1206
Chem. Res. Toxicol. 2004, 17, 1206-1216
A Novel Approach for Predicting Acyl Glucuronide Reactivity via Schiff Base Formation: Development of Rapidly Formed Peptide Adducts for LC/MS/MS Measurements Jianyao Wang,* Margaret Davis, Fangbiao Li, Farooq Azam, JoAnn Scatina, and Rasmy Talaat Department of Drug Safety and Metabolism, Division of Biotransformation, Wyeth Research, 500 Arcola Road, Collegeville, Pennsylvania 19426 Received April 6, 2004
A novel technique to study the reactivity of acyl glucuronide metabolites to protein has been developed and is described herein. Considered here are acyl glucuronide metabolites, which have undergone the rearrangement of the glucuronic acid moiety at physiological temperature and pH. The investigation of the reactivity of these electrophilic metabolites was carried out by measuring the rate of reaction of rearranged AG metabolites in forming the corresponding acyl glucuronide-peptide adduct in the presence of Lys-Phe. This differs from the parallel technique used in forming AG adducts of proteins that have been previously reported. In the study described here, the Schiff base adduct, diclofenac acyl glucuronide-Lys-Phe product, was generated and structurally elucidated by liquid chromatography tandem mass spectrometry (LC/MS/MS) analysis. The product structure was proved to be a Schiff base adduct by chemical derivatization by nucleophilic addition of HCN and chemical reduction with NaCNBH3, followed by LC/MS/MS analysis. It is proposed here that the degree of reactivity of acyl glucuronides as measured by covalent binding to protein is proportional to the amount of its peptide adduct generated with the peptide technique described. The application of this technique to the assessment of the degree of reactivity of acyl glucuronide metabolites was validated by developing a reactivity rank of seven carboxylic acid-containing drugs. Consistency was achieved between the ranking of reactivity in the peptide technique for these seven compounds and the rankings found in the literature. In addition, a correlation (R2 ) 0.95) was revealed between the formation of a peptide adduct and the rearrangement rate of the primary acyl glucuronide of seven tested compounds. A structure effect on the degree of reactivity has demonstrated the rate order: acetic acid > propionic acid > benzoic acid derivatives. A rational explanation of this order was proposed, based on the inherent electronic and steric properties of each specific aglycone. In addition, adaptation of this technique to automation in order to more rapidly assess the ranking of reactivity of acyl glucuronide covalent binding to proteins by new chemical entities is proposed.
Introduction 1
AG metabolites are major metabolites for many drugs bearing a carboxylic acid moiety, and their properties have been studied extensively (1-7). Observed clinical adverse effects of such drugs may arise from the formation of covalent adducts of AG to proteins. This stable binding causes chemical alterations in proteins that are thought to contribute to drug toxicity, either through alteration of the functionality of the modified protein or through antigen formation, causing subsequent immune reactions. The covalent binding may occur via two * To whom correspondence should be addressed. Tel: 484-865-8792. Fax: 484-865-9404. E-mail:
[email protected]. 1 Abbreviations: CH OH, methanol; ESI, electrospray ionization; 3 UDPGA, uridine 5′-diphosphoglucuronic acid; MS, mass spectrometry; LC/MS, HPLC interfaced with mass spectrometry; LC/MS/MS, HPLC interfaced to tandem mass spectrometry; AG, acyl glucuronide; AGP, acyl glucuronide-peptide; DCL, diclofenac; FEN, fenoprofen; FUR, furosemide; IBU, ibuprofen; KET, ketoprofen; TOM, tolmetin; ZOM, zomepirac; HSA, human serum albumin; HLM, human liver microsomes; NaOAc, sodium acetate; SIM, selected ion monitoring; CID, collision-induced dissociation; NAC, N-acetylated cystein.
different mechanisms. The first is a transacylation mechanism, where a nucleophilic group (-NH2, -OH, or -SH) on a protein attacks the carbonyl group of the primary AG leading to the formation of an acylated protein and free glucuronic acid. The second is a mechanism of Schiff base formation where condensation occurs between the aldehyde group of a rearranged AG with the amine group of the N terminus and/or the -NH2 of a lysine residue of a protein, leading to the formation of a glycated protein (8) (Figure 1). An in vitro assessment of AG reactivity regarding covalent binding to protein is likely to be of significant importance for predicting the toxicity of drug candidates. Several procedures have been developed for assessing the extent of AG covalent binding to proteins. For example, Benet et al. (9) have summarized, for six drugs, a correlation between the extent of covalent binding to protein and the “global degradation rate constant” (hydrolysis and intramolecular rearrangement) of the 1-Oβ-glucuronide adduct. The conclusion reached was that
10.1021/tx049900+ CCC: $27.50 © 2004 American Chemical Society Published on Web 08/06/2004
Covalent Binding to Proteins by Acyl Glucuronide
Chem. Res. Toxicol., Vol. 17, No. 9, 2004 1207
Figure 1. Schematic illustration of 1-O-J-AG and its successive reactions: transacylation to form acylated protein from primary AG and rearrangement by intramolecular acyl migration to form the 2-, 3-, and 4-C position isomers, which allow Schiff base formation between amine groups of proteins leading to the formation of glycated proteins. Nucleophilic addition of HCN at the imine bond forms R-aminonitrile adducts, and reduction of the imine bond by NaCNBH3 forms a secondary amine.
