Introduction

Top Abstract Introduction Results Discussion References

Prostaglandin G/H synthase (PGHS²), also known as cyclooxygenase (COX), catalyzes the conversion of arachidonic acid to prostaglandin H₂. Nonsteroidal anti-inflammatory drugs exert their anti-inflammatory activity through inhibition of PGHS. The recent discovery of a second form of PGHS (COX-2), expressed primarily in inflamed tissues, has generated much interest in developing isozyme-selective nonsteroidal anti-inflammatory drugs. Selective inhibition of COX-2 should prove beneficial in treating a variety of inflammatory conditions (Ehrich et al., 1999; Lefkowith, 1999; Schwartz et al., 1999), while reducing the potential for gastrointestinal toxicity associated with inhibition of the physiological function of COX-1 (Lanza et al., 1999; Lefkowith, 1999). One such selective and potent inhibitor of COX-2 is DFU, [5,5-dimethyl-3-(3-fluorophenyl)-4-(4-methylsulphonyl)phenyl-2(5H)-furanone] (Riendeau et al., 1997).

An important component in support of COX-2 inhibitor candidate selection has been the discovery-stage metabolism studies that are directed at understanding the metabolic fate of candidate compounds in safety species, and the likely metabolic pathways in humans. The structures of novel metabolites of drug candidate compounds, generated in incubations with hepatic microsomes or with suspensions of freshly isolated hepatocytes, are routinely examined using HPLC with UV-diode array or mass spectrometric detection (Chauret et al., 1995; Li et al., 1995). Changes in UV spectra, molecular weight, and mass spectral fragmentation patterns are often sufficient to identify the substructure that has been metabolically altered. In particular, capillary HPLC/continuous-flow liquid secondary ion mass spectrometry (CF-LSIMS) has been used to identify metabolites of DFU, in a manner similar to previous reports (Chauret et al., 1995; Li et al., 1995).

In most cases, metabolism leads to discreet increments in molecular weight (e.g., oxidation or glucuronidation), such that accurate mass determination is not particularly useful in the identification process. However, for DFU, products formed in hepatocyte incubations did not have apparent molecular weights that could be easily rationalized. The metabolites were, therefore, characterized using HPLC/CF-LSIMS at high resolution to generate accurate masses and elemental compositions and were isolated for NMR structural characterization. The accurate mass CF-LSIMS and NMR data suggested that a novel rearomatization was occurring following the addition of glutathione to an epoxide intermediate produced in hepatocyte incubations of DFU.

Experimental Procedures

Chemicals and Reagents. DFU and p-OH-DFU (Fig. 1) were synthesized at Merck Frosst (QC, Canada). The latter was prepared from a solution of [3] [5,5-dimethyl-3-(3-fluoro-4-methoxyphenyl)-4-(4-methylsulphonyl)phenyl-2(5H)-furanone] (Fig. 1) (100 mg, 0.26 mmol) in CH₂Cl₂ (2 ml) at 78°C, to which was added BBr₃ (0.51 ml, 0,51 mmol, 1.0 M in CH₂Cl₂). The cooling bath was removed, more BBr₃ (0.25 ml, 0.25 mmol) was added, and the mixture was stirred for 1 h at room temperature. Water (20 ml) was added, and the mixture was extracted with CH₂Cl₂ (3 × 10 ml). The combined organic layers were washed with brine and then dried over MgSO₄. Evaporation of the solvent and trituration from Et₂O gave p-OH-DFU, 90 mg (93%) as a colorless solid: mp 232-233°C, ¹H NMR (400 MHz, acetone-d₆) d 4'-OH 8.86 (d, J = 1.2 Hz, 1H), 3",5" 8.06 (d, J = 8.2 Hz, 2H), 2",6" 7.66 (d, J = 8.2 Hz, 2H), 2' 7.21 (dd, J = 12.6, 2.0 Hz, 1H), 6' 6.97 (dd, J = 8.4 Hz, 2.0 Hz, 1H), 5' 6.84 (dd, J = 8.8 Hz, 8.4 Hz, 1H), 3.17 (s, 3H), 1.59 (s, 6H). Deuterated NMR solvents were obtained from MSD Isotopes (Montreal, Canada). All other chemicals were of analytical, reagent, or HPLC grade, as required.

