Can In Vitro Metabolism-Dependent Covalent Binding Data in Liver

Aug 9, 2008 - ... Adducted by a Reactive Metabolite Using Shotgun Proteomics. .... Michael Zawistoski, Jianke Li, Robert Oliver, Angel Guzman-Perez, K...
3 downloads 0 Views 403KB Size
1814

Chem. Res. Toxicol. 2008, 21, 1814–1822

Can In Vitro Metabolism-Dependent Covalent Binding Data in Liver Microsomes Distinguish Hepatotoxic from Nonhepatotoxic Drugs? An Analysis of 18 Drugs with Consideration of Intrinsic Clearance and Daily Dose R. Scott Obach,* Amit S. Kalgutkar, John R. Soglia, and Sabrina X. Zhao Pharmacokinetics, Dynamics and Metabolism Department, Pfizer Global Research and DeVelopment, Groton, Connecticut 06340 ReceiVed May 8, 2008

In vitro covalent binding assessments of drugs have been useful in providing retrospective insights into the association between drug metabolism and a resulting toxicological response. On the basis of these studies, it has been advocated that in vitro covalent binding to liver microsomal proteins in the presence and the absence of NADPH be used routinely to screen drug candidates. However, the utility of this approach in predicting toxicities of drug candidates accurately remains an unanswered question. Importantly, the years of research that have been invested in understanding metabolic bioactivation and covalent binding and its potential role in toxicity have focused only on those compounds that demonstrate toxicity. Investigations have not frequently queried whether in vitro covalent binding could be observed with drugs with good safety records. Eighteen drugs (nine hepatotoxins and nine nonhepatotoxins in humans) were assessed for in vitro covalent binding in NADPH-supplemented human liver microsomes. Of the two sets of nine drugs, seven in each set were shown to undergo some degree of covalent binding. Among hepatotoxic drugs, acetaminophen, carbamazepine, diclofenac, indomethacin, nefazodone, sudoxicam, and tienilic acid demonstrated covalent binding, while benoxaprofen and felbamate did not. Of the nonhepatotoxic drugs evaluated, buspirone, diphenhydramine, meloxicam, paroxetine, propranolol, raloxifene, and simvastatin demonstrated covalent binding, while ibuprofen and theophylline did not. A quantitative comparison of covalent binding in vitro intrinsic clearance did not separate the two groups of compounds, and in fact, paroxetine, a nonhepatotoxin, showed the greatest amount of covalent binding in microsomes. Including factors such as the fraction of total metabolism comprised by covalent binding and the total daily dose of each drug improved the discrimination between hepatotoxic and nontoxic drugs based on in vitro covalent binding data; however, the approach still would falsely identify some agents as potentially hepatotoxic. Introduction Safety-related issues continue to be a significant contributor to overall attrition statistics in the pharmaceutical industry. It has been estimated that as many as 50% of all new chemical entities that enter preclinical animal toxicological evaluations fail to advance to human trials, and as much as 30% of all drugs that enter clinical testing are terminated because of unanticipated safety problems (1). Also, over the years, there have been wellpublicized instances of drugs known to exhibit rare but serious adverse drug reactions (ADRs)1 after approval and marketing that have ultimately resulted in their withdrawal (e.g., tienilic acid and troglitazone) (2-4). These ADRs [also referred to as idiosyncratic ADRs (IADRs)] are not related to known drug pharmacology, and although they are dose-dependent in susceptible individuals, they can occur at any dose within the usual therapeutic range (5). Conventional animal models of toxicity are poor predictors of IADRs in humans (6). Considering the * To whom correspondence should be addressed. Tel: 860-441-6122. E-mail: [email protected]. 1 Abbreviations: ADRs, adverse drug reactions; IADRs, idiosyncratic adverse drug reactions, DILI, drug-induced liver injury; NSAID, nonsteroidal anti-inflammatory drug; GSH, reduced glutathione; UDPGA, uridine-5′diphosphoglucuronic acid; UGT, uridine 5′-diphosphoglucuronosyl transferase; LC-MS/MS, liquid chromatography-tandem mass spectrometry.

