Article pubs.acs.org/crt
Preclinical Strategy to Reduce Clinical Hepatotoxicity Using in Vitro Bioactivation Data for >200 Compounds Melanie Z. Sakatis,*,† Melinda J. Reese,‡ Andrew W. Harrell,† Maxine A. Taylor,† Ian A. Baines,† Liangfu Chen,§ Jackie C. Bloomer,† Eric Y. Yang,§ Harma M. Ellens,§ Jeffrey L. Ambroso,⊥ Cerys A. Lovatt,# Andrew D. Ayrton,† and Stephen E. Clarke† †
Drug Metabolism and Pharmacokinetics, GlaxoSmithKline, Park Road, Ware, Hertfordshire SG12 0DP, United Kingdom Drug Metabolism and Pharmacokinetics, GlaxoSmithKline, Five Moore Drive, P.O. Box 13398, Research Triangle Park, North Carolina, United States § Drug Metabolism and Pharmacokinetics, GlaxoSmithKline, Upper Merion, 709 Swedeland Road, King of Prussia, Pennsylvania 1905, United States ⊥ Safety Assessment, GlaxoSmithKline, Five Moore Drive, P.O. Box 13398, Research Triangle Park, North Carolina, United States # Safety Assessment, GlaxoSmithKline, Park Road, Ware, Hertfordshire SG12 0DP, United Kingdom ‡
ABSTRACT: Drug-induced liver injury is the most common cause of market withdrawal of pharmaceuticals, and thus, there is considerable need for better prediction models for DILI early in drug discovery. We present a study involving 223 marketed drugs (51% associated with clinical hepatotoxicity; 49% non-hepatotoxic) to assess the concordance of in vitro bioactivation data with clinical hepatotoxicity and have used these data to develop a decision tree to help reduce late-stage candidate attrition. Data to assess P450 metabolism-dependent inhibition (MDI) for all common drug-metabolizing P450 enzymes were generated for 179 of these compounds, GSH adduct data generated for 190 compounds, covalent binding data obtained for 53 compounds, and clinical dose data obtained for all compounds. Individual data for all 223 compounds are presented here and interrogated to determine what level of an alert to consider termination of a compound. The analysis showed that 76% of drugs with a daily dose of 5-fold change in IC50) or metabolismdependent covalent binding ≥200 pmol eq/mg protein. The percentage of hepatotoxic compounds identified by high dose
resulting in the intracellular accumulation of bile salts); mitochondrial dysfunction (resulting in the depletion of cellular energy supply and the generation of damaging reactive oxygen species); cell damage from oxidative stress (caused by reactive oxygen or reactive nitrogen species); and local inflammatory effects.6 DILI can also be immunologically mediated, resulting in idiosyncratic toxicity. Idiosyncratic reactions pose a particular challenge for drug development since they are not always detected in standard toxicology testing and occur in a very small proportion of patients (making it difficult to detect preregistration, even in large clinical trials). All of these mechanisms are often interconnected and have, under various circumstances, been associated with the formation of chemically reactive metabolites. Although there is much evidence associating chemically reactive metabolites with adverse drug reactions (including hepatotoxicity), the relationship is complex. Reactive metabolites produced within the liver may bind essential macromolecules and block their function leading to acute cytotoxicity. They may also produce oxidative stress by depleting antioxidant molecules such as GSH, rendering the cells more susceptible to other environmental stresses. Another way reactive metabolites can lead to DILI is through covalent modification of cellular proteins to form haptens which can illicit an immune or autoimmune response in the liver, and this is the most prevalent hypothesis for the mechanism behind DILI that is idiosyncratic in nature.7 Autoantibody production has been reported in patients dosed with drugs such as erythromycin and tienilic acid, which are associated with a high incidence of DILI.8,9 Current hypotheses, however, postulate that the production of reactive metabolites and autoantibodies alone is insufficient to illicit a pathological response, and it is currently believed that the immunological response can only occur in the presence of a second signal or “danger signal” which involves modulation of cytokine release.10 The inherent chemical instability of chemically reactive metabolites means the direct measurement of reactive metabolites in a routine manner is not feasible. By understanding the theoretical processes governing the production and fate of reactive metabolites, however, various signals can be detected that are associated with their existence, as depicted in Figure 1. Under normal circumstances, the majority of molecules will be eliminated by innocuous routes, either through harmless metabolite pathways or by direct renal or biliary secretory mechanisms. For some molecules, however, a
Figure 1. Scheme depicting the hypothetical production and fate of reactive metabolites, detailing various measurable signals that may alert one to their existence during drug development. 2068
dx.doi.org/10.1021/tx300075j | Chem. Res. Toxicol. 2012, 25, 2067−2082
Chemical Research in Toxicology DEF: diethoxyfluorescein.25
409/530
MMC: 4-methylaminoethyl-7-methoxycoumarin. BMC: 3-butyryl-7-methoxycoumarin.
