Abstract
Midazolam is a potent benzodiazepine derivative with sedative, hypnotic, anticonvulsant, muscle-relaxant, and anxiolytic activities. It undergoes oxidative metabolism catalyzed almost exclusively by the CYP3A subfamily to a major metabolite, 1′-hydroxymidazolam, which is equipotent to midazolam. 1′-Hydroxymidazolam is subject to glucuronidation followed by renal excretion. To date, the glucuronidation of 1′-hydroxymidazolam has not been evaluated in detail. In the current study, we identified an unreported quaternary N-glucuronide, as well as the known O-glucuronide, from incubations of 1′-hydroxymidazolam in human liver microsomes enriched with uridine 5′-diphosphoglucuronic acid (UDPGA). The structure of the N-glucuronide was confirmed by nuclear magnetic resonance analysis, which showed that glucuronidation had occurred at N-2 (the imidazole nitrogen that is not a part of the benzodiazepine ring). In a separate study, in which midazolam was used as the substrate, an analogous N-glucuronide also was detected from incubations with human liver microsomes in the presence of UDPGA. Investigation of the kinetics of 1′-hydroxymidazolam glucuronidation in human liver microsomes indicated autoactivation kinetics (Hill coefficient, n = 1.2–1.5). The apparent S50 values for the formation of O- and N-glucuronides were 43 and 18 μM, respectively, and the corresponding apparent Vmax values were 363 and 21 pmol/mg of microsomal protein/min. Incubations with recombinant human uridine diphosphate glucuronosyltransferases (UGTs) indicated that the O-glucuronidation was catalyzed by UGT2B4 and UGT2B7, whereas the N-glucuronidation was catalyzed by UGT1A4. Consistent with these observations, hecogenin, a selective inhibitor of UGT1A4, selectively inhibited the N-glucuronidation, whereas diclofenac, a potent inhibitor of UGT2B7, had a greater inhibitory effect on the O-glucuronidation than on the N-glucuronidation. In summary, our study provides the first demonstration of N-glucuronidation of 1′-hydroxymidazolam in human liver microsomes.
Midazolam is a potent benzodiazepine derivative with sedative, hypnotic, anticonvulsant, muscle-relaxant, and anxiolytic activities. It is widely used in the clinic for induction of anesthesia and for sedation of patients who are artificially ventilated in intensive care units (Dundee et al., 1984). Midazolam is rapidly eliminated from the body, almost exclusively by metabolism (Heizmann et al., 1983). It undergoes oxidative metabolism catalyzed mainly by the CYP3A subfamily to a major metabolite, 1′-hydroxymidazolam, which is equipotent to midazolam, and two minor metabolites, 4-hydroxymidazolam and 1′,4-dihydroxymidazolam, which are quantitatively unimportant (Dundee et al., 1984). 1′-Hydroxymidazolam is subject to further glucuronidation, followed by renal excretion. In humans, urinary recovery of 1′-hydroxymidazolam glucuronide accounted for 60 to 70% of an administered dose of [14C]midazolam (Heizmann and Ziegler, 1981). It has been reported that elevated serum levels of 1′-hydroxymidazolam glucuronide were found in patients with renal failure after administration of midazolam and may account for the prolonged sedation observed in those patients (Bauer et al., 1995; Hirata et al., 2003).
Glucuronidation represents one of the major phase II conjugation reactions in the conversion of both exogenous and endogenous compounds to polar and water-soluble metabolites that can be eliminated from the body in urine or bile (Sipes and Gandolfi, 1991). The reaction is catalyzed by a family of enzymes, UDP-glucuronosyltransferases (UGTs), which transfer glucuronic acid from UDP-glucuronic acid to the aglycone substrate. UGTs comprise two subfamilies, UGT1 and UGT2. The substrate specificity of individual UGTs has been partially characterized. UGT subfamily 1 is responsible for glucuronidation of bilirubin, amines, and planar and bulky phenols, whereas subfamily 2 enzymes catalyze glucuronidation of a diverse chemical base including steroids, bile acids, and opioids (King et al., 2000).
