Abstract
Midazolam (MDZ) is one of the most commonly used in vivo and in vitro CYP3A4 probe substrates for drug-drug interactions (DDI) studies. The major metabolic pathway of MDZ in humans consists of the CYP3A4-mediated 1′-hydroxylation followed by urinary excretion as 1′-O-glucuronide derivative. In the present study, following incubation of MDZ with human liver microsomes supplemented with UDP-glucuronic acid, two major high-performance liquid chromatography (HPLC) peaks were isolated. HPLC and liquid chromatography/tandem mass spectrometry analyses identified these two metabolites as quaternary direct N-glucuronides of MDZ, thus revealing an additional metabolic pathway for MDZ. 1H NMR spectrometry studies were performed showing that these two glucuronides were β-N-glucuronides, which could be considered as two different conformers of the same molecule. According to molecular modeling experiments, the two glucuronide derivatives could be involved in atropoisomerism equilibrium. The formation of MDZ N-glucuronide exhibited moderate intersubject variability (at most 4.5-fold difference, n = 10). Among the recombinant human UDP glucuronosyltransferase (UGT) isoforms tested, only isoform UGT1A4 catalyzed the N-glucuronidation of MDZ fitting a Michaelis-Menten model. Km and Vmax values were 29.9 ± 2.4 μM and 659.6 ± 19.0 pmol/min/mg protein, respectively. The N-glucuronide derivative was found in human hepatocytes incubated under control conditions but also in the presence of the well known CYP3A4 inhibitor, ketoconazole. In the context of the in vitro study of CYP3A4-mediated DDI using MDZ and ketoconazole, direct MDZ N-glucuronidation may partly compensate the decrease in MDZ metabolic clearance caused by the addition of the inhibitor, thus potentially leading to underestimation, at least in vitro, of the extent of DDI.
Because a large number of currently available drugs and future drugs will be metabolized by the members of the CYP3A subfamily, the potential for drug-drug interaction (DDI) is substantial. DDIs involving the inhibition and induction of CYP3A4 are of great scientific and clinical relevance. Indeed, drugs with potent CYP3A4 inhibitory properties have been implicated in significant CYP3A4-mediated DDIs. Interactions of the benzodiazepines with the azole antifungal agents and especially the inhibition of CYP3A-mediated midazolam (MDZ) metabolism by ketoconazole have been widely studied. Concomitant administration of both drugs results in large, variable, and highly significant increases (5–16-fold) in MDZ exposure, depending on the dose regimen of ketoconazole used. Table 1 summarizes available clinical data on the effects of ketoconazole on MDZ plasma levels following p.o. or i.v. administration of MDZ.
MDZ is a short-acting water-soluble imidazobenzodiazepine (Fig. 1) extensively used in clinical practice mainly for induction and maintenance of anesthesia, sedation for diagnostic and therapeutic procedures, and also as an oral hypnotic agent (Reves et al., 1985). MDZ is a well known CYP3A substrate because its metabolism has been the focus of many in vitro investigations (Fabre et al., 1988b; Kronbach et al., 1989; Wrighton and Ring, 1994; Ghosal et al., 1996; Maenpaa et al., 1998; Hosea et al., 2000; Wang et al., 2000). MDZ biotransformation is mediated by at least three different CYP3A isoenzymes: CYP3A4, CYP3A5, and CYP3A7 (Gorski et al., 1994; Kuehl et al., 2001). Because CYP3A7 is principally expressed in fetal tissues, CYP3A4 and CYP3A5 represent the main cytochrome P450 (P450) isoforms in adult liver and intestine (Guengerich, 1995).
MDZ biotransformation yields two primary hydroxylated metabolites: 1′-hydroxy-MDZ (1′-OH-MDZ) and 4-hydroxy-MDZ. 1′-OH-MDZ represents the main metabolite because it accounts for 95% of net intrinsic clearance of MDZ in human liver microsomes (von Moltke et al., 1996). In vivo, when administered p.o. to humans, MDZ is rapidly absorbed, and the amount of 1′-OH-MDZ excreted as conjugate in the urine of healthy volunteers reaches 75% of the initial administered dose versus 4% for the minor primary metabolite, 4-hydroxy-MDZ, and 6% for a minor secondary metabolite, 1′,4-dihydroxy-MDZ. Both metabolic routes are also catalyzed by CYP3A4, and hydroxylated metabolites are excreted as glucuronoconjugates (Heizmann and Ziegler, 1981; Dundee et al., 1984).
In recent years, MDZ has emerged as one of the best and most widely used in vitro and in vivo metabolic probes for prediction of CYP3A activity (von Moltke et al., 1996; Thummel and Wilkinson, 1998; Galetin et al., 2005) because it meets most, if not all, of the necessary criteria suggested for such applications. Indeed, MDZ is a substrate of one well known P450 isoform; it is highly sensitive to changes in status/activity of the respective P450 enzyme; it is unaffected by P-glycoprotein or other known transporters; and it presents negligible pharmacodynamic effects (adverse effects) at the dose used for probe studies (Bjornsson et al., 2003). In addition, MDZ is commercially available, exhibits suitable pharmacokinetic profile, and can be administered both p.o. and i.v. MDZ can provide a measure of CYP3A4/5 activity relative to both intestinal and hepatic metabolism. Tables 2 and 3 list the preferred in vitro and in vivo CYP3A4 probe substrates, inhibitors, and inducers recommended by the U.S. Food and Drug Administration Guidance for Industry (Drug Interaction Studies—Study Design, Data Analysis, and Implications for Dosing and Labeling, http://www.fda.gov/cder/guidance/6695dft.htm) released in September 2006.
