Drug-induced toxicity is a major impediment of drug therapy and drug development. Although toxic drugs should be screened out during preclinical safety studies, in some cases, rare but serious toxicities, so-called idiosyncratic drug toxicities (IDTs) (Li, 2002; Séguin and Uetrecht, 2003; Kaplowitz, 2005), become apparent after the launch of new chemical entities (NCEs). Many drugs, including troglitazone, tienilic acid, clozapine, flutamide, and ticlopidine, have been withdrawn from the market or given a “black box” warning as a result of adverse events that had not been predicted. Currently, it is suggested that the metabolic activation of a drug to a reactive metabolite and its covalent binding to cellular macromolecules are involved in the occurrence of IDTs (Evans et al., 2004; Walgren et al., 2005; Zhou et al., 2005). Therefore, it is important to assess the potential of NCEs to generate reactive metabolites and form drug-protein covalent adducts in the preclinical stages. Several in vitro techniques to evaluate reactive metabolite formation have been developed recently, including the qualitative liquid chromatography-tandem mass spectrometry detection of the reactive metabolites trapped with nucleophiles such as glutathione (GSH) and cyanide (Baillie and Davis, 1993). Because these techniques can provide us with structural information regarding the reactive metabolites trapped with the nucleophiles, they are helpful in understanding the mechanism of metabolic bioactivation and in modifying the structure to minimize metabolic bioactivation. The quantitative assessment of reactive metabolite formation using radiolabeled GSH and fluorophore-tagged GSH has also been reported and utilized (Gan et al., 2005; Masubuchi et al., 2007).

Radiolabeled tracers of NCEs enable us to conduct the study of covalent binding so that the amount of radioactivity covalently bound to proteins can be determined. Covalent binding studies are considered to be the most general and definitive assessments of bioactivation because they can provide direct evidence of the covalent binding of NCEs to proteins. To interpret the covalent binding data in terms of risk assessment, it is important to collect sufficient background data on a variety of marketed drugs by a standard method. Evans et al. (2004) in Merck Research Laboratories have provided protocols for in vitro covalent binding studies in a liver microsomal system and in an in vivo study using rats and a decision tree for assessing the suitability of NCEs for development based on bioactivation considerations. Liver microsomes are employed to assess the intrinsic covalent binding yields catalyzed primarily by oxidative enzymes. Rat in vivo studies can be used to assess the metabolic bioactivation that occurs in systems other than liver microsomal systems. Based on these protocols, the covalent binding data of some drugs known to undergo bioactivation, such as acetaminophen, furosemide, and tienilic acid, were collected by Masubuchi et al. (2007). For further consideration of the relationship between the incidence of drug-induced toxicities and the amount of covalent binding to liver proteins, we have collected the covalent binding data in human and rat liver microsomal systems and rats in vivo of 16 ¹⁴C-labeled “problematic” drugs associated with drug-induced toxicities, including IDTs and “safe” drugs.

Whole-body quantitative autoradiography (ARG) allows us to acquire the tissue distribution/retention properties of ¹⁴C-labeled-NCEs. If tissue retention properties are correlated with covalent binding yields, ARG can be useful for risk assessment. In this study, the tissue distribution/retention of a number of drugs was investigated by a rat ARG to investigate the relationship between covalent binding and tissue retention. We also focused on the relationship between the toxicity profile and drug distribution based on the ARG data.