the degree of the R-carbon substitution of the carboxylic acid group of the parent drug could affect the reactivity of the AG. With eight drugs, Bolze et al. (10) established a correlation between the extent of covalent binding to protein by AG with the reappearance of the aglycone, weighted by the percentage of rearrangement of the primary glucuronide via complete hydrolysis of the protein adducts, followed by LC/MS/MS of released aglycone. It is relevant that the above-mentioned studies used HSA as the binding target. In the procedures described, the protein adduct was precipitated and centrifuged to form protein pellets. The pellets required exhaustive washing to remove reversible, noncovalently bound AG from the protein. The extent of covalent binding for the radiolabeled compounds studied was measured by scintillation counting. For studies with nonlabeled compounds, the protein pellets were subjected to hydrolysis in a strong alkaline solution to release the aglycone prior to its quantitative analysis (8). With respect to the study of AG reactivity in the drug discovery process, the above-mentioned procedures have some inherent difficulties. The experimental procedures for the HSA studies described are tedious and time consuming. The radiolabeled compounds of interest in most cases are not available at the early stage of drug discovery. Also, the results obtained lacked the specificity of the structures of the covalent adducts. To attempt to adapt such procedures to automation (a widely applied technique in drug discovery) would be a formidable, if not impossible, task using the techniques described above. The objective of the work described here was to develop a novel, simple, and mechanistically well-defined technique that could be easily adapted for the automation of determining the reactivity of AG metabolites toward the
formation of AGP. The developed technique involved the reaction of rearranged AG with small peptides instead of proteins, such that the amount of AGP formed could be directly quantified by LC/MS/MS of the reaction mixture without any pretreatment. Experimentally, two incubation steps were involved. The first step consisted of the biosynthesis of AG metabolites by incubation of carboxylic acid-containing compounds with HLM in the presence of UDPGA. The second step was the formation of AGP adducts by the addition of a small peptide into the first step’s incubation supernatant after the microsomal proteins had been removed. The quantitative and qualitative information regarding the AGP adduct could then be determined by LC/MS and LC/MS/MS of the final reaction mixture. In the present study, seven carboxylic acid-bearing drugs that have been reported to bind covalently to protein, in vitro and in vivo, with different degrees of reactivity (10-17) were selected to evaluate the procedure. The structures of the compounds used are shown in Figure 2 and include DCL, FEN, FUR, IBU, KET, TOM, and ZOM. Among these seven drugs, TOM and ZOM have been withdrawn from the market for their clinical toxicity; DCL was reported to cause hepatocelluar damage; IBU is considered to be the safest nonsteroidal antiinflammatory drug; FUR was observed to have the least covalent binding to proteins. Structurally, DCL, ZOM, and TOM bear an acetic acid moiety; IBU, FEN, and KET bear an iso-propionic acid moiety; and FUR bears a benzoic acid moiety.
Materials and Methods Materials. HLM was purchased from In Vitro Technologies (Baltimore, MD); Lys-Phe, β-glucuronidase (363 200 units), UDPGA, MgCl2, KH2PO4, acetic acid, NaCN, NaCNBH3, DCL,
1208
Chem. Res. Toxicol., Vol. 17, No. 9, 2004
Wang et al.
Figure 2. Structures of seven model drugs for validating the newly developed technique (AA, acetic acid derivative; IPA, isopropionic acid derivative; and BA, benzoic acid derivative). FEN, FUR, IBU, KET, TOL, and ZOM were purchased from Sigma-Aldrich (St. Louis, MO). Biosynthesis of AGs. The general procedure for AG biosynthesis involved incubation of 300 µM (final concentration) each substrate with 1.5 mg/mL HLM, 4 mM UDPGA, and 1 mM MgCl2 in 0.1 M K2HPO4 buffer (pH 7.4). The final volume was brought to 2.0 mL and then incubated at 37 °C for 1.5 h. The resulting biosynthesized AGs were then subjected to the following experimental steps. Measurement of Total AG by LC/MS/MS. A 100 TL aliquot of the microsomal reaction mixture described above was transferred into a 1.5 mL eppendorf tube containing 200 µL of 50/50 CH3OH/water (v/v). The pH was adjusted to 3.5 using 10% acetic acid. The reaction mixture was vortexed for 1 min, followed by centrifugation at 14000 rpm and 5 °C for 10 min using a Beckman, Allegra 21R centrifuge (Beckman Instruments Inc., Palo Alto, CA). An aliquot of 100 µL of the supernatant of each was transferred into a sample vial containing 100 µL of an internal standard [phenylpropionic acid, 10 µg/mL in 20/80 CH3OH/H2O (v/v)] for HPLC. The samples were used for both qualitative and quantitative measurement of total AG formation by LC/MS/MS. Identification of 1-O-β-AG Using β-Glucuronidase. The primary AG, 1-O-J-AG, is the only substrate for which β-glucuronidase can hydrolyze an AG to glucuronic acid and the corresponding aglycone. The enzymatic hydrolysis was conducted by first transferring 400 µL of each microsomal reaction mixture to a 2.0 mL eppendorf tube containing 1200 µL of acetonitrile. The resulting mixture was vortexed for 1 min, followed by centrifugation at 14000 rpm and 5 °C for 10 min. The supernatant was transferred into another test tube and dried under a stream of N2. The dried residue was reconstituted into 400 µL of 0.8 M NaOAc buffer at pH 5.0. One aliquot (200 µL) of this mixture was mixed with 400 TL of 20 mg/mL β-glucuronidase in NaOAc buffer. A second aliquot of 200 µL was mixed with 400 µL of NaOAc buffer to serve as a negative control. Both solutions were incubated at 37 °C for 24 h. The reaction was quenched by adding 5 volumes of acetone. The mixtures were vortexed for 1 min, followed by centrifugation at 14000 rpm and 5 °C for 10 min. The supernatant was transferred into another test tube and dried under a stream of N2. The dried residues were reconstituted into 200 µL of 20/80 CH3OH/water (v/v). The solutions were analyzed qualitatively and quantitatively using LC/MS/MS to identify the primary AG in the microsomal reaction mixture.
