Numerous drugs containing a carboxylic acid group, and especially certain nonsteroidal anti-inflammatory drugs (NSAIDs), have been associated with diverse clinical toxicities (Griffin and Scheiman, 2001). Thus hepatotoxicity, to divergent extents, is known for nearly all NSAIDs, although the proportion of hepatic versus nonhepatic side effects varies considerably between the drugs and hepatotoxicity is not exclusive to the carboxylate compounds (Merlani et al., 2001). The liver damage is in most cases considered to derive from an idiosyncratic metabolic and/or immune reaction that is essentially independent of dose (Boelsterli, 2002; Bailey and Dickinson, 2003). Some carboxylate drugs are also linked with rarer hypersensitivity reactions. In general, the symptomatic hepatotoxicity is uncommon, mild, and reversible; only in exceptional instances does it result in fulminant liver failure.

SB-209247 [(E)-3-[6-[[(2,6-dichlorophenyl)-thio]methyl]-3-(2-phenylethoxy)-2-pyridinyl]-2-propenoic acid; Fig. 1], an LTB₄ receptor antagonist with oral anti-inflammatory activity in mice (Daines et al., 1996; Davis et al., 2000), was developed for the treatment of dermal inflammatory diseases. LTB₄ is a potent, proinflammatory neutrophil activator, chemotactic agent, and apoptosis regulator that is released from neutrophils and other leukocytes during physiological and pathophysiological inflammatory reactions (Crooks and Stockley, 1998). Development of SB-209247 was discontinued when administration to beagle dogs at 60 to 1000 mg/kg/day (0.13-2.17 mmol/kg/day) for 28 days was associated with non-dose-dependant inflammatory hepatopathy, with minimal hepatocellular necrosis being observed in the more severe cases (GlaxoSmithKline, Uxbridge, Middlesex, UK, unpublished data). Male rats were given doses of SB-209247 that produced maximum plasma concentrations similar to those attained in the dogs, but no pathological changes were found in their livers. The etiology of the hepatopathy in dogs is unknown: an influx of inflammatory cells is a common response to hepatocellular injury, although it is superficially paradoxical in this case because LTB₄ receptor antagonists inhibit, particularly, the chemotaxis of neutrophils and, inter alia, their accumulation in skin (Brooks and Summers, 1996). Hepatic adverse reactions attributable to the various types of leukotriene antagonist are not unknown in experimental animals and humans, but there is presently no clear evidence for a generic association related to either pharmacological activity or the presence of a carboxylic acid function (Fretland et al., 1995; Chambers et al., 1999; Reinus et al., 2000).

Several mechanisms have been proposed by which a carboxylic acid might cause hepatotoxicity (Boelsterli, 2002). Some carboxylic acid drugs have the potential to initiate cell injury by acting directly on mitochondria (Boelsterli, 2003). Additional cytotoxic actions might require bioactivation, mediated by either oxidative metabolism (Bort et al., 1999) or the conjugation reactions forming acyl-CoA thioesters and acyl glucuronides (Boelsterli, 2002; Bailey and Dickinson, 2003). Ester-linked glucuronides have received particular attention because of their well documented ability to react with cellular proteins and DNA through nucleophilic addition or displacement. This adduction can result in perturbation of molecular and organellar activity (Sallustio and Holbrook, 2001) or the generation of neoantigens, which instigate immune-mediated cytotoxicity (Bailey and Dickinson, 2003). The cytotoxic or cytostatic actions of acyl glucuronides have in certain cases suggested direct effects of the conjugate (Sallustio et al., 1997; Seitz and Boelsterli, 1998; Bailey and Dickinson, 2003).

Conjugation with d-glucuronic acid is a common route of elimination for carboxylic acids and is catalyzed by a number of hepatic UGTs (Soars et al., 2001). Biosynthetic (β-1-O-) acyl glucuronides (Bailey and Dickinson, 2003) are generically reactive, electrophilic compounds capable of undergoing facile hydrolysis to the agylcone, rearrangement via reversible acyl migration and subsequent mutarotation, and covalent binding to proteins (King and Dickinson, 1991; Corcoran et al., 2001; Kenny et al., 2004). However, there is considerable variation in the intrinsic stability and reactivity of such conjugates (Boelsterli, 2002). Protein adducts are formed either by nucleophilic displacement of the glucuronic acid moiety, which produces aglycone-protein combinations, or by a glycation pathway, in which intramolecular migration of the acyl residue from C-1 allows opening of the glucuronic acid ring to create an aldehyde intermediate and subsequent formation of glucuronide-protein adducts (Bailey and Dickinson, 2003).

The propenoic acid moiety of SB-209247 and the absence of conspicuous, alternative conjugable groups make this drug a potential substrate for extensive acyl glucuronidation in vivo. We have considered the possibility that it might be bioactivated in dogs via this pathway: there are instances of carboxylic acid drugs undergoing species-selective acyl glucuronidation by hepatic microsomes (Soars et al., 2001; Prueksaritanont et al., 2002). The metabolism of SB-209247 and the compound's hepatocytotoxicity in vitro have been investigated with particular reference to comparative rates of glucuronidation and covalent binding to hepatocellular proteins.