Materials and Methods

Materials.¹⁴C-Labeled amodiaquine, benzbromarone, clozapine, flutamide, imipramine, nevirapine, sulfamethoxazole, tacrine, valsartan, zafirlukast, and zomepirac were obtained from BlyChem Ltd. (Billingham, UK). ¹⁴C-Labeled aminopyrine, caffeine, erythromycin, phenacetin, procainamide, and valproic acid were purchased from American Radiolabeled Chemicals (St. Louis, MO). ¹⁴C-Labeled phenytoin was purchased from PerkinElmer Life and Analytical Sciences (Waltham, MA). ¹⁴C-Labeled warfarin was purchased from GE Healthcare (Chalfont St. Giles, UK). ¹⁴C-Labeled amlodipine was purchased Moravek Biochemicals (Brea, CA). Unlabeled amlodipine, amodiaquine, benzbromarone, clozapine, erythromycin, flutamide, imipramine, phenytoin, sulfamethoxazole, tacrine, warfarin, and zomepirac were obtained from Sigma-Aldrich (St. Louis, MO). Unlabeled aminopyrine and valproic acid were purchased from Wako Pure Chemicals (Osaka, Japan). Unlabeled caffeine, nevirapine, phenacetin, and zafirlukast were obtained from Fluka (Buchs, Switzerland), Toronto Research Chemicals Inc. (North York, ON, Canada), ULTRA Scientific (North Kingstown, RI), and Cayman Chemical (Ann Arbor, MI), respectively. Unlabeled procainamide and valsartan were synthesized in-house. Pooled human (n = 50, mixed gender) and male Sprague-Dawley rat liver microsomes were purchased from XenoTech, LLC (Lenexa, KS). NADP and glucose-6-phosphate dehydrogenase were purchased from Oriental Yeast Co. Ltd, (Tokyo, Japan) and glucose 6-phosphate was obtained from Sigma-Aldrich. All other reagents and solvents were of the highest grade commercially available.

Animals. Male Crj:CD (SD) IGS rats (4–8 weeks) were obtained from Charles River Laboratories Japan, Inc. (Kanagawa, Japan). The rats were acclimated for 1 week to a 12-h light/dark cycle in a humidity- and temperature-controlled environment and allowed access to food and water until experimental use, whereupon food was withdrawn for 16 to 18 h before administration of the compounds. The rats were cared for and treated in accordance with the National Institute of Health Guidelines for Laboratory Animal Welfare. The protocols were approved by our Institutional Animal Care and Use Committee.

In Vitro Covalent Binding Study. The experimental procedure was based on that previously reported (Evans et al., 2004; Masubuchi et al., 2007). Briefly, the complete incubation of the ¹⁴C-labeled test compounds with HLMs or RLMs containing the following: 10 μM ¹⁴C-labeled test compound (substrate), 2 mg/ml HLMs/RLMs, 100 mM potassium phosphate buffer, pH 7.4, 25 mM glucose 6-phosphate, 2 units/ml glucose-6-phosphate dehydrogenase, and 10 mM MgCl₂. This mixture was preincubated for 3 min at 37°C. A reaction was initiated by the addition of β-NADP⁺ to reach final concentration of 2.5 mM. The final incubation volume was 0.5 ml. Incubation samples without β-NADP⁺ were used as (–) NADP control systems, respectively. As the substrates were dissolved in acetonitrile, the final incubation mixture contained 1% (v/v) acetonitrile. After 1-h incubation, the reaction was terminated by the addition of 0.5 ml of ice-cold acetonitrile. After vortexing and centrifugation, the precipitated protein was washed with 80% (v/v) aqueous methanol containing 10% (w/v) trichloroacetic acid, diethyl ether-methanol [1:1 (v/v)], and 80% (v/v) aqueous methanol (twice for each solvent). The resulting precipitated protein was dissolved in 0.5 ml of 1.0 N NaOH, and aliquots were taken for the protein assay with a DC protein assay kit (Bio-Rad, Hercules, CA) and determination of the radioactive content by liquid scintillation counting with Hionic-fluor scintillation fluid cocktail (PerkinElmer Life and Analytical Sciences). The amount of the test compound-related material as radioactivity covalently bound to the microsomal protein was determined, and the covalent binding (picomoles per milligram of protein) was calculated. As a background, the radioactivity of the samples without incubation was also determined.