View larger version (16K): [in this window] [in a new window]   Fig. 1.   Structures of DFU[1] and p-OH-DFU [2], and [3].

Microsome Preparation and Incubations. Hepatic microsomes were prepared according to published procedures (Lu and Levine, 1972) from fresh Sprague-Dawley rat livers. Oxidative incubations were conducted with approximately 1 mg of microsomal protein and 200 µM substrate in a final incubation medium of 100 mM sodium phosphate (pH 7.4) containing 2.0 mM MgCl₂, 2.0 mM NADP, 20 mM glucose 6-phosphate, and 2 units of glucose-6-phosphate dehydrogenase. The total incubation volume was 0.5 ml and contained 2.5% DMSO (v/v). Incubations with preboiled microsomes or without the NADPH-generating system were used as negative controls. For examining microsomal glucuronidation of p-OH-DFU, incubations contained 10 mM UDP-glucuronic acid and 16 mM D-saccharic acid-1,4-lactone in 50 mM phosphate buffer (pH 6.6). Control incubations also either used boiled microsomes or contained all of the cofactors except UDP-glucuronic acid. After 1 h at 37°C, all incubations were quenched with 0.5 ml of acetonitrile to precipitate the proteins, vortexed, and centrifuged. The resulting supernatant was diluted with aqueous HPLC mobile phase before HPLC/UV or HPLC/CF-LSIMS characterization.

Rat Hepatocyte Incubations. Hepatocytes were isolated from male Sprague-Dawley rats by collagenase perfusion of the fresh liver and the cell viability was assessed by trypan blue exclusion, using light microscopy, as previously detailed (Silva et al., 1998). DFU (50 µM) was incubated for 3 h with hepatocytes (1 × 10⁶ cells/0.5 ml of incubation, 0.6% DMSO), suspended in Krebs-Henseleit buffer containing 12.5 mM HEPES under an atmosphere of 95% O₂/5% CO₂ in a continuously shaking 48-well plate at 37°C (Nicoll-Griffith et al., 1999). All incubations were quenched with an equal volume of acetonitrile, vortexed, centrifuged, and the resulting supernatant diluted with aqueous HPLC mobile phase before characterization.

Analytical HPLC Characterization. Analytical HPLC characterizations were performed on a Waters HPLC system (Waters, Milford, MA) consisting of a model 660 MS gradient pump, a model 715 autosampler, a Waters Nova-Pak C₁₈ column (0.46 × 15 cm) and a model 996 photodiode array detector operated by Waters Millennium software. A flow rate of 1 ml/min was used for the mobile phase consisting of methanol and 20 mM ammonium acetate (pH 7.0). Incubation mixtures were separated using gradient elution from 20 to 90% methanol in 25 min.

Mass Spectrometric Instrumentation and Methods. All CF-LSIMS determinations were performed using a JEOL HX-110A mass spectrometer (JEOL USA, Inc., Peabody, MA), as previously described (Li et al., 1995). The mobile phase consisted of methanol and 20 mM ammonium acetate (pH 6.8), with 1.5% glycerol added to each as a necessary matrix for the LSIMS ionization. Gradient elution from 20 to 80% methanol in 40 min was used for all LC/MS separations. Full-scan mass spectra were obtained by scanning linearly from 0 to 1000 Da every 4 s and are presented as averages across the chromatographic peak tops, with subtraction of the adjacent chromatographic background.

Accurate Mass Determinations. Accurate mass data were obtained for metabolites using the same capillary HPLC/CF-LSIMS procedures, but using electric-field scans at 5000 resolution (10% valley definition), spanning the molecular ion of interest and the closest glycerol adduct ions. For metabolite M1 and the biosynthetic p-OH-DFU-glucuronide, this corresponds to a mass range of 540 to 660 Da, and for metabolites M2 and M3 the mass range was 630 to 770 Da. To increase sensitivity, an aliquot of the rat hepatocyte incubation supernatant was first concentrated 10-fold by solid-phase extraction, using a 1-ml C₁₈ cartridge, before injection and acquisition of accurate mass CF-LSIMS data.

