Indirect Cytotoxicity of Flucloxacillin toward Human Biliary Epithelium

Flucloxacillin, an isoxazolyl-penicillin, causes cholestasis and biliary epithelium injury. The aim of the study was to determine whether flucloxacill...
0 downloads 0 Views 90KB Size
Download PDF
694

Chem. Res. Toxicol. 2001, 14, 694-701

Indirect Cytotoxicity of Flucloxacillin toward Human Biliary Epithelium via Metabolite Formation in Hepatocytes Fatima Lakehal,†,‡,§ Patrick M. Dansette,| Laurent Becquemont,⊥ Elisabeth Lasnier,†,# Roland Delelo,†,3 Pierre Balladur,†,3 Raoul Poupon,†,O Philippe H. Beaune,‡ and Chantal Housset*,†,O Unite´ INSERM U402, Faculte´ de Me´ decine Saint-Antoine, Unite´ INSERM U490, CNRS UMR 8601, Universite´ Rene´ Descartes, Services de Pharmacologie, Biochimie, Chirurgie Ge´ ne´ rale and Centre de Recherches Chirurgicales, He´ pato-Gastroente´ rologie, Hoˆ pital Saint-Antoine, Paris, France Received November 27, 2000

Flucloxacillin, an isoxazolyl-penicillin, causes cholestasis and biliary epithelium injury. The aim of the study was to determine whether flucloxacillin, either directly or through metabolite formation, may induce cytotoxicity in hepatic or biliary cells. Cytotoxicity was assessed by lactate dehydrogenase release in primary cultures of human hepatocytes and of gallbladderderived biliary epithelial cells (BEC). Metabolite production in microsome and cell preparations was analyzed by chromatography, nuclear magnetic resonance spectroscopy, and mass spectrometry. While flucloxacillin induced no direct cytotoxicity in any of the hepatocyte (n ) 12) and BEC (n ) 19) preparations, the conditioned media from cultured hepatocytes preincubated with flucloxacillin (50-500 mg/L) triggered a significant increase in lactate dehydrogenase release over controls in ∼50% of BEC preparations (7/12), and this effect depended upon flucloxacillin concentration. Remaining BEC preparations exhibited no toxic response. Cytotoxicity in BEC preparations (9/13) was also induced by the supernatants of human liver microsomes and of recombinant human cytochrome P450 (CYP)3A4 preincubated with flucloxacillin (500 mg/L). Supernatants from both liver microsome and CYP3A4 preparations contained one major metabolite which was identified as 5′-hydroxymethylflucloxacillin. The production of this metabolite was inhibited following CYP3A4 inhibition by troleandomycin in human liver microsomes, and markedly enhanced following CYP3A induction by dexamethasone in rat liver microsomes. As opposed to BEC, cultured hepatocytes displayed significant CYP3A activity and produced low amounts of this metabolite. The purified metabolite (0.01-5 mg/L) exerted toxic effects in BEC but not in hepatocytes. In conclusion, hepatocytes mainly via CYP3A4 activity, generate flucloxacillin metabolite(s) including 5′-hydroxymethylflucloxacillin that may induce cytotoxicity in susceptible BEC. These metabolic events may contribute to the pathogenesis of drug-induced cholangiopathies.

Introduction Evidence indicates that drug-induced prolonged cholestasis is associated with bile duct destruction (13), while the pathogenesis underlying this type of adverse reaction remains poorly understood. The most commonly reported drug to cause severe cholestatic liver injury so far is an isoxazolyl-penicillin called flucloxacillin (4), which has been widely prescribed since the mid 1980s to treat staphylococcal infections in many countries, with the exception of the United States. In 1994, approximately 600 cases of flucloxacillin-associated cholestatic * To whom correspondence should be addressed. Phone: 33-1-40 01 13 59. Fax: 33-1-40 01 14 99. E-mail: chantal.housset@ st-antoine.inserm.fr. † Unite ´ INSERM U402. ‡ Unite ´ INSERM U490. § Present address: Fatima Lakehal, Ph.D., University of Washington, Health Sciences building, Department of Pharmaceutics, Seattle, WA. | CNRS UMR 8601. ⊥ Service de Pharmacologie. # Service de Biochimie. 3 Chirurgie Ge ´ ne´rale and Centre de Recherches Chirurgicales. O He ´ pato-Gastroente´rologie.

hepatitis had been reported to the Australian Adverse Drug Reaction Advisory Committe and to the Collaborative Center for International Drug Monitoring (World Health Organization) in Sweden (5). High daily doses and prolonged intake of flucloxacillin have been identified as risk factors for developing the reaction (6, 7). Flucloxacillin-induced hepatitis is usually cholestatic in nature. The main histological features are those of cholestasis, with minimal or no hepatic necrosis. Abnormalities of interlobular bile ducts including degenerative lesions of the biliary epithelium are seen in a majority of cases, while inflammatory reaction is moderate or absent. Most often, repair of bile ducts manifested by ductular proliferation results in healing. In some cases, however, severe bile duct damage leads to paucity of bile ducts and to biliary cirrhosis (8, 9). The mechanisms responsible for this reaction are unknown. The relatively high frequency of the reaction has suggested that a reactive metabolite may be commonly produced or that flucloxacillin itself is intrinsically toxic (8). It was previously shown by the analysis of rat and human body fluids, that flucloxacillin as well as other isoxazolyl-penicillins, give rise to 5-hy-

10.1021/tx0002435 CCC: $20.00 © 2001 American Chemical Society Published on Web 05/10/2001

Biliary Cytotoxicity of Flucloxacillin

droxymethyl derivatives, presumably formed by hydroxylation of the 5-methyl group in the isoxazole ring (1013). It is postulated, although not yet proven, that this reaction results from the monooxygenase activity of cytochrome P4501 (CYP) enzymes. The aim of the present study, was to determine whether flucloxacillin may induce a cytotoxic response in cells of the liver and biliary tract, either directly or indirectly through the formation of toxic metabolites. Cell responses were examined in primary cultures of human hepatocytes and of gallbladder-derived biliary epithelial cells1 (BEC) that share with intra-hepatic biliary epithelial cells major common functional and phenotypic features (14, 15), including drug-metabolizing enzyme expression (16). Gallbladder-derived BEC may also be a target of injury in patients treated with penicillin derivatives (17) including flucloxacillin (18). The formation of toxic metabolites and flucloxacillin metabolic pathways were further investigated using different types of microsome preparations, by means of chromatographic and spectrometric analyses.