In Vivo Covalent Binding Study. The experimental procedure was based on that previously reported (Evans et al., 2004; Masubuchi et al., 2007). Briefly, ¹⁴C-labeled and unlabeled compounds were dissolved or suspended in 0.5% methylcellulose (400 cP) to prepare a solution at a concentration of 2 mg/ml as free form for oral administration to fasted rats. Following a single oral administration of each test compound at a dose of 20 mg/kg to fasting male rats, the rats were exsanguinated at 2, 6, and 24 h (n = 3 animals per time point), and their plasma and liver samples were collected and stored frozen until analysis. The obtained liver samples were weighed and homogenated with aqueous 1.15% (v/v) KCl. In the same way as in the in vitro covalent binding study, the liver homogenate and plasma samples were washed with the organic solvents, followed by a protein assay and the determination of the radioactivity covalently bound to the protein.

Rat Whole-Body Autoradiography. For oral administration to fasted rats, ¹⁴C-labeled and unlabeled compounds were dissolved or suspended in 0.5% (w/v) aqueous methylcellulose (400 cP) to prepare a solution at a concentration of 0.5 mg/ml as a free form. Following a single oral administration of each test compound at a dose of 3 mg/kg (4 MBq/kg) to fasting male rats, the rats were killed by ether inhalation, deep-frozen, and embedded in carboxy-methylcellulose at –70°C. Sagittal sections (30 μm) were cut using a cryomicrotome (CM3600; Leica, Wetzlar, Germany) and lyophilized. Selected sections were exposed to Imaging Plates (TYPB-BAS SR2040; Fuji Photo Film Ltd., Tokyo, Japan) with plastic standards for quantification in cassettes. After exposure, the Imaging Plates were scanned using an imaging analyzer (FUJIX BAS2500; Fuji Photo Film Ltd.) to obtain the autoradiograms. The radioactivity concentrations in the tissues of interest were calculated from the standard curves derived from the plastic standards.

Results

In Vitro Covalent Binding Study. The in vitro covalent binding yields of 16 problematic and four safe drugs were determined by incubation for 1 h with HLMs and RLMs. The structures, the proposed reactive metabolite species, and the reported toxicities caused by the reactive metabolites of the drugs tested are shown in Table 1. The covalent binding yields of radioactivity from the ¹⁴C-labeled drugs tested incubated with HLMs are shown in Table 2. In the complete incubations with HLMs, benzbromarone exhibited the highest level of covalent binding (389.9 ± 18.9 pmol/mg protein), and amodiaquine, flutamide, tacrine, and imipramine exhibited relatively high levels of covalent binding (208.1 ± 13.4, 178.0 ± 10.9, 137.0 ± 7.5, and 133.8 ± 7.0 pmol/mg protein, respectively). The NADPH-dependent covalent binding values of benzbromarone, flutamide, and tacrine were also at a high level (>100 pmol/mg protein), indicating that their covalent binding to liver microsomal proteins occurred via NADPH-dependent metabolism. On the other hand, amodiaquine and imipramine exhibited relatively high levels of covalent binding in the (–)NADP incubation with HLMs (258.2 ± 14.7 and 91.0 ± 15.7 pmol/mg protein), showing that their covalent binding contained NADPH-independent processes. Covalent binding studies of the drugs tested using RLMs showed results similar to HLMs (Table 3). In the complete incubations with RLMs, benzbromarone exhibited the highest level of covalent binding (313.0 ± 79.4 pmol/mg protein), and imipramine, amodiaquine, clozapine, and flutamide exhibited relatively high levels of covalent binding (234.7 ± 11.2, 125.6 ± 11.4, 117.6 ± 4.3, and 106.5 ± 0.1 pmol/mg protein, respectively). Amodiaquine and imipramine also exhibited high levels of covalent binding in the (–)NADP incubation with RLMs (94.3 ± 2.7 and 130.6 ± 21.0 pmol/mg protein). The safe drugs amlodipine, caffeine, valsartan, and warfarin gave levels of covalent binding lower than 50 pmol/mg protein in both HLMs and RLMs.

