0022-3565/02/3002-399–407$3.00 THE JOURNAL OF PHARMACOLOGY AND EXPERIMENTAL THERAPEUTICS Vol. 300, No. 2 Copyright © 2002 by The American Society for Pharmacology and Experimental Therapeutics 4360/960505 JPET 300:399–407, 2002 Printed in U.S.A.

Amodiaquine Clearance and Its Metabolism to N- Desethylamodiaquine Is Mediated by CYP2C8: A New High Affinity and Turnover -Specific Probe

XUE-QING LI, ANDERS BJORKMAN,¨ TOMMY B. ANDERSSON, MARIANNE RIDDERSTROM,¨ and COLLEN M. MASIMIREMBWA Drug Metabolism and Pharmacokinetics and Bioanalytical Chemistry, AstraZeneca Research and Development, Mo¨lndal, Sweden (X.-Q.L., T.B.A., M.R., C.M.M.); and Unit of Infectious Diseases, Karolinska Institute Hospital, Stockholm, Sweden (X.-Q.L., A.B., C.M.M.) Received July 25, 2001; accepted October 5, 2001 This paper is available online at http://jpet.aspetjournals.org Downloaded from

ABSTRACT Amodiaquine (AQ) metabolism to N-desethylamodiaquine activity factor method. Correlation analyses between AQ me- (DEAQ) is the principal route of disposition in humans. Using tabolism and the activities of eight hepatic P450s were made

human liver microsomes and two sets of recombinant human on 10 different HLM samples. Both the formation of DEAQ and jpet.aspetjournals.org isoforms (from lymphoblastoids and yeast) the clearance of AQ showed excellent correlations (r2 ϭ 0.98 we performed studies to identify the CYP isoform(s) involved in and 0.95) with 6␣-hydroxylation of , a marker sub- the metabolism of AQ. CYP2C8 was the main hepatic isoform strate for CYP2C8. The inhibition of DEAQ formation by quer-

that cleared AQ and catalyzed the formation of DEAQ. The cetin was competitive with Ki values of 1.96 for CYP2C8 and extrahepatic P450s, 1A1 and 1B1, also cleared AQ and cata- 1.56 ␮M for HLMs. Docking of AQ into the homology

lyzed the formation of an unknown metabolite M2. The Km and models of the CYP2C isoforms showed favorable interactions ␮ Vmax values for AQ N-desethylation were 1.2 M and 2.6 pmol/ with CYP2C8, which supported the likelihood of an N-desethy- min/pmol of CYP2C8 for recombinant CYP2C8, and 2.4 ␮M lation reaction. These data show that CYP2C8 is the main at ASPET Journals on September 30, 2017 and 1462 pmol/min/mg of for human liver microsomes hepatic isoform responsible for the metabolism of AQ. The (HLMs), respectively. Relative contribution of CYP2C8 in the specificity, high affinity, and high turnover make AQ desethy- formation of DEAQ was estimated at 100% using the relative lation an excellent marker reaction for CYP2C8 activity.

Amodiaquine (AQ) is a 4-aminoquinoline derivative that DEAQ) (Churchill et al., 1985, 1986; Mount et al., 1986). has been widely used for treatment of malaria over the past Whereas the formation of DEAQ is rapid, its elimination is 50 years. It is intrinsically more active than the other 4-ami- very slow with a terminal half-life of over 100 h (Winstanley noquinoline, chloroquine, against Plasmodium falciparum et al., 1987; Laurent et al., 1993). AQ and DEAQ both have parasites, which are moderately chloroquine resistant. The antimalarial activity, but AQ is 3 times more active drug is therefore increasingly being considered as a replace- (Churchill et al., 1985). However, since AQ is rapidly cleared ment for chloroquine as a first line drug in Africa because of and the formed DEAQ attains high plasma concentrations for widespread chloroquine resistance. Because of major side a long time, AQ is considered a prodrug, which is bioactivated effects, mainly agranulocytosis, observed during prophylactic to DEAQ. use of the drug, AQ is now only recommended for treatment In vivo pharmacokinetic studies have shown that the pri- of malaria, for which no serious cases of toxicity have been mary route of systemic elimination of AQ in humans is via reported (Laurent et al., 1993). extensive first-pass biotransformation to the active DEAQ Upon oral administration, AQ is rapidly absorbed and ex- (White et al., 1987; Laurent et al., 1993). Jewell et al. (1995) tensively metabolized such that very little of the parent drug found that DEAQ and bisDEAQ were formed by human liver is detected in the plasma. The main metabolite of AQ is N-desethylamodiaquine (DEAQ) with other minor metabo- microsomes and postulated that the liver was the major site lites being 2-hydroxyl-DEAQ and N-bisdesethylAQ (bis- of AQ first-pass metabolism. Studies with isolated neutro- phils and lymphocytes showed that agranulocytosis caused

X.-Q.L. is a recipient of the Wenner-Gren Foundation postdoctoral fellow- by AQ might be due to metabolism to a reactive quinoneimine ship (Stockholm, Sweden). (Naisbitt et al., 1998). All the pharmacokinetic studies on AQ

