Abstract
A series of fluoroquinolone compounds (compounds 1–9), which contain a common quinolone scaffold, inactivated the metabolic activity of CYP3A. The purpose of this study was to identify mechanism-based inhibition (MBI) among these fluoroquinolone compounds by metabolite profiling to elucidate the association of the substructure and MBI potential. Reversibility of MBI after incubation with potassium ferricyanide differed among the test compounds. Representative quasi-irreversible inhibitors form a metabolite-intermediate (MI) complex with the heme of CYP3A4 according to absorption analysis. Metabolite profiling identified the cyclopropane ring-opened metabolites from representative irreversible inhibitors, suggesting irreversible binding of the carbon-centered radical species with CYP3A4. On the other hand, the oxime form of representative quasi-irreversible inhibitors was identified, suggesting generation of a nitroso intermediate that could form the MI complex. Metabolites of compound 10 with a methyl group at the carbon atom at the root of the amine moiety of compound 8 include the oxime form, but compound 10 did not show quasi-irreversible inhibition. The docking study with CYP3A4 suggested that a methyl moiety introduced at the carbon atom at the root of the primary amine disrupts formation of the MI complex between the heme and the nitroso intermediate because of steric hindrance. This study identified substructures of fluoroquinolone compounds associated with the MBI mechanism; introduction of substituted groups inducing steric hindrance with the heme of P450 can prevent formation of an MI complex. Our series of experiments may be broadly applicable to prevention of MBI at the drug discovery stage.
Introduction
Some drug–drug interactions (DDIs) involve time-dependent inhibition (TDI) of drug-metabolizing enzymes in which a reactive intermediate generated by a metabolic enzyme interacts with the enzyme quasi-irreversibly or irreversibly, thereby inactivating the enzyme. The TDI phenomenon, which involves enzymatic activity loss induced by incubating enzymes with inhibitors before the addition of substrates, is used in kinetic experiments. Mechanism-based inhibition (MBI) refers to a subset of TDI focusing on a chemical mechanism in which reactive intermediates lead to the inactivation of enzymes. A prior study pointed out that the distinction between TDI and MBI must be appreciated (Grimm et al., 2009). In this report, we focus on the mechanism of enzyme inactivation caused by reactive intermediates, and we thus use the term MBI. Cytochrome P450 (P450) is an enzyme that is responsible for the metabolism of many drugs in human (Guengerich, 2001). Thus, MBI of P450 may cause a clinically severe DDI because the enzymatic activity is recovered only by synthesis of a new enzyme; thus, the inhibition continues even after the inhibitor is eliminated from the body. Indeed, many drugs have been withdrawn from the market because of P450-related DDIs (Wienkers and Heath, 2005). Pharmaceutical companies have attempted to develop in vitro test systems to attenuate the MBI potential of drug candidates for P450 at the early stage of drug discovery (Watanabe et al., 2007; Grime et al., 2009; Grimm et al., 2009; Zientek et al., 2010; Yates et al., 2012).
MBI of P450 involves the irreversible or quasi-irreversible binding of a reactive intermediate metabolite to the metabolizing enzyme (Lin and Lu, 1998). The irreversible inhibition is caused by covalent binding of reactive intermediates to the heme or apoprotein of the active site of P450, whereas the quasi-irreversible inhibition is caused by formation of a stable metabolite-intermediate (MI) complex with the ferrous form of the heme iron atom (Ullrich and Schnabel, 1973). A large number of compounds including methylenedioxybenzenes, alkylamines, and hydrazines have been reported to form MI complexes (Murray, 1997; Lin and Lu, 1998; Orr et al., 2012). The MI complexes can dissociate after treatment with potassium ferricyanide, which oxidizes iron to the ferric form and recovers the enzymatic activity (Buening and Franklin, 1976; Muakkassah et al., 1982); on the basis of that mechanism, we established an assay to distinguish between irreversible and quasi-irreversible inhibition in our previous report (Watanabe et al., 2007). Furthermore, incubation with potassium ferricyanide has also been applied to identification of metabolites that form an MI complex. A nitroso intermediate generated from lapatinib, which can cause quasi-irreversible inhibition of CYP3A4, increased in abundance after the addition of potassium ferricyanide to the reaction solution of recombinant CYP3A4 and lapatinib (Barbara et al., 2013). Identification of the MBI mechanism will also be useful for analysis of the risk of drug-induced toxicities caused by covalent binding of reactive intermediates to proteins and lipids, in addition to DDIs (Zhou et al., 2004; Fontana et al., 2005; Kalgutkar et al., 2005; Walgren et al., 2005; Takakusa et al., 2011).
