![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Discovery DMPK & Bioanalytical Chemistry, AstraZeneca R&D Mölndal, Mölndal, Sweden (B.B., C.M.M.); Department of Chemistry, Medicinal Chemistry, University of Gothenburg, Gothenburg, Sweden (B.B.); African Institute of Biomedical Science & Technology, Harare, Zimbabwe (C.M.M.); Department of Chemistry, Laboratory for Chemometrics and Chemoinformatics, University of Perugia, Perugia, Italy (Y.A.); Lead Molecular Design, Sociedad Limitada, San Cugat del Vallés, Spain (I.Z.); and Institution Municipal d'Investigació Medica, Universitat Pompeu Fabra, Barcelona, Spain (I.Z.)
(Received May 15, 2008; Accepted August 22, 2008)
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
). In drug discovery research it is of great interest to identify whether a compound is a substrate or not, and considerable effort has been put into the development of models to identify substrates and/or inhibitors of CYP2D6. Several pharmacophore models have been published identifying common features in CYP2D6 substrates and inhibitors (Koymans et al., 1992; de Groot et al., 1997, 1999; Lewis et al., 1997). The pharmacophore comprises basic nitrogen atom 5-7 or 10 Å from the site of oxidation, a flat hydrophobic region near the site of oxidation and a negative molecular electrostatic potential coplanar with this hydrophobic region (see examples in Fig. 1). The intramolecular distances of 5 and 7 Å were combined by Koymans et al. (1992) in a model assuming a carboxylate group as an anchor point in the protein with one of the oxygen atoms interacting with the "5-Å substrates" and the other with the "7-Å substrates." Since several substrates with approximately 10 Å between the nitrogen atom and the site of oxidation were also known, refined models were constructed including this intramolecular distance (Lewis et al., 1997). Even though the early pharmacophore models are rather crude, they have been successful in predicting the influence of CYP2D6 in the metabolism of xenobiotics as well as predicting possible sites of metabolism [e.g., 75% predictability (de Groot et al., 1999)].
|
; De Rienzo et al., 2000) and a crystal structure of ligand-free CYP2D6 (Rowland et al., 2006) revealed two acidic amino acids, Glu216 and Asp301, and two phenylalanine residues, Phe120 and Phe483, in the active site cavity. This was compatible with the previous pharmacophore models identifying the basic nitrogen and a planar hydrophobic area as characteristic features in CYP2D6 substrates. Site-directed mutagenesis studies also confirmed the importance of the acidic amino acids for substrate specificity and catalytic activity (Mackman et al., 1996; Guengerich et al., 2003; Paine et al., 2003). Similar experiments also suggest that the hydrophobic and steric characters of the phenylalanine residues influence the regiospecific metabolism in a substrate-dependent manner (Flanagan et al., 2004; Lussenburg et al., 2005). Despite the well characterized CYP2D6 substrate/inhibitor pharmacophores, prediction of the regiospecificity of CYP2D6-mediated metabolism is a complex matter with many different interactions to account for. However, in the present study, the main focus has been on the electrostatic interactions and the influence of the acidic amino acids.
Considering the common structure of CYP2D6 substrates with a basic nitrogen 5-7 Å from the site of oxidation interacting with the acidic amino acid Glu216, it would be unlikely that N-dealkylation reactions occur. The main reason is that the electrostatic interaction between the basic nitrogen in the substrate and the carboxylic acid group of Glu216 on top of the active site cavity would favor an orientation of the substrate with the nitrogen directed away from the heme. Nevertheless, N-dealkylation does occur in CYP2D6 but is far less common than in, for example, CYP3A4 (Coutts et al., 1994). In a previous study, the influence and importance of the positively charged nitrogen on the orientation of the substrates in the active site cavity was explored (Upthagrove and Nelson, 2001). It was shown that addition of substituents that increase the basicity of the compound also increased the affinity for the enzyme.