ABBREVIATIONS: AQ, amodiaquine; DEAQ, N-desethylamodiaquine; bisDEAQ, N-bisdesethylamodiaquine; P450, cytochrome P450; HLM,

human liver microsome; RAF, relative activity factor; LC, liquid chromatography; HPLC, high-performance LC; MS, mass spectroscopy; tR, retention time. 399 400 Li et al. and its major metabolite, DEAQ, show that there is a great (Hewlett-Packard, Palo Alto, CA), and a Finnigan LCQ ion trap mass interindividual variability in kinetic parameters (Cmax, t1/2, spectrometer (ThermoFinnigan MAT, San Jose, CA) employing an and area under the curve). Such variability probably reflects atmospheric pressure ionization interface. Chromatography was per- ϫ ␮ variability in metabolic capacity and could imply different formed on a Symmetry C18 column (3.9 150 mm i.d.; 5 m, Waters, therapeutic and toxicological responses to AQ. Milford, MA). Compounds were eluted with a linear gradient of acetonitrile (5–40%, v/v, over 7 min) in 5 mM ammonium acetate Besides the report by Jewell et al. (1995) on the possible using formic acid to bring the pH to 3.3, at a flow rate of 1 ml/min. role of CYP3A4 in AQ metabolism based on inhibition studies The effluent was split with approximately 0.3 ml/min introduced into with ketoconazole, no detailed study has been performed to the mass spectrometer. Source parameters of mass spectrometer identify P450s that metabolize AQ. The aim of this study was (e.g., spray voltage, temperature, gas flow rates, etc.) were individ- therefore to investigate the metabolic clearance of AQ and ually optimized for each compound, and the MS/MS spectra were identify the cytochromes P450 responsible for its metabolism. obtained for precursor ions through incidental collision with neutral Our results show that the clearance of AQ and its hepatic gas (Helium) molecules in the ion trap. The metabolites of AQ were metabolism to DEAQ is catalyzed mainly by CYP2C8. Molec- analyzed on-line, qualitatively by ion trap-based MS, and quantita- ular modeling studies also demonstrated the substrate spec- tively by diode array detector detection set at 342 nm. Instrument ificity of CYP2C8 for AQ metabolism. control, data acquisition, and data evaluation were performed using Xcalibur software (version 1.2, ThermoFinnigan MAT). Materials and Methods Contribution of CYP2C8 to HLM AQ Metabolism and Esti- mation of in Vivo Drug Clearance. The percentage contribution Downloaded from Chemicals. Amodiaquine dihydrochloride and desethylamodia- of CYP2C8 to AQ N-desethylation was estimated by applying the quine hydrochloride were obtained from Karolinska Institute (Stock- relative activity factor (RAF) values as proposed by Crespi (1995) holm, Sweden). dihydrate was obtained from Ald- using the values of the activities (RAFv). The RAFv of CYP2C8 was rich Chemical Co. (Milwaukee, WI). Ketoconazole was purchased determined as the ratio of the activity of paclitaxel 6␣-hydroxylation from Janssen Biotech (Flander, NJ). NADP and glucose 6-phosphate (at the substrate concentration of 100 ␮M), a specific metabolic were obtained from Sigma Chemical Co. (St. Louis, MO). Glucose-6- reaction mediated by CYP2C8 (Rahman et al., 1994), in HLMs to the jpet.aspetjournals.org phosphate dehydrogenase was purchased from ICN Pharmaceuticals activity for the recombinant CYP2C8. Using RAFv, the N-desethyla- Biochemicals Division (Aurora, OH). Paclitaxel and 6␣-hydroxypa- tion clearance of AQ by CYP2C8 in HLMs was calculated using clitaxel were obtained from Gentest (Woburn, MA). All other re- equations described previously (Nakajima et al., 1999). agents used were of analytical or HPLC grade. Estimation of in vivo clearance was made from in vitro data Human Liver Microsomes (HLMs) and Recombinant Cyto- according to a venous equilibrium model (well stirred model) using chromes P450 (rCYP). Ten HLMs were obtained from an in-house equations described previously (Houston, 1994; Obach et al., 1997). bank of liver microsomes maintained at AstraZeneca Research and The liver was taken as the main site of drug metabolic clearance. Development (Mo¨lndal, Sweden) (A¨ belo¨ et al., 2000), and the CYP Previous studies have shown that AQ does not bind to human mi- at ASPET Journals on September 30, 2017 activities (CYP1A2, 2A6, 2C8, 2C9, 2C19, 2D6, 2E1, and 3A4) in crosomal (Jewell et al., 1995), and DEAQ but not AQ does individual HLMs were determined using diagnostic marker sub- accumulate in blood, preferentially in lymphocytes at an average strates (C. M. Masimirembwa, M. E. S. Lutz, R. A. Thompson, and T. blood/plasma distribution ratio of 3.35 (Pussard et al., 1987; Laurent B. Andersson, submitted for publication). The following marker re- et al., 1993). The clearance we predicted in vitro was compared to the actions were used: phenacetin demethylation (CYP1A2), coumarin in vivo clearance obtained using plasma concentrations of AQ. 7-hydroxylation (CYP2A6), paclitaxel 6␣-hydroxylation (CYP2C8), Michaelis-Menten and Inhibition Kinetics. Linear conditions 4-hydroxylation (CYP2C9), S-mephenytoin 4Ј-hydroxy- for the formation of metabolites were established with respect to lation (CYP2C19), bufuralol 1-hydroxylation (CYP2D6), chlorzoxa- protein content and incubation time for the HLM and for CYP2C8. zone 6-hydroxylation (CYP2E1), and midazolam 1-hydroxylation The optimum protein concentrations of HLMs and CYP2C8 for ki- (CYP3A4). The 10 HLM samples were selected from a bank of 21 netic analysis were 0.1 mg/ml and 3 pmol of P450/incubation, respec- HLM samples, ensuring that there was minimal cross P450 activity tively. The formation rates of DEAQ were linear at 35°C for incuba- correlations and that the activities of each P450 covered a wide tion times of up to 30 min. The Michaelis-Menten kinetics of AQ range. Pooled HLMs were prepared from a pooled set of liver N-desethylation by HLMs and CYP2C8 were determined using nine pieces of patients undergoing liver resections. Recombinant human substrate concentrations in the range of 0.5 to 67 ␮M. The substrates P450 isoforms