To date the glucuronidation of 1′-hydroxymidazolam has not been studied in detail. In the literature, 1′-hydroxymidazolam glucuronide is often assumed to be an O-glucuronide. In the current study, we have identified an unreported quaternary N-glucuronide, as well as the known O-glucuronide, from incubations of 1′-hydroxymidazolam in human liver microsomes enriched with uridine 5′-diphosphoglucuronic acid (UDPGA). Further studies were conducted to determine the isozymes responsible for the formation of N- and O-glucuronic acid conjugates of 1′-hydroxymidazolam in human liver microsomes.
Materials and Methods
Materials. 1′-Hydroxymidazolam was synthesized by the Labeled Compound Synthesis Group, Department of Drug Metabolism, Merck Research Laboratories (Rahway, NJ). Recombinant human UGTs were purchased from BD Gentest (Woburn, MA). UDPGA, alamethicin, hecogenin, and diclofenac were purchased from Sigma-Aldrich (St. Louis, MO). All solvents were of HPLC grade and were obtained from Fisher Scientific Co. (Pittsburgh, PA).
Glucuronidation of 1′-Hydroxymidazolam. Pooled human liver microsomes (batch no. 452161) were purchased from BD Gentest. A pool of 50 male Sprague-Dawley rat (∼225–250 g, ∼8 weeks old) liver microsomes was prepared in-house following procedures described in the literature (Raucy and Lasker, 1991). Microsomal protein concentrations of 0.25 to 1 mg/ml and incubation times of 10 to 60 min were used to optimize the conditions of the assay. The reaction mixture consisted of 1′-hydroxymidazolam (5–200 μM), 100 mM potassium phosphate buffer (pH 7.5) with 2 mM MgCl2, and human/rat liver microsomes (0.5 mg of protein/ml) or recombinant UGTs (1 mg of protein/ml) treated with alamethicin at 50 μg/mg of microsomal protein. The reactions were initiated by the addition of UDPGA (2 mM), incubated at 37°C, and terminated with ice-cold acetonitrile containing 0.2% formic acid and phenolphthalein-β-d-glucuronide as the internal standard. Samples were centrifuged at 3220g for 15 min, and supernatants were subjected to LC-MS/MS analysis.
Enzyme Kinetics Analysis. The apparent kinetic parameters Km/S50, Vmax, and n (Hill coefficient; where appropriate) were calculated using nonlinear regression analysis (Sigma Plot; Systat Software, Inc., San Jose, CA). Each set of data were fitted to both the Michaelis-Menten and the Hill equations. The quality of fit to a particular model was determined by evaluation of three criteria that are listed in decreasing order of importance: 1) the randomness of the residuals; 2) the size of the sum of the square of the residuals; and 3) the standard error of the parameter estimates (Soars et al., 2003). The maximal intrinsic clearance due to autoactivation kinetics was calculated on the basis of the following equation (Houston and Kenworthy, 2000):
Inhibition with Chemical Inhibitors of UGTs. The incubations of 1′-hydroxymidazolam at a substrate concentration close to the apparent Km/S50 values (25 or 50 μM) were performed in human liver microsomes, as described above, in the presence of diclofenac (9.3–500 μM) or hecogenin (2.5–400 μM). The reaction was allowed to proceed at 37°C for 45 min and was terminated with ice-cold acetonitrile containing 0.2% formic acid and phenolphthalein-β-d-glucuronide as the internal standard. The assay was performed in duplicate. Samples were centrifuged before LC-MS/MS analysis. Similar inhibition studies were conducted using recombinant UGT1A4 or UGT2B4/2B7 in the presence of hecogenin or diclofenac (0.4–100 μM), respectively. The IC50 values were calculated using nonlinear regression analysis (KaleidaGraph; Synergy Software, Reading, PA).
Isolation of 1′-Hydroxymidazolam Glucuronides. The glucuronides of 1′-hydroxymidazolam (Glu-A and Glu-B) were isolated from human liver microsomal incubations. Briefly, the reaction mixture was precipitated with an equal volume of acetonitrile, followed by centrifugation. The supernatant was concentrated under N2 and subjected to several centrifugations in preparation for injection onto an HPLC column for further purification. Two columns were used sequentially for isolating the glucuronides, Synergi Polar-RP column (4.6 × 250 mm, 4 μm; Phenomenex, Torrance, CA), and Zorbax RX C8 column (4.6 × 250 mm, 5 μm; Agilent Technologies, Wilmington, DE).
