev. 17, 303–326.inhibition of human, drug-metabolizing cytochromes CYP. Anal. van Breemen, R.B., Nikolic, D., Bolton, J.L., 1998. Metabolic screeningBiochem. 248, 188–190. using on-line ultrafiltration mass spectrometry. Drug Metab. Dispos.

de Waziers, I., Bouguet, J., Beaune, P.H., Gonzalez, F.J., Ketterer, B., 26, 85–90.Barouki, R., 1992. Effects of ethanol, dexamethasone and RU 486 onexpression of cytochromes CYP 2B, 2E, 3A and glutathione transfer-ase in a rat hepatoma cell line (Fao). Pharmacogenetics 2, 12–18.

Documents

Rapid assessment of drug metabolism in the drug discovery process