low frequency of occurrence of IADRs (1 in 10000 to 1 in 100000 patients), large clinical trials, exposing perhaps up to 10000 patients to a new therapeutic agent prior to registration, may not suffice in detecting IADRs reliably. An additional complication is that IADRs usually manifest as overt or symptomatic disease and can occur with intermediate (1-8 weeks) or long (1 year) periods of latency. Drugs can adversely affect almost any organ in the body; however, potentially life-threatening IADRs noted for several drugs include hepatotoxicity, toxic epidermal necrolysis, anaphylaxis, and blood dyscrasias. Among these, drug-induced liver injury (DILI) is the most common cause for the withdrawal of a drug from the market, accounts for 50% of the cases of acute liver failure, and mimics all forms of acute and chronic liver disease (7). Following oral administration, the liver is the primary accumulation site for drugs and serves as the central facility for their metabolism; therefore, it is not surprising that this organ is particularly vulnerable to damage by drugs and their metabolites. In fact, an estimated 1400 drugs have been implicated in causing liver damage on more than one occasion (8). For some drugs, DILI can be readily recognized as drugassociated, can occur in higher numbers of subjects, generally follows a dose response, and in most cases can be observed in preclinical species under toxicological evaluation. Hepatotoxicity

10.1021/tx800161s CCC: $40.75  2008 American Chemical Society Published on Web 08/09/2008

Metabolism-Dependent CoValent Binding of 18 Drugs

Chem. Res. Toxicol., Vol. 21, No. 9, 2008 1815

Figure 1. Structures of the 18 drugs examined in this analysis. The asterisk indicates the position of the radiolabel.

associated with the anti-inflammatory agent acetaminophen is a good example of this phenomenon (9). For most drugs, hepatotoxicity is observed in very few subjects, does not have a readily discernible dose-response relationship, and is not observed in laboratory animal models. The idiosyncratic hepatotoxicity associated with the nonsteroidal anti-inflammatory drug (NSAID) diclofenac is a good example of this behavior (10). Among several possible biochemical mechanisms, which could lead to DILI (11, 12), much attention has been focused on the phenomena of metabolic activation (bioactivation) of drugs to electrophilic, reactive intermediates. The concept of bioactivation of xenobiotics to chemically reactive metabolites, which form adducts with essential tissue macromolecules, resulting in disruption of necessary cellular processes has been a hallmark of molecular toxicology for over 50 years since the Millers proposed concepts of metabolism-dependent carcinogenicity/hepatotoxicity of aminoazo dyes (13). Extension of these concepts to human drug-induced hepatotoxicity was provided from studies in the 1970s on the covalent binding to hepatic tissue by structurally diverse drugs including acetaminophen (14, 15). The availability of in vitro approaches to evaluate bioactivation and covalent binding of drugs to target tissue(s) has provided useful insights to the role of drug metabolism in the retrospective analysis of underlying mechanism(s) of hepatotoxic drugs (16, 17). Likewise, the concept of structural alerts, that is, lists of chemical substituents prone to bioactivation by drug metabolizing enzymes to chemically reactive intermediates, has been developed from these efforts

and are now a consideration in the design of new drugs (18). Although the detection of a bioactivation process in vitro is relatively straightforward, the downstream consequence(s) of this process remain poorly understood; the identities of macromolecular targets to which bioactivated drugs bind and for which binding results in toxicity have not been routinely established and remain an area of active research. Given the inability to predict whether a bioactivation phenomena detected in vitro will ultimately translate to DILI in the clinic, a general strategy adopted within the pharmaceutical community involves the assessment of electrophilic reactive metabolite formation as early as possible in the selection of drug candidates, with the goal of eliminating or minimizing the formation of reactive metabolites by rational structural modifications of lead chemical matter prone to bioactivation. To this end, various in vitro approaches have been proposed as screens for reactive metabolites, such as assays in which nucleophilic trapping agents [e.g., glutathione (GSH), cyanide, and/or amines] are included in metabolic incubations with human hepatic tissue and the trapped electrophilic metabolites are detected (18). In addition, the utility of radiolabeled drug candidates to quantitatively measure the formation of macromolecule adducts to liver microsomal proteins in vitro and rat liver in vivo has been proposed, with cutoff values for when concerns should be raised and included with other important factors termed “qualifying considerations” (16). However, to date, there is no experimental evidence suggestive of a quantitative relationship between the amount of covalent adduct or trapped electrophile formed and the potential for toxicity. Consequently, the utility of this