CYP1A1, 1A2, 2C8, 2C9, 2C19, 2D6, and 3A4 were were dissolved in water. For inhibition studies, quercetin was dis- from lymphoblastoid cell lines (Gentest) and yeast (Masimi- solved in methanol and added to each microsomal suspension (final rembwa et al., 1999). P450 isoforms available from lymphoblastoid solvent content of 1%) while on ice. After a 5-min preincubation, cell lines only and not yeast were CYP2A6, 1B1, 2B6, 2E1, 3A5, and NADP was added to initiate the reaction. Kinetic studies were un- 4A11, which were obtained from Gentest. The microsomal prepara- dertaken to determine the mechanism of inhibition and to calculate Ϫ tions were stored at 80°C until use. the apparent inhibition constant (Ki). AQ N-desethylase activity was Incubation Conditions with HLMs and rCYP Isoforms. The measured at six inhibitor concentrations at each of three substrate basic incubation medium contained 0.1 mg/ml HLMs, 3.3 mM MgCl2, concentrations (approximately 1/3 Km, Km, and 3 Km). 1 mM NADP, 3.3 mM glucose 6-phosphate, 1 U/ml glucose-6-phos- Data Analysis. All data points represent the means of duplicate phate dehydrogenase, 100 mM potassium phosphate buffer (pH 7.4), estimations. Km and Vmax values for HLMs and CYP2C8 were deter- and AQ (0.5–67 ␮M), in a final volume of 200 ␮l. The mixture was mined by nonlinear least-squares regression analysis using GraFit incubated at 35°C for 20 min. The reaction was initiated by the software (version 3.0, Erithacus Software Limited, Middlesex, UK). addition of NADP after a preincubation period of 5 min. The incu- Correlations between the activities of respective P450 isoform and bation was terminated by the addition of 150 ␮l of ice-cold acetoni- DEAQ formation or clearance of AQ were determined by least- trile. The mixture was then centrifuged at 4500g for 20 min, and 20 squares linear regression. P Ͻ 0.05 was considered statistically ␮l of supernatant was analyzed by HPLC as described below. Incu- significant. bation conditions used for the 13 different recombinant human P450 Docking AQ into CYP2C Isoform Homology Models. The isoforms were essentially similar to those used for HLMs, except for results from the in vitro experiments described above showed that the quantity of enzyme used (10 pmol of P450/incubation). CYP2C8 was the main enzyme responsible for the metabolism of AQ. Chromatography and Metabolite Structural Analysis. The In our laboratory, we have generated homology models for human HPLC/UV and LC/MS system consisted of an HP 1100 system CYP2C8, 2C9, 2C18, and 2C19 (Ridderstro¨m et al., 2001) based on Amodiaquine N-Desethylation Is Mediated by Cytochrome P450 2C8 401 the crystal structure of the rabbit CYP2C5 (Williams et al., 2000). To Identification of P450s Responsible for AQ and understand the molecular basis of CYP2C8 substrate specificity for DEAQ Metabolism. Figure 2 shows the catalytic activities AQ N-desethylation compared to the closely related CYP2C9, 2C18, of the 13 recombinant human P450 isoforms with respect to and 2C19, AQ was docked into the active site of the homology models the elimination of AQ (1 ␮M) and the N-desethylation of AQ of these . The substrate-binding cavity was defined at var- (50 ␮M). In light of the fact that one can get different RAF ious radii (10, 12, or 15Å) from the oxygen atom of the water coor- factors when using different expression systems and that dinated to the heme. The two-dimensional structure of AQ was built in SYBYL 6.7.6 rates of metabolism of a test compound could differ between (Tripos Associates Inc., St. Louis, MO). The structure was energy systems due to differences in cytochrome 450/cytochrome b5 minimized using the MMFF94s force field and MMFF94 atom ratios and other physiologic parameters (Venkatakrishnan et charges in vacuo conditions. The docking program, GOLD 1.1 (Dr. al., 2000), we used two sets of recombinant P450s from dif- Gareth Jones, University of Oxford, UK), was used in the docking ferent expression systems to see if we could arrive at the (Jones et al., 1997) of AQ to homology models of CYP2C8, 2C9, 2C18, same conclusions. Figure 2 shows that the qualitative role of and 2C19. The genetic algorithm implemented in GOLD was used to the P450s from the different systems in the clearance of AQ optimize the orientation of the ligand in the active site. During this and the formation of DEAQ are similar between yeast- and optimization process the ligand was considered flexible (movement of the flexible bonds), whereas the active site of the enzyme was con- lymphoblastoid-expressed P450s. Figure 2 and Table 1 show sidered rigid (no movement of the amino acid side chains was al- that there are some quantitative differences with yeast lowed). For each experiment 10 dockings were allowed with an early CYP1A1 and CYP2C8 having a greater capacity to clear AQ Downloaded from termination if the root mean square distances were within 1.5 Å for and to form DEAQ. An unknown metabolite M2 was found to the top three solutions. be the major metabolite formed by P450s 1A1 and 1B1 (Fig. 1). Metabolite M2 has the protonated molecular ion at m/z ϭ 299 and a long retention time (tR 6.65 min) compared with Results ϭ AQ (tR 4.51 min). This is not consistent with the other reported AQ metabolites, AQ quinoneimine, bisDEAQ, or AQ and DEAQ Identification and Quantitation. jpet.aspetjournals.org LC/MS and HPLC/UV analysis were used for the qualifica- 2-hydroxyDEAQ, and hence represents a new metabolite tion and quantitation of the metabolites of AQ after incuba- whose identity has yet to be established. tion with rCYP isoforms and HLMs. The only metabolite The P450 isoforms involved in the metabolism of DEAQ found after AQ incubation with HLMs is DEAQ, and it was were also studied at 1 ␮M of the substrate. CYP1A1 com- characterized by comparing the chromatographic retention pletely cleared all the DEAQ, and CYP1B1 also contributes time and multistage mass spectra with the reference sub- to a significant extent in the clearance of the DEAQ. CYP1A1