In the present study, this was investigated by comparing the N-dealkylation of dextromethorphan (DXM) at different pH values to see how different ratios of unprotonated/protonated forms of the ligand affect the formation of the N-dealkylated metabolite. Besides the effect of the degree of ionization, which may be affected by the microenvironment in the enzyme active site cavity, another working hypothesis was that N-dealkylation in CYP2D6 occurs, for example, when aromatic hydroxylation or O-demethylation reactions are not possible. That is, when the preferred site of metabolism is prevented from participating in an oxidative reaction, the substrate will be N-dealkylated to a more significant extent. To study this hypothesis, the literature was searched in order to select CYP2D6 substrates that were reported to be N-dealkylated and substrates that were not. The preferred site of metabolism for these compounds was then identified by aligning the compounds to CYP2D6 probe substrates, using the program FLAP (Baroni et al., 2007). The site of metabolism (SOM) prediction program MetaSite (Zamora et al., 2003) was also used and compared with the results from FLAP. In addition, some compounds were selected for experimental determinations of the major metabolites formed, the rate of formation of N-dealkylated product, as well as the inhibition potential in recombinant CYP2D6.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Computer Hardware and Software. The calculations were performed in a Linux (Ogdensburg, NY) environment on a Hewlett Packard (Palo Alto, CA) xw6200 Workstation with a Pentium 4 processor (Intel, Santa Clara, CA). The software used was GRID version 22.2.2, MetaSite version 2.7.5, MOKA version 1.0, Mizer version 1.0, and a command line version of FLAP beta from Molecular Discovery (http://www.moldiscovery.com) and ISIS draw version 2.4 (MDL Information Systems Inc., http://www.mdl.com).
Data Set. From a literature search on CYP2D6 substrate, compounds were chosen for which the published data clearly showed which metabolites were formed via CYP2D6 catalysis. Since the basic nitrogen was of interest as a pharmacophoric feature in this study, compounds with a predicted pKa below 8 were filtered away, ending up with a set of CYP2D6 substrates being protonated at physiological pH. To get a comparable environment around the basic nitrogen and thereby simplify the analysis, an additional filtering of the data set was done to only include methyl- and ethylamines. The final data set of 43 compounds was divided into two groups: 1) substrates with N-dealkylation as a major metabolic pathway in CYP2D6 and 2) substrates undergoing other metabolic reactions (e.g., aromatic hydroxylation, O-dealkylation) as primary routes in CYP2D6.
From this data set, 20 compounds were selected for experimental determination of metabolism and inhibition potential using recombinant CYP2D6 enzyme. As mentioned above, the focus of this study was on major metabolites formed by CYP2D6. For data on compounds found in the literature the major metabolite formed by CYP2D6 was rationalized from the available references. For the compounds classified as not being N-dealkylated by this enzyme some were still reported to be N-dealkylated but only to a minor extent and were therefore not considered of importance in this study. For the experimental data set, metabolites with an abundance <20% of all metabolites formed were considered minor and >20% were considered to be major metabolites. This estimation was based on the LC-MS response areas. For these estimations it was assumed that the ionization efficiencies for the metabolites were similar to that of the parent drug, which is not always the case. However, since no authentic standards were available for the metabolites, the LC-MS response areas were used for a rough quantification and distribution of metabolites.
For the computational part, all structures were drawn in ISIS and exported as SDF files. Because the basic CYP2D6 substrates would be protonated at physiological pH and the positive charge is suggested to be of importance for affinity of the substrate to the enzyme (Upthagrove and Nelson, 2001), the compounds were protonated prior to further analysis. This was done with the program MOKA (Milletti et al., 2007), which protonates the molecules at a selected pH based on predicted pKa values. The 2D-3D conversion was made with the program Mizer, generating one conformer minimized with the MM3 force field.
pH-Dependent N-Dealkylation of DXM. DXM was incubated with recombinant CYP2D6 and buffers at different pH values. Since it is difficult to establish the pH of the microenvironment in the active site cavity, it was assumed that this pH resembles that of the surrounding buffer, which is also used to determine the protonation of DXM. The buffers used were potassium phosphate at pH 6.0, 6.6, and 7.4 and Tris-HCl at pH 8.0 and 8.8. Stock solutions of DXM were prepared in 50% acetonitrile at 50 times the concentration in the incubation. The incubations were performed in triplicates at two different concentrations based on the Km values for formation of the two metabolites, dextrorphan (DXO) and 3-methoxymorphinan (MEM), in human liver microsomes. The compound (3 and 200 µM) was mixed with recombinant CYP2D6 (50 pmol/ml) and buffer (0.1 M). After 10-min preincubation, the reaction was initiated by the addition of NADPH (1 mM) and incubated for 7, 15, 20, and 30 min. Zero time-point samples were taken just before the addition of NADPH. All incubations were performed in glass vials in a water bath at 37°C. The reaction was quenched with two parts of ice-cooled acetonitrile with 0.8% formic acid, the samples were centrifuged at 2737g for 20 min at 4°C, and the supernatant was diluted 1:2 with water prior to analysis. For analysis of DXM at high concentration, the samples were diluted 1:400 in water. Standard curves were prepared in the incubation matrix for DXM, DXO, and MEM (30, 10, 3.3, 0.37, 0.12, 0.04, 0.014, and 0.0046 µM). The standard samples were precipitated with acetonitrile and diluted in water in the same way as the incubation samples.
