Article pubs.acs.org/crt
Benchmarking in Vitro Covalent Binding Burden As a Tool To Assess Potential Toxicity Caused by Nonspecific Covalent Binding of Covalent Drugs Upendra P. Dahal,*,†,‡ R. Scott Obach,† and Adam M. Gilbert‡ †
Pharmacokinetics, Dynamics and Metabolism and ‡Worldwide Medicinal Chemistry, Pfizer Inc., Groton, Connecticut 06340, United States S Supporting Information *
ABSTRACT: Despite several advantages of covalent inhibitors (such as increased biochemical efficiency, longer duration of action on the target, and lower efficacious doses) over their reversible binding counterparts, there is a reluctance to use covalent inhibitors as a drug design strategy in pharmaceutical research. This reluctance is due to their anticipated reactions with nontargeted macromolecules. We hypothesized that there may be a threshold limit for nonspecific covalent binding, below which a covalent binding drug may be less likely to cause toxicity due to irreversible binding to off-target macromolecules. Estimation of in vivo covalent binding burden from in vitro data has previously been used as an approach to distinguish those agents more likely to cause toxicity (e.g., hepatotoxicity) via metabolic activation to reactive metabolites. We have extended this approach to nine covalent binding drugs to determine in vitro covalent binding burden. In vitro covalent binding burden was determined by incubating radiolabeled drugs with pooled human hepatocytes. These data were scaled to an estimate of in vivo covalent binding burden by combining the in vitro data with daily dose. Scaled in vivo daily covalent binding burden of marketed covalent drugs was found to be under 10 mg/day, which is in agreement with previously reported threshold value for metabolically activated reversible drugs. Covalent binding was also compared to the intrinsic reactivities of the covalent inhibitors assessed using nucleophiles glutathione and N-α-acetyl lysine. The intrinsic reactivity did not correlate with observed in vitro covalent binding, which demonstrated that the intrinsic reactivity of the electrophilic groups of covalent drugs does not exclusively account for the extent of covalent binding. The ramifications of these findings for consideration of using a covalent strategy in drug design are discussed.
■
INTRODUCTION Drugs produce their activities by binding to macromolecules to inhibit/activate downstream biological pathways. The effects of drug molecules are usually driven by reversible binding to the macromolecules, but a number of key drugs elicit their effects via covalent binding. The inhibitor containing an electrophilic group that forms a covalent bond with a nucleophile in an active site of the targeted macromolecule is called a covalent inhibitor. Covalent inhibitors can possess advantages over their reversible binding counterparts, such as increased biochemical efficiency, longer duration of action when target resynthesis rate is slow, and/or the potential for improved therapeutic indices due to lower efficacious doses.1 However, covalent inhibitors also have the potential disadvantage of causing hypersensitivity, particularly if the covalent inhibitor lacks specificity.2,3 This concern has prevented many pharmaceutical research organizations from pursuing them as clinical agents. A number of reviews have recently appeared highlighting efforts to achieve safe covalent inhibitors.4−7 The major concern for irreversible inhibitors is that the electrophilic moieties, often called “warheads”, can react covalently with unintended biological targets and cause toxicity.8 However there are a number of covalent inhibitors © 2013 American Chemical Society
used in clinical practice that do not appear to be associated with high frequencies of toxicity, e.g., aspirin and omeprazole. While covalent binding of the inhibitors with unintended targets cannot be completely ruled out, a tolerance limit for nonspecific covalent binding of covalent drugs can be envisioned, below which a covalent drug may be less likely to cause toxicity. Such a benchmark could serve as a decision criterion for further progression or discontinuation of a drug candidate and better ensure safety of covalent inhibitors in clinical research and use in medical practice. Drug-related toxicity assessment and mitigation is one of the challenging research fields and has gained considerable interest by various research groups.9−15 These approaches were primarily focused on idiosyncratic adverse drug reactions and drug-induced liver toxicity of noncovalent (reversible) inhibitors. Noncovalent inhibitors are not usually toxic in the form in which they are administered but may become toxic16 after bioactivation by drug-metabolizing enzymes that convert the parent drug into reactive electrophilic metabolites that can covalently bind to proteins and nucleotides. In addition, Received: August 19, 2013 Published: October 10, 2013 1739
dx.doi.org/10.1021/tx400301q | Chem. Res. Toxicol. 2013, 26, 1739−1745
Chemical Research in Toxicology
Article
was removed. The pellet was resuspended, and the extraction was repeated at least six times with organic solvent (CH3CN/CH3OH, 4:1). Supernatant fractions were monitored until no radioactivity was detected. Sodium hydroxide (0.5 mL, 1 M) was added to the remaining protein pellet and placed in a shaking water bath at 37 °C overnight to dissolve the pellet, and the total radioactivity of the resulting mixture was determined by liquid scintillation counting. Radioactivity remaining in pellets was used to calculate in vitro covalent binding (pmol/mg protein) using eq 1. Percent covalent binding was determined by applying eq 2. Tritium can be removed by metabolism, and hence metabolic lability for tritiated compound was considered and the percentage loss of the tritium was determined in separate assay (experimental and results were presented in Supporting Information) to correct the amount of available drug used in eq 2. Daily covalent binding burden was determined from daily dose using eq 3.