stance (Fig. 1). The limits of quantitation for both AQ and and 1B1 are however constitutively extrahepatic. The con- at ASPET Journals on September 30, 2017 DEAQ were 0.05 ␮M by UV detection at 342 nm. In our centration of DEAQ after incubation with the other 11 P450s incubation conditions, no detectable bis-desethylated metab- and HLMs were similar to the control samples in which there olite of AQ was found, which could be due to the low protein was no NADPH. This result indicated that DEAQ was rela- concentration (0.1 mg/ml) of HLMs and short incubation time tively metabolically stable in HLMs, which is consistent with (30 min) we used compared with the high concentration of 2 its long terminal half-life in humans. mg/ml and longer incubation time (60 min) used by Jewell et Reaction Kinetics Using Liver Microsomes and al. (1995). CYP2C8. A typical Eadie-Hofstee plot for the formation of

Fig. 1. HPLC chromatograms of me- tabolites of AQ in human liver micro- somes and recombinant CYP1A1, 1B1, and 2C8 after incubation with 10 ␮M of AQ for 30 min. Peaks: 1 ϭ DEAQ, ϭ ϭ ϭ tR 4.30 min; 2 AQ, tR 4.51 min; ϭ ϭ 3 metabolite M2, tR 6.65 min. AQ was incubated with human liver mi- crosomes (a); AQ was incubated with CYP2C8 (b); AQ was incubated with CYP1A1 (c); AQ was incubated with CYP1B1 (d); and AQ in human liver microsomes without NADPH (e). 402 Li et al. Downloaded from

Fig. 2. Amodiaquine clearance (left panel, at 1 ␮M AQ) and N-desethylation (right panel, at 50 ␮M AQ) activities of recombinant human P450 isoforms expressing from lymphoblastoid cell lines (lined column) versus yeast (open column).

TABLE 1 1.2 ␮M and 2.6 pmol/min/pmol with CYP2C8 from lympho- jpet.aspetjournals.org Michaelis-Menten kinetic parameters of amodiaquine N-desethylation blastoid cell lines (Fig. 3), and 0.9 ␮M and 3.9 pmol/min/ and inhibition K values of AQ metabolism by quercetin in HLMs and i pmol, respectively, with CYP2C8 from yeast. Accordingly, CYP2C8 expressed in human lymphoblastoid cell lines vs. yeast the CL of AQ for HLMs and CYP2C8 were 608.2 ␮l/min/ Vmax data shown are in units of picomoles per minute per milligram for HLMs and int picomoles per minute per picomole for CYP2C8; CLint data shown are in units of mg, and 2.1 or 4.4 ␮l/min/pmol of CYP2C8 from lymphoblas- microliters per minute per milligram for HLMs and microliters per minute per toid cell lines and yeast, respectively. These data show that picomole for CYP2C8; Km and Ki data shown are in units of micromoles. AQ is a high clearance drug with high affinity to the associ- Microsomes Vmax Km CLint Ki

ated enzyme(s). at ASPET Journals on September 30, 2017 HLM 1462 2.4 608.2 1.56 Correlation Study. Correlations between AQ metabolic CYP2C8Lymphoblastoid cell lines 2.6 1.2 2.1 1.96 CYP2C8Yeast 3.9 0.9 4.4 2.35 stability and the N-desethylation reaction by a panel of 10 HLMs with activities of eight marker substrates for specific DEAQ from AQ shown in Fig. 3 exhibits monophasic behav- P450s were performed at AQ concentration of 1 ␮M. Figure 4 2 ior, suggesting that a single isoform of P450s may be involved shows an excellent correlation (r ϭ 0.98, P Ͻ 0.01) between in the N-desethylation of AQ in HLMs. Accordingly, a simple AQ desethylase and CYP2C8 activities (determined as the Michaelis-Menten kinetic analysis was used to estimate the activity of 6␣-hydroxylation of paclitaxel) in 10 different 2 ϭ Ͻ affinity constant (Km), the maximum enzyme velocity (Vmax), HLMs. A good relationship (r 0.95, P 0.01) was also and the intrinsic clearance (CLint), defined as Vmax/Km (Table observed (Fig. 4) between the clearance of AQ and CYP2C8 1). The apparent Km and Vmax values for AQ desethylation activities in those liver microsomes. There were no signifi- were 2.4 ␮M and 1462 pmol/min/mg of protein with HLMs, cant correlations between the desethylation of AQ and cata-

Fig. 3. Plot of velocity versus amodiaquine concentration for the formation of desethylamodiaquine in human liver microsomes (left panel) and in CYP2C8 expressed from lymphoblastoid cell lines (right panel). Inset, Eadie-Hofstee plot for the desethylation of amodiaquine in human liver microsomes. Amodiaquine N-Desethylation Is Mediated by Cytochrome P450 2C8 403

desethylase activity was very close to the measured liver micro- somal N-desethylation activity for AQ, indicating that CYP2C8 plays a major role in the N-desethylation of AQ. In Vitro-in Vivo Correlation of AQ Clearance. Using the well stirred model, the clearance of AQ calculated from ␮ the in vitro t1/2 (15.6 min, at 1 M of AQ) predicted an in vivo clearance of 19.1 ml/min/kg. The clearance of AQ calculated