Experimental Procedures Cell Isolation and Primary Culture. Liver and gallbladder samples were obtained from patients who underwent liver or pancreas surgery. The procedure was in agreement with current French legislation. Samples were all taken at a distance from tumors and displayed no significant histological abnormality. BEC were isolated from gallbladders by incubation in 0.075% (w/v) protease type XIV from Streptomyces griseus (Sigma, Saint-Quentin Fallavier, France) for 12 h at 4 °C, as described (19). Hepatocytes were isolated by an established method (20) with modifications. Liver samples weighing 15-60 g were perfused with 0.5 mM ethylenediaminetetraacetic acid and subsequently with 0.8 IU/mL collagenase D (Boehringer Mannheim, Mannheim, Germany), at 37 °C and a flow rate of 8 mL min-1. Isolated hepatocytes were then purified by centrifugation through a 1.13 g/mL-density Percoll solution (21). Cell viability exceeded 90% as tested by erythrosin exclusion. BEC were plated at a density of 106 cells/mL, in 24-well culture dishes coated with type IV collagen (Sigma), and hepatocytes, at 8 × 105 cells/mL, in 12-well culture dishes coated with type I collagen (Becton Dickinson, Le Pont de Claix, France). Unless otherwise specified, BEC were maintained in Dulbecco’s modified Eagle medium and Ham’s F12 (1:1) mixture (DMEM/Ham’s F12, Life Technologies, Cergy Pontoise, France) supplemented with 2% (w/v) Ultroser G (Biosepra, Villeneuve-la-Garenne, France), 2.5 mM glutamine (Eurobio, Les Ulis, France), 5.35 g/L D-glucose, 14 mM NaHCO3, and 100 000 IU-100 mg/L penicillin G-streptomycin, while hepatocytes were maintained in Leibovitz L-15 medium (Sigma) containing 2 g/L bovine serum albumin, 2 mM glutamine, 26 mM NaHCO3, 100 IU/L insulin (Novo Nordisk Pharmaceutique, Boulogne Billancourt, France), 10-6 M hydrocortisone hemisuccinate (Sigma), and 100 000 IU10 mg/L penicillin G-streptomycin. The culture medium was renewed every 24 h. At the beginning of cell treatments described below, cells were washed twice and thereafter maintained in an antibiotic-free culture medium that consisted in DMEM/Ham’s F12 supplemented with the same additives as for both BEC and hepatocyte culture (i.e., 2% Ultroser G, 2 g/L bovine serum albumin, 100 IU/L insulin, 10-6 M hydrocortisone hemisuccinate, 2.5 mM glutamine, 5.35 g/L D-glucose, 14 mM NaHCO3). Cell incubations were all performed under air/CO2 95/5, at 37 °C. Cell Treatments. To test flucloxacillin cytotoxicity and metabolite production in cultured cells, hepatocytes and BEC 1 Abbreviations: BEC, biliary epithelial cell; CYP, cytochrome P450; HLM, human liver microsome; LDH, lactate dehydrogenase.

Chem. Res. Toxicol., Vol. 14, No. 6, 2001 695 at days 2-3 and 2-4 of primary culture, respectively, were incubated in antibiotic-free culture medium with increasing concentrations of flucloxacillin (a gift from SmithKline Beecham Pharmaceuticals, Worthing, United Kingdom), for 24 h. Because serum levels of flucloxacillin peak at 40-230 mg/L in treated patients (12, 22), tested concentrations ranged between 0 and 500 mg/L (∼1 mM). Some of the cell preparations were incubated in parallel with amoxicillin (0-1000 mg/L, i.e. 2.5 mM) instead of flucloxacillin, and in some experiments, hepatocyte incubations were performed in the presence of 50 µM rifampicin, used as an inducer of CYP3A4. Cell supernatants were harvested and centrifuged 5 min at 450g in order to eliminate cell debris before analysis or further use in cytotoxic experiments. To investigate indirect mechanisms of cytotoxicity, BEC at days 2-5 of primary culture, were incubated for 24 h with the conditioned media from flucloxacillin-treated hepatocytes or with the supernatants of flucloxacillin-treated microsomes prepared as below. In addition, hepatocytes and BEC in primary culture were incubated with flucloxacillin purified metabolite (0.01-5 mg/L, i.e., 20 nM to 10 µM). Microsome Treatments. Previously characterized microsome preparations that were from dexamethasone-, methylcholanthrene-, or phenobarbital-treated rat liver (23), from human liver (16), and from yeasts expressing human recombinant CYP isoforms (1A2, 2C9, 3A4, and 3A5) (24), were used to test flucloxacillin metabolite production and/or cytoxicity. Rat liver, human liver or yeast microsomes were added to 100 µL of 0.1 M phosphate buffer (pH 7.4) containing 100 mg/L (i.e., 200 µM) flucloxacillin, to obtain final CYP concentrations of 2 nmol/mL, 0.5 nmol/mL and 50 pmol/mL, respectively. The reaction was started by the addition of a NADPH-generating system, composed of 2 units/mL glucose 6-phosphate dehydrogenase, 10 mM glucose 6-phosphate, and 1 mM NADP. Incubations with liver microsomes and recombinant CYP were carried out under agitation in ambient air, at 37 and 28 °C, respectively. After a time course of 0-60 min, the reaction was interrupted by the addition of 100 µL of acetonitrile in 0.1 M phosphate buffer (pH 7.4), on ice. Proteins were discarded after centrifugation at 19000g for 5 min at 4 °C. NADPH-generating system and/or flucloxacillin were omitted in negative controls. In addition, in some experiments, incubations of human liver microsomes with 100 mg/L flucloxacillin and NADPH-generating system, were carried out for 60 min in the presence of 4% (v/v) tetrahydrofuran (SDS, Peypin, France), used as a specific inhibitor of CYP1A2, of 25 µM sulfaphenazole (Sigma), used as a specific inhibitor of CYP2C9, of 50 µM troleandomycin (TAO) or 25 µM miconazole, as specific inhibitors of CYP3A4. When microsome supernatant was to be tested in BEC cultures, incubation with flucloxacillin was conducted for 2 h at 37 °C in antibiotic-free culture medium and the supernatant was filtersterilized through 0.22 µm filters before use. Cytotoxicity Assay. Lactate dehydrogenase1 (LDH) release, a sensitive test of cytotoxicity in most cell types including BEC (25), was assessed in parallel with morphology. In the end of experiments, the cells were examined under light microscopy. The supernatants were then harvested and the cells were washed twice in phosphate-buffered saline and lysed in 0.2% (v/v) Triton X-100. LDH activities were assayed extemporaneously by monitoring the reduced form of nicotinamide adenine dinucleotide consumption during the conversion of pyruvate to lactate. The rate at which absorbance at 340 nm diminished was monitored at 37 °C on a synchron CX4 analyzer (Beckman Instruments, Gagny, France). LDH release was calculated as the ratio of LDH activity in the supernatant to total activity in the supernatant and cell pellet. Morphological signs of cell death were visible above the level of 20% which was referred to as a threshold to define a cytotoxic response. High-Performance Liquid Chromatography. Cell conditioned media to analyze were first submitted to a solid-phase extraction procedure. The samples were applied to Sep-Pak C18 columns (Waters, Saint-Quentin en Yvelines, France). A wash step with 2 mL of water was followed by elution with 3 mL of