Materials and Methods

Chemicals. NADPH, sodium estrone β-d-glucuronide, UDPGA, and H-2 β-glucuronidase-arylsulfohydrolase (Helix pomatia) were obtained from Sigma-Aldrich (Poole, Dorset, UK). Williams' medium E, Waymouth's MB 752/1 medium, hepatocyte-grade (type IV) collagenase, and other cell culture reagents were obtained from Invitrogen (Paisley, Scotland, UK). SB-209247 (Daines et al., 1996), SB-209247 sulfoxide (SB-215244), and [¹⁴C]SB-209247 (41.5 mCi/mmol; radiochemical purity by HPLC, 99.5%) were synthesized at GlaxoSmithKline (King of Prussia, PA). HPLC-grade solvents were products of Fisher Scientific (Loughborough, Leicestershire, UK). All other chemicals were purchased from BDH (Poole, Dorset, UK).

Animals. Adult male Wistar rats were obtained from a breeding colony in the Biomedical Services Unit, The University of Liverpool. The adult male Sprague-Dawley rats and beagle dogs used for the in vitro toxicity (liver slices) study were bought from Charles River Laboratories, Inc. (Wilmington, MA) and Marshall Farms (North Rose, NY), respectively; they were maintained at GlaxoSmithKline (King of Prussia, PA). The beagle dogs used for hepatocyte preparation were obtained from Huntingdon Life Sciences (Cambridgeshire, UK) and were maintained at GlaxoSmithKline (The Frythe, Hertfordshire, UK).

Human Livers. Histologically normal livers were obtained from four white male transplant donors (aged 19-41 years). Approval for the study was granted by the relevant ethical committees, and consent was obtained from the donors' relatives. The certified cause of death was either cerebrovascular injury or traumatic injury consequent to a road traffic accident. The livers were removed and transported from the hospital to the laboratory within 30 min of death. They were then portioned, frozen in liquid nitrogen, and stored at -80°C.

Metabolism of [¹⁴C]SB-209247 in the Rat. Adult male Wistar rats (200-250 g) were anesthetized with urethane (1.4 g/ml in isotonic saline; 1.0 ml/kg i.p.) and cannulated via the trachea, jugular vein, and common bile duct. [¹⁴C]SB-209247 (100 μmol/kg; specific activity 0.069 μCi/μmol) freshly dissolved in dimethyl sulfoxide (200 μl) was injected intravenously over 10 min. Bile was collected as hourly fractions for 5 h into preweighed Eppendorf tubes cooled on ice, acidified with 4% (v/v; pH 1.8) orthophosphoric acid (10:1 bile/acid, v/v), and protected from light in amber-glass vials (SB-209247 is photolabile in daylight). Aliquots were either assayed for radioactivity by liquid scintillation counting (10 μl, two times) or analyzed by parallel LC-MS and radiochromatography (50 μl). The bile was stored at -80°C. Enzymatic hydrolysis of glucuronides in bile samples (200 μl) was achieved by incubation with H-2 preparation (20 μl; β-glucuronidase, 131 units/μl) in 0.1 M sodium acetate buffer (pH 5) at 37°C for 4 h. The hydrolysate was analyzed by radiometric HPLC. Residues of radiolabeled material in tissues (brain, heart, kidney, liver, lung, and spleen) were measured as described previously (Maggs et al., 2000). Portions of liver (approximately 100 mg) were homogenized in Hanks' balanced salt solution, protein was precipitated from the homogenates with 3 volumes of acetonitrile (3 ml), and radiolabeled material bound irreversibly to the pelleted protein (750g for 15 min) was estimated by exhaustive solvent extraction as described below.

Stability of Acyl Glucuronide Metabolites. Bile (0- to 1-h collection) from a rat given [¹⁴C]SB-209247 (100 μmol/kg; specific activity 0.069 μCi/μmol) was adjusted to pH 5.0 or 7.4 with 0.1 M sodium phosphate buffer and mixed with a solution of estrone glucuronide in water (200 μg/ml; 3:2 v/v). An aliquot of the resulting mixture was analyzed immediately (time = 0) by LC-MS with selected ion monitoring of the glucuronides of estrone (16.5 min), SB-209247 (32 min), and SB-209247 sulfoxide (32.5 min) at m/z 445, 634, and 650 ([M - 1]^-), respectively. The remainder was kept at 37°C. Samples were removed and analyzed at intervals for 24 h. The mass chromatogram peak area for each acyl glucuronide was expressed as the ratio metabolite/estrone glucuronide.