intrinsically electrophilic noncovalent inhibitors, including drugs that form protein adducts, have been developed inadvertently.17,18 For example TAK-242, a novel small molecule Toll-like receptor (TLR) 4 signal transduction inhibitor, was found to be reactive with glutathione, human serum albumin, and rat serum albumin in both in vitro and in vivo systems.18 Removing and/or altering the bioactivation site can remove/alleviate the toxicity of noncovalent inhibitors. On the other hand, it is not possible to remove a reactive electrophilic group from a covalent inhibitor because the electrophilic moiety is the key and integral motif for biological activity. However, one could more confidently advance covalent inhibitors as drug candidates if guided by benchmarked tolerance limit for in vitro covalent binding burden for covalent drugs that translated into a safe amount of binding in vivo. In vitro covalent binding has previously been used to estimate in vivo covalent binding burden in association with hepatotoxicity.14,15 Similarly, many of the toxicity assessment approaches9,12,14,15 have applied in vitro covalent binding as a major tool to distinguish drugs with regard to toxicity. Nakayama et al.9 developed the zonal classification by plotting covalent binding of the drugs against the dose of the drugs to distinguish idiosyncratic drug toxicity. Similarly, Bauman et al.14 used covalent binding burden to distinguish liver-toxic drugs from nontoxic drugs. In this study, we have determined the in vitro covalent binding and corresponding scaled in vivo covalent binding burden of a cohort of marketed covalent inhibitor drugs in an attempt to determine the tolerance limit of nonspecific covalent binding associated with toxicity. In addition we determined intrinsic reactivities of the inhibitors to determine if there is any correlation between intrinsic reactivity of electrophilic groups and covalent binding burden of the covalent drugs.
■
covalent binding (pmol/mg protein) =
total radioactivity in pellets (dpm) × 106 cells/mg protein specific activity (dpm/pmol) × no. of cells (1)
percent covalent binding =
covalently bound drug (pmol) × 100% total available drug (pmol)
covalent binding burden (mg/day) % covalent binding = × daily dose (mg/day) 100
(2)
(3)
Intrinsic Reactivity with Glutathione (GSH) and N-α-Acetyl Lysine. Reactions were carried out with the ReactArray Liquid Handler System coupled with an Agilent 1100 HPLC/MS for quantitative analysis. The temperature setting for all reactions was 37 °C. All stock solutions and solvents were bubbled with nitrogen for at least 6 h, and the reactions were performed under nitrogen. A 250μL aliquot of 10.0 mM solution of electrophile in N,Ndimethylacetamide was manually transferred to a reaction vial, and 250 μL of a 2.0 mM solution of indoprofen (used as internal standard in MS analysis) in N,N-dimethylacetamide was automatically transferred to the reaction vial using the Liquid Handler System of the ReactArray. A 4.5-mL aliquot of nucleophile solution (11.1 mM solution of GSH in 100 mM potassium phosphate buffer pH 7.4 for GSH reactivity or 55.5 mM solution of N-α-acetyl lysine in 100 mM borate buffer pH 10.2 for lysine reactivity) was automatically transferred to the reaction vial using ReactArray’s Liquid Handler System. The reactions were initiated upon addition of nucleophile. Control incubations were run in the absence of nucleophile. The liquid chromatography was carried out using mobile phases solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in methanol) and an Atlantis T3 column (3.0 mm × 75 mm, 3 μm) with 1 mL/min flow rate. The liquid chromatography was initiated with 95% mobile phase A and was held for 2 min followed by linear increase to make the solvent B 95% in 6 min. The 95% of solvent B was held for 2 min and then rapidly decreased to 5% in 0.1 min and held for 2 min. The masses of parent covalent drug, GSH or N-α-acetyl lysine adduct and internal standard were monitored in positive ion mode using electro spray ionization technique (mass range 150−900 Da, ionization switch delay 0.05 s, polarity switch delay 0.3 s, capillary voltage 3000 V, nebulizer pressure 45 psig). Pseudo-first-order rate constants were determined by plotting the natural log of the relative concentration of electrophile against time. The negative slope of the straight line is the pseudo-first-order rate constant. The pseudo-first-order rate constants were used to determine half-life using eq 4.