Metabolite Isolation for NMR. Metabolites M1 through M3 were isolated for NMR characterization from a large-scale rat hepatocyte incubation mixture. DFU (200 µM) was incubated for 3 h with 25 ml of freshly isolated rat hepatocytes (5 × 10⁶ cells/ml). The reaction was terminated with an equal volume of acetonitrile, vortexed, and centrifuged. Except as noted, the isolation procedures were as previously described (Li et al., 1995), involving a crude solid-phase extraction step to remove microsomal protein, preparative HPLC for isolation of the metabolites, and a final solid-phase extraction step before drying the isolated material. Preparative HPLC isolations were performed with a Waters NovaPak C₁₈ column (19 × 300 mm), a mobile phase consisting of methanol and 20 mM ammonium acetate (pH 6.8), and gradient elution from 20 to 80% methanol in 40 min.

NMR Instrumentation and Methods. Metabolite NMR samples were prepared by dissolving each metabolite in 160 µl of deuterated DMSO and placing the solution in a 5-mm Shigemi symmetrical NMR microtube (Shigemi, Inc., Allison Park, PA) matched to the solvent. All ¹H-detected NMR spectra were acquired on a Bruker AMX 500 spectrometer (Bruker Spectrospin, Milton, ON, Canada) equipped with a 5-mm inverse broadband gradient probehead (Bruker, Newark, DE). The spectra were acquired at 295K and referenced to the residual solvent line of DMSO (2.49 ppm for ¹H and 39.5 ppm for ¹³C). ¹⁹F NMR spectra were acquired at 282.4 MHz on a Bruker AMX 300 spectrometer (Bruker Spectrospin) at 295K. ¹⁹F chemical shifts were reported relative to CF₃COOH. Assignment of the metabolites was carried out by comparison to the previously assigned ¹H NMR spectrum of DFU and by a combination of one-dimensional nuclear Overhauser enhancement (NOE) and two-dimensional heteronuclear one bond and multiple bond (HMBC) experiments.

	Results

Top Abstract Introduction Results Discussion References

Oxidative Microsomal Incubations. A single metabolite (M4, 13.6 min) was produced in rat microsomal incubations of DFU (16.1 min), as observed by analytical HPLC with UV-diode array detection. Too little metabolite M4 was produced (< 1% of the parent DFU) to obtain a reliable UV-diode array spectrum for comparison with DFU. The retention time of M4 was identical to that of the synthetic p-OH analog of DFU (p-OH-DFU). DFU was also incubated with microsomes under identical conditions, but in the presence of 5 mM glutathione. HPLC/UV examination revealed a smaller M4 peak, but no additional peaks were observed.

Microsomal Glucuronidation of p-OH-DFU. Microsomal incubation of p-OH-DFU, using glucuronidation conditions, led to the very efficient (98% in 1 h) production of a metabolite having the same chromatographic properties as M1 generated in rat hepatocytes. Capillary HPLC/CF-LSIMS characterization of the p-OH-DFU glucuronidation incubation in full-scan positive-ion mode provided a strong mass spectrum for the metabolite with an apparent MNH₄⁺ ion at 570 Da.

Rat Hepatocyte Incubations. HPLC/UV characterization of rat hepatocyte incubations containing DFU, compared with blank incubations without compound and control incubations conducted with denatured (boiled) hepatocytes, indicated that three metabolites (M1-M3) were formed (Fig. 2), as has been previously reported (Nicoll-Griffith et al., 1999). Approximately 90% of parent DFU remained following the 3-h rat hepatocyte incubation. Capillary HPLC/CF-LSIMS characterization of rat hepatocyte metabolites in full-scan mode produced clear mass spectra for metabolites M2 and M3 only, with apparent MH⁺ ions at 664 and 666 Da in positive-ion mode and apparent (MH) ions at 662 and 664 Da in negative-ion mode, respectively. Metabolite M1 did not give a strong spectrum in full-scan CF-LSIMS mode, producing only a weak ion at 361 Da, corresponding to the MH⁺ for the parent DFU.