Determination of the Time Course of AGP Formation. A third aliquot (1.5 mL) of the microsomal reaction mixture was centrifuged at 14000 rpm and 5 °C for 15 min to precipitate the microsomal proteins. The supernatant was transferred into another test tube containing an equal volume of 20 mM KF in 0.1 M K2HPO4, pH 7.4, buffer. The reaction was carried out at 37 °C. An aliquot of 50 TL of the reaction mixture was removed at each time: 0, 1.5, 3, 18, 27, 42, and 67 h of incubation. The samples were transferred into a sample vial containing 50 µL of an internal standard [phenylpropionic acid, 10 µg/mL in 20/ 80 CH3OH/H2O (v/v)]. The AGP adducts formed were analyzed qualitatively and quantitatively using LC/MS/MS. Chemical Derivatization. The following two chemical derivatization reactions were carried out using DCL-AGP as a model to confirm the presence of Schiff base in the AGP adduct. Nucleophilic Addition of NaCN to Schiff Base Adducts. An aliquot of 200 µL of the AGP reaction mixture from the 67 h time point was transferred into another test tube containing 100 µL of NaCN (100 mM in 0.1 M K2HPO4, pH 7.4). The reactants were maintained at 37 °C for 3 h. The product of this reaction was directly analyzed by LC/MS/MS for structural elucidation. Reduction of Schiff Base with NaCNBH3. An aliquot of 200 µL of the AGP reaction mixture from the 67 h time point was evaporated under a stream of N2 and reconstituted in 200 µL of CH3OH. An equal volume of NaCNBH3 (100 mM in CH3OH) was added. The reduction reaction was carried out at 37 °C for 2 h with stirring. The reaction was stopped by the addition of 100 µL of 10% aqueous acetic acid. The mixture was dried under a stream of N2 and reconstituted to 200 µL of 20/80 CH3OH/H2O (v/v) for structural elucidation using LC/MS/MS analysis. LC/MS/MS Analysis. Samples generated from the above experiments were analyzed using LC/MS/MS. Qualitative LC/ MS/MS was performed on a Surveyor-LCQ deca HPLC/MS system (ThermoFinnigan, San Jose, CA), including an autosampler and diode array detector, operated by Xcalibur version 1.3. Quantitative analysis was performed on a ThermoFinnigan TSQ Quantum mass spectrometer, also operated by Xcalibur version 1.3. The mass spectrometer was interfaced to an Agilent 1100 HPLC system (Agilent Technologies, Palo Alto, CA). The HPLC conditions for all qualitative analyses were set up with the UV detector monitoring 200-600 nm. The column used was a Luna C8 2 mm × 150 mm, 5 µm (Phenomenex,
Covalent Binding to Proteins by Acyl Glucuronide
Chem. Res. Toxicol., Vol. 17, No. 9, 2004 1209
Figure 3. Molecular ion chromatograms of deprotonated DCL-AG, m/z 470, in negative ion ESI MS mode. (A) A 1.5 h biosynthesis solution showing primary AG and its rearranged isomers. (B) AG biosynthesis solution treated with J-glucuronidase showing enzyme hydrolysis resistant rearranged AG isomers.
Figure 4. Product ion spectra of the protonated molecular ion of DCL-AG from LC/MS/MS analysis on 1.5 h biosynthesis solution. (A) Primary DCL-AG at a RT of 15.88 min, (B) rearranged DCL-AG at a RT of 15.12 min, and (C) rearranged DCL-AG at a RT of 16.30 min. Torrance, CA) using a flow rate of 200 µL/min. The mobile phase consisted 5 mM aqueous ammonium acetate (A) and acetonitrile (B). The separation of different AGs of seven drugs was achieved using the gradient elution mode. The total run time was 30 min. Negative ion ESI LC/MS analysis was performed on a TSQ Quantum/Agilent 1100 HPLC system. The HPLC effluent was introduced directly into the instrument. To prevent in-source
fragmentation, the operation parameters were adjusted as softly as possible. The heated capillary inlet temperature was set at 250 °C. The auxiliary gas flow was set at 20, and the sheath gas was 60 psi. The electrospray voltage was -3.5 kV. The ionization source operational parameters were carefully adjusted, which were optimized not only to minimize in-source fragmentation but also to maximize the detection of the (M -
1210
Chem. Res. Toxicol., Vol. 17, No. 9, 2004
Wang et al.
Figure 5. LC/MS/MS analysis of DCL-AGP of Lys-Phe in the second incubation solution. (A) Ion chromatogram of the deprotonated molecular ion of DCL-AGP, m/z 745. (B) Positive ion full scan mass spectrum of the LC/MS peak at 18.07 min. (C) Product ion spectrum of m/z 747 at 18.07 min. (D) Product ion spectrum of ions at m/z 747 at a RT of 18.87 min (the 3-AG-Lys-Phe adduct was used as an example for the fragmentation illustration). H)- ions of AG and AGP of each compound to be analyzed. The SIM mode was applied to the quantitative analysis of all seven AGs, AGPs, and internal standards by monitoring their respective molecular ions. The peak areas of all analytes were integrated and compared using Xcalibur version 1.3 quantitative software. Positive ion ESI LC/MS/MS product ion scanning was performed on the LCQ deca for structural elucidation. The mass spectrometer acquisition time was 25 min. HPLC effluent (200 µL/min) was introduced directly into the ionization source of the MS. The heated capillary inlet temperature was set at 350 °C. The auxiliary gas flow was set at 40, and the sheath gas
was 90 psi. The electrospray voltage was 4.5 kV. The operational parameters of the mass analyzer were optimized for the detection of each of the AGs and AGPs. The full scan mass range was set from 150 to 1000 Da. The MS/MS collision energy was set at 28 eV, 25% of Q, and 30 ms activation.