EXPERIMENTAL SECTION
Materials. Radiolabeled compounds were obtained from three sources. [14C]-Aspirin (55 mCi/mmol specific activity and 99.0% radiochemical purity) was purchased from American Radiolabeled Chemicals, Saint Louis, MO. [3H]-Orlistat (320 mCi/mmol specific activity and 97.2% radiochemical purity), [3H]-vildagliptin (510 mCi/ mmol specific activity and 97.9% radiochemical purity), [3H]phenoxybenzamine (5.35 Ci/mmol specific activity and 95.0% radiochemical purity), [3H]-exemestane (330.85 Ci/mmol specific activity and 98.9% radiochemical purity), [3H]-compound 8 (300 mCi/mmol specific activity and 96.1% radiochemical purity), and [3H]-canertinib (32.68 Ci/mmol specific activity and 95.5% radiochemical purity) were custom synthesized by PerkinElmer Health Sciences, Boston MA. [3H]-Ethacrynic acid (6 Ci/mmol specific activity and 99.6% radiochemical purity) and [3H]-cladribine (10 Ci/ mmol specific activity and 99.1% radiochemical purity) were custom synthesized by Quotient Bioresearch, Cardiff, U.K. The compounds were used without further purification. Radioactivity was measured using liquid scintillation counter (PerkinElmer Wallac 1409 DSA). Cryopreserved pooled human hepatocytes from 10 donors were purchased from In Vitro Technologies (Baltimore, MD), and Williams’ E media was from Invitrogen Corporation. In Vitro Covalent Binding Burden Determination. Pooled cryopreserved human hepatocytes were suspended in Williams’ E media at a final concentration of 0.75 × 106 viable cells/mL. Cell viability was based on the tryptan blue staining test. The cell suspensions (2 mL) were incubated at 37 °C in 95% oxygen and 5% carbon dioxide environment for 4 h with radiolabeled substrate (1 μM). One milliliter of cell suspension was quenched by adding it into 5 mL of organic solvent (CH3CN/CH3OH, 4:1). The mixture was vortexed (5 min), sonicated (5 min, 25 °C; Fisher Scientific FS220H, 8 A, 881 W), and centrifuged (3000 rpm, 5 min), and the supernatant
half life = 1740
ln(2) pseudo‐first‐order rate constant
(4)
dx.doi.org/10.1021/tx400301q | Chem. Res. Toxicol. 2013, 26, 1739−1745
■
Chemical Research in Toxicology
Article
RESULTS Selection of Covalent Inhibitors. There are 39 covalent drugs approved by the U.S. FDA.4 Drugs used in this study are shown in Figure 1 and were selected on the basis of the nature
orlistat, and aspirin, are currently in clinical use with low frequencies of reported toxicities, whereas two inhibitors, compound 8 and canertinib, are not in clinical practice. Compound 8 was selected on the basis of similarity to agents being investigated as kinase inhibitors for cancer indications. Canertinib was stopped for further development due to lack of efficacy and adverse side effects in clinical trials.19,20 In Vitro Covalent Binding of the Drugs in Human Hepatocytes. In vitro covalent binding parameters are presented in Table 1. To assess variability in protein concentration and hepatocyte activity between hepatocyte pools, a positive control was carried out with radiolabeled 14 