In vitro inhibition of human liver microsomal metabolism of MDZ to 1′-OH-MDZ by ketoconazole was initially studied by Gascon and Dayer (1991). Wrighton and Ring (1994) described ketoconazole as a potent, selective, and noncompetitive inhibitor with Ki ranging between 0.0037 and 0.18 μM according to the authors (Bourrié et al., 1996; von Moltke et al., 1996; Galetin et al., 2005). Therefore, ketoconazole, for which plasma Cmax concentrations can be as high as 25 μM in humans, is expected to be a potent inhibitor of metabolism of all the CYP3A substrates, thus causing very significant CYP3A4-mediated DDIs in vivo.
Therefore, MDZ and ketoconazole are extensively used both in vitro and in vivo for the prediction of CYP3A-mediated DDIs, the former as a phenotypic marker of CYP3A4/5 metabolic capacity, the latter as a reference as specific and potent CYP3A4/5 inhibitor. The central hypothesis for the prediction of DDIs involving CYP3A4 inhibition by a compound of interest, both in vitro and in vivo, is that the decrease in the overall observed clearance of MDZ is caused by the coadministration of (or in vitro coincubation with) a CYP3A4 inhibitor (i.e., ketoconazole or other novel compounds exhibiting CYP3A4 inhibitory potency). This can be assigned entirely and exclusively to the CYP3A4/5 enzymes and therefore can be used to predict the magnitude of the clinical interaction expected between the potential inhibitor under investigation and any other coadministered CYP3A4 substrate.
However, in vitro studies showed that MDZ could also be directly conjugated with UDP-glucuronic acid (UDPGA) (Siddle et al., 2003). Moreover, additional experiments conducted in our laboratory showed that ketoconazole did not fully inhibit the in vitro metabolic clearance of MDZ in human hepatocytes. Such results would suggest that, at least in vitro, the metabolism of MDZ is not exclusively dependent on CYP3A4/5, thus warranting further investigation of MDZ metabolic pathways and of the ketoconazole/MDZ metabolic interaction.
The purpose of the present work was to study in more detail the in vitro metabolic pathways of MDZ in humans and to elucidate whether an alternative metabolic route such as direct N-glucuronidation could be responsible for a metabolic shift between CYP3A4/5 and UDP glucuronosyltransferase (UGT), thus possibly leading to a certain extent of misinterpretation of DDI study results both in vitro and in vivo.
Materials and Methods
Compounds. MDZ and its hydroxylated 1′-OH-MDZ metabolite were purchased from Ultrafine Chemicals (Manchester, UK). The 1′-OH-MDZ glucuronide (Glu-O-MDZ) was synthesized by the Isotope Chemistry and Metabolites Synthesis Department of Sanofi Aventis Recherche (Chilly-Mazarin, France). Dimethyl sulfoxide (DMSO), ketoconazole, UDPGA, alamethicin, ethanolamine, transferrin, linoleic acid, ascorbic acid, insulin, l-arginin, and glucagon were obtained from Sigma (St. Louis, MO). Ham's F-12 and Williams' E media, l-glutamine, HEPES, sodium pyruvate, penicillin, and streptomycin were purchased from Invitrogen (Carlsbad, CA). The other chemicals and solvents used were all of analytical grade.
Reagents. Recombinant human UGTs (supersomes: UGT1A1, UGT1A3, UGT1A4, UGT1A6, UGT1A7, UGT1A8, UGT1A9, UGT1A10, UGT2B4, UGT2B7, UGT2B15, and UGT2B17) expressed in baculovirus-infected insect cells and control microsomes from insect cells infected with wild-type baculovirus were all obtained from BD Gentest (Woburn, MA).
Identification of MDZ N-Glucuronides by High-Performance Liquid Chromatography/Electrospray Ionization/Mass Spectrometry. Chromatography was performed with a high-performance liquid chromatography (HPLC) system Agilent 1100 (Agilent, Santa Clara, CA), which was equipped with a reverse-phase column (SunFire C18 3.5-μm particle size, 3.0 i.d. × 150-mm length) (Waters, Milford, MA). The column temperature was set at 35°C using a column oven. The mobile phase was composed of distilled water with 0.1% (v/v) formic acid (A) and acetonitrile/methanol (70:30, v/v) (B). Elution was performed using the following gradient of solvent B: 0 to 5 min at 15% and then 55% over 20 min. The flow rate was 0.4 ml/min. Under these analytical conditions, the retention time for MDZ was 14.8 min. The column outlet was connected to a Finnigan linear ion trap mass spectrometer (Thermo Finnigan MAT, San Jose, CA) operating in electrospray positive ionization mode.
Preparation of M1 and M2. After a 3-h incubation period of 98 μg of MDZ with recombinant UGT1A4 in the presence of UDPGA, the HPLC eluate containing all the MDZ-N-glucuronides converted from MDZ was collected. At the end of this process, the total amount estimated for metabolites M1 and M2 was approximately 50 μg. These metabolites were in solution in a water/acetonitrile (50:50, v/v) mixture.
On-Line NMR. The HPLC/UV/solid-phase extraction (SPE)/NMR measurements were carried out using an Agilent 1100 series chromatographic system (autosampler, degasser, quaternary pump, oven, and UV variable wavelength detector), a Prospekt 2 SPE robot (Bruker Biospin, Rheinstetten, Germany and Spark Holland, Emmen, Holland), and an AVANCE 500 NMR spectrometer (Bruker Biospin, Wissembourg, France). Injections were performed by autosampler into a reverse-phase HPLC column. M1 and M2 were automatically detected by UV absorbance and trapped onto SPE cartridges after postcolumn addition of water. The metabolites present in M1 and M2 peaks then were flushed to the NMR flow cell to record NMR spectra.
