Abstract
2-[6-(4-Chlorophenyl)-2,2-dimethyl-7-phenyl-2,3-dihydro-1H-pyrrolizin-5-yl] acetic acid (licofelone) is a dual inhibitor of both cyclooxygenase isoforms and 5-lipoxygenase and under development for treatment of osteoarthritis. In conventional in vitro assays using liver microsomes and NADPH as cosubstrate, a high metabolic stability of licofelone was observed. In the presence of UDP-glucuronic acid, licofelone is rapidly converted into the corresponding acyl glucuronide, M1. These results are in conflict with data from clinical studies. After administration of licofelone to humans, M1 plasma concentrations were negligibly low, whereas the exposure of the hydroxy-metabolite M2 achieved values of approximately 20% compared with that of the parent drug. Metabolism studies with human hepatocytes and dual-activity assays with microsomes, which allowed the simultaneous monitoring of hydroxylation and glucuronidation reactions, were performed, and the metabolic pathway of licofelone was elucidated. After glucuronidation, predominantly catalyzed by UDP glucuronosyltransferase (UGT) isoforms UGT2B7, UGT1A9, and UGT1A3, M1 is converted into the hydroxy-glucuronide M3 in a CYP2C8-dependent reaction. The enzyme specificities were investigated using recombinant human cytochrome P450 and UGT isoforms as test systems. In vitro drug-interaction studies using the 6α-hydroxylation of paclitaxel as control reaction confirmed that neither licofelone nor M1 is a relevant inhibitor of CYP2C8. The formation of M3 was also observed with liver microsomes from cynomolgus monkeys, but in incubations with mouse and rat liver microsomes, M1 remained unchanged. The clinical relevance of these findings is discussed.
Licofelone (ML3000) is a dual inhibitor of cyclooxygenases (COX-1 and COX-2) and 5-lipoxygenase (5-LOX). Anti-inflammatory drugs with this mode of action are considered as attractive therapeutics for the treatment of arthritic diseases (Skelly and Hawkey, 2003; Clària and López-Parra, 2005). Almost 3 decades ago, the very first COX/5-LOX inhibitor benoxaprofen entered the market, and at that time, compared with the conventional nonsteroidal anti-inflammatory drugs (NSAIDs), this compound not only showed equivalent efficacy but also an advantageous gastrointestinal safety profile. Unfortunately, this compound was hepatotoxic, which led to its withdrawal from the market (Bakke et al., 1995). In the late 1980s, several companies were running drug-discovery programs to develop dual COX/LOX inhibitors as NSAIDs with an improved gastrointestinal safety. Most projects were terminated after the discovery of the inducible COX-2 and its identification as that isoform that is directly linked with the inflammatory process. After showing advantageous gastrointestinal safety in a series of clinical trials, COX-2-selective drugs (coxibs) achieved blockbuster-drug status very soon after entering the market. Today, it is known that all the coxibs are burdened with a very low but unequivocal risk to induce life-threatening or fatal cardiovascular events (Grosser et al., 2006).
Apart from the first clinical experience with benoxaprofen, there is a lot of nonclinical and clinical evidence that dual inhibition of the COX and 5-LOX pathway combines the advantages of conventional NSAIDs and coxibs (i.e., good anti-inflammatory, analgesic activity and gastrointestinal and cardiovascular safety). As a promising candidate, licofelone has been developed for treatment of arthritic conditions. The mechanism of action of licofelone (i.e., inhibition of both COX-1/-2 and 5-LOX) was shown in pharmacological studies and in experimental disease models (Tries et al., 2002; Kulkarni and Singh, 2007). Human and veterinary clinical trials confirmed the efficacy of licofelone for treatment of osteoarthritis (Alvaro-Gracia, 2004; Moreau et al., 2007), as well as its advantageous gastrointestinal tolerability (Bias et al., 2004; Moreau et al., 2005).
To date, no data regarding clinical and nonclinical pharmacokinetics and metabolism have been published. In humans, after p.o. administration of immediate-release tablets, licofelone is rapidly absorbed from the gastrointestinal tract, and maximum plasma concentrations are achieved approximately 2 to 3 h after administration. Systemic elimination follows biphasic characteristics with a rapid initial decrease of plasma concentration [t1/2(α) = 1 h] and a slow terminal elimination [t1/2(β) = 7–9 h]. In plasma, after single dosing, ML3000–1-O-acyl glucuronide (M1) and hydroxy-ML3000 (M2) were detected as metabolites. Relative to the parent drug, the systemic exposure remained less than 2%. After repeated administration, at steady state, the exposure of M2 increased to approximately 20% relative to that of the parent drug, and the rate of systemic elimination was below that of the parent drug (monophasic, t1/2 = 10–12 h). In contrast, M1 remained at the level of a trace metabolite. In that respect, the disposition of licofelone in humans is different from all the standard animal species (mouse, rat, dog, monkey) in which systemic concentrations of M2 were negligible even on chronic dosing. A further hydroxy-metabolite of licofelone is M4. This compound was initially identified in microsomal experiments but not in plasma samples from humans after single and repeated administration of therapeutic doses (i.e., 200 or 400 mg b.i.d.). Relevant concentrations were determined in plasma samples from subjects who were treated with increasing doses to determine the maximum tolerated dose. The chemical structures of licofelone and its metabolites are shown in Fig. 1. In the present article, results from in vitro metabolism studies are presented that show that in humans hydroxylation of the glucuronide M1 represents the pivotal step in the biosynthesis of M2. Although the cytochrome P450 (P450)-dependent hydroxylation of glucuronides has been described in the literature (Kumar et al., 2002; Delaforge et al., 2005), the formation of M2 represents a unique example as the systemic exposure of humans to this major metabolite is based on the glucuronidation of the parent drug followed by hydroxylation of the glucuronide.