409/508
ER: ethoxyresorufin. FCA: 7-methoxy-4-trifluoromethylcoumarin-3-acetic acid. 7BQ: 7-benzyloxyquinoline. MDMA: 3,4-methylenedioxymethamphetamine.
a
probe substrate (final concentrations) positive control bactosome mix final concentrations (mg/mL) bactosomes expressing specific P450 + control bactosomes wavelength (nm): excitation/emission
22
530/590
409/460
23
409/485
485/530
24
0.1 + 0 0.06 + 0.04 0.06 + 0.04 0.0006 + 0.994 0.1 + 0 0.01 + 0.09
BMC (10 μM in 1% DMSO) ticlopidine
3A4 2C9
FCA (50 μM in 1% acetonitrile) tienilic acid ER (0.5 μM in 1% acetonitrile) furafylline
Table 1. Incubation Conditions for Cytochrome P450 Inhibition Assaysa 2069
1A2
2C19
MATERIALS AND METHODS
Experimental Procedures. Materials. A set of 223 test compounds were identified for evaluation. Nonradioactive compounds were obtained from standard commercial suppliers and were reagent grade or higher. 14C-caffeine was supplied by Chemical Development, GlaxoSmithKline. All other 14C-radiolabeled compounds were obtained from Amercian Radiolabeled Chemicals (Missouri, USA) and GE Healthcare (Amersham, UK). Specific P450 probe substrates and P450 MDI positive controls were obtained from Sigma-Aldrich Company Ltd., (Dorset, UK), Molecular Probes (Oregon, USA), or Gentest Corporation (Massachusetts, USA) or were supplied by Medicinal Chemistry, GlaxoSmithKline (Harlow, UK). GSH, 3methylindole, and the chemical constituents of the NADPH regenerating system were purchased from Sigma-Aldrich Company Ltd., (Dorset, UK). All other chemicals used were obtained from standard commercial suppliers. Bactosomes (derived from E. coli) containing individually overexpressed human cytochrome P450 enzymes and either low level (LR) or high level (HR or R) human NADPH-P450 reductase (1A2LR, 2C9HR, 2C19R, 2D6R, 3A4R) and control bactosomes containing empty expression plasmids were obtained from Cypex Ltd. (Dundee, UK). Pooled mixed-gender human liver microsomes (>15 individuals) were obtained from XenoTech LLC (Kansas, USA). P450 MDI Experiments. Each P450 enzyme was assessed in a separate assay, using appropriate conditions detailed in Table 1. Incubations were conducted in a 96-well plate in 50 mM phosphate buffer, pH 7.4, using the final concentrations stated below, in a final incubation volume of 250 μL. Test compounds (0.1−100 μM, 2% (v/ v) methanol), appropriate positive control (0.01−10 μM, 2% (v/v) methanol), or solvent only (2% methanol) were incubated with the appropriate bactosome mix (0.1 mg/mL total protein) and appropriate fluorescent probe substrates in a Cytofluor fluorescence plate-reader at 37 °C. The plate was preincubated for ca. 10 min, and the reaction initiated by the addition of cofactor solution (NADPH regenerating system containing final concentrations of 0.22 mM β-NADP, 2.8 mM glucose-6-phosphate, and 0.6 Units/mL of glucose-6-phosphate dehydrogenase in 0.2% sodium hydrogen carbonate). The fluorescence in each well was measured for 30 min at 1 min intervals using the appropriate excitation and emission wavelengths. The rates of fluorescent metabolite production were expressed as a percentage of the uninhibited control rates and IC50 for each test compound
DEF (1 μM in 1% acetonitrile) troleandomycin
3A4 2D6 P450 enzyme
■
MMC (10 μM in 1% methanol) MDMA
or each assay alone (∼65% in each case) increased significantly (to 80−100%) when bioactivation data were combined with dose.20 The summary publication, however, did not include the individual data for the molecules examined, and strategy was not defined or discussed in any detail. Here, we present the individual reactive metabolite data for all 223 compounds (P450 MDI, GSH adduct formation, and covalent binding data where available) and, furthermore, present MDI data for the individual P450 enzymes, which have not been discussed previously. We consider these data to be a powerful reference resource for anyone investigating metabolism-dependent inhibition or GSH adduct formation of marketed drugs, or for anyone wishing to correlate these alerts with any other paradigm. We then present the evaluation of these data in the form of Spotfire plots to determine what level of an alert to consider the termination of a compound and have used this evaluation to develop an evidence-based strategy in the form of a decision tree to help reduce late-stage candidate attrition due to hepatotoxicity. Finally, after excluding compounds with insufficient data, 196 of these compounds (96 hepatotoxic, 100 non-hepatotoxic) were used to test the decision tree, together with an independent test set of 10 GSK compounds with known clinical hepatotoxicity status. These outcome analyses are presented and demonstrate the value of the decision tree developed.
