Exploration of catalytic properties of CYP2D6 and CYP3A4 through metabolic studies of levorphanol and levallorphan

Article abstract of DOI:10.1124/dmd.109.028670

CYP2D6 and CYP3A4, two members of the cytochrome P450 superfamily of monooxygenases, mediate the biotransformation of a variety of xenobiotics. The two enzymes differ in substrate specificity and size and characteristics of the active site cavity. The aim of this study was to determine whether the catalytic properties of these isoforms, reflected by the differences observed from crystal structures and homology models, could be confirmed with experimental data. Detailed metabolite identification, reversible inhibition, and time-dependent inhibition were examined for levorphanol and levallorphan with CYP2D6 and CYP3A4. The studies were designed to provide a comparison of the orientations of substrates, the catalytic sites of the two enzymes, and the subsequent outcomes on metabolism and inhibition. The metabolite identification revealed that CYP3A4 catalyzed the formation of a variety of metabolites as a result of presenting different parts of the substrates to the heme. CYP2D6 was a poorer catalyst that led to a more limited number of metabolites that were interpreted in terms to two orientations of the substrates. The inhibition studies showed evidence for strong reversible inhibition of CYP2D6 but not for CYP3A4. Levallorphan acted as a time-dependent inhibitor on CYP3A4, indicating a productive binding mode with this enzyme not observed with CYP2D6 that presumably resulted from close interactions of the N-allyl moiety oriented toward the heme. All the results are in agreement with the large and flexible active site of CYP3A4 and the more restricted active site of CYP2D6. Copyright

Full text of DOI:10.1124/dmd.109.028670

Supplemental Material can be found at:
http://dmd.aspetjournals.org/content/suppl/2009/10/01/dmd.109.028670
.DC1.html
0090-9556/10/3801-187–199$20.00
DRUG METABOLISM AND DISPOSITION
Copyright © 2010 by The American Society for Pharmacology and Experimental Therapeutics
DMD 38:187–199, 2010
Vol. 38, No. 1
28670/3540905
Printed in U.S.A.
Exploration of Catalytic Properties of CYP2D6 and CYP3A4
Through Metabolic Studies of Levorphanol and Levallorphan
□
S
Britta Bonn, Collen M. Masimirembwa, and Neal Castagnoli, Jr.
Department of Chemistry, Medicinal Chemistry, University of Gothenburg, Gothenburg, Sweden (B.B.); Discovery DMPK,
AstraZeneca Research and Development, Mo¨ lndal, Sweden (B.B., C.M.M.); African Institute of Biomedical Science and
Technology, Harare, Zimbabwe (C.M.M.); and Department of Chemistry, Virginia Tech and the Edward Via College of
Osteopathic Medicine, Blacksburg, Virginia (N.C.)
Received May 31, 2009; accepted September 29, 2009
ABSTRACT:
CYP2D6 and CYP3A4, two members of the cytochrome P450 su- tion revealed that CYP3A4 catalyzed the formation of a variety of
perfamily of monooxygenases, mediate the biotransformation of a metabolites as a result of presenting different parts of the sub-
variety of xenobiotics. The two enzymes differ in substrate speci- strates to the heme. CYP2D6 was a poorer catalyst that led to a
ficity and size and characteristics of the active site cavity. The aim
more limited number of metabolites that were interpreted in terms
of this study was to determine whether the catalytic properties of to two orientations of the substrates. The inhibition studies
these isoforms, reflected by the differences observed from crystal showed evidence for strong reversible inhibition of CYP2D6 but not
structures and homology models, could be confirmed with exper- for CYP3A4. Levallorphan acted as a time-dependent inhibitor on
imental data. Detailed metabolite identification, reversible inhibi- CYP3A4, indicating a productive binding mode with this enzyme
tion, and time-dependent inhibition were examined for levorphanol not observed with CYP2D6 that presumably resulted from close
and levallorphan with CYP2D6 and CYP3A4. The studies were interactions of the N-allyl moiety oriented toward the heme. All the
designed to provide a comparison of the orientations of sub- results are in agreement with the large and flexible active site of
strates, the catalytic sites of the two enzymes, and the subsequent
outcomes on metabolism and inhibition. The metabolite identifica-
CYP3A4 and the more restricted active site of CYP2D6.
CYP3A4 and CYP2D6 are two members of the cytochrome P450 cently (Kjellander et al., 2007). This comparison involved docking
(P450) superfamily of monooxygenases that catalyze the biotransfor-
mation of a variety of xenobiotics (Guengerich, 2003). CYP3A4 has
a large, flexible active site (Ekroos and Sjo¨gren, 2006) resulting in
broad substrate selectivity. CYP2D6 is more substrate-selective be-
cause of its smaller active site and the presence of specific amino acid
residues that may contribute to substrate-enzyme interactions (Koy-
mans et al., 1992; de Groot et al., 1997; Lewis et al., 1997).
To gain further structural information of the P450s, efforts have
been directed toward the crystallization of these enzymes. The crys-
tallization of soluble bacterial P450s (Poulos et al., 1987) provided an
opportunity to conduct homology modeling studies of human P450s
using these bacterial crystal structures as templates (Lewis, 1999; De
Rienzo et al., 2000). The crystallizations of human CYP3A4 in 2004
(Williams et al., 2004; Yano et al., 2004) and CYP2D6 in 2006
(Rowland et al., 2006) allow for an extension of homology modeling
to include direct comparisons with the human enzymes. A comparison
of crystal structures and homology models has been published re-
solutions of a set of compounds for which specific interactions with
the surrounding amino acid residues had been identified by compu-
tational methods. In addition, the opportunity to predict sites of
metabolism for the different structures was investigated. The com-
pound class used for this study was a family of morphinanyl opioid
analgesics. These compounds are known substrates for both CYP2D6
and CYP3A4 that presumably adopt different orientations in the
active sites of the two isoforms (Kirkwood et al., 1997; Yue and Sawe,
1997; von Moltke et al., 1998; Benetton et al., 2004; Hutchinson et al.,
2004). CYP3A4 preferentially catalyzes the oxidative N-demethyl-
ation, and CYP2D6 catalyzes the oxidative O-demethylation of se-
lected members of this class of compounds. Attempts to dock some of
these probe molecules into the active sites of CYP3A4 as defined by
its X-ray crystal structure led to ambiguous results. The large and
flexible active site cavity of this enzyme resulted in orientations with
different parts of the molecules in close proximity to the heme. This
in turn led to a variety of solutions for the predicted sites of metab-
olism. Because the experimentally determined sites of metabolism
reported in the article by Kjellander et al. (2007) considered major
metabolic pathways only, a more detailed characterization of the
metabolic profiles of two compounds of the series has been pursued.
The experimental results were expected to provide information about
Article, publication date, and citation information can be found at
http://dmd.aspetjournals.org.
doi:10.1124/dmd.109.028670.
□S The online version of this article (available at http://dmd.aspetjournals.org)
contains supplemental material.
ABBREVIATIONS: P450, cytochrome P450; TDI, time-dependent inhibition; AMMC, 4-aminomethyl-7-methoxycoumarine; BFC, 7-benzyloxy-4-
trifluoromethylcoumarine; HPLC, high-performance liquid chromatography; MS, mass spectrometry; XIC, extracted ion chromatogram; TIC, total
ion current chromatogram; NR, normalized ratio; Clint, intrinsic clearance.
187