from Vmax/Km in vitro predicted in vivo clearance of 19.3 ml/min/kg. Inhibition Study. Quercetin was previously shown to be a potent diagnostic inhibitor for CYP2C8. Figure 5 shows that quercetin is a competitive inhibitor of DEAQ formation catalyzed by CYP2C8 from lymphoblastoid cell lines and ␮ HLMs, with Ki values of 1.96 and 1.56 M, respectively. The ␮ Ki value for CYP2C8 from yeast was 2.35 M. Inhibition of AQ N-desethylation by ketoconazole (10 ␮M) was also ob- served in HLMs at an AQ concentration of 5 ␮M. Ketocon- Fig. 4. Relationship between amodiaquine metabolism (clearance and Downloaded from desethylation of amodiaquine) and paclitaxel 6␣-hydroxylase activities in azole inhibited the formation of DEAQ by about 60%, microsomes obtained from 10 different human livers. whereas the effect by quercetin (10 ␮M) was 70%. However, ketoconazole has been shown to also inhibit CYP2C8 cata- lytic activity for P450s 1A2, 2A6, 2C19, 2D6, 2E1, and 3A4 lyzed paclitaxel 6␣-hydroxylation by more than 50% (Ma- 2 Ͻ (r 0.35). Although CYP2C9 showed a fair relationship simirembwa et al., 1999) at 10 ␮M. The inhibition of AQ 2 ϭ (r 0.59) between its activity and AQ desethylase activity, N-desethylation caused by ketoconazole could therefore be a

2 ϭ jpet.aspetjournals.org it also had some correlation (r 0.64) with the activity of reflection of the unspecificity of the inhibitor at high concen- CYP2C8. Since no detectable amount of DEAQ has been tration. found after incubation with CYP2C9, it is unlikely that this Docking AQ into CYP2C Isoforms. When AQ was enzyme is involved in the metabolism of AQ. docked into the active site cavities of the four CYP2C iso- Contribution of CYP2C8 to AQ N-Desethylase Activity forms, the docking program, GOLD, found several preferred in HLMs. From the P450 identification studies (Fig. 2) of the AQ-CYP2C protein interactions. In the docking query, the hepatic isoforms, CYP2C8 had the major role in the clearance of program was asked to provide the 10 best solutions. Inter- AQ and the formation of its major metabolite, DEAQ. To esti- pretation of the results was based on the favorability of at ASPET Journals on September 30, 2017 mate the relative contribution of this P450 in the hepatic me- interactions as deduced by GOLD and the proximity of po- tabolism of AQ, the relative activity factor approach was used. tential sites of metabolism to oxygen of the water coordinat- The contribution of CYP2C8 to the AQ N-desethylase activity in ing to the heme. Table 3 shows that in all cavity sizes chosen, ϭ pooled HLMs was estimated using the clearance (CL Vmax/ the greatest number of favorable solutions was found with Km)ofAQN-desethylation by recombinant CYP2C8 from lym- CYP2C8 and that the cavity with radii of 12 and 15 Å defined ␮ phoblastoid cell lines (CLCYP2C8, 2.1 l/min/pmol of CYP2C8) the best substrate binding cavity for AQ. AQ docked in the ␮ and by pooled HLMs (CLHLM, 608.2 l/min/mg of protein), different P450s indicating possible regions for metabolism ␣ respectively (Table 2). The paclitaxel 6 -hydroxylase activity in (Table 3, Fig. 6). microsomes from CYP2C8 was 0.35 pmol/min/pmol of CYP2C8, and for pooled HLMs it was 94.2 pmol/min/mg of protein. Thus, Discussion RAFv, CYP2C8(lymphoblastoid) was estimated to be 268.8 (pmol of CYP2C8/mg of HLMs). Using the same approach, a The results of our study show that the clearance of AQ and RAFv,CYP2C8(yeast) of 151.1 was obtained using yeast-expressed metabolism to its main metabolite, DEAQ, is catalyzed by CYP2C8. Table 2 shows that using either the lymphoblastoid- CYP2C8. They also show that there is an unidentified me- or yeast-expressed CYP2C8, we get similar percentage of con- tabolite (M2), which is a product of AQ metabolism by extra- tributions of CYP2C8 to the clearance of AQ. The data in Table hepatic CYP1A1 and 1B1. The role of the extrahepatic 2 show that the predicted contribution of CYP2C8 to AQ N- CYP1A1 and 1B1 in the clearance of AQ or the formation of M2 could have toxicological implications in individuals ex- TABLE 2 posed to inducers (e.g., polyaromatic hydrocarbons) of these Contribution of CYP2C8 expressed in human lymphoblastoid cell lines vs. yeast to AQ N-desethylation in HLMs by RAF method extrahepatic P450s if M2 is reactive. Measured CL data shown are in units of microliters per minute per milligram for The identification of CYP2C8 as the major enzyme respon- HLM and microliters per minute per picomoles for CYP2C8; predicted CL data sible for the hepatic metabolism of AQ was derived from shown are in units of microliters per minute per milligram. several lines of evidence: 1) of several recombinant cyto- AQ-Desethylase Activity chrome P450 isoforms tested, CYP2C8 showed a dominant capacity for AQ N-desethylation; 2) CYP2C8 was an efficient Microsomes CYP2C8 HLM catalyst of the reaction, as demonstrated by a turnover of 2.6 Lymphoblastoid Cell Lines Yeast pmol/min/pmol of CYP2C8; 3) AQ N-desethylase activity, as CL (measured) 2.1 4.4 608.2 well as AQ elimination rate, correlated well with hepatic 2 ϭ RAFv, CYP2C8 268.8 151.1 CYP2C8 activity (r 0.98 and 0.95, respectively); 4) when CL (predicted) 562.3 660.7 the activity of CYP2C8 used in this study was adjusted for its % Contribution 92.5 108.6 content of HLMs, the predicted contribution of CYP2C8 to 404 Li et al. Downloaded from

Fig. 5. Dixon plots for inhibition of CYP2C8 catalyzed amodiaquine desethylation by quercetin in human liver microsomes (left panel) and CYP2C8 expressed from lymphoblastoid cell lines (right panel). Concentrations of substrate (AQ) are shown in each plot. Each point represents the mean of duplicate measurements.

