Abstract
NSC686288 [aminoflavone (AF)], a candidate chemotherapeutic agent, possesses a unique antiproliferative profile against tumor cells. Metabolic bioactivation of AF by drug-metabolizing enzymes, especially CYP1A monooxygenases, has been implicated as an underlying mechanism for its selective cytotoxicity in several cell culture-based studies. However, in vivo metabolism of AF has not been investigated in detail. In this study, the structural identities of 13 AF metabolites (12 of which are novel) in mouse urine or from microsomal incubations, including three monohydroxy-AFs, two dihydroxy-AFs and their sulfate and glucuronide conjugates, as well as one N-glucuronide, were determined by accurate mass measurements and liquid chromatography-tandem mass spectrometry fragmentation patterns, and a comprehensive map of the AF metabolic pathways was constructed. Significant differences between wild-type and Cyp1a2-null mice, within the relative composition of urinary metabolites of AF, demonstrated that CYP1A2-mediated regioselective oxidation was a major contributor to the metabolism of AF. Comparisons between wild-type and CYP1A2-humanized mice further revealed interspecies differences in CYP1A2-mediated catalytic activity. Incubation of AF with liver microsomes from all three mouse lines and with pooled human liver microsomes confirmed the observations from urinary metabolite profiling. Results from enzyme kinetic analysis further indicated that in addition to CYP1A P450s, CYP2C P450s may also play some role in the metabolism of AF.
The diaminoflavone compound, NSC686288 [5-amino-2-(4-amino-3-fluorophenyl)-6,8-difluoro-7-methylchromen-4-one; aminoflavone (AF)] (Fig. 1), is entering phase I clinical trials as a candidate for chemotherapy based on its antiproliferative potency proven in both in vitro cell culture and in vivo tumor xenograft studies (Akama et al., 1998; Loaiza-Perez et al., 2004a, b). Results from a comprehensive National Cancer Institute (NCI) screening test showed that AF selectively inhibited the growth of diverse types of tumor cells among a panel of 60 widely used cell lines (Weinstein et al., 1997), and the difference between sensitive and resistant cell lines with regard to the responses to AF treatment was very distinctive and significant (http://www.dtp.nci.nih.gov/docs/dtp_search.html).
Mechanistic studies on AF's cytotoxicity have demonstrated that AF can cause DNA damage through DNA single-strand breaks and DNA-protein cross-linking and subsequently induce histone H2AX phosphorylation and S phase arrest (Meng et al., 2005; Pobst and Ames, 2006). As for its unique selectivity, significant effort has focused on the potential bioactivation of AF by the P450, CYP1A1. Recombinant CYP1A1 oxidized AF to multiple metabolites in vitro (Kuffel et al., 2002). AF treatment also induced a much higher level of CYP1A1 protein in sensitive cell lines than in resistant cell lines through the aryl hydrocarbon receptor-mediated pathway (Kuffel et al., 2002; Loaiza-Perez et al., 2004a, b). Recently, a relationship between sensitivity to AF treatment and expression level of cytosolic sulfotransferase (SULT) was identified through analyzing the gene expression pattern of AF-sensitive cell lines, and stable transfection of SULT1A1 converted AF-resistant cells into sensitive cells (Meng et al., 2006). Therefore, it is highly possible that multiple xenobiotic-metabolizing phase I and II enzymes, either collectively or separately, contribute to the metabolic activation of AF.
To date, information on the identities of AF metabolites was limited to the detection of one N4′-acetylated metabolite in plasma (Phillips et al., 2000) and one N4′-hydroxylated metabolite after incubations with microsomal membranes (Kuffel et al., 2002). It should also be noted that current knowledge on the bioactivation of AF was mainly derived from in vitro incubation and cell line-based experiments. In human and animal studies, liver, kidney, intestine, and other extrahepatic tissues will likely participate in the absorption, distribution, metabolism, and excretion of AF. Indeed, it is well known that the major circulating forms of chemotherapeutic agents that are destined for uptake into tumors are not the parental compounds in the most cases, and the expression levels of major drug-metabolizing enzymes in tumor tissues are significantly different from those in liver. For example, CYP1A1 may be the major AF-metabolizing enzyme in tumor cells, but in liver, its expression level is far lower than CYP1A2 (Shimada et al., 1996). The literature is replete with examples illustrating significant differences between cell culture models and human and animals studies with regard to the metabolism of specific drugs.
To understand the in vivo metabolism of AF and the role of CYP1A2, AF metabolites were identified and their structures elucidated, and urinary metabolites of AF from wild-type, Cyp1a2-null, and CYP1A2-humanized mice were profiled after oral dosing of AF. The regioselectivity of CYP1A2-mediated oxidation and interspecies differences in the metabolism of AF were confirmed after microsomal incubation, and the contribution of major P450s to the metabolism of AF was determined by incubations with recombinant human P450s and enzyme kinetics.