7BQ (25 μM in 1% acetonitrile) Troleandomycin
Article
dx.doi.org/10.1021/tx300075j | Chem. Res. Toxicol. 2012, 25, 2067−2082
Chemical Research in Toxicology
Article
calculated. If a decrease in IC50 was observed over time, further experiments were conducted to assess if this was time-dependent inhibition or metabolism-dependent inhibition, the latter requiring the presence of NADPH. IC50 values were compared following a 10 and 40 min preincubation in the absence of cofactor. A decrease in the initial IC50 between the 10 and 40 min preincubation in the absence of cofactor indicated time-dependent inhibition (or NADPH independent), probably due to slow equilibration effects and not a true metabolic effect. No decrease in the initial IC50 between 10 and 40 min of preincubation in the absence of cofactor (NADPH) indicated that the original decrease in IC50 was due to MDI, i.e., time- and NADPHdependent inhibition that is potentially associated with P450 inactivation. As the mechanism of inhibition was not determined in this study (i.e., true irreversibility was not established) we have not used the term mechanism-based inhibition. GSH Adduct Assay. Incubations were conducted in 96-deepwell refill tubes in 50 mM phosphate buffer, pH 7.4, using the final concentrations stated below, in a final incubation volume of 500 μL. Test compounds (100 μM, 0.5% (v/v) DMSO) were incubated with GSH (10 mM) and pooled human liver microsomes (1 mg/mL) in a shaking water bath at 37 °C. After 3 min of preincubation, the reactions were initiated by the addition of cofactor solution (NADPH regenerating system containing final concentrations of 0.44 mM βNADP, 5.5 mM glucose-6-phosphate, and 1.2 units of glucose-6phosphate dehydrogenase, together with 3 mM MgCl 2 ). 3Methylindole was used as positive control for the formation of both cofactor-dependent and cofactor-independent GSH adducts. Negative control incubations were also performed in the absence of compound, GSH, cofactor, or microsomes. After 90 min, the reactions were terminated by the addition of 0.2 volumes of 6% acetic acid in acetonitrile, placed on ice for ca. 10 min, and then centrifuged at 2250g at 4 °C for 20 min. Supernatants were then analyzed by HPLC-MS, using a Rheous HPLC system and a Sciex API 4000 triple quadrupole mass spectrometer. A volume of 2−10 μL of supernatant was directly injected onto the HPLC system equipped with a reversed phase column, (50 × 2 mm, 3 μM particle size), using 10 mM ammonium formate (pH 3) and acetonitrile as mobile phases. The typical run-time was approximately 9 min per injection. GSH adducts were detected by neutral loss scan of pyroglutamic acid (129 Da) and confirmed with selective reaction monitoring. The relative intensity of a GSH adduct was measured by the signal/noise ratio, obtained from the selective reaction monitoring experiment. Covalent Binding Assay. Incubations were conducted in incubation tubes in 100 mM phosphate buffer, pH 7.4, using the final concentrations stated below, in a final incubation volume of 500 μL. Pooled human liver microsomes (1 mg/mL) were incubated with cofactor solution (NADPH regenerating system containing final concentrations of 0.44 mM β-NADP, 5.5 mM glucose-6-phosphate, and 5 units of glucose 6-phosphate dehydrogenase, in 0.4% sodium hydrogen carbonate) or with 0.4% sodium hydrogen carbonate only in a shaking water bath at 37 °C. After 5 min of preincubation, the reactions were initiated by the addition of [14C]-radiolabeled compound (10 μM). [14C]-Acetaminophen was used as the positive control. After 60 min, the reactions were terminated by the addition of 8 volumes (4 mL) of 90% (v/v) methanol and briefly vortexed. Negative controls were also conducted whereby the reaction was terminated before the addition of radiolabeled compound (t = 0 controls). Samples were then briefly vortexed before loading onto Whatman GF-B filters presoaked in 100 mM phosphate buffer at pH 7.4 using a Millipore vacuum manifold. Samples were filtered to isolate the protein and then the filters washed five times with 4 mL of 90% (v/v) methanol. Filters were then transferred to scintillation vials and incubated overnight at ambient temperature in 4 mL of liquid scintillation cocktail to solubilize the protein from the filters before being radioassayed using a Beckman 6700 scintillation counter with background subtraction. Covalent binding at 60 min in the presence of cofactor was expressed as pmol eq/mg protein and was corrected for any apparent binding observed in the t = 0 controls. Data Analysis and Strategy Development. Hepatotoxicity Classification of Drugs. Of the 223 compounds identified for
evaluation, 113 compounds (51%) were defined as hepatotoxic, on the basis of at least 50 reports of clinical hepatotoxicity or 3 reports of life-threatening hepatotoxicity,26 or warnings or precautions on the label.27 All forms of hepatotoxicity (idiosyncratic and nonidiosyncratic) were included. The other 110 compounds (49%) were defined as non-hepatotoxic, on the basis of no reports of clinical hepatotoxicity or only one report of mild hepatotoxicity (transient mildly elevated liver enzymes).26,27 Clinical Dose Categorization. Compounds were categorized into 4 dosage groups: < 10 mg, 10 to