View larger version (12K): [in this window] [in a new window]   Fig. 2.   HPLC/UV-diode array chromatogram for the rat hepatocyte incubation of DFU, with UV data plotted at 275 nm (see Experimental Procedures for chromatographic system).

Accurate Mass Determinations. A portion of the rat hepatocyte incubation was concentrated 10-fold relative to the original using solid-phase extraction. This concentrated sample and the microsomal incubation of p-OH-DFU conducted under glucuronidation conditions were examined using capillary HPLC/CF-LSIMS at a resolving power of 5000. Figure 3 illustrates the capillary HPLC/high-resolution CF-LSI mass spectra for metabolites M1 to M3, and Fig. 4 illustrates the high resolution CF-LSI mass spectrum of the biosynthetic glucuronide of p-OH-DFU. Table 1 summarizes the observed and expected masses and elemental compositions for the metabolites.

View larger version (22K): [in this window] [in a new window]   Fig. 3.   Capillary HPLC/CF-LSIMS determination of 10-fold solid-phase extraction concentrated rat hepatocyte incubation (10 µl injected). Gradient separation: 20 to 80% solvent A (methanol) in 40 min. High resolution (5000, 10% valley definition) unsubtracted CF-LSI mass spectra. a, metabolite M1; b, metabolite M2; c, metabolite M3. Refer to Table 1 for elemental compositions and theoretical accurate masses.

View larger version (19K): [in this window] [in a new window]   Fig. 4.   Capillary HPLC/CF-LSIMS determination of p-OH-DFU glucuronidation incubation (10 µl injected). Gradient separation: 20 to 80% solvent A (methanol) in 40 min. High resolution (5000, 10% valley definition) unsubtracted CF-LSI mass spectrum. Refer to Table 1 for elemental composition and theoretical accurate mass.

NMR Characterization. Metabolites M1 to M3 were isolated using preparative HPLC from a large-scale rat hepatocyte incubation and characterized using NMR. The ¹H NMR of M1 suggested that the addition of glucuronic acid had occurred; however, the sample was not pure enough for complete structural studies to be carried out. In the case of M2, the fluorine was not observed in the ¹⁹F spectrum, indicating a loss of the aromatic fluorine, which is in agreement with the mass spectrometric results. Analysis of the proton NMR spectrum (Fig. 5) showed that this compound also was a glutathione adduct and that the glutathione had added to the fluorine-substituted ring. This ring now bore only three protons at 7.09 ppm (d, J = 8.4 Hz), 6.85 ppm (d, J = 1.9 Hz), and 6.66 ppm (dd, J = 8.4, 1.9 Hz). A one-dimensional NOE experiment showed a contact between protons 2",6" and protons 2',6'. Finally, protons 2' and 6' as well as the Cys beta protons on the glutathione moiety had strong three-bond correlations with a single carbon at 123.0 ppm whereas H-5' had a strong three-bond correlation with a carbon that resonated at 154.1 ppm in the long-range heteronuclear HMBC experiment.

View larger version (20K): [in this window] [in a new window]   Fig. 5.   1H NMR spectrum of M2 at 500 MHz.

View larger version (20K): [in this window] [in a new window]   Fig. 6.   1H NMR spectrum of M3 at 500 MHz.

	Discussion

Top Abstract Introduction Results Discussion References

A single metabolite (M4) was produced in oxidative microsomal incubations of DFU. Based on the compound structure and our discovery metabolism studies of related compounds, the p-OH analog was immediately synthesized (p-OH-DFU). The HPLC and CF-LSIMS data for p-OH-DFU were identical to those for metabolite M4. An attempt to trap the potential epoxide intermediate leading to M4, by addition of glutathione to microsomal incubations, was unsuccessful. While a smaller M4 peak was observed in incubations containing glutathione, no additional peaks were observed. The relatively low turnover in this system would make observation of a small additional peak quite difficult, however.