Materials and Methods
Chemicals and Reagents. Aminoflavone was provided by the Developmental Therapeutics Program, NCI. α-Naphthoflavone, sulfaphenazole, mephenytoin, high-performance liquid chromatography-grade acetonitrile, methanol, and formic acid as well as β-NADPH were purchased from Sigma (St. Louis, MO). Pooled human liver microsomes (HLMs) and recombinant human P450s (CYP1A1; 1A2; 1B1; 2A6; 2B6; 2C9; 2C19; 2D6; 2E1; 3A4) from baculovirus-infected insect cells were purchased from BD Biosciences (San Jose, CA).
Animals. Both the Cyp1a2-null mouse line (mCyp1a2-/-) and CYP1A2-humanized mouse line on a Cyp1a2-null background (hCYP1A2+/+, mCyp1a2-/-) were described previously (Pineau et al., 1995; Cheung et al., 2005). Wild-type mice on a 129/Sv strain background were used in this study. All animals were maintained in the NCI's animal facilities under a standard 12-h light/dark cycle with food and water ad libitum. Handling and treatment procedures were in accordance with animal study protocols approved by the NCI Animal Care and Use Committee.
In Vivo Metabolism of AF in Mice. Four female mice from each genotype group, age 8 to 12 weeks, were used for urinary metabolite profiling. Urine samples were collected by housing mice for 24 h in glass metabolic bowls (Jencons, Leighton Buzzard, UK), and stored at -80°C before usage. Control urine samples were collected 2 days before AF treatment, and then a 50 mg/kg dose of AF (suspended in corn oil) was administrated by oral gavage. After urine collection, mice were killed, and livers were harvested and snap-frozen in liquid nitrogen before storage at -80°C. Samples for ultraperformance liquid chromatography (UPLC)-time-of-flight mass spectrometry (TOFMS) analysis were prepared by mixing 50 μl of urine with 200 μl of 50% acetonitrile and centrifuging at 18,000g for 5 min to remove protein and particulates. A 200-μl aliquot of the supernatant was transferred to an autosampler vial for further analysis.
Preparation of Mouse Liver Microsomes. Livers from untreated wild-type (mCyp1a2+/+), Cyp1a2-null (mCyp1a2-/-), and transgenic (hCYP1A2+/+, mCyp1a2-/-) mice were homogenized, and microsomes (MLMs) were prepared as described previously (Yu et al., 2003).
In Vitro Metabolism of AF in MLM and HLM. Microsomal incubations were carried out in 20 mM phosphate-buffered saline, pH 7.4, containing 0.5 mg/ml microsomal protein, 2 mM MgCl2,20 μM AF, and 1 mg/ml freshly prepared β-NADPH in a final volume of 200 μl. After a 30-min incubation at 37°C, the reaction mix was directly loaded into a pretreated Oasis column (Waters, Milford, MA), and metabolites were eluted with 1 ml of methanol. The extract was dried by N2 flow and reconstituted in 50% aqueous acetonitrile for LC-MS/MS analysis.
In Vitro Metabolism of AF by Recombinant Human P450s. Reactions were carried out in 20 mM phosphate-buffered saline, pH 7.4, containing 20 nM recombinant human P450s (CYP1A1; 1A2; 1B1; 2A6; 2B6; 2C9; 2C19; 2D6; 2E1; 3A4), 2 mM MgCl2,20 μM AF, and 1 mg/ml freshly prepared β-NADPH in a final volume of 200 μl. After 30 min of incubation at 37°C, the reaction mix was directly loaded into a pretreated Oasis column (Waters), and metabolites were eluted with 1 ml of methanol. The extract was dried by N2 flow and reconstituted in 50% aqueous acetonitrile for LC-MS/MS analysis.
Enzyme Kinetics of Recombinant Human CYP1A2, CYP2C9, CYP2C19, and CYP3A4. AF ranging from 200 nM to 80 μM was incubated with recombinant human CYP1A2, CYP2C9, CYP2C19, and CYP3A4 for 20 min under the same condition as described above. Kinetic parameters for three major monohydroxylated metabolites, Km and relative Vmax (represented as the percentage of the highest activity detected in all reactions), were evaluated by nonlinear regression (Prism GraphPad 2.0; GraphPad Software Inc., San Diego, CA).
Chemical Inhibition of HLM-Mediated AF Metabolism. Different from the aforementioned condition for microsomal incubations, chemical inhibitors of CYP1A2, CYP2C9, and CYP2C19 were preincubated with HLMs for 5 min before adding AF and NADPH. The reactions were terminated at 20 min, and samples were processed under the same conditions as described above.