C-raloxifene. The in vitro covalent binding of the raloxifene was found to be consistent (41.1 ± 5.4 pmol/mg of protein) throughout the study. Vildagliptin, cladribine, and aspirin showed single digit in vitro covalent binding, whereas exemestane, phenoxybenzamine, ethacrynic acid, and orlistat showed in vitro covalent binding of 10−50 pmol/mg of protein. Compound 8 and canertinib showed covalent binding of 102 and 179 pmol/mg of protein, respectively. The covalent binding data was converted to percent covalent binding using eq 2. The percent covalent binding was used to estimate the daily covalent binding burden as a product of percent covalent binding and daily dose of the covalent inhibitors. Estimated Daily Covalent Binding Burden of the Drugs in Human Hepatocytes. Estimated daily covalent binding burdens scaled from in vitro data are presented in Table 1. Three covalent inhibitors, vildagliptin, exemestane, and cladribine, showed daily covalent binding burden less than 1 mg/day; four covalent inhibitors, phenoxybenzamine, ethacrynic acid, orlistat, and aspirin, showed daily covalent binding burden of 1−10 mg/day; and two compounds, compound 8 and canertinib, showed daily covalent binding burden above 10 mg/day. Aspirin has various dosing regimens. It has been reported that aspirin can be taken up to 4 g/day for various indications including pain and fever, whereas a low dose daily regimen (80 mg/day) is also recommended for long-term use to reduce heart attack. The commonly recommended maximum daily dose of 1950 mg/day was used in daily covalent binding burden calculation. Canertinib was calculated to have the highest daily covalent binding burden of 44.4 mg/ day. Intrinsic Reactivity of Covalent Inhibitors with Glutathione and N-α-Acetyl Lysine. To determine if there is any correlation between covalent binding and intrinsic reactivities of the drugs with biological nucleophiles such as thiol and amine nucleophiles, intrinsic reactivity assessments using GSH and N-α-acetyl lysine were carried out. GSH is one of the abundantly found nucleophiles in liver. N-α-Acetyl lysine was used as a model compound to determine the intrinsic reactivity of the electrophiles with lysine residues of the proteins. For GSH reactivity, 0.5 mM electrophile was reacted with 10 mM GSH in phosphate (0.1 M) buffer pH 7.4 at 37 °C. Since the Lys-ε-NH2 moiety is protonated at pH 7.4, N-αacetyl-lysine/electrophile reactivity was not detected at pH 7.4. As a result, the N-α-acetyl lysine reactivity studies were carried out in borate buffer (0.1 M) at pH 10.2 using 0.5 mM electrophile and 50 mM N-α-acetyl-L-lysine at 37 °C. In both cases, pseudo-first-order rate constants were determined by plotting the natural log of the relative concentration of electrophile versus time. Aspirin and phenoxybenzamine were hydrolyzed quantitatively, so the reactivity with nucleophiles could not be determined (Table 2). Vildagliptin, cladribine,
Figure 1. Structures of the covalent inhibitors used in the study. Arrows indicate the electrophilic sites.