696

Chem. Res. Toxicol., Vol. 14, No. 6, 2001

Lakehal et al.

Figure 1. Chemical structure of flucloxacillin (R ) CH3), and of 5′-hydroxymethylflucloxacillin (R ) CH2OH). acetonitrile, and evaporation to dryness under nitrogen. The solid residue was then dissolved in 200 µL of 0.1 M phosphate buffer (pH 7.4)-acetonitrile (2:1) mixture and centrifuged for 5 min at 12000g, before analysis. The samples derived from this preparation of cell conditioned media and samples of microsome supernatants were analyzed by high-performance liquid chromatography on a Chromatem 380 (Touzart et Matignon, Courtaboeuf, France) gradient system using a C8 MOS Hypersil column (5 µm, 250 × 4.6 mm) (Thermoquest, Les Ulis, France), at a flow rate of 1 mL min-1, using a linear gradient of 0 to 70% B in A + B in 20 min (solution A, 0.1 M ammonium acetate (pH 4.6); solution B, acetonitrile/methanol/H2O, 70:20:10 v/v/ v). UV and visible spectra were recorded on-line by a SpectraFocus scanning spectrophotometer (Thermoquest) over a 30-min time interval at room temperature. Isolation and Characterization of Flucloxacillin Major Metabolite. Flucloxacillin major metabolite was isolated from the supernatant of dexamethasone-treated rat liver microsomes incubated with flucloxacillin (100 mg/L) as described above, in a reaction volume of 20 mL, for 40 min. After acidification with 0.4 mL acetic acid and centrifugation at 3000g, the supernatant was loaded on two in-line Sep-Pak C18 columns and eluted as described above, and 150-µL aliquots were purified on a C8 MOS Hypersil column (5 µm, 250 × 7 mm) at a flow rate of 2.3 mL min-1, using the same gradient as above. Collected fractions containing flucloxacillin and metabolites were lyophilized and then dissolved in 400 µL and D2O before NMR spectroscopy and mass spectrometry. NMR spectra were obtained on a Bruker WM 250 (Wissembourg, France). The H2O peak in D2O is at 4.8 ppm, and the chemical shifts, given in parts per million relative to (CH3)4Si and J (Hz), were as follows. Flucloxacillin: 7.66 (1H, dd, J5′′4′′ ) 8.5 Hz, J5′′F ) 15 Hz, H5′′), 7.55 (1H, d, J3′′4′′ ) 8 Hz, H3′′), 7.39 (1H, dd, J3′′4′′ ) 8 Hz, J4′′5′′ ) 8.5 Hz, H4′′), 5.65 (1H, d, J ) 4 Hz, H5 or H6), 5.57 (1H, d, J ) 4.2 Hz, H5 or H6), 4.22 (1H, s, H3), 2.77 (3H, s, 5′CH3), 1.57 (3H, s, 2CH3), 1.53 (3H, s, 2CH3). 5′-hydroxymethylflucloxacillin: 7.66 (1H, dd, J5′′4′′ ) 8.5 Hz, J5′′F ) 15 Hz, H5′′), 7.54 (1H, d, J3′′4′′ ) 8 Hz, H3′′), 7.39 (1H, dd, J3′′4′′ ) 8 Hz, J4′′5′′ ) 8.5 Hz, H4′′), 5.64 (1H, d, J ) 4 Hz, H5 or H6), 5.60 (1H, d, J ) 4 Hz, H5 or H6), 5.10 (2H, s, 5′CH2OH), 4.26 (1H, s, H3), 1.59 (3H, s, 2CH3), 1.54 (3H, s, 2CH3). Mass spectra were obtained on a JEOL magnetic sector mass spectrometer (Tokyo, Japan), at Ecole Normale Supe´rieure, Paris. Flucloxacillin, EIMS: m/z (%) 453 (0.1, M+), 238 (17, ArCO+), 196 (80, C6H4ClFNO), 160 (100, C6H10NO2S), 114 (65, C6H8NS); CIMS (NH3): 471 (70, M + NH4+), 428 (30, M - CH3), 272 (100, ArCONH2 + NH4+), 255 (35, ArCONH2 + H+). 5′-hydroxymethylflucloxacillin, EIMS: m/z (%); 469 (0.2, M+) 254 (8, ArCO+), 196 (40, C6H4ClFNO), 160 (100, C6H10NO2S), 114 (45, C6H8NS); CIMS (NH3): 487 (35, M + NH4+), 444 (100, M CH3), 288 (17, ArCONH2 + NH4+), 271 (40, ArCONH2 + H+). Figure 1 shows the structure of flucloxacillin and of 5′-hydroxymethylflucloxacillin.