LC-Electrospray Ionization-MS/MS Analysis of AF Metabolites. For LC-MS/MS analysis, a 5-μl aliquot of samples from mouse urine or in vitro incubation was injected into a Waters UPLC-QT-OFMS system. An Acquity UPLC BEH C18 column (Waters) was used to separate AF and its metabolites at 30°C. Flow rate of mobile phase was 0.6 ml/min with a gradient ranging from water to 95% aqueous acetonitrile containing 0.1% formic acid in a 10-min run. The QTOF Premier mass spectrometer was operated in the positive electrospray ionization mode. Capillary voltage and cone voltage was maintained at 3 kV and 40 V, respectively. Source temperature and desolvation temperature were set at 120 and 350°C, respectively. Nitrogen was applied as the cone (50 l/h) and desolvation gas (600 l/h) and argon as collision gas. For accurate mass measurement, the QTOFMS was calibrated with sodium formate solution (range m/z, 100-1000) and monitored by the intermittent injection of the lock mass leucine enkephalin ([M + H]+ = 556.2771 m/z) in real time. Mass chromatograms and mass spectral data were acquired by MassLynx software (Waters) in centroid format and were further processed by MetaboLynx software (Waters) to screen and identify potential AF metabolites through the mass difference from AF ion (m/z, 321.0851+). The structure of each AF metabolite was elucidated by tandem mass spectrometry (MS2) fragmentation with collision energy ranging from 10 to 40 eV.
Statistics. Experimental values are expressed as mean ± S.D. Statistical analysis was performed with independent Student's t tests, with P < 0.01 considered as statistically significant.
Results
Identification and Structural Elucidation of Phase I Metabolites of AF by Mouse Liver Microsomes. Previously, a number of potential AF metabolites were identified from high-performance liquid chromatography chromatograms of extracted samples derived from microsomal incubations, but only one phase I metabolite of AF was structurally elucidated (Kuffel et al., 2002). To obtain comprehensive structural information on the phase I metabolites of AF and set a basis for further in vivo metabolism studies, AF was incubated with MLMs from untreated mice. Extracted samples containing AF and its metabolites were separated by the UPLC system, and their accurate masses were measured by a time of flight mass spectrometer. By comparing chromatographic differences between control (without NADPH) and normal (with NADPH) samples and by examining the mass difference between AF and new ions generated by the MLM incubation, three monohydroxylated and two dihydroxylated metabolites were identified (Table 1). Their structures were further elucidated by MS2 fragmentation (Fig. 2).
As the reference compound for the structural elucidation of AF metabolites, the fragmentation pattern of the parent AF ion (I) was examined first to show the formation of major fragment ions (186+ and 136+) via a retro Diels-Alder (rDA) reaction in C-ring (Fig. 2A). Further detachment of the carbonyl group (-C = O) from ion 186+ generated ion 158+. In addition, the breaking of the C-C bond between the B-ring and C-ring yielded ion 110+ (C6H6NF+ by accurate mass measurement) from the B-ring.
AF metabolite II was a monohydroxy derivative of AF based on accurate mass measurement and the loss of free radical HO· (-17). The presence of ion 202+ instead of ion 186+ indicated that the oxidation had occurred in the A-ring (Fig. 2B). Therefore, the identity of this metabolite was assigned as N5-hydroxy-AF (N5-OH-AF). However, it is recognized that hydroxylation might also have occurred on the 7-methyl group, but this was deemed unlikely due to the ready loss of HO· (from the pseudomolecular ion, a known characteristic of N-hydroxy compounds (Saito et al., 1985).
The structure of AF metabolite III was previously reported as N4′-hydroxy-AF (N4′-OH-AF) (Kuffel et al., 2002). As shown in Fig. 2C, both the presence of ions 186+ and 158+ provided evidence that the oxidation did not occurred in the A-ring, and loss of the -CH2NO group from the parent ion (-44 from 337+ to 293+) supported this structure.
Generation of ions 186+ and 152+ by fragmenting the AF metabolite IV ion excluded the possibility of oxidation in the A-ring (Fig. 2D). The high abundance of fragment ion 110+ (C6H6NF+) as a result of breaking of the C-C bond between the B- and C-rings also excluded the presence of a hydroxyl group in the B-ring (inlaid spectrum in Fig. 2D). Therefore, the only possible oxidation site is the carbon-3 in the C-ring, which confirms the structure of metabolite IV as 3-hydroxy-AF (3-OH-AF). In this connection, it should be noted that both quercetin and kaempferol, the two most abundant natural flavonols in human diet, have the 3-hydroxy group in their structure (Havsteen, 2002).