TABLE 3 CYP2C isoforms was mainly driven by the geometric struc- Docking of AQ to CYP2C8, 2C9, 2C18, and 2C19 ture of the binding cavity. The many solutions of docking into jpet.aspetjournals.org Hits were defined as a number of dockings (of 10) with the hydrogen atoms on AQ at CYP2C8 indicated that the possibility of hydrogen abstrac- a maximum distance of 4 Å from the oxygen atom on the heme. tions on the carbons bonded to the nitrogen where AQ is N-desethylated (Table 3 and Fig. 6) and most of the distances from those hydrogens to the oxygen atom were around 4 Å,a distance consistent with that of sites of oxidation of other substrates when docked into active sites of P450s that me-

tabolize them. AQ desethylation is likely to proceed in two at ASPET Journals on September 30, 2017 steps, a hydrogen abstraction and hydroxylation at the adja- cent carbon, forming an unstable carbinolamide that rapidly hydrolyzes to DEAQ and acetaldehyde. The results from AQ clearance studies using the t and the Distance 1/2 P450s Cavity Radius Hits a (Å) Vmax/Km approach were comparable further indicating that the CYP2C8 15 Å 5 2.47–7.94 AQ desethylation is the major hepatic metabolic route for the 12 Å 7 2.65–4.10 clearance of the drug. The predicted in vivo clearance value of 10 Å 2 2.29–3.44 19.1 ml/min/kg estimated from in vitro experiments greatly CYP2C9 15 Å 1 12 Å 0 underestimated the observed clearance of 216 ml/min/kg (ob- 10 Å 1 tained after intravenous drug administration). There is, how- CYP2C18 15 Å 0 ever, a wide interindividual variation in the in vivo clearance 12 Å 0 10 Å 2 rates ranging from 78 to 943 ml/min/kg (White et al., 1987). In CYP2C19 (many different 15 Å 0 general, clearances predicted from in vitro data underestimate solutions of binding 12 Å 2 the observed values for many drugs (Carlile et al., 1999). The modes) 10 Å 0 possible reasons for this trend have been reviewed by Iwatsubo a The distance from the hydrogen on the carbon adjacent to site of N-desethyla- et al. (1997). The reason for underestimating the in vivo clear- tion to the oxygen ligated to the heme. ance of AQ might have to do with the fact that AQ is probably AQ N-desethylase activity (562.3 ␮l/min/mg of protein) was also metabolized in the blood since CYP1A1 and 1B1 have been similar to that found with HLMs (608.2 ␮l/min/mg of pro- found in this fluid (Baron et al., 1998; Nguyen et al., 2000). Our tein); 5) the inhibitory effects for quercetin were comparable results show that these extrahepatic enzymes extensively me- tabolize AQ (Fig. 2). Another reason could be that during the for expressed CYP2C8 and for HLMs with Ki values of 1.96 and 1.56 ␮M, respectively, which are similar to those ob- preparation of HLMs, the recovery of the specific enzyme me- tained using another CYP2C8 substrate, paclitaxel (Rahman tabolizing AQ, CYP2C8, is low and not reflected in the HLM et al., 1994); and 6) docking of AQ into CYP2C isoforms recovery factor we used in estimating hepatic clearance. showed best solutions with CYP2C8, which also suggested Using recombinant P450s from different sources (lympho- potential metabolism in the proximity of the nitrogen atom blastoid and yeast), similar relative contributions (93 and 109%, on the diethylamino group of AQ. Furthermore, in the corre- respectively) of CYP2C8 to the metabolism of AQ were ob- lation study, the amount of the loss of AQ was almost equal tained. Similar Km and Ki (for inhibition by quercetin) values to that of the formation of DEAQ, indicating that DEAQ is were also obtained, indicating that for AQ metabolism, the the dominant metabolite in HLMs and catalyzed by CYP2C8. enzymes from the two different expression systems are equally The selectivity of AQ-CYP2C8 interaction among the predictive of the role of CYP2C8. For the metabolism of other Amodiaquine N-Desethylation Is Mediated by Cytochrome P450 2C8 405 Downloaded from jpet.aspetjournals.org at ASPET Journals on September 30, 2017

