Abstract
Idiosyncratic drug toxicity is a major challenge for the pharmaceutical industry since complex and multifactorial steps are involved, the dose-dependency is unclear, and its occurrence is not reliably predictable. Whereas the exact mechanisms leading to idiosyncratic toxicity remain elusive in many cases, there are often hints at the involvement of reactive metabolites, such as acyl glucuronides formed by conjugation of carboxylic acids with glucuronic acid. Because the patient-related susceptibilities leading to idiosyncratic toxicity are not sufficiently understood, the best option for the pharmaceutical industry is to minimize drug-related risk factors such as potential acyl glucuronide formation. Here, we describe a rapid in vitro assay for the assessment of the reactivity of acyl glucuronides, on the basis of acyl glucuronide migration, that can support the selection of low-risk drug candidates in the drug discovery phase. Twenty marketed compounds with a wide range of half-lives were tested, their acyl glucuronide migration rates were determined and compared with the half-lives of the respective acyl glucuronides. Ranking of acyl glucuronide stability using this method compared well with the results from existing methodologies. With this method, migration rates >20% would indicate higher risk of reactivity. This simpler approach using the acyl glucuronide migration rate is not dependent on authentic standards, therefore eliminating the requirement for either lengthy chemical synthesis or in vitro biosynthesis and purification of the 1-O-β-glucuronide. This methodology provides a rapid in vitro assay to assess acyl glucuronide stability and reactivity that is well suited for use early in the drug discovery phase.
Introduction
Idiosyncratic drug toxicity (IDT) is a major challenge for the pharmaceutical industry because of its poor predictability, unclear dose-dependency, and complex, multifactorial steps. Whereas the exact mechanisms leading to idiosyncratic toxicity remain elusive in many cases, there are often hints that reactive metabolites and/or the adaptive immune system play a role (Kaplowitz, 2013). One group of potentially reactive metabolites is acyl glucuronides formed by the conjugation of carboxylic acids with glucuronic acid. Reactive acyl glucuronides can bind covalently to proteins, thus potentially disrupting protein function, changing cellular signaling pathways or acting as haptens (Bailey and Dickinson, 2003). Formation of new antigens by haptenization is considered to be the first step in the immune response cascade, which subsequently can lead to an idiosyncratic toxicity-like drug-induced liver injury (DILI), skin reactions (e.g., Steven-Johnson syndrome), or anaphylaxis. Examples of carboxylic acid drugs that form acyl glucuronides and that had to be withdrawn from the market include benoxaprofen, ibufenac, and zomepirac (Boelsterli, 2002; Sawamura et al., 2010). Since the patient-related susceptibilities leading to idiosyncratic toxicity are currently not sufficiently understood, the best option for the pharmaceutical industry is to minimize drug-related risk factors like the potential to form reactive acyl glucuronides.
Acyl glucuronide conjugates are able to undergo migration. This is an enzyme-independent intramolecular rearrangement process in which the drug moiety migrates from the original 1-O position of the glucuronic acid to the 2-O-, 3-O-, and 4-O-positions. During this migration process there is the potential to produce chemically reactive species that can covalently interact with proteins and other biomolecules. This chemical reactivity can therefore be linked to the stability of the acyl glucuronide, such that the greater the migration rate the greater the potential reactivity of the acyl glucuronide. (Bailey and Dickinson, 2003). Generally, two types of reaction can occur, starting with the initial 1-O-β-isomer, that can be investigated to draw conclusions about the reactivity of the acyl glucuronide: first, direct reaction with nucleophiles, such as cysteine residues in human serum albumin (HSA), or with peptides, leading to displacement of the acyl residue; or second, acyl migration to the 2-, 3-, and 4-acyl isomers, resulting in formation of a potentially reactive aldehyde group on the ring-open tautomer of the glucuronic acid moiety, which can subsequently condense with primary amino functional groups of proteins (Fig. 1; Fenselau, 1994, Bailey and Dickinson, 2003).