Both AF metabolites V and VI are the dihydroxylated AFs based on their accurate masses (Table 1). Similar to metabolite II, the fragmentation of metabolite V yielded the characteristic ion 202+, which indicated a N5-hydroxylation reaction. Moreover, neutral loss of mass 44, similar to metabolite III, proved the existence of N4′-hydroxy group (Fig. 2E). By combining the aforementioned evidence, the structure of metabolite V was designated as N5,N4′-dihydroxy-AF. As for metabolite VI, the presence of ions 202+ and 152+ resembled the MS2 spectra of metabolites II and IV (Fig. 2F). Therefore, the structure of metabolite VI was designated as 3,N5-dihydroxy-AF.
Identification and Structural Elucidation of Urinary Metabolites of AF. The observation of five major phase I metabolites after MLM incubation suggested that AF might also be extensively metabolized in vivo. To test this hypothesis, urine samples from the AF-treated mice (50 mg/kg) were analyzed under the same chromatographic and spectroscopic conditions as extracts from in vitro incubation. To achieve an exhaustive detection and comprehensive coverage of major urinary metabolites, a software-assisted metabolite screening and detection (as described under Materials and Methods) was conducted based on the data from accurate mass measurements of all urinary ions and the theoretical mass changes caused by the common phase I- and II-metabolizing reactions. The results revealed the presence of AF parent ion (I), N5-OH-AF (II), and eight new conjugated metabolites (VII-XIV) in mouse urine samples (Table 2). Structural identities of those urinary metabolites were further elucidated by MS2 analysis. Their major fragment ions are shown in Table 2, and MS2 spectra of the six most abundant phase II conjugates (VIII-XII and XIV) are presented in Fig. 3, A to F.
AF metabolite VII as a sulfate conjugate was confirmed by the neutral loss of mass 80 (SO3). Appearance of fragment ions 202+ and 136+ matched the fragmentation pattern of N5-OH-AF (II) (Table 2). Therefore, it was designated as N5-hydroxy-AF sulfate.
Metabolite VIII was also a sulfate conjugate of monohydroxylated AF because of its accurate mass and neutral loss of mass 80. Fragment ions 186+ and 152+ indicated its resemblance to 3-OH-AF (III) (Fig. 3A), so its chemical identity was designated as 3-hydroxy-AF sulfate (VIII).
Formation of ion 353+ after neutral loss of sulfate group (-80) from metabolite IX indicated a dihydroxylated AF moiety (Fig. 3B). The presence of ion 202+ and 152+ validated the structure as 3,N5-dihydroxy-AF sulfate. Since the retention time (RT) shift caused by sulfation at the 3-OH group (0.19 min) was far less than that derived from sulfation at N5-OH group (0.90 min) (Table 1 and 2), and RT shift of metabolite IX from its parent compound 3,N5-dihydroxy-AF (V) was 0.11 min, the sulfate group of metabolite IX was putatively conjugated to the 3-OH group.
Metabolite X was a direct glucuronidation product of AF because of neutral loss of mass 176 (glucuronic acid moiety) and characteristic fragment ions of 186+ and 136+ from deconjugated AF (Fig. 3C). Furthermore, the presence of ion 363+ from potential rDA reaction revealed that glucuronidation occurred on the A-ring, in which the N5 atom is the only possible conjugation site. Therefore, metabolite X was designated as AF-N5-glucuronide.
Formation of ion 337+ after neutral loss of the glucuronic acid group (-176) from metabolite XI indicated a monohydroxylated AF moiety (Fig. 3D). Ions 320+ and 291+ were generated by consecutive loss of hydroxyl free radical (HO·) and carbonyl group (HCO) from the deconjugated molecule. It was noted that the deconjugated ion (337+) and parent ion (513+) coeluted at RT 4.74 min because of in-source fragmentation under the described MS condition, so a MS2 scan was performed on ion 337+ to obtain further structural information. As shown in the inlaid spectrum of Fig. 3D, the deconjugated ion possessed a very similar fragmentation pattern as N5-OH-AF (II). In addition, since it is known that glucuronidation of N-hydroxylamine mainly occurs at the nitrogen rather than oxygen atom (Kadlubar et al., 1977; Kaderlik et al., 1994), metabolite XI was putatively designated as N5-OH-AF-N5-glucuronide.
Ions 513+, 337+ (loss of glucuronic acid)n and 320+ (loss of HO·) in MS2 spectra of metabolite XII indicated that it was another glucuronide of monohydroxylated AF (Fig. 3E). The presence of ions 293+ and 186+ was consistent with the fragmentation pattern of N4′-OH-AF (III). Therefore, metabolite XII was putatively designated as N4′-OH-AF-N4′-glucuronide.
Formation of ion 353+ after neutral loss of glucuronic acid group (-176) from metabolite XIII indicated a dihydroxylated AF moiety (Table 2). The presence of ions 309+ and 202+, which were the characteristic fragments of N5,N4′-diOH-AF (V), indicated metabolite XIII as N5,N4′-diOH-AF glucuronide.