HPLC/UV conditions. The chromatography was controlled using the HyStar 2.3 software (Bruker Daltonik, Rheinstetten, Germany). Elution was performed using eluent A [water (Thermo Electron Corporation, Waltham, MA) + 0.01% trifluoroacetic acid] and eluent B (acetonitrile, Riedel de Haen, Germany). The initial conditions consisted of 20% eluent B and 80% eluent A. The flow rate was set at 0.4 ml/min through an Inertsil ODS-3 C18 2.1 mm i.d., 150-mm length, 5-μm particle size HPLC column (GL Sciences, Torrance, CA). The column temperature was maintained in an oven at 20°C. Gradient elution was performed using eluents A and B with the following gradient: 20% B to 25% B in 20 min, 25% B to 90% B in 2 min, 90% B held for 3 min, and post-time of 2 min under initial conditions. The M1/M2 solution was dried under N2 stream. The total amount of M1/M2 mixture recovered was estimated to be approximately 25 μg (estimated recovery of 50%). The dried extract was then dissolved in DMSO (100 μl). Ten microliters of this M1/M2 solution was injected. UV wavelength detection was set at 254 nm. Retention times observed for metabolites M1 and M2 were 6.62 and 6.92 min, respectively.
SPE conditions. The Prospekt 2 SPE robot was controlled using the Bruker HyStar 2.3 software. The type of SPE cartridge used was Hysphere GP (2 mm i.d., 10-mm length, 10–12-μm particle size). Before use, the cartridge was cleaned with acetonitrile and conditioned with water. The postcolumn water make-up flow was set at 0.8 ml/min and was performed using a K100 HPLC pump (Knauer, Berlin, Germany). The total flow going through the trapping cartridge was 1.2 ml/min. A four-step trapping process (4 × 10 μl injected) was carried out for each metabolite M1 and M2 to increase the amount of compound available for NMR experiments. The peak of interest was automatically detected by UV absorbance at 254 nm for trapping. Trapped M1 and M2 were dried under a stream of N2 for 40 min to remove residual solvents. Pure deuterated acetonitrile (“100% grade,” Euriso-Top, Saint-Aubin, France) was used to flush M1 and M2 from the SPE cartridges directly into the NMR cell for spectroscopic analysis.
NMR conditions. From the SPE cartridges, trapped compounds were eluted into a Bruker AVANCE 500-MHz spectrometer operating at 500.13 MHz for proton and equipped with a 3-mm 1H/13C probe, inverse and gradient z, fitted with a 60-μl flow cell (30-μl active volume). The NMR was controlled using XwinNMR and IconNMR software (Bruker Daltonik). The concentrated metabolites M1 and M2 flushed in the flow probe were of sufficient purity and quantity to provide 1H one-dimensional (1D) data and 1H two-dimensional (2D) homonuclear polarization transfer spectrum.
The 1H 1D proton spectrum was recorded in 1024 scans using a 64-K transient and a spectral width of 10,330 Hz (acquisition time 3.17 s). The data were apodized with an exponential window function (0.3 Hz). The COSY-GPQF spectrum was obtained with F2 and F1 spectral widths of 6666 Hz in 2-K data points for 177 t1 increments and 128 scans/t1. The data were apodized with a sine bell window function in both dimensions and zero-filled in the F1 dimension to 512 data points. The 1H NMR spectra were calibrated on acetonitrile signal (1.95 ppm).
Off-Line NMR. HPLC/UV/mass spectrometry (MS) experiments were performed to collect M1 and M2 metabolites. These compounds were collected using an Agilent 1100 series chromatographic system (autosampler, binary pump, oven, UV variable wavelength detector, and fraction collector), a mass spectrometer (quadrupole Agilent 6110), and an AVANCE 600 NMR spectrometer (Bruker Biospin). Metabolites M1 and M2 were automatically detected by MS, collected into vials, and dried under N2. Metabolites M1 and M2 then were put in solution, and NMR spectra were recorded.
HPLC conditions. The chromatography was controlled using the Agilent liquid chromatography (LC)/mass selective detector Chemstation software. Elution was performed using eluent A (water) and eluent B (acetonitrile/methanol, 70:30). The initial conditions consisted of 10% eluent B and 90% eluent A. The flow rate was set at 0.4 ml/min through an YMC-Pack J'Sphere H80 2.1 mm i.d., 250-mm length, 4-μm particle size HPLC column (GL Sciences). The column temperature was maintained in an oven at 30°C. Gradient elution was performed using eluent A and B with the following gradient: 10% B to 28% B in 2 min, 28% B held for 16 min, 28% B to 50% B in 1 min, and post-time of 5 min under initial conditions. M1/M2 solution underwent an evaporation process to remove the major part of acetonitrile. The total amount of M1/M2 was estimated to be 50 μg in solution in water. Fifty microliters (containing approximately 2.5 μg of M1/M2) of this solution was injected for the chromatographic run.
M1 and M2 collection conditions. HPLC, MS, and fraction collector were controlled using the Agilent LC/mass selective detector Chemstation software. After the HPLC column, the flow was split in a ratio of 10:90 (10% MS/90% fraction collector). Metabolites M1 and M2 were detected using the MS trace in single ion monitoring mode (m/z 502) recorded by an Agilent quadrupole 6110 mass spectrometer equipped with electrospray ionization (ESI) source. Retention times observed for metabolites M1 and M2 were 11.2 and 11.8 min, respectively. Elution peaks of metabolites M1 and M2 were collected using an Agilent 1200 fraction collector. This purification process was repeated 20 times. At the end of this process, M1 and M2 were isolated and kept in solution (28% B and 72% A). The M1 and M2 solutions were dried under N2 stream. The amount of each metabolite M1 and M2 was estimated to be 10 μg.
