The FDA has approved 31 small-molecule kinase inhibitors (KIs) for human use as of November 2016, with six having black box warnings for hepatotoxicity (BBW-H) in product labeling. The precise mechanisms and risk factors for KI-induced