Ions 529+, 353+ (loss of glucuronic acid) and 336+ (loss of HO·) in MS2 spectra of metabolite XIV indicated it was another glucuronide of dihydroxylated AF (Fig. 3F). The presence of ions 202+ and 152+ resembled the fragmentation pattern of 3,N5-diOH-AF (VI). Since glucuronidation was the preferred conjugation reaction for N5-OH-AF and sulfation for 3-OH-AF, as proven by the high abundance of N5-OH-AF-N5-glucuronide (XI) and 3-OH-AF sulfate (VIII), metabolite XIV was putatively designated as 3,N5-diOH-AF-N5-glucuronide.
Relative Composition of Urinary AF Metabolites in Wild-Type, Cyp1a2-Null, and CYP1A2-Humanized Mice. Previous studies on the metabolism of AF by using tumor cell lines have shown that CYP1A1 and CYP1A2 contributed to the metabolic bioactivation of AF through an aryl hydrocarbon receptor-mediated pathway (Kuffel et al., 2002; Loaiza-Perez et al., 2004a, b). However, it is well known that CYP1A2 plays a much more prominent role in xenobiotic metabolism in vivo than does CYP1A1 based on its much higher expression level and catalytic activity in liver (Shimada et al., 1996; Turesky et al., 2002). In addition, CYP1A2 is also an important enzyme in the bioactivation of heterocyclic aromatic amines, a chemical class to which AF belongs (Butler et al., 1989). To investigate the role of CYP1A2 in the in vivo metabolism of AF, three mouse lines including wild-type, Cyp1a2-null (mCyp1a2-/-), and CYP1A2-humanized (hCYP1A2+/+, mCyp1a2-/-) mice were used for oral dosing of AF. Although all of the urine samples from wild-type, Cyp1a2-null, and CYP1A2-humanized mice contained AF (I) and its nine urinary metabolites (II, VII-XIV), significant differences in the relative abundance of several metabolites among three mouse lines became evident after pooling all the peak areas of AF and its urinary metabolites and calculating the percentage of each metabolite in this metabolite cluster (Fig. 4; Table 2). In wild-type mice (Fig. 4), AF was mainly converted to the 3,N5-diOH-AF sulfate (IX) and N5-OH-AF-N5-glucuronide (XI), but in the Cyp1a2-null mice (Fig. 4), the major metabolites were N5-OH-AF (II) and its glucuronide (XI). Comparing the metabolic profiles of wild-type and Cyp1a2-null mice, it was apparent that the removal of CYP1A2 from the metabolic system diminished the 3-hydroxylation reaction, which subsequently led to significantly decreased levels of the 3-hydroxy metabolites including 3-OH-AF sulfate (VIII), 3,N5-di-OH-AF sulfate (IX) and 3,N5-diOH-AF-N5-glucuronide (XIV), and to the accumulation of N5-hydroxy metabolites, such as N5-OH-AF (II), N5-OH-AF sulfate (VII), and N5-OH-AF-N5-glucuronide (XI). Therefore, 3-hydroxylation of AF was a selective CYP1A2-mediated reaction. This conclusion was further strengthened by the relative composition of urinary AF metabolites in CYP1A2-humanized mice. Introduction of the human CYP1A2 gene into Cyp1a2-null mice resulted in recovery of the 3-hydroxylation reaction of AF and its downstream metabolites and led to comparable levels of metabolites II, VII, IX, and XIV as found in wild-type mice (Fig. 4). However, there still existed significant differences between wild-type and CYP1A2-humanized mice with regard to the metabolites VIII and XI. The higher level of 3-OH-AF sulfate (VIII) and the lower level of N5-OH-AF glucuronide (XI) in CYP1A2-humanized mice implied that 3-hydroxylation of AF was a more preferred reaction for human CYP1A2 than for mouse CYP1A2. Besides the major metabolites from 3-hydroxylation and N5-hydroxylation, an AF-N5-glucuronide (X) and two N4′-hydroxylation products, N4′-OH-AF-N4′-glucuronide (XII) and N5,N4′-diOH-AF-N-glucuronide (XIII), were also detected as minor metabolites found at comparable levels in the three mouse lines. It should also be noted that all three mouse lines including Cyp1a2-null mice retained similar levels of the AF parent compound (I) in their urine, which comprised about 10 to 15% of total urinary AF metabolites. This observation indicated that, even without the mouse CYP1A2 enzyme, AF was still extensively metabolized. Therefore, some other P450 enzymes were also involved in the metabolism of AF.