Results Characterization of AG Metabolites and AGP Adducts Using LC/MS/MS. Structural elucidation of AG metabolites and AGP adducts for all compounds were obtained from the interpretation of product ion scanning
Covalent Binding to Proteins by Acyl Glucuronide
Chem. Res. Toxicol., Vol. 17, No. 9, 2004 1211
Figure 6. Product ion spectra of protonated AGP-Lys-Phe adducts of seven compounds: (A) DCL-AGP, parent ion m/z 747; (B) FEN-AGP, parent ion m/z 694; (C) FUR-AGP, parent ion m/z 782; (D) IBU-AGP, parent ion m/z 658; (E) KET-AGP, parent ion m/z 706; (F) TOM-AGP, parent ion m/z 709; and (G) ZOM-AGP, parent ion m/z 743.
LC/MS/MS data. The prototypical procedure, using drug, HLM, and UDPGA at physiological temperature and pH for 1.5 h, was used. As an example, the AG isomers of DCL eluted with retention times of 15.12, 15.88, and 16.30 min, as are shown in Figure 3. The product ions of the protonated molecular ions of the three isomeric DCL glucuronides (m/z 472) are presented in Figure 4A-C. It is noteworthy that the product ion scanning of 1-O-DCL glucuronide produced an ion at m/z 296, corresponding to the neutral loss of the glucuronic acid moiety as a base peak (Figure 4A). The rearranged AGs, under CID, displayed the loss of water as the base peak with only a minimal loss of the glucuronic acid moiety (Figure 4B,C). This fragmentation behavior could be rationalized by the fact that the O-C1 bond is relatively weak to be preferably cleaved among four O-C bonds in the glucuronic acid moiety since C1 is an acetal carbon. The structural characterization of the Schiff base adduct of DCL-AGP byproduct ion scanning LC/MS/MS was carried out by detection of two isomeric peptide adducts in the second step incubation solution. These had retention times of 18.07 and 18.87 min (Figure 5A). The full scan mass spectrum of the peak that eluted at 18.07 min displays that the adduct contains two chlorine atoms, equaling the number of chlorine atoms in DCL. This adduct displayed protonated molecular ions at m/z 747 and m/z 749 with an intensity ratio of 3:2. Data corresponding to the peak at 18.87 min show a similar spectrum (Figure 5B). The product ion mass spectra of the protonated molecular ions (m/z 747 and m/z 749) of two isomeric DCL-AG-Lys-Phe adducts are presented in Figure 5C,D. The fragmentation pathway is illustrated in the figure, and from the interpretation of the spectrum, it was proposed that the Schiff base product formed via
reaction of the rearranged AG of DCL with the -NH2 of the dipeptide. The product ions at m/z 729 and m/z 711 correspond to the consecutive neutral loss of water molecules (18 Da) from the glucuronic acid moiety. The ion at m/z 582 corresponds to a neutral loss of the phenylanaline moiety (165 Da). The ion at m/z 470 is derived from a neutral loss of dehydrated DCL (277 Da). The ions at m/z 452, m/z 434, and m/z 416 display consecutive losses of water from m/z 470. The ion at m/z 294 is the protonated molecular ion of Lys-Phe, which could be formed via mediating a ring close structure through C-5 oxygen attacking on the imine carbon. The Schiff base AGPs from seven drugs were structurally elucidated by product ion scanning LC/MS/MS. The MS/MS analyses for structure elucidation are presented in Figure 6. Proposed origins of structural diagnostic product ions utilized are summarized in Table 1. Characterization of Chemically Derivatized AGP Adducts Using LC/MS/MS. Because AGP adducts appear to arise via Schiff base formation, further evidence was sought for the presence of the characteristic imine bond of Schiff base adducts. This was done using chemical derivatization via nucleophilic addition of HCN to the imine bond to form an R-aminonitrile and chemical reduction of NaCNBH3 to form a secondary amine (Figure 1). The R-aminonitrile product of the DCL-AGP adduct has molecular ions of m/z 774 and m/z 776 that are 27 Da higher than that of nonderivatized DCL-AGP (Figure 7A). The reduced peptide adduct showed molecular ions at m/z 749 and m/z 751, 2 Da higher than those of the DCL-AGP adduct (Figure 7B). The product ion spectra of the molecular ions of the chemically derivatized products are shown in Figure 8A,B. It was observed that most of the diagnostic product ions were formed by the
1212
Chem. Res. Toxicol., Vol. 17, No. 9, 2004
Wang et al.
Table 1. Product Ion Identities of the LC/MS/MS Analysis of AGPs of Seven Drugs ion’s origin MH+
-H2O -F -H2O/-Phe -H2O/-drug KFH+
DCL-AGP
FEN-AGP
FUR-AGP
IBU-AGP
KET-AGP
TOM-AGP
ZOM-AGP
747 729 582
694 676 529 511 434 294
782 764 617
658 640 493 475 434 294
706 688 541
709 691
743 725
526 434
560 434
434 294
434 294
434 294
Figure 7. Positive ESI full scan mass spectra of DCL-AGP of Lys-Phe derivatives: (A) nucleophilic addition with NaCN addition; (B) chemical reduction with NaCNBH3 (the 3-AG-Lys-Phe adduct was used as an example for the fragmentation illustration).