NMR conditions. The dried extracts containing the metabolites M1 and M2 were reconstituted in deuterated DMSO (DMSO-D6, 100% grade, Euriso-Top). Approximately 10 μg of each metabolite M1 and M2 was dissolved in 30 μl of DMSO-D6, and 1.7-mm capillary NMR tubes were filled up with these solutions to perform NMR experiments. Ten milligrams of MDZ was also dissolved in 600 μl of DMSO-D6 and put into 5-mm NMR tubes to compare M1/M2 metabolites and MDZ NMR data. Spectra were recorded using a Bruker AVANCE 600-MHz spectrometer operating at 600.13 MHz for proton (150.91 MHz for 13C) and equipped with a 5-mm TCI 1H/13C, inverse and gradient z, cryoprobe. The concentrated metabolites in the cryoprobe were of sufficient purity and quantity to provide 1H data, as well as 2D homonuclear and heteronuclear polarization transfer spectra.
For MDZ, the 1H 1D proton spectrum was recorded in 64 scans using a 48-K transient and a spectral width of 7861 Hz (acquisition time 3.12 s). The data were apodized with a Gaussian window function (-0.3 Hz, 0.1) and zero-filled (64 K). The 1H 2D COSY-GPQF spectrum was obtained with an F2 and F1 spectral width of 1796 Hz in 6-K data points for 256 t1 increments and 4 scans/t1. The data were apodized with a sine bell window function in both dimensions and zero-filled in the F1 dimension to 2048 data points. The 1H/13C 2D HSQCEDETGPSI spectrum was obtained with F2 spectral width of 4807 Hz in 3-K data points for 256 t1 increments and 4 scans/t1. The data were apodized with a shifted (π/2) sine bell window function in both dimensions. NMR spectra were calibrated on DMSO-D6 signal (2.5 ppm for 1H and 39.6 ppm for 13C).
For M1 and M2 metabolites, the 1H 1D proton spectra were recorded in 2048 scans using a 48-K transient and a spectral width of 7861 Hz (acquisition time 3.12 s). The data were apodized with an exponential window function (0.3 Hz) and zero-filled (64 K). The 1H 2D COSY-quantum-filtered spectrum was obtained with an F2 and F1 spectral width of 5388 Hz in 4-K data points for 512 t1 increments and 64 scans/t1. The data were apodized with a sine bell window function in both dimensions and zero-filled in the F1 dimension to 2048 data points. The 1H/13C 2D HSQCEDETGPSI spectrum was obtained with an F2 spectral width of 4807 Hz in 2-K data points for 256 t1 increments and 112 scans/t1. The data were apodized with a shifted (π/2) sine bell window function in both dimensions. NMR spectra were calibrated on DMSO-D6 signal (2.5 ppm for 1H and 39.6 ppm for 13C).
Molecular Modeling. All the molecular modeling work was done with the MMF94 force field, which is part of SYBYL environment (Tripos, St. Louis, MO).
Human Liver Microsomes and Human Hepatocytes. Liver samples were obtained either from whole livers coming from human donors that were unfit for organ transplant or from donors undergoing partial hepatectomy. These liver samples were termed HTL-x, FH-x, or HL-x. Available demographic information for patients including gender and age is reported in Tables 4 and 5.
Liver Microsome Preparation. Microsomes were prepared from 10 donors using standard techniques adapted by Fabre et al. (1988a). Briefly, microsomes were prepared from frozen liver samples by differential ultracentrifugation. Microsomal pellets were resuspended in 0.1 M potassium phosphate buffer containing 0.1 mM EDTA and 20% glycerol (v/v), aliquoted, and stored at -80°C until use. Protein content was determined by the method of Bradford as described by Pollard et al. (1978). P450 content was estimated from the CO difference spectrum as described by Omura and Sato (1964). Frozen microsomes were thawed only once just before use.
Cell Isolation Procedure for Human Hepatocyte Primary Culture. Tissue samples were rapidly transported from the operating theater in ice-cold University of Wisconsin solution at 4°C. Hepatocytes were obtained according to the two-step collagenase perfusion technique described by Fabre et al. (1988b). This perfusion technique allows several billions of cells to be obtained, up to 4 × 109 hepatocytes depending on the size of the liver sample. Following different washing steps (filtration through 150- and 250-μm nylon mesh and low-speed centrifugation at 50g for 5 min, 3-fold), freshly isolated hepatocytes were plated on collagen-coated plastic dishes in a chemically defined medium adapted from Isom and Georgoff (1984), consisting of a 50/50 (v/v) mixture of Ham's F-12/Williams' E media supplemented with 10% decomplemented fetal calf serum, 10 mg/l insulin, 0.8 mg/l glucagon, and antibiotics (100 IU penicillin and 100 μg/ml streptomycin). Viability for all the human hepatocyte preparations used was greater than 85% as measured by trypan blue exclusion test. Hepatocytes were seeded in six-well plates with 1.4 × 106 hepatocytes in a final volume of 1 ml. After 4 to 6 h of incubation at 37°C in a 5% CO2 and 100% humidified atmosphere, a period during which hepatocytes attached to the collagen matrix, the plating medium was removed and replaced by the same serum-free culture medium supplemented with HEPES (3.6 g/l), ethanolamine (4 mg/l), transferrin (10 mg/l), linoleic acid-albumin (1.4 mg/l), glucose (252 mg/l), sodium pyruvate (44 mg/l), ascorbic acid (50 mg/l), arginine (104 mg/l), and l-glutamine (0.7 g/l).
MDZ N-Glucuronidation Assay in Human Liver Microsomes. The assay mixture was composed of human liver microsomes (1 mg/ml), UDPGA (3 mM), 100 mM potassium phosphate, pH 7.4, containing 5 mM MgCl2, alamethicin (25 μg/ml), and 100 μM MDZ dissolved in DMSO in a final volume of 500 μl. The final concentration of DMSO in the reaction mixture was 0.1% (v/v). Initial rate conditions with respect to time and protein concentration for the formation of the MDZ-N-glucuronides were established in preliminary studies. The reaction was initiated by addition of 50 μl of an aqueous solution of 30 mM UDPGA, incubated at 37°C, and stopped after 60 min by addition of 1 volume of ice-cold acetonitrile. After removal of the protein by centrifugation at 10,000 rpm for 5 min, a portion of the supernatant was analyzed by HPLC.