188
BONN ET AL.
observed on the Sciex API 4000 quadrupole mass spectrometer (Applied
Biosystems). The collision energy was ramped from 20 to 40 eV, and the cone
voltage was set to 35 V. For chromatographic separation, a Waters (Milford,
MA) ACQUITY UPLC was used with a Waters ACQUITY UPLC C18 column
(50 ϫ 2.1 mm, 1.18 ␮m). The same mobile phases as on the Agilent 1100
series instrument (Hewlett-Packard GmbH) were used. The gradient of 5 to
90% B was achieved in 5 min; the flow rate was 0.75 ml/min. MassLynx
version 1.4 software (Waters) was used for analysis and storage of data.
The analysis of HO-bufuralol and 1Ј-HO-midazolam for the IC50 deter-
minations was performed with HPLC/MS on the Sciex API 4000 instru-
ment (Applied Biosystems) described above. The multiple reaction moni-
toring transitions used were 2778.1/185.7 for HO-bufuralol and 342.0/
202.7 for HO-midazolam. The column used for chromatographic separation
was a HyPURITY C18 column (50 ϫ 2.5 mm, 5 ␮m; Thermo Fisher
Scientific).
Metabolite Identification in Incubations with Recombinant Enzymes.
The metabolic profiles of levorphanol (1) and levallorphan (2) were investi-
gated with recombinant CYP2D6 and CYP3A4. The compounds (10 ␮M) were
mixed with enzyme (100 pmol/ml) and phosphate buffer (0.1 M, pH 7.4). Each
mixture was preincubated for 10 min at 37°C, and the reaction was then
initiated by addition of NADPH (1 mM). Zero time samples were taken before
the addition of NADPH. After 60 min, the incubations were quenched with the
addition of an equal amount of ice-cooled solution of 0.8% formic acid in
acetonitrile. The samples were centrifuged, and the supernatant was diluted 1:1
in water before analysis. The 0- and 60-min samples were used for identifi-
cation of metabolites.
FIG. 1. Structures of levorphanol (1) and levallorphan (2).
the different orientations the molecules adopt in CYP2D6 and
CYP3A4 as part of an effort to characterize further the differences
between the two isoforms, regarding interactions and active site
cavities, observed with computational methods.
The two compounds selected for this study (Fig. 1) were levorpha-
nol (1) and levallorphan (2). Levorphanol, an N-methylmorphinanyl
derivative, serves as the prototype for this family of opioid analgesics.
The second compound, levallorphan (2), was selected because of the
presence of the N-allyl side chain that might be involved in hydro-
phobic interactions in the active site cavity that could result in an
improved affinity. Furthermore, compounds bearing the allylaminyl
group present in levallorphan have been reported to undergo meta-
bolic activation leading to time-dependent inhibition (TDI) of the
enzyme (Fontana et al., 2005). Consequently, the TDI properties of
levorphanol and levallorphan were compared using the two enzymes.
In addition, the reversible inhibition potential was determined for the
compounds as a measure of the affinity for the two P450 isoforms. It
was anticipated that these results would give further information on
the productive orientations of these two molecules in the active sites
of CYP3A4 and CYP2D6. Also included in this article are dockings
of the two compounds based on the crystal structures of CYP3A4 and
CYP2D6.
In the case of levallorphan (2), incubations were also performed in the
presence of GSH to capture any reactive species. The same incubation condi-
tions as above were used with the addition of GSH (5 mM).
To quantify the turnover of the parent compound, intrinsic clearance (Clint
)
values for levorphanol and levallorphan were determined at 1 ␮M. The
compounds were mixed with recombinant P450 (50 pmol/ml) and potassium
phosphate buffer (0.1 M, pH 7.4). After 10-min preincubation, the reaction was
initiated by addition of NADPH (1 mM), and the incubations were continued
for 7, 15, 20, and 30 min. Zero time point samples were taken just before
addition of NADPH. All the incubations were performed in microtiter plates at
37°C. The reaction was quenched with two parts ice-cooled acetonitrile con-
taining 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 before analysis.
IC50 Determination. For determination of the potential of the test com-
pounds to inhibit reversibly CYP2D6 and CYP3A4, full IC50 curves were
obtained with recombinant P450s. The inhibition of the catalytic activity of the
enzymes was studied using bufuralol and midazolam as probe substrates for
CYP2D6 and CYP3A4, respectively. Serial dilutions of the test compounds
were done in 50% acetonitrile (50.00, 16.70, 5.50, 1.90, 0.60, 0.20, 0.07, and
0.02 ␮M). These solutions were added to a master mix containing enzyme (5
pmol/ml), potassium phosphate buffer (0.2 M, pH 7.4), and the probe sub-
strates bufuralol (10 ␮M) for CYP2D6 and midazolam (3 ␮M) for CYP3A4.
Also included were blank samples, without test compound and substrate, and
incubations containing only the substrate (100% activity). The incubation
mixtures were preincubated for 10 min, and the reactions were initiated by
addition of NADPH (1 mM), except for the blank samples. After incubating for
15 min, the reactions were quenched with ice-cooled acetonitrile containing
0.8% formic acid. All the incubations were performed in microtiter plates at
37°C. The plates were centrifuged at 2737g for 20 min at 4°C, and the
supernatants were diluted 1:2 in water before analysis. The CYP2D6 samples
were diluted again 1:10. Standard curves of 1Ј-HO-midazolam and HO-
bufuralol were prepared in the incubation matrix and were precipitated in the
same way as the samples. The formation of 1Ј-HO-midazolam and HO-
bufuralol were analyzed with HPLC/MS.
Materials and Methods
Materials. The recombinant enzymes used in these studies (human P450s
CYP2D6 and CYP3A4 coexpressed with NADPH-P450 reductase in bacterial mem-
branes) were supplied from Cypex Ltd. (Dundee, Scotland, UK). Levorphanol tartrate,
levallorphan tartrate, NADPH, and GSH were purchased from Sigma-Aldrich (St.
Louis, MO), and 4-aminomethyl-7-methoxycoumarine (AMMC) and 7-benzyloxy-4-
trifluoromethylcoumarine (BFC) were from BD Gentest (Woburn, MA). Propranolol
and troleandomycin were supplied from Sigma-Aldrich. HO-Bufuralol, bufuralol, and
1Ј-HO-midazolam were purchased from SAFC Corp. (St. Louis, MO), and midazolam
was from Lipomed AG (Arlesheim, Switzerland). All of the other chemicals were of
analytical grade and of the highest quality available.
Bioanalytical Equipment and Analytical Methods. High-performance
liquid chromatography/mass spectrometry (HPLC/MS) and HPLC/MS2 were
conducted for the metabolite identification studies. The HPLC system was an
Agilent 1100 Series (Hewlett-Packard GmbH, Waldbronn, Germany) coupled
to an HTC PAL autosampler (CTC Analytics AG, Zwingen, Switzerland). The
analytical column used for chromatographic separation of the metabolites was
a HyPURITY C18 column (100 ϫ 2.5 mm, 5 ␮m; Thermo Fisher Scientific,
Waltham, MA). The mobile phases consisted of 0.1% formic acid in water (A)
and 0.1% formic acid in acetonitrile (B). The gradient was increased linearly
from 5% solvent B to 50% solvent B in 8 min at a flow rate of 0.75 ml/min.
The HPLC system was connected to a Sciex API 4000 quadrupole mass
spectrometer with electrospray ionization interface (Applied Biosystems, Con-
cord, ON, Canada). Positive ion mass spectra were recorded over the mass
range m/z 80 to 400, followed by product ion scans on peaks of interest. The
collision energy used was 40 eV. The Analyst 1.4.1 software (Applied Bio-
systems) was used for analysis and storage of data.
Time-Dependent Inhibition. Levorphanol and levallorphan were tested for
their TDI properties. A time-dependent inhibitor is a compound that inhibits
the enzyme in an irreversible or quasi-irreversible manner after metabolic
activation. In this assay the activities of recombinant CYP2D6 and CYP3A4
were measured with the fluorescent probe substrates AMMC and BFC after 0,
Additional MS2 experiments using a Waters (Wythenshawe, Manchester, 5, 10, and 20 min of preincubation with the test compound. For the determi-
UK) quadrupole time-of-flight Premiere instrument were performed to obtain nation of the extent of enzyme inactivation, the enzymes were incubated at
accurate mass data. The MS2 analyses were performed on the metabolites 37°C with the test compound (1.5, 12.5, and 25 ␮M) in the presence and