NMR Analysis. NMR spectra were acquired using a Varian Inova 600 MHz spectrometer with CD3CN as solvent. Signal assignments were obtained using 1H-1H correlation spectroscopy, nuclear Overhauser enhancement spectroscopy, and 1H-13C heteronuclear single quantum correlation two-dimensional NMR experiments. The 13C chemical shifts were obtained from the heteronuclear single quantum correlation spectra. Chemical shifts are expressed in parts per million downfield from tetramethylsilane.
LC-MS/MS Analysis. LC-MS/MS was carried out on a API-3000 (PerkinElmerSciex, Concord, ON, Canada) triple quadrupole mass spectrometer, interfaced to a HPLC system (PerkinElmer Life and Analytical Sciences, Norwalk, CT) equipped with two Series 200 pumps and a Series 200 autosampler. The instrument was operated in the positive ion mode using a TurboIon-Spray interface. The chromatography was performed using a Polar-RP column (4.6 × 250 mm, 4 μm particle size) purchased from Phenomenex. The mobile phase consisted of 5 mM ammonium acetate in water (A) and acetonitrile-methanol (70:30; v/v) (B). The analytes were eluted at 1 ml/min using a linear gradient. Solvent B started at 20%, then increased as follows: 40% (25 min), 50% (26 min), 60% (40 min), and 95% (45 min). The column was washed at 95% B for 3 min, before equilibration with starting conditions. The retention times for the O- and N-glucuronides (Glu-A and Glu-B, respectively) were 10 and 21 min, respectively. For quantitation purposes, another gradient was used. Solvent B started at 20% and increased as follows: 50% (8 min) and 95% (17 min). The relative amounts of the glucuronides of 1′-hydroxymidazolam were determined using multiple reaction monitoring of the transitions m/z 518 → m/z 324 for the O-glucuronide of 1′-hydroxymidazolam, m/z 518 → m/z 342 for the N-glucuronide, and m/z 495 → m/z 319 for the internal standard. The amounts of the isolated O- and N- glucuronides were determined by quantitative NMR analysis (Pauli et al., 2005). The quantitation was accomplished by comparing the absolute integral values of a well resolved proton signal in each glucuronide sample to the integral of the same signal in an external standard (1′-hydroxymidazolam). The precision and accuracy of the method were evaluated using known amounts of 1′-hydroxymidazolam, with coefficient of variation of <5% and accuracy of 97.4%. The analysis of the standards and the glucuronides was done back-to-back under identical conditions (i.e., using the same NMR spectrometer, receiver gain, probe tuning, and acquisition and processing parameters). The isolated metabolites were used to construct standard curves for later LC-MS/MS quantitation.
Results
Glucuronidation of 1′-Hydroxymidazolam and Midazolam in Human and Rat Liver Microsomes. When 1′-hydroxymidazolam was incubated with human liver microsomes in the presence of UDPGA, two putative glucuronides (Glu-A and Glu-B) were detected, as confirmed by full scan analysis that showed addition of 176 Da to the parent (m/z 518). Only Glu-A was detected in incubations with rat liver microsomes (Fig. 1a). A subsequent product ion scan at m/z 518 revealed different fragmentation patterns for these two glucuronides. A fragment ion at m/z 324 (loss of 194 Da) was the most prominent fragment observed with Glu-A, whereas a fragment ion at m/z 342 (loss of 176 Da) was the major ion for Glu-B (Fig. 1a). When midazolam was incubated with liver microsomes enriched with UDPGA, one glucuronide (m/z 502) was detected in human liver microsomes but not in rat liver microsomes (Fig. 1b).