|
For the interpretation of the metabolic stability data, the following classification was used. Compounds with Clint values below 0.7 µl/min/pmol P450 were considered to be low clearance compounds in CYP2D6, whereas Clint values above 3 µl/min/pmol P450 were classified as high clearance. These cut-off values were estimated from values used in metabolic stability screens in human liver microsomes (low Clint <10 µl/min/mg and high Clint >45 µl/min/mg, corresponding to hepatic clearance predictions of <30% of liver blood flow and >70% of liver blood flow in vivo). In product sheets on pooled human liver microsomes from BD Gentest, it was declared that the amount CYP2D6 was 15 pmol/ml microsomes (BD Gentest Inc.). This value was used to get the corresponding values for Clint in recombinant CYP2D6.
IC50 Determinations. For determination of inhibition potential of CYP2D6, full IC50 curves were obtained for each compound using a fluorescence endpoint assay (Crespi et al., 1997). A serial dilution of the test compounds was done in 50% acetonitrile (50, 16.667, 5.556, 1.852, 0.617, 0.206, 0.069, and 0.023 µM) and added to a master mix containing enzyme (20 pmol/ml), potassium phosphate buffer (0.1 M, pH 7.4), and the fluorescent probe substrate AMMC (15 µM). Also included were blank samples without test compound and substrate and incubations only containing the substrate, representing 100% activity. The incubation mixtures were preincubated for 10 min and initiated by addition of NADPH (0.4 mM), except for blank samples. All incubations were performed in microtiter plates at 37°C. After 30 min of incubation, the reactions were quenched with 80% acetonitrile and 20% TrisBase. The fluorescence was measured at 390/460 nm using a SpectraMax Gemini XS (Molecular Devices, Sunnyvale, CA). The IC50 values were determined by nonlinear least-squares regression analysis using XLfit4 version 4.1.1 (idbs, Guildford, UK). In this assay, an IC50 value below 10 µM was considered to be an indication of potential P450 inhibition.
Bioanalytical Equipment and Analysis. Analysis and quantification of DXM and the two metabolites DXO (O-demethylation) and MEM (N-demethylation) from the incubation of DXM at different pH were done using LC-MS. The HPLC system was an Agilent 1100 Series (Hewlett Packard GmbH, Waldbronn, Germany) coupled to a HTC PAL auto sampler (CTC Analytics AG, Zwingen, Germany). The analytical column used for chromatographic separation was a HyPurity C18 column (50 x 2.2 mm, 5 µm), and the mobile phases consisted of water and acetonitrile acidified with 0.1% formic acid. The gradient used was 5 to 90% acetonitrile in 4.5 min at a flow rate of 0.75 ml/min. The HPLC system was connected to a Sciex API 4000 quadrupole mass spectrometer with ESI interface (Applied Biosystems, Foster City, CA). The analytes were recorded with multiple reaction monitoring (MRM) in positive mode with Q1/Q3 masses 272/171 for DXM, 258/171 for DXO, and 258/157 for MEM. The Analyst 1.4.1 software (Applied Biosystems) was used for analysis and storage of data. All samples were analyzed regarding disappearance of DXM and formation of the two metabolites DXO and MEM. Authentic standards were used for quantification.