EXPLORATION OF CATALYTIC PROPERTIES OF CYP2D6 AND CYP3A4
189
absence of NADPH (1 mM for CYP3A4 and 0.4 mM for CYP2D6). Control
incubations were included without any test compound present. After preincu-
Sunnyvale, CA) and exported as sdf-files. The two- to three-dimensional
conversion was made with CORINA (Molecular Networks GmbH, Erlangen,
bation, a 20-␮l aliquot of the preincubation mixture was transferred to a well Germany; http://www.mol-net.de). For the dockings in CYP2D6 substrate
molecules were protonated because the electrostatic interactions are believed
to be important for the orientation in the CYP2D6 active site cavity. As in the
study by Kjellander et al. (2007), neutral molecules were used for the dockings
in CYP3A4.
containing BFC (50 ␮M) or AMMC (60 ␮M), NADPH (1 mM for CYP3A4
and 0.4 mM for CYP2D6), and phosphate buffer (0.2 M) at a final volume of
200 ␮l. To decrease the impact of reversible inhibition, substrate concentra-
tions at 4ϫ Km were used to saturate the enzyme. A 1/20 dilution of the
preincubation mixture into the incubation with substrate was also tested to
further rule out a reversible inhibition component. The final protein concen-
tration was 1.25 pmol/ml CYP3A4 and 20 pmol/ml CYP2D6. The higher
enzyme concentration with CYP2D6 was used to obtain a sufficient signal-to-
noise ratio when measuring the fluorescence of the AMMC metabolite. The
incubations were quenched after 15 min with a stop solution containing 80%
acetonitrile and 20% Tris. The catalytic activities were determined from the
amount of metabolite formed from the probe substrates by measuring the
fluorescence at 390/460 nm for AMMC and at 405/535 nm for BFC. This was
done using a SpectraMax Gemini XS fluorescence detector (Molecular De-
vices, Sunnyvale, CA). The percentage of remaining activity was calculated by
dividing the counts in the samples with the counts in the control incubations
(100% activity). Included in the assays were positive controls known to act as
time-dependent inhibitors on the two enzymes. These were propranolol for
CYP2D6 (Narimatsu et al., 2001) and troleandomycin for CYP3A4 (Yamazaki
and Shimada, 1998).
Dockings of Levorphanol and Levallorphan into Crystal Structures of
CYP2D6 and CYP3A4. Dockings of levorphanol and levallorphan in
CYP2D6 and CYP3A4 were performed to visualize possible orientations in the
active site cavities. The protein structures used were the ligand-free CYP2D6
crystal structure (Protein Data Bank code 2f9q) (Rowland et al., 2006), and the
CYP3A4 crystal structure cocrystallized with erythromycin (Protein Data
Bank code 2j0d) (Ekroos and Sjo¨gren, 2006). In the study by Kjellander et al.
(2007), these two crystal structures were evaluated regarding site of metabo-
lism predictions of compounds in the same family as levallorphan and levor-
phanol. Consequently, these structures were chosen for the purpose of this
study. The CYP2D6 crystal structure has some unresolved regions; however,
these are assumed not to affect the environment of the active site. Because the
aim was to visualize possible orientations of the compounds, settings were
used to restrict the dockings based on knowledge of the two enzyme active site
The dockings were performed with the docking module GLUE in program
GRID version 22.2.2 (Molecular Discovery, Pinner, Middlesex, UK; http://
moldiscovery.com). This program is based on calculations of GRID-molecular
interaction fields (Goodford, 1985). Specific probes related to the two com-
pounds studied were chosen for analysis of the active site of the enzyme. In
CYP2D6 these were H (hydrogen), OH2 (water), DRY (hydrophobic),
N1ϩ (sp3 aminyl cation), C3 (methyl), C1 ϭ (sp2 CH aromatic or vinylic),
and OH (phenol or carboxyl). For CYP3A4 the same probes were used
except for N1ϩ, which was exchanged for N: (sp3 N with a lone pair). For
the analysis a distance of 1 Å between the grid points was used. To restrict
the dockings, a box estimated to include the heme and active site was used
(ϳ16 ϫ 20 ϫ 23 Å). In the dockings possible interactions with surrounding
water molecules were not considered. For the binding affinity estimation,
contributions from electrostatic interactions were included, as well as a
penalty for loss of rotatable torsions. Different conformations of the ligands
were generated in GLUE using random search by rotating a maximum of
five rotatable bonds. A maximum number of binding sites was set to 100,
and an energy cutoff to Ϫ100 kcal/mol was included. Because the aim was
to provide possible productive binding modes, the resulting docking poses
were filtered, and only those that were within 6 Å from any atom of the
heme were selected for further analysis.
Results
Identification of Levorphanol Metabolites. An analysis of the
product ion spectrum (Fig. 2) of protonated levorphanol (1H؉) was
undertaken to provide comparative data that would be useful in the
elucidation of the structures of metabolites derived from this com-
pound. This spectrum (as well as the corresponding spectra of lev-
allorphan and the derived metabolites) could be obtained only at a
cavities. The substrates were drawn in ISISdraw (Symyx Technologies Inc., high collision dissociation energy (40 eV) because of the stability of
FIG. 2. HPLC/electrospray ionization/MS product ion spectrum of protonated levorphanol (1H؉) obtained at 40 eV.