To observe any species differences in covalent binding, the covalent binding yields of the drugs tested in the complete incubation with HLMs were plotted versus that with RLMs, as shown in Fig. 1. The log-linear regression line was calculated by the least-squares method. The correlation coefficient (R) between HLMs and RLMs, the in vitro covalent binding of the drugs tested, was calculated to be 0.84. In the case of tacrine, the covalent binding in HLMs was approximately 7-fold higher than that in RLMs. In the cases of aminopyrine, clozapine, nevirapine, phenytoin, procainamide, sulfamethoxazole, and warfarin, on the contrary, the covalent binding in RLMs was 2- to 5-fold higher than that in HLMs.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Correlation between HLM and RLM in vitro covalent binding of the drugs tested.

In Vivo Covalent Binding Study. The potential of the in vivo metabolic bioactivation of the drugs was assessed in rats. After a single oral administration of 20 mg/kg doses of ¹⁴C-labeled drugs to rats (n = 3), radioactivity irreversibly bound to proteins in the liver, and plasma homogenates were detected using a liquid scintillation counter. The relatively high dose (20 mg/kg) was chosen to highlight the potential of metabolic bioactivation and to balance maximizing analytical sensitivity with a standardizing protocol. The maximum levels of radioactivity covalently bound to rat liver and plasma proteins are shown in Table 4. Aminopyrine, imipramine, clozapine, valproic acid, and amodiaquine exhibited relatively high levels of covalent binding to liver proteins, with levels of more than 100 pmol/mg protein. Aminopyrine, imipramine, and valproic acid also gave high levels (>100 pmol/mg protein) of covalent binding to plasma proteins, whereas most of the other drugs showed only trace levels (<10 pmol/mg protein). Thus, the levels of covalent binding to liver and plasma protein did not correspond exactly.

To investigate the in vitro-in vivo correlation of the covalent binding, the rat in vivo covalent binding to liver proteins was plotted to the in vitro RLM covalent binding, as shown in Fig. 2. The correlation coefficient (R) between RLM in vitro and rat in vivo covalent binding yields of the drugs tested was calculated to be 0.47 from a log-linear regression analysis; thus, a quite weak correlation between them was observed. Valproic acid, aminopyrine, and zomepirac exhibited significantly high levels of rat in vivo covalent binding relative to those of RLM in vitro covalent binding, with levels of approximately 30-, 6-, and 5-fold, respectively. Conversely, benzbromarone showed a markedly low level of rat in vivo covalent binding relative to those of RLM in vitro covalent binding.

Rat Quantitative Whole-Body Autoradiography. The distribution and retention of total radioactivity were investigated by means of rat quantitative whole-body autoradiography. Eleven drugs, including amodiaquine, benzbromarone, caffeine, clozapine, flutamide, nevirapine, imipramine, sulfamethoxazole, tacrine, zafirlukast, and zomepirac, were tested. Autoradiograms were taken at 3, 6, 24, 72, and 168 h after a single oral administration of ¹⁴C-labeled compounds (3 mg/kg) to fasted male rats. Because washout of the free and reversibly bound fraction of radioactivity is critical for detecting the covalently bound fraction of radioactivity in this experiment, we chose the relatively low dose (3 mg/kg). Typical autoradiograms at 72 and 168 h postdose are shown in Fig. 3. The temporal changes of the concentration (percentage to the maximum concentration) in some tissues are shown in Fig. 4. In most of the tissues, the maximum levels of radioactivity were observed at 3 or 6 h after administration, and then the levels of radioactivity were decreased. Amodiaquine, clozapine, flutamide, and tacrine gave residual radioactivity, 2 to 6% of the maximum radioactivity concentration, in the liver up to 168 h after administration. Amodiaquine and clozapine, which are associated with agranulocytosis, also showed the retention of radioactivity in the bone marrow and spleen up to 168 h after administration. In the cases of benzbromarone, nevirapine, zafirlukast, and zomepirac, slight residual radioactivity, 1 to 3% of the maximum radioactivity concentration, was detected in the liver at 72 h after administration and was eliminated at 168 h. Imipramine exhibited the retention of radioactivity in almost all the tissues observed up to 168 h after administration. In the case of sulfamethoxazole, the radioactivity was almost completely eliminated from all tissues at 72 h after administration.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Plots of rat in vivo covalent binding to liver proteins versus RLM in vitro covalent binding of the drugs tested.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Typical whole-body autoradiograms (left axis aspect) 72 and 168 h after a single oral administration of ¹⁴C-labeled drugs at a dose of 3 mg/kg to rats. A, amodiaquine; B, clozapine; C, flutamide.

