Chem. Res. Toxicol. 1992,5, 211-219
211
A Predictive Model for Substrates of Cytochrome P450-Debrisoquine (2D6) Luc Koymans,’ Nico P. E. Vermeulen,*J Saskia A. B. E. van Acker,* Johan M. te Koppele,* Jos J. P. Heykants,§ Karel Lavrijsen,g Willem Meuldermans,! and Gabrielle M. Donne-Op den Kelders Divisions of Drug Design and Molecular Toxicology, Department of Pharmacochemistry, Faculty of Chemistry, Free University, De Boelelaan 1083, 1081 HV Amsterdam, The Netherlands, and Department of Drug Metabolism and Pharmacokinetics, Janssen Research Foundation, B-2340 Beerse, Belgium Received August 14,1991
Molecular modeling techniques were used to derive a predictive model for substrates of cytochrome P450 2D6, an isozyme known to metabolize only compounds with one or more basic nitrogen atoms. Sixteen substrates, accounting for 23 metabolic reactions, with a distance of either 5 A (y5-%L substrates”, e.g., debrisoquine) or 7 A (“7-A substrates”, e.g., dextromethorphan) between oxidation site and basic nitrogen atom were fitted into one model by postulating an interaction of the basic nitrogen atom with a negatively charged carboxylate group on the protein. This acidic residue anchors and neutralizes the positively charged basic nitrogen atom of the substrates. In case of “5-A substrates” this interaction probabl occurs with the carboxylic oxygen atom nearest to the oxidation site, whereas in the case of “7- substrates” this interaction takes place a t the other oxygen atom. Furthermore, all substrates exhibit a coplanar conformation near the oxidation site and have negative molecular electrostatic potentials (MEPs) in a part of this planar domain approximately 3 A away from the oxidation site. No common features were found in the neighbourhood of the basic nitrogen atom of the substrates studied so that this region of the active site can accommodate a variety of N-substituents. Therefore, the substrate specificity of P450 2D6 most likely is determined by the distance between oxidation site and basic nitrogen atom, by steric constraints near the oxidation site, and by the degree of complementarity between the MEPs of substrate and protein in the planar region adjacent to the oxidation site. The predictive value of the model was evaluated by investigating the P450 2D6 mediated metabolism of four new compounds comprising a t least 14 oxidative metabolic routes. According to our model, 4 of the metabolic routes were predicted to be mediated by P450 2D6, whereas 10 were not. The involvement of P450 2D6 in these 14 metabolic reactions was investigated in man in vivo and/or in vitro. From these experimental results it appeared that 3 of the 4 predicted metabolic routes were mediated by P450 2D6 and 11 were not, closely matching the predictions from the model. Thus, the computer-assisted predictions seem to correlate well with the experimental results, and hence the presented model may be useful in identifying metabolic pathways that might be subject to the “debrisoquine/sparteine” type of polymorphism in a very early stage of the development of drugs.
K
I ntroductlon Cytochromes P450 (P450) are enzymes involved in the oxidative metabolism of a wide variety of endogenous and exogenous compounds. Multiple isozymes of P450 have been shown to exist. Each isozyme exhibits its own, usually overlapping, substrate specificity. In recent years, genetic polymorphism of drug oxidations related to specific isozymes of cytochromes P450 has been described (1,2).The isozyme P450 2D6l is responsible for the human defect in drug oxidation known as the “debrisoquine/sparteine” polymorphism, called after its two independently discovered marker substrates debrisoquine and sparteine (4,5). The clinical relevance of this polymorphism has been established through epidemiological and familial studies. It was shown that in extensive metabolizers (EMs)~debrisoquine and related compounds are metabolized at a normal rate, whereas in poor metabolizers (PMs) the ‘This work was supported by the foundation SURF from the National Fund Supercomputers. * To whom correspondence should be addressed. *Free University. 1Janssen Research Foundation.
metabolism for these compounds is defective. PMs comprise about 1-30% of various ethnic populations and are homozygous for the recessive gene (6). They may develop toxic plasma levels of the parent compound or they may be unable to bioactivate a parent drug such as encainide to its therapeutically active metabolites (I). Recently, the primary gene defect at the cytochrome P450 CYP2D locus responsible for the metabolic defect has been identified (7). The more than 20 compounds for which the PMs have shown impaired oxidation comprise antihypertensive drugs, @-adrenergicblocking agents, and antidepressants (Figure 1). A remarkable characteristic of substrates of P450 2D6 is that they all possess one or more basic nitrogen atoms. In contrast to other P450 isozymes cytochrome P450 2D6 only accepts basic compounds as substrates. Recently, two models have been developed indewhich describe common characteristics pendently (8,9), Systematic nomenclature according to ref 3. Abbreviations: EMS, extensive metabolizers;PMs, poor metabolizers; MNDO, modified neglect of differential overlap;GAMESS,general atomic and molecular electronic structure system; MEP, molecular electrostatic potential; MR, metabolic ratio; AUC, area under the curve.