Metabolism of [¹⁴C]SB-209247 by Hepatocytes. Hepatocytes were isolated from whole livers of adult male Wistar rats (180-200 g) by two-step collagenase perfusion (Tettey et al., 1999) and from the cordate lobe of adult male beagle dogs by a three-step perfusion technique derived from the method of Strom et al. (1982). The viability of the cell suspensions, typically ≥88% and ≥98%, respectively, was determined by trypan blue exclusion.

Freshly isolated rat hepatocytes (2.4 × 10⁶ viable cells/ml) suspended in Krebs-Henseleit buffer (pH 7.4) were incubated with [¹⁴C]SB-209247 (final concentration, 100 μM; 1 μCi) dissolved in dimethyl sulfoxide (final concentration, 0.1% v/v). The total volume was 5 ml. Incubations were carried out on four separate occasions in rotating 50-ml round-bottom flasks at 37°C under an atmosphere of O₂ and CO₂ (95:5 v/v). [¹⁴C]SB-209247 was also incubated in the absence of hepatocytes. After 1 or 3 h, ice-cold acetonitrile (15 ml) was added to the incubations and they were placed on ice for 15 min. The resulting precipitate was sedimented at 750g, and the unbound radiolabeled material associated with it was extracted with acetonitrile (3 ml, two times). The supernatant and extracts were combined, evaporated to dryness under N₂ at 40°C, and reconstituted in acetonitrile (200 μl) for analysis by LC-MS and radiochromatography.

Freshly isolated dog hepatocytes in Williams' medium supplemented with newborn calf serum (10% v/v), insulin (0.1 μg/ml), and antibiotics (penicillin, streptomycin, and neomycin sulfate; each 100 μg/ml) were seeded in six-well collagen-coated plates (5 × 10⁵ cells/ml, 3 ml). They were allowed to adhere under an atmosphere of O₂ and CO₂ (95:5 v/v) at 37°C. The medium was changed to serum- and insulin-free Williams' medium, and the cells were incubated with [¹⁴C]SB-209247 (final concentration, 100 μM; 0.6 μCi) dissolved in dimethyl sulfoxide (final concentration, 0.1% v/v) for 4 h. Incubations were terminated with ice-cold acetonitrile (12 ml) and acidified with 17% (v/v) orthophosphoric acid (1:10 v/v). The culture medium was removed and discarded. The hepatocytes were recovered from the well plate with a cell scraper, left on ice for 15 min, and centrifuged. The supernatant and two acetonitrile extracts (3 ml) of the cell pellet were combined, evaporated to dryness under N₂ at 40°C, and reconstituted in acetonitrile (200 μl) for analysis. The cell pellets from both sets of incubations were retained for determination of irreversibly bound radiolabeled material.

Toxicity of SB-209247 to Rat and Dog Liver Slices. Precision-cut liver slices were prepared from one male beagle dog (10.5 kg) and two male Sprague-Dawley rats. The dog was anesthetized with acepromazine and pentobarbital and then exsanguinated. The rats were euthanized with carbon dioxide and exsanguinated. Livers were kept in ice-cold Krebs-Henseleit buffer (pH 7.4) until they were processed. Briefly, 8-mm cores produced with a stainless steel coring tool were sliced with a Krumdieck precision slicer (Krumdieck et al., 1980). Slices (approximately 250 μm thick, 11-15 mg wet weight) were collected, placed individually upon stainless steel mesh screens, and incubated in Waymouth's medium containing 25 mM HEPES and 10 mM glucose in a rolling incubator under an atmosphere of O₂ and CO₂ (95:5 v/v) at 37°C. Slices were incubated in prewarmed medium (1.7 ml) for 1 h to equilibrate. This medium was replaced with prewarmed medium (1.7 ml) containing either dimethyl sulfoxide (1% v/v), SB-209247 in dimethyl sulfoxide (final drug concentration, 10 μM, 100 μM, or 1 mM), or acetaminophen dissolved directly in medium (final concentration, 27 or 90 mM). Slices were incubated in duplicate for 2, 4, or 24 h, after which the medium was removed and assayed for LDH leakage spectrophotometrically (Bergmeyer and Bernt, 1974). Each slice was sonicated in phosphate-buffered saline (pH 7.4; 1 ml) on ice for 30 s to liberate the remaining LDH.

Preparation of Microsomes. Livers were removed from adult male Wistar rats immediately after they were killed by cervical dislocation and homogenized individually in 2 volumes of ice-cold 67 mM potassium phosphate buffer (pH 7.5) containing 0.15 M potassium chloride. Samples (10-20 g) of frozen human and dog liver stored at -80°C were also homogenized individually. Microsomal fractions were prepared according to the method of Gill et al. (1995). Protein concentrations were determined by the method of Lowry et al. (1951). Equal amounts of protein from four rat, dog, or human livers were pooled.