MDZ N-Glucuronidation by Human Recombinant UGT Enzymes. The assay mixture was composed of the microsomes (1 mg/ml) of baculovirus-infected cells expressing human UGTs, 100 mM potassium phosphate, pH 7.4, containing 5 mM MgCl2, alamethicin (25 μg/ml), UDPGA (3 mM), and 100 μM MDZ in a final volume of 500 μl. The reaction was initiated by addition of 50 μl of an aqueous solution of UDPGA, incubated at 37°C, and stopped after 60 min by addition of 1 volume of ice-cold acetonitrile. After removal of the protein by centrifugation at 10,000 rpm for 5 min, a portion of the supernatant was analyzed by HPLC.
Metabolic Studies Using Fresh Human Hepatocytes in Primary Culture. Experiments were performed in six-well plastic plates coated with rat-tail collagen type I. Once medium was renewed, 5 μM MDZ was directly added to the incubation medium in the absence or the presence of 10 μM ketoconazole. Regardless of the final concentration investigated, the final solvent (DMSO) concentration never exceeded 0.2% (v/v). To determine the metabolism of MDZ, kinetic studies were performed over 6 to 8 h based on primary determinations on the rate of MDZ biotransformation. For each time point, 1 ml of acetonitrile was added to the specific well for protein precipitation, and both extracellular medium and cell compartment were scraped together. Cell extracts were transferred to a glass tube and stored at -20°C until analysis. Before analysis, cell homogenates were sonicated for a few seconds, homogenized, and centrifuged at 6000g for 30 min. Supernatant fluids were then analyzed for unchanged MDZ and its specific metabolites.
HPLC Instruments and Analytical Conditions. HPLC analysis was performed using a Waters 2795 Alliance HPLC system (Waters) equipped with a Waters UV detector 2487 and a Waters Symmetry C18 column (250 mm length × 4.6 mm i.d., 5-μm particle size). The column temperature was 35°C, and the eluate was monitored by UV absorbance at 250 nm. Solvent A was 0.1% (v/v) formic acid with 2% ammonium acetate (w/v) in distilled water. Solvent B was a mixture of acetonitrile/methanol (70:30, v/v), 0.2% formic acid (v/v), 1% water (v/v), and 0.15% ammonium acetate (w/v). A linear gradient of 15 to 55% mobile phase B over 23 min was used for analysis of MDZ and its metabolites at a flow rate of 0.9 ml/min. The retention times of MDZ, 1′-OH-MDZ, Glu-O-MDZ, and MDZ-N-glucuronide conjugates were 20.6, 22.6, 18, 16.9, and 17.8 min, respectively. Because of the absence of authentic standards for MDZ-N-glucuronides, these conjugates were semiquantified using an MDZ calibration curve after verification that their UV absorbance was equivalent to that of MDZ. To do so, the HPLC peaks containing MDZ-N-glucuronides obtained from the quantitative conversion of MDZ by recombinant UGT1A4 incubates were collected. After hydrolysis of MDZ-N-glucuronides to MDZ, the peak area ratio of MDZ to the converted MDZ-N-glucuronides was found to be 1, thus allowing the quantification of MDZ-N-glucuronides using MDZ standard curves.
Data Analysis. The kinetic parameters Km and Vmax were calculated using GraFit version 5 software (Erithacus Software, East Grinstead, West Sussex, UK). The intrinsic in vitro metabolic clearances (Clint) were calculated using WinNonLin PK analysis software (Pharsight, Mountain View, CA). All the disappearance kinetics data were fitted with a model using the equation Clint = dose/AUC(0-clast).
Results
Identification of MDZ N-Glucuronides Formed by Human Liver Microsomes by HPLC/ESI/MSn Analysis. Enzymatic formation of MDZ N-glucuronides by human liver microsomes was first characterized by HPLC/UV and HPLC/ESI/MSn. Figure 2 shows representative HPLC/UV chromatograms of MDZ and enzymatically formed MDZ-N-glucuronides (M1 and M2) by human liver microsomes in the presence of UDPGA. Figure 3, A through C, shows the mass spectra of MDZ and both N-glucuronides with a pseudomolecular ion [MH]+ at m/z = 326 and molecular ions [M]+ at m/z = 502 (addition of 176 Da), respectively. The MS2 mass spectra of both glucuronides (Fig. 3, D and E) show only one fragment ion at 326 m/z (loss of 176 Da) corresponding to the aglycone ion. The MS3 mass spectra of both glucuronides (Fig. 3, F and G) are similar and showed the same major fragment ions at m/z = 244, 291, and 325, characteristic of MS2 mass spectra of MDZ (Fig. 3H).
From these observations, it was confirmed that the peaks labeled M1 and M2 formed by incubation of MDZ with human liver microsomes in the presence of UDPGA were consistent with MDZ direct glucuronide derivatives. Furthermore, in the absence of free hydroxyl group in the structure of MDZ, these MDZ glucuronides were most likely MDZ-N-glucuronide derivatives.