Materials and Methods
Chemicals and Enzyme Preparations. Licofelone and the metabolites M1, M2, and M4, as well as ML3000 formic acid, diclofenac, and 4′-hydroxydiclofenac, were from Merckle (Ulm, Germany). Alamethicin, paclitaxel, β-NADPH, β-NADP, glucose-6-phosphate (G6P), G6P dehydrogenase, UDP-glucuronic acid (UDPGA), and recombinant human UDP glucuronosyltransferases (UGTs) UGT1A1, UGT1A6, UGT1A7, and UGT1A10 were obtained from Sigma (Taufkirchen, Germany). Montelukast was extracted from SIN-GULAIR (Merck & Co., Inc., Whitehouse Station, NJ) 10-mg tablets. Recombinant human P450 supersomes CYP1A2, CYP2A6, CYP2C8, CYP2C9*1, CYP2C19, CYP2D6*1, CYP3A4, CYP2J2, and CYP4F12, lymphoblast-expressed human CYP2B6 and CYP2E1, recombinant human UGT1A3, UGT1A9, UGT2B7, and UGT2B15, as well as pooled liver microsomes from male beagle dogs (DLM) and male New Zealand White rabbits (NZLM), were purchased from BD Gentest (Heidelberg, Germany). Pooled human intestine (HIM), kidney (HKM), and lung (HLuM) (smokers) microsomes were purchased from In Vitro Technologies (Baltimore, MD). All the other chemicals used were of the highest purity grade available.
Preparation of Liver Microsomes. Pooled liver microsomes from humans (HLM), cynomolgus monkeys (CLM), Sprague-Dawley rats (RLM), and CD-1 mice (MLM) were prepared by differential centrifugation of homogenized liver samples according to previously described procedures (Pearce et al., 1996; Wilson et al., 2003). Cynomolgus monkey intestine microsomes (CIM) were prepared from monkey mucosa homogenate. Protein content was determined according to the Bradford assay (Bradford, 1976) with bovine serum albumin (BSA) as standard.
Biotransformation in Human Hepatocytes. Primary human hepatocytes were isolated from liver samples obtained from patients undergoing partial liver resections according to a two-step collagenase perfusion technique and cultivated in six-well plates (1 × 106 cells/well). After the attachment period, the culture medium was changed, and 30 and 100 μM licofelone dissolved in methanol were added (final solvent concentration 0.5%). Plates were incubated at 37°C (5% CO2 in air, 95% relative humidity), and samples (100 μl) were withdrawn after 1 and 4 h. For final sampling, reactions were terminated after 24 h by addition of 0.5 volume of ice-cold methanol. The cells were carefully scraped off the plates, and the suspensions were transferred into reaction tubes. Each experiment was performed in triplicate with cells isolated from liver samples of three different donors. After addition of ML3000 formic acid as internal standard (ISTD), protein was precipitated with acetonitrile (ACN). Samples were centrifuged, and the supernatant was analyzed by liquid chromatography/tandem mass spectrometry (LC/MS/MS).
Phase I in Vitro Metabolism Experiments and Dual-Activity Assays. Phase I metabolism experiments and dual-activity assays were carried out in 0.1 M Tris-HCl buffer, pH 7.4 (37°C), containing 0.1% BSA, 25 μg/ml alamethicin, 4 mM MgCl2, 5 mM UDPGA, and 1 mM β-NADPH at a final volume of 250 μl. For phase I metabolism assays no UDPGA was added. All the incubations were performed with HLM, CLM, RLM, MLM, DLM, NZLM, HIM, HKM, HLuM, and CIM. Final microsomal protein concentration was 1 mg/ml each. For activation of microsomal UGTs, samples were preincubated on ice for 15 min. After prewarming to 37°C, reactions were started by addition of licofelone dissolved in methanol (final solvent concentration 1%). Final concentrations were 10, 30, and 100 μM. After 30 min, an aliquot was withdrawn and mixed with ISTD and 1.0 volume of cold ACN. Precipitated protein was removed by centrifugation, and the supernatant was analyzed with LC/MS/MS. All the experiments were performed in triplicate. In a preliminary experiment, no influence on phase I metabolism reactions in case of preincubation on ice in the presence of alamethicin could be observed (data not shown).