Acyl glucuronide migration, starting with the initial 1-O-β-isomer: 1) direct reaction with nucleophiles, such as cysteine residues in HSA or peptides, leading to displacement of the acyl residue; 2) acyl migration to the 2-, 3-, and 4-acyl isomers resulting in the formation of a potentially reactive aldehyde group on the ring-open tautomer of the glucuronic acid moiety, which can subsequently condense with primary amino functional groups of proteins.
The stability and reactivity of acyl glucuronides varies widely; for example, the 1-O-β- acyl glucuronide of tolmetin has a very short first-order half-life of 0.26 hours, whereas the valproic acid acyl glucuronide has a long half-life of 79 hours (Fenselau, 1994). The extent of covalent binding to HSA and the apparent first-order degradation rate of acyl glucuronides have been shown to correlate well (Benet et al., 1993). It has been further demonstrated that the chemical stability of the acyl glucuronides correlates with the IDT risk, the shorter the half-life of the acyl glucuronide formed the higher the risk of induced IDT (Sawamura et al., 2010). Therefore, given this relationship between the rate of degradation of 1-O-β-acyl glucuronides and the extent of in vitro covalent binding (and IDT risk), the direct assessment of the stability of the 1-O-β-acyl glucuronide provides a simple indication of the potential reactivity of the acyl glucuronide product.
Multiple assays for the assessment of acyl glucuronide stability and reactivity are described in the literature. The gold-standard methods used to investigate acyl glucuronide stability typically assess the half-life of an acyl glucuronide in buffer (Ebner et al., 1999; Ebner et al., 2010), buffer with human serum albumin (Benet et al., 1993), or plasma (Sawamura et al., 2010). These methods require authentic standard of the 1-O-β-acyl glucuronide. Compound optimization early in drug discovery requires rapid assessment methods that do not depend on authentic standard, since its generation and purification is too time-consuming for application to multiple compounds. Therefore, methods using acyl glucuronides produced by liver microsomes for determining the half-lives in buffer, HSA solution, or fresh plasma have been developed and previously documented (Bolze et al., 2002; Chen et al., 2007; Jinno et al., 2013).
Here, we describe a simpler approach using the acyl glucuronide migration rate in human liver microsomes to assess their reactivity in a rapid screening assay that can support the selection of low-risk drug candidates in the drug discovery phase. Twenty marketed compounds (Supplemental Table 1) with a wide range of half-lives were tested, and their migration rates were determined.
Materials and Methods
Chemicals and Reagents.
Uridine 5′-diphosphoglucuronic acid trisodium salt (UDPGA), β-glucuronidase from bovine, zomepirac sodium salt, diclofenac sodium salt, ibuprofen sodium salt, telmisartan, furosemide, gemfibrozil, flufenamic acid, repaglinide, indomethacin, mefenamic acid, R-naproxen, probenecid, clofibric acid, valproic acid, S-naproxen, and diflunisal were purchased from Sigma-Aldrich (St. Louis, MO). Montelukast sodium hydrate and sodium meclofenamate monohydrate were obtained from Tokyo Chemical Industry (Chuo-ku, Tokyo, Japan). (RS)-Benoxaprofen, dabigatran, levofloxacin hemihydrate, and ibufenac were ordered from Toronto Research Chemicals (Toronto, ON, Canada). Tolmetin sodium was purchased from Santa Cruz Biotechnology (Dallas, TX). Pooled human liver microsomes were obtained from Gentest (Woburn, MA). Acetic acid glacial and ammonium acetate were ordered from Tedia (Fairfield, OH). Gradient-grade acetonitrile was purchased from Merck KGaA (Darmstadt, Germany). All the other chemicals and reagents used in this work were of analytical grade.
Acyl Glucuronide Formation and Acyl Migration.