0893-228x/92/2705-0211$03.00/00 1992 American Chemical Society
212 Chem. Res. Toxicol., Vol. 5, No. 2, 1992
Debrisoquine
" Guanoxan
I NH2
Koymam et al.
T
II
NH
Desmethylimipramine
Mexiletine N,C-
2
Perhexiline
Sparteine
u --. .
Encainide
t'
b,O*
-\ NH2
/ Phenformine
I:y"
NH2
\
Dextromethorphan NH2
4-methoxyamphetamine
NH
OCH3
t
OCH,
t
Methoxyphrnaminc
Figure 1. Chemical structures of 16 compounds with deficient metabolism in poor metabolizers of debrisoquine. The major oxidation sites thought to be involved are indicated with arrows. Asterisks indicate a chiral center.
of substrates of P450 2D6. Assuming a common oxidation site and a similar location for the basic nitrogen atom on the enzyme, space-filling substrate models were used to search for common structural features defining these substrates. However, the two models have contradictive elements: in the model of Wolff et al. (8) the distance between the basic nitrogen atom of the compounds and the oxidation site was suggested to be 5 A (the so-called 5-A substrates), whereas Meyer et al. (9) proposed this distance to be 7 A (7-A substrates). Knowledge of the common structural features of substrates of P450 2D6 might be useful in determining whether metabolism of a drug is controlled by this specific type of polymorphism. Quantum chemical calculations and molecular modeling techniques already have been used with success in elucidating specific features of substrates of cytochromes P450, and implications for the active site structure of cytochromes P450 have been discussed (10-13). In the present study a model has been derived in which both 5-A and 7-A substrates can be accommodated and which describes the main characteristics of P450 2D6 Substrates. The predictive value of the model was evaluated by predicting the involvement of P450 2D6 in the metabolism of four new compounds which were not used to derive the model. Subsequently, the metabolism of these compounds was studied in PMs and EMS and compared to the predicted role of cytochrome P450 2D6 in the at least 14 oxidative pathways involved.
Methods Molecular Modeling. The experimental data used to construct the theoretical model for substrates of P450 2D6 were taken exclusively from human studies (in vivo, with liver microsomal fractions, or with purified isozymes). These data constitute the 23 oxidative reactions of 16 known substrates of P450 2D6 presented in Figure 1 and Table I.
The structural data used for the initial geometries of the substrates in the geometry optimization studies were taken from the Cambridge Structural Database (32) and were assumed to be low-energy conformations. When necessary, changes were made using the routines of the molecular modeling package ChemX (33) implemented on a rVAX-11. The initial geometries of the substrates were fully optimized using the semiempirical MNDO method (34)(MOPAC version 4.0). Since conjugation effects are underestimated by the MNDO method (35),conjugated systems like guanidine were fully optimized with the ST03G minimal basis set (36)using the ab initio program package GAMESS (37). This was done for the NHC(NH,)=NR moiety of phenformine and guanoxan and the NHC(=O)C,$I40CH3moiety of encainide. The optimized ST03G geometries of these parts were pasted on the remainder of the molecules using the routines of ChemX. The torsional angles of these moieties were restrained in the subsequent MNDO optimalizations. The series of compounds studied contains two fairly rigid molecules, debrisoquine and dextromethorphan, in which the distance between the oxidative site and basic nitrogen atom is fixed: 5 and 7 A, respectively. Furthermore, debrisoquine is known to be stereoselectively hydroxylated in man at the 4position to (S)-(+)-4-hydroxydebrisoquine(38),which was accounted for in the strategy followed: the prochiral oxidation site of debrisoquine was fitted onto the oxidation site of dextromethorphan while keeping the conformation near the oxidation site coplanar. In a subsequent flexible fitting procedure the basic nitrogen atom of debrisoquine was fitted as close as possible to the basic nitrogen atom of dextromethorphan while keeping the oxidation site fiied and the conformation near this site coplanar. The fully MNDO-optimized structures of the remaining 14 compounds were fitted either onto debrisoquine or onto dextromethorphan; the basic nitrogen atoms and oxidation sites were matched. After fitting some of these molecules, it turned out to be possible to create a planar region adjacent to the oxidation site. This feature was used as an additional restraint. During the subsequent flexible fitting procedures only van der Waals interactions were taken into account. Only those fits were accepted in which the MNDO energy of the fitted substrate was no more than 10 kcaI/moI above the minimum energy obtained with
Predictive Model of P450 206
Chem. Res. Toxicol., Vol. 5, No. 2, 1992 213