NMR Studies. Approximately 1 h after purification of M1 and M2 (off-line NMR), these metabolites were analyzed using LC/MS/MS to check the purity of each sample. Interestingly, the sample corresponding to metabolite M1 (>90% purity on collection) already contained a mixture of both M1 and M2 in equal proportions (50:50). Similarly, the sample corresponding to metabolite M2 (also >90% purity on collection) was also composed of a M1/M2 mixture, also in equal proportions (50:50). These results seemed to indicate that the two metabolites M1 and M2 were in equilibrium and could not be kept separated in solution for a long period. On the other hand, 1H1D NMR spectra recorded for each purified metabolite M1 and M2 (off-line NMR) were very similar and corresponded to the superimposition of M1 and M2 NMR spectra in equal proportions (50:50). Moreover, the observation of 1H 1D NMR spectrum of the M1/M2 mixture (on-line NMR and off-line NMR) showed that these two compounds shared the same basic structure. This result confirmed that metabolites M1 and M2 were in equilibrium, and because their structure was likely to be the same, it was postulated that M1 and M2 were two conformers of the same N-glucuronide derivative of MDZ.
The 1H 1D NMR spectra of M1/M2 metabolites (Fig. 4, on-line NMR) showed significant signals of glucuronide moiety (Ismail et al., 2002) at δ = 5.46 ppm and δ = 5.50 ppm (anomeric protons 17 of the two conformers), at δ = 3.61 (18), δ = 3.52/3.54 (19), δ = 3.61 (20), and δ = 4.1 (21). The β-anomeric configuration was assigned to M1 and M2 glucuronide conjugates because the coupling constants of their anomeric protons were approximately 8.5 Hz. Coupling constants in the range of 7 to 10 Hz have been reported to be characteristic of the β-anomers of various glucuronides, whereas the α-anomers have coupling constants in the range of 2 to 4 Hz (Kemp et al., 2002).
The comparison of 1H 1D NMR spectra of M1/M2 and MDZ (Table 4, off-line NMR) showed that resonances corresponding to each proton of MDZ were also present in the structures of M1 and M2. This information suggested that the glucuronide group was linked to one of the N atoms (2, 5, and 11) present in the chemical structures of M1 and M2.
Assignments of the resonances of MDZ and M1/M2 metabolites are shown in Table 6. The positioning of the glucuronide ring on N atom (5) was not in agreement with the chemical shifts observed for proton (4) in metabolites M1 and M2. Indeed, a positive charge located on this N atom (5) should have induced a movement of proton (4) chemical shifts toward lower field. This phenomenon was not observed as seen in Table 6, where proton (4) chemical shifts of metabolites M1 and M2 and MDZ are all similar.
The major differences observed between MDZ and metabolites M1/M2 NMR chemical shifts were focused on the imidazole proton (3) and methyl group (12). We noticed that proton (3) and methyl group (12) chemical shifts were located at lower field for metabolites M1/M2. Furthermore, proton (3) and methyl group (12) both showed duplicate signals, one for each supposed conformer M1 and M2 (Fig. 5). Finally, we assumed that the hybridization type for atoms N (11) and N (2) located in the imidazole ring was sp2 as it is usually supposed. Thus, a conjugation of the glucuronide ring with the N atom (11) would be unlikely. These observations taken into account, we assumed that the glucuronide ring was located on N atom (2) for metabolites M1 and M2. Moreover, because of the interconversion described above between M1 and M2, it was also proposed that these two metabolites M1 and M2 should be considered as two conformers in equilibrium.
Variable-temperature NMR experiments were performed to probe the torsional barrier. Unfortunately, metabolites M1/M2 underwent thermal decomposition during this study, and under these conditions we could not observe NMR signal collapses for protons (3), (17), and (12), as it was expected. Subsequently, nuclear Overhauser experiments (1H 2D nuclear Overhauser effect spectroscopy) were also performed to help discriminate the two conformers M1 and M2. Unfortunately, poor sensitivity as a result of the low amount of metabolites did not allow observing any signals.
To understand better the likely equilibrium that affected NMR and LC/MS/MS spectra during this study (i.e., the presence of some duplicate NMR signals and the impossibility to keep M1 and M2 separate after LC purification), molecular modeling calculations were carried out. During this work, we assumed that the hybridization types for atoms N (2) and N (11) in the imidazole ring were sp2. NMR data recorded during this study suggested that metabolites M1 and M2 should be considered as conformers in equilibrium (50:50). The glucuronide ring is linked to the imidazole ring with covalent nitrogen-carbon bond, a type of chemical bond that is known to undergo atropoisomerism. Atropoisomerism (Schantl, 1995; Campos et al., 1999; Voitenko et al., 2002) is a type of stereoisomerism that may arise in systems where free rotation around a single covalent bond is sufficiently impeded to allow different stereoisomers to be isolated. For metabolites M1 and M2, we assumed that the glucuronide ring located on atom N (2) of the imidazole ring underwent a restricted rotational motion around the nitrogen-carbon bond. Molecular modeling experiments confirmed the presence of a torsional barrier around the nitrogen-carbon bond that connects the imidazole ring to the glucuronide ring. Profiles for changes in energy rotation displayed maxima when the glucuronide ring was passing through the region of the methyl group (12) located on the imidazole ring. Minima energy profiles were found for two energy conformers when the glucuronide ring approximately bisected the imidazole ring plane, being perpendicular to this plane. Furthermore, the calculated relative populations of these two low-energy conformers were in reasonably good agreement with the corresponding abundances observed for each conformer M1 and M2 (atom 3, about 50:50) (Fig. 5). Finally, we would stress that the energy of the torsional barrier was estimated to be 20 kcal/mol, which was in agreement with atropoisomerism phenomena previously described in the literature (Srivastav et al., 1993).
In conclusion, the whole set of NMR data and molecular modeling calculation were in accordance with the proposed structure shown below for metabolites M1 and M2 (Fig. 6). Altogether, these results allowed us to propose a new metabolic scheme for MDZ as shown in Fig. 7.