of the electrophile and availability of radiolabeled material. The selected drugs contain both soft electrophilic moieties (α,βunsaturated ketones, α,β-unsaturated amides, nitriles) as seen in exemestane, vildagliptin, ethacrynic acid, compound 8 and canertinib and hard electrophilic moieties (heteroaryl chlorides, alkyl chlorides, lactones and esters) as seen in cladribine, orlistat, phenoxybenzamine, and aspirin. Daily doses of the drugs were also taken into consideration during inhibitor selection given the hypothesis regarding the relationship between dose and idiosyncratic adverse drug reaction.11 The selected drugs have daily maximum doses ranging from 25 mg for exemestane to 1950 mg for aspirin. Among the nine compounds selected for the study, seven inhibitors, vildagliptin, exemestane, cladribine, phenoxybenzamine, ethacrynic acid, 1741
dx.doi.org/10.1021/tx400301q | Chem. Res. Toxicol. 2013, 26, 1739−1745
Chemical Research in Toxicology
Article
Table 1. In Vitro Covalent Binding Parameters of the Drugs with Human Hepatocytes, Daily Dose, and Therapeutic Area of the Covalent Drugs Selected for the Study (average ± SD; n = 3)a drug vildagliptin exemestane cladribine phenoxybenzamine ethacrynic acid orlistat aspirin compound 8 canertinib
covalent binding (pmol/mg of protein) 1.80 14.6 5.18 11.4 47.6 19.7 4.81 102 179
± ± ± ± ± ± ± ± ±
0.10 0.51 0.62 1.0 1.8 1.5 0.031 9.1 7.3
percent covalent binding
daily dose (mg)
± ± ± ± ± ± ± ± ±
50 25 70 120 50 360 1950 320g 225
0.18 1.5 0.55 1.2 5.2 2.2 0.48 12 20
0.02 0.06 0.06 0.1 0.2 0.3 0.01 1.0 0.8
daily covalent binding (mg)
therapeutic area
± ± ± ± ± ± ± ± ±
metabolic cancer cancer cardiovascular cardiovascular gastrointestinal inflammation cancer cancer
0.090 0.38 0.38 1.5 2.6 7.9 9.4 37 44
0.01 0.01 0.04 0.1 0.1 0.9 0.08 3 1
year drug launched
FDA drug label
2008 1999 1993 1953 1965 1999 1899
b c d d d e f
a
The results are presented in increasing order of estimated daily covalent binding burden value. bVildagliptin was launched in Europe; EMA label: should not be used in people who may be hypersensitive (allergic) to drug. cContraindication for those with known hypersensitivity to drug; generally well tolerated and adverse events were usually mild to moderate. dContraindication for those with known hypersensitivity to drug. eAdverse reactions (hypersensitivity reactions) from post marketing experience. fWarning (hypersensitivity reactions). gDose of compound 8 was taken from neratinib due to structural similarity.
Table 2. Half Lives of Covalent Inhibitors with GSH and N-α-Acetyl Lysine (average ± SD; n = 3)a drug vildagliptin aspirin cladribine phenoxybenzamine exemestane orlistat ethacrynic acid compound 8 canertinib
percent covalent binding
half life with GSH (h)b
half life with N-α-acetyl lysine (h)b
electrophilic moiety
hard/soft category
± ± ± ± ± ± ± ± ±
>90 ND >90 ND 13 ± 0.06 >90 0.061 ± 0.01 1.1 ± 0.2 0.15 ± 0.01
>90 ND >90 ND >90 >90 0.77 ± 0.03 13 ± 3 0.42 ± 0.01
nitrile ester hetaromatic halide alkyl halide unsaturated ketone lactone unsaturated ketone acrylamide acrylamide
soft hard hard hard soft hard soft soft soft
0.18 0.48 0.55 1.23 1.53 2.19 5.19 11.5 19.7
0.02 0.01 0.06 0.11 0.06 0.34 0.19 1.0 0.8
a
The electrophilic moiety present in the drugs is indicated along with hard-soft category of the electrophilic group. The results are presented in increasing order of percent covalent binding value. ND = not determined, due to high hydrolysis rate of the compound in aqueous solution. bAssay parameters are in the Experimental Section.
exemestane, and orlistat reacted slowly with N-α-acetyl lysine; the half-lives were found to be longer than 90 h. Half lives of three inhibitors, ethacrynic acid, compound 8, and canertinib, with N-α-acetyl lysine were determined. The reactivity of vildagliptin, cladribine, and orlistat with GSH were monitored for 50 h, and little reactivity was seen. Half lives of exemestane, ethacrynic acid, compound 8, and canertinib were found to be less than 20 h.