We developed novel fluoroquinolone antibacterial agents, compound 1 (also known as DX-619 or 7-[(3R)-3-(1-aminocyclopropyl)pyrrolidin-1-yl]-1-[(1R,2S)-2-fluorocyclopropyl]-8-methoxy-4-oxoquinoline-3-carboxylic acid) and compound 8 (also known as DK-507k or 7-[(7S)-7-amino-5-azaspiro[2.4]heptan-5-yl]-6-fluoro-1-[(1R,2S)-2-fluorocyclopropyl]-8-methoxy-4-oxoquinoline-3-carboxylic acid), at the clinical stage (Otani et al., 2003; Fujikawa et al., 2005). These compounds show significant MBI potential toward CYP3A (Imamura et al., 2013; Odagiri et al., 2013). Indeed, on phase I samples, it was shown that DX-619 causes a significant reduction in apparent 6β-hydroxycortisol formation clearance, an index of CYP3A4 activity in the liver, during DX-619 administration (Imamura et al., 2013). Because CYP3A is responsible for most of the P450-mediated drug metabolism (Wienkers and Heath, 2005), DX-619 will cause moderate DDIs with various CYP3A substrate drugs. Compounds 1 and 8 have the same scaffold except for substructures of positions C6 and C7 in the quinolone ring (Table 1). Compound 10 (also known as DC-159a [(+)-7-[(7S)-7-amino-7-methyl-5-azaspiro[2.4]heptan-5-yl]-6-fluoro-1-[(1R,2S)-2-fluoro-1-cyclopropyl]-1,4-dihydro-8-methoxy-4-oxo-3-quinolinecarboxylic acid hemihydrate]), in which a methyl group was introduced at the carbon atom at the root of the C7-amino moiety of compound 8, was designed for prevention of MBI and actually turned out to be MBI negative, sustaining the pharmacological activity (Hoshino et al., 2008; Odagiri et al., 2013). This study aimed to clarify the MBI mechanism caused by our fluoroquinolones and to then investigate the effect of the methyl group introduced into compound 10 on prevention of MBI. In addition, we analyzed a series of fluoroquinolone antibacterial candidates containing a common quinolone scaffold (Table 1) to determine the possible association of the substructure with MBI potential.
Materials and Methods
Materials.
All of the tested fluoroquinolone compounds shown in Table 1 were synthesized by Daiichi Sankyo Co., Ltd. (Tokyo, Japan). The synthesis method of compound 10 as a representative fluoroquinolone compound was described in a previous report (Odagiri et al., 2013). Midazolam maleate salt was purchased from Sigma-Aldrich (St. Louis, MO). 1′-Hydroxymidazolam was purchased from Ultrafine (Manchester, UK); 0.5 M potassium phosphate buffer, [13C3]hydroxymidazolam, NADPH Regenerating System Solution A and B, and recombinant human CYP3A4 Supersomes containing P450 reductase and cytochrome b5 were acquired from Corning (Woburn, MA). Potassium ferricyanide was purchased from Kanto Chemical Co. Inc. (Tokyo, Japan). Fifty mixed-sex, donor-pooled human liver microsomes (HLMs) were acquired from XenoTech LLC (Lenexa, KS). All other reagents and solvents were of the highest grade commercially available.
Screening for Reversibility of MBI of CYP3A Using HLMs.
The screening was performed as described previously (Watanabe et al., 2007).
Reversibility of MBI Using Recombinant Human CYP3A4 Supersomes.
A total of 270 μl preincubation solution contained 10 pmol/ml recombinant CYP3A4 Supersomes in 0.1 M potassium phosphate buffer with or without test compounds (final concentration of 10, 30, or 100 μM). The final solvent concentration in the preincubation solutions was 1% (v/v) dimethylsulfoxide. Preincubation reactions were initiated by the addition of 30 μl of an NADPH-generating system consisting of NADPH Regenerating System Solutions A and B. After a 0-minute or 30-minute preincubation, 50 μl of each preincubation solution was added to 50 μl of the solutions containing 0.1 M sodium phosphate buffer with or without 2 mM potassium ferricyanide and was then incubated for 10 minutes. After the 10-minute reaction, each reaction mixture was diluted 5-fold with incubation solution containing 0.1 M potassium phosphate buffer and 25 µM midazolam as a substrate [final concentration 1% (v/v) acetonitrile]. At the end of the 10-minute incubation reactions, 100-μl aliquots of each incubation solution were added to the mixture of 50 μl acetonitrile containing 2 μM [13C3]hydroxymidazolam as an internal standard and 100 μl methanol. The samples were centrifuged at 2000g for 3 minutes, and the supernatants were transferred to other plates. A standard curve of 1′-hydroxymidazolam was constructed to determine its concentration in the samples. The concentration range of the standard curve was 8–1000 nM. All samples were analyzed by liquid chromatography (LC)–tandem mass spectrometry (MS/MS) using a Waters Acquity UPLC and TQD system (Waters, Manchester, UK). Chromatographic separation was performed on a SunFire C18 column (3.5-μm particle size, 2.1 × 100 mm; Waters). The mass spectrometer was operated in positive electrospray ionization mode. The mass-to-charge ratio (m/z) (precursor → product) values of 1′-hydroxymidazolam and [13C3]hydroxymidazolam were 342 → 324 and 345 → 327, respectively. Concentrations of 1′-hydroxymidazolam in the samples were calculated using MassLynx software (version 4.1; Waters). The percentage of metabolic activity [% of control(0 minutes) and % of control(30 minutes)] was obtained as follows and is detailed in a previous report (Watanabe et al., 2007).v(0 minutes, ±inhibitor) and v(30 minutes, ±inhibitor) indicate the metabolic activity after 0-minute or 30-minute preincubation with (+) or without (−) an inhibitor, respectively.