MDZ N-Glucuronidation in Human Liver Microsomes. Formation rates of MDZ N-glucuronide conjugates were determined in liver microsomes of 10 different human donors at an initial substrate concentration of 100 μM. A moderate variability between individual donors was observed as shown in Fig. 8. All the liver samples were able to glucuronoconjugate MDZ, but preparation HTL19 exhibited the highest glucuronidation capacity, 4.5-fold faster than HL23, the slowest one. Because additional physicochemical and 1H NMR experiments showed a very rapid equilibrium between both conformers, N-glucuronide quantification was performed by summing the two conformers M1 and M2 together.
MDZ N-Glucuronidation by Recombinant Human UGT Enzymes. All the human recombinant UGT isoforms expressed in baculovirus-infected insect cells (UGT1A1, UGT1A3, UGT1A4, UGT1A6, UGT1A7, UGT1A8, UGT1A9, UGT1A10, UGT2B4, UGT2B7, UGT2B15, and UGT2B17) that were commercially available were compared with regard to their ability to catalyze MDZ N-glucuronidation. Among them, only recombinant UGT1A4 showed significant MDZ N-glucuronosyltransferase activity (Fig. 9).
Kinetic Parameters for MDZ N-Glucuronidation by Human Liver Microsomes and Recombinant UGT1A4. Kinetic analyses of MDZ N-glucuronidation were performed using human liver microsomes (from donor HTL19) and recombinant UGT1A4. MDZ N-glucuronidation displayed Michaelis-Menten kinetics. Figure 10 shows the concentration-dependent formation rate of MDZ N-glucuronide in human liver microsomes and recombinant UGT1A4, respectively. Observed apparent Km and Vmax values for microsomal MDZ N-glucuronidation were 37.8 ± 3.6 μM and 276.0 ± 10.3 pmol/min/mg protein, respectively. An apparent Km value similar to that obtained with human liver microsomes was observed for N-glucuronidation by recombinant UGT1A4 (29.9 ± 2.4 μM) with a Vmax value at 659.6 ± 19.0 pmol/min/mg protein (Table 7).
MDZ Metabolic Studies Using Freshly Isolated Human Hepatocytes in Primary Culture. The metabolism of MDZ was investigated in fresh human hepatocytes in primary culture. Based on preliminary results on the rate of biotransformation of MDZ (5 μM initial concentration) in the absence and the presence of ketoconazole (10 μM), kinetic studies were performed over 6 to 8 h.
Figure 11 illustrates the variability and the extremes in ketoconazole effect on MDZ disappearance kinetics on two different human hepatocyte preparations. Whereas MDZ clearance is significantly decreased in the presence of ketoconazole in preparation HTL190, there is almost no inhibitory effect observed in preparation HTL234.
The inhibitory effect of ketoconazole on MDZ clearance was evaluated on a set of 29 human hepatocyte preparations, which showed that this inhibitory effect of ketoconazole was highly variable from one subject/preparation to another. The mean inhibition value calculated on MDZ clearance decrease was 42 ± 25% with a range from 0 to 75% (Fig. 12).
In addition to the effect of ketoconazole on MDZ clearance, the inhibitory effect of ketoconazole on the formation of MDZ CYP3A4/5-mediated metabolites (i.e., 1′-OH-MDZ and its conjugate Glu-1′-O-MDZ) was investigated further. Data are illustrated in Fig. 13 for two different human hepatocyte preparations. Results showed that ketoconazole completely abolished the MDZ 1′-hydroxylation metabolic pathway.
In Fig. 14 the effect of ketoconazole on the formation of CYP3A4/5-mediated metabolites of MDZ on a set of 18 human hepatocyte preparations is illustrated. The mean inhibition value calculated on the 1′-hydroxylation pathway metabolite formation rate was 98 ± 5% with a range from 83 to 100%. Given that ketoconazole often did not fully inhibit MDZ clearance and also that MDZ showed the potential to be glucuronoconjugated directly with UDPGA, the formation of the direct N-glucuronide conjugate was further investigated in two human hepatocyte preparations (Fig. 15).
In the absence of ketoconazole, the N-glucuronide formation represented a very minor metabolic pathway and confirmed the predominant contribution of CYP3A4/5 to the overall in vitro clearance of MDZ with the formation of 1′-OH-MDZ further conjugated with glucuronic acid.
However, in the presence of ketoconazole, the complete inhibition of CYP3A4/5 was observed, but the N-glucuronide conjugate formation increased and accounted for almost all the MDZ clearance in these conditions. The increase in N-glucuronidation under ketoconazole coincubation conditions was shown in all the human hepatocyte preparations investigated.
Inhibition of MDZ N-Glucuronidation by Ketoconazole. Like oxidative metabolism, drug glucuronidation may be inhibited by the coadministration of other chemicals (Miners and Mackenzie, 1991). Because ketoconazole has been described to inhibit glucuronidation of several drugs (Satoh et al., 2004; Yong et al., 2005; Takeda et al., 2006), the rate of direct N-glucuronidation of MDZ formation using human liver microsomes and human recombinant UGT1A4 in the presence of the inhibitor has been investigated (Fig. 16).
For these experiments, using either human liver microsomes or rhUGT1A4, MDZ concentration was set to 30 μM (i.e., close to the observed Km), and under these conditions we observed no significant inhibition of direct MDZ N-glucuronidation rate in the presence of ketoconazole.
Discussion
When incubating MDZ, a well known in vitro and in vivo CYP3A4/5 probe substrate, with human liver microsomes in the presence of UDPGA, we observed the formation of two N-glucuronide conjugates. Looking at MDZ structure, three distinct direct N-glucuronidation positions were plausible. It was confirmed by LC/MS/MS analysis that the peaks observed in HPLC were direct MDZ N-glucuronides because the MS2 and MS3 spectra of the metabolite peaks were consistent with the MS2 spectrum of MDZ. Additional 1H NMR characterization showed that both conjugates were β-N-glucuronides of MDZ. Those studies also showed that because the 1H NMR spectra of both glucuronides were very close, both conjugates were probably conformers of the same β-N-glucuronide. Molecular modeling studies suggested that these conjugates could be involved in atropoisomerism equilibrium and determined the glucuronidation position on the methyl-imidazole moiety.