Metabolism Experiments with M1. HLM, CLM, RLM, and MLM (1 mg/ml each) were incubated with 3 μM M1 in 0.1 M Tris-HCl buffer, pH 7.4 (37°C), containing 0.1% BSA, 1 mM β-NADPH, and 1 mM MgCl2 at a final volume of 250 μl. Reactions were initiated by addition of substrate dissolved in methanol (final solvent concentration 1%). Each experiment was performed in duplicate. Sample preparation was performed as described above. In a control experiment, M1 was incubated in 0.1 M Tris-HCl buffer only.
Phase I Assays with Licofelone and Recombinant P450 Isoforms. Incubations with CYP1A2, CYP1A6, CYP2B6 (62.5 pmol/ml each), CYP2C8, CYP2C9*1, CYP2C19, CYP2D6*1, CYP2E1 (127.5 pmol/ml each), CYP2J2, CYP4F12, and CYP3A4 (50 pmol/ml each) were carried out in 0.1 M Tris-HCl buffer, pH 7.4 (37°C), containing 1 mM MgCl2, 1 mM β-NADPH, and 0.1% BSA. Reactions were started by addition of 10 μM licofelone or 3 and 10 μM M1 dissolved in methanol (final solvent concentration 1%). After 30 min, reactions were terminated as described above. Additional assays with CYP2C8, CYP2C9*1, CYP2C19, CYP2D6*1, CYP2J2, and CYP3A4 were performed with 100 μM licofelone in the presence of 50 and 200 pmol/ml P450. Sample preparation was performed as described above.
Phase II Assays with Licofelone and Recombinant UGT Isoforms. Experiments with UGT1A1, UGT1A3, UGT1A6, UGT1A7, UGT1A9, UGT1A10, UGT2B7, and UGT2B15 were carried out in 0.1 M Tris-HCl buffer, pH 7.4, in the presence of 4 mM MgCl2, 5 mM UDPGA, and 25 μg/ml alamethicin at a final protein concentration of 1.0 mg/ml each. UGTs were activated by incubation on ice in the presence of alamethicin and MgCl2 for 15 min. After addition of UDPGA, samples were prewarmed to 37°C, and reactions were initiated by addition of 100 μM licofelone dissolved in methanol (final solvent concentration 1%). Samples were incubated for 30 min and processed as described above.
Inhibition Experiments.Montelukast. A dual-activity assay was performed with 30 μM licofelone as described above. Immediately before the addition of the substrate, montelukast was added to the incubation mixtures. Final concentrations were 0, 1, 10, and 30 μM at a final methanol concentration of 1.5%. After incubation for 30 min, samples were analyzed by LC/MS/MS.
M1. Ten and 30 μM M1 were incubated with HLM (1 mg/ml) and CYP2C8 (50 pmol/ml) as described above. In a modified experiment, M1 was preincubated at 37°C in the presence of HLM and CYP2C8, respectively. After 15 min, reactions were started by addition of β-NADPH, and samples were incubated for 30 min.
Paclitaxel assay. The paclitaxel-6α-hydroxylation was investigated in HLM and with CYP2C8. Standard assays were performed in 0.1 M Tris-HCl buffer, pH 7.4, with addition of 0.1% BSA and in the presence of 4 mM MgCl2. An NADPH-regenerating system was established by addition of 5 mM G6P, 1 mM NADP, and 1 U/ml G6P dehydrogenase. Final substrate concentration was 50 μM in a final volume of 250 μl. After 30 min, reactions were terminated by addition of 125 μl of cold ACN. Samples were vortexed and centrifuged, and the supernatant was subjected to high-performance liquid chromatography (HPLC)/UV analysis.
Metabolite production for paclitaxel was linear with respect to time, microsomal protein, and P450 concentration. The assay was validated using montelukast as CYP2C8 inhibitor (data no shown). With 100 μM montelukast, a remaining enzyme activity of 15% was determined. Interaction experiments were performed with 10, 30, and 100 μM licofelone (dissolved in methanol) in comparison with a control assay with addition of solvent only. All the assays were performed in triplicate, and mean values and S.D. were calculated.