Microsomal Oxidation of [¹⁴C]SB-209247. Incubations were carried out in 4-ml amber-glass vials. [¹⁴C]SB-209247 (final concentration, 100 μM or 250 μM; 0.2 μCi) dissolved in dimethyl sulfoxide (1 μl) was added to microsomes (1 mg of protein) in 67 mM phosphate buffer (pH 7.5) containing 10 mM MgCl₂ to give a final volume of 1 ml. Following preincubation at 37°C for 2 min in a shaking water bath, the reaction was initiated by addition of NADPH (final concentration, 1 mM). NADPH was omitted from control incubations. After 30 min, the reaction was terminated with ethyl acetate (3 ml). The supernatants of two 15-min extractions of the protein pellet (750g for 15 min) were pooled and evaporated to dryness under a stream of N₂ at 40°C. The residue was reconstituted in acetonitrile (300 μl) and stored at -80°C until analyzed by radiometric HPLC and/or LC-MS. The protein pellets were retained for determination of irreversibly bound radiolabeled material.

Microsomal Glucuronidation of [¹⁴C]SB-209247. Incubations were carried out in 4-ml amber-glass vials. [¹⁴C]SB-209247 (final concentration, 100 μM or 250 μM; 0.2 μCi) dissolved in dimethyl sulfoxide (1 μl) was added to microsomes (1 mg of protein) in 50 mM Tris-HCl buffer (pH 7.5) containing 10 mM MgCl₂ and Brij 58 surfactant (0.1 mg/ml). Following preincubation at 37°C for 2 min in a shaking water bath, the reaction was initiated by addition of UDPGA (final concentration, 3 mM). The final volume was 1 ml. UDPGA was omitted from control incubations.

After 30 min, the reaction was terminated with ice-cold acetonitrile (3 ml) and the solution acidified with 17% (v/v) orthophosphoric acid (1:10 v/v) to stabilize the acyl glucuronide. The mixture was placed on ice for 15 min. The protein pellet was extracted twice with acetonitrile, and the pooled extracts were evaporated to dryness under N₂ at 40°C. The residue was reconstituted in acetonitrile (200 μl) and stored at -80°C until analyzed by radiometric HPLC and/or LC-MS.

Kinetics of [¹⁴C]SB-209247 Glucuronidation. Incubations of drug (1-250 μM; 0.23 μCi) and UDPGA (3 mM) were carried out for 30 min as described above, using rat, dog, and human liver microsomes (1 mg/ml protein) in a final volume of 200 μl. Reactions were terminated with ice-cold acetonitrile (600 μl), and the glucuronide was stabilized with 17% (v/v) orthophosphoric acid (1:10, v/v). Dry residues of the pooled acetonitrile extracts were reconstituted in acetonitrile (200 μl) for radiometric HPLC. The area of the radioactivity peak corresponding to [¹⁴C]SB-209247 glucuronide was expressed as a fraction of eluted radioactivity. Apparent K_m and V_max values were calculated by fitting the glucuronidation activities (duplicates for nine substrate concentrations between 1 and 250 μM, inclusive) to the Michaelis-Menten equation using nonlinear least-squares regression analysis (Grafit software; Sigma-Aldrich, St. Louis, MO). Incubations were performed on four occasions. The microsomal protein pellets from these incubations were retained for determination of irreversibly bound radiolabeled material.

Covalent Binding of [¹⁴C]SB-209247. Radiolabeled material bound irreversibly to hepatocytes, hepatic microsomes, and pellets of liver protein was measured after removal of the unbound drug by exhaustive solvent extraction. An aliquot (3 ml) of acetonitrile was added to the protein pellet and mixed thoroughly, initially by vortexing and then on a rotary mixer for 20 min. The protein was separated by centrifugation at 750g for 15 min. The extraction procedure was repeated with 3-ml volumes of acetonitrile (three times), methanol, and finally methanol/water (7:3 v/v) until negligible amounts of radioactivity were detected in the supernatant by liquid scintillation counting. The protein was then dissolved in 1 M NaOH (0.3 ml) by warming for 16 h at 50°C. Aliquots (50 μl) of the solution were assayed for radioactivity by scintillation counting. Microsomal and precipitated liver protein was determined by the method of Lowry et al. (1951) using bovine serum albumin as a standard, and hepatocyte protein by the method of Bradford (1976).

High-Performance Liquid Chromatography. Analyses were performed with a Zorbax 5-μm C-18 column (25 × 0.46 cm; Phenomenex, Macclesfield, Cheshire, UK) connected to a Kontron 360 automated sample injector (Kontron Instruments, Watford, Hertfordshire, UK) and a Kontron 325 system pump. The eluate was acetonitrile (10-70% over 55 min) in 50 mM ammonium acetate (pH 6.9) at a flow rate of 0.9 ml/min. The eluent was monitored with a Kontron 332 UV detector (λ = 254 nm). Radiolabeled analytes were quantified with a Radiomatic A250 flow detector (PerkinElmer Life and Analytical Sciences, Pangbourne, Berkshire, UK) as described previously (Maggs et al., 2000).