Table I. Drugs with Impaired Metabolic Oxidation Associated with the Debrisoquine/Sparteine Type of Cytochrome P460 2D6 Polymorphism drug defective metabolic reaction preferred isomer fit energy" minimum potentialb (R)-(+IC 4.8 -14 propranolol 4-hydroxylationC noned 8.9 -14 metoprolol a-hydroxylationdve (R)-(+Id 9.8 -14 O-demethylationd*' timolol morpholino ring oxidation' ndu 6.0 -20 (R)-(+)' 8.3 -33 bufuralol 1'-hydroxylationf (S)-(-P 7.8 -35 4-hydroxylation8 (S)-(-)8 8.4 -35 6-hydroxylation8 5.3 -9 debrisoquine 4-hydroxylationh 9.1 -16 guanoxan 6-hydroxylationf 8.4 -16 7-hydroxylation' 7.8 -18 amiflamine N-demethylatiod 6.7 -14 desmethylimipramine 2-hydroxylationk 5.6 -14 encainide O-demethylation' perhexilline ring 4-hydroxylationm 8.7 -10 8.7 -15 mexiletine 2/-hydroxylation" 8.5 -16 4/-hydroxylation" 9.0 -18 sparteine Az-dehydrogenationosp 9.0 -18 A6-dehydrogenation0G 0.4 -12 4-methoxyamphetamine O-demethylation* phenformine 4-hydroxylation' 9.4 -14 0.0 -10 dextrometorphan O-demethylations 5.0 -11 methoxyphenamine O-demethylatiod 5.1 -12 5-hydroxylation' Energy difference between fitted conformation and minimum energy obtained with MNDO (kcal/mol). *Minimum molecular electrostatic potential (kcal/mol) in the region next to the oxidation site (see Figure 5). Ward et al. (14). dotton et al. (15). 'Lennard et al. (16). fDayer et al. (17). gDayer et al. (18). "Ritchie et al. (19). 'Idle and Smith (20). jAlvan et al. (21). kVon Bahr et al. (22). 'McAllister et al. (23). "Gould et al. (24). "Brolv et al. (15). "Eichelbaum et al. (26). PGuengerich (27). '?Kitchen et al. (28). 'Eichelbaum (29). "ayer et &. (30). 'Roy et al. ( 3 j ) . Und,hot determined. MNDO. Molecular electrostatic potentials (MEPs) were calculated on a molecular surface at a distance of 1.7 times the van der Waals radius using Mulliken populations generated by MNDO. I t is fully recognized that the above-described methods are rather simple and not fully objective. Therefore, the strategy followed is only acceptable when it is validated by experimental evidence. Experimental Procedures. (A) In Vitro Metabolism of Risperidone. Preparation of human hepatocytes and incubation of the cells in suspension culture were performed as described elsewhere (39). The hepatocytes (IO6 cells/mL) were incubated with [14C]risperidone(labeled in the 6- and 10-positions of the 6,7,8,9-tetrahydro-2-methyl-4H-pyrido[ 1,2-a]pyrimidin-4-one group, specific activity 44.44 mCi/mmol, final risperidone concentration 1 pM) for 210 min. In order to investigate the role of cytochrome P450 2D6 in the metabolism of risperidone, quinidine (5 pM, final concentration), a potent inhibitor of cytochrome P450 2D6 (40),was added. At the end of the incubations, the hepatocyte suspensions were frozen in dry ice until HPLC analysis. HPLC analysis was performed on stainless steel columns packed with Hypersil c-18 (5 pM, Shandon) bound phase. A mixture of authentic reference compounds, postulated as possible metabolites, was cochromatographed together with the samples. A linear gradient from 100% 0.1 M ammonium acetate containing 0.2% diethylamine, pH 6.0 (solvent system A), to 50% of solvent system A and 50% of 1.0 M ammonium acetate containing 2% diethylamine, pH &O/methanol/acetonitrile(101080 v/v) (solvent system B) over 60 min was applied to the column at a flow rate of 1mL/min. The latter solvent composition was held for 5 min, and then a short gradient to 100% of solvent system B was applied over 5 min. UV detection of the eluates emerging from the HPLC columns was performed at 280 nm by a Varichrom (Varian) spectrophotometric detector, with on-line radioactivity detection by a Berthold Radioactivity monitor LB 504 system. (B) In Vivo Metabolism. Astemizole was given at 10 mg/day for 6 weeks to 12 healthy (6 male, 6 female) subjecta; two of them were identified as PMs of debrisoquine [metabolic ratio (MR) of debrisoquine: 4-hydroxydebrisoquine in 0-8-h urine > 121. Plasma levels at steady state of astemizole and of its O-demethylated metabolite, desmethylastemizole, were measured by validated radioimmunoassay procedures (41). Areas under the plasma concentration-time curves (AUCs) to the last time point
were calculated by trapezoidal summation and extrapolated to infinity by addition of the residual AUC obtained as the plasma concentration at time t divided by 6,the slope of the log linear plasma concentration-time curve. Three male healthy subjects with debrisoquine MRs of 0.98, 6.5, and 20, respectively, received a single oral dose of 1 mg of [14C]risperidone(specific activity 24.0 mCi/mmol). Urine and feces were collected up to 168 h after dosing. Relative amounts of unchanged risperidone and of its major metabolites in urine samples were determined by radio-HPLC analysis as described above. Metabolic ratios of unchanged risperidone to 9hydroxyrisperidone and to 7-hydroxyrisperidone were determined in the 0-48-h urine.