Figure 2. Hepatocyte and BEC toxic response to flucloxacillin. Preparations of (A) human hepatocytes (n ) 12) and of (B) BEC (n ) 19) at days 2-3 and 2-4 of primary culture, respectively, were incubated with increasing concentrations of flucloxacillin for 24 h. Cytotoxicity was assessed by LDH release. Results of experiments performed in triplicate are expressed as means ( SEM. Difference between each concentration and control is not significant. CYP3A Activity Assay. Testosterone 6β-hydroxylase activity was used as an index of CYP3A activity, as previously reported (16, 26). Freshly isolated and cultured cells were incubated in antibiotic-free culture medium containing 100 µM testosterone (Sigma) for 12 h at 37 °C. The surpernatants were then collected, centrifuged to eliminate cell debris, and stored at -80 °C until analysis. The quantity of 6β-hydroxytestosterone in the supernatants was determined by high-performance liquid chromatography, as described (16). Statistics. Comparisons were made using nonparametric Kruskal-Wallis analysis of variance followed by multiple comparison Scheffe test, and the paired Wilcoxon signed rank test, n as appropriate. Linear regression analysis was performed by the method of least squares. p < 0.05 was considered as significant.

Results Hepatocyte and BEC Toxic Responses to Flucloxacillin. To examine flucloxacillin direct toxic effects, we submitted human hepatocytes and gallbladderderived BEC in primary culture to 24-h incubations with increasing concentrations of flucloxacillin. Concentrations up to 500 mg/L (1 mM) caused no significant increase in LDH release over basal level, in 12 hepatocyte (Figure 2A), and 19 BEC preparations (Figure 2B), providing no evidence for direct toxicity. We next tested the hypothesis that flucloxacillin might cause BEC injury indirectly through cell interactions. The conditioned media from nine different hepatocyte preparations incubated with flucloxacillin were prepared and transferred onto autologous and/or heterologous BEC in culture. In 5 of 12 BEC preparations, incubation with hepatocyte conditioned media caused no significant increase in LDH release as compared to controls (Figure 3A, broken line, and B). In these BEC preparations, LDH release never exceeded 20%, irrespective of the concentration of flucloxacillin tested (Figure 3A, broken line). In the remaining prepa-

Biliary Cytotoxicity of Flucloxacillin

Figure 3. BEC toxic response to the conditioned media of flucloxacillin-treated hepatocytes. Human hepatocytes (n ) 9) at days 2-3 of primary culture were incubated with increasing concentrations of flucloxacillin for 24 h. BEC (n ) 12) at days 2-4 of primary culture, were incubated with the conditioned media from flucloxacillin-treated hepatocytes or with flucloxacillin directly, for 24 h. Experiments were performed in triplicate and cytotoxicity was assessed by LDH release. (A) Toxic response of individual BEC preparations to hepatocyte conditioned media. Based on the criteria listed in Experimental Procedures, responders (unbroken line) can be distinguished from nonresponders (broken line). (B) In nonresponding BEC preparations (n ) 5), LDH release (means ( SEM) following incubation with hepatocyte conditioned media (solid bar) was not significantly different from LDH release following direct incubation with flucloxacillin of the same cells (shaded bar), and of four hepatocyte preparations used to generate conditioned media (open bar), (ANOVA, NS); (C) in responding BEC preparations (n ) 7), the conditioned media from flucloxacillintreated hepatocytes (solid bar) induced a dose-dependent increase in LDH release (means ( SEM; r ) 0.6; p < 0.001), while no significant change in LDH release resulted from direct incubation with flucloxacillin of the same cells (shaded bar), and of seven hepatocyte preparations used to generate conditioned media (open bar). (*) p < 0.05 vs incubation with conditioned media from flucloxacillin-untreated hepatocytes.