190
BONN ET AL.
the MH؉ species. The molecular formulas for the major product ions metabolites of levorphanol and levallorphan display product ions
equivalent to ii؉ and iii؉.
were obtained from accurate mass analysis (Table 1). Structure
Levorphanol Metabolites Formed in the Presence of CYP3A4
and CYP2D6. The extracted ion chromatograms (XICs) obtained
from incubations of levorphanol with CYP3A4 and CYP2D6 of all the
metabolite peaks are shown in Fig. 3. (See supplemental data for all
the total ion current chromatograms.) The product ions are recorded in
Table 2. Peaks corresponding to six metabolites (3-8) were observed
in the XICs for CYP3A4, and three of these metabolites (4, 6, and 8)
were formed in the presence of CYP2D6. The m/z for MHϩ of 3, 4,
and 5 (m/z 274) was 16 m/z units greater than 1H؉ (m/z 258); the m/z
for MHϩ of 6 (m/z 244) was 14 m/z units less than that of 1H؉; and
the m/z for MHϩ of 7 and 8 (m/z 260) was 2 m/z units greater than
1H؉. The suggested metabolic pathways of levorphanol with
CYP3A4 and CYP2D6 are summarized in Scheme 2.
Based on the principal product ions, the structure of the ion with
MHϩ at m/z 274.1799 is assigned to the C(8)-hydroxylated species 3
(Scheme 2). The product ion present at m/z 256.1709, corresponding
to loss of water, supports an aliphatic, as opposed to aromatic, site of
hydroxylation. The proposed position of the hydroxylation at C(8) is
based on the structures of the product ions at m/z 215.1067 (xii؉),
173.0596 (xiii؉), and 157.0659 (v؉ via xi؉) (see Supplemental Data
Scheme S2 for details).
assignments and attempts to rationalize the formation of the major
product ions are detailed in the supplemental data.
Two informative product ions present in this spectrum, at m/z
201.1277 (C14H17Oϩ) and 199.1120 (C14H15Oϩ), are likely to cor-
respond to ii؉ and iii؉, respectively. Pathways to rationalize the
formation of these product ions (Scheme 1) start with ring-opening
and proton rearrangements of 1H؉. Migration of a proton from the
N-methyl group to C(9) with simultaneous ring opening leads to 1a؉
that loses ethene and formaldimine to yield ii؉ [path (a)]. Alterna-
tively, cleavage of the C(9)-N bond accompanied by proton migration
from the benzylic C(10) atom to nitrogen leads to ib؉. Subsequent
fragmentation with neutral losses of ethene and methylamine leads to
iii؉ [path (b)]. The product ions observed in this spectrum are also
present in the product ion spectrum of protonated levallorphan (see
below). Furthermore, the corresponding spectra obtained from several
TABLE 1
Comparison of calculated and experimentally determined masses for the product
ions observed in the product ion spectrum of protonated levorphanol (1H؉)
Found
Mass
Suggested
Formula
Difference
(ppm)
Product Iona
Exact Mass
The m/z values of the ions observed in the product ion spectrum of
the compound assigned the structure 4H؉ correspond to those ob-
served in the product ion spectrum of 1H؉ with the addition of 16 m/z
units. The structure of this metabolite is suggested to be a dihydroxy-
phenyl species (4 in Scheme 2, unknown regiochemistry) because the
product ion spectrum shows no loss of water. Proposed structures of
the other product ions are consistent with this assignment (see Sup-
plemental Data Scheme S3).
iaϩ and ibϩ
258.1842
201.1277
199.1120
159.0813
157.0646
145.0659
133.0647
C
C
C
C
C
C
17H24NOϩ
14H17Oϩ
14H15Oϩ
11H11Oϩ
11H9Oϩ
258.1858
201.1279
199.1123
159.0810a
157.0653a
145.0653a
133.0653a
6.2
1.0
1.5
1.9
4.5
4.1
4.5
iiϩ
iiiϩ
ivϩ
vϩ
viiϩ
ixϩ
10H9Oϩ
C9H9Oϩ
a Proposed structures and fragmentation pathways to rationalize product ions with these
formulas will be found in the supplemental data.
H
HNCH2
C2H4
CH2
HN+
H
(a)
H
H
H+
H
HO
C+
NH+
HO
HO
ia+
ii+
C14H17O+
Exact mass: 201.1279
Found: 201.1277 (1.0 ppm)
1H+
C17H24NO+
Exact mass: 258.1858
Found: 258.1842 (6.2 ppm)
H2NCH3
C2H4
H
CH3
HN+
H
(b)
H+
H
H
HO
C+
NH+
H
HO
HO
ib+
1H+
iii+
C14H15O+
Exact mass: 199.1123
Found: 199.1120 (1.5 ppm)
SCHEME 1. Pathways proposed to account for the product ions at m/z 201.1277 (ii؉) and 199.1120 (iii؉).

EXPLORATION OF CATALYTIC PROPERTIES OF CYP2D6 AND CYP3A4
191
The product ion spectrum of the third metabolite assigned with retention time of this metabolite compared with that of the parent.
MHϩ 274.1793 shows the same ions as observed with 1H؉ together Both 3 and 5 may be unresolved diastereomeric mixtures.
with two additional ions at m/z 257.1777 and m/z 256.1703. The ion
The metabolite with MHϩ at m/z 244.1696 is a compound with 14 m/z
at m/z 256.1703, corresponding to loss of water, is consistent with a units less than that of protonated levorphanol (1H؉). This loss of 14 m/z
second aliphatic hydroxylation product. However, the ion at m/z units corresponds to loss of a methylene group and suggests that the
257.1777 corresponds to a loss of 17 m/z units from the parent. The structure of this metabolite is the secondary amine (6 in Scheme 2). This
only reasonable assignment for a neutral loss of 17 m/z units from this proposal is supported by the observed product ions (Table 2) that corre-
system is the hydroxyl radical (ꢀOH). Loss of ꢀOH suggests that the spond to those found in the product ion spectrum of 1H؉ and all of which
oxidation of 1 has taken place on a position that can fragment to give involve pathways (a) and (b) outlined in Scheme 1.
a relatively stable radical cation. Therefore, the suggested structure of
The remaining two metabolites have an MHϩ at m/z 260.1636,
this metabolite is the N-oxide (5 in Scheme 2) (see Supplemental Data 2 m/z units greater than that of 1H؉. Based on the proposed
Scheme S4 for details). This assignment is supported by the longer structures of 3, 4, and 6, these metabolites are likely to result from
FIG. 3. XICs obtained from extracts of incubation mixtures of levorphanol (10 ␮M) in the presence of NADPH-supplemented recombinant CYP3A4 (black) and CYP2D6
(red).

192
BONN ET AL.
compound may be the N-demethylated dihydroxyphenyl species 8
TABLE 2
ϩ
(Scheme 2).
Molecular mass, responsible P450, and MH -generated product ions of
levorphanol (1), levallorphan (2), and their metabolites
Because no authentic standards were available, only approximate
estimates of the quantities of the metabolites formed, from integrated
ion intensities, were possible. Even if this is not a proper way of
quantification it gives an opportunity to compare the distribution of
metabolites for the compound studied. The MS response area and a
calculation of the relative MS area (%MS) of all the metabolites and
the parent are presented in Table 3. CYP3A4 proved to be a more
effective catalyst with approximately 43% of the substrate converted
to six metabolites, including 20% conversion to the aliphatic carbinol
3 as a major route. CYP2D6 was less active (33% conversion), with
aromatic hydroxylation to give 4 as the dominant pathway (27%). The
Clint values at 1 ␮M were Ͻ0.2 ␮l/min/pmol P450 (t1/2 Ͼ40 min) in
both CYP2D6 and CYP3A4, indicating a low turnover for the com-
pound in the two enzymes.
Identification of Levallorphan Metabolites. The product ion
spectrum of protonated levallorphan (2H؉ at m/z 284.1986; see Sup-
plemental Data Fig. S9) was essentially identical to that of protonated
levorphanol (1H؉) shown in Fig. 2. This behavior makes clear that
collision-induced dissociation of 1H؉ and 2H؉ proceeds via the same
two initial fragmentation reactions [(a) and (b)] shown for 1H؉ in
Scheme 1. The only unique product ions formed from 2H؉ are the
ring-opened ions xxxa؉ and xxxb؉ (Scheme 3) that replace ia؉ and
ib؉ formed from 1H؉.
MHϩ at
Compound
P450
Product Ions Ͼ10% Relative Abundance
m/z
Levorphanol (1)
258
274
274
274
244
260
260
284
244
300
300
316
316
258
201, 199, 159, 157,a 145, 133
256, 215, 199, 197, 173, 157,a 145
217, 215, 175, 173,a 161, 149
257, 256, 201, 199, 159, 157,a 145, 133
201, 199, 159, 157,a 145, 133
242, 215, 199, 173, 157,a 145
202, 200, 158,a 157, 134, 133
201, 199, 159, 157,a 145, 133
201, 199, 159, 157,a 145, 133
282, 215, 199, 197, 173, 157,a 145
217, 215, 175, 173,a 161, 149
298, 231, 215,a 213, 189, 173, 161
231, 215, 213, 197,a 169
3
4
5
6
7
8
3A4
3A4/2D6
3A4
3A4/2D6
3A4
3A4/2D6
Levallorphan (2)
6
9
10
11
12
13
3A4/2D6
3A4
3A4/2D6
3A4/2D6
3A4
3A4
215, 187, 159, 146, 145,a 133, 107
a Base peak.
a combination of N-demethylation (Ϫ14 m/z units) and hydroxy-
lation (ϩ16 m/z units). The product ions formed from the proton-
ated 7 species resemble those observed from 3H؉, suggesting that
this is the N-demethylated metabolite of 3 (7 in Scheme 2). The
product ions at m/z 242.1613 (loss of water), 215.1097 (losses of
ethene and methylamine), and 173.0591 (subsequent loss of cyclo-
propane) are consistent with structure 7. The product ion spectrum
of the second species with MHϩ at m/z 260 was very weak.
Incubations of Levallorphan with CYP3A4 and CYP2D6. The
Because there is no evidence for loss of water in the spectrum, this total ion current chromatograms (TICs) of incubation extracts ob-
H
H
N
N
H
3A4
H
OH
HO
HO
7
6
3A4
2D6
O-
N+
N
N
3A4
H
H
3A4
H
OH
HO
HO
HO
5
Levorphanol (1)
3
3A4
2D6
HO
H
N
H
HO
N
3A4
2D6
H
HO
HO
4
8
SCHEME 2. Proposal for the in vitro metabolic pathways of levorphanol (1) catalyzed by CYP3A4 and CYP2D6.