Results and Discussion The therapeutic and toxic responses to a wide variety of drugs primarily are determined by the hepatic metabolism by the cytochrome P450 enzyme system. Genetic polymorphisms in human cytochrome P450,resulting in markedly altered biotransformation of drugs,can therefore be of considerable clinical relevance. A reliable theoretical predictive model might be very useful in screening drugs for a possible major involvement of P450 2D6 in the metabolism of a drug in a very early stage of its development. The airn of the present study is to derive and validate such a predictive "in computro" model. The "Bidentate Carboxylate Hypothesis". As yet, two models have been postulated describing common structural elements of cytochrome P450 2D6 substrates. Studies with space-filling substrate models led Wolff et al. (8)to propose a distance of about 5 A between the site of oxidation and the basic nitrogen atom as a common structural feature of substrates of P450 2D6. Furthermore, all substrates should have a hydrophobic domain near the site of oxidation, and a carboxylate group on the protein was suggested to neutralize the charge of a basic nitrogen atom of the substrates (8). In contrast, Meyer et al. (9) proposed a distance of 7 8, between the site of oxidation and the basic nitrogen atom. In addition, they concluded that the substrates of cytochrome P450 2D6 should have nearly coplanar aromatic or aliphatic rings adjacent to the
214 Chem. Res. Toxicol., Vol. 5, No. 2,1992
Koymans et al.
-
-
BASIC NITROCEN
CiI
II
e--
Figure 2. (Upper panel) Conformation of the templates used in this study, dextromethorphan (“7-8Lsubstrate”) and debrisoquine (“5-A substrate”). (Lower panel) The template molecules debrisoquine and dextromethorphan are oriented such that the oxidation sites are fitted while keeping the region near this site coplanar. The basic nitrogen atoms of the two compounds are separated by 2.5 8L and can interact with either one of the carboxylic oxygen atoms of an aspartic or glutamic acid.
metabolic oxidation site. Surprisingly, the marker substrates debrisoquine and sparteine could not be fitted into the “7-A model” of Meyer (9),whereas in the “5-Amodel” of Wolff et al. (8) the rigid substrate dextromethorphan could not be accommodated. The distance between the basic nitrogen atom and the site of oxdiation in debrisoquine is about 5 A, whereas in the rigid substrate dextzomethorphanthis distance is about 7 A (Figure 2A). When the oxidation sites of debrisoquine and dextromethorphan are fitted, while keeping the region close to these sites coplanar, the basic nitrogen atoms of the two compounds are separated by a distance of about 2.5 A (Figure 2B). Meyer et al. (9) took this apparent discrepancy into account and proposed different binding sites for these substrates at the enzyme. However, to our opinion, it is possible to fit both debrisoquine and dex-
tromethorphan into one binding site by postulating an interaction between a basic nitrogen atom of these two compounds and a negatively charged carboxylate group on the protein (Figure 2B); in this way the negatively charged carboxylate neutralizes the positively charged nitrogen atom. The presence of a negatively charged residue at the active site of P450 2D6 is in agreement with a recently speculated role of P450 2D1,the rat analogue of human P450 2D6, as a piperazine acceptor site. This neuronal P450 2D1 is suggested to play a role in the catabolism and processing of neurotransmitters subsequent to their reuptake (42). Our model suggests that debrisoquine and dextromethorphan bind to the same site on P450 2D6 but interact with a different oxygen atom of the same carboxylate group. The distance between the two oxygen atoms in a carboxylate is about 2.2 A, which is in
Chem. Res. Toxicol., Vol.5, No. 2, 1992 215
Predictive Model of P450 206
f
Figure 3. Active site model of P450 2D6 when all 16 substrates (23 metabolic reactions) used in this study are fitted to the template molecules debrisoquine and dextromethorphan.