Chem. Res. Toxicol., Vol. 14, No. 6, 2001 697

rations, by contrast, hepatocyte conditioned media triggered an increase in LDH release which was related to the concentration of flucloxacillin tested, and which constantly exceeded 20% when the concentration was g200 mg/L (400 µM) (Figure 3A, unbroken line). Morphological signs of cell death including cytoplasm vacuolization, loss of nuclear boundaries, cell shrinkage and detachment were visible above the level of 20%, which was subsequently used as a threshold to define a cytotoxic response. In BEC preparations thus defined as responders, LDH release was significantly higher in cells exposed to the conditioned media from flucloxacillin treated hepatocytes than in controls (Figure 3C) and was constantly g35% when the concentration of flucloxacillin tested was of 500 mg/L (Figure 3A, unbroken line). In two cases, the conditioned medium derived from the same hepatocyte preparation was distributed onto two different BEC preparations and, in both cases, induced a cytotoxic response in only one them, suggesting that susceptibility to toxic response was dictated by BEC rather than by hepatocytes. In addition, no toxic response occurred in any of the BEC preparations when, in the same experimental protocol, amoxicillin up to 1000 mg/L (2.5 mM) was substituted for flucloxacillin (not shown), suggesting that flucloxacillin had a specific triggering effect in BEC toxic response. BEC Toxic Response to Microsome-Derived Flucloxacillin Metabolite(s). To determine whether toxic products released by hepatocytes resulted from flucloxacillin biotransformation, similar experiments were conducted, using microsome instead of hepatocyte preparations. The supernatant from a unique preparation of human liver microsomes1 (HLM) incubated with flucloxacillin (500 mg/L) and NADPH-generating system, was collected and distributed onto eight different BEC preparations. In four of them (Figure 4A), this resulted in LDH release which exceeded 20% and which was significantly higher than in controls including cells exposed to the supernatant of HLM incubated with NADPH-generating system alone. In the other four BEC preparations, LDH release remained below 10% and was not significantly different from controls (not shown). Similarly, the supernatant of human recombinant CYP3A4 preincubated with flucloxacillin (500 mg/L) and NADPH-generating system, was toxic in four out of five BEC preparations, as ascertained by LDH release beyond 20% (Figure 4B). None of the five preparations exhibited a cytotoxic response to the supernatant of other recombinant CYP isoforms incubated with flucloxacillin (500 mg/L), including CYP1A2 (Figure 4C), CYP2C9 and CYP3A5 (not shown). These results indicated that flucloxacillin metabolite(s) may be generated by microsomal enzymes including the CYP3A4 isoform, and cause cytotoxicity in susceptible BEC. Flucloxacillin Metabolic Pathways. Since both our findings and the structure of flucloxacillin major metabolite, 5′-hydroxymethylflucloxacillin (Figure 1) previously identified in rat and human body fluids, were consistent with CYP-mediated biotransformation, we further analyzed flucloxacillin metabolic pathways in microsome preparations. We first used CYP-induced rat liver microsomes to produce large amounts of flucloxacillin metabolites in vitro. Inductions mainly of CYP3A, CYP1A, and CYP2B/CYP3A families, were achieved by treating rats with dexamethasone, with methylcholanthrene, and with phenobarbital, respectively. Following incubations

698

Chem. Res. Toxicol., Vol. 14, No. 6, 2001

Figure 4. BEC toxic response to microsome-derived flucloxacillin metabolite(s). Preparations of human liver microsomes (HLM) and of human recombinant CYP isoforms were incubated for 2 h with or without flucloxacillin (500 mg/L), in the presence of NADPH-generating system. Human BEC at days 2-4 of primary culture were incubated for 24 h either with microsome supernatants, with flucloxacillin (500 mg/L) directly, or with culture medium alone. Cytotoxicity was assessed by LDH release. (A) Toxic effects of HLM supernatant in four BEC preparations, (B) toxic effects of CYP3A4 supernatant in 4 preparations, and (C) absence of toxic effect of CYP1A2 supernatant in five preparations, are shown. Results of experiments performed in triplicate are expressed as means ( SEM. (***) P < 0.001 vs other conditions.

of rat liver microsomes with flucloxacillin, a unique peak that depended upon time and NADPH-generating system, became detectable by HPLC in the supernatants. The highest peak was detected in the supernatant of dexamethasone-treated rat liver microsomes (Figure 5A), suggesting that flucloxacillin metabolite production resulted predominantly from CYP3A activity. The metabolite generated by dexamethasone-treated rat liver microsomes was purified by semipreparative HPLC, and identified as 5′-hydroxymethylflucloxacillin by mass spectrometry and NMR spectroscopy. In particular, the 5′-

Lakehal et al.

Figure 5. HPLC analysis of flucloxacillin metabolite production in microsome preparations. Preparations of (A) dexamethasone-treated rat liver microsomes, (B) human liver microsomes (HLM), and (C) human recombinant CYP3A4, were incubated with flucloxacillin (100 mg/L) and NADPH-generating system for 0-60 min. The supernatants were then analyzed by HPLC as described in methods. Chromatograms following (A) 0-min (t-0) and 40-min (t-40) incubations; (B) 60-min (t-60) incubation, with or without 50 µM troleandomycin (TAO); and (C) 60-min (t-60) incubation, are shown. UV spectra at 260 nm of flucloxacillin (F) and metabolite (M) are shown.

CH3 signal of flucloxacillin at 2.77 ppm in the 1H NMR spectrum was changed into a 5′-CH2OH signal at 5.10 ppm, in agreement with previous reports (9, 13). Purified 5′-hydroxymethylflucloxacillin subsequently served as a standard in HPLC analyses of human preparations. Human liver microsomes incubated with flucloxacillin also generated a major chromatographic peak depending upon time and NADPH-generating system (Figure 5B). This major metabolite exhibited the same time profile and UV spectrum as 5′-hydroxymethylflucloxacillin standard. Its formation was by more than 70% inhibited in the presence of CYP3A4 specific inhibitors, troleandomycin (Figure 5B) or miconazole, while it was not significantly affected by tetrahydrofuran or sulfaphenazole, used as specific inhibitors of CYP1A2 and CYP2C9, respectively (not shown). Identification of the same chromatographic peak in the supernatant of human recombinant CYP3A4 (Figure 5C) further indicated that the formation of flucloxacillin major metabolite is catalyzed by the CYP3A4 isoenzyme.

Biliary Cytotoxicity of Flucloxacillin

Figure 6. Hepatocyte and BEC toxic response to 5′-hydroxymethylflucloxacillin. Preparations of autologous hepatocytes (open bar) and BEC (shaded bar) from one individual, at day 2 of primary culture, were incubated for 24 h with increasing concentrations of metabolite purified from the supernatant of dexamethasone-treated rat liver microsomes (Figure 5A), and identified as 5′-hydroxymethylflucloxacillin by NMR spectroscopy and mass spectrometry, as described in methods. Cytotoxicity was assessed by LDH release. Results are the mean of triplicate values from one experiment.