Comparison of in Vitro Metabolism of AF by MLMs from Wild-Type, Cyp1a2-Null, and CYP1A2-Humanized Mice and by HLMs. Urinary metabolite profiling revealed a significant influence of CYP1A2 in the metabolism of AF through a 3-hydroxylation reaction and the potential interspecies difference with regard to the catalytic activity of CYP1A2 (Fig. 4; Table 2). To confirm these observations, 20 μM AF was incubated with MLMs from wild-type, Cyp1a2-null, and CYP1A2-humanized mice and pooled HLMs, respectively. As shown in Fig. 5, A and B, two monohydroxylamines, N5-OH-AF (II) and N4′-OH-AF (III), were generated by all three MLMs from wild-type, Cyp1a2-null, and CYP1A2-humanized mice, as well as by pooled HLMs, and both metabolites could be further converted to N5,N4′-diOH-AF (V) in all microsomal incubations. Consistent with the in vivo data (Fig. 4; Table 2), both 3-OH-AF (IV) and 3,N5-diOH-AF (VI) were absent after incubation of AF with the MLMs from Cyp1a2-null mice because of the lack of the CYP1A2-mediated 3-hydroxylation reaction. Furthermore, the interspecies difference between mouse and human CYP1A2 on the 3-hydroxylation activity was observed since the incubation of AF with MLMs from CYP1A2-humanized mice and HLMs resulted in much higher yields of 3-OH-AF (IV) than from MLMs of wild-type mice.
Comparison of in Vitro Metabolism of AF by Recombinant P450s. Extensive metabolism of AF in Cyp1a2-null mice implied the potential involvement of other P450s in the oxidation of AF (Table 2). Therefore, 20 μM AF was incubated with a panel of recombinant human P450s, including CYP1A1, CYP1A2, CYP1B1, CYP2A6, CYP2B6, CYP2C9, CYP2C19, CYP2D6, CYP2E1, and CYP3A4. Chromatograms generated by three most reactive P450s (CYP1A1, CYP1A2, and CYP2C19) are presented in Fig. 6A, and the relative activities of the whole P450 panel on the generation of three monohydroxylated metabolites (II-IV) and two dihydroxylated metabolites (V and VI) are illustrated in Fig. 6, B to F. As shown in Fig. 6, CYP1A2 was more selective in the conversion of AF to 3-OH-AF (IV), whereas CYP1A1 was more active than CYP1A2 in the production of N5-OH-AF (II) and N4′-OH-AF (III). The catalytic activities of CYP1A1 and CYP1A2 for the formation of N5,N4′-diOH-AF (V) were comparable, but CYP1A2 was far more active than CYP1A1 for the formation of 3,N5-diOH-AF (VI) since it required CYP1A2-mediated 3-hydroxylation. Besides CYP1A P450s, two CYP2C P450s, CYP2C9 and CYP2C19, were also active in the N-hydroxylation of AF to generate metabolites II and III. In addition, both CYP3A4 and CYP2D6 also had low but measurable levels of AF-metabolizing activity (Fig. 6, B-D).
Relative Contribution of CYP1A2, CYP2C, and CYP3A Enzymes in the Metabolism of AF. Since both CYP2C and CYP3A P450s are two of the most abundant P450s in liver (Guengerich, 2005), their relative contribution to the formation of the three major monohydroxylated metabolites (II-IV) were evaluated by an enzyme kinetic experiment. As shown in Fig. 7, A and B, it is clear that CYP2C19 may be an important enzyme for the N-hydroxylation of AF in humans, especially at high substrate concentrations, although the contribution from CYP2C9 and CYP3A is likely to be negligible. Nevertheless, compared with CYP1A2 (Km values for II and III are 4.0 and 5.7 μM, respectively), AF-metabolizing activity of CYP2C19 (Km values for II and III are 14.9 and 17.5 μM, respectively) was still much lower. By calculating the ratio between relative Vmax and Km, enzyme efficiencies of CYP1A2 for the formation of metabolites II and III were more than 6- and 5-fold greater than CYP2C19, respectively. In addition, consistent with other in vivo and in vitro data (Table 2; Figs. 4, 5 and 6), enzyme kinetic analysis further confirmed that 3-hydroxylation is a CYP1A2-dominant reaction, in which the Km value of CYP1A2 is 3.3 μM (Fig. 7C). Furthermore, results from the preincubations of HLMs with widely used chemical inhibitors showed that α-naphthoflavone (CYP1A2 inhibitor) can significantly inhibit the formation of all three monohydroxylated AF metabolites, especially to 3-OH-AF (IV), although sulfaphenazole (CYP2C9 inhibitor) and mephenytoin (CYP2C19 inhibitor) failed to affect AF metabolism (Fig. 7D).