Determination of KI and kinact values.
A total of 180 μl preincubation solution contained 10 pmol/ml recombinant CYP3A4 Supersomes in 0.1 M potassium phosphate buffer with or without test compounds (five or six concentrations per test compound). The preincubation reactions were initiated by the addition of 20 μl of an NADPH-generating system. After a 0-, 2-, or 5-minute preincubation for compound 1 and 0-, 5-, or 15-minute preincubation for compound 6, 20 μl of each preincubation mixture was added to 180 μl incubation solution containing 0.1 M potassium phosphate buffer and 25 µM midazolam. After the 10-minute incubation reactions, samples were prepared for LC-MS/MS analysis and concentrations of 1′-hydroxymidazolam in the samples and percentages of control values for each preincubation time with each inhibitor concentration were determined as described above. The natural logarithm of the percentage of the control was plotted against the preincubation times for each concentration of a test compound. The slope from the linear regression analysis provided the observed inactivation rate constant (kobs) for each concentration; kobs and the inhibitor concentration (I) were fitted into the following expression by using Phoenix WinNonlin 6.1 software (Certara G.K., Princeton, NJ).
Absorption Analysis for MI Complex Formation.
A total of 180 μl reaction mixture containing 100 pmol/ml recombinant human CYP3A4 Supersomes, 50 μM test compound [final concentration 1% (v/v) dimethylsulfoxide], and 0.1 M potassium phosphate buffer was transferred to the microplate well and preincubated at 37°C for 5 minutes. For the reference well, the solvent was added to the reaction mixture in place of the test compound. The reaction was initialized by adding 20 μl of the NADPH-generating system to the reaction mixture. Absorbance at 455 nm and 490 nm was monitored for 20 minutes at 37°C using a VersaMax microplate reader (Molecular Devices, Sunnyvale, CA) controlled by SoftMax Pro software (version 5.4.6; Molecular Devices). Using the absorbance data at 455 nm and 490 nm, the absorbance difference between 455 nm and 490 nm was calculated.
Structural Elucidation of Metabolites of Test Compounds after Incubation with Recombinant Human CYP3A4 Supersomes.
Five microliters of a test compound solution (5 mM in dimethylsulfoxide) was added to 445 μl incubation solution consisting of 100 pmol/ml recombinant human CYP3A4 and 0.1 M potassium phosphate buffer. For the 0-minute incubation sample, 500 μl acetonitrile was added to the incubation mixture, followed by the addition of 50 μl of the NADPH-generating system. For the 30-minute incubation sample, 50 μl of the NADPH-generating system was added to the incubation mixture, and the mixture was then incubated at 37°C for 30 minutes. Afterward, the reaction was terminated by the addition of 500 μl acetonitrile. Incubation samples were centrifuged at 9000g for 3 minutes, and the supernatants were concentrated using a centrifugal evaporator. The concentrated samples were analyzed by means of a LTQ-Orbitrap XL (Thermo Fisher Scientific, San Jose, CA) equipped with an Acquity UPLC PDA system (Waters). Solvents A and B were based on H2O containing 0.1% (v/v) trifluoroacetic acid and acetonitrile (LC–mass spectrometry grade) with 0.1% (v/v) trifluoroacetic acid, respectively. In hydrogen-deuterium exchange experiments, D2O with 0.1% (v/v) trifluoroacetic acid was used as solvent A. Analyte separation was achieved using an Acquity UPLC BEH C18 column (100 × 2.1 mm, 1.7-μm particle size) at a flow rate of 0.5 ml/min under a linear gradient from 10% to 70% B (0–8 minutes). The data were acquired in full-scan and MS/MS modes. Mass spectrometry conditions were as follows: electrospray voltage, 4.2 kV; capillary temperature, 275°C; sheath gas flow rate, 30 arbitrary units; auxiliary gas flow rate, 15 arbitrary units; resolution, 30,000 for full-scan mode and 7500 for MS/MS mode; ionization, electrospray ionization in positive ion mode; activation type for MS/MS, higher-energy collision dissociation; and collision energy for MS/MS, 40% and 50%.
