Abstract
Clinical outcomes of vincristine therapy, both neurotoxicity and efficacy, are unpredictable, and the reported pharmacokinetics of vincristine have considerable interindividual variability. In vitro and in vivo data support a dominant role for CYP3A enzymes in the elimination of vincristine. Consequently, genetic polymorphisms in cytochrome P450 (P450) expression may contribute to the interindividual variability in clinical response, but the contributions of individual P450s and the primary pathways of vincristine metabolism have not been defined. In the present study, vincristine was incubated with a library of cDNA-expressed P450s, and the major oxidative metabolites were identified. CYP3A4 and CYP3A5 were the only P450s to support substantial loss of parent drug and formation of the previously unidentified, major metabolite (M1). The structure of M1, arising as a result of an oxidative cleavage of the piperidine ring of the dihydro-hydroxycatharanthine unit of vincristine, was conclusively established after conversion to suitable derivatives followed by spectroscopic analysis, and a new pathway for vincristine metabolism is proposed. CYP3A5 was more efficient in catalyzing the formation of M1 compared with CYP3A4 (9- to 14-fold higher intrinsic clearance for CYP3A5). The formation of M1 was stimulated (3-fold) by the presence of coexpressed cytochrome b5, but the relative efficiencies of M1 formation by CYP3A4 and CYP3A5 were unaffected. Our findings demonstrate that in contrast to most CYP3A biotransformations, the oxidation of vincristine is considerably more efficient with CYP3A5 than with CYP3A4. We conclude that common genetic polymorphisms in CYP3A5 expression may contribute to the interindividual variability in the systemic elimination of vincristine.
The treatment of pediatric acute lymphoblastic leukemia (ALL) includes vincristine as one of the core drugs in induction. For these patients, high interindividual variability in efficacy and toxicity with significant differences in outcomes between ethnic groups have been reported. In one ALL study, controlling for compliance, African-American children had a 42% increased mortality rate compared with Caucasian children (Pollock et al., 2000). Event-free survival was only 54% for African-American children compared with a Caucasian event-free survival of 82% in another ALL study (Lange et al., 2002). Unpredictable but treatment-limiting neurotoxicity has resulted in the arbitrary use of a maximum dose of 2 mg despite known differences in drug exposure among patients (McCune and Lindley, 1997). Early vincristine clinical studies showed up to 11-fold variability in drug exposure (dose-corrected area under the curve) among adult patients (Van den Berg et al., 1982). More recent pediatric clinical pharmacokinetic studies also reported a 19-fold difference in the dose-corrected area under the curve (Frost et al., 2003). Understanding determinants of the variation in vincristine exposure may allow identification of patient-specific risk factors for neurotoxicity or individualized dosing strategies to decrease the risk of patient relapse.
Evidence from in vitro and in vivo studies supports a determinant role for CYP3A-mediated metabolism in the systemic elimination of vincristine. The metabolism of structurally related compounds, vinblastine and vindesine, with human liver microsomes was inhibited by vincristine and other CYP3A-selective inhibitors, which suggests that CYP3A participates in the metabolism of vincristine (Rahmani and Zhou, 1993; Zhou-Pan et al., 1993). In addition, numerous clinical drug-drug interactions with itraconazole and nifedipine have been reported for vincristine, and the observed increases in neurotoxicity are consistent with the CYP3A-mediated metabolism of vincristine (Fedeli et al., 1989; Bohme et al., 1995; Kamaluddin et al., 2001; Sathiapalan and El-Solh, 2001). The genetically polymorphic expression of the CYP3A5 enzyme may contribute to the clinically observed interindividual and interracial variability in vincristine response. More than 70% of African-Americans have at least one CYP3A5*1 allele which allows expression of active CYP3A5, but the CYP3A5*1 allele is only present in 10 to 20% of Caucasians (Xie et al., 2004). CYP3A4 is normally considered the major metabolizing enzyme in the CYP3A family (Williams et al., 2002). However, for individuals with the CYP3A5*1 allele, CYP3A5 can represent more than 50% of the total CYP3A content in human liver microsomes (Huang et al., 2004). CYP3A5 genotype-dependent drug clearance is well established for tacrolimus, a substrate oxidized by CYP3A4 and CYP3A5 with relatively equal efficiencies in vitro (MacPhee et al., 2002; Kamdem et al., 2005).
Structure of vincristine.
Despite the evidence implicating CYP3A enzymes in the primary metabolic elimination of vincristine, the relative contributions of the drug-metabolizing P450s to the oxidation of vincristine along with the structural identities of the primary oxidative metabolites remain to be determined (Zhou et al., 1994). These deficiencies preclude an understanding of the genetic and environmental factors that may contribute to the interindividual variability in vincristine disposition. In addition, the roles of primary and subsequent metabolites to the efficacy and toxicity of this important chemotherapeutic agent are unknown. Therefore, the objectives of this study were to identify the major metabolites from the P450-mediated metabolism of vincristine and to determine the relative contributions of the major drug-metabolizing P450s to vincristine metabolism in vitro.