The analysis of the experimental data set of CYP2D6 substrates regarding disappearance of parent compound and formation of metabolites was done on a Waters (Milford, MA) Micromass LCT premiere TOF mass spectrometer with an ESI interface. For chromatographic separation, this was connected to a Waters Acquity UPLC system. The analytical column used was an Acquity UPLC C18 column (2.1 x 50 mm, 1.7 µm). The mobile phases consisted of water and acetonitrile acidified with 0.2% formic acid, and the gradient used was 5 to 90% acetonitrile in 5 min with a flow rate of 0.75 ml/min. A generic MS method was used for all compounds with a capillary voltage of 3000 V and the cone voltage set to 35 V. Detailed metabolite identification was performed with LC-MS/MS using a Waters Q-TOF Premiere instrument. The MS/MS analysis was done on the metabolites identified in the LCT premiere TOF spectrometer. The collision energy was ramped from 15 to 40 eV, and the cone voltage was set to 30 V. For chromatographic separation, the same system and settings were used as above. The MassLynx version 1.4 (Waters) software was used for analysis and storage of data.
Computational Analysis. FLAP. For identification of the SOM based on the CYP2D6 pharmacophore all compounds were aligned to DXM (Fig. 1b), a rigid molecule representing a 7-Å CYP2D6 substrate, and to the larger substrate tropisetron (Fig. 1c) to cover the 10-Å substrates. The alignments were done using the program FLAP (Baroni et al., 2007). This program uses GRID-derived molecular interaction fields (GRID-MIFs) (Goodford, 1985) to create four-point pharmacophores for the substrates. For these calculations, a hydrophobic probe (DRY), a hydrogen bond donor (N1), and a hydrogen bond acceptor (O) probe were used to describe the substrates. For each ligand, a maximum of 25 conformers were generated using random search by rotating a maximum of 10 rotable bonds. The calculated four-point pharmacophores for each conformer were used to determine pair wise similarity with the four-point pharmacophore for DXM or tropisetron resulting in a best conformer overlap for each compound. After confirming that the basic nitrogen atoms in DXM/tropisetron and the test compound were aligned, the parts of the test molecule that were overlapping with the known SOM in the templates were identified to be the preferred SOM. In the cases where solutions were obtained with both DXM and tropisetron as templates, the solution with the best alignment of the molecules and the shortest distance between the basic nitrogen atoms was selected.
For the interpretation of FLAP results, the following rules were used: 1) the FLAP predicted SOM is within two atoms from the group overlapping with the SOM in the template molecule; 2) N-dealkylation is suggested to be the metabolic reaction if the SOM in the template molecule is overlapping with or is within one atom away from a halogen atom; and 3) N-dealkylation is suggested to be the metabolic reaction if the SOM in the template molecule is overlapping with other atoms unable to undergo oxidative metabolism (e.g., exact overlap with an hydroxyl group).
MetaSite. The identification of SOM with the FLAP methodology was compared with the results from the previously known SOM prediction software MetaSite (Zamora et al., 2003). This methodology considers structural complementarity between the active site of the enzyme and the ligand as well as reactivity of the ligand. To compare the enzyme and ligand, two sets of descriptors are calculated. For the enzyme these are based on flexible GRID-MIFs. To generate the descriptor set of the ligand each atom of the molecule is classified as a GRID probe, and the distances between them are calculated. The resulting fingerprint of the ligand is compared with the description of the enzyme active site and the most optimal orientations are obtained. Based on this, all atoms in the molecule are ranked according to their accessibility to the heme. In addition, a reactivity factor based on fragment recognition is added. In summary, the site of metabolism is described by a probability index, which is the product of the similarity between ligand and protein (i.e., accessibility of one atom to the heme) and the reactivity.