CYP3A4 Docking of Compound 6 and Its Nitroso Metabolite.
Molecular docking of compound 6 and its nitroso metabolite to CYP3A4 was performed using the crystal structures of CYP3A4 [Protein Data Bank (PDB) code 2V0M (Ekroos and Sjögren, 2006) for compound 6 and PDB code 4I4G (Sevrioukova and Poulos, 2013) for the nitroso metabolite of compound 6]. The substrates were prepared using the LigPrep module of Maestro (version 3.5; Schrödinger, LLC, New York, NY). Starting with two-dimensional structures, LigPrep produces a three-dimensional structure with ionization states at pH 7. In compound 6, a formal charge of the amine group, which is an iron-coordinating moiety, was manually modified to 0. Docking was carried out using the Glide docking module of Maestro (version 6.8; Schrödinger, LLC) with the positional constraint on the reference crystal structures [Cambridge Structural Database code CICNEH (Munro et al., 1999) for compound 6 and PDB code 4M4A (Yi et al., 2013) for the nitroso metabolite of compound 6] to keep typical iron-coordinate geometry, because Glide did not recognize nitrogen atoms of the -NH2 or nitroso moiety as iron coordinators. After Glide docking, methylated compounds were manually modeled by adding a methyl group to the representative docking models of each compound.
Results
Reversibility of MBI for CYP3A.
To distinguish between quasi-irreversible and irreversible binding to CYP3A by fluoroquinolones, the screening of MBI reversibility was performed in HLMs using these compounds (Supplemental Table 1). The activity of 1′-hydroxymidazolam formation from midazolam by HLMs was monitored as the CYP3A activity. The enzymatic activity of CYP3A inactivated after 30-minute preincubation with the compound containing cyclopropylamine (compounds 1–5) in the pyrrolidine ring of the 7′ position of fluoroquinolone did not recover after oxidation with potassium ferricyanide. These compounds were assumed to be irreversible inhibitors. In contrast, enzymatic activity that was reduced after 30-minute preincubation with the compound containing an amine moiety (compounds 6–9) in the ring form of the 7′ position of fluoroquinolone tended to recover after oxidation with potassium ferricyanide. It was shown that these compounds could bind to CYP3A quasi-irreversibly. Unlike compounds 1–9, the decrease in enzymatic activity after 30-minute preincubation with compound 10 was very low.
The assay of MBI reversibility was also performed in recombinant human CYP3A4 Supersomes by using these compounds. Enzymatic activities reduced by 30-minute preincubation with each of compounds 6–9 were restored more than 20% with the addition of potassium ferricyanide and were close to the activity after 0-minute preincubation; however, those with each of compounds 1–5 were not fully restored, although the percentage of control data of compounds 3–5 after 30-minute preincubation followed by incubation with potassium ferricyanide was statistically higher than that without potassium ferricyanide (Fig. 1). This study clearly distinguished between compounds 1–5 as irreversible inhibitors and compounds 6–9 as quasi-irreversible inhibitors. CYP3A4 was shown to be one of the CYP3A isozymes responsible for MBI observed in HLMs. On the basis of this result, recombinant human CYP3A4 Supersomes were used in the following experiments.
KI and kinact values of compounds 1 and 6 as representative irreversible and quasi-irreversible inhibitors, respectively, were obtained for recombinant CYP3A4 Supersomes (Supplemental Fig. 1; Supplemental Table 2). These two compounds showed different characteristics indicated as kinetic parameters, with higher KI (395 ± 47 μM) and higher kinact (0.459 ± 0.024 min−1) for compound 1 and lower KI (7.21 ± 1.38 μM) and lower kinact (0.190 ± 0.009 min−1) for compound 6.
Absorption Analysis for MI Complex Formation.
To determine whether inhibition of CYP3A4 enzymatic activity by the fluoroquinolone compounds assumed to be quasi-irreversible inhibitors occurs via the formation of an MI complex, absorbance at 455 nm and 490 nm was monitored for 20 minutes after the addition of the NADPH-generating system in the reaction mixture containing recombinant CYP3A4 Supersomes and each test compound. The absorbance difference between 455 nm and 490 nm was plotted against time (Fig. 2). Compounds 6 and 8 were used as representative quasi-irreversible inhibitors of CYP3A4, and compound 10 was used as a noninhibitor or weak inhibitor. The absorbance difference increased in a time-dependent manner and reached a plateau after approximately 15 minutes for incubation with compound 6 or 8, but the increase was not observed with compound 10.