Materials and Methods
Chemicals and cDNA-Expressed Human P450s. Vincristine sulfate (98%; see Fig. 1 for structure) and peroxidase (type VI; 298 purpurogallin units/mg) were purchased from Sigma Chemical Co. (St. Louis, MO). [G-3H]Vincristine sulfate (3.30 Ci/mmol) was obtained from GE Healthcare Bio-Sciences (Buckinghamshire, UK). NADPH (98%) was purchased from Roche Diagnostics (Indianapolis, IN). cDNA-expressed P450s coexpressed with P450-reductase (CYP1A1, 1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 3A4, 3A5, and 4A11), P450s coexpressed with P450-reductase and cytochrome b5 (CYP2J2, 2E1, 3A4, 3A5, and 3A7), and insect cell control microsomes (Supersomes) were purchased from the BD Gentest Corporation (Woburn, MA). Cytochrome b5 was purchased from PanVera Corporation (Madison, WI). The manufacturers provided the P450-reductase activities, protein concentrations, cytochrome b5 content, and P450 content as appropriate. All other reagents were of HPLC grade and were purchased from Fisher Scientific (Pittsburgh, PA).
Purification of Vincristine. For key experiments, vincristine was purified by HPLC before use because vincristine-related impurities were present in the commercially available vincristine, including a compound that coeluted with the major metabolite M1 (see below). A vincristine standard curve with vinorelbine as an internal standard was used to quantify the concentration of vincristine in the HPLC fractions. Before use, the fractions containing purified vincristine were evaporated to dryness at room temperature. 3H-vincristine sulfate was also purified using this procedure.
Incubations with cDNA-Expressed P450 Enzymes. All incubations were performed in duplicate. Vincristine (5 μM) was preincubated with CYP1A1, 1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 2E1, 2J2, 3A4, 3A5, 3A7, and 4A11 (50 pmol of P450 for CYP2A6, 2C8, and 2E1; 25 pmol of all P450s) and insect cell microsomes (control; matched protein content to highest P450) in a suitable buffer (final volume 250 μl). Coexpressed cytochrome b5 was present with CYP2J2, 2E1, and 3A7. In accord with the recommendation of the manufacturer, CYP2A6, 2C9, and 4A11 incubations were performed in 100 mM Tris buffer with 5 mM MgCl2, pH 7.4 at 37°C and the remainder in 100 mM Na2HPO4 buffer with 5 mM MgCl2, pH 7.4. The reaction was initiated with the addition of NADPH (0.5 mM). After 1 h, the incubation was quenched with an equal volume of acetonitrile, chilled, and centrifuged. The supernatant (50 μl) was directly assayed by HPLC with UV detection (see below).
To determine the Michaelis-Menten parameters, vincristine (1–45 μM) was preincubated with the appropriate P450 and 100 mM Na2HPO4 with 5 mM MgCl2, pH 7.4. The reaction was initiated with the addition of NADPH (0.5 mM). After a set incubation time, the incubation was quenched with an equal volume of acetonitrile, chilled, and centrifuged. The supernatant (25–50 μl) was directly assayed by HPLC with UV detection. CYP3A activity can be stimulated by the addition or coexpression of cytochrome b5, and this effect may be P450-selective such that activities of CYP3A4 and CYP3A5 may be altered (McCune et al., 2005). Thus, the Michaelis-Menten parameters of CYP3A4 and CYP3A5 were determined with enzyme preparations containing cDNA-expressed enzyme alone, with coexpressed cytochrome b5, and with added cytochrome b5 (PanVera Corp.) at a molar ratio of 3:1 immediately preceding the incubations. To determine any effect of added cytochrome b5, the same lot numbers of CYP3A4 or CYP3A5 were used with and without cytochrome b5 to minimize variability. The incubation conditions for CYP3A4 (25 or 50 pmol, 15 min, 500-μl incubation) and CYP3A5 (25 pmol, 3 min, 500-μl incubation) reactions were determined based on the HPLC assay limits of quantitation and linear conditions. To determine linearity, different P450 concentrations and incubation times were tested, and the best conditions were selected in which less than 15% of the parent was metabolized. Microsomes without P450s (insect cell control Supersomes) were used as negative controls.
For the radiolabeled vincristine incubations, vincristine at approximately 10 μM (2 × 106 dpm per incubation) was used. CYP3A4 with coexpressed b5 (12.5 pmol), CYP3A5 with coexpressed b5 (25 pmol), and control insect microsomes (matched by CYP3A5 protein content) were incubated with the radiolabeled vincristine for 15 min at a final volume of 250 μl.