The acyl migration protocol was adapted from previously published methods (Bolze et al., 2002; Chen et al., 2007; Jinno et al., 2013). The initial mixture contained potassium phosphate buffer (100 mM, pH 7.4), pooled human liver microsome (2 mg/ml), MgCl2 (1 mM), and test compound (100 μM). After preincubation at 37°C in a shaker for 5 minutes, the reaction was initiated by adding UDPGA (4 mM). The final reaction volume was 1 ml. The reaction was incubated at 37°C with shaking at 1000 rpm for 1.5 hours (Supplemental Method 1, Supplemental Table 2, and Supplemental Fig. 1), and then the incubation solution was divided into two equal fractions (A and B).
Fraction A (500 μl) was stopped by adding 500 μl of ice-cold acetonitrile containing 10% (v/v) of acetic acid, as acyl migration may be stabilized under an acidic condition (pH < 4.0) according to previous publications (Walker et al., 2007; Di Meo et al., 2013). The sample was mixed thoroughly and then centrifuged at 16,000g at 4°C for 10 minutes. The supernatant obtained from fraction A was directly injected onto a liquid chromatography coupled to mass spectrometry (LC-MS) system for analysis.
Fraction B (500 μl) was quenched by adding the same volume of ice-cold acetonitrile. Neutral acetonitrile without acid was used to quench the sample for 1-O-β-acyl glucuronide identification (fraction B) as the efficiency of the following β-glucuronidase hydrolysis study may be compromised in a low pH environment. The centrifugation conditions described above were applied. The supernatant from fraction B was vacuum-evaporated at 25°C and the residue was reconstituted into 400 μl of NH4OAc buffer (0.8 M, pH 5). The reconstituted solution was divided into two equal fractions (200 μl) and both incubated at 37°C for 120 minutes without or in parallel with 200 μl of β-glucuronidase (2.0 mg/ml) in the NH4OAc buffer. Reaction was subsequently quenched with 400 μl of ice-cold acetonitrile containing 10% (v/v) of acetic acid. After centrifugation at 16,000g and 4°C for 10 minutes, the supernatant was analyzed by an LC-MS system for data comparison.
LC-MS Analysis of Acyl Glucuronide.
The separation and qualitative determination of the primary acyl glucuronide (1-O-β-acyl glucuronide) and its isomeric products in the incubated samples were performed using an Agilent QTOF 6530 high resolution mass spectrometer (Agilent Technologies, Santa Clara, CA) coupled to an Agilent 1290 Infinity series HPLC system (Agilent Technologies, Waldbronn, Germany). Sample acquisition and data processing were carried out using MassHunter Workstation Data Acquisition version B.04.00 and MassHunter Workstation Qualitative Analysis revision B.04.00 (Agilent Technologies, Santa Clara), respectively. The Dual AJS ESI ionization was performed in positive (tolmetin, zompirac, telmisartan, montelukast, repaglinide, dabigatran, and levofloxacin) or negative (diclofenac, ibuprofen, furosemide, gemfibrozil, flufenamic acid, indomethacin, mefenamic acid, R-naproxen, probenecid, meclofenamate, (RS)-benoxaprofen, ibufenac, clofibric acid, valproic acid, S-naproxen, and diflunisal) ionization mode. The isomeric acyl glucuronides and the corresponding aglycone were detected by monitoring their molecular ions under full scan mode (m/z 100–1000). The HPLC column used to conduct separation was a core-shell C18 column (Kinetex 2.6-mm XB-C18 100Å, 2.1 mm × 100 mm; Phenomenex, Torrance, CA). Mobile phase A was water and mobile phase B was acetonitrile, both containing 10 mM of ammonium acetate. The flow rate was 0.5 ml/min. The HPLC solvent gradients were well optimized for each individual drug to maintain the separation of 1-O-β-acyl glucuronide from the corresponding isomers as well as the corresponding aglycone (for gradients, please see Supplemental Table 1). All primary acyl glucuronides (1-O-β-acyl glucuronide) and their isomeric products from the incubated samples were completely resolved within a 20-minutes run.
Data Analysis.
The migration rate of the acyl group from the 1-O-β-acyl glucuronide to its isomers was determined by the following equation:
(1)The glucuronidation rate of parent drug was also evaluated:
(2)All the peak areas used in this study were obtained from mass spectrometry data as previously discussed.