Structural Elucidation of Metabolites of Irreversible Inhibitors after Incubation with Recombinant CYP3A4 Supersomes.
To gain mechanistic understanding of the irreversible inhibition of fluoroquinolone compounds, we performed a structural analysis by LC-MS/MS of metabolites after incubation of compound 1, a representative irreversible inhibitor, with recombinant CYP3A4 Supersomes. Using full-scan conditions, four metabolites related to oxidation reactions of the cyclopropylamine moiety of compound 1 [cpd1-M1 [retention time (Rt) = 2.91 minutes, m/z 418.1772], cpd1-M2 (Rt = 3.58 minutes, m/z 418.1771), cpd1-M3 (Rt = 4.41 minutes, m/z 419.1611), and cpd1-M4 (Rt = 5.26 minutes, m/z 416.1615)] were detected as shown in the mass chromatograms (Fig. 3A). The mass spectrometric data are summarized in Supplemental Table 3. The molecular composition was estimated to be C21H24N3O5F for cpd1-M1 and cpd1-M2, C21H23N2O6F for cpd1-M3, and C21H22N3O5F for cpd1-M4 by accurate mass measurements. Hydrogen-deuterium exchange measurements using D2O as an eluent revealed the molecular ions [M + D]+ at m/z 423.2079 (cpd1-M1),m/z 423.2065 (cpd1-M2), m/z 422.1797 (cpd1-M3), and m/z 418.1734 (cpd1-M4), indicating that the numbers of exchangeable protons of the metabolites were 4, 4, 2, and 1, respectively. The proposed structure of each metabolite based on the MS/MS fragmentation patterns, molecular compositions, and number of exchangeable protons are shown in Fig. 3B. cpd1-M1 and cpd1-M2 were found to be oxidative (+O) metabolites according to the estimated molecular composition. The product ions (m/z 305, m/z 287, and m/z 229), which correspond to the quinolone moiety of cpd1-M1 and cpd1-M2, were identical to those of the unchanged form, pointing to the monooxygenation in the pyrrolidinyl cyclopropylamine side chain. In addition, the increase (+1) of the exchangeable protons in cpd1-M1 and cpd1-M2 indicated hydroxylation on carbon atoms, not N-oxidation or hydroxylamine formation. cpd1-M3 was proposed to be the hydroxyethyl carbonyl form in the side chain on the basis of its molecular compositional change (+2O-N-H), fragmentation pattern, and number of exchangeable protons (n = 2). cpd1-M4 was proposed to be the dihydroisoxazole form in the side chain based on its molecular compositional change (+O-2H), fragmentation pattern, and number of exchangeable protons (n = 1).
Structural Elucidation of Metabolites of Quasi-Irreversible Inhibitors after Incubation with Recombinant CYP3A4 Supersomes.
To gain mechanistic understanding of quasi-irreversible inhibition of fluoroquinolone compounds, we carried out a structural analysis by LC-MS/MS of metabolites after incubation of compound 6, a representative quasi-irreversible inhibitor, with recombinant CYP3A4 Supersomes. The full-scan analysis detected five metabolites related to oxidation reactions of the amino azaspiro[4.4]nonan moiety of compound 6 [cpd6-M1 (Rt = 3.45 minutes, m/z 450.1834), cpd6-M2 (Rt = 3.88 minutes, m/z 450.1833), cpd6-M3 (Rt = 6.50 minutes, m/z 448.1677), cpd6-M4 (Rt = 6.50 minutes, m/z 435.1729), and cpd6-M5 (Rt = 7.16 minutes, m/z 433.1569)], which were detected as shown in the mass chromatograms (Fig. 4A). The mass spectrometric data are summarized in Supplemental Table 3. The molecular composition was estimated to be C22H25N3O5F2 for cpd6-M1 and cpd6-M2, C22H23N3O5F2 for cpd6-M3, C22H24N2O5F2 for cpd6-M4, and C22H22N2O5F2 for cpd6-M5 by accurate mass measurements. Hydrogen-deuterium exchange measurements using D2O as an eluent revealed the molecular ions [M + D]+ at m/z 455.2144 (cpd6-M1), m/z 455.2142 (cpd6-M2), m/z 451.1863 (cpd6-M3), m/z 438.1912 (cpd6-M4), and m/z 435.1690 (cpd6-M5), indicating that the numbers of exchangeable protons of the metabolites were 4, 4, 2, 2, and 1, respectively. The proposed structure of each metabolite based on the MS/MS fragmentation patterns, molecular compositions, and numbers of exchangeable protons are shown in Fig. 4B. cpd6-M1 and cpd6-M2 were found to be oxidative (+O) metabolites according to the estimated molecular composition. The product ion (m/z 279), which corresponds to the quinolone moiety of cpd6-M1 and cpd6-M2, was identical to that of the unchanged form, pointing to monooxygenation in the amino azaspiro[4.4]nonan side chain. In addition, the increase (+1) of the exchangeable protons in cpd6-M1 and cpd6-M2 indicated hydroxylation on carbon atoms, not N-oxidation or hydroxylamine formation. cpd6-M3 was found to be the oxime form in the side chain on the basis of its molecular compositional change (+O-2H), fragmentation pattern, and number of exchangeable protons (n = 2). cpd6-M4 and cpd6-M5 were proposed to be the hydroxyl form and keto form via oxidative deamination in the side chain based their molecular compositional changes +O-N-H and +O-N-3H, respectively. These proposed metabolite structures were consistent with the fragmentation patterns and the number of exchangeable protons of cpd6-M4 and cpd6-M5.