LC/UV/Radioactivity Analysis. The HPLC system used was an Agilent 1100 Series (Agilent Technologies, Wilmington, DE) and a Hewlett Packard 1050 Series UV detector (Hewlett Packard, Wilmington, DE). Chromatographic separation of M1, M2, vincristine, and vinorelbine (internal standard) was achieved with a C18 column (Inertsil ODS3, 3.0 × 150 mm, 5-μm particle size; MetaChem Technologies Inc., Torrance, CA) at a flow rate of 0.4 ml/min. The mobile phase consisted of 0.2% formic acid (mobile phase A) and methanol (mobile phase B). Analytes were eluted using a series of linear gradients: 0 min/20% B, 7 min/20% B, 42 min/56% B, 42.1 min/80% B, 52 min/80% B, 52.1 min/20% B. The parent drug and metabolites were detected by ultraviolet absorbance at a wavelength of 254 nm. To allow quantification of M1 by HPLC using a vincristine standard curve, the extinction coefficients of M1 and vincristine were determined by synthesizing M1 from radiolabeled vincristine. The extinction coefficients were not statistically different (p = 0.78).
The metabolites were also quantified after incubation of radiolabeled vincristine with the P450s followed by HPLC. Fractions were collected in 20-s aliquots, and the radioactivity of each fraction was determined by liquid scintillation counting (LS3801 scintillation counter; Beckman, Fullerton, CA) with 5 ml of scintillation cocktail (ScintiVerse; Fisher Scientific Co., Fair Lawn, NJ).
Preparation of Vincristine Metabolites and Derivatives.M1 and M1-acetate. Vincristine (not purified) at 30 μM was preincubated with recombinant CYP3A5 (100 pmol) and 100 mM Na2HPO4 with 5 mM MgCl2, pH 7.4 (total volume 1 ml). The reaction was started with the addition of NADPH (0.5 mM final concentration). After 40 min, the incubation was extracted with methylene chloride (1 ml). M1 was then isolated by HPLC (see above), and the appropriate M1 eluent fractions were pooled. The product was extracted with methylene chloride (6 ml), and the solvent was evaporated at room temperature to yield M1. M1-acetate was prepared by mixing M1 and 100 μl of neat acetic anhydride at room temperature. After 30 min, the reaction was quenched with 200 μl of water and extracted with methylene chloride (3 × 1 ml). The organic extract was dried and reconstituted in the HPLC mobile phase before LC/MS/MS analysis.
M2 with Peroxidase. Larger quantities of M2 were prepared by the oxidation of vincristine by horseradish peroxidase and hydrogen peroxide, a procedure described previously (Ahn et al., 1997) with slight modifications. Briefly, vincristine sulfate (1.5 mg) was dissolved in 7 ml of 100 mM Na2HPO4 at 37°C. A solution of horseradish peroxidase (1.2 mg), 100 mM Na2HPO4 buffer (350 μl), and hydrogen peroxide (125 μl, final reaction concentration 0.4 mM) was added to the vincristine solution and incubated at 37°C in a shaking water bath for 2 h. The final mixture was extracted with methylene chloride (6 ml), and the organic extract was dried at room temperature. The residue was purified by HPLC, and the eluent fractions were dried to yield M2 (0.5 mg).
M3. Vincristine (not purified) at 30 μM was preincubated with recombinant CYP3A5 (100 pmol) and 100 mM Na2HPO4 with 5 mM MgCl2, pH 7.4 (total volume 1 ml). The reaction was initiated with the addition of NADPH (0.5 mM final concentration). After 40 min, the incubation was quenched with ethyl acetate (0.5 ml) and 5 N NaOH (100 μl). Additional ethyl acetate (3 ml) was added, and the organic layer was evaporated. The resulting residue was purified by HPLC as described above to yield M3. M3 was also prepared directly from the purified M1 by mixing M1 with water (1 ml) and 5 N NaOH (20 μl) followed by extraction with ethyl acetate (1 ml).
LC/MS Analysis. LC/MS analysis was carried out on a Shimadzu VP Series HPLC (Shimadzu Scientific Instruments Inc., Columbia, MD) interfaced with either a triple quadrupole or ion trap mass spectrometer (TSQ Quantum or LTQ mass spectrometer; Thermo Electron Corporation, Waltham, MA). Chromatographic separations were performed on a C18 column (Inertsil ODS-3, 2.1 × 150 mm, 5-μm particle size; MetaChem Technologies, Inc.) with a flow rate of 0.2 ml/min. The mobile phase consisted of 0.2% formic acid (mobile phase A) and methanol (mobile phase B). The analytes were eluted using a series of linear gradients: 0 min/20% B, 3 min/20% B, 28 min/50% B, 45 min/80% B, 50.1 min/20% B. Full-scan mass spectra were obtained between 150 and 1050 Da. Positive ion MS/MS was conducted for the ions of interest using argon as the collision gas at 1.5 mTorr and collision energy of –40 V. Accurate mass measurements were performed using a Waters Micromass Q-TOF II quadrupole/orthogonal time-of-flight mass spectrometer (Waters Corporation, Milford, MA). The protonated ion (m/z 311.0814) of sulfadimethoxine (Sigma-Aldrich, St. Louis, MO) was used as the lock mass in all accurate mass determinations.