LC/MS/MS Analysis. Chromatographic separation of licofelone metabolites was achieved on an Inertsil ODS-2 column (60 × 4.6 mm, dp = 5 μm) with a WiCom OptiGuard C18, 1-mm guard column (Wicom GmbH, Heppenheim, Germany) using an Agilent 1100 HPLC system (Agilent Technologies, Palo Alto, CA). Mobile phase A consisted of ACN containing 1% formic acid, and mobile phase B consisted of deionized water adjusted to pH 4.0 with formic acid. The gradient profile was percent B [t(min)] 60(0)-60(1)-40(2)-40(7)-25(8)-25(10)-10(11)-10(14)-60(14.1)-60(20) at a flow rate of 0.4 ml/min. Analytes were detected using a Thermo LCQDuo ion trap mass spectrometer (Thermo Electron Corporation, Waltham, MA) with electrospray ionization in positive ion mode. Electrospray ionization conditions were set as follows: spray voltage of 4.5 kV, sheath gas flow 90 units, auxiliary gas flow rates 20 units, and capillary temperature 280°C with a voltage of 20 V. Column effluent was split 1:3 between mass spectrometer and waste. The mass transitions used were m/z 380.3 → 334.1 for licofelone, 556.1 → 380.0 for M1, 396.1 → 350.1 for M2 and M4, 572.1 → 396.0 for M3 and M5, and 366.3 → 348.2 for the ISTD. Collision energy was set to 35% each. Rates of M2 and M4 formation were quantified by comparison of peak area ratios of the incubation samples with those of known concentrations of reference compounds. Calibration samples were processed in the same way as described for incubation samples. Calibration curves were linear over the concentration range used (r2 > 0.99). Because of the lack of adequate reference compounds for quantitation of M1 and M3, a mean glucuronide conversion coefficient was established by comparison of free and total M2 concentration in selected microsomal incubation samples and samples from hepatocyte cultures (i.e., M2 concentrations were determined before and after hydrolysis with 5% NaOH; incubation at 50°C for 90 min). This coefficient was also applied to estimate M1 concentrations. In samples incubated with UGT isoforms only peak area ratios of M1 were determined.
HPLC/UV Analysis. Paclitaxel and 6α-hydroxy-paclitaxel were separated on a Phenomenex Synergi (Torrance, CA) 4U MAX-RP 150 × 4.6-mm column with a mixture of methanol (A) and 0.01 M KH2PO4 buffer, pH 2.3 (B), with a flow rate of 1.5 ml/min at 40°C using an HP1090 instrument (Hewlett Packard, Palo Alto, CA). The gradient profile was percent B [t(min)] 35(0)-35(10)-10(14)-10(17)-35(19)-35(25), and the column effluent was monitored at 230 nm. Retention times of paclitaxel and metabolites were as follows: 3.60 min (3α-hydroxy-paclitaxel), 5.70 min (6α-hydroxy-paclitaxel), and 7.30 min (paclitaxel).
6α-Hydroxy-paclitaxel concentrations were determined by means of external calibration using a calibration curve that was established by linear regression of concentration/peak area data pairs of paclitaxel reference standards. Similar extinction coefficients for paclitaxel and its hydroxy-metabolites were concluded based on the fact that the sum of peak areas determined for paclitaxel and its metabolites 6α-hydroxy-paclitaxel and 3-hydroxy-paclitaxel remained essentially unchanged during microsomal metabolism assays.
Results
Phase I and Phase II Assays. Initial in vitro phase I biotransformations using liver microsomes from rat and human showed a high metabolic stability of licofelone. Two hydroxy-metabolites, M2 and M4, were detected, but concentrations, determined after 30 or 60 min incubation of 10 to 100 μM licofelone, reflected that less than 2% of the substrate was converted. Assays with diclofenac were performed as control reactions (Table 1). In incubations with HLM, only 0.06% of the initial substrate concentrations was converted into M2 compared with 1.8% M4. With RLM as test system, 1.1% M2 and 1.8% M4 were found. The formation of M4 was somewhat surprising because in plasma samples from laboratory animals, this metabolite was not detected, and in humans, quantifiable concentrations were determined only in plasma samples, which were collected during a clinical study after administration of supratherapeutic doses to determine the maximum tolerated dose.
To identify the metabolizing enzymes involved in the formation of hydroxy-metabolites, experiments with different P450 isoforms were performed. Although the direct hydroxylation of licofelone has been considered as a quantitatively negligible pathway, a primary screening experiment, conducted with 10 μM licofelone, gave evidence that CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2J2, and CYP3A4 are involved in the oxidation of licofelone. Thus, in a second experiment, 100 μM licofelone was incubated in the presence of 200 pmol/ml of these isoenzymes. In all the incubations, both hydroxy-metabolites were detected, but M4 concentrations always exceeded those of M2 (Fig. 2). With more than 600 ng/ml M4, which corresponds to a substrate turnover of approximately 1.6%, CYP2C9*1 showed the highest activity toward formation of M4 followed by CYP2J2 > CYP3A4 > CYP2C8 > CYP2C19 > CYP2D6*1. However, M4 concentrations formed with CYP2J2, CYP3A4, CYP2C8, CYP2C19, and CYP2D6*1 were less than the third compared with the amount found in incubations with CYP2C9*1. With 67.8 ng/ml, only CYP2J2 formed significant amounts of M2. CYP2C9*1 and CYP2C8 also formed M2 but to a much lower extent.
In vitro glucuronidations using microsomes showed a rapid conversion of licofelone into the corresponding 1-O-acyl glucuronide, M1 (Table 1). The chemical structure of this metabolite was previously shown by analysis of biological samples and chemical synthesis (Kirschning et al., 1997). Again, glucuronidation of diclofenac was used as a positive control (Table 1). To identify the UGT isoforms involved in the formation of M1, 100 μM licofelone was incubated with different human recombinant UGT isoforms. M1 concentrations were expressed as M1/ISTD peak area ratios, and results are summarized in Table 2. UGT isoforms UGT2B7, UGT1A9, and UGT1A3 were the most important enzymes. Minor activities were also observed with UGT1A7 and UGT2B15. The determination of kinetic constants was not the objective of this UGT screening assay.