In addition, we analyzed metabolites after incubation of compound 10, which has the methyl moiety in the side chain but does not cause quasi-irreversible inhibition for CYP3A4, with recombinant CYP3A4 Supersomes. The full-scan analysis detected four metabolites related to oxidation reactions of the amino azaspiro[2.4]heptan moiety of compound 10 [cpd10-M1 (Rt = 3.97 minutes, m/z 434.1514), cpd10-M2 (Rt = 4.09 minutes, m/z 436.1674), cpd10-M3 (Rt = 5.86 minutes, m/z 466.1415), and cpd10-M4 (Rt = 6.88 minutes, m/z 450.1470)], which were detected as shown in mass chromatograms (Fig. 5A). The mass spectrometric data are summarized in Supplemental Table 3. The molecular compositions were estimated to be C21H22N3O5F2 for cpd10-M1, C21H24N3O5F2 for cpd10-M2, C21H22N3O7F2 for cpd10-M3, and C21H22N3O6F2 for cpd10-M4 by accurate mass measurements. Hydrogen-deuterium exchange measurements using D2O as an eluent revealed the molecular ions [M + D]+ at m/z 436.1653 (cpd10-M1), m/z 440.1935 (cpd10-M2), m/z 469.1617 (cpd10-M3), and m/z 452.1601 (cpd10-M4), indicating that the numbers of exchangeable protons of the metabolites were 1, 3, 2, and 1, respectively. The proposed structure of each metabolite according to the MS/MS fragmentation patterns, molecular compositions, and numbers of exchangeable protons is shown in Fig. 5B. cpd10-M1, cpd10-M2, and cpd10-M4 were found to be the nitroso form, hydroxylamine form, and nitro form in the side chain according to its molecular compositional changes +O-2H, +O, and +2O-2H, respectively. These proposed metabolite structures were consistent with the fragmentation patterns and the number of exchangeable protons. On the basis of its molecular compositional change (+2O-2H), cpd10-M3 was found to be an oxidative metabolite of the amino azaspiro[2.4]heptan side chain of the nitro form (cpd10-M4). It was shown that compound 10 is also metabolized near the amino moiety in the side chain by CYP3A4.
CYP3A4 Docking of Compound 6 and Its Nitroso Metabolite.
To study the effect of a methyl group introduced at the carbon atom at the root of the C7-amino moiety of the quasi-irreversible inhibitors for binding to CYP3A4, molecular docking of compound 6 as a representative quasi-irreversible inhibitor and its nitroso metabolite to CYP3A4 was performed using the crystal structure of CYP3A4 (PDB code 2V0M and 4I4G for each; Ekroos and Sjögren, 2006; Sevrioukova and Poulos, 2013) as shown in Figs. 6 and 7. We obtained the possible docking mode of compound 6 and its nitroso metabolite against CYP3A4, which maintain typical iron-coordinate geometry, as shown in Fig. 6A and Supplemental Fig. 5, and Fig. 7A and Supplemental Fig. 6, respectively. The distance between amine nitrogen and the heme iron was 2.25 Å for compound 6, and the angle θ1 (C-N-Fe) was 110.0° (Fig. 6A). The distance between amine nitrogen and the heme iron was 2.35 Å for the nitroso metabolite, and the angle θ2 (C-N-Fe) was 130.0° (Fig. 7A). It was assumed that these docking models were plausible especially around the heme iron and these structures were used for further analysis. Methyl moieties manually added to these docking models indicated that there would be a severe steric clash between added methyls and hemes as shown in Figs. 6B and 7B.