NMR Analysis. NMR spectra were acquired on an Inova 500 MHz NMR system equipped with either a 5-mm cold triple-resonance probe or a 3-mm IFC indirect detection probe (Varian Inc., Palo Alto, CA). Compounds were dissolved in either CD3OD or DMSO-d6, transferred to a 3-mm NMR tube, and sealed before NMR analysis. Proton and carbon chemical shifts were referenced to the residual solvent signals at 3.3 and 49 ppm, respectively, in CD3OD and at 2.49 and 39.5 ppm, respectively, in DMSO-d6. Two-dimensional NMR experiments including TOCSY, DQFCOSY, HSQC, and HMBC were performed using Varian standard pulse sequences. A selective 1D ZQFTOCSY experiment (Bradley et al., 2005) was performed to identify proton chemical shifts in the extremely crowded regions.
Data Analysis. For incubations with cDNA-expressed CYP3A4 or CYP3A5, the Michaelis-Menten constants (Km) and maximal rates of metabolism (Vmax) were determined by fitting the data to a one-enzyme model using nonlinear least squares regression analysis (WinNonlin 4.0; Pharsight, Mountain View, CA). Statistical significance (p < 0.05) was determined using Student's t test or one-way analysis of variance for multiple comparisons.
Vincristine disappearance and M1 formation for various P450s. All enzymes were tested in duplicate (25 or 50 pmol, 5 μM vincristine, 250 μl final volume). Vincristine and M1 concentrations were compared with an internal standard and normalized to the average of an insect microsome control. The enzymes were tested in the absence of cytochrome b5 with the exception of coexpressed b5 in CYP2J2, 2E1, and 3A7. Error bars represent data ranges (n = 2) for each P450. Statistical differences from the control incubations were calculated for vincristine disappearance using one-way analysis of variance; *, p < 0.05, **, p < 0.01, ***, p < 0.001.
Results
P450 Screen. Screening experiments with a panel of recombinant enzymes revealed that CYP3A4 and CYP3A5 are the major P450s that metabolize vincristine as shown by parent drug disappearance and formation of M1, the major metabolite (Fig. 2). The vincristine disappearance was normalized to an insect microsome control, and no M1 was detected in the control. CYP3A4, 3A5, 3A7, and 2E1 showed a statistical difference in vincristine disappearance from the control, but only CYP3A4 and CYP3A5 depleted the substrate more than 10%. M1 accounted for 70% and 50% of parent drug loss in the CYP3A5 and CYP3A4 incubations, respectively. M2 accounted for 7% product loss with CYP3A5 and 2% with CYP3A4. M4 formation was also increased over the control for both CYP3A5 and CYP3A4 but only accounted for 1% and 4% of the product loss, respectively.
Vincristine Metabolite Profiles with CYP3A4 and CYP3A5.Figure 3 shows the metabolite profile of vincristine with CYP3A5 in which baseline separation was achieved for M1 (21 min), vincristine (VCR; 23 min), and M2 (35 min). M3, a compound formed from M1 under basic conditions, has a retention time of 38 min (peak not shown). Radiolabeled vincristine (10 μM) was also used to characterize the metabolite formation from 15-min incubations with CYP3A4 and CYP3A5. As shown in Fig. 4, the major metabolite for both CYP3A4- and CYP3A5-mediated metabolism of vincristine was M1. This peak was not present in the insect microsome control. M2 was present in all samples, but only the CYP3A5 reaction produced levels higher than that of the control, approximately 10% of M1. Another earlier eluting compound, M4 (19.3–19.7 min), was also present for CYP3A4 and CYP3A5 in larger quantities than that of the control, but the total amount was only 2% of the total radioactivity for both samples. M5 (21.0–21.3 min) was present in the control but was not detectable as a discrete peak in the other incubations because the M1 peak retention time (20.5–20.7 min) was too close to M5 to allow resolution. For the HPLC assay used to determine the kinetic parameters with UV detection, M5 can be distinguished from M1 by retention time. The radioactivity measured from 10 to 45 min represented at least 95% of the radioactivity present in the samples, and no additional peaks with retention times before 10 min or after 45 min were identified.
Structure Determination of Vincristine Metabolites.M2. M2 showed a protonated molecular ion peak at m/z 839, 14 Da higher than that of the corresponding vincristine ion peak. The MS/MS spectrum showed product ions at m/z 471, representing a protonated N-formylvindoline (NFV) segment, and m/z 369 due to the loss of the NFV segment (a similar ion was observed at m/z 355 for vincristine; Table 1), revealing metabolic changes only in the dihydro-hydroxycatharanthine (DHC) part of the molecule. The molecular mass of M2, 14 Da higher than that of vincristine, and its molecular formula C46H54N4O11, determined by the accurate mass measurement, led us to believe that M2 could be the same as the previously known metabolite of vincristine obtained by its biotransformation catalyzed either by horseradish peroxidase or ceruloplasmin (Ahn et al., 1997). This hypothesis was further substantiated by a direct comparison of the metabolic products derived from CYP3A5 and horseradish peroxidase reactions of vincristine, which showed identical LC/MS profiles (retention time and MS/MS data) and 1H NMR data. Hence, the structure of M2 is depicted as shown in Fig. 5A. As noted previously by Ahn et al. (1997), the formation of M2 could be rationalized by the oxidation of the carbon α to the piperidine nitrogen followed by the fission of the C-13 and C-14 bond (Fig. 5A). An analogous oxidative biotransformation pathway has also been reported for leurosine, another dimeric Vinca alkaloid, when it was catalyzed by horseradish peroxidase (Goswami et al., 1988). Furthermore, a similar compound, catharinine, has been reported as a natural product from several Catharanthus species (Andriamialisoa et al., 1978) as a result of the C-13 and C-14 bond fission of the piperidine ring, probably following the same metabolic reaction from an appropriate precursor in plants.