Although the enzymes involved in the direct hydroxylation and glucuronidation of licofelone were identified, these data also showed that standard assays with microsomes were not appropriate to explain the high exposure of M2 in humans. Therefore, experiments with freshly isolated human hepatocytes were performed.
Biotransformation of Licofelone in Human Hepatocytes. Licofelone (30 and 100 μM) was added to primary human hepatocytes, and aliquots of the supernatant were analyzed by LC/MS. The rate of metabolite formation was not linear throughout the observation period but significantly decreased after 4 h. Biotransformation rates, which were derived from metabolite concentrations determined at 4 and 24 h after incubation of 30 and 100 μM licofelone, are given in Table 3. The glucuronide M1 accounted for 85% of the sum of metabolites and was clearly dominating after 24 h. Both hydroxy-metabolites M2 and M4 were present, but in contrast to phase I metabolism experiments using microsomes, M2 concentrations substantially exceeded those of M4, in particular if corresponding glucuronides were considered. The observation that M3 concentrations substantially exceeded those of M2 was interpreted as a result of the high glucuronidation activity of hepatocytes. M5 was detected, but throughout the observation period concentrations remained below the limit of quantification.
The metabolite profile in experiments performed with 100 μM licofelone was almost identical; however, the total biotransformation was only about half the rate observed with 30 μM. In addition, oxidative metabolism was less pronounced because M1 accounted for even more than 95% of the overall metabolism.
Dual-Activity Assays Using Liver Microsomes. The metabolism experiments using human hepatocytes showed that both biotransformation activities (i.e., hydroxylation and glucuronidation) were required to achieve an M2/M4 ratio that at least qualitatively reflects the human in vivo metabolism. However, the integrated phase I/II metabolic capacity of hepatocytes does not allow the elucidation of metabolic pathways. Furthermore, this test system is less suitable to identify the enzymes responsible for biotransformations of interest. Finally, hepatocytes from dogs, mice, or monkeys are not ubiquitously available, which limits their use for interspecies comparison of drug metabolism. Glucuronidation of xenobiotics is catalyzed by UGTs, a superfamily of membrane-bound enzymes. UGTs largely reside at the luminal face of the endoplasmic reticulum, and both the substrate and the cofactor UDPGA have limited access to the active center. In vitro, high glucuronidation activities are generally achieved by addition of membrane-disrupting polyoxyethylene-based detergents like Brij or Triton X. However, this procedure has a substantial impact on P450 activity, and the simultaneous investigation of P450- and UGT-catalyzed reactions was not possible. Substitution of these conventional ingredients with the channel-forming antibiotic alamethicin results in an activation of UGTs without affecting P450-dependent biotransformations (Fisher et al., 2000). Thus, this dual-activity assay system was established and applied to investigate the metabolism of licofelone.
The experiments were performed with HLM, CLM, DLM, NZLM, RLM, and MLM. As cofactors, NADPH and UDPGA were added. To determine the effect of the glucuronidation activity of each microsomal preparation, control experiments without UDPGA were included. The results obtained after incubation of 10 μM licofelone for 30 min are summarized in Table 4. In the absence of UDPGA, no glucuronides were detected, and the extent of hydroxylation of licofelone was as low as observed in initial phase I assays. M2 was detected in liver microsomes from all the species except for DLM, but concentrations were below the limit of quantification in HLM and MLM. The most relevant biotransformation was observed in NZLM with a relative content of 8% M2 and 0.18% M4. Supplementation of UDPGA resulted in a completely different metabolite pattern. In HLM, the overall biotransformation increased from 2.0% to 34.9%, and the main metabolite was M3 (i.e., the glucuronide of M2; relative content: 26.8%). The content of M1 was 7.3%. With microsomes from animal species, M1 was dominating the other metabolites. M3 was either not detected (DLM, RLM, MLM) or at comparably low concentrations (CLM, NZLM). This experiment strongly indicated principal differences between humans and animal species. Considering the M1 concentration, an ultrarapid glucuronidation of M2 appeared unlikely. Therefore, the only plausible explanation of this result was that M1 was hydroxylated to M3 followed by partial hydrolysis of the acyl glucuronide, which led to the determined concentrations of M2. In contrast to the experiments with HLM, with microsomes from monkey, rabbit, and rat, M2 concentrations were lower in the presence of UDPGA, thus suggesting the opposite metabolic pathway (i.e., hydroxylation followed by glucuronidation).