Next, we investigated the relationship between the retention profiles of radioactivity in the liver and the covalent binding of radioactivity from the drugs to liver proteins. The plots of the residual radioactivities in liver (% to the maximum concentration) of the 11 drugs observed by means of a rat ARG versus the covalent binding yields to liver proteins obtained by rat in vivo covalent binding studies are shown in Fig. 5. The retention of the radioactivity in the liver was found to be well correlated with the rat in vivo covalent binding to liver proteins. The correlation coefficients between the residual radioactivities at 24, 72, and 168 h after administration and the covalent binding yields were 0.82, 0.91, and 0.91, respectively.

Discussion

Because covalent binding studies are presently considered to provide the most definitive data for the risk assessment of IDT, the covalent binding data of NCEs usually become available at a late preclinical stage and influence their fates as candidates for further development. To make decisions based on the covalent binding data, it is important to accumulate the reference data of a variety of marketed drugs by a standard method. Most of the drugs used in the present study are known to form various reactive metabolites (Kalgutkar et al., 2005), as shown in Table 1. This includes several types of reactive metabolites, such as quinone imine derived from amodiaquine (Christie et al., 1989); ortho-quinone from benzbromarone (McDonald and Rettie, 2007); quinone methide from tacrine (Madden et al., 1993); nitrenium ion from clozapine (Pirmohamed et al., 1995); iminium ion from nevirapine (Kalgutkar et al., 2005) and zafirlukast (Kassahun et al., 2005); nitroso species from erythromycin (Larrey et al., 1983), phenacetin (Koymans et al., 1990), procainamide (Uetrecht, 1985), and sulfamethoxazole (Naisbitt et al., 2002); arene oxide from imipramine (Masubuchi et al., 1996); acyl glucuronide from valproic acid (Baillie, 1988) and zomepirac (Wang et al., 2001); and radical species from aminopyrine (Uetrecht et al., 1995), flutamide (Kang et al., 2007), and phenytoin (Cuttle et al., 2000). Quinone imine formation from amodiaquine was reported to be caused by autoxidation (Maggs et al., 1987); thus, the high levels of covalent binding without NADPH observed in this study seem to be reasonable. In the case of imipramine, as far as we know, no NADPH-independent bioactivation pathway has been reported, and the reason for the high levels of covalent binding without NADPH remains unclear. Presumably, some unknown NADPH-independent pathway might contribute to the covalent binding of imipramine. The formation of reactive metabolites is thought to cause a variety of adverse events (Kaplowitz, 2005; Zhou et al., 2005). Amodiaquine, benzbromarone, erythromycin, flutamide, nevirapine, tacrine, valproic acid, zafirlukast, and zomepirac cause hepatotoxicity. As a result of their hepatotoxicity, amodiaquine and zomepirac were withdrawn from the market, and benzbromarone, flutamide, nevirapine, and valproic acid received black box warnings. Aminopyrine, amodiaquinee, and clozapine are known to cause severe and life-threatening agranulocytosis; thus, aminopyrine was withdrawn from the market, and clozapine received a warning. Imipramine, phenacetin, phenytoin, procainamide, and sulfamethoxazole are associated with hypersensitivity reactions. Thus, the drugs in the present study are useful in investigating the relationship between the covalent binding profile and the occurrence of drug-induced toxicities.