when examined under the same conditions, as well as to overcome pragmatic analytical issues due to background noise. The authors also acknowledged that the cutoff value was not an absolute threshold above which a compound should not be advanced into development but other factors weighing risks against benefits should be taken into consideration. Daily dose was also an important factor in the occurrence of adverse drug reactions. Lammert et al.11 reported a statistical relationship between daily dose and idiosyncratic drug-induced liver injury. Similarly Uetrecht24 reported that the occurrence of adverse drug reactions was rare with drugs administered at doses lower than 10 mg/day. Studies from our lab14,15 incorporated daily dose to determine in vivo covalent binding burden of 18 drugs in various biological systems to identify the best in vitro biological matrix to carry out such studies as well as to distinguish toxic and nontoxic drugs using covalent binding burden approach. These studies identified that hepatocytes offered a superior means (vs microsomes and S-9 fractions) to assess drug toxicity due to metabolic bioactivation and proposed 1−10 mg/day in vivo covalent binding burden as a cutoff value to distinguish toxic and nontoxic reversibly binding drugs. Similarly, studies from Nakayama et al.9 and Thompson et al.12 supported that covalent binding in hepatocytes represented a reasonable approach for predicting the toxicities of the drugs. Nakayama et al. proposed a zonal classification system for risk assessment of idiosyncratic drug toxicity using daily dose and covalent binding. Thompson et al. incorporated cellular affects along with covalent binding in their in vitro
■
DISCUSSION Adverse drug reaction is a leading cause of attrition of drug candidates during the discovery, preclinical, and clinical development phases of pharmaceutical research and development. Although exact reasons for adverse drug reactions are often difficult to predict, it has been widely appreciated that reactive metabolites, as opposed to the parent molecules from which they are derived, are responsible for the pathogenesis of some adverse drug reactions.21−23 More specifically, covalent binding of the reactive moiety with biomolecules is a necessary, though not sufficient, step for toxicity. Many pharmaceutical research organizations have developed and implemented procedures9,12−15,23 to evaluate drug candidates for reactive metabolite formation and use such data for rational structural modification to reduce/avoid reactive metabolite formation. Evans et al.23 proposed a target upper limit of covalent binding as 50 pmol drug/mg of protein, which was approximately 20fold less than covalent binding of prototypic hepatotoxins such as acetaminophen, bromobenzene, furosemide, or 4-ipomeanol 1742
dx.doi.org/10.1021/tx400301q | Chem. Res. Toxicol. 2013, 26, 1739−1745
Chemical Research in Toxicology
Article
Figure 2. Daily covalent binding burden of clinically used covalent drugs were found to be under 10 mg/day.
approach to assess potential risk of idiosyncratic adverse reactions caused by candidate drugs. None of the aforementioned reports included any covalent drugs in their studies but focused on predicting idiosyncratic adverse drug reactions from metabolic activation of noncovalent drugs. In this report, we attempted to benchmark daily covalent binding burden of the covalent drugs using radiolabeled covalent drugs and human hepatocytes. To our knowledge, this is the first study to determine daily covalent binding burden of clinically used covalent drugs. Covalent drugs may provide treatment opportunities in some therapeutic areas where reversible counterparts have difficulty inhibiting/activating the biological targets (e.g., kinases),1 but due to possible toxicity resulting from nonspecific covalent binding, covalent drugs are not frequently considered as a viable design strategy. Covalent drugs have a reactive motif that forms a covalent attachment with the intended targeted macromolecules. While it is always preferred to make covalent drugs as selective as possible for the biological target, the potential exists for irreversible reactions with other macromolecules that can lead to hepatotoxicity, blood dyscrasia, and immune-mediated toxicities. This concern has reduced interest in covalent inhibitors as a general drug design strategy.4 Recently, covalent inhibitors have gained considerable interest in biopharmaceutical research laboratories due to their superior efficacy and success in a number of mid to late stage clinical trials: dacomitinib, (Pfizer, phase III), afatinib (Boehringer Ingelhelm, a recent launch), ibrutinib (Pharmacyclics, phase III), neratinib (Puma Biotech, phase III), and carfilzomib (Onyx, phase II). Considering this growing interest in covalent drugs, covalent binding burden of covalent drugs was determined to define tolerance limit for nonspecific covalent binding for covalent drugs. Seven of the selected inhibitors (aspirin, orlistat, vildagliptin, phenoxybenzamine, exemestane, cladribine, and ethacrynic acid) which are currently in clinical use, are not associated with drug-related hypersensitivity based on a literature search. Thus, these drugs can serve as a panel by which the amount of covalent binding in vitro (and extrapolated to in vivo) can be assessed and deemed to be “well tolerated”. The percent covalent binding of the drugs was determined using radiolabeled drugs in human hepatocytes. The percent binding was multiplied with daily dose to estimate the daily covalent binding burden. As seen in Figure 2, the daily covalent binding burdens of the clinically used drugs were under 10 mg/day. It was
observed that the daily dose can have a profound effect in the covalent binding burden. Vildagliptin, exemestane, and cladribine, all of which have daily dose