Proposed product ion assignments for vincristine (VCR) and its metabolites
Representative HPLC chromatogram of CYP3A5-mediated metabolism of vincristine (VCR) (50 pmol CYP3A5, 20 μM vincristine, 500 μl final volume). The retention times of M1, the major metabolite, and M2, a minor metabolite, are identified. Vinorelbine (VRL) is used as an internal standard.
Radiochromatograms of 10 μM 3H-vincristine incubations for 15 min with insect control microsomes, matched to CYP3A5 protein content (A), cDNA-expressed CYP3A4 with coexpressed b5, 50 pmol/ml (B), and CYP3A5 with coexpressed b5, 100 pmol/ml (C). At least 95% of the radioactivity is represented from 10 to 45 min, and no distinctive peaks different from the control are present outside of this time range. All dpm values are normalized to total radioactivity recovered.
In the 1997 publication, Ahn et al. have reported only partial NMR chemical shift assignments for M2. Herein, we report the complete (with very few exceptions) proton and carbon chemical shift assignments of M2 in two solvents, namely, CD3OD and DMSO-d6 (Tables 2 and 3). As noted by Ahn et al. (1997), both with vincristine and M2, we also observed two sets of resonances for M2 in these solvents, but in approximately a 4:1 ratio as opposed to the 3:2 ratio that was observed by Ahn et al. (1997) in CDCl3. The presence of rotamers in CDCl3 solutions of 18′-desacetylvincristine was postulated previously as a result of a restricted rotation involving the N-formyl group of the NFV moiety (Rao et al., 1989). In contrast, in the present investigations, we observed this phenomenon only with M2, and not with vincristine and M3 in CD3OD solutions. In addition, with M2, significant line broadening for most of the resonances was observed in DMSO-d6 at 25°C; however, upon heating to 100°C, almost all resonances returned to normal line width, with the exception of the formyl proton (H-22′), H-2′, and H-8′ of the NFV segment, which showed relatively sharper lines at 25°C. The fact that two sets of resonances were still observed when the sample was heated to 120°C may possibly reveal a significantly higher energy barrier between the two rotamers than is normally observed with rotamers due to restricted rotation.
Proton chemical shift assignments of vincristine, M2, and M3
Carbon chemical shift assignments of vincristine, M2, and M3.
M1. The initial MS analysis of M1, either on an ion trap or a triple quadrupole mass spectrometer, showed a probable protonated molecular ion peak at m/z 793. However, when analyte was introduced at higher concentrations into the source, an ion at m/z 811 (Fig. 6) was observed corresponding to a molecular formula C45H55N4O10 as determined by accurate mass measurements. Thus, we concluded that the ion at m/z 793 is probably a source-induced fragment ion from the protonated molecular ion at m/z 811 due to the loss of water. The MS/MS spectra of the ion at m/z 811 and the source-induced ion at m/z 793 showed product ions at m/z 341 and 323, respectively, due to the loss of the NFV component of vincristine, suggesting metabolic changes, again in the DHC part of the molecule. Attempts to characterize M1 further by NMR were complicated by the fact that M1 underwent a spontaneous decomposition during the dry down process after preparatory isolation. As described under Materials and Methods, however, a basic workup of the enzyme reaction mixture produced a stable product, M3. M1, the major component of the CYP3A5-mediated metabolism of vincristine, was observed to a lesser extent in this case, suggesting probable formation of M3 from M1 upon base treatment. The formation of M3 from M1 was additionally confirmed by treating purified M1 with a base, which yielded M3 as the dominant product.
Proposed biotransformation pathways of vincristine catalyzed by CYP3A4 and CYP3A5. A, formation of M2; B, formation of M1 and M3.
The structure of M3 was unambiguously elucidated by using a combination of MS and NMR analyses. The molecular formula was determined to be C44H50N4O9 by accurate mass measurements (protonated molecular ion at m/z 779). This formula represents 32 (CH4O) and 46 (C2H6O) Da less than M1 and vincristine, respectively. The presence of a product ion at m/z 309 as a result of the loss of NFV from the parent reinforced an unaltered NFV segment in M3 as well. Because vincristine and its related analogs are arguably the most complex and highly functionalized members of the indole alkaloid family, an extensive suite of NMR experiments was required to accurately assign all the proton and carbon chemical shifts and unequivocally establish the structure of its metabolites. The NMR results of M3 reported in Tables 2 and 3 were accomplished after an extensive analysis of 1H, DQFCOSY, TOCSY, HSQC, HMBC, and 1D ZQFTOCSY spectra. For reliable chemical shift comparisons, vincristine recovered from the CYP3A5 reaction mixture under HPLC conditions identical to those of M3 was used for the NMR experiments to minimize salt-induced effects because vincristine is only available commercially as a sulfate salt.