For this analysis, the compounds were loaded into MetaSite as single conformers and protonated at the nitrogen atom according to the predicted pKa of the compounds. When analyzing the results the top two averaged ranked solutions with reactivity factor enabled were considered.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
) and, for the pH range used in the experiments (pH 6.0 to 8.8), this would theoretically correspond to a variation of the amount of ionized form of the molecule from 99.9 to 66.6%. The amount of metabolites formed after incubation of DXM was plotted against the time of incubation, and the rate of formation was calculated from the slope of the curve. This analysis was done for the different incubation conditions to see how the rate of formation varied with pH (Fig. 2, a and b). Based on the Km values for the two reactions, the incubations were done at DXM concentrations of 200 and 3 µM, respectively. In the incubations at 3 µM, only minute amounts of MEM (formed via N-demethylation) were formed; therefore, the results obtained from this concentration were not included in further analysis. The O-demethylation of DXM to DXO had a maximal rate at pH 7.4, and at higher pH values the reaction rate decreased (Fig. 2a). The N-demethylation (formation of MEM) also peaked at pH 7.4, but the rate did not decrease as much at increasing values (Fig. 2b). Since the change in pH probably affected the protein and its catalytic activity, the comparison is difficult to evaluate. For a more proper evaluation of the influence of pH on the N- and O-demethylation of DXM, the ratio between the rates of formation of the two metabolites was calculated (DXO/MEM). To use this ratio would diminish the effect on the protein assuming that the decreased catalytic activity at high and low pH was the same for both reactions. The O-demethylation/N-demethylation ratio clearly decreased at higher pH (Fig. 2c), indicating that more N-demethylation occurred when the pH increased and DXM was present to a higher extent as the free base. FLAP Alignment and MetaSite Analysis of Literature Compounds. Table 1 summarizes the reported SOM, the preferred SOM based on the FLAP alignment, and the predicted SOM using MetaSite for a literature data set of 23 compounds. Also included in the table are the predicted pKa values from MOKA. The majority of the compounds gave the best alignments with DXM (7-Å substrate), but eletriptan, brofaromine, buflomedil, carteolol, tolperisone, bufuralol, ibogaine, propafenone, and bunitrolol were better aligned with tropisetron (10-Å substrate). For five compounds (dexfenfluramine, hydromorphone, morphine, chlorpheniramine, and eletriptan) in the literature data set, N-dealkylation was reported as one of the major metabolic pathways in CYP2D6. For four of these compounds (dexfenfluramine, hydromorphone, morphine, and chlorpheniramine), the part identified as the preferred site of metabolism by the FLAP alignments was blocked toward oxidative metabolism according to the FLAP prediction criteria (see Materials and Methods). The fifth N-dealkylated compound, eletriptan, did not follow the working hypothesis since the molecule did not contain any groups unreactive toward oxidation. On the other hand, the electron withdrawing sulfon would make the aromatic moiety in eletriptan less prone to undergo oxidative metabolism. For this group of compounds, MetaSite predicted N-dealkylation in one of five cases (Table 3). It is worth mentioning that MetaSite predicted SOM for morphine and eletriptan to be at a carbon atom in the alfa-position to the nitrogen but not the carbon in the N-methyl group. A hydroxylation at the MetaSite predicted carbon atom would eventually result in a cleavage of the nitrogen-carbon bond. However, this will not result in the N-demethylated metabolite reported in the literature and is not considered a correct prediction.
|
|
In the larger set comprising the compounds reported to be metabolized by CYP2D6 via other metabolic routes such as aromatic hydroxylation and/or O-demethylation the FLAP alignment did not identify any stable groups at the preferred site of metabolism. In 14 of 18 cases, the FLAP-predicted site of metabolism was correct compared with what was reported in the literature. One of the four wrongly predicted compounds, bisoprolol, was not successfully aligned to any of the template structures. The prediction rate for MetaSite was 16 of 18 correctly predicted compounds for this set. A summary of the prediction rates is presented in Table 3. It was difficult to find a large number of methyl-/ethylaminyl compounds metabolized through N-dealkylation as major metabolic route by CYP2D6; however, the compounds identified support the working hypothesis stated in this study.
Metabolite Identification, Clint, and IC50 Determinations of CYP2D6 Substrates. Twenty compounds were selected for experimental determinations of major metabolites, intrinsic clearance (Clint), and inhibition potential in recombinant CYP2D6. The selection was done with the aim to represent both N-dealkylated and non-N-dealkylated substrates. Another goal was to find some compounds with similar core structures in order to compare the effects of stable groups in the molecule. The compounds were incubated with recombinant CYP2D6, and the Clint was determined from disappearance of the parent compound. These incubations were also analyzed with LC-MS/MS in order to rationalize the structure of major metabolites. The inhibition potential of CYP2D6, the IC50 value, was determined with an endpoint assay using a fluorescent probe substrate. Most of the compounds used were racemates but, due to the availability, (S,S)-sertraline and (R)-tomoxetine were used as pure stereoisomers. Tramadol and O-desmethyltramadol were used as the cis-racemates.