The HPLC/UV-diode array data for the glucuronidation of the biosynthetic p-OH-DFU provided chromatographic evidence that metabolite M1 formed in rat hepatocytes (Fig. 2) was the p-OH-DFU-glucuronide. Using capillary HPLC/CF-LSIMS in full-scan mode, the biosynthetic p-OH-DFU glucuronide produced an abundant MNH₄⁺ ion at 570 Da, but no corresponding molecular ion at 570 Da was visible in the spectrum of metabolite M1 from the hepatocyte mixture. However, 570 Da corresponds to the mass/charge ratio for an abundant glycerol adduct ion (Glycerol₆NH₄⁺) observed in the CF-LSI mass spectra, and as such, this mass/charge ratio will be subtracted from the full-scan spectra. Although the molecular ion for M1 was not observed in these full-scan data, a fragment corresponding to the aglycone was observed at m/z 361, and the molecular ion for M1 was observed in the high-resolution experiment, as detailed below.

Characterization of metabolites M2 and M3, produced in rat hepatocyte incubations, was less straightforward. The UV spectra for metabolites M2 and M3 were not drastically different from that of DFU, but as with the microsomal metabolite M4, the UV spectra were not strong enough to observe subtle changes. The capillary HPLC/CF-LSI full-scan positive- and negative-ion CF-LSI mass spectra included strong molecular ions, but little fragmentation. These data suggested molecular weights for M2 and M3 of 663 and 665, respectively. These masses correspond to the addition of glutathione to an oxidized DFU, with a subsequent loss of 20 and 18 Da, respectively. The 18-Da mass difference would suggest a loss of water after addition of glutathione to a DFU epoxide. However, the only elemental composition that could account for the corresponding loss of 20 Da following glutathione addition would be HF, an unexpected loss.

Metabolites M1 to M3 were also examined mass spectrometrically at high resolution, to help assign elemental compositions, and were isolated for NMR structure determination. The high-resolution mass spectrometric data are summarized in Table 1 and are consistent with elemental compositions for metabolites M2 and M3 that correspond to addition of glutathione to an oxidized DFU with a subsequent loss of HF and H₂O, respectively. The accurate mass data for biosynthetic p-O-glucuronide prepared from p-OH-DFU further confirmed the identity of the glucuronide. The accurate mass data for metabolite M1 also confirm its elemental composition as the p-O-glucuronide. The data also clearly indicate, since the metabolite peak is a small shoulder on the large glycerol adduct ion, why the molecular ion for M1 was not observed under full-scan conditions. The mass spectrometric studies on M2 suggested that fluorine had been lost in an oxidized glutathione adduct of DFU. This hypothesis is supported by 1) the loss of the single fluorine signal in the ¹⁹F NMR spectrum, and in the ¹H NMR spectrum the observation of glutathione; 2) the loss of one proton on the fluorine-substituted ring; 3) and an upfield shift in the chemical shifts of the remaining three lower ring protons. These results indicated that both the OH and glutathione ligands were bound to this ring. Furthermore, the coupling constants of the remaining protons and the observed NOE between 2",6" and 2',6' indicated that this ring was 1,2,4-trisubstituted. Thus, only the proposed structure of M2 (Fig. 7) and one in which the OH and SG are reversed are possible. These two possible structures were distinguished by the observation of strong three-bond correlations in the HMBC spectrum between carbon 4' and protons 2',6' and the Cys beta protons of the glutathione. Only the structure shown for M2 in Fig. 7 satisfies this requirement. Metabolite M3 was easily identified as a glutathione adduct on the fluorine-substituted ring of DFU by observation of its ¹H NMR spectrum. As was the case with M2, all of the protons of glutathione were identified, as were the methylsulfone ring and geminal methyl protons. Therefore, the glutathione must have added to the lower ring. Further evidence for this is that one of the four protons of this ring could no longer be found. The substitution pattern of this ring was determined by analysis of the proton-proton and proton-fluorine coupling constants. Examination of the ¹⁹F spectrum showed that the lone fluorine had two coupling constants of 9.9 and 9.6 Hz, which are consistent with three-bond proton-fluorine aromatic coupling. Thus, there must be a proton adjacent to each side of the fluorine, indicating that the glutathione substituent can be at neither the 2' nor the 4' positions. The remaining coupling constants, each less than or equal to 2.5 Hz, must then arise from proton-proton coupling. The fact that these values are so small necessitates them all to be four-bond aromatic couplings requiring the ring to be 1,3,5-substituted. The only possible structure to satisfy these heteronuclear and homonuclear couplings is that shown in Fig. 7.