Liquid Chromatography-Mass Spectrometry. A Quattro II mass spectrometer (Micromass MS Technologies, Manchester, UK) fitted with the standard coaxial electrospray source was used in the negative-ion mode except as indicated. The LC system consisted of two Jasco PU980 pumps (Jasco UK, Great Dunmow, Essex, UK) and a Jasco HG-980-30 mixing module. Analytes (≥50 × 10³ dpm for radiolabeled material) were resolved on a Zorbax 5-μm C-18 column with a gradient of acetonitrile (10-70% over 55 min) in 50 mM ammonium acetate, pH 6.9. The flow rate was 0.9 ml/min. Eluate split-flow to the LC-MS interface was ca. 40 μl/min. Nitrogen was used as the nebulizing and drying gas. The interface temperature was 70°C; the capillary voltage, 3.9 kV. Spectra were acquired between m/z 100 and 1050 over a scan duration of 5 s. Fragmentation of analyte ions was achieved at a cone voltage of 70 V or by collision-induced decomposition at 50 eV using argon as collision gas. Selected ion monitoring was performed with a dwell time of 200 ms and an interchannel delay of 20 ms. Data were processed with MassLynx 2.0 software (Micromass MS Technologies).

Results

Metabolism of [¹⁴C]SB-209247 in the Rat. Following i.v. administration of [¹⁴C]SB-209247 (100 μmol/kg), 48.85 ± 5.8%, 24.98 ± 1.9%, and 6.05 ± 1.8% (mean ± S.E.M., n = 4) of the radioactivity was recovered in the first, second, and third hourly collections of bile, respectively; the recovery over 0 to 5 h was 81.95 ± 6.9%.

The liver contained 2.8 ± 1.5% of the dose at 5 h. Of this percentage, 36.8 ± 0.1% (1.03 ± 0.14% of the dose) remained associated with the precipitated protein fraction after exhaustive solvent extraction and was therefore judged to be bound irreversibly, representing approximately 25 nmol Eq/g liver. Less than 0.5% of the dose was found in each of the following organs: brain, heart, kidneys, lungs, and spleen.

Two major radiolabeled biliary metabolites (II and IV) were resolved by HPLC (Fig. 2A; Table 1). The same metabolites were found in bile following administration of [¹⁴C]SB-209247 at 10 μmol/kg (data not shown). Parallel LC-MS analysis located coeluting chlorine isotope peaks in the negative-ion mass chromatograms for m/z 650/652 (II) and 634/636 (IV), which represented [M - 1]^- for the acyl glucuronides of [¹⁴C]SB-209247 sulfoxide and [¹⁴C]SB-209247, respectively (Fig. 1); corresponding protonated molecules were obtained at m/z 652/654 and 636/638. The fraction of the dose eliminated as the two acyl glucuronides in bile over 3 h was 56.8 ± 1.8% (n = 4). They were identified as glucuronides by diagnostic elimination of dehydroglucuronic acid ([M -1 - 176]^-) at elevated cone voltages, yielding aglycone anions at m/z 474/476 and 458/460 (Fig. 3). LC-tandem mass spectrometry of the protonated sulfoxide conjugate (m/z 652) gave the expected product ion at m/z 476. Both metabolites underwent the pH-dependent isomerization, i.e., acyl group migration, characteristic of acyl glucuronides (Fig. 4), the transacylation of SB-209247 sulfoxide glucuronide at pH 7.4 being somewhat faster than that of SB-209247 glucuronide (Fig. 5). The single peaks in the mass chromatograms of bile incubated at pH 5.0 (Fig. 4) verified that the β-1-O-acyl glucuronides were stable under acidic conditions. The products of transacylation seen at pH 7.4 are likely to have comprised unresolved α- and β-anomeric pairs (Corcoran et al., 2001); C-1 anomerization of the positional isomers of β-1-O-acyl glucuronides occurs at low pH values but increases with pH (King and Dickinson, 1991). Enzymatic hydrolysis of the biliary conjugates with H. pomatia preparation liberated radiolabeled aglycones (Fig. 2A) that were identified by coelution with authentic standards as SB-209247 sulfoxide (III) and the parent drug (V). Essentially complete hydrolysis confirmed the glucuronides' stability in acidified bile; acyl migration isomers are resistant to β-glucuronidase (Bailey and Dickinson, 2003).

The most polar of the annotated radiolabeled metabolites (I; Fig. 2A) coeluted with putative chlorine isotope peaks in the negative-ion mass chromatograms for m/z 745/747, which corresponded to [M - 1]^- for a derivative of oxygenated SB-209247 in which one of the chlorines had been substituted with glutathione. These peaks were absent from rat bile collected immediately before administration of the drug, but they did not yield diagnostic fragments for glutathione adducts at higher cone voltages. This metabolite was not observed in either rat or dog hepatocyte incubations.