A key, obvious feature observed in the NMR spectrum of M3 compared with vincristine was the absence of a methoxy resonance (H3-23) of the methyl ester in the DHC part of the molecule. In addition, several deviations in the chemical shifts of key protons (Table 2) and carbons (Table 3) relative to vincristine were also observed. Among them, rather significant departures were observed in the values for both the ethyl side chains. As shown in Table 2, the most dramatic departure was observed for the methylene resonance (H2-20) of the DHC ethyl side chain (1.37 ppm in vincristine versus 2.44 ppm in M3) and the methyl resonance (H3-21′) of the NFV ethyl side chain (0.79 ppm in vincristine versus 0.07 ppm in M3). The above chemical shift assignments of the ethyl side chains of the respective DHC and NFV segments were established through long-range HMBC correlations observed from the methyl protons at 0.98 ppm (H3-21) to a methylene carbon at 36.7 ppm (C-20) and a carbonyl carbon at 212.5 ppm (C-14), and from the other methyl protons at 0.07 ppm (H3-21′) to a quaternary carbon at 43.3 ppm (C-16′) which, in turn, showed correlations to the double bond protons at 5.76 (H-14′) and 5.14 (H-15′) ppm. The presence of the unique keto ethyl functionality in the DHC segment of M3 revealed that both M2 and M3 must stem from the same metabolic reaction involving the oxidative cleavage of the piperidine ring.
Initially, the structural relationships between M2 and M1, the putative parent of M3 (28 Da), and M2 and M3 (60 Da) were difficult to reconcile based on the mass differences alone. However, inspection of the diol intermediate, which led to M2 as a result of the cleavage of the C-13 and C-14 bond (Fig. 5A), prompted us to consider an alternative cleavage pathway involving the N-12 and C-13 bond. In fact, such a cleavage should be favored over the C-C bond fission, which led to the formation of M2. As outlined in Fig. 5B, fission of the N-12 and C-13 bond would lead to M1. It is reasonable to assume that the nucleophilic secondary amine thus formed is presumably reactive and, under basic conditions, underwent an intramolecular amidation reaction, leading to the formation of M3 (Fig. 5B). This proposal is consistent with the loss of 32 Da, a mass equivalent of methanol, from M1 during the transition to M3. Furthermore, the absence of a methoxy resonance, as alluded to earlier, in the NMR spectrum of M3 compared with vincristine is also in accord with this proposal. The formation of the new amide ring and the structure of M3 were fully characterized by the long-range HMBC correlations observed from H-11, H-17, and H-18 to a carbonyl carbon at 185.1 ppm (C-22) and other selected correlations as shown in Fig. 7. Accordingly, in the MS/MS spectrum of M3, distinct product ions were observed at m/z 309, 281, and 237 for the modified DHC moiety (Table 1).
The structure proposed for M1 (precursor of M3) was further supported by the acetylation of M1 immediately after isolation with acetic anhydride, which showed, as expected, the molecular ion peak at m/z 853 and a product ion at m/z 383, 14 Da higher than M2, due to the loss of the NFV group. This change in the mass signifies the replacement of the formyl group in M2 by an acetyl group in the acetylated product of M1. The facile loss of water from the protonated molecular ion of M1 (m/z 811) in the mass spectrometer ion source could be explained by the formation of an aminol intermediate, as shown in Fig. 8, which then could loose water rapidly in the source to afford the product ion at m/z 793. The reactivity of this newly formed amine could, in part, explain the instability of M1 during the dry down process.
The transformation of vincristine to M3 led to significant structural changes between the two molecules in the DHC unit, and one might anticipate a dramatic conformational departure for M3 from vincristine. The molecular mechanics force field minimized structures of vincristine, and M3 revealed a difference in the disposition of the indole ring of the DHC unit with respect to the ethyl group of the NFV unit. Thus, the pseudo axial distance measured from the methyl of the ethyl group to the indole ring in vincristine is 3.83 Å and in M3, 2.16 Å (C. Krishna, R. Barbuch, and P. Kulanthaivel, manuscript in preparation). The unusually large diamagnetic shift (Δ –0.72 ppm) observed in the NMR spectrum of M3 for the methyl group (H3-21′) of the ethyl side chain compared with vincristine (Table 2) is consistent with the modeling experiments, which suggested a close proximity of the methyl group positioned perpendicular to the indole ring in M3. A similar diamagnetic shift, but to a lesser extent (Δ –0.34 ppm), was observed in M2 for the same methyl group compared with vincristine, suggesting a similar conformational change in M2 due to the fission of the C-13 and C-14 bond.