Role of Flucloxacillin Metabolic Pathways in the Formation of Hepatocyte-Derived Toxic Product(s). CYP3A-associated monooxygenase activity as tested by testosterone 6β-hydroxylation was maintained in primary cultures of hepatocytes, at levels of 72 ( 12% (mean ( SEM, n ) 3) at day 2 as compared with freshly isolated cells, while the activity was at the limit of detection in cultured BEC. Consistent with these data, HPLC analyses of hepatocyte and BEC conditioned media following 24-h incubation with flucloxacillin revealed a number of chromatochraphic peaks among which one was restricted to hepatocytes, and identified as 5′-hydroxymethylflucloxacillin, based on retention time and UV spectrum (not shown). Since the estimated amount of this metabolite in hepatocyte conditioned media was approximately 1% that of the parent compound, we tested the toxic effects of purified 5′-hydroxymethylflucloxacillin at low concentrations in the preparations of autologous hepatocytes and BEC from three different individuals. While LDH release never exceeded 15% in hepatocytes, levels over 20% were recorded in two BEC preparations when the cells were exposed to a concentration of 5 mg/L (10 µM), and were observed in the third one in response to concentrations as low as 0.01 mg/L (20 nM) (Figure 6). Evidence that toxicity induced by hepatocyte supernatants was due to CYP3A4-derived metabolite(s) was further provided by the results of CYP3A4 induction in hepatocytes. Hepatocyte treatment with rifampicin (50 µM) in addition to flucloxacillin (500 mg/L) resulted in LDH release from responsive BEC, at levels of 273 ( 8% (mean ( SEM, n ) 4, p < 0.05) as compared with hepatocyte treatment with flucloxacillin alone.

Discussion This study provides first evidence that flucloxacillin metabolite(s) are produced in hepatocytes and may be toxic to BEC. Our data also demonstrate that the major metabolite identified in patients, 5′-hydroxymethylflu-

Chem. Res. Toxicol., Vol. 14, No. 6, 2001 699

cloxacillin (12), results from CYP3A4-mediated oxidative metabolism. On the basis of mass spectrometry and NMR spectroscopy, 5′-hydroxymethylflucloxacillin was identified in the supernatant of dexamethasone-treated rat liver microsomes, and one major metabolite exhibiting the same chromatographic characteristics as those of purified 5′-hydroxymethylflucloxacillin was detected in the supernatant of human liver microsomes and of recombinant CYP3A4. Furthermore, the production of this metabolite was inhibited after CYP3A4 inhibition by troleandomycin in human liver microsomes, and markedly enhanced after CYP3A induction by dexamethasone in rat liver microsomes. Even though flucloxacillin was not intrinsically toxic to BEC nor to hepatocytes as previously noted in a preliminary report (27), the conditioned media from flucloxacillin-treated hepatocytes and liver microsomes induced cytotoxicity in approximately 50% of BEC preparations. CYP3A4 induction in hepatocytes resulted in exacerbation of cytotoxicity, suggesting that toxic metabolites were synthesized in hepatocytes through CYP3A4 activity. This was further supported by the toxic response of BEC to the supernatant of flucloxacillin-treated recombinant CYP3A4 and to purified 5′-hydroxymethylflucloxacillin. Cytotoxicity in BEC was induced by concentrations of the purified metabolite that were in the range of those produced by hepatocytes. In comparing the chromatograms of conditioned media from hepatocytes and BEC, it appeared clearly that the production of 5′hydroxymethylflucloxacillin was restricted to hepatocytes. This finding was consistent with the observation that only cultured hepatocytes displayed significant CYP3A activity and showed ability to generate toxic metabolite(s). However, we cannot exclude that additional products in hepatocyte conditioned media contributed to their toxic effects, inasmuch as one chromatographic peak other than 5′-hydroxymethylflucloxacillin was observed in hepatocyte supernatants only (not shown). Its retention time (16.4 min) was compatible with that expected for the penicilloic acid of 5′-hydroxymethylflucloxacillin, another metabolite detected in patients, that results from spontaneous hydrolysis of the 5-hydroxymethyl derivative (28, 29). The reason this peak was not identified in microsome supernatants may be that the incubations of microsome with flucloxacillin were notably shorter than those of hepatocytes (60 min vs 24 h, respectively). Participation of this metabolite to toxicity would explain that LDH release from BEC tended to be higher in response to hepatocyte conditioned media as compared to microsome supernatants. Hepatocytes were clearly more resistant than BEC to their own metabolite production and to exogenous 5′hydroxymethylflucloxacillin. Other biliary toxicants, menadione and cumene hydroperoxide, were previously shown to exert toxic effects in rat intrahepatic BEC at concentrations that were not toxic to hepatocytes (30). The mechanisms of cell death induced by these latter compounds in BEC involve glutathione depletion and oxidation of protein thiols leading to conditions of oxidative stress. BEC sensitivity to these toxicants has been attributed to the fact that their glutathione content is only one-third that usually found in hepatocytes (31). Another finding in the present study was that the different BEC preparations exhibited variable response to similar toxic stimuli, as previously observed