1816

Chem. Res. Toxicol., Vol. 21, No. 9, 2008

methodology as a predictive tool for toxicity remains as an unproven hypothesis. While the combination of in vitro studies to detect reactive metabolite formation followed by covalent binding assessments has been extremely valuable in unraveling bioactivation mechanisms, as a property itself, covalent binding in vitro has not been rigorously tested for its ability to distinguish between hepatotoxic and nonhepatotoxic drugs, primarily because nonhepatotoxic drugs have not been tested in covalent binding studies. Furthermore, failing to consider aspects such as the potential for detoxication of reactive intermediates, the relative importance of the metabolic pathway that leads to covalent binding vs total metabolism, and the total daily dose can confound the utility of in vitro covalent binding data in the prediction of hepatotoxicity. To this end, we have undertaken a study of covalent binding of 18 drugs (nine well-established hepatotoxins and nine drugs considered as nonhepatotoxic) (Figure 1) in human liver microsomes, applying concepts of fractional clearance and dose to the interpretation of the data. Our findings demonstrate that in vitro covalent binding studies, while potentially useful to help understand mechanisms of hepatotoxicity, are themselves not predictive of hepatotoxicity.

Experimental Procedures Chemicals. Radiolabeled materials were from four sources. [14C]Buspirone, [14C]benoxaprofen, [14C]carbamazepine, [3H]diphenhydramine, [3H]felbamate, [14C]meloxicam, [14C]nefazodone, [14C]propranolol, [14C]raloxifene, [14C]simvastatin, and [14C]sudoxicam were custom synthesized by Nerviano Medical Sciences (Nerviano, Italy). [14C]Acetaminophen, [14C]diclofenac, and [14C]ibuprofen were purchased from GE-Healthcare Biosciences Corp. (Piscataway, NJ). [3H]Paroxetine was purchased from Perkin-Elmer (Boston, MA). [3H]Indomethacin and [14C]theophylline were obtained from American Radiochemicals (Columbia, MO). Positions of the radionuclides are shown in Figure 1. Acetaminophen, alamethicin, carbamazepine, diclofenac, diphenhydramine, GSH, ibuprofen, indomethacin, NADPH, paroxetine, propranolol, raloxifene, saccharolactone, theophylline, and uridine-5′-diphosphoglucuronic acid (UDPGA) were obtained from Sigma (St. Louis, MO). Meloxicam, felbamate, nefazodone, simvastatin, and buspirone were purchased from Sequoia (Oxford, United Kingdom). Benoxaprofen, tienilic acid, and sudoxicam were available in the chemical bank at Pfizer Global Research and Development (Groton, CT). Pooled human liver microsomes (pool of 56 donors) were obtained from BD-Gentest (Woburn, MA). Intrinsic Clearance Determinations for Covalent Binding to Liver Microsomes. Initial determinations were made to define the incubation time that provided linearity for covalent binding (Table 1). Radiolabeled substrates at varying concentrations (Table 1) were incubated with pooled human liver microsomes (2.0 mg/mL) in 1.0 mL of 0.1 M potassium phosphate buffer, pH 7.45, containing 3.3 mM MgCl2. Incubations were commenced with the addition of NADPH (1.3 mM) and conducted in a shaking water bath at 37 °C open to air. All incubations were conducted in duplicate. Microsomal incubations were terminated by the addition of 5 mL of acetonitrile to precipitate protein and salts. The mixtures were spun in a centrifuge (1700g) for 5 min, the supernatants were decanted, and the pellets were resuspended in acetonitrile by mixing on a vortex mixer and using metal wire to aid in the disruption of the pellets. This cycle of spinning in a centrifuge, discarding the supernatant, and resuspending the pellet was repeated six more times. The final pellet was dissolved in 1 mL of 2 M

Obach et al. Table 1. In Vitro Incubation Conditions Used for Covalent Binding Intrinsic Clearance Determinationsa drug

incubation time (min)

substrate concentration range (µM)

acetaminophen buspirone carbamazepine diclofenac diphenhydramine indomethacin meloxicam nefazodone paroxetine propranolol raloxifene simvastatin sudoxicam tienilic acid

30 20 30 20 30 20 20 20 10 30 20 20 20 10

25-10000 1.0-200 1.0-200 1.0-200 1.0-200 1.0-200 1.0-200 1.0-200 0.2-2100 1.0-200 1.0-200 1.0-200 1.0-200 1.0-200

a Benoxaprofen, felbamate, ibuprofen, and theophylline are not listed since binding was not observed.