Metabolism of [¹⁴C]SB-209247 by Hepatocytes. Radiometric HPLC analysis of combined supernatant and extracts from 1- and 3-h rat hepatocyte incubations with [¹⁴C]SB-209247 (100 μM) revealed extensive turnover (Fig. 2B; Table 2). No transformations occurred in the absence of cells. The two major metabolites corresponded to peaks in the negative-ion mass chromatograms for m/z 474/476 and 634/636 that were assigned to SB-209247 sulfoxide (III) and SB-209247 acyl glucuronide (IV), respectively. Recovery of incubated radioactivity as [¹⁴C]SB-209247 glucuronide and sulfoxide from the rat hepatocyte incubations decreased between 1 h and 3 h (Table 2). The sulfoxide glucuronide was present in only very small proportions. A number of unidentified metabolites of greater polarity than the sulfoxide glucuronide, which were only seen as minor/trace metabolites in rat bile, were more prominent in hepatocyte incubations. Dog hepatocytes transformed [¹⁴C]SB-209247 during 4-h incubations to a lesser extent than did rat hepatocytes over 3 h but still produced SB-209247 acyl glucuronide and sulfoxide as the principal metabolites (Fig. 2C; Table 2). However, no sulfoxide glucuronide was detected in the incubations of dog hepatocytes.

Toxicity of SB-209247 to Rat and Dog Liver Slices. Control precision-cut dog and rat liver slices remained viable in Waymouth's medium for 24 h as assessed by low (5-10% of total) levels of LDH leakage into the medium. SB-209247 caused dose- and time-dependant cellular injury to both the rat and dog tissue over 24 h (Fig. 6). Similar results were obtained when leakage of aspartate aminotransferase was monitored (data not shown). Although LDH leakage, especially from the rat liver tissue, may have occurred within 2 h at the highest concentration of SB-209247 (1 mM), severe toxicity, entailing leakage of 93 to 94% of the LDH (mean of 87 and 100% for rat slices; 89 and 96% for dog slices), was only evident with SB-209247 at 1 mM over 24 h. In a positive-control incubation, acetaminophen, at 27 mM, caused only 16 to 21% leakage of LDH from rat and dog liver slices over 24 h, but 100% leakage was attained at 90 mM.

Microsomal Metabolism of [¹⁴C]SB-209247. Rat, dog, and human liver microsomes in rank order in the presence of NADPH metabolized [¹⁴C]SB-209247 (100 μM) to one radiolabeled product (Fig. 7A; Table 3). This product was identified by LC-MS ([M - 1]^- at m/z 474/476) and cochromatography as SB-209247 sulfoxide (III). No metabolism was observed in the absence of added NADPH. The recovery of incubated radioactivity in the ethyl acetate extracts was 79 to 90%.

Rat, dog, and human microsomes supplemented with UDPGA metabolized [¹⁴C]SB-209247 (100 or 250 μM) to a single glucuronide (Fig. 7B; Table 4). The identity of this conjugate was confirmed by LC-MS ([M - 1]^- at m/z 634/636). Hydrolysis with the H-2 β-glucuronidase preparation yielded only [¹⁴C]SB-209247. No glucuronidation was observed in the absence of added UDPGA. The recovery of incubated radioactivity in the acetonitrile extracts was 74 to 89%.

SB-209247 Glucuronidation Kinetics. The glucuronidation of [¹⁴C]SB-209247 conformed to Michaelis-Menten kinetics in hepatic microsomes from all three species (Fig. 8A): the initial rate was linear for substrate concentration (1-50 μM), time (5-35 min), and protein concentration (0.1-2 mg/ml) (data for rate versus time and rate versus protein concentration not shown). For pooled dog, rat, and human microsomes, the mean V_max values were 2.6 ± 0.1, 1.2 ± 0.1, and 0.4 ± 0.0 nmol/min/mg protein (mean ± S.E.M., n = 4), respectively; the corresponding apparent K_m values were 69 ± 11.2, 44.3 ± 7.6, and 31.1 ± 3.8 μM. The V_max for dog liver microsomes was significantly greater than that for either rat or human (p < 0.05 by Mann-Whitney test). There was also a significant difference (p < 0.05) in apparent K_m of SB-209247 glucuronidation between the dog and human microsomes and the dog and rat microsomes.

Covalent Binding of [¹⁴C]SB-209247 to Hepatic Protein. In rat and dog hepatocytes incubated with [¹⁴C]SB-209247 for 1 to 4 h, a small fraction of the radiolabeled material became irreversibly bound to cellular material: in rat hepatocytes, 1.5 ± 0.3 and 2.2 ± 1.0% (mean ± S.E.M., n = 4) of the incubated radioactivity after 1 and 3 h, respectively; in dog hepatocytes, 1.0 ± 0.2% after 4 h. Neither dog, rat, nor human liver microsomes catalyzed NADPH-dependent irreversible binding of [¹⁴C]SB-209247 (100 or 250 μM) to microsomal protein (data not shown).