Predicted pKa values, major metabolites formed in CYP2D6, calculated Clint values, and IC50 values for the 20 compounds are reported in Table 2. Due to difficulties in determining specific regioselectivity for hydroxylation of some compounds, larger and more unspecific areas were suggested as the major SOM. However, from the results it is still possible to discriminate whether the metabolism occurred in the proximity of the nitrogen or not, which actually was the aim of this study. Metabolite identification data from LC-MS/MS analysis of the parent compounds and major metabolites are available as Supplemental Information.
|
The compounds were divided into two groups—those with N-dealkylation as one major metabolic route and compounds with other metabolic reactions responsible for the biotransformation by CYP2D6. The group with N-dealkylation as a major metabolic pathway consisted of eight compounds (nisoxetine, fluoxetine, Bionet-BB 6N-710, Bionet-BB 6N-708, lidocaine, (S,S)-sertraline, maprotiline, and citalopram). The major metabolite of citalopram was the N-oxide and not the N-dealkylated compound. Despite this observation, citalopram was, however, included in this group due to the fact that the oxidation takes place on the basic nitrogen. Of these eight compounds, FLAP alignment identified stable groups at the preferred site of metabolism for fluoxetine, Bionet-BB 6N-708, (S,S)-sertraline, and citalopram (Table 2). The other compounds did not contain any stable groups in the molecule but, on the other hand, Bionet-BB 6N-710 and lidocaine were both stable according to their Clint values. Even though metabolites were found, the compounds were not metabolized by CYP2D6 to any greater extent. The high IC50 value for lidocaine also indicates poor affinity for the isoenzyme. Comparing the predicted pKa of the compounds, lidocaine also had the lowest pKa value (pKa, 7.99), which would further support that N-dealkylation occurs due to a larger fraction present as the free base (
20%). The metabolism of nisoxetine and maprotiline could not be explained by our working hypothesis. Interestingly the IC50 values were in general within the lower range for this group (all except for citalopram had IC50 values below 10 µM with most of them below 5 µM), indicating a high inhibition potential. The high inhibition potential could be an indication that the compounds adopt a binding mode in CYP2D6 based on the "traditional" pharmacophore, which does not lead to efficient formation of any metabolite due to blocking groups at the predicted SOM.
The last group, containing the compounds not N-dealkylated by CYP2D6 to any major extent, included the highest Clint values. Only 3/12 compounds were classified as low Clint compared with the first group where 5/8 were low Clint and neither of the compounds classified as high clearance drugs in CYP2D6. This indicates that the compounds in the second group were more efficiently metabolized by CYP2D6 compared with those metabolized through N-dealkylation. The IC50 values ranged between <1 µM up to approximately 40 µM. Thus, many of these compounds can adopt productive binding modes that lead to metabolites that do not cause any major inhibition. To rationalize these differences in Clint values and inhibition potential, a general trend observed for these compounds was that when introducing stable groups in the molecule the metabolic stability was improved while the inhibition potential was increased. This effect can be high-lighted with the example of Bionet-BB 4N-726 and Bionet-BB 6N-708 (Table 2), two compounds with the same core structure but with different substituents. Bionet-BB 4N-726 had a Clint value of 4.3 µl/min/pmol P450 and an IC50 value of 3.33 µM. Bionet-BB 6N-708, in which stable groups are introduced, was stable in the Clint assay (<0.2 µl/min/pmol P450) but the IC50 value was almost 10 times lower (0.38 µM) than that of Bionet-BB 4N-726.
As for the literature data set, the majority of the compounds in the experimental data set were successfully aligned with DXM, but fluoxetine, lidocaine, and alprenolol were better aligned with tropisetron. For compounds not metabolized by N-dealkylation as a major pathway, the success rate for predicting the SOM using the FLAP alignment was 9/12. Of the three wrongly predicted compounds, O-desmethyltramadol was not successfully aligned with either of the two template molecules. For this group, MetaSite predicted correct sites of metabolism in 10 of 12 cases. On the other hand, MetaSite was not successful in predicting compounds that were N-dealkylated (1 of 8). A summary of the prediction rates is presented in Table 3.
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
; Lussenburg et al., 2005), as well as the hypothesis that different oxygenating species are responsible for different reactions (Keizers et al., 2005; Newcomb and Chandrasena, 2005), have been previously reported. This study, however, focuses on another determinant: the influence of the electrostatic interaction between the acidic amino acids in the active site cavity and the basic nitrogen in the substrate.