same neutral losses as those of the DCL-AGP adduct. Briefly, as illustrated in Figure 9A, the molecular ion of the R-aminonitrile derivative of DCL-AGP, m/z 774, under CID conditions, produces the product ion m/z 747 that results from the neutral loss of HCN (27 Da). The product ion at m/z 758 corresponds to the neutral loss of water from the glucuronic acid moiety. The neutral loss of the phenylalanine moiety gives rise to the ion at m/z 609, and the product ion at m/z 479 is derived from the neutral loss of the elements of DCL (295 Da). m/z 470, 452, 434, and 294 could be the same origin described above for nonderivatized DCL-AGP. In the product ion spectrum of the reduced AGP (Figure 8B, [M + H]+ ) m/z 749), the most structural characteristic ions were 2 Da higher than those of DCL-AGP. The product ion at m/z 731 arises from a neutral loss of water (18 Da), and the product ion at m/z 584 corresponds to the neutral loss of the phenylanaline (165 Da). The ion appearing at m/z 472 is due to a neutral loss of dehydrated DCL (277 Da). On the basis of the above LC/MS/MS analysis results, it can be confirmed that the DCL-AGP adduct was generated via Schiff base formation. This was due to the condensation of the aldehyde of the rearranged AG isomer with the N-terminal or -amine of Lys-Phe. Determination of the Rearrangement Percentage of AG Metabolites. The extent of rearrangement of the primary AG metabolite in 1.5 h microsomal incubations was quantified, based on molecular ion current peak areas of AGs extracted from LC/MS analysis. For ex-
ample, Figure 3A shows the ion current chromatogram of the DCL-AG (M - H)- ion (m/z 470) of three AGs. In Figure 3B, the two peaks shown are the DCL-AGs (rearrangement products) that were resistant to Jglucuronidase hydrolysis. Thus, the primary AG metabolite was identified as peak “b”. The AG rearrangement percentage (AGr%) of DCL-AG was arrived at by dividing the sum of the ion current peak areas of “a” and “c” by the sum of the peak areas of a, b, and c (Figure 3A), multiplied by 100. Peak c may contain two isomers coeluted, which could be separated in a longer HPLC gradient. The AGr% values of the seven compounds tested are listed in Table 2 and were calculated on the basis of the assumption that peaks a, b, and c have the same ionization potential. Determination of the Reactivity Index (RI) of AG Metabolites. The amount of AG covalently bound to protein is dependent on the time, the pH, the concentration of AG present in the reaction system, and the degree of reactivity of the AG metabolite to proteins/peptides. In fact, AG reactivity has been expressed as the percentage of total AG involved in the reaction with protein. In the technique described here, the AG reactivity was determined as a RI. This was calculated as the ratio of ion current chromatographic peak areas of AGPs to those of the corresponding AGs from the same sample, multiplied by 100. The following equation was derived for the determination of RI for tested compounds: RI ) (AAGPt × 100)/AAG0, where AAGPt represents the peak area of the
Covalent Binding to Proteins by Acyl Glucuronide
Chem. Res. Toxicol., Vol. 17, No. 9, 2004 1213
Figure 8. Positive MS/MS spectra of DCL-AGP of Lys-Phe derivatives: (A) nucleophilic addition with NaCN addition, m/z 774; (B) chemical reduction with NaCNBH3, m/z 749. Table 2. Rearrangement Percentage, RI of a Single Time Point (67 h), and RI h-1 compd name
AGr%
RI
RI h-1 (R2)
TOM ZOM DCL KET FEN IBU FUR
80.0 67.5 58.4 36.4 33.8 15.8 2.0
1.5 1.33 0.88 0.47 0.46 0.33 0.03
0.0224 (0.98) 0.0208 (0.98) 0.0124 (0.98) 0.0090 (0.99) 0.005 (0.96) 0.0067 (0.99) 0.0006 (0.90)
Figure 9. RI of a single time of 67 h of seven drugs.
molecular ion of AGPs measured at a specified time following incubation with Lys-Phe, while AAG0 is the peak area of the total AG measured at the beginning of incubation with Lys-Phe. For example, the RI for DCLAG was calculated after a 67 h of incubation with LysPhe and determined to be 0.88%. This was determined by dividing the sum of the peak areas of a and b shown in Figure 5A, which represent DCL-AGP present in the 67 h incubation, by the sum of the peak areas in Figure 3A, which is equivalent to the total peak area of DCLAG metabolites at time 0 prior to the addition of LysPhe. The RI values of seven drugs are listed in Table 2.
Reactivity Ranking of the AGs of Seven Drugs. The AG reactivities of seven drugs were ranked based on two criteria. The first was on the determination of the RI measured at a single time point. The second was based on the measurement of the RI over the 67 h incubation, RIh-1. Figure 9 displays the diagram of the RIs of the seven model compounds achieved by measurement at a single time point (67 h). The RIs for the seven compounds measured over the course of time of the incubation are shown in Figure 10. The rate of reactivity (RIh-1) can then be expressed as the slope of a regression curve for each compound as shown in Figure 10 and summarized in Table 2.
1214
Chem. Res. Toxicol., Vol. 17, No. 9, 2004
Figure 10. Time course of RI from 0 to 67 h and their regression curves.