M4 and M5. Both M4 and M5 showed identical protonated molecular ion peaks at m/z 823, 2 Da less than the corresponding vincristine ion peak. The structures of M4 and M5 were tentatively proposed as the isomeric epoxides, probably arising as a result of dehydration followed by epoxidation of the resulting C14-C15 double bond of the DHC unit. This is supported by the co-occurrence of vincristine and leurosine in the plant, Catharanthus roseus, most likely as a result of similar metabolic changes. The product ion observed at m/z 353 (Table 1) was in concert with the proposed changes in the DHC unit. No additional efforts were taken to further characterize these metabolites.
Enzyme Kinetics. The rates of M1 formation were determined for cDNA-expressed CYP3A4 and CYP3A5 (Fig. 9). The Km and Vmax values were determined for CYP3A4 and CYP3A5 with and without cytochrome b5 (Table 4). The presence of cytochrome b5 consistently increased the Vmax values for both CYP3A4 and CYP3A5 incubations, with the highest Vmax achieved with coexpressed cytochrome b5. The intrinsic clearance values indicated that CYP3A5 selectively metabolized vincristine compared with CYP3A4 for all preparations (9- to 14-fold higher for CYP3A5). The Km was not statistically different for preparations without b5 and with coexpressed b5. The kinetic parameters were not determined for M2 or M4 because the control incubations contained small amounts of these metabolites, and for incubations in the linear range for M1 at the lowest concentrations of vincristine, the amounts of M2 and M4 formed were below the limits of detection. However, the amounts of M2 and M4 were higher than the control amounts at the highest concentration tested, 48 μM. Correcting for the amounts in the controls, M2 was less than 10% of the M1 values for both CYP3A5 and CYP3A4 reactions. M4 was approximately 5% of M1 for CYP3A5 incubations and 15% of M1 for CYP3A4 reactions. The levels of M5 for both CYP3A4 and CYP3A5 reactions were not measurably higher than the controls.
Michaelis-Menten parameters for M1 formation with CYP3A4 and CYP3A5
Units are as follows: Vmax, pmol M1/pmol P450/min; Km, μM. Fitted values ± 1 standard deviation. CLint, Vmax/Km.
Discussion
Our findings show that vincristine oxidation by P450s is predominantly mediated by CYP3A4 and CYP3A5. This result is in agreement with clinical reports of drug-drug interactions upon coadministration of vincristine with CYP3A inhibitors such as itraconazole and nifedipine (Fedeli et al., 1989; Bohme et al., 1995; Kamaluddin et al., 2001; Sathiapalan and El-Solh, 2001). The metabolism of vincristine primarily by CYP3A is also consistent with the physicochemical features of vincristine; in this case, a large molecular volume. In addition, CYP3A4 and CYP3A5 metabolize vincristine to one major metabolite, M1, a novel compound with chemical modifications to the dihydro-hydroxycatharanthine (DHC) part of the molecule, and two minor metabolites, M2 and M4. Previous in vitro experiments with human hepatocytes described the rapid formation of three or four unknown compounds from the metabolism of vincristine, vinblastine, and vindesine (Zhou et al., 1994). These products may include M1 and the minor metabolites (M2 and M4) that we describe herein or other secondary metabolites. Previously, in vitro studies with structurally related Vinca alkaloids, vinblastine and vindesine, showed one major metabolite in human microsomal studies, but the investigators were unable to obtain enough pure material to allow structural identification of the metabolite (Zhou et al., 1993; Zhou-Pan et al., 1993). Beyond scale limitations, the chemical instability of M1 that we observed during standard isolation methods may have hampered previous attempts to identify M1 or its related Vinca analogs. We determined the structure of M1 unequivocally by modifying the reactive secondary amine moiety to form two chemically stable derivatives: M3, formed by the base-catalyzed intramolecular amidation of M1, and M1-acetate, generated by acetylation with acetic anhydride. The structures of these compounds as determined by LC/MS/MS and/or NMR analyses in combination with the previously described structure and formation of M2 (Ahn et al., 1997) led to the proposal of a new oxidation pathway leading to the formation of M1. In this pathway, M1 shares a common intermediate with M2, formed by the oxidation of C13, α to N-12. Although cleavage of the C-13 and C-14 bond resulted in the formation of M2, as documented previously (Ahn et al., 1997), fission of the N-12 and C-13 bond followed by subsequent oxidation and elimination, as shown in Fig. 5B, resulted in the formation of M1. We believe that the new pathway proposed should be favored over the previously proposed pathway leading to the formation of M2, since chemical cleavage of the C-N bond would be favored over the cleavage of the C-C bond.
Full-scan mass spectrum of M1. The ions at m/z 406 and 397 are the doubly charged molecular ion and the source-induced dehydration ion, respectively.
Selected 1H-13C HMBC correlations of M3. The J-filter was optimized for 1JC-H = 140 Hz and N bond delay was set to 0.063 s, corresponding to NJC-H = 8 Hz.