The correlation between acyl migration rates versus reported half-lives of 1-O-β-acyl glucuronide of selected carboxylic acid–containing drugs was evaluated by linear regression analysis (GraphPad Prism 6.04 software; GraphPad Software Inc., La Jolla, CA).
Results
Identification of Glucuronides of Selected Carboxylic Acid–Containing Drugs.
Twenty-three carboxyl acid–containing drugs were incubated with pooled human liver microsomes. Glucuronidation was confirmed as the major metabolic reaction in these incubations as only 1-O-β-acyl glucuronide and its isomeric products were detected.
The 1-O-β-acyl glucuronide was identified from its isomeric products by treatment with or without the β-glucuronidase. The peak that significantly decreased or disappeared in the presence of β-glucuronidase was identified as 1-O-β-acyl glucuronide (Wang et al., 2004). For tolmetin, telmisartan and ibuprofen, reference standards of the 1-O-β-acyl glucuronide were available.
Representative chromatograms of gemfibrozil incubated with or without β-glucuronidase are presented in Fig. 2. In these illustrations, three acyl glucuronides were detected from the gemfibrozil incubation samples that were quenched by acetonitrile containing 10% (v/v) of acetic acid (Fig. 2A). By comparing the chromatograms of samples incubated without (Fig. 2B) and with β-glucuronidase (Fig. 2C), it was confirmed that the peak at retention time of 5.22 minutes was from the 1-O-β-acyl glucuronide, whereas those at retention times of 6.76 minutes and 7.15 minutes were from the isomeric products of the 1-O-β-acyl glucuronide, as determined by the disappearance of the peak at 5.22 minutes when incubated in the presence of β-glucuronidase.
Representative chromatograms of gemfibrozil acyl glucuronide samples. (A) Gemfibrozil acyl glucuronide samples. (B) Gemfibrozil acyl glucuronide samples incubated without β-glucuronidase. (C) Gemfibrozil acyl glucuronide samples incubated with β-glucuronidase.
LC gradients for each individual test compound were optimized to obtain baseline separation of the 1-O-β-acyl glucuronide from the corresponding isomers in the incubation samples. The detailed LC separation results are shown in Table 1.
Retention time of parent drug and its corresponding acyl glucuronide metabolites
All compounds incubated with human liver microsomes and UDPGA showed a number of detected acyl glucuronide metabolites in the incubations, with the exception of those incubations with furosemide, dabigatran, and levofloxacin. For furosemide only the 1-O-β-acyl glucuronide was detected, whereas for dabigatran and levofloxacin no acyl glucuronide was detected in incubation samples.
The migration rate and glucuronidation rate of the different carboxylic acid drugs were calculated with eqs. 1 and 2, respectively, under the assumption that all isomeric acyl glucuronides and corresponding aglycone have similar ionization efficiency on the MS (Table 2).
Migration and glucuronidation rate of carboxylic acid–containing drugs
Samples were prepared in triplicate; the values represent mean ± S.D. (n = 3).
Compound Selection.
In this study, migration rates of 23 carboxylic acid–containing drugs covering a broad chemical space and different indications were evaluated.
Half-life data of 1-O-β-acyl glucuronide in phosphate buffer at pH 7.4 of tested compounds were collected from the literature (Fenselau, 1994; Williams and Dickinson, 1994; Ebner et al., 2010; Sawamura et al., 2010) and ranged from 0.3 to 79.0 hours. Detailed information on tested drugs is listed in Tables 1, 2, and 3.
Indications of selected carboxylic acid–containing drugs
Migration Rate and Stability of 1-O-β-Acyl Glucuronides.