Fig. 6. Docking of amodiaquine into CYP2C8, 2C9, 2C18, and 2C19. The heme is magenta-colored and active site amino acids are in green. The docked compound is amodaiquine. GOLD 1.1 was used for the docking. compounds, use of P450s from different expression systems Our results conclusively show that CYP3A4 is not in- have been shown to result in inconsistent relative contributions volved in the metabolism of AQ to DEAQ as had been of specific P450s (Venkatakrishnan et al., 2000). previously proposed by Jewell et al. (1995) using ketocon- Contrary to in vivo studies and the in vitro studies of azole as a diagnostic inhibitor. That erroneous conclusion Jewell et al. (1995) in which two other metabolites, 4-hy- is common when a single enzyme identification procedure droxyDEAQ and bisDEAQ were observed, we did not ob- is used without caution with respect to its shortcomings. In serve these metabolites with both HLMs and rCYPs. The this case (Jewell et al., 1995), ketoconazole is probably only M2 produced by CYP1A1 and 1B1 has a protonated molec- selective for CYP3A in the nanomolar range as we have ular ion at m/z 299, which is not consistent with these shown that at 10 ␮M, it also inhibited CYP2C8 by over 50% reported metabolites of AQ. This highlights the risk of (Masimirembwa et al., 1999). In other methods like the failing to identify minor and slowly formed metabolites correlation analysis, use of HLM samples with P450 activ- when using low microsomal proteins and short incubation ities which cocorrelate can also lead to erroneous P450 times as used in our study. Another limitation pointed out identifications. Following the formation of a known metab- by Jewell et al. (1995) is that the in vitro assays underes- olite can also be misleading since that might not be the timate the likely bioactivation of AQ to the reactive qui- main or only route of elimination. A simultaneous investi- noneimine since HLMs rapidly reduce it. gation of substrate disappearance is therefore recom- 406 Li et al. mended. A multifaceted approach as employed in this high affinity and turnover enzyme-specific marker reaction study ensures valid P450 identification and better estima- for assaying CYP2C8 activity. tions of drug clearance. Pharmacokinetic studies have shown a large variation in References A¨ belo¨ A, Andersson TB, Antonsson M, Naudot AK, Skånberg I, and Weidolf L (2000) the kinetic parameters of AQ and DEAQ. This variation Stereoselective metabolism of omeprazole by human cytochrome P450 enzymes. could have implications in the therapeutic and toxicologi- Drug Metab Dispos 28:966–972. cal response to the drug. In a subject with acute hepatic Baron JM, Zwadlo-Klarwasser G, Jugert F, Hamann W, Rubben A, Mukhtar H, and Merk HF (1998) Cytochrome P450 1B1: a major P450 isoenzyme in human blood failure, the concentration of AQ was found to be a quarter monocytes and macrophage subsets. Biochem Pharmacol 56:1105–1110. that of the DEAQ (Pussard et al., 1987). This was very Carlile DJ, Hakooz N, Bayliss MK, and Houston JB (1999) Microsomal prediction of in vitro clearance of CYP2C9 substrates in humans. Br J Clin Pharmacol 47:625– usual since in most subjects, very low concentrations of AQ 635. itself are detected in blood and urine in the first hours Churchill FC, Mount DL, and Patchen LC (1986) Isolation, characterization and standardization of a major metabolite of amodiaquine by chromatographic and following oral administration. These results are interest- spectroscopic methods. J Chromatogr 377:307–318. ing in light of the findings of our study. First, the large Churchill FC, Patchen LC, Campbell CC, Schwartz IK, Nguyen-Dinh P, and Dick- inson CM (1985) Amodiaquine as a pro-drug: the importance of metabolite(s) in the interindividual variation in pharmacokinetics is consis- antimalarial effect of amodiaquine in humans. Life Sci 36:53–62. tent with the up to 38-fold variation among HLMs in Crespi CL (1995) Xenobiotic-metabolizing human cells as tools for pharmacological and toxicological research. Adv Drug Res 26:179–235. metabolizing CYP2C8 substrates (reviewed in Ong et al., Dai D (2001a) Allelic frequencies of human CYP2C8 and genetic linkage among different ethnic populations. FASEB 15:A575. 2000). Variability in CYP2C8 either due to coadministered Downloaded from Dai D (2001b) Cardiovascular effects of polymorphic human CYP2C8s and the drug inhibitors or coingested dietary inhibitors/inducers, metabolism of . FASEB 15:A918. or genetic polymorphism could explain differential suscep- Houston JB (1994) Utility of in vitro drug metabolism data in predicting in vivo metabolic clearance. Biochem Pharmacol 47:1469–1479. tibility of individuals to AQ toxicity and maybe explain Iwatsubo T, Hirota N, Ooie T, Suzuki H, Shimada N, Chiba K, Ishizaki T, some of the fatalities reported (Larrey et al., 1986; Rouveix Green CE, Tyson CA, and Sugiyama Y (1997) Prediction of in vivo drug metabolism in the human liver from in vitro metabolism data. Pharmacol Ther et al., 1989). Regulation of CYP2C8 expression and activity 73:147–171. Jewell H, Maggs JL, Harrison AC, O’Neill PM, Ruscoe JE, and Park BK (1995) Role by these factors need to be investigated as they might jpet.aspetjournals.org of hepatic metabolism in the bioactivation and detoxication of amodiaquine. Xe- assist in the optimal use of AQ. nobiotica 25:199–217. Genetic variants of CYP2C8 have been reported (http:// Jones G, Willett P, Glen RC, Leach AR, and Taylor R (1997) Development and validation of a genetic algorithm for flexible docking. J Mol Biol 267:727–748. www.imm.ki.se/CYPalleles; Ree et al., 1999). The CYP2C8*2, Larrey D, Castot A, Pessayre D, Merigot P, Machayekhy JP, Feldmann G, Lenoir A, CYP2C8*3, and CYP2C8*4 variants associated with an in- Rueff B, and Benhamou JP (1986) Amodiaquine-induced hepatitis. A report of ␣ seven cases. Ann Intern Med 104:801–803. creased Km, decreased activity for paclitaxel 6 -hydroxyla- Laurent F, Saivin S, Chretien P, Magnaval JF, Peyron F, Sqalli A, Tufenkji AE, tion and uncharacterized effects, respectively (Dai, 2001a,b; Coulais Y, Baba H, and Campistron G (1993) Pharmacokinetic and pharmacody- namic study of amodiaquine and its two metabolites after a single oral dose in