Discussion
Structure-activity relationship protocols guided the development of AF as a chemotherapeutic agent. Based on the fact that natural or synthetic flavonoids can elicit antiestrogenic and/or estrogenic activities, AF was derived from the structure of flavone (Havsteen, 2002). By adding two amino groups as electron donors and three fluoride atoms plus a methyl group for blocking metabolism, AF possesses much higher cytotoxicity against the estrogen receptor-positive cell line MCF-7 than its parent compounds (Akama et al., 1996, 1997). Recent results from the NCI 60 cell line screenings indicated that the bioactivity of AF may surpass the original expectation of developing an antiestrogen for breast cancer therapy (http://www.dtp.nci.nih.gov/docs/dtp_search.html). Interestingly, although one of the original medicinal chemistry strategies for developing this compound was to decrease the potential metabolism in the flavone moiety (Akama et al., 1997), which was thought to have a deactivating effect on its bioactivity, the results from this study clearly demonstrated that AF was still extensively metabolized since multiple metabolites were detected after in vitro incubation and oral administration, and the parent compound only accounted for about 10 to 15% of all detected urinary metabolites in the three mouse lines used in this study.
To identify and profile of phase I and phase II metabolites of AF, the UPLC system coupled with a TOFMS were adopted to achieve high-resolution chromatographic separation and accurate mass measurements (in the 5-10 ppm range of real mass). By combining the information on the retention time and the mass of each ion, computer-assisted analysis can automatically interrogate the mass chromatograms and spectra for identifying the potential AF metabolites. Compared with a manual search, this strategy was more efficient and exhaustive. It should be noted that in-source fragmentation occurred for several sulfate and glucuronide metabolites, presumably because of weak chemical bonds between the parent compound and the conjugated moiety. With careful examination and adjustment of conditions on the mass spectrometer, the identities of those fragments can be unambiguously resolved. Structures of major AF metabolites were elucidated through analyzing the fragmentation pattern recorded by tandem mass spectrometry. Like typical flavone compounds, the rDA reaction contributes to the formation of major daughter ions of AF and its metabolites by breaking the C-ring (Mullen et al., 2002). Examining the product ions from the rDA reaction and the breaking of the C-C bond between the B- and C-rings can pinpoint whether the phase I reaction happens on the A-, B-, or C-rings. As for phase II metabolites, the conjugation sites of sulfation and glucuronidation were putatively assigned based on the common rules of drug metabolism and previous knowledge of the N-glucuronidation of hydroxylamine (Kadlubar et al., 1977; Kaderlik et al., 1994).
As mentioned above, previous studies on the AF bioactivation and metabolism have focused on the CYP1A P450s, CYP1A1 and CYP1A2 (Kuffel et al., 2002; Loaiza-Perez et al., 2004b). Results from urinary metabolite profiling (Table 2), in vitro incubation (Figs. 5 and 6), and chemical inhibition by α-naphthoflavone (Fig. 7D) in this study also support the view that CYP1A2 plays an important role in AF metabolism, especially for the 3-hydroxylation reaction. However, as shown in Table 2, mice without CYP1A2 can still extensively metabolize AF, which suggests that CYP1A2 was not the sole enzyme responsible for AF metabolism in vivo. Indeed, incubation of AF with a panel of recombinant human cytochromes P450 revealed that CYP2C9 and CYP2C19 have significant N-hydroxylation activities (Fig. 6), and enzyme kinetic analysis further indicated that CYP2C19 has a pharmacologically relevant Km value for the formation of hydroxylamines (II and III) (Fig. 7). Although mephenytoin, a CYP2C19 substrate and inhibitor, did not inhibit the HLM-mediated metabolism, which is possibly attributed to the ineffective competitive inhibition mechanism, a preliminary inhibition experiment showed that anti-rat CYP2C11 antibody can significantly inhibit the N-hydroxylation reactions catalyzed by MLMs (data not shown). It is known that there is some degree of overlap in the substrate specificity of CYP1A and CYP2C enzymes (Venkatakrishnan et al., 1999; Otake and Walle, 2002). Because of their prominence in hepatic xenobiotic metabolism system, CYP2C enzymes could contribute to the N-hydroxylation of AF, although the 3-hydroxylation reaction is highly CYP1A2-selective. Further in vivo and in vitro studies are required to understand the role of CYP2C P450s in AF metabolism and the potential interspecies difference.