EXPLORATION OF CATALYTIC PROPERTIES OF CYP2D6 AND CYP3A4
193
TABLE 3
that of levallorphan, suggesting formation of hydroxylated metabo-
lites. The structure proposed for 9 is the C(5)-cyclohexylcarbinol (9,
Scheme 4). This assignment is based on the same analysis that led to
the assignment of 3 for the levorphanol metabolite because the prod-
uct ions of 9H؉ correspond to the product ions observed for 3H؉. In
particular, the loss of 18 m/z units (water) to give m/z 282.1799
supports the proposed aliphatic hydroxylation site. The product ion
spectrum of the second ϩ 16 metabolite displays ions corresponding
to those observed for 4, the aromatic hydroxylated metabolite of
levorphanol. That leads to structure 10 for this metabolite (Scheme 4).
Compounds assigned structures 11 and 12 (MHϩ at m/z 316.1920)
have a mass difference compared with levallorphan of ϩ32 m/z units,
indicating bis-hydroxylations. The MS2-generated product ions of 11
correspond to the product ions observed with the carbinol metabolite
9 of levallorphan except for the increase of 16 Da for the second
hydroxyl group. The presence of a product ion at m/z 161.0612 in the
MS2 of 11, corresponding to v؉ (see Supplemental Data Scheme S2)
with the addition of 16 m/z units, suggests that the second hydroxy-
lation is positioned in the aromatic ring [C(10) or C(13)] (11, Scheme
4). The proposed structure of 12 is shown in Scheme 4. This com-
pound is also a bis-hydroxylated metabolite of levallorphan that, in
this case, seems to involve two oxidations of the cyclohexyl ring (ring
C). This conclusion is based on comparisons of the product ions
present in the MS2 tracing of 12 with those formed in the MS2 of 9
and 10 (see Supplemental Data Scheme S7 for details).
MS response areas and %MS of levorphanol, levallorphan, and the observed
metabolites after 60-min incubations with recombinant CYP2D6 and CYP3A4
MS Response Area
%MSa
(107 cps)
Compound
3A4
2D6
3A4
2D6
67
Levorphanol (1)
7.40
2.54
0.40
0.77
0.72
0.16
0.39
5.59
1.90
2.41
1.29
0.819
0.53
0.417
7.91
3.15
57
20
3.1
6
5.6
1.2
3.1
39
13
17
9
3
4
5
6
7
8
27
0.41
0.41
3.4
0.38
7.71
0.299
3.2
84
3.2
Levallorphan (2)
6
9
10
11
12
13
0.240
0.52
2.6
5.6
5.7
3.7
2.9
a %MS calculated from MS response area divided by the summed areas of all the metabolites
and parent.
tained with levallorphan (2) in the presence of CYP3A4 revealed six
metabolite peaks; the corresponding chromatograms with CYP2D6
showed only three metabolite peaks, all of which were present in the
CYP3A4 tracing. The XICs of these analytes are shown in Fig. 4.
The retention time and product ion spectra of the CYP3A4- and
CYP2D6-generated species observed at m/z 244 were identical to
those of the N-demethylated metabolite 6 of levorphanol. Conse-
quently, this compound must be the N-deallylated metabolite of le-
vallorphan (6, Scheme 4). The other five levallorphan metabolites
were identified as 9 through 13 (see below). In addition to 6, com-
pounds 10 and 11 were common metabolites of CYP3A4 and
CYP2D6. The ions obtained from the product ion spectra are reported
in Table 2.
The compound assigned structure 13 has MHϩ at m/z 258.1472,
26 m/z units less than the parent 2H؉. This mass corresponds to
N-deallylation and hydroxylation of the parent followed by further
oxidation of a carbinol group to a carbonyl group. The MS2
fragmentation pattern differs from those of other compounds in the
series. This leads to the suggestion that the carbonyl group is positioned
␣ to the nitrogen atom resulting in the lactam 13 (Scheme 4). This
proposal is supported by the observed product ions that are discussed in
A summary of the metabolic pathways of levallorphan including detail in the supplemental data (Scheme S8).
structures of metabolites is depicted in Scheme 4. Metabolites 9 and
In addition to the aforementioned metabolites, three weak signals
10 have MHϩ at m/z 300.1934/300.1955, 16 m/z units greater than were observed at m/z 314 (data not shown). The mass difference of 30
H
(a)
H
H+
NH+
HO
H
HN+
xxxa+
H
HO
SCHEME 3. Formation of product ions xxxa؉ and
xxxb؉ from 2H؉.
2H+
(b)
H
H+
NH+
H
HO
xxxb+
C19H26NO+
Exact mass: 284.2014
Found: 284.1986 (9.9 ppm)

194
BONN ET AL.
FIG. 4. XICs obtained from extracts of in-
cubation mixtures of levallorphan (10 ␮M)
in the presence of NADPH-supplemented
recombinant CYP3A4 (black) and CYP2D6
(red).
m/z units compared with the parent compound suggests double hy-
Time-Dependent Inhibition. The two test compounds also were
droxylations and further oxidation to a carbonyl group. No structural analyzed for their potential TDI properties. After preincubation with
assignments are offered for these three secondary metabolites. each test compound in the presence and absence of NADPH, the
The absolute MS response area for each metabolite of levallorphan enzyme activity was measured in incubations with the probe sub-
together with the %MS are presented in Table 3. The main species strates AMMC (CYP2D6) and BFC (CYP3A4). The results for lev-
detected in the HPLC/MS analysis is the parent drug, particularly in allorphan showed a decrease in CYP2D6 activity at higher concen-
the CYP2D6 incubations. Based on the %MS, the major metabolite trations both in samples preincubated with and without NADPH. This
formed with CYP3A4 results from hydroxylation in the cyclohexyl result suggests that levallorphan is a strong, reversible inhibitor. No
ring. The three CYP2D6-generated metabolites appear to be formed in changes were obtained with the 1/20 dilution compared with the results
approximately equal amounts. The Clint values at 1 ␮M were Ͻ0.2 from the 1/10 dilution that are reported here. From the results of lev-
␮l/min/pmol P450 (t1/2 Ͼ40 min) in CYP2D6 and 0.82 ␮l/min/pmol allorphan incubated with CYP2D6, a concentration-dependent loss of
P450 (t1/2 8.5 min) in CYP3A4, indicating a higher turnover for activity was observed but it was not time-dependent. To diminish this
levallorphan in CYP3A4.
time-independent component the activity was normalized to the incuba-
IC50 Determinations. To determine the reversible inhibition po- tions carried out in the absence of NADPH. The normalized ratios (NRs)
tentials of levorphanol (1) and levallorphan (2) on CYP3A4 and were calculated according to the following equation:
CYP2D6, full IC50 curves were obtained. The IC50 values for levor-
͑Ainh/Actrl
͒
phanol were Ͼ50 ␮M in the CYP3A4 assay and 11.5 ␮M in the
CYP2D6 assay. The corresponding values for levallorphan were Ͼ50
␮M in the CYP3A4 assay and 0.65 ␮M in the CYP2D6 assay. These
NR ϭ
͑Ainh/Actrl
͒
wo NADPH
results indicate that both compounds act as relatively strong reversible where Ainh is the catalytic activity after preincubation with inhibitor
inhibitors of CYP2D6, but neither of them shows any inhibition of present and Actrl is the activity in the control incubations without
CYP3A4.
inhibitor. A compound is considered to be a time-dependent inhibitor