With HLM, the experiment was repeated with licofelone concentrations of 30 and 100 μM. In Fig. 3, the biotransformation rates, relative to the initial substrate concentrations, and the concentrations are given. With 10 and 30 μM licofelone, the sum of glucuronides (M1 + M3) was 3.41 and 9.92 μM, which correspond to biotransformation rates of 34 and 33%. With 100 μl of licofelone, the relative glucuronide content was 25%. The biotransformation rate of M2 and M3 decreased with ascending substrate concentrations. The highest M3 concentration (5.02 ± 0.11 μM) was observed at 30 μM, whereas only 2.41 ± 0.56 μM was determined after incubation of 100 μM licofelone. These data indicate that the glucuronidation capacity of the test system was limited at concentrations greater than 30 μM. In contrast, hydroxylation of M1 was partly inhibited in incubations with 100 μM. M4 concentrations increased with ascending substrate concentrations in an almost proportional manner. The biotransformation rate to M4 was essentially constant.
In Vitro Metabolism of M1. To confirm the proposed metabolic pathway, conventional phase I metabolism experiments with HLM, CLM, RLM, and MLM at a concentration of 10 μM M1 were performed. M1 was prepared by enzymatic glucuronidation and purified by preparative HPLC. As summarized in Table 5, a quantifiable extent of M1 hydroxylation into M3 was obtained with HLM (>90%) and CLM (35%). In contrast, in incubations with MLM and RLM, only traces of M3 were detected. In control incubations in Tris-HCl buffer, M1 remained stable and did not show significant hydrolysis, and a chemical oxidation was not observed at all. M5, the glucuronide of M4, was not detected in any experiment. In addition, in vitro assays were performed with several P450 isoforms, but the formation of M3 was only observed in incubations with CYP2C8 (50%).
Inhibition of CYP2C8. Montelukast, a selective and potent CYP2C8 inhibitor (Walsky et al., 2005a), was used to verify the observed CYP2C8 specificity of M1 hydroxylation. Dual-activity assays with HLM were repeated without and with increasing montelukast concentrations. The effect of 1, 10, and 30 μM montelukast on the formation of M3 is shown in Table 6. For M2 and M3, a concentration-dependent inhibition was obtained. In contrast, M1 concentrations increased with ascending inhibitor concentrations, which reflects that only the hydroxylation of M1 rather than its UGT-mediated formation is affected by montelukast. To achieve a substantial effect on M2 and M3, at least 10 μM montelukast had to be added to the incubation mixture, which is substantially above the reported Ki value of 20 nM (Walsky et al., 2005a). However, the inhibition potency of montelukast is strongly dependent on the protein concentration (Walsky et al., 2005b). Using amodiaquine N-deethylation as model reaction, the IC50 increased from 0.23 μM at a protein concentration of 0.025 mg/ml to 18 μM at 2 mg/ml. In the dualactivity assays, the protein concentration was also 2 mg/ml (1 mg/ml microsomal protein in buffer containing 0.1% BSA), and the observed 75% inhibition of M2 formation (0.11% instead of 0.45%) and the 69% inhibition of M3 formation (8.1% instead of 25.8%) in the presence of 10 μM montelukast are in line with the published inhibition potency. In the presence of 30 μM montelukast, M2 concentrations were below the limit of quantification. The formation of M4 was also inhibited in a concentration-dependent manner by montelukast, which can be explained by the involvement of CYP2C8 in this reaction but also by the inhibition of CYP2C9 at high montelukast concentrations (Walsky et al., 2005a).
Influence of M1 on CYP2C8. In dual-activity assays using HLM, as well as in human hepatocytes, ascending substrate concentration led to a shift of relative concentrations of metabolites. At high licofelone concentrations, the relative content of M1 was higher compared with incubations at lower substrate concentrations. Accordingly, the relative concentrations of hydroxy-metabolites decreased with ascending substrate concentrations. M1 is a 1-O-acyl glucuronide, and these metabolites are generally known to be chemically unstable. Dependent on the nature of the aglycon, a shift to C2, C3, and C4 of the glucuronic acid may occur, and as a final reaction, a transesterification of the aglycon to proteins was described (Spahn-Langguth et al., 1996; Bailey and Dickinson, 2003; Stachulski, 2007). As shown above, as far as quantifiable with the applied analytical method, M1 remained stable during in vitro metabolism assays. To confirm this observation in a functional setting, 10 and 30 μM M1 were preincubated with HLMs and recombinant CYP2C8 microsomes for 15 min before the addition of NADPH, which triggers the hydroxylation activity. As shown in Fig. 4, in incubation with microsomes, M3 concentrations increased proportionally with the initial substrate concentration, and preincubation had no impact on the reaction. In the analogous experiment with CYP2C8, M3 concentrations were similar at 10 and 30 μM, but again preincubation had no effect on the hydroxylation activity.