In this study we have explored how structural factors influence the occurrence of the less common N-dealkylation reactions by CYP2D6. N-Dealkylation is a common metabolic pathway for drugs containing secondary and tertiary amines and, in many cases, CYP3A4, CYP1A2, or CYP2C19 are involved in the catalysis (Coutts et al., 1994). This reaction also occurs in CYP2D6 but it is far less common than in the above-mentioned isoforms. To focus on how CYP2D6 catalyzes N-dealkylation reactions, this study only considered the in vitro situation using recombinant enzyme. From the results it could be seen that the few compounds that were identified to or reported to be metabolized through N-dealkylation had a low Clint value and were poorly cleared from the system. In the in vivo situation, N-dealkylation via CYP2D6 is probably of minor importance and CYP3A4 with less substrate specificity will dominate the metabolism at high concentrations of the substrate.
Early pharmacophore models and the crystal structure of CYP2D6 have pointed out one major characteristic of CYP2D6 substrates, a basic nitrogen present in the molecule 5-7 or 10 Å from the site of oxidation, which can interact with an acidic amino acid residue on top of the active site. At physiological pH the positive charge of the basic nitrogen would favor an orientation of the substrate with the nitrogen away from the heme due to electrostatic interactions with the acidic amino acid in the active site cavity. Even though a major fraction of the amines will be protonated there is still a small neutral fraction whose orientation would not be restrained by the electrostatic interactions in the cavity. Thus, increasing the fraction of neutral molecules would theoretically increase the probability for N-dealkylation to occur. The results from incubation of DXM with CYP2D6 at different pH values clearly support this hypothesis. The ratio between the rate of O-demethylation and N-demethylation shows a pH dependence with an increase in N-demethylation at higher pH values. The influence of electrostatic interactions in CYP2D6 has previously been studied by Upthagrove and Nelson (2001), where it was shown that an increase of the basicity of CYP2D6 substrates by introducing different substituents in the molecule increased the affinity for the enzyme (Upthagrove and Nelson, 2001). These studies show a relationship between protonation of substrate with affinity and orientation of CYP2D6 substrates in the active site cavity of the enzyme. In a study by Miller et al. (2001), the issues of protonation of substrates in binding and catalysis were studied in CYP2D6. From these results it was suggested that the basicity of the amine was decreased upon binding resulting in enhanced availability of the unprotonated form.
Searching the literature for compounds that are N-dealkylated by CYP2D6 and compounds not N-dealkylated gave the opportunity to investigate another working hypothesis that N-dealkylation occurs in CYP2D6 if the preferred SOM is blocked in one way or another. Since the amounts of CYP2D6 substrates that are N-dealkylated are rather limited in the literature it is difficult to do a comparison with statistical significance. Nevertheless, this study gives the opportunity to evaluate the general trends observed. There are, however, difficulties in drawing conclusions based only on literature data since experimental conditions often differ with different enzyme sources, different concentrations, etc. In this study, 20 compounds were selected for experimental determination of the metabolism by CYP2D6 in order to obtain comparable data. The Km values for the metabolic reactions studied vary over a wide range; however, the experiments performed to study and compare the formation of metabolites and disappearance of parent compound were run at a concentration of 1 µM in all cases. In incubations at higher concentrations, N-dealkylated metabolites could be detected for almost all compounds. This indicates that, at some point, with an increased substrate concentration, there will be some molecules oriented with the nitrogen toward the heme even if this is not the most favorable orientation.
Analyzing the results, several of the substrates that are N-dealkylated have a stable blocking group in the molecule. On the other hand all of these compounds have low Clint values and are thus poorly metabolized by CYP2D6. This could be explained by the above-mentioned discussion that only the noncharged fraction of the molecules is N-dealkylated. When the preferred SOM is blocked, other metabolic reactions like O-demethylation and hydroxylation will not occur and mainly N-dealkylated metabolites are observed. Regarding the metabolic properties of the compounds, adding blocking groups at sites of metabolism seemed to increase the metabolic stability but also resulted in an increased inhibition potential. This has important implications in solving metabolic issues, in that minor changes in the molecular structure in order to improve metabolic stability might increase the enzyme inhibition potential of the compound and result in undesirable drug-drug interactions. This phenomenon of improved metabolic stability but retained or increased inhibition potential as a consequence of the addition of metabolically stable groups in the molecules has also been observed by Ahlström et al. (2007) in the case of CYP2C9.