Structure Identification of 1′-Hydroxymidazolam Glucuronides (Glu-A and Glu-B) by NMR. Glu-A and Glu-B were isolated, and their structures were determined by comparing their NMR spectra to that of 1′-hydroxymidazolam (Fig. 2a). The major glucuronide (Glu-A) was confirmed to be the O-glucuronide (Fig. 2c). The minor glucuronide (Glu-B), a hitherto unknown glucuronide, was shown to be a quaternary N-glucuronide (Fig. 2b). The distinction between O- versus N-glucuronide structures was evident from both 1H and 13C chemical shifts of the anomeric position (4.46 and 101.0 ppm in Glu-A versus 5.53 and 87.8 ppm in Glu-B, respectively; see Supplemental Figs. 1–6). The exact location of the N-glucuronide moiety was determined on the basis of the nuclear Overhauser enhancement observed between the anomeric proton and the imidazole proton (as indicated by the arrow in Fig. 3).
Enzyme Kinetics of 1′-Hydroxymidazolam Glucuronidation in Human Liver Microsomes. The rates of formation of O- and N-glucuronides were proportional to microsomal protein concentration from 0.25 to 0.5 mg of protein/ml and were approximately linear with time of incubation from 5 to 60 min (data not shown). For most of the microsomal incubations, 0.5 mg/ml protein and a 30-min incubation time were used. The effect of substrate concentration (5–200 μM) on the glucuronidation of 1′-hydroxymidazolam is shown in Fig. 4. Both O- and N-glucuronidation seemed to best fit the Hill model, yielding n values of ∼1.2 and 1.5, respectively. The apparent S50 values for the formation of O- and N-glucuronides were ∼43 and 18 μM, respectively, in human liver microsomes. The corresponding apparent Vmax values were 363 and 21 pmol/mg/min (Fig. 4). Thus, the maximal intrinsic clearance for the O-glucuronidation was ∼9-fold higher than that for the N-glucuronidation.
Glucuronidation of 1′-Hydroxymidazolam by Recombinant UGTs. Incubations of 1′-hydroxymidazolam with 12 recombinant human UGTs (UGT1A1, 1A3, 1A4, 1A6, 1A7, 1A8, 1A9, 1A10, 2B4, 2B7, 2B15, and 2B17) revealed that O-glucuronidation was catalyzed by UGT2B4 and UGT2B7, whereas N-glucuronidation was catalyzed by UGT1A4. No metabolism was observed with other UGT enzymes (the limit of assay sensitivity was ∼0.01 μM). The kinetic parameters of the glucuronidation catalyzed by UGT2B4, UGT2B7, and UGT1A4 are presented in Table 1. Incubations of various concentrations of 1′-hydroxymidazolam with UGT2B4 and UGT2B7 demonstrated a similar apparent Km value of ∼35 μM for the formation of the O-glucuronide. The corresponding Vmax values were 37 and 27 pmol/mg of protein/min. The N-glucuronidation by UGT1A4 exhibited non-Michaelis-Menten kinetics, with an apparent S50 value of 40 μM, apparent Vmax value of 19 pmol/mg of protein/min, and an n value of 1.5 (Table 1).
Inhibition of 1′-Hydroxymidazolam Glucuronidation in Human Liver Microsomes. The inhibition study was conducted at a substrate (1′-hydroxymidazolam) concentration of 25 or 50 μMinthe presence of UGT inhibitors, hecogenin or diclofenac. Hecogenin showed potent and selective inhibitory effects on the N-glucuronidation (IC50 value = 1.1 μM) but had little or no inhibitory effect on the O-glucuronidation in human liver microsomes (IC50 value of >500 μM) (Fig. 5a). This inhibitory effect of hecogenin was further confirmed using recombinant UGT1A4 (IC50 value = 0.77 μM) (Fig. 5b). Diclofenac, on the other hand, had a greater inhibitory effect on the formation of O-glucuronide than on the formation of N-glucuronide by human liver microsomes (IC50 values of 16 versus 151 μM) (Fig. 6a). Further inhibition studies with recombinant UGT isozymes indicated that diclofenac exhibited similar inhibitory effects toward UGT2B4- and UGT2B7-mediated O-glucuronidation of 1′-hydroxymidazolam, with IC50 values of 7.4 and 17.2 μM, respectively (Fig. 6b).