Another interesting observation on the relative composition of urinary metabolites was the much lower levels of N4′-hydroxylation products in urine than N5- and 3-hydroxylation products (Table 2), although the relative abundance of N4′-OH-AF (III) after microsomal incubations was comparable with that of N5-OH-AF (II) and 3-OH-AF (IV) (Fig. 5). During the course of this study, it was noted that when stored at room temperature or 4°C, the half-life of N4′-OH-AF (III) in the microsomal incubation extracts was much shorter than other monohydroxylated AF metabolites (II and IV) (data not shown). In addition, the abundance of the parent ion of N4′-OH-AF (III) in the MS2 fragmentation experiment was also much lower than N5-OH-AF (II) and 3-OH-AF (IV) with the same collision energy (Fig. 2), and it disappeared completely with the increased collision energy. Therefore, this hydroxylamine metabolite is likely to be unstable and chemically reactive. Since reactive hydroxylamine compounds are destined for conversion into genotoxic electrophiles through acetylation and sulfation (Turesky, 2002), it is possible that N4′-OH-AF (III) can be further transformed to more reactive species that can covalently bind to DNA and protein, as shown in a recent study on the bioactivation of AF by sulfotransferase (Meng et al., 2006). It would be interesting to investigate whether N4′-OH-AF (III) is the major reactive and cytotoxic metabolite of AF in future studies. Furthermore, N4′-acetyl-AF, an AF metabolite found in rat plasma (Phillips et al., 2000), has not been detected in urine, although its presence in plasma was also observed in this study. A possible explanation is that acylases, which exist in high abundance in kidney, may deacetylate N4′-acetyl-AF back to AF (Yamauchi et al., 2002).
One of the drawbacks of using mouse urine samples for metabolism studies, especially for the purpose of quantitation, is the high degree of variance in urine volume and quality caused by dosing, water intake, evaporation, and contamination from diet and feces. By calculating the relative percentage of each metabolite in the whole metabolite cluster, the aforementioned variability problem among the samples can be significantly reduced since the concentrations of all urinary components change proportionally with the alteration in experimental conditions. As shown in Table 2, the results from this study demonstrated that urinary metabolite profiling based on the relative composition was able to establish a data set with small variance; therefore, it can function as an efficient tool to investigate the influence of genetic modification on drug metabolism. This strategy may also be applicable to drug-drug interaction studies.
The use of genetically modified animal models, including gene knockout and transgenic mice, provides a powerful tool for studying xenobiotics metabolism. Cyp1a2-null mice have been used to demonstrate the important role of CYP1A2 on the pharmacokinetics and metabolism of caffeine (Buters et al., 1996) and CYP1A2-humanized mice to the interspecies difference in metabolism of the carcinogen 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (Cheung et al., 2005). This study further proved the value of both animal models for studying the metabolism of CYP1A2 substrates as well as interspecies differences between human and mouse. Besides their application to drug metabolism, these animal models can also be used to determine the role of drug-metabolizing enzymes in the metabolism of endogenous chemicals and the modulation of normal homeostasis and for understanding the role of a candidate enzyme in the carcinogenicity of known carcinogens (Gonzalez, 2003).
In summary, by combining the accurate mass measurement and the software-assisted global metabolite profiling strategy, 13 AF metabolites, from CYP-mediated oxidation and phase II glucuronidation and sulfation, were identified, and a comprehensive in vivo metabolic pathway of AF was constructed (Fig. 8). Besides one known metabolite, 12 novel metabolites were structurally elucidated by LC-MS/MS analysis. Significant differences in the relative composition of urinary metabolites among three mouse lines demonstrated that CYP1A2 plays an important role in the AF metabolism, although it is not the sole enzyme capable of oxidizing AF; 3-hydroxylation is a CYP1A2-selective reaction; and there is a clear interspecies difference between mouse and human CYP1A2 in the metabolism of AF. In vitro incubations of AF with microsomes and recombinant P450s as well as enzyme kinetics confirmed these observations and further indicated that CYP2C enzymes might also contribute to AF metabolism.
Acknowledgments
We thank the NCI Developmental Therapeutics Program for providing aminoflavone.
Footnotes
-
doi:10.1124/jpet.106.105213.
-
This work was supported by the NCI Intramural Research Program of the National Institutes of Health and by the U.S. Smokeless Tobacco Company (grant for collaborative research to J.R.I.).
-
ABBREVIATIONS: NSC686288, 5-amino-2-(4-amino-3-fluorophenyl)-6,8-difluoro-7-methylchromen-4-one; AF, aminoflavone; NCI, National Cancer Institute; SULT, sulfotransferase; HLM, human liver microsome; UPLC, ultraperformance liquid chromatography; TOFMS, time-of-flight mass spectrometry; MLM, mouse liver microsome; LC-MS/MS, liquid chromatography-tandem mass spectrometry; MS2, tandem mass spectrometry; rDA, retro Diels-Alder reaction; N5-OH-AF, N5-hydroxy-AF; N4′-OH-AF, N4′-hydroxy-AF; 3-OH-AF, 3-hydroxy-AF; RT, retention time.
- Received March 24, 2006.
- Accepted June 13, 2006.
- The American Society for Pharmacology and Experimental Therapeutics