View larger version (23K): [in this window] [in a new window]   Fig. 7.   Proposed pathway for the in vitro metabolism of DFU.

In summary (see Fig. 7), the available HPLC/UV, HPLC/MS, and NMR data are consistent with the proposal that DFU is initially metabolized to an epoxide, which then either rearranges to the p-OH and is conjugated with glucuronic acid or is trapped with glutathione, with a subsequent rearomatization with net loss of HF (M2) or H₂O (M3). Either a 3',4'-epoxide or 4',5'-epoxide could lead to the formation of the p-OH (M4) and p-O-glucuronide (M1) metabolites. However, two distinct epoxides are likely involved in the production of M2 and M3, involving attack of glutathione on a 3',4'-epoxide to produce metabolite M2 and on a 4',5'-epoxide to produce metabolite M3.

The loss of fluoride in the production of M2 represents a novel rearrangement following glutathione addition to an aromatic epoxide. Glutathione has been shown in many cases to add to ring systems such as hydroquinones, or quinones, resulting in a structure with glutathione conjugated directly to an aromatic system. Examples include glutathione adducts to acetaminophen (Mitchell et al., 1974; Hinson et al., 1977; Miner and Kissinger, 1979; Nelson et al., 1981) and anthrapyrazoles (Renner et al., 1995). However, the Michael addition in these cases results directly in an aromatic system and does not require the loss of a small neutral molecule to bring the system back to aromaticity. On the other hand, when glutathione adds to an arene epoxide, the net result is typically a dihydrohydroxy-S-glutathionyl system, such as that seen in the glutathione addition observed for verlukast (Nicoll-Griffith et al., 1993) or that observed in the metabolism of carbamazepine (Amore et al., 1997). Carbamazepine was also shown to produce an aromatic glutathione conjugate, purportedly via nonenzymatic dehydration of a dihydrohydroxy-S-glutathionyl metabolite (Amore et al., 1997), as is proposed for DFU metabolite M3. Clozapine has similarly been shown to produce aromatic glutathione conjugates, including a minor biliary metabolite in the rat and mouse corresponding to a des-chloro-glutathione adduct (Maggs et al., 1995). In their study, however, Maggs et al. proposed that the mechanism of formation involved a radical cation or diimine intermediate, rather than an arene epoxide. A quinone or peroxy radical intermediate was proposed to lead to the S-methyl replacement of fluorine in flumezapine metabolism (Sullivan and Franklin, 1985). None of the proposed mechanisms for formation of thiol conjugates of clozapine or flumezapine are likely to be involved in the glutathione conjugation of DFU. Aromatic epoxidation, followed by rearrangements and conjugations, including defluorination and fluorine NIH shifts, were described for metabolism of GW420867X (Dear et al., 2000). A single thiol conjugate, identified as a desfluoro-hydroxycysteine adduct, was found in rabbit urine, but the mechanism leading to this product was not established. We believe our article is the first to show direct evidence of glutathione conjugation to an epoxide involving loss of fluoride upon rearomatization. The use of hepatocytes, as a more complete metabolic system than microsomes for generating metabolites, coupled with accurate mass determination and NMR spectroscopy, provided a powerful combination for characterizing low-level metabolites of DFU.