Discussion
After the incubations of 1′-hydroxymidazolam with human liver microsomes in the presence of UDPGA, two glucuronides (Glu-A and Glu-B) were detected. Based on the different fragmentation patterns, we assumed that Glu-A, which had a protonated aglycone with further loss of water (m/z 324) as the prominent fragment ion, was an O-glucuronide and that Glu-B, which had a protonated aglycone corresponding to 1′-hydroxymidazolam (m/z 342) as the major fragment ion, was an N-glucuronide, given that more energy would be required to cleave a C–N bond than a C–O bond. Glu-A and Glu-B were isolated from microsomal incubations, and their structures were confirmed by NMR analysis. The NMR data indicated that the exact location of the glucuronic acid moiety in 1′-hydroxymidazolam N-glucuronide was at N-2. To the best of our knowledge, this is the first demonstration of a quaternary N-glucuronide of 1′-hydroxymidazolam in human liver microsomes and the first rigorous structural proof of the O-glucuronide, which is often referred to as the 1′-hydroxymidazolam glucuronide. The maximal intrinsic clearance for O-glucuronidation was ∼9-fold higher than that for N-glucuronidation, suggesting that O-glucuronidation could be the major conjugation pathway in human liver microsomes.
In a separate study, in which midazolam was used as the substrate, an analogous N-glucuronide also was detected in human liver microsomal incubations in the presence of UDPGA (Fig. 1b). It was reported previously that a quaternary N-glucuronide of midazolam was detected from human liver preparations in vitro, but no structure elucidation was presented (Siddle et al., 2003). On the basis of the structure of 1′-hydroxymidazolam N-glucuronide presented here, it is reasonable to assume that glucuronic acid was conjugated to midazolam at the same N-2 position.
The O- and N-glucuronidation of 1′-hydroxymidazolam in human liver microsomes exhibited non-Michaelis-Menten kinetics consistent with autoactivation, which has been reported for in vitro glucuronidation of several substrates (Fisher et al., 2000; Soars et al., 2003) as well as in vivo (Wong et al., 2007). The mechanism of autoactivation kinetics observed for UGTs is currently unknown.
In vitro incubations with 12 different recombinant human UGT isoforms showed that UGT1A4 was the only isoform catalyzing N-glucuronidation of 1′-hydroxymidazolam. UGT1A4 has been identified in human liver and stomach. It is a major UGT isoform involved in the glucuronidation of many tertiary amines (Green and Tephly 1996; King et al., 2000). But there is no UGT1A4 ortholog in rat tissues (King et al., 2000). Accordingly, when 1′-hydroxymidazolam was incubated with rat liver microsomes enriched with UDPGA, only the O-glucuronide (Glu-A) but not the N-glucuronide (Glu-B) was detected, consistent with the observation that UGT1A4 is the only UGT isoform catalyzing the N-glucuronidation. Similar studies with midazolam also showed that an analogous N-glucuronide formed in human liver microsomes was not observed in rat liver microsomes enriched with UDPGA and that only UGT1A4 was able to catalyze the formation of midazolam N-glucuronide.
Hecogenin is a known substrate and a selective inhibitor of UGT1A4 (Green and Tephly 1996; Al-Zoughool and Talaska 2006; Uchaipichat et al., 2006). In the current inhibition study, hecogenin essentially abolished the N-glucuronidation of 1′-hydroxymidazolam in human liver microsomes, whereas it had little or no inhibitory effect on the O-glucuronidation. In a follow-up inhibition study using recombinant UGT1A4, hecogenin inhibited UGT1A4-mediated N-glucuronidation at an IC50 value of <1 μM. Taken together, these data confirmed the involvement of UGT1A4 in the formation of 1′-hydroxymidazolam N-glucuronide.