Effect of Licofelone and M1 on Paclitaxel-6α-Hydroxylation. To evaluate the drug interaction potential of licofelone and M1 on CYP2C8-dependent reactions, the effect of both compounds on the 6α-hydroxylation of paclitaxel was investigated using the CYP2C8 standard probe substrate paclitaxel (Dai et al., 2001). The results are illustrated in Fig. 5. In HLM and CYP2C8, licofelone inhibited the formation of 6α-hydroxy-paclitaxel in a concentration-dependent manner. In the presence of 10 μM licofelone, only a weak inhibition was observed, and a licofelone concentration of 100 μM was required to achieve a substantial inhibition (approximately 80%) when compared with the control experiment. The effect of M1 on CYP2C8 activity was even less pronounced. A relevant inhibition was only observed at 100 μM. Control reactions with the CYP2C8-specific inhibitor montelukast verified the validity of this inhibition experiment. Systemic concentrations of licofelone and M1 at or more than 30 μM are not achieved in humans in vivo; therefore, a clinically relevant inhibition of CYP2C8 during therapeutic use of licofelone can be excluded.
Extrahepatic Metabolism of Licofelone. Results of dual-activity assays performed with pooled microsomes from tissue samples of human intestine, kidney, and lung and cynomolgus monkey intestine are summarized in Table 7. M1 was the major extrahepatic metabolite. Both hydroxy-metabolites of licofelone could not be detected in any of these incubation mixtures. With 33.0%, the highest extrahepatic biotransformation was observed in HKM followed by HIM (21%). Whereas in HKM traces of M3 also could be detected, in HIM only M1 could be detected. HLuM did not show significant metabolism activity toward licofelone. Microsomes from monkey intestinal mucosa also showed substantial glucuronidation of licofelone. Low amounts of M3 were also determined. The total biotransformation in CIM was almost 2-fold higher compared with HIM. With increasing licofelone concentrations, biotransformation rates in microsomes from extrahepatic tissue also decreased as observed with liver microsomes.
Discussion
The results of in vitro metabolism experiments with licofelone showed that glucuronidation of the carboxylic acid followed by CYP2C8-catalyzed hydroxylation of the acyl glucuronide M1 represents the primary elimination pathway of this compound. Direct hydroxylation to M2 and M4 was also observed in these metabolism experiments, but at least in humans the contribution of these pathways to systemic clearance is negligible. A screening with recombinant P450 isoforms provided qualitative information regarding licofelone hydroxylation enzymes, but determination of the enzyme kinetic constants was not possible. At a substrate concentration of 100 μM (= 38,000 ng/ml), maximum metabolite concentrations of 600 ng/ml were observed with CYP2C9, which correspond to a relative conversion of 1.6%. At an initial concentration of 10 μM, only traces of hydroxylated metabolites were detected, and the determination of valid conversion rates at a concentration range that would allow the determination of Vmax and KM was not possible. Therefore, the analysis was limited to the identification of all those P450 isoforms that catalyze the formation of M2 and M4. Aryl hydroxylation of licofelone to M4, which was predominant in microsomal phase I assays, was catalyzed in decreasing order of importance by CYP2C9 >> CYP2J2 > CYP3A4 > CYP2C8 > CYP2C19 > CYP2D6. M4 was also found in incubations with liver microsomes from animal species. In plasma samples this metabolite was not detected. In vivo, a high glucuronidation-biliary excretion capacity may prevent a systemic exposure of this metabolite. M2 was detected, again in decreasing order, in incubations with CYP2J2 > CYP2C9 > CYP2C8. CYP2C isoforms and CYP3A4 belong to the most important drug-metabolizing enzymes, but CYP2J2 is mainly known for its role in the metabolism of arachidonic acid into epoxyeicosatrienoic acid derivatives. The antihistamines astemizole and ebastine have been described as the first and only xenobiotic compounds that are metabolized by CYP2J2 (Hashizume et al., 2002; Matsumoto et al., 2002). The metabolism of licofelone by CYP2J2 also supports the hypothesis that the mechanism of dual inhibition of COX and 5-LOX is based on conformation similarities between licofelone and arachidonic acid (Laufer et al., 1994).
In further functional assays using HLM and P450-specific substrates [midazolam and testosterone for CYP3A4, diclofenac for CYP2C9, (S)-mephenytoin for CYP2C19, and dextromethorphan for CYP2D6], it was shown that, up to concentrations of 30 μM, licofelone had no significant impact on the activity of these P450-specific biotransformations (Albrecht W, internal report). The effect of licofelone on the activity of CYP2J2 still remains to be investigated.
In vivo, the formation of M2 via the acyl glucuronide M1 represents the main pathway. The UGTs UGT2B7, UGT1A9, and UGT1A3 were identified as the most relevant isoforms involved in the formation of M1, an observation that is in line with results of UGT screenings with other carboxylic acids. Recently Kuehl et al. (2005) identified the UGT isoforms involved in the acyl glucuronidation of eight different NSAIDs. Apart from individual differences between the substrates, the isoforms UGT2B7, UGT1A3, UGT1A9, and UGT1A8 used all tested compounds as substrates. The rank order UGT2B7 >> UGT1A3 ≈ UGT1A9 for glucuronidation of carboxylic acids was also reported by Sakaguchi et al. (2004). UGT1A8 was not included in the present investigation, but because of its low substrate specificity it may also be involved in the formation of M1.