EXPLORATION OF CATALYTIC PROPERTIES OF CYP2D6 AND CYP3A4
195
H
N
H
N
H
3A4
H
O
HO
HO
M11 (13)
M4 (6)
3A4
2D6
N
N
N
H
3A4
H
3A4
OH
H
HO
HO
OH
HO
Levallorphan (2)
M7 (9)
M10 (12)
OH
3A4
2D6
HO
N
H
HO
N
3A4
2D6
H
HO
OH
HO
M8 (10)
M9 (11)
SCHEME 4. Proposed in vitro metabolic pathways of levallorphan (2) catalyzed by CYP2D6 and CYP3A4.
if the NR is less than 0.7. This is an empirical value based on the mass of 2 ϩ 307 Ϫ H2 ϩ Hϩ, consistent with a glutathionyl conjugate
understanding that the value for a compound not acting as a time- of levallorphan that had undergone an initial 2-electron (net loss of
dependent inhibitor should be approximately 1.0. Allowing for an H2) oxidation (Scheme 5). The suggested structure, 16, reflects the
experimental error of 20%, it could be higher or lower. Hence there is proposed 1,4-addition reaction of GSH with the eneiminiumyl metab-
a gray zone for NR ϭ 0.7 to 0.9.
olite 15؉ formed by initial oxidation (hydroxylation followed by
The NRs for the two test compounds, after 20 min of preincubations, dehydration of the protonated carbinol 14H؉ oxygen atom) of the
together with those of the positive controls, are presented in Fig. 5. Complete carbon atom ␣ to the nitrogen atom of levallorphan. Less likely would
time curves are presented in the Supplemental Data Figs. S17 and be 1,2-addition because the terminal carbon atom of the eneiminiumyl
S18. From the NR values it was apparent that levallorphan acts as a moiety is sterically less hindered than the internal ␲-bond. In addition,
time-dependent inhibitor on CYP3A4 (NR 0.45 at 12.5 ␮M and NR the proposed 1,4-addition product 16 will be more stable than the
0.27 at 25 ␮M) but not on CYP2D6. Levorphanol, which lacks the corresponding 1,2-addition product because of the conjugation of the
allylic side chain, did not act as a time-dependent inhibitor on either nitrogen lone pair with the ␲-bond. The principal fragment ion ob-
CYP3A4 or CYP2D6.
served in the MS2 of MHϩ at m/z 589 was at m/z 282 as expected
It is noteworthy that levallorphan shows strong reversible inhibition because the glutathionyl group is in conjugation with the eneaminyl
(IC50 ϭ 0.65 ␮M) but no TDI of CYP2D6. The opposite results observed system and should readily eliminate following protonation of sulfur to
with CYP3A4 could indicate differences in how the two isoenzymes interact give the eneiminiumyl fragment (m/z 282) with glutathione as the
with this substrate. The difference in the TDI properties with CYP3A4 for neutral loss (Scheme 6). The second diagnostic fragment ion observed
levallorphan and levorphanol supports the involvement of the allylic side at m/z 460 corresponds to the expected neutral loss of pyroglutamic
chain of levallorphan in the mediation of the observed TDI.
(129 Da) that is characteristic of glutathionyl adducts (see Supple-
Studies with Glutathione. The presence of the allylic side chain in mental Data Scheme S9 for details).
levallorphan and the fact that this compound acts as a time-dependent
In addition to 16H؉ two minor peaks at 0.7 and 5.14 min are
inhibitor of CYP3A4 led to glutathione trapping studies that were present in both the CYP3A4 and CYP2D6 incubations. These peaks
designed to trap possible metabolically generated reactive species. have MHϩ at m/z 605, corresponding to a glutathionyl conjugate of
The result could indicate a possible relationship between the TDI and hydroxylated levallorphan (2 ϩ 305 ϩ 16 ϩ Hϩ). The MS2 tracings
the anticipated bioactivation of levallorphan to reactive species.
of these peaks were weak, and no structural assignments were at-
TICs obtained with levallorphan incubations in the presence of tempted. Possible metabolites could result from epoxidation of either
GSH and CYP3A4 or CYP2D6 are presented in Fig. 6. In MS2 the allyl or the aromatic moiety followed by conjugation with GSH.
tracings of the CYP3A4 incubations, a large peak with the retention These peaks are considered to be minor metabolites and are not
time 4.18 min at m/z 589 was observed. This mass corresponds to the considered further.

196
BONN ET AL.
FIG. 5. Normalized ratios from the TDI assay with CYP3A4 (a) and CYP2D6 (b). The values are averages of data obtained from two incubations with the S.D. represented
by error bars. The area between the dotted lines is considered to be a gray zone (see definition under Results). ء for troleandomycin the corresponding concentrations used
were 0.3, 2.5, and 5 ␮M. For propranolol the concentrations used were 0.6, 5, and 10 ␮M.
FIG. 6. TIC from the HPLC/MS experi-
ment of the incubation mixture of levallor-
phan (10 ␮M) supplemented with NADPH
and GSH with recombinant CYP3A4 (a)
and recombinant CYP2D6 (b).

EXPLORATION OF CATALYTIC PROPERTIES OF CYP2D6 AND CYP3A4
197
H
O+
H
N
CYP3A4
H
N
HO
H
HO
2
14H+
H2O
SCHEME 5. Proposed mechanism for the
formation of the glutathionyl conjugate 16
of levallorphan.
-SG
SG
N+
N
H
H
HO
HO
15+
16
gluthionyl conjugate of
levallorphan
Dockings of Levorphanol and Levallorphan into Crystal orientation of each docking pose are available in the Supplemental
Structures of CYP2D6 and CYP3A4. In CYP2D6 three poses each Data Table S1.
of levorphanol and levallorphan were obtained. Comparing the cal-
Interaction energies between each docking pose and the surround-
culated interaction energies, two poses are indicated to be the most ing amino acid residues were calculated with the energy calculation
favorable. These are presented in Fig. 7, a and b, and represent an method described previously (Kjellander et al., 2007). These data are
orientation of levallorphan and levorphanol with the aromatic ring/ available in Supplemental Data Tables S2 and S3.
phenol toward the heme. For levallorphan the distance between the
Discussion
positively charged nitrogen and the negatively charged Glu216 was
4.64 Å, and the distance to the oxy species in the heme was 3.55 Å.
The rationales for the present study included the possibility of exploring
For levorphanol these distances were 7.50 and 2.87 Å, respectively. The orientations of the two selected compounds, levallorphan and levorphanol, in
other two poses of levallorphan were with the allyl group pointing toward the active sites of CYP2D6 versus CYP3A4 in vitro and in silico. Our
the heme, and an additional pose is with the aromatic moiety toward the approach involved metabolic profiling studies on levallorphan and levorpha-
heme. Levorphanol also showed an additional pose with the aromatic ring nol with the key tool being the characterization of metabolites using recom-
closest to the heme and one with the pyridine ring pointing toward the binant enzymes. The inhibition potentials of levallorphan and levorphanol on
heme. Docking energies and information about the orientation of each CYP2D6 and CYP3A4 were also examined in an effort to assess their
docking pose are available in the Supplemental Data Table S1.
affinities for these enzymes. In addition, their TDI properties were explored
In contrast to the limited number of docking poses in CYP2D6, the in part because of the presence of the allylic side chain in levallorphan, a
dockings in CYP3A4 revealed 19 poses of levallorphan and levor- structural feature often associated with TDI (Fontana et al., 2005). We
phanol. The calculated interaction energies did not differ much be- anticipated that the results of these experiments would provide useful insights
tween these poses, and there was no obvious ranking among them. into how these compounds would be oriented in the enzyme active site
The poses represent orientations of the substrates presenting many cavities of the two P450 isoforms to give productive binding modes. Finally,
different parts of the molecule to the heme. Some examples are glutathionyl trapping experiments were performed on levallorphan to deter-
visualized in Fig. 8. In general, most of the levallorphan poses were mine whether the formation of glutathionyl conjugates correlated with the
far away from the heme compared with levorphanol. These poses observed TDI of this allylamine derivative.
might be favored by hydrophobic interactions with the phenyl alanine
CYP2D6 catalyzed only the aromatic hydroxylation and N-demeth-
cluster in CYP3A4. Docking energies and information about the ylation/N-deallylation pathways, consistent with an orientation of the
H
S+
GSH
N
N+
H
G
H
SCHEME 6. Proposed mechanism to account for the observed frag-
HO
ment xxxxxiii؉ in the MS2 of 16H؉. ء no accurate mass available
HO
for the GSH adduct and its product ion.
16H+
xxxxxiii+
C19H24NO+
Exact mass: 282.1858
Found: 282*