The objective of this study was to evaluate the correlation between the migration rates of primary acyl glucuronide in carboxylic acid–containing drugs to the gold-standard half-life method of the 1-O-β-acyl glucuronides reference standard in phosphate buffer (Sawamura et al., 2010). The half-lives obtained from buffer with HSA (Castillo and Smith, 1995; Bolze et al., 2002) or buffer with fresh human plasma (Castillo and Smith, 1995; Chen et al., 2007) can also be used to categorize the acyl glucuronide reactivity of carboxylic acid–containing drugs. However, it was shown (Sawamura et al., 2010) that the use of either HSA or fresh human plasma may overestimate the reactivity of 1-O-β-acyl glucuronide.
Twenty-three carboxylic acid–containing drugs were analyzed and the corresponding migration rates were determined as described above. Furosemide showed a nonmeasureable migration rate (0%) because of its acyl glucuronide metabolite’s low signal intensity in MS analysis. With respect to levofloxacin and dabigatran, the migration rate could not be estimated mainly owing to the low glucuronidation of levofloxacin (Jinno et al., 2014) and potential low UDP-glucuronosyltransferase (UGT) activity to dabigatran. Accordingly, these three compounds were not included in the correlation analysis.
Figure 3 shows that a good correlation (R2 = 0.727) was obtained with reciprocals of migration rate plotted versus half-lives of the 1-O-β-acyl glucuronide for the remaining 20 carboxylic acid–containing drugs; the correlation improved (R2 = 0.831) when data were converted into log-log units.
Correlation between reciprocals of migration rate to half-lives of 1-O-β-acyl glucuronides. (A) Regression curve in linear scale. (B) Regression curve in logarithmic scale; origin was set at (–0.6, 0).
Discussion
FDA 2008 Guidance for Industry, Safety Testing of Drug Metabolites for small-molecule metabolite safety testing (Food and Drug Administration, 2008) stated specifically that additional reactivity evaluation of acyl glucuronide formed in human would be needed. The half-life approach is currently considered as the gold standard for acyl glucuronide reactivity assessment (Benet et al., 1993; Sawamura et al., 2010). Although this assay is robust and relatively easy to conduct, its application at early candidate selection stage is rather challenging as the method requires the synthesis of the 1-O-β-acyl glucuronide authentic standard. Thus, the objective of this study was to develop a fast, robust and predictive screening method to rank and flag compounds on the basis of their acyl glucuronide reactivity without synthesizing the 1-O-β-acyl glucuronide authentic standard.
This work demonstrated that the migration rate assay can nicely predict the buffer stability of 1-O-β-acyl glucuronides in carboxylic acid–containing drugs. Figure 3 shows that a good correlation (R2 = 0.831) was obtained with reciprocals of migration rate versus half-lives of 1-O-β-acyl glucuronides for the 20 carboxylic acid–containing drugs tested.
An in-depth analysis of the plot reveals that the correlations were beyond the 95% confidence interval of regression curve for meclofenamate and montelukast, and that reciprocals of migration rate of these two drugs underestimated the reported corresponding half-lives. It has been reported that the acyl migration played a more important role in reactivity compared with the hydrolysis of acyl glucuronide to the aglycone (Baba and Yoshioka, 2009; Di Meo et al., 2013). Both reactions were comprehensively monitored by the half-life approach, whereas the migration process was the only reaction evaluated in the acyl migration method. This discrimination may lead to underestimation of half-life values of some drugs for which the hydrolysis of 1-O-β-acyl glucuronide to aglycone is the dominant pathway. Therefore the application of the acyl migration method to assess reactivity of those compounds for ranking and selection may provide misestimated results. However, the acyl migration method was still applicable to a majority of drugs that have acyl glucuronide issues, as it has been reported that acyl migration was identified as the predominant reaction occurring in vitro (Walker et al., 2007). More importantly, the acyl migration method can provide preliminary reactivity information in a rapid screening assay that can be used to support the selection of low-risk drug candidates in the drug discovery phase that do not depend on authentic standards. Because of the metabolic properties or potentially poor UGT enzyme activity in pooled human liver microsome, trace or no acyl glucuronide metabolite was detected for some drugs. This could also lead to underestimation of migration rate.