Hepatically formed acyl glucuronides are generally excreted by carrier-mediated processes across either the canalicular or basolateral membrane (Zamek-Gliszczynski et al., 2006), and it is hypothesized that this route also dominates in the disposition of licofelone—both in humans and animal species. In vitro metabolism assays showed that in humans M1 is also hydroxylated to M3 in a CYP2C8-dependent reaction. This enzymatic step represents the pivotal step for the systemic availability of M2. The site of hydrolysis of M3 is not unequivocally clear. However, in biotransformations of licofelone with hepatocytes, M3 was the predominant metabolite, which suggests that hepatocytes contain a low enzymatic activity toward M3 hydrolysis. Furthermore, in circulating blood of humans, only traces of M3 were found (Albrecht W, internal report), which excludes the possibility that M3 enters the systemic circulation and is hydrolyzed by plasma esterases. Therefore, the biliary excretion of M3 and its hydrolysis in the intestinal tract followed by absorption of M2 remains the most likely pathway. This hypothesis is supported by the observation that after single administration of licofelone, M2 plasma concentrations were low but became relevant after repeated, twice-daily drug administration.
Hydroxylation of either licofelone to M2 or M1 to M3 introduces a chiral center (quaternary carbon in the pyrrolizine ring) into the molecule. However, the stereospecificity of this biotransformation was not addressed in this study.
CYP2C8, which was identified as the only P450 isoform involved in M1 hydroxylation, was previously reported as the enzyme responsible for hydroxylation of diclofenac acyl glucuronide (Kumar et al., 2002) and estradiol-17β-glucuronide (Delaforge et al., 2005). Despite the similarity of the elimination pathway of diclofenac and licofelone, the results published by Kumar et al. (2002) show an important difference between these two drugs. In incubations with microsomes, a rapid degradation of diclofenac acyl glucuronide was observed, irrespective of the presence or absence of NADPH. This result can be most likely attributed to the known chemical instability of the acyl glucuronide (Grillo et al., 2003). In contrast, M1 remained stable throughout the incubation period of 60 min with dog and rat liver microsomes with or without supplemented NADPH.
Incubation experiments with different P450 isoforms and with liver microsomes in the presence of the CYP2C8-specific inhibitor montelukast showed that hydroxylation of M1 is solely catalyzed by this isoform. In the liver, CYP2C8 constitutes approximately 7% of the total microsomal P450 content, but its expression was also shown in kidney, intestine, adrenal gland, brain, mammary gland, ovary, and heart, as well as in breast cancer tumors (Totah and Rettie, 2005). In dual-activity assays with licofelone and HKM, M3 was identified as a metabolite that is in line with the reported CYP2C8 expression in kidneys. In contrast, with HIM, another potential CYP2C8 source, only the formation of M1 was observed.
CYP2C8 is involved in the metabolism of many drugs (Totah and Rettie, 2005); therefore, CYP2C8 substrates should be investigated for their potential to cause clinically relevant drug interactions. For example, based on in vitro inhibition studies, it has been suggested that CYP2C8 inhibition by gemfibrozil acyl glucuronide substantially contributed to the severe toxicity of the lipid-lowering drug cerivastatin when coadministered with gemfibrozil (Shitara et al., 2004; Ogilvie et al., 2006). Based on the results of the present study, neither licofelone nor M1 modulates the activity of CYP2C8 at clinically relevant concentrations. Therefore, drug interactions with CYP2C8 substrates are unlikely. Vice versa, the inhibition of CYP2C8 by coadministered drugs such as montelukast may lead to decreased systemic exposure of M2, which, however, is without any consequence for the safety and efficacy of licofelone.
Footnotes
-
The hepatocyte work was partially supported by the Federal Ministry of Education and Research (BMBF 0313081B).
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.108.020347.
-
ABBREVIATIONS: licofelone, 2-[6-(4-chlorophenyl)-2,2-dimethyl-7-phenyl-2,3-dihydro-1H-pyrrolizin-5-yl] acetic acid, ML3000; COX, cyclooxygenase; 5-LOX, 5-lipoxygenase; NSAID, nonsteroidal anti-inflammatory drug; P450, cytochrome P450; G6P, glucose-6-phosphate; UDPGA, UDP-glucuronic acid; UGT, UDP glucuronosyltransferase; DLM, beagle dog liver microsomes; NZLM, New Zealand White rabbit liver microsomes; HIM, human intestinal microsomes; HKM, human kidney microsomes; HLuM, human lung microsomes; HLM, human liver microsomes; CLM, cynomolgus monkey liver microsomes; RLM, Sprague-Dawley rat liver microsomes; MLM, CD-1 mouse liver microsomes; CIM, cynomolgus monkey intestinal microsomes; BSA, bovine serum albumin; ISTD, internal standard; ACN, acetonitrile; LC/MS/MS, liquid chromatography/tandem mass spectrometry; HPLC, high-performance liquid chromatography.
-
↵1 Current affiliation: Department of Traumatology, TU Munich, Munich, Germany.
- Received January 2, 2008.
- Accepted February 8, 2008.
- The American Society for Pharmacology and Experimental Therapeutics