[¹⁴C]SB-209247 (1-250 μM) incubated with dog, human, and rat hepatic microsomes in the presence of UDPGA underwent irreversible binding that was both time-dependent (5-35 min; data not shown) and substrate concentration-dependent (Fig. 8B). At a substrate concentration of 100 μM, the binding was only significantly (p < 0.05 by the Mann-Whitney test) increased over binding in the absence of UDPGA with dog microsomes, but it was increased with microsomes from all three species at 250 μM. There was a significant difference between the binding of [¹⁴C]SB-209247 to dog and human microsomes at a substrate concentration of 50 μM (p < 0.05), but none of this binding appeared to be UDPGA-dependent. Otherwise, there were no significant differences of binding among the three species, notwithstanding the apparent trend toward greater binding to dog than either rat or human microsomes.

Discussion

We have shown that SB-209247, a carboxylic acid NSAID associated in beagle dogs with a predominantly inflammatory hepatopathy, is metabolized extensively to an unstable acyl glucuronide in the rat and by rat, dog, and human hepatic microsomes. SB-209247 was also metabolized in rats to the acyl glucuronide of its sulfoxide. In rat bile, under mild alkaline conditions, the sulfoxide conjugate was the more labile of the two glucuronides. Isolated rat and dog hepatocytes replicated most of these pathways, but sulfoxide glucuronidation was much lower in rat hepatocytes and not detected in dog hepatocytes. Although a great variety of xenobiotics with alkylcarboxyl or alkenylcarboxyl moieties (Wieland et al., 2000; Soars et al., 2001; Prueksaritanont et al., 2002) are substrates for acyl glucuronidation, this would appear to be the first report of glucuronidation of a propenoic acid group. LC-MS profiling of rat bile for indications of an alternative bioactivation pathway failed to detect S-glutathione thioester conjugates of SB-209247 and its sulfoxide derived from putative acyl-CoA intermediates (Boelsterli, 2002).

SB-209247 has potent oral anti-inflammatory activity, having an ED₅₀ of 32 μmol/kg in the murine arachidonic acid-induced ear inflammation model (Daines et al., 1996). Inasmuch as the two acyl glucuronide metabolites were eliminated in rat bile when doses of 10 and 100 μmol/kg SB-209247 were given i.v., it seems likely that they are both formed at pharmacologically active doses.

The kinetic parameters of glucuronidation indicated that dog liver microsomes have a greater capacity to form SB-209247 acyl glucuronide than either rat or human microsomes. The rank order dog > human for V_max of this reaction conformed broadly with the finding of Soars et al. (2001) that the maximum rates of glucuronidation of a structurally diverse group of drugs, including five carboxylic acids (one benzoic and four alkylcarboxylic acids), were at least 5-fold greater in dog than in human liver microsomes. They also found that the affinity of microsomal UGT for these drugs did not exhibit this relationship but instead was specific to the compound. Consequently, their estimates of higher intrinsic clearance by dog microsomes, as calculated from V_max/K_m, were principally a reflection of differences in V_max. However, when the intrinsic clearances of SB-209247 by dog, rat, and human liver microsomes, namely 37, 27, and 13 μl/min/mg, respectively, were estimated, it was found that the disparities in V_max values were partly counterbalanced by an opposite trend in K_m. Other, smaller scale comparisons of glucuronidation in hepatic microsomes from these three species have not found a consistent species-related rank order of UGT activity against phenolic and carboxylate substrates (Ward et al., 1999; Prueksaritanont et al., 2002). Additionally, a species-related difference between the extents of acyl glucuronidation of a benzoic acid drug in rat and dog liver slices is reversed in vivo (Howell et al., 2001). An instance of a benzoic acid acyl glucuronide being formed in humans but in neither dogs nor rats has been reported (Dahms et al., 1997).

At a drug concentration (100 μM) favoring both S-oxidation and acyl glucuronidation in hepatocytes, SB-209247 incubated with rat and dog hepatocytes underwent irreversible reaction with cellular material. From the results of the microsomal studies, it was deduced that this binding was likely to have been a consequence of bioactivation via glucuronidation rather than oxidative transformation of either the parent compound or sulfoxide metabolite. Whereas S-alkenyl sulfoxides can be chemically reactive metabolites, as in the case of the α,β-unsaturated 1,2-dichlorovinyl-cysteine sulfoxide, which acts as a Michael acceptor for glutathione in vivo (Sausen and Elfarra, 1991), extensive formation of SB-209247 sulfoxide was not associated with measurable irreversible binding in microsomal incubations, although a potentially analogous metabolite of SB-209247 (100 μmol/kg) resulting apparently from S-oxidation and substitution by glutathione of one of the chlorines was detected in rat bile. Nevertheless, the significantly higher rate of SB-209247 glucuronidation by dog liver microsomes did not translate into a commensurately higher rate of binding of radiolabel to microsomal protein, although the apparent trend was toward a greater extent of binding in the dog as opposed to rat and human liver microsomes. Nor was the disparity in microsomal glucuronidation kinetics represented by a significant difference between the extents of binding of the drug in rat and dog hepatocytes.