In addition to identifying the major metabolite M1, we discovered that CYP3A5 selectively metabolizes vincristine versus CYP3A4, regardless of cytochrome b5 content; for M1 formation, CYP3A5 had a 9- to 14-fold higher intrinsic clearance than CYP3A4. For most tested substrates, the CYP3A5 intrinsic clearance is similar to or less than that of CYP3A4 (Williams et al., 2002; Patki et al., 2003; Huang et al., 2004; McConn et al., 2004). For example, some studies show a modest selectivity (3-fold) for CYP3A5 with midazolam (Huang et al., 2004), but for other studies, CYP3A5 has equivalent or less activity than CYP3A4 in the formation of 1′-OH-midazolam (Williams et al., 2002; Patki et al., 2003; Walsky and Obach, 2004). Recently, another study on the metabolism of buprenorphine showed the selective metabolism by CYP3A5 versus CYP3A4 in a screening experiment, but rigorous enzyme kinetics were not presented (Chang et al., 2006). Thus, our findings with vincristine, which show an order of magnitude greater intrinsic clearance values, provide an unusual example of a very high selectivity for CYP3A5.
Mechanism of formation of M1 fragment ion at m/z 793 in MS ion source. R = N-formylvindoline.
Rate of M1 formation normalized to incubation time during CYP3A4- and CYP3A5-mediated metabolism of vincristine with coexpressed cytochrome b5 (•), exogenous b5 (3:1) (▾), and no b5 (○). Data were fit to a one-enzyme model to determine Michaelis-Menten parameters. Replicate data were not available for M1 formation with CYP3A5 and exogenous b5.
The selective metabolism of vincristine by CYP3A5 may be clinically important because CYP3A5 is polymorphically expressed. CYP3A5 is expressed in approximately 70% of African-Americans, who have at least one CYP3A5*1 allele, which allows expression of significant quantities of active CYP3A5, but the CYP3A5*1 allele is only present in 10 to 20% of Caucasians (Xie et al., 2004). The most common allelic variants include CYP3A5*3, CYP3A5*6, and CYP3A5*7. For the CYP3A5*3 genotype, a single nucleotide polymorphism in intron 3 creates a consensus splice site that after processing leads to a premature termination codon (Kuehl et al., 2001). The less common CYP3A5*6 and CYP3A5*7 allelic variants also result in the expression of little to no active CYP3A5 (Kuehl et al., 2001). Because such a small amount of active CYP3A5 is produced in individuals without the CYP3A5*1 allele, the other genotypes described are effectively void of active CYP3A5.
Polymorphic CYP3A5 expression may, in part, explain interracial differences in vincristine efficacy (Pollock et al., 2000; Lange et al., 2002). Clinical studies that evaluated the impact of CYP3A5 genotype on drug disposition have been confounded by the variability in CYP3A4 expression and often show substrate dependence. Studies with midazolam, a common CYP3A probe, have shown mixed results. Some clinical studies showed increased metabolism for CYP3A5*1 carriers (Wong et al., 2004; Yu et al., 2004), but other studies did not find a statistical difference between the groups (Shih and Huang, 2002; Floyd et al., 2003). For other CYP3A substrates, such as tacrolimus, drug disposition is dependent upon CYP3A5 genotype. For example, in an organ transplant study, patients with CYP3A5*3/*3 required less drug to reach the required therapeutic trough concentrations compared with CYP3A5*1 carriers (Hesselink et al., 2003). This result is consistent with the clinical evidence that African-Americans have a poorer outcome after solid organ transplantation (Nair et al., 2002). Interestingly, in vitro studies showed that the intrinsic clearances of tacrolimus with recombinant CYP3A4 and CYP3A5 were similar, with less than a 2-fold selectivity by CYP3A5 (Kamdem et al., 2005) compared with the 9- to 14-fold selectivity we report for vincristine. Therefore, if the in vitro model is representative of human metabolism, the CYP3A5 contribution to vincristine metabolism may be even more clinically relevant than tacrolimus. Based on our results of selective metabolism by CYP3A5, it seems particularly important to evaluate the impact of CYP3A5 genotype on the pharmacokinetics of vincristine. This type of study may allow more individualized dosing strategies, which could improve clinical outcomes in cancers that include vincristine in their treatment.
Acknowledgments
We thank Jeffrey J. Alberts for the accurate mass measurements and David R. Jones for technical assistance.
Footnotes
-
This study was funded by the National Institutes of Health (1 K23 RR019956-01). J.B.D. and P.K. contributed equally to this work.
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.106.009902.
-
ABBREVIATIONS: ALL, acute lymphoblastic leukemia; P450, cytochrome P450; DQFCOSY, double-quantum filtered correlation spectroscopy; HPLC, high performance liquid chromatography; HMBC, heteronuclear multiple bond correlation; HSQC, heteronuclear single-quantum coherence; LC, liquid chromatography; MS, mass spectrometry; MS/MS, tandem mass spectrometry; TOCSY, total correlation spectroscopy; 1D ZQFTOCSY, one-dimensional zero-quantum filtered total correlation spectroscopy; b5, cytochrome b5; NFV, N-formylvindoline; DHC, dihydrohydroxycatharanthine; OH, hydroxy.
- Received February 26, 2006.
- Accepted May 3, 2006.
- The American Society for Pharmacology and Experimental Therapeutics