700

Chem. Res. Toxicol., Vol. 14, No. 6, 2001

in human intrahepatic BEC (25). It was previously demonstrated in Clara cells, that glutathione levels are highly heterogeneous within the cell population, leading to a wide range of cytotoxic response to bioactivated xenobiotic compounds (32). Since glutathione is critical to cellular defense, and also because one of the key pathways of glutathione depletion is the conjugation of metabolites produced by CYP monooxygenases, we may hypothesize that susceptibility of BEC to injury was determined by different basal levels of GSH content. An alternative hypothesis which was previously proposed to explain haloalkene nephrotoxicity (33), is that a two-step metabolism occurs, and that the metabolite produced in hepatocytes is toxic after further metabolism in BEC. Polymorphic expression of the metabolizing enzyme in BEC would, in that case, account for susceptibility. Both idiosyncratic metabolic and immune-mediated mechanisms have been postulated to explain flucloxacillin-induced liver disease. Lymphocyte sensitization to flucloxacillin as assessed by in vitro lymphocyte transformation test, has been reported in one patient with flucloxacillin-induced cholestatic hepatitis (34). However, immunoallergic clinical manifestations are virtually absent in most patients and in liver biopsies, bile duct epithelium typically shows degenerative changes and only occasional infiltration by inflammatory cells (9). Our findings suggest that BEC damage may be caused by metabolites produced by hepatocytes, raising the question of how these metabolites are conveyed to the biliary epithelium. Flucloxacillin is eliminated in human bile (35), and induces hypercholeresis followed by delayed cholestasis in the isolated perfused rat liver (27). These effects on bile secretion are similar to those produced by other organic acids including xenobiotics such as SC2644 (36) and sulindac (37), which undergoes a cholehepatic circulation, inducing hypercholeresis (37). Should this transport pathway also apply to flucloxacillin and its metabolites, the concentrations and exposure time required for the metabolite(s) to be injurious may be rarely achieved in vivo, accounting for the low incidence of symptomatic biliary epithelial cell damage in treated patients (8). Biliary epithelial cells are a target for chemical toxicants that must be bioactivated to exert their toxic effects, and a number of drugs that can cause cholangiopathies are metabolized by cytochromes P450 (2). Our findings support the concept that exposure of biliary epithelial cells to bile-borne drug metabolites synthesized in hepatocytes may ultimately lead to drug-induced cholangiopathy.

Acknowledgment. This work was supported by grants from the European Union Biomed 2 Contract BMH4-CT96-0658, the Program Hospitalier de Recherche Clinique 94-95, the re´seau He´patox, the Institut National de la Sante´ et de la Recherche Me´dicale, SmithKline Beecham France, and a fellowship from Rhoˆne-Poulenc Rorer R&D (as part of the Bioavenir Program, to Fatima Lakehal). The authors are very grateful to Franc¸ ois Ballet, Michel Biour, Marie-The´re`se Bonnefis, Yves Chre´tien, and Patrice Jaillon for their contributions, and also wish to acknowledge Marcel Delaforge for helpful suggestions.

Lakehal et al.

References (1) Zimmerman, H. J., and Lewis, J. H. (1987) Drug-induced cholestasis. Med. Toxicol. 2, 112-160. (2) Degott, C., Feldmann, G., Larrey, D., Durand-Schneider, A.-M., Grange, D., Machayekhi, J.-P., Moreau, A., Potet, F., and Benhamou, J.-P. (1992) Drug-induced prolonged cholestasis in adults: A histological semiquantitative study demonstrating progressive ductopenia. Hepatology 15, 244-251. (3) Desmet, V. J. (1997) Vanishing bile duct syndrome in druginduced liver disease. J. Hepatol. 26, 31-35. (4) McNeil, J. J., Grabsch, E. A., and McDonald, M. M. (1999) Postmarketing surveillance: strengths and limitations. The flucloxacillin-dicloxacillin story. Med. J. Aust. 170, 270-273. (5) Devereaux, B. M., Crawford, D. H. G., Purcell, P., Powell, L. W., and Roeser, H. P. (1995) Flucloxacillin associated cholestatic hepatitis. An Australian and Swedish epidemic. Eur. J. Pharmacol. 49, 81-85. (6) Olsson, R., Wiholm, B.-E., Sand, C., Zettergren, L., Hultcrantz, R., and Myrhed, M. (1992) Liver damage from flucloxacillin, cloxacillin and dicloxacillin. J. Hepatol. 15, 154-161. (7) Fairley, C. K., McNeil, J. J., Desmond, P., Smallwood, R., Young, H., Forbes, A., Purcell, P., and Boyd, I. (1993) Risk factors for development of flucloxacillin associated jaundice. Br. Med. J. 306, 233-235. (8) Desmond, P. V. (1995) Flucloxacillin hepatitis - an Australian epidemic. Aust. NZ. J. Med. 25, 195-196. (9) Eckstein, R. P., Dowsett, J. F., and Lunzer, M. R. (1993) Flucloxacillin induced liver disease: histophatological findings at biopsy and autopsy. Pathology 25, 223-228. (10) Thijssen, H. H. W. (1979) Identification of the active metabolites of the isoxazolyl-penicillins by means of mass-spectrometry. J. Antibiot. 32, 1033-1037. (11) Thijssen, H. H. W. (1980) Analysis of isoxalyl penicillins and their metabolites in body fluids by high-performance liquid chromatography. J. Chromatogr. 183, 339-345. (12) Thijssen, H. H. W., and Wolters, J. (1982) The metabolic disposition of flucloxacillin in patients with impaired kidney function. Eur. J. Clin. Pharmacol. 22, 429-434. (13) Everett, J. R., Tyler, J. W., and Wooddnutt, G. (1989) A study of flucloxacillin metabolites in rat urine by two-dimensional 1H, 19F COSY NMR. J. Pharm. Biomed. Anal. 7, 397-403. (14) Dray-Charier, N., Paul, A., Veissie`re, D., Mergey, M., Scoazec, J.-Y., Capeau, J., Brahimi-Horn, C., and Housset, C. (1995) Expression of cystic fibrosis transmembrane conductance regulator in human gallbladder epithelial cells. Lab. Invest. 73, 828836. (15) Scoazec, J.-Y., Bringuier, A.-F., Medina, J. F., Martinez-Anso, E., Veissie`re, D., Feldmann, G., and Housset, C. (1997) The plasma membrane polarity of human biliary epithelial cells: in situ immunohistochemical analysis and functional implications. J. Hepatol. 26, 543-553. (16) Lakehal, F., Wendum, D., Barbu, V., Becquemont, L., Poupon, R., Balladur, P., Hannoun, L., Ballet, F., Beaune, P. H., and Housset, C. (1999) Phase I and phase II drug-metabolizing enzymes are expressed and heterogeneously distributed in the biliary epithelium. Hepatology 30, 1498-1506. (17) Parry, S. W., Pelias, M. E., and Browder, W. (1988) Acalculous hypersensitivity cholecystitis: hypothesis of a new clinicopathologic entity. Surgery 104, 911-916. (18) Richard, B., Nadal, D., Meuli, M., and Braegger, C. P. (1993) Acute acalculous cholecystitis in infective endocarditis. J. Pediatr. Gastroenterol. Nutr. 17, 215-216. (19) Housset, C., Carayon, A., Housset, B., Legendre, C., Hannoun, L., and Poupon, R. (1993) Endothelin-1 secretion by human gallbladder epithelial cells in primary culture. Lab. Invest. 69, 750-755. (20) Ballet, F., Bouma, M.-E., Wang, S.-R., Amit, N., Marais, J., and Infante, R. (1984) Isolation, culture and characterization of adult human hepatocytes from surgical liver biopsies. Hepatology 4, 849-854. (21) Kreamer, B. L., Staecker, J. L., Sawada, N., Sattler, G. L., Hsia, M. T. S., and Pitot, H. C. (1986) Use of a low-speed, iso-density Percoll centrifugation method to increase the viability of isolated rat hepatocyte preparations. In Vitro Cell. Dev. Biol. 22, 201211. (22) Adam, D., Koeppe, P., and Heilmann, H.-D. (1983) Pharmacokinetics of amoxicillin and flucloxacillin following the simultaneous intravenous administration of 4 and 1 g, respectively. Infection 11, 150-154. (23) Bensoussan, C., Delaforge, M., and Mansuy, D. (1995) Particular ability of cytochromes P450 3A to form inhibitory P450-iron-