The HLM in vitro covalent binding yields of these drugs indicated that the levels of covalent binding of most of the problematic drugs were higher than those of the safe drugs. Evans et al. (2004) proposed a rationale for compound evaluation; both in vitro HLM and rat in vivo covalent binding yields of <50 pmol/mg protein are desirable. Considering the HLM in vitro covalent binding data in light of these criteria, the covalent binding yields of the safe drugs were lower than 50 pmol/mg protein, and the drugs exhibiting higher levels of covalent binding than 50 pmol/mg protein were all problematic. However, the drugs that exhibited levels of in vitro HLM covalent binding lower than 50 pmol/mg protein include some problematic drugs containing the “withdrawn” drugs such as aminopyrine and zomepirac and black box warning drugs such as clozapine, valproic acid, and nevirapine. Therefore, it appears to be difficult to completely predict the risk of toxicity only by judgments based on the in vitro HLM covalent binding yields. The rat in vivo covalent binding yields of the drugs were different from those of the HLM in vitro covalent binding yields. Meanwhile, the safe drugs exhibited a level of covalent binding lower than 50 pmol/mg protein, and the drugs exhibiting levels of covalent binding higher than 50 pmol/mg protein were all problematic drugs. In particular, aminopyrine, clozapine, nevirapine, and valproic acid showed high levels (>50 pmol/mg protein) of rat in vivo covalent binding relative to those of HLM in vitro covalent binding (<50 pmol/mg protein). Therefore, most (seven of eight) of the withdrawn or black box warning drugs exhibited levels of covalent binding higher than 50 pmol/mg protein in the in vitro HLM assay and/or rat in vivo assay. As shown in Fig. 2, the plots of the in vitro-in vivo correlation revealed large differences between the in vitro and in vivo covalent bindings of aminopyrine and valproic acid. From these observations, it is suggested that aminopyrine and valproic acid have a major bioactivation pathway other than P450-mediated oxidation in liver microsomal systems. In fact, it has been reported that aminopyrine could be metabolized to a reactive species by myeloperoxidase in activated neutrophils (Uetrecht et al., 1995). The bioactivation pathways of valproic acid are known to be glucuronidation of the acyl group and mitochondrial β-oxidation (Baillie, 1988; Bailey and Dickinson, 1996). Because it has been proven that a rat in vivo covalent binding study can find drugs with a risk of bioactivation other than P450-mediated oxidation, such as aminopyrine and valproic acid, this type of study will be essential for risk assessment, in addition to an HLM in vitro study. Also, for HLM in vitro and rat in vivo covalent binding studies, a covalent binding study using hepatocytes would be feasible and effective because it can serve to reveal the metabolic bioactivation processes that depend on the presence of a full component of cellular enzyme systems.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Temporal changes of the concentration (percentage to the maximum concentration) in some tissues. A, amodiaquinee; B, benzbromarone; C, caffeine; D, clozapine; E, flutamide; F, imipramine. G, sulfamethoxazole; H, nevirapine; I, tacrine; J, zafirlukast; K, zomepirac.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Plots of residual radioactivity ratios of the 11 drugs observed in the liver in rat quantitative whole-body ARG versus rat in vivo covalent binding to liver proteins.

In rat in vivo covalent binding study, some drugs exhibited higher levels of covalent binding to plasma proteins than those to liver proteins. The level of covalent binding to plasma proteins is considered to be dependent on the distribution, reactivity, and lifetime of reactive metabolites in plasma. As in the case with aminopyrine, the drugs bioactivated by reactive oxygen species generated from leukocytes in blood seem to exhibit the high levels of covalent binding to plasma proteins. Also, if a reactive metabolite generated in liver is relatively stable, such as the acyl glucuronide derived from valproic acid, it can be translocated from liver to blood and can bind to plasma proteins.