NaOH overnight, and the radioactivity was determined by mixing the alkaline digest with 1 mL of water and 18 mL of TruCount scintillation fluid (In/US Systems Inc., Tampa, FL). The following equation was used to calculate radiolabeled covalent binding (pmol/mg protein):

radioactivity in pellet (dpm) · substrate concentration (nmol/mL) total radioactivity added to incubation (dpm) · protein concentration (mg/mL)

× 1000

The covalent binding data were further converted to reaction velocity values corrected for protein concentration. The velocity vs substrate concentration values were fit to the Michaelis-Menten equation or the Michaelis-Menten equation with an additional nonsaturable term (CLint,cb,2):

V)

Vmax · [S] Vmax · [S] or V ) + CLint,cb,2 · [S] KM + [S] KM + [S]

Intrinsic clearance of covalent binding (CLint,cb,1) was calculated as Vmax/KM, with total covalent binding intrinsic clearance being the sum of CLint,cb,1 and CLint,cb,2 when the second nonsaturable term was obtained from the velocity vs substrate concentration curve. In some experiments, the effects of various cofactors and additives on covalent binding were also assessed. Thus, in human liver microsomes, covalent binding was determined in the presence of 5 mM GSH, 1 mM semicarbazide, 1 mM potassium cyanide, or conditions to favor the activity of uridine glucuronosyl transferase (UGT) enzymes (5 mM UDPGA, 0.02 mg/mL alamethicin, and 1 mM saccharolactone). Total Intrinsic Clearance Determinations. In vitro metabolic lability determinations, in duplicate, were done using the same conditions as the covalent binding incubations. Incubations at a substrate concentration of 1 µM were commenced with the addition of NADPH, and aliquots of the incubation mixture were removed at 0, 2, 5, 10, 20, and 30 min and added to acetonitrile to terminate the reactions. The precipitated protein was removed by spinning in a centrifuge, and the supernatant was analyzed by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) for the remaining parent drug to yield in vitro t1/2 values used in the calculation of CLint. The system consisted of two pumps (Jasco Inc., Easton, MA), a Gilson 215 multiprobe liquid handling system (Gilson Instruments, Middeleton, WI), and a Sciex API3000 mass spectrom-

Metabolism-Dependent CoValent Binding of 18 Drugs

Chem. Res. Toxicol., Vol. 21, No. 9, 2008 1817

Table 2. Determination of the Enzyme Kinetics for Covalent Binding and Total Metabolic Intrinsic Clearance in Pooled Human Liver Microsomesa drug

KM

Vmax

CLint,cb,1

CLint,cb,2

total CLint,cb

metabolismCLint

fCL,cb

acetaminophen benoxaprofen buspirone carbamazepine diclofenac diphenhydramine felbamate ibuprofen indomethacin meloxicam nefazodone paroxetine propranolol raloxifene simvastatin sudoxicam theophylline tienilic acid

160 ND 24 51 19 1.9 ND ND ND 84 10 0.48 3.6 76 ND 250 ND 4.8

14 ND 4.1 2.0 0.42 0.23 ND ND ND 8.0 37 3.6 2.0 110 ND 14 ND 13

0.088 ND 0.17 0.039 0.022 0.12 ND ND ND 0.095 3.7 7.5 0.54 1.5 ND 0.056 ND 2.7

0.030 ND 0.048 ND 0.013 0.0033 ND ND 0.053 ND ND 0.63 0.049 0.040 0.067 ND ND 0.19

0.12 ND 0.22 0.039 0.035 0.12 ND ND 0.053 0.095 3.7 8.1 0.59 1.5 0.067 0.056 ND 2.9

3.6 1.5 200 0.49 48 7.8 ND 26 0.56 2.7 2700 50 13 0.60 420 3.7 0.028 190

0.031