The program FLAP was used to identify the preferred SOM according to the CYP2D6 pharmacophore. To cover the range of CYP2D6 substrates, DXM, a "7-Å substrate," and tropisetron, a "10-Å substrate," were used as templates. The prediction success rates were encouraging and, when summarizing the results from the literature data set and the experimental data set, SOM was successfully predicted in 77% of the compounds not metabolized by N-dealkylation. Of these 30 compounds, two were not successfully aligned (bisoprolol and O-desmethyltramadol) with either template, thus not fitting the CYP2D6 pharmacophore. This method is a ligand-based approach, and a comparison of the prediction rate with MetaSite, an approach that considers both protein-ligand complementarity and reactivity, was done. For the compounds not metabolized by N-dealkylation, MetaSite predicted the correct site of metabolism in 87% of the cases. This is somewhat better than the result obtained from FLAP alignments, but still this simple ligand-based approach works well for CYP2D6. Afzelius et al. (2007) compared different computational approaches for SOM predictions in CYP2C9 and CYP3A4 and showed that MetaSite is superior, but that also the pure ligand-based approaches perform well (Afzelius et al., 2007). It was argued that due to the large and flexible active site cavity of CYP3A4, reactivity is thought to be the driving force in metabolism of the substrates of this isoform. The study also indicated that CYP2C9 is reactivity driven, despite its smaller active site cavity resulting in more specific interactions. In the case of CYP2D6, steric hindrance is also believed to be of importance for selectivity, thus indicating that structure based methods would be successful for SOM predictions. Nevertheless, with the well characterized pharmacophoric features of CYP2D6 substrates it is not surprising that a pharmacophore based alignment method is successful in predicting the site of metabolism of this isoform. To identify compounds that are N-dealkylated by CYP2D6 the FLAP alignment can be a useful method. For all compounds that were N-dealkylated and contained stable groups in the molecule FLAP identified the stable group at the preferred site of metabolism; however, there were some false negatives where N-dealkylation occurs even though no blocking group was present. MetaSite did not perform as well in identifying N-dealkylation as site of metabolism and the FLAP alignment could therefore be a good complement to this methodology.
In conclusion, the hypothesis that N-dealkylation occurs in CYP2D6 when the preferred site of metabolism is blocked seems to be correct in most cases studied here. It is also evident that the fraction of unprotonated molecules affects the extent to which N-dealkylation will occur. In addition, an important finding is that blocking of the sites of metabolism can retain or increase problems with enzyme inhibition. This study also includes an investigation of a novel computational tool, FLAP, for site of metabolism predictions. First, using this tool to spot stable groups at preferred SOMs to identify N-dealkylation reactions by CYP2D6 was shown to be successful and could be a valuable complement to other available SOM prediction tools such as MetaSite. Second, this fairly simple approach to predict SOM for CYP2D6 substrates by alignment to a template molecule worked satisfactorily, which highlights the importance and influence of the CYP2D6 pharmacophore. Nevertheless, further work needs to be done to explore ways of disrupting the CYP2D6 pharmacophore to address both metabolic stability issues (SOM) and enzyme inhibition issues while retaining the pharmacological activity of compound classes.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
ABBREVIATIONS: P450, cytochrome P450; Clint, intrinsic clearance; SOM, site of metabolism; DXM, dextromethorphan; DXO, dextrorphan; MEM, 3-methoxymorphinan; AMMC, 4-aminomethyl-7-methoxycoumarine; GRID-MIFs, GRID-derived molecular interaction fields; FLAP, Fingerprints for Proteins and Ligands; HPLC, high-performance liquid chromatography; ESI, electrospray ionization; LC-MS, liquid chromatography-mass spectrometry; MRM, multiple reaction monitoring.
The online version of this article (available at http://dmd.aspetjournals.org) contains supplemental material. 
Address correspondence to: Britta Bonn, Discovery DMPK & Bioanalytical Chemistry, AstraZeneca R&D Mölndal, SE-431 81 Mölndal, Sweden. E-mail: kjelland{at}chem.gu.se
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
glutamate, asparagine, and glycine mutants. Arch Biochem Biophys 331: 134-140.[CrossRef][Medline]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
 |
 |
|
 |
 |
 |