Discussion As a major metabolite of most carboxylic acid-bearing compounds, AGs have been demonstrated to be labile electrophiles that can react chemically with the nucleophilic functional groups of tissue macromolecules, leading to their covalent binding. The chemical reactivity of these conjugates is suspected to be responsible to the toxicity observed for drugs containing a carboxylic acid group (18-21). It has been suggested that long-lived, drugaltered proteins may act as immunogens and produce cytotoxic T-cell-mediated or antibody-dependent, cellmediated toxicity in susceptible patients (22). Imine formation is reversible but may be followed by an Amadori rearrangement of the imino sugar to the more stable 1-amino-2-keto product. However, this rearrangement was not defined in this experiment. In the current studies, Schiff base AGPs demonstrated a steady stability in physiologic pH and temperature and displayed a linear increase in AGP formation over the entire time frame from 0 to 67 h. The combination of information gained from the use of the new technique described here to measure the covalent binding of AGs with a small peptide, along with that information gained through protein-targeted processes, may lead to an enhanced understanding of the clinical adverse effects of carboxylic acid-containing drugs. The two Schiff base AGP isomers displayed in Figure 5A with retention times (RTs) of 18.07 and 18.87 min could be formed via the aldehyde group of different isomeric AG condensed with I-NH2 of the Lys residue, which was proven by NMR when Lys-Phe was used for another reaction system as a nucleophilic trapping agent (the results will be shown in a later publication). The higher reactivity corresponding to the higher nucleophilicity of I- over M-NH2 could be understood based on their pKa values of -NH3+ forms, which is 9.2 for I-NH2 while it is 10.8 for M-NH2 (23), which means that there are about 40 time folds of free amine formed in the I chain over in the M chain in the given buffer. The free amine is believed to be acting as a reactive chemical form for the Schiff base formation. The Schiff base mechanism, one of the two processes proposed for the covalent binding of AGs to proteins, was specifically investigated to assess the reactivity of seven model drugs using a new technique herein described. Using this method, there was consistency found between the reactivity rank of the seven AGs, where a single time RI was applied for ranking, with those reported by Benet et al. (9) and Bolze et al. (10) where they measured the total amount of covalent protein binding by the two mechanisms to rank the reactivity of various carboxylic
Wang et al.
Figure 11. Correlation of the RI with the rearrangement rate (the carboxylic acid moiety was grouped with the dashed line).
acid-bearing drugs. The consistency of findings here and the findings of Benet et al. and Bolze et al. proved the practicality of the proposed technique to study Schiff base reactivity. Among these seven drugs studied, Schiff base adducts with proteins have been reported for TOL, ZOM, and DCL. Ding et al. identified the glycated peptide fragment using LC/MS/MS analysis on tryptic-digested HSA incubated with TOL-AG (24-25). Smith et al. (26) and Kretz-Rommel et al. (27) confirmed the occurrence of Schiff base formation given the observation that the extent of covalent binding increased during the addition of NaCN to the reaction systems of ZOM- and DCL-AGs incubated with proteins. The work described here demonstrates also the relationship between the rate of AG rearrangement and the AGP formation. As demonstrated in Figure 1, the rearrangement of AGs is a prerequisite for Schiff basepeptide linkage. It was proposed that the extent of formation of Schiff base-peptide adducts from AGs should be proportional to the rearrangement rate of the primary AG. A correlation (R2 ) 0.95) between the RI, C%, and percentage of rearrangement of the primary AG of seven compounds is displayed in Figure 11. On the basis of the results reported here from use of a small peptide, it may eventually be claimed that the covalent binding reactivity of AG to protein can be predicted based on its rearrangement rate. Furthermore, these studies investigate the relationship between the structure of the aglycone and the reactivity of AGs. In addition to well-known environmental factors, such as pH and temperature, the rate of AG rearrangement and its transacylation interaction with protein can be affected by the structure of the aglycone. Chemically, the rearrangement of an AG is an intramolecular transesterification where the acyl group migrates between the hydroxyl groups of the glucuronic acid moiety, driven by the nucleophilic attack from each adjacent -OH group, whereas transacylation can be considered as an intermolecular transesterification, where the acyl group is attacked by the nucleophilic groups of protein/peptide. The rearrangement and transacylation rate of AG could be affected by the inherent electronic and steric properties of the aglycone. The consistency on the reactivity ranking for seven model compounds demonstrated by the peptide technique (Schiff base only) and by protein binding (transacylation and Schiff base) could be rationalized as the facts to affect AG reactivity are same for both mechanisms. Although the number of compounds tested was limited to only seven, the order of rate of rearrangement observed was acetic acid > propionic acid > benzoic acid, which implies that inherent electronic and
Covalent Binding to Proteins by Acyl Glucuronide