Rat ARG visually gives us information regarding the tissue distribution and retention properties of the drugs. Because the residual radioactivity observed in the liver at 72 or 168 h postdose was found to be well correlated with the rat in vivo covalent binding to liver proteins, as shown in Fig. 5, the major fraction of the long-term retained radioactivity was thought to be covalently bound to liver proteins. Thus, it seems possible that the in vivo covalent binding yields of the drugs should be extrapolated from their retention profiles observed by means of ARG. From the obtained data, drugs exhibiting residual radioactivity of more than 5% to the maximum radioactivity in liver at 72 h postdose or that more than 2% at 168 h postdose will show high levels (>50 pmol/mg protein) of covalent binding yields to the liver with a high probability. It is exceptional that caffeine exhibited residual radioactivity in several tissues, including the liver, although it does not undergo bioactivation and form covalent binding. The reason for this remains unclear, but presumably, caffeine is metabolized to nucleobases such as xanthines and is assimilated into endogenous molecules in tissues.

Regarding tissues other than the liver, long-term retention of radioactivity in the bone marrow was observed with some drugs associated with severe agranulocytosis, such as amodiaquine and clozapine, implying that the covalent binding of their reactive species to the proteins in bone marrow might cause agranulocytosis. The mechanism of drug-induced agranulocytosis has not been fully defined. However, it has been proposed that the covalent binding of reactive metabolites to neutrophil proteins could lead to agranulocytosis, either by direct toxicity or through an immune-mediated reaction (Pumford and Holmes, 1997; Gardner et al., 1998). Alternatively, it also has been proposed that the stromal cells in the bone marrow, which are required for the growth and maturation of neutrophil precursors, rather than neutrophils or their precursors, are the targets of toxicity (Guest and Uetrecht, 1999). A long-lived drug covalently bound to proteins can be a hapten and lead to an immune response when stimulated with “danger signals” caused by cellular damage or stress. From the point of view of the hapten and danger hypotheses (Uetrecht, 2008), the long-term retention of drug-protein adducts is considered to be a risk factor for IDT. It appears to be important and useful as a risk assessment for drug-induced agranulocytosis to examine the retention of radioactivity in bone marrow by means of ARG.

In conclusion, we have determined the HLM in vitro and rat in vivo covalent binding yields of a variety of drugs by a standard method. Most of problematic drugs, including withdrawn and “warning” drugs, exhibit higher HLM in vitro covalent binding yields than safe drugs, and some problematic drugs that undergo bioactivation other than P450-mediated oxidation exhibited only trace levels of HLM covalent binding, like safe drugs. Because rat in vivo covalent binding studies can assess the bioactivation of such drugs, they are essential for risk assessment, in addition to an HLM in vitro study. Although the criteria (50 pmol/mg protein) from Merck have some false negatives, they are likely to be useful because most (seven of eight) of the withdrawn or black box warning drugs exhibited levels of covalent binding higher than 50 pmol/mg protein in the in vitro HLM assay and/or rat in vivo assay. Accumulation of safe drug data is still important to improve the reliability of the risk border for covalent binding studies. Furthermore, by means of ARG, we have also shown the strong correlation between the covalent binding yields of the drugs and the long-term retention of radioactivity in autoradiograms and the spatial relationship between the toxicity profile and drug distribution/retention. ARG enables us to estimate the covalent binding yields in various tissues at a time without extensive washing procedures, suggesting that ARG would be useful for IDT risk assessment. Taken together, the covalent binding and tissue distribution/retention data of the various marketed drugs obtained in this study are quite informative for the interpretation of data in terms of risk assessment.