A. Daly and G. P. Aithal, submitted for publication). Knowl- human volunteers. Arzneim-Forsch 43:612–616. at ASPET Journals on September 30, 2017 edge of frequencies of distribution of CYP2C8 variants in the Masimirembwa CM, Otter C, Berg M, Jo¨nsson M, Leidvik B, Jonsson E, Johan- sson T, Bo¨ckman A, Edlund A, and Andersson TB (1999) Heterologous expres- population likely to be exposed to the CYP2C8 substrate drug sion and kinetic characterization of human cytochromes P-450: validation of a amodiaquine will contribute to our understanding of how pharmaceutical tool for drug metabolism research. Drug Metab Dispos 27: 1117–1122. patients will respond to this drug. Mount DL, Patchen LC, Nguyen-Dinh P, Barber AM, Schqartz IK, and Churchill FC Due to the changing level of knowledge, the number and (1986) Sensitive analysis of blood for amodiaquine and three metabolites by high-performance liquid chromatography with electrochemical detection. J Chro- relative importance of individual P450s is continuously matogr 383:375–386. changing. Once it was only 2, then 5 and now over 10 P450s Naisbitt DJ, Williams DP, O’Neill PM, Maggs JL, Willock DJ, Pirmohamed M, and Park BK (1998) Metabolism-dependent neutrophil cytotoxicity of amodiaquine: a that are thought to be important for drug metabolism. comparison with pyronaridine and related antimalarial drugs. Chem Res Toxicol These claims are partly to do with the biased studies with 11:1586–1595. Nakajima M, Nakamara S, Tokudome S, Shimada N, Yamazaki H, and Yokoi T respect to some P450s and the expanding chemical space in (1999) Azelastine N-demethylation by cytochrome P-450 (CYP)3A4, CYP2D6, and the drug discovery process leading to new chemistries pre- CYP1A2 in human liver microsomes: evaluation of approach to predict the contri- bution of multiple P450s. Drug Metab Dispos 27:1381–1391. ferring previously “minor” P450s. Currently, there are few Nguyen LT, Ramanathan M, Weinstock-Guttman B, Dole K, Miller C, Planter M, well characterized CYP2C8 marker substrates and diag- Patrick K, Brownscheidle C, and Jacobs LD (2000) Detection of cytochrome P450 and other drug-metabolizing enzyme mRNAs in peripheral blood mononuclear nostic inhibitors. Because of CYP2C8 selectivity for AQ cells using DNA arrays. Drug Metab Dispos 28:987–993. desethylation, high affinity, and turnover, AQ could be an Obach RS, Baxter JG, Liston TE, Silber BM, Jones BC, MacIntyre F, Rance DJ, and Wastall P (1997) The prediction of human pharmacokinetic parameters excellent in vitro probe drug for CYP2C8 activity. AQ and from preclinical and in vitro metabolism data. J Pharmacol Exp Ther 283:46– DEAQ are relatively cheap compared with paclitaxel and 58. Ong CE, Coulter S, Birkett DJ, Rhasker B, and Miners JO (2000) The xenobi- its metabolite. Although CYP2C8 is the major hepatic en- otic inhibitor profile of cytochrome P450 2C8. Br J Clin Pharmacol 50:573– zyme responsible for both the clearance of AQ and the 580. Pussard E, Verdier F, Faurisson F, Scherrmann JM, Bras JL, and Blayo MC (1987) formation of DEAQ, paclitaxel is mainly metabolized to Disposition of monodesethylamodiaquine after a single oral dose of amodiaquine 6␣-OH paclitaxel by CYP2C8 and to other metabolites by a and three regimens for prophylaxis against plasmodium falciparum malaria. Eur J Pharmacol 33:409–414. major hepatic enzyme, CYP3A4 (Sonnichsen et al., 1995). Rahman A, Korzekwa KR, Gorgan J, Gonzalez FJ, and Harris JW (1994) Selective These factors make AQ desethylation a relatively better biotransformation of taxol to 6␣-hydroxytaxol by human cytochrome P450 2C8. Cancer Res 54:5543–5546. probe. Ree HC, McSorley LC, and Daly AK (1999) A novel genetic polymorphism in the In conclusion, we have shown that CYP2C8 is the exclusive cytochrome P450 CYP2C8. ISSX Proc 14:72. Ridderstro¨m M, Zamora I, Fjellstro¨m O, and Andersson TB (2001) Analysis of isoform contributing to the N-desethylation of AQ in HLMs selective regions in the active sites of human cytochromes P450, 2C8, 2C9, 2C18, from among a panel of recombinant human hepatic P450 and 2C19 homology models using GRID/CPCA. J Med Chem 44:4072–4081. Rouveix B, Coulombel L, Aymard JP, Chau F, and Abel L (1989) Amodiaquine- isoforms. This knowledge will give us a better understanding induced immune agranulocytosis. Br J Haematol 71:7–11. of the basis of the interindividual variability in AQ pharma- Sonnichsen DS, Liu Q, Schuetz EG, Schuetz JD, Pappo A, and Relling MV (1995) Variability in human cytochrome P450 paclitaxel metabolism. J Pharmacol Exp cokinetics and probably therapeutic and toxicological re- Ther 275:566–575. sponses to the drug. AQ desethylation also represents a new Venkatakrishnan K, Von Moltke LL, Court MH, Harmatz JS, Crespi CL, and Amodiaquine N-Desethylation Is Mediated by Cytochrome P450 2C8 407

Greenblatt DJ (2000) Comparison between cytochrome P450 (CYP) content and Winstanley P, Edwards G, Orme M, and Breckenridge A (1987) The disposi- relative activity approaches to scaling from cDNA-expressed P450s to human liver tion of amodiaquine in man after oral administration. Br J Clin Pharmacol microsomes: ratios of accessory proteins as sources of discrepancies between the 23:1–7. approaches. Drug Metab Dispos 28:1493–1504. White NJ, Looareesuwan S, Edwards G, Phillips RE, Karbwang J, Nicholl DD, Bunch C, and Warrell DA (1987) Pharmacokinetics of intravenous amodiaquine. Br J Clin Pharmacol 23:127–135. Address correspondence to: Dr. Collen Masimirembwa, DMPK and Bio- Williams PA, Cosme J, Sridhar V, Johnson EF, and McRee DE (2000) Mammalian analytical Chemistry, AstraZeneca R&D Mo¨lndal, S-431 83 Mo¨lndal, Sweden. microsomal cytochrome P450 monooxygenase: structural adaptations for mem- E-mail: [email protected] brane binding and functional diversity. Mol Cell 5:121–131. Downloaded from jpet.aspetjournals.org at ASPET Journals on September 30, 2017