steric properties may play an important role in affecting the rate of primary AG rearrangement. It could be hypothesized that the drug bearing the carboxylic acid group, bound to an aromatic group, demonstrates the lowest extent of rearrangement due to resonance stabilization provided by the aromatic moiety. Isopropionic acid derivatives display a slower rearrangement rate than those of acetic acid derivatives, possibly due to the higher steric hindrance capacity of the isopropyl group over the acetyl group. Smith et al. (28) have reported that etodolac (an acetic acid derivative) produced minimal covalent binding of proteins to AGs. On the basis of Benet’s assertion (9) that the reactivity of carboxylic acidcontaining compounds depends on the degree of substitution at the R-carbon of the carboxylic acid, it would be expected that etodolac would generate high covalent binding given that the R-position of the carboxylic acid is saturated. Apparently, the degree of substitution at the R-carbon position does not solely explain reactivity. It might even be proposed that the low rate of reactivity of etodolac to proteins might arise from the low degradation rate of its primary AG, which could be caused by steric hindrance from the ethyl group on the β-carbon to the carboxylic acid moiety. Recently, Sidenius et al. have revealed that the substitution of the β-carbon can affect the reactivity of carboxylic acyl-CoA thioesters with GSH (29). Because of their low rate of covalent binding to protein, it might be suggested that compounds bearing benzoic acid or isopropionic acid substitutions at the R-carbon of the drug should be considered when designing new chemical entities, provided these functional groups do not undermine desirable pharmacological activity. Regardless, any conclusion based on a correlation between AG reactivity and aglycone structure requires further study. A transacylation mechanism, suggested as another process for AG covalent binding to protein, was not addressed here experimentally. There is literature to support a role for this mechanism, including Stogniew et al. (30), who first reported that the formation clofibrate mercapturate in humans via transacylation; Van Breeman et al. (31), who proposed that transacylation occurred in the interaction of flufenamic glucuronide and HSA; and Ruelius et al. (32), who reported later that only the transacylation process could be observed in oxaprozin covalent binding to protein. Recently, Grillo et al. (3335) reported the formation of ZOM and DCL-acyl-GSH formation clearly resulting from a transcylation process by incubating primary ZOM-AG and DCL-AG with GSH in buffer. In addition, the DCL-SG conjugate was found to be more reactive to NAC than was NAC to DCL-AG, further supporting a transacylation mechanism. The study of transacylation-related protein covalent binding of AGs is underway. Automation of many laboratory techniques is used to help satisfy the high throughput requirement for drug discovery. As it has been described, the most distinguishable advantage of the new technique described herein is its simplicity. Analysis of a reaction mixture by LC/MS without any pretreatment is already a process that is extensively automated. It has been taken into consideration that a screening technique using LC/MS analysis to determine the tendency of the AG of a carboxylic acidcontaining drug to form an AGP should include at least two reference compounds with differing degrees of reactivity, such as IBU and DCL. The RIs determined for new
Chem. Res. Toxicol., Vol. 17, No. 9, 2004 1215
chemical entities as compared to those of the reference compounds should reveal the relative reactivity of the candidate to the reference compounds, perhaps appearing as high as DCL or as low as IBU. The time required to perform each experiment is as little as 1 day. It will entail 1.5-2.0 h for biosynthesis of the AG and overnight incubation for AGP formation. Adaptation of this new procedure for automation to increase throughput of the determination of the reactive natures of carboxylic acid-containing new chemical entities is underway. In addition, to gain a more comprehensive understanding of the reactivity of AG metabolites in the formation of covalent bonds with proteins, studies have been initiated to evaluate the involvement of a possible transacylation pathway of AGP formation, using small peptides and LC/MS/MS analytical approaches. To date, this is the first report to demonstrate the Schiff base formation in AGPs by covalent binding of AGs with a very small peptide. Short peptides used to trap the reactive rearranged aldehyde from the glucuronic acid moiety need not be limited to KF, as we have successfully characterized the AGP adducts formed by DCL-AG with Lys-Tyr-Lys, Val-Tyr, leucine-enkephalin, Ala-Gly-Gly, and GSH by LC/MS/MS (data not shown.). The Lys-Phe used in this experiment showed advantages over all of the other small peptides with regard to analytical considerations.
References (1) Bailey, M. J., and Dickinson, R. G. (2003) Acyl glucuronide reactivity in perspective: biological consequences. Chem.-Biol. Interact. 145 (2), 117-137. (2) Akira, K., Uchijima, T., and Hashimoto, T. (2002) Rapid internal acyl migration and protein binding of synthetic probenecid glucuronides. Chem. Res. Toxicol. 15 (6), 765-772. (3) Dansette, P. M., Bonierbale, E., Minoletti, C., Beaune, P. H., Pessayre, D., and Mansuy, D. (1998) Drug-induced immunotoxicity. Eur. J. Drug Metab. Pharmacokinet. 23, 443-451. (4) Faed, E. M. (1984) Properties of acyl glucuronides: implications for studies of the pharmacokinetics and metabolism of acidic drugs. Drug Metab. Dispos. 15, 1213-1249. (5) Park, B. K., Coleman, J. W., and Kitteringham, N. R. (1987) Drug disposition and drug hypersensitivity. Biochem. Pharmacol. 36, 581-590. (6) Riley, R. J., and Leeder, J. S. (1995) In vitro analysis of metabolic predisposition to drug hypersensitivity reactions. Clin. Exp. Immunol. 