Biliary Cytotoxicity of Flucloxacillin

(24)

(25)

(26)

(27) (28)

(29) (30)

metabolite complexes upon metabolic oxidation of aminodrugs. Biochem. Pharmacol. 49, 591-602. Belloc, C., Baird, S., Cosme, J., Lecoeur, S., Gautier, J.-C., Challine, D., de Waziers, I., Flinois, J.-P., and Beaune, P. H. (1996) Human cytochromes P450 expressed in Escherichia coli: production of specific antibodies. Toxicology 106, 207-219. Ayres, R. C. S., Shaw, J., Mills, C. O., Coleman, R., and Neuberger, J. M. (1991) A 51Cr release cytotoxicity assay for use with human intrahepatic biliary epithelial cells. J. Immunol. Methods 141, 117-122. Kostrubsky, V. E., Ramachandran, V., Venkataramanan, R., Dorko, K., Esplen, J. E., Zhang, S., Sinclair, J. F., Wrighton, S. A., and Strom, S. C. (1999) The use of human hepatocyte cultures to study the induction of cytochrome P-450. Drug Metab. Dispos. 27, 887-894. Frost, L., Tapner, M., Field, J., and Farrell, G. C. (1994) Flucloxacillin stimulates choleresis and cholestasis in perfused rat liver: relationship to hepatotoxicity. Hepatology 20, 349A. Murai, Y., Nakagawa, T., Yamaoka, K., and Uno, T. (1983) Comparative pharmacokinetics of metabolism and urinary excretion of isoxazolylpenicillins in man. Chem. Pharm. Bull. 31, 32923301. Everett, J. R., Jennings, K., and Woodnutt, G. (1985) 19F NMR spectroscopy study of the metabolites of flucloxacillin in rat urine. J. Pharm. Pharmacol. 37, 869-873. Parola, M., Cheeseman, K. H., Biocca, M. E., Dianzani, M. U., and Slater, T. F. (1990) Menadione and cumene hydroperoxide induced cytotoxicity in biliary epithelial cells isolated from rat liver. Biochem. Pharmacol. 39, 1727-1734.

Chem. Res. Toxicol., Vol. 14, No. 6, 2001 701 (31) Parola, M., Cheeseman, K. H., Biocca, M. E., Dianzani, M. U., and Slater, T. F. (1990) Biochemical studies on bile duct epithelial cells isolated from rat liver. J. Hepatol. 10, 341-345. (32) West, J. A., Chichester, C. H., Buckpitt, A. R., Tyler, N. K., Brennan, P., Helton, C., and Plopper, C. G. (2000) Heterogeneity of clara cell glutathione. A possible basis for differences in cellular responses to pulmonary cytotoxicants. Am. J. Respir. Cell Mol. Biol. 23, 27-36. (33) Birner, G., Bernauer, U., Werner, M., and Dekant, W. (1997) Biotransformation, excretion and nephrotoxicity of haloalkenederived cysteine S-conjugates. Arch Toxicol 72, 1-8. (34) Victorino, R. M. M., Maria, V. A., Correia, A. P., and de Moura, M. C. (1987) Floxacillin-induced cholestatic hepatitis with evidence of lymphocyte sensitization. Arch. Intern. Med. 147, 987989. (35) Bergan, T. (1984) Pharmacokinetics of beta-lactam antibiotics. Scand. J. Infect. Dis. (Suppl. 42), 83-98. (36) Barnhart, J. L., and Combes, B. (1978) Characterization of SC2644-induced choleresis in the dog. Evidence for canalicular bicarbonate secretion. J. Pharmacol. Exp. Ther. 206, 190-197. (37) Bolder, U., Trang, N. V., Hagey, L. R., Schteingart, C. D., TonNu, H.-T., Cerre`, C., Oude Elferink, R. P. J., and Hofmann, A. F. (1999) Sulindac is excreted into bile by a canalicular bile salt pump and undergoes a cholehepatic circulation in rats. Gastroenterology 117, 962-971.

TX0002435