Discussion
In this study, we focused on the effect of substructures of a series of fluoroquinolone antibacterial compounds with a common scaffold on the MBI potential toward CYP3A. To obtain information on the binding mechanism of the fluoroquinolone compounds, we performed MBI reversibility screening using HLMs established in our previous study (Watanabe et al., 2007) and we examined the reversibility using recombinant human CYP3A4 Supersomes. The fluoroquinolone compounds (compounds 1–9) were classified into irreversible and quasi-irreversible inhibitors (compounds 1–5 and compounds 6–9, respectively). The classification appears to depend on substructures of the 7′ position of the quinolone ring, the pyrrolidine ring bearing cyclopropylamine (compounds 1–5), and the pyrrolidine or azaspiro ring bearing a primary amine (compounds 6–9). Compound 10, which contains a methyl group at the carbon atom at the root of the primary amine of compound 8, did not show the inhibitory effect in both matrices. MBI could be reproduced in recombinant human CYP3A4 Supersomes. In the subsequent experiments clarifying the detailed MBI mechanism of these compounds, we used CYP3A4 Supersomes.
Absorption analysis of the reaction mixtures of recombinant CYP3A4 Supersomes with these quasi-irreversible inhibitors was performed to confirm MI complex formation. This assay is based on previous reports showing that the absorbance difference between 455 and 490 nm increases in a time-dependent manner and reaches a plateau after incubation of P450 with compounds forming an MI complex (Ullrich and Schnabel, 1973; Buening and Franklin, 1976; Franklin, 1991). The absorbance difference increased with incubation time in the presence of compound 6 or 8 but not compound 10 (Fig. 2). This result indicates that compounds 6 and 8 cause quasi-irreversible inhibition via MI complex formation with CYP3A4, and compound 10 does not. The order of the absorbance difference was consistent with that of CYP3A4 inhibition after 30-minute preincubation with these compounds (Fig. 1).
The structural analysis of metabolites after incubation of irreversible and quasi-irreversible inhibitors with recombinant human CYP3A4 Supersomes was performed to study the mechanistic difference between irreversible and quasi-irreversible inhibition by the fluoroquinolone compounds from the viewpoint of drug metabolism. The proposed metabolic pathways for compounds 1 and 6 are shown in Figs. 8 and 9, respectively. In the case of compound 1, cpd1-M3 was found to be the hydroxyethyl carbonyl form, which indicates that compound 1 undergoes ring-opening metabolism. In line with the existing literature on the metabolism of cyclopropylamines (Cerny and Hanzlik, 2005, 2006), cpd1-M3 formation can be explained by the hydrogen abstraction from the primary amine group of compound 1, followed by ring opening of the aminyl radical to form carbon-centered radical species, hydroxylation, and hydrolysis of the imine into a keto form. The radical intermediates are likely to irreversibly modify CYP3A4 and to cause irreversible inhibition. A precursor of cpd1-M3, the hydroxylated imine intermediate, was presumed to be oxidized into the dihydroisoxazole form (cpd1-M4). The same ring-opening metabolites of cyclopropylamines were also detected after incubation of the other irreversible inhibitors, compound 2 or 5, with recombinant CYP3A4 (Supplemental Figs. 2 and 3).
In contrast with irreversible inhibitors, ring-opened metabolites were not detected with compound 6. The key metabolite of compound 6 was the oxime form, cpd6-M3, which suggests the formation of a nitroso intermediate. This is because alkylnitroso intermediates are known to be generally unstable and to tautomerize to more stable oxime forms (Mansuy et al., 1977). The primary amine group in the side chain of compound 6 is likely to be oxidized to form the hydroxylamine, followed by formation of the nitroso intermediate, which appears to form an MI complex with the heme of CYP3A4, and consequently to cause quasi-irreversible inhibition. In the case of compound 8, we detected the oxime and nitro form in the side chain, which are assumed to be formed via the nitroso intermediate (Supplemental Fig. 4).
Metabolite profiling suggests that oxidation of the primary amine in the side chain is the initial step of MBI for both irreversible and quasi-irreversible inhibitors; however, the subsequent ring-opening radical reaction was observed only for the irreversible inhibitors. Because the three-membered ring has higher distortion energy than the five-membered ring does, the aminyl radical of the cyclopropane is presumed to be less stable than that of cyclopentane. Therefore, the ring structure bearing a primary amine group in quinolones seems to account for the difference in the MBI mechanism between irreversible and quasi-irreversible binding with CYP3A4.