99, 1-6. (7) Spahn-Langguth, H., Dahms, M., and Hermening, A. (1996) Acyl glucuronides. covalent binding and its potential relevance. Adv. Exp. Med. Biol. 387, 313-328. (8) Spahn-Langguth, H., and Benet, L. Z. (1992) Acyl glucuronides revisited: is the glucuronidation process a toxification as well as a detoxification mechanism? Drug Metab. Rev. 24, 5-48. (9) Benet, L. Z., Spahn-Langguth, H., Iwakawa, S., Volland, C., Mizuma, T., Mayer, S., Mutschler, E., and Lin, E. T. (1993) Predictability of the covalent binding of acidic drugs in man. Life Sci. 53, 141-146. (10) Bolze, B., Bromet, N., Gay-Feutry, C., Massiere, F., Boulieu, R., and Hulot, T. (2002) Development of an in vitro screening model for the biosynthesis of acyl glucuronide metabolites and the assessment of their reactivity toward human serum albumin. Drug Metab. Dispos. 30, 404-413. (11) Castillo, M., and Smith, P. C. (1991) Covalent binding of Ibuprofen acyl glucuronide to human serum albumin in vitro. Pharmacol. Res. 10 (Suppl.), S242. (12) Dubois, N., Lapicque, F., Magdalou, J., Abiteboul, M., and Netter, P. (1994) Stereoselective binding of the glucuronide of ketoprofen enantiomers to human serum albumin. Biochem. Pharmacol. 48, 1693-1699. (13) Rachmel, A., Hazelton, G. A., Yergey, A. L., and Liberato, D. J. (1985) Furosemide 1-O-acyl glucuronide. In vitro biosynthesis and
1216
(14) (15)
(16) (17)
(18)
(19)
(20)
(21) (22)
(23) (24)
Chem. Res. Toxicol., Vol. 17, No. 9, 2004
pH-dependent isomerization to beta-glucuronidase-resistant forms. Drug Metab. Dispos. 13 (6), 705-710. Ojingwa, J. C., Spahn-Langguth, H., and Benet, L. Z. (1988) Studies of Tolmetin reversible and irreversible binding to proteins via its glucuronide conjugates. Pharmacol. Res. 5 (Suppl.), S214. Ojingwa, J. C., Spahn-Langguth, H., and Benet, L. Z. (1994) Irreversible binding of tolmetin to macromolecules via its glucuronide: Binding to blood constituents, tissue homogenates and subcellular fractions in vitro. Xenobiotica 24, 495-506. Tang, W. (2003) The metabolism of diclofenac enzymology and toxicology perspectives. Curr. Drug Metab. 4 (4), 319-329. Volland, C., Sun, H., Dammeyer, J., and Benet, L. Z. (1991) Stereoselective degradation of the fenoprofen acyl glucuronide enantiomers and irreversible binding to plasma protein. Drug Metab. Dispos. 19 (6), 1080-1086. Bailey, M. J., Worrall, S., de Jersey, J., and Dickinson, R. G. (1998) Zomepirac acyl glucuronide covalently modifies tubulin in vitro and in vivo and inhibits its assembly in an in vitro system. Chem.Biol. Interact. 115 (2), 153-166. Kretz-Rommel, A., and Boelsterli, U. A. (1993) Diclofenac covalent protein binding is dependent on acyl glucuronide formation and is inversely related to P450-mediated acute cell injury in cultured rat hepatocytes. Toxicol. Appl. Pharmacol. 120 (1), 155-161. Dickinson, R. G., and King, A. R. (1993) Studies on the reactivity of acyl glucuronides- -V. Glucuronide-derived covalent binding of diflunisal to bladder tissue of rats and its modulation by urinary pH and beta-glucuronidase. Biochem. Pharmacol. 46 (7), 11751182. Seitz, S., and Boelsterli, U. A. (1998) Diclofenac acyl glucuronide, a major biliary metabolite, is directly involved in small intestinal injury in rats. Gastroenterology 115 (6), 1476-1482. Boelsterli, U. A., Zimmerman, H. J., and Kretz-Rommel, A. (1995) Idiosyncratic liver toxicity of nonsteroidal antiinflammatory drugs: Molecular mechanisms and pathology. Crit. Rev. Toxicol. 25 (3), 207-235. Edsall, J. T., and Wyman, J. (1958) Biophysical Chemistry, Chapter 8, Academic Press, New York. Ding, A., Ojingwa, J. C., McDonagh, A. F., Burlingame, A. L., and Benet, L. Z. (1993) Evidence for covalent binding of acyl glucuronides to serum albumin via an imine mechanism as revealed by tandem mass spectrometry. Proc. Natl. Acad. Sci. U.S.A. 90 (9), 3797-3801.
Wang et al. (25) Ding, A., Zia-Amirhosseini, P., McDonagh, A. F., Burlingame, A. L., and Benet L. Z. (1995) Reactivity of tolmetin glucuronide with human serum albumin. Identification of binding sites and mechanisms of reaction by tandem mass spectrometry. Drug Metab. Dispos. 23 (3), 369-376. (26) Smith, P. C., McDonagh, A. F., and Benet, L. Z. (1986) Irreversible binding of zomepirac to plasma protein in vitro and in vivo. J. Clin. Invest. 77, 934-939. (27) Kretz-Rommel, A., and Boelsterli, U. A. (1994) Mechanism of covalent adduct formation of Diclofenac to rat hepatic microsomal proteins. Retention of the glucuronic acid moiety in the adduct. Drug Metab. Dispos. 22, 956-961. (28) Smith, P. C., Song, W. Q., and Rodriguez, R. J. (1992) Covalent binding of etodolac acyl glucuronide to albumin in vitro. Drug Metab. Dispos. 20 (6), 962-965. (29) Sidenius, U., Skonberg, C., Olsen, J., and Hansen, S. H. (2004) In vitro reactivity of carboxylic acid-CoA thioesters with glutathione. Chem. Res. Toxicol. 17 (1), 75-81. (30) Stogniew, M., and Fenselau, C. (1982) Electrophilic reactions of acyl-linked glucuronides. Formation of clofibrate mercapturate in humans. Drug Metab. Dispos. 10, 609-613. (31) Van Breeman, R. B., and Fenselau, C. (1985) Acylation of albumin by 1-O-acyl glucuronides. Drug Metab. Dispos. 13, 318-320. (32) Ruelius, H. W., Kirkman, S. K., Young, E. M., and Janssen, F. W. (1986) Reactions of oxaprozin-1-O-acyl glucuronide in solutions of human plasma and albumin. Adv. Exp. Med. Biol. 197, 431441. (33) Grillo, M. P., Knutson, C. G., Sanders, P. E., Waldon, D. J., Hua, F., and Ware, J. A. (2003) Studies on the chemical reactivity of diclofenac acyl glucuronide with glutathione: Identification of diclofenac-S-acyl-glutathione in rat. Drug Metab. Dispos. 31, 1327-1336. (34) Grillo, M. P., Hua, F., Knutson, C. G., Ware, J. A., and Li, C. (2003) Mechanistic studies on the bioactivation of diclofenac: Identification of diclofenac-S-acyl-glutathione in vitro in incubations with rat and human hepatocytes. Chem. Res. in Toxicol. 16 (11), 1410-1417. (35) Grillo, M. P., and Hua, F. (2003) Identification of zomepirac-Sacyl-glutathione in vitro in incubations with rat hepatocytes and in vivo in rat bile. Drug Metab. Dispos. 31 (11), 1429-1436.
TX049900+