The concentration-dependent toxicity of SB-209247 against rat and dog liver slices, assessed as cellular necrosis via cytoplasmic enzyme leakage, did not exhibit the clear species difference in hepatotoxicity obtained during chronic (28-day) toxicity studies: exposure over 24 h to drug concentrations (10-1000 μM) that were approximately 3 to 150 times the maximum plasma concentrations measured in the dogs (6.8-32.9 μM from doses of 0.13-2.17 mmol/kg/day; GlaxoSmithKline, unpublished data) revealed that hepatocytes in slices from these species are equally sensitive to SB-209247. This is not irreconcilable with the drug's toxicity in vivo because the dominant feature of the hepatopathy seen in dogs, but not in rats, was neutrophil infiltration unrelated to dose, with only minimal necrosis being observed even in the more severely affected dogs. Neither of these effects, although not necessarily the toxicity toward liver slices, can be easily related to a combination of hepatic microsomal glucuronidation (dog > rat) and protein adduction (dog ≅ rat) as measured here. It is, of course, possible that the drug's toxicity in vivo is a function of a particular subcellular fraction of the protein binding. In this context, it is notable that the cytotoxicity of diclofenac toward isolated rat hepatocytes has been linked with oxidative bioactivation rather than protein binding by the drug's acyl glucuronide (Kretz-Rommel and Boelsterli, 1993; Bort et al., 1999). However, SB-209247 was not a substrate for NADPH-dependent activation to protein-binding intermediates in either dog, rat, or human hepatic microsomes. Acyl glucuronides might also be allowed a hypothetical causal role in the adverse reactions of SB-209247 if it is assumed the complete mechanism was not represented at the hepatocellular level in these experiments. Thus, mechanisms requiring accumulation of drug-protein adducts during chronic exposure to elicit the inflammatory hepatopathy would not have been responsible for the cytotoxicity of SB-209247 observed over 24 h.

A local inflammatory response to chemically induced hepatocellular injury, which commonly includes neutrophil recruitment by chemokines, may seemingly contribute to the tissue damage (Smith et al., 1998) or promote regeneration (James et al., 2003). With acetaminophen, neutrophils infiltrate the liver parallel to or slightly after the appearance of necrosis, but in this and other instances, it has been suggested that the inflammation is sufficient to maintain recruitment of neutrophils but not to cause additional injury (Copple et al., 2003; James et al., 2003). In the case of SB-209247, however, the inflammation apparently neither requires induction by persistent hepatocellular necrosis nor results in such damage in most cases. As a high-affinity selective antagonist of LTB₄ binding to neutrophils (Daines et al., 1996), SB-209247 would be expected to block any LTB₄-mediated neutrophilic sequestration in response to drug-induced liver injury (Fretland et al., 1995; Brooks and Summers, 1996; Davis et al., 2000). The inhibitory activities of the drug's metabolites are unknown; the ether glucuronide of one LTB₄ receptor antagonist (BIIL 284) does retain considerable pharmacological activity (Birke et al., 2001). The possibility of a pharmacological component in the species-selective neutrophilic inflammatory response to SB-209247 is suggested by the appreciably lower affinity of canine than rat neutrophil LTB₄ receptors for BIIL 284 (Birke et al., 2001). Nevertheless, a hypothetical initiating or modulating role for acyl glucuronides of SB-209247 in the hepatic inflammation can be constructed by analogy with the actions of other drugs. Thus, the acyl glucuronide of mycophenolic acid, but not the aglycone, induces release of proinflammatory cytokines from leukocytes (Wieland et al., 2000). Certain of these mediators can up-regulate neutrophil adhesion molecules involved in sequestration in the liver (Essani et al., 1997).

In conclusion, although acyl glucuronidation of SB-209247 is a major hepatocellular biotransformation that leads to protein binding, and SB-209247 is toxic in canine liver slices, these findings alone are insufficient to propose that glucuronidation and/or protein adduction initiate the drug's distinctive hepatopathy in dogs. The possibility that the cytotoxicity obtained with both dog and rat liver slices is connected with irreversible binding of SB-209247 to hepatocellular protein is not excluded. However, the conventional hypothesis of the role of acyl glucuronides in hepatotoxicity, protein adduction leading directly or indirectly to cellular death (Boelsterli, 2002; Bailey and Dickinson, 2003), appears not to apply to SB-209247 because the predominant feature of its hepatotoxicity is accumulation of neutrophils in the apparent absence of tissue damage. The minimal necrosis observed in the more severely affected livers might even have been a consequence of cytotoxic neutrophil activity rather than the drug's hepatocytotoxicity. Alternatively, the inflammatory response associated with chronic dosing might have been triggered by damaged hepatocytes which, in most of the dogs, were removed within the period of drug administration. Clearly, the possibility that SB-209247 or its acyl glucuronide can initiate a species-selective hepatic recruitment of inflammatory cells by pathways not involving tissue damage merits further consideration.