--
B M C NITROCEN OXIDRTION U T E c1sI
, I
Figure 4. The substrates debrisoquine and dextromethorphan as in Figure 2. The red dot surface represents the region in the flat part of the molecules where all substrates exhibit negative molecular electrostatic potentials induced by a .Ir-system, an 0-atom, or an N-atom.
close agreement with the above-mentioned distance of 2.5
A between the basic nitrogen atoms of debrisoquine and
dextromethorphan. Therefore, we propose that in the case of 5-A substrates the basic nitrogen-carboxylate interaction occurs at the carboxylic oxygen atom nearest to the oxidation site, whereas the interaction of 7-A substrates proceeds at the carboxylic oxygen atom furthest away from the oxidation site (Figure 2B). Modeling Procedures. The 16 substrates listed in Table I were unambiguously reported in literature to be oxidized by human cytochrome P450 2D6. On the basis of the assumption that a protein carboxylate group of aspartic or glutamic acid interacts with both the basic
nitrogen atom of debrisoquine and that of dextromethorphan, all compounds were fitted either onto the rigid template structure of debrisoquine or onto the structure of dextromethorphan. The substrate molecules were fitted in such a way that the oxidation site and the basic N-atom coincided with correspondingsites on one of the templates. Furthermore, the part of the substrates adjacent to the oxidation site was kept coplanar with the corresponding part of the templates. As a result, all fitted substrates exhibited a planar conformation near the site of oxidation with a distance of either 5 or 7 A between oxidation site and basic nitrogen atom. The energies of the fitted substrates were no more than 10 kcal/mol above the minimum
216
Koymans et al.
Chem. Res. Toxicol., Vol. 5, No. 2, 1992
1
T
fYF
Figure 5. Chemical structures of alfentanil (I), astemizole (II), risperidone (III), and nebivolol (IV). The arrow8 indicate the sites of oxidative metabolic attack. These metabolic reactions were used to evaluate the predictive value of the model but were not used to derive the model. energy conformations obtained with MNDO (Table I). It is recognized that the modeling procedures are limited. For example, conformations close to the crystal structure are not necessarily “binding”conformations. However, the great divergence in chemical structures of the substrates used (flexible ring structures, aliphatic chains, conjugated structures, the incomplete parameter lists in molecular mechanics programs) makes it difficult to perform an absolutely reliable conformational analysis. Several different techniques would have to be used to perform a reliable conformational analysis of all the compounds. However, a comparison of results obtained from different computational techniques has many disadvantages. Therefore, the rather simple strategy was followed as outlined in Methods. The correctness of this strategy has to be validated by a comparison between predictions and experimental results. Overall Shape of the Model. Figure 3 depicts the overall shape of our model after fitting all substrates according to the above-defined criteria. The planar domain near the oxidation site has dimensions of approximately 6.5 X 7.5 x 2.5 A3. In the region of the basic nitrogen atoms steric hindrance was not found to be an important factor. Stereoselectivity of P450 2D6. It is known that P450 2D6 metabolizes several @-blockersin a stereoselective manner (Table I). The substrate selectivity for bufuralol 1’-hydroxylation in man has been studied extensively in vivo (43),in liver microsomal fractions (17,44),and with 1’purified enzymes (45). As judged from the V,, hydroxylation is only about 2-fold more favorable with (R)-(+)-bufuralolthan with the (S)-(-)-isomer. The ap-
parent Michaelis-Menten constant K, was not found to be significantly different for the two enantiomers. This suggests that the binding affinities of (R)-and (8-bufuralol to P450 2D6 are not different and that the site of oxygen insertion rather than binding of the substrate is under stereochemical control (17, 18, 44, 45). In the current model the small stereoselectivityobserved with @-blockers could not be explained since the small difference in V,, of (R)-and (5’)-bufuralol corresponds to a difference in binding energy of only f0.5 kcal/mol (compare to energy differences in Table I). Molecular Electrostatic Potentials (MEPs). Evaluation of the MEP distributions of all the substrates investigated revealed a common feature for all compounds. Negative MEP values (minimum potentials