Our study demonstrated that UGT2B4 and UGT2B7 are the two isoforms responsible for the in vitro formation of 1′-hydroxymidazolam O-glucuronide in human liver microsomes. UGT2B7 is a very important human UGT isoform in that it appears to be expressed in many tissues besides liver, and it catalyzes the glucuronidation of a wide range of xenobiotics, including polycyclic aromatic hydrocarbons, phenols, opioids, aliphatic alcohols, carboxylic acids, and tetrazoles (King et al., 2000). Recently, UGT2B7 was reported to catalyze the N-glucuronidation of an amide and a primary amine (Staines et al., 2004; Zhang et al., 2004). In contrast, UGT2B4 catalyzes only a limited number of substrates, and the data so far are not consistent among different laboratories. For example, the glucuronidation of hyodeoxycholic acid was reported to be catalyzed by UGT2B4; yet in another study, UGT2B4 showed no activity for hyodeoxycholic acid (King et al., 2000). The best substrate for this isoform has yet to be identified. Diclofenac is a substrate of UGT2B7 and also inhibits the glucuronidation of dihydrocodeine and morphine catalyzed by UGT2B7 (King et al., 2001). In the present study, diclofenac selectively inhibited O-glucuronidation of 1′-hydroxymidazolam in human liver microsomes and also showed similar inhibitory effect toward UGT2B4- and UGT2B7-mediated O-glucuronidation of 1′-hydroxymidazolam.
Available data suggest that UGT enzymes exhibit distinct but overlapping substrate selectivities (Burchell et al., 1995; King et al., 2000). Thus, identifying selective substrates of certain UGT isoforms could be very useful when evaluating drug-drug interaction potential of a given compound at the UGT level. For this purpose, 1′-hydroxymidazolam could be used as an in vitro probe substrate for UGT1A4 and UGT2B4/2B7 by monitoring the formation of N- and O-glucuronide, respectively. In clinical drug-drug interaction studies, midazolam is often used as a probe substrate for CYP3A (Thummel et al., 1994a,b; Huang et al., 2007). The exposure to midazolam, if altered, will provide information on the potential of an investigational drug as a perpetrator of CYP3A. On the other hand, given that the levels of 1′-hydroxymidazolam in plasma are the result of formation by CYP3A and subsequent metabolism to glucuronides by UGTs, by monitoring the plasma concentration of 1′-hydroxymidazolam from the same study, we may be able to get additional insights into the potential of the investigational drug to affect selected UGT isoforms (UGT1A4, UGT2B4, and UGT2B7), assuming the exposure to midazolam is not altered.
In summary, we have demonstrated, for the first time, the formation of a quaternary N-glucuronide of 1′-hydroxymidazolam, in addition to the well known O-glucuronide in human liver microsomes enriched with UDPGA. Also, we identified the corresponding enzymes responsible for the glucuronidations as UGT1A4 and UGT2B4/2B7, respectively.
Acknowledgments
We thank Dr. Matt Braun, Dr. Magang Shou, and Regina Wang for helpful discussions, Brian Cato for synthesizing 1′-hydroxymidazolam, and Dr. Ralph Stearns for critical review of this work.
Footnotes
-
B.Z. and D.B. contributed equally to this work.
-
This work was presented in part in N-Glucoronidation of 1′-Hydroxymidazolam in Human Liver Microsomes (B. Zhu, D. Bush, G. A. Doss, S. Vincent, R. B. Franklin, and S. Xu) at the 3rd Pharmaceutical Sciences World Congress, 2007 April 22–25; Amsterdam, The Netherlands.
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.107.017962.
-
ABBREVIATIONS: UGT, uridine diphosphate glucuronosyltransferase; UDPGA, uridine 5′-diphosphoglucuronic acid; HPLC, high performance liquid chromatography; LC-MS/MS, liquid chromatography-tandem mass spectrometry; NMR, nuclear magnetic resonance.
-
↵s⃞ The online version of this article (available at http://dmd.aspetjournals.org) contains supplemental material.
-
↵1 Current affiliation: Department of Chemistry, The University of Arizona, 1306 E. University Blvd., Tucson, AZ 85721-0041.
-
↵2 Current affiliation: Department of Drug Metabolism, Array Biopharma Inc., 3200 Walnut St., Boulder, CO 80301.
- Received August 2, 2007.
- Accepted November 7, 2007.
- The American Society for Pharmacology and Experimental Therapeutics