We also evaluated whether there is benefit in factoring in compound-specific determinants affecting the 1-O-β-acyl glucuronide formation (e.g., UGT enzyme binding rate, test compound concentration, UGT isoform activity, etc.). Therefore, the migration rate data for each compound was also normalized by the respective glucuronidation rate calculated using eq. 2 (Fig. 4). However, the correlation of normalized migration rates versus half-lives (R2 = 0.543 in linear scale) showed no improvement when compared with the migration rates calculated using eq. 1 (R2 = 0.727 in linear scale). The possible explanations of this observation could be: 1) Non-cytochrome P450–dependent drug metabolism may lead to an overestimation of glucuronidation rate for drugs that could be metabolized by non-P450 enzymes; 2) different ionization efficiency between acyl glucuronide isomers and parent drug. Ionization efficiency of parent drug and metabolites has been extensively studied (Schadt et al., 2011), revealing that the MS signal intensity could vary remarkably, especially in cases of substantial structural differences between the parent drug and the metabolites. It is thus possible that the glucuronidation rate could not be properly assessed in absence of an analytical standard.
Correlation between reciprocals of migration rate normalized with glucuronidation rate to half-lives of 1-O-β-acyl glucuronides. (A) Regression curve in linear scale. (B) Regression curve in logarithmic scale; origin was set at (–1.0, –2.5).
The classification value of the half-life in potassium phosphate buffer that separated the perceived safe drugs from the drugs withdrawn from the market was calculated to be 3.6 hours (Sawamura et al., 2010). On the basis of this cut-off value, half-lives of the 1-O-β-acyl glucuronide of safe drugs were larger than 7.2 hours (flufenamic acid) and that of “warning” or “withdrawn” drugs were shorter than 3.2 hours (furosemide) (Table 2). Likewise, with some exceptions, the migration rate of the 1-O-β-acyl glucuronide of safe carboxylic acid compounds should be lower than 10%, whereas that of potentially unsafe compounds might be greater than 20% on the basis of our results (Table 2). With respect to the migration rates of the 1-O-β-acyl glucuronide in carboxylic acid drugs that were tested in the range of 10 to 20%, additional consideration should be made for the reactivity assessment.
The methods used to investigate the acyl glucuronide stability typically assess the half-life of an acyl glucuronide in buffer (Ebner et al., 2010), buffer with human serum albumin (Benet et al., 1993), or plasma (Sawamura et al., 2010), which require the authentic standard of the 1-O-β-acyl glucuronide. However, for compound optimization early in drug discovery, rapid methods are needed that do not depend on authentic standards. Additionally, methods have been documented in which the acyl glucuronide metabolite was first biosynthesized in vitro, and subsequently 1) incubated with human serum albumin (Bolze et al., 2002), or with a small peptide (Wang et al., 2004) to monitor covalent adducts; 2) incubated with plasma (Chen et al., 2007) or buffer (Jinno et al., 2013) to assess the half-life of the acyl glucuronide. Our simpler approach relying on the acyl glucuronide migration rate to evaluate reactivity of 1-O-β-acyl glucuronide can be used to support the selection of low-risk drug candidates in the drug discovery phase.
Acknowledgments
The authors thank Jon K. Bodnar for language editing and David Moore for helpful discussions.
Authorship Contributions
Participated in research design: Zhong, Lu, Ottaviani, Jones, Schadt.
Conducted experiments: Zhong.
Performed data analysis: Zhong, Lu, Ottaviani.
Wrote or contributed to the writing of the manuscript: Zhong, Lu, Ottaviani, Jones, Schadt.
Footnotes
- Received July 2, 2015.
- Accepted August 12, 2015.
↵
This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- HSA
- human serum albumin
- IDT
- idiosyncratic drug toxicity
- LC-MS
- liquid chromatography coupled to mass spectrometry
- UGT
- UDP-glucuronosyltransferase
- Copyright © 2015 by The American Society for Pharmacology and Experimental Therapeutics
References
In this issue
Assessment of Acyl Glucuronide Reactivity