To obtain direct evidence that the nitroso intermediate of compound 6 binds to the heme of CYP3A4 quasi-irreversibly, we compared the abundance of the metabolites of compound 6 in the reaction mixture of CYP3A4 with or without the treatment with potassium ferricyanide, according to one report (Barbara et al., 2013); however, the difference in the abundance of the metabolites could not be detected in our study (data not shown). We assumed that the abundance of the metabolite binding to proteins nonspecifically could not be negligible compared with that binding to CYP3A4 heme quasi-irreversibly.
We hypothesized that compound 10 (showing no inhibitory effect on CYP3A4) was not metabolized to the reactive intermediate because of the steric hindrance of the methyl group introduced at the carbon atom at the root of the primary amine group. Unexpectedly, the nitroso form of compound 10 (cpd10-M1) and the related metabolites generated via the nitroso form—the nitro form (cpd10-M4) and the oxidized form of the nitro form (cpd10-M3)—were also detected after incubation of compound 10 with recombinant CYP3A4 Supersomes (Fig. 5). The level of the peak area of the metabolites generated from compound 10 in the mass chromatogram was comparable with that of compound 6 (data not shown), suggesting that the amount of the reactive intermediate may be sufficient to inhibit CYP3A4. We speculated that the methyl group may hinder formation of an MI complex with the heme of CYP3A4.
To test this hypothesis, we conducted molecular docking studies of compound 6 as a representative quasi-irreversible inhibitor and its nitroso metabolite to CYP3A4. The plausible docking models of compound 6 and the nitroso metabolite against CYP3A4 were obtained (Figs. 6A and 7A), and the methyl moieties were then manually added to these docking models (Figs. 6B and 7B). It was demonstrated that there would be a severe steric clash between the added methyl group and heme in both docking models. Therefore, the amine nitrogen and heme iron cannot keep the optimal distance and the angle (C-N-Fe) to form an MI complex. A previous study reported that the potential energy of coordinate binding between the heme iron of myoglobin and nitric oxide depends on the distance (Negrerie et al., 2006). The potential energy reached the maximum at a distance of approximately 1.8 Å and then decreased dramatically according to dissociation between the heme iron and nitric oxide. If we apply this scheme to the binding between the heme iron of CYP3A4 and the nitroso intermediate, then the potential energy of binding between the heme and the nitroso intermediate may be decreased when the distance is extended by introduction of the methyl moiety, preventing formation of the MI complex.
The methyl moiety–conjugated compound, such as compound 10, is still recognized and oxidized by CYP3A4 regardless of the steric hindrance of the methyl moiety according to the docking experiment. It is possible that electron transfer from the heme to drugs tolerates extension of the distance between the heme and drugs.
In conclusion, we elucidated the key structures that are responsible for MBI of the fluoroquinolone antibacterial compounds, and the difference in the ring structure bearing a primary amine group in quinolones accounts for the difference between irreversible and quasi-irreversible inhibition of CYP3A4. Moreover, we assume that quasi-irreversible inhibition via MI complex formation can be avoided after introduction of substituted groups causing steric hindrance with the heme of P450. Our study may be widely applicable to the discovery of drugs free of the DDI risk via MBI.
Authorship Contributions
Participated in research design: Watanabe, Kusuhara.
Conducted experiments: Watanabe, Takakusa, Kimura.
Performed data analysis: Watanabe, Takakusa, Kimura.
Wrote or contributed to the writing of the manuscript: Watanabe, Takakusa, Kimura, Inoue, Kusuhara, Ando.
Footnotes
- Received May 19, 2016.
- Accepted July 27, 2016.
↵This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- DC-159a
- (+)-7-[(7S)-7-amino-7-methyl-5-azaspiro[2.4]heptan-5-yl]-6-fluoro-1-[(1R,2S)-2-fluoro-1-cyclopropyl]-1,4-dihydro-8-methoxy-4-oxo-3-quinolinecarboxylic acid hemihydrate
- DDI
- drug–drug interaction
- DK-507k
- 7-[(7S)-7-amino-5-azaspiro[2.4]heptan-5-yl]-6-fluoro-1-[(1R,2S)-2-fluorocyclopropyl]-8-methoxy-4-oxoquinoline-3-carboxylic acid
- DX-619
- 7-[(3R)-3-(1-aminocyclopropyl)pyrrolidin-1-yl]-1-[(1R,2S)-2-fluorocyclopropyl]-8-methoxy-4-oxoquinoline-3-carboxylic acid
- HLM
- human liver microsome
- LC
- liquid chromatography
- m/z
- mass-to-charge ratio
- MBI
- mechanism-based inhibition
- MI
- metabolite-intermediate
- MS/MS
- tandem mass spectrometry
- P450
- cytochrome P450
- PDB
- Protein Data Bank
- Rt
- retention time
- TDI
- time-dependent inhibition
- Copyright © 2016 by The American Society for Pharmacology and Experimental Therapeutics