198
BONN ET AL.
FIG. 7. Levallorphan (a) and levorphanol (b) docked in active site cavity of
CYP2D6 crystal structure (Protein Data Bank code 2f9q).
FIG. 8. Levallorphan and levorphanol docked in active site cavity of CYP3A4
crystal structure (Protein Data Bank code 2j0d).
substrate with the aromatic moiety or the nitrogen atom pointing
toward the heme. These orientations were also those obtained from the
docking experiments. The orientation with the aromatic moiety to-
ward the heme was the most favorable with respect to calculated
interaction energies. The CYP2D6 substrate/inhibitor pharmacophore
is proposed to consist of interactions with a basic nitrogen atom of the
substrate and acidic amino acid residues of the enzyme (Glu216 and
Asp301) that fix the substrate at a defined distance from the site of
metabolism. In addition, a hydrophobic region is suggested to be
present near the site of oxidation. Site-directed mutagenesis studies
have documented the influence of the hydrophobic amino acids,
Phe120 and Phe483, on the regioselectivity of some CYP2D6 sub-
strates (Flanagan et al., 2004; Lussenburg et al., 2005). The impor-
tance of the interactions between the basic nitrogen and acidic amino
acid residues in the active site also have been supported by site-
ligand is rather far away from the acidic amino acid residue in the
active site. However, this picture is a static view of the protein, and in
reality the side chains of the amino acids are flexible and are allowed
to move in the active site.
Comparing the extent of metabolism of levorphanol and levallor-
phan with the selected isoenzymes made clear that both compounds
were more extensively metabolized by CYP3A4 than CYP2D6. The
CYP3A4 metabolites observed indicated that both compounds could
be oriented in different ways in the active site, resulting in the
presentation of different regions of the molecule to the heme.
CYP3A4 has a larger and flexible active site in which hydrophobic
interactions with the ligand will probably dominate (see Fig. 8). In
contrast to CYP2D6, N-dealkylation reactions are more common with
CYP3A4 as a result of an orientation with the nitrogen atom in the
directed mutagenesis studies (Guengerich et al., 2003; Paine et al., ligand oriented toward the heme (Coutts et al., 1994). These results
2003). In addition, Upthagrove and Nelson (2001) have shown a also are consistent with the large number of energetically expectable
relationship between substrate affinity and pKa of the substrate. From docking solutions obtained with CYP3A4.
this knowledge of CYP2D6 and its substrates it could be hypothesized
Besides metabolic profiling of compounds in different P450s, the
that an orientation of the ligand with the nitrogen pointing away from inhibition potentials of the compounds are of interest. P450 inhibition
the heme is favored, and consequently N-dealkylation is unlikely to can result from a nonproductive binding mode of a substrate and may
occur. The orientation with the aromatic ring in levorphanol/levallor- give an indication of the orientation in the active site. The potential to
phan toward the heme is consistent with the proposed CYP2D6 inhibit CYP2D6 and CYP3A4 in a reversible manner was determined
pharmacophore, whereas the orientation with the nitrogen toward the for levorphanol and levallorphan. From the resulting IC50 values
heme does not fit this analysis. However, considering the amount of neither of the two compounds proved to inhibit CYP3A4. On the other
metabolites formed, the N-dealkylation is only a minor pathway in hand, both compounds inhibited CYP2D6. If applying the CYP2D6
CYP2D6. In the best pose obtained from the dockings of levallorphan substrate/inhibitor pharmacophore to the interactions of these two
the basic nitrogen is 4.6 Å from the Glu216. In addition, Phe120 compounds, the preferred orientation would be with the oxygen atom
seems to interact with the hydrophobic core structure of the ligand. of the aromatic moiety pointing toward the heme and the nitrogen
For levorphanol a similar picture is obtained, but the nitrogen in the atom pointing away from the heme. Interactions between the basic