Moderate interindividual variability in MDZ N-glucuronidation was observed among the 10 human liver microsome preparations used (at most 4.5-fold, n = 10). Further characterization studies using commercially available baculovirus-infected insect cells showed that recombinant UGT1A4 was the single human UGT isoform able to glucuronidate MDZ among all the recombinant UGT isoforms available. Km value determined with rhUGT1A4 was similar to that observed with human liver microsomes, 29.9 ± 2.4 versus 37.8 ± 3.6 μM, respectively. These results suggest that the major liver UGT isoform catalyzing MDZ direct glucuronidation is UGT1A4, in accordance with the results recently published by Zhu et al. (2008). Because of the comparable apparent Km of UGT1A4 relative to the Km value obtained in human liver microsomes, the contribution of UGT1A4 to MDZ clearance is subject to its relative abundance in tissues such as liver and gut.
Complementary studies were performed on fresh human hepatocytes in primary culture. The presence of the MDZ N-glucuronide was also evidenced in samples in control conditions as well as, and to a much greater extent, in the presence of the specific CYP3A4 inhibitor ketoconazole. These observations suggest a metabolic shift phenomenon between the CYP3A4 1′-hydroxylation pathway and direct N-glucuronidation of MDZ evidenced in the presence of specific and complete CYP3A4 inhibition, such as that exerted by ketoconazole.
Similar observations were also made when looking for MDZ N-glucuronidation in plasma samples of patients treated with MDZ and a potent proprietary CYP3A4/5 inhibitor. Like in human hepatocyte preparations, a large intersubject variability was observed in glucuronidation efficiency in vivo (data not shown, further investigations are ongoing).
In the present study, an additional metabolic pathway to the well known CYP3A4/5-mediated oxidative process of MDZ has been revealed. Indeed, a potential metabolic shift from MDZ 1′-hydroxylation to direct N-glucuronidation has been evidenced in vitro principally in the presence of ketoconazole. Those findings provide evidence that direct N-glucuronidation could play an important role in MDZ clearance in vitro and in vivo. The apparent Km and Vmax values obtained could result in quite high glucuronidation efficiency in vivo.
Thus, the results obtained in this study should be taken into account when trying to predict the magnitude of a DDI that could occur in patients under ketoconazole treatment. Indeed, the occurrence of MDZ direct N-glucuronidation could lead to a certain extent of underestimation of the ketoconazole inhibitory effect at least in vitro and maybe also in vivo because in the clinic, the magnitude of a DDI is mainly evaluated based on the MDZ exposure, the -fold increase in the presence of a CYP3A4 inhibitor, and not on the decrease of the specific CYP3A4-mediated MDZ-1′-hydroxylation route. Thus, particularly for patients exhibiting a large glucuronidation capacity via the MDZ direct N-glucuronidation pathway (most likely linked to a high UGT1A4 capacity), the overall increase in MDZ plasma exposure observed is a contribution of partial or total suppression of the MDZ 1′-hydroxylation pathway and the partly compensating enhanced direct N-glucuronidation pathway.
Such findings could account, at least in part, for the large variability observed in the ketoconazole effect from one patient to another because they suggest that MDZ clearance would not be exclusively associated with CYP3A4/5. Indeed, intersubject variability could not only be related to CYP3A4 variability but also to variability in N-glucuronidation (UGT1A4) capacity. For example, N-glucuronidation could be an alternative metabolic pathway when poor CYP3A4 activity is shown. This means that intersubject variability in glucuronidation (Ehmer et al., 2004) combined with interindividual variability in CYP3A4 metabolism should be considered at this stage when using MDZ as an in vitro and in vivo probe for DDI studies.
Acknowledgments
We thank Dr. J.-L. Reversat (Sanofi-Aventis, Department of Chemical and Analytical Sciences, Montpellier, France) for expert molecular modeling work, Dr. E. Sultan (Sanofi-Aventis, Department of Global Metabolism Pharmacokinetics, Montpellier, France) for supplying human plasma samples, and Dr. Chandra Natarajan (Sanofi-Aventis, Department of Discovery Metabolism, Pharmacokinetics, and Safety, Bridgewater, NJ) for a careful reading of the document and its scientific expertise.
Footnotes
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.107.019539.
-
ABBREVIATIONS: DDI, drug-drug interaction; MDZ, midazolam, 8-chloro-6-(2-fluorophenyl)-1-methyl-4H-imidazo [1,5a] [1,4] benzodiazepine; P450, cytochrome P450; 1′-OH-MDZ, 1′-hydroxymidazolam, 8-chloro-6-(2-fluorophenyl)-1-hydroxymethyl-4H-imidazo [1,5a] [1,4] benzodiazepine; UDPGA, UDP-glucuronic acid; UGT, UDP glucuronosyltransferase; Glu-O-MDZ, 1′-hydroxymidazolam glucuronide; DMSO, dimethyl sulfoxide; HPLC, high-performance liquid chromatography; SPE, solid-phase extraction; 1D, one-dimensional; 2D, two-dimensional; COSY, correlation spectroscopy; COSY-GPQF, correlation spectroscopy using gradient pulse for selection; MS, mass spectrometry; LC, liquid chromatography; ESI, electrospray ionization; DMSO-D6, deuterated dimethyl sulfoxide.
- Received October 30, 2007.
- Accepted January 31, 2008.
- The American Society for Pharmacology and Experimental Therapeutics