EXPLORATION OF CATALYTIC PROPERTIES OF CYP2D6 AND CYP3A4
199
nitrogen atom and the acidic amino acid residues in the active site determinants for ligand-enzyme affinity, whereas CYP2D6 has a more
cavity may favor a nonproductive orientation of the substrate that restricted active site with electrostatic interactions with the ligand as
could, in theory, result in enzyme inhibition. From the dockings, an important factor.
however, it is difficult to draw any conclusions about the inhibitory or
Acknowledgments. We thank Ismael Zamora, Professor Kristina
productive binding modes of the ligands. In this case the high affinity
Luthman, and Richard Thompson for valuable discussions and critical
of levallorphan for CYP2D6, reflected by its low IC50 value, might be
reading of the manuscript.
the result of hydrophobic interactions between the allyl chain and the
protein, as well as electrostatic interactions with the acidic amino
acids. In CYP3A4, on the other hand, both compounds were exten-
sively metabolized but did not inhibit the substrate in a reversible
manner. This result is clearly consistent with the different character of
the active sites of CYP2D6 and CYP3A4 as discussed above.
The potential TDI properties of levallorphan and levorphanol also
References
Benetton SA, Borges VM, Chang TK, and McErlane KM (2004) Role of individual human
cytochrome P450 enzymes in the in vitro metabolism of hydromorphone. Xenobiotica 34:
335–344.
Coutts RT, Su P, and Baker GB (1994) Involvement of CYP2D6, CYP3A4, and other cyto-
chrome P-450 isozymes in N-dealkylation reactions. J Pharmacol Toxicol Methods 31:177–
186.
de Groot MJ, Bijloo GJ, Martens BJ, van Acker FA, and Vermeulen NP (1997) A refined
substrate model for human cytochrome P450 2D6. Chem Res Toxicol 10:41–48.
De Rienzo F, Fanelli F, Menziani MC, and De Benedetti PG (2000) Theoretical investigation of
were examined. As anticipated, levorphanol, which lacks the allylic
side chain, did not show any TDI properties, whereas the allylaminyl
substrate specificity for cytochromes P450 IA2, P450 IID6 and P450 IIIA4. J Comput Aided
Mol Des 14:93–116.
difference in the TDI properties of levallorphan with the two isoen-
Ekroos M and Sjo¨gren T (2006) Structural basis for ligand promiscuity in cytochrome P450 3A4.
substrate levallorphan proved to be a time-dependent inhibitor. A
zymes should give an indication of binding modes in CYP2D6 and
CYP3A4 because the assay reports on the TDI properties of the test
compound that presumably involve the formation of reactive species.
That is, a metabolite is formed that in turn can form adducts with the
protein. The mechanism of inactivation is not considered here, only
the formation of the reactive metabolite. Studies on the TDI properties
established that levallorphan inactivated CYP3A4 but not CYP2D6,
suggesting that a reactive species is formed in the presence of
CYP3A4 but not in the presence of CYP2D6. From the glutathionyl
trapping experiment a large peak, corresponding to a glutathionyl
adduct (16) at the allylic side chain of levallorphan, was present in
incubations with CYP3A4. The suggested adduct position of the
glutathionyl moiety is consistent with that previously reported in the
literature for allylamines that cause time-dependent inactivation of
P450 (Fontana et al., 2005). The absence of any large peak corre-
sponding to a glutathionyl adduct of levallorphan in CYP2D6 incu-
bations correlates with the lack of TDI properties of levallorphan on
this enzyme. This is in agreement with the discussion that the envi-
ronment in the active site cavity of CYP2D6 orients the ligands with
the basic nitrogen pointing away from the heme.
Proc Natl Acad Sci U S A 103:13682–13687.
Flanagan JU, Mare´chal JD, Ward R, Kemp CA, McLaughlin LA, Sutcliffe MJ, Roberts GC,
Paine MJ, and Wolf CR (2004) Phe120 contributes to the regiospecificity of cytochrome P450
2D6: mutation leads to the formation of a novel dextromethorphan metabolite. Biochem J
380:353–360.
Fontana E, Dansette PM, and Poli SM (2005) Cytochrome p450 enzymes mechanism based
inhibitors: common sub-structures and reactivity. Curr Drug Metab 6:413–454.
Goodford PJ (1985) A computational procedure for determining energetically favorable binding
sites on biologically important macromolecules. J Med Chem 28:849–857.
Guengerich FP (2003) Cytochromes P450, drugs, and diseases. Mol Interv 3:194–204.
Guengerich FP, Hanna IH, Martin MV, and Gillam EM (2003) Role of glutamic acid 216 in
cytochrome P450 2D6 substrate binding and catalysis. Biochemistry 42:1245–1253.
Hutchinson MR, Menelaou A, Foster DJ, Coller JK, and Somogyi AA (2004) CYP2D6 and
CYP3A4 involvement in the primary oxidative metabolism of hydrocodone by human liver
microsomes. Br J Clin Pharmacol 57:287–297.
Kirkwood LC, Nation RL, and Somogyi AA (1997) Characterization of the human cytochrome
P450 enzymes involved in the metabolism of dihydrocodeine. Br J Clin Pharmacol 44:549–
555.
Kjellander B, Masimirembwa CM, and Zamora I (2007) Exploration of enzyme-ligand interac-
tions in CYP2D6 & 3A4 homology models and crystal structures using a novel computational
approach. J Chem Inf Model 47:1234–1247.
Koymans L, Vermeulen NP, van Acker SA, te Koppele JM, Heykants JJ, Lavrijsen K, Meul-
dermans W, and Donne´-Op den Kelder GM (1992) A predictive model for substrates of
cytochrome P450-debrisoquine (2D6). Chem Res Toxicol 5:211–219.
Lewis DF (1999) Homology modelling of human cytochromes P450 involved in xenobiotic
metabolism and rationalization of substrate selectivity. Exp Toxicol Pathol 51:369–374.
Lewis DF, Eddershaw PJ, Goldfarb PS, and Tarbit MH (1997) Molecular modelling of cyto-
chrome P4502D6 (CYP2D6) based on an alignment with CYP102: structural studies on
specific CYP2D6 substrate metabolism. Xenobiotica 27:319–339.
Lussenburg BM, Keizers PH, de Graaf C, Hidestrand M, Ingelman-Sundberg M, Vermeulen NP,
and Commandeur JN (2005) The role of phenylalanine 483 in cytochrome P450 2D6 is
strongly substrate dependent. Biochem Pharmacol 70:1253–1261.
Narimatsu S, Arai T, Masubuchi Y, Horie T, Hosokawa M, Ueno K, Kataoka H, Yamamoto S,
Ishikawa T, and Cho AK (2001) Inactivation of rat cytochrome P450 2D enzyme by a further
metabolite of 4-hydroxypropranolol, the major and active metabolite of propranolol. Biol
Pharm Bull 24:988–994.
Paine MJ, McLaughlin LA, Flanagan JU, Kemp CA, Sutcliffe MJ, Roberts GC, and Wolf CR
(2003) Residues glutamate 216 and aspartate 301 are key determinants of substrate specificity
and product regioselectivity in cytochrome P450 2D6. J Biol Chem 278:4021–4027.
Poulos TL, Finzel BC, and Howard AJ (1987) High-resolution crystal structure of cytochrome
P450cam. J Mol Biol 195:687–700.
Rowland P, Blaney FE, Smyth MG, Jones JJ, Leydon VR, Oxbrow AK, Lewis CJ, Tennant MG,
Modi S, Eggleston DS, et al. (2006) Crystal structure of human cytochrome P450 2D6. J Biol
Chem 281:7614–7622.
von Moltke LL, Greenblatt DJ, Grassi JM, Granda BW, Venkatakrishnan K, Schmider J,
Harmatz JS, and Shader RI (1998) Multiple human cytochromes contribute to biotransforma-
tion of dextromethorphan in-vitro: role of CYP2C9, CYP2C19, CYP2D6, and CYP3A.
J Pharm Pharmacol 50:997–1004.
Williams PA, Cosme J, Vinkovic DM, Ward A, Angove HC, Day PJ, Vonrhein C, Tickle IJ, and
Jhoti H (2004) Crystal structures of human cytochrome P450 3A4 bound to metyrapone and
progesterone. Science 305:683–686.
Yamazaki H and Shimada T (1998) Comparative studies of in vitro inhibition of cytochrome
P450 3A4-dependent testosterone 6beta-hydroxylation by roxithromycin and its metabolites,
troleandomycin, and erythromycin. Drug Metab Dispos 26:1053–1057.
Yano JK, Wester MR, Schoch GA, Griffin KJ, Stout CD, and Johnson EF (2004) The structure
of human microsomal cytochrome P450 3A4 determined by X-ray crystallography to 2.05-A
resolution. J Biol Chem 279:38091–38094.
In summary, studies of the metabolism and inhibition potentials of
levorphanol and levallorphan with CYP2D6 and CYP3A4 have led to
information on how these compounds interact with these two isoen-
zymes of P450. Rationalizations of the fragmentation pathways ob-
served in the product ion spectra for the metabolites formed from
levorphanol and levallorphan have provided an opportunity to char-
acterize the orientations of these molecules in the respective enzyme
active sites. From the metabolites formed it may be concluded that
substrates may have several productive orientations in the active site
of CYP3A4. In CYP2D6, on the other hand, the compounds orient
with the aromatic moiety or the nitrogen atom pointing toward the
heme. The N-dealkylated metabolites, not following the CYP2D6
pharmacophore, were only minor compared with those formed in
CYP3A4. Furthermore, the TDI properties observed with levallorphan
and the formation of a glutathionyl conjugate of this compound with
CYP3A4 but not with CYP2D6 strongly indicate that levallorphan
adopts a productive binding mode in CYP3A4 that is not allowed in
CYP2D6. These results support the hypothesis that the acidic amino
acid residues in the CYP2D6 active site cavity favor an orientation of
the substrate with the nitrogen pointing away from the heme, whereas
orientations with the nitrogen atom toward the heme are commonly
observed in CYP3A4. These results are also consistent with the
dockings and support previous conclusions reached from computa-
tional studies of the enzymes in which CYP3A4 is suggested to have
a more flexible active site with hydrophobic interactions as major
Yue QY and Sa¨we J (1997) Different effects of inhibitors on the O- and N-demethylation of
codeine in human liver microsomes. Eur J Clin Pharmacol 52:41–47.
Address correspondence to: Britta Bonn, Discovery DMPK, AstraZeneca
R&D Mo¨ lndal, SE-43181 Mo¨ lndal, Sweden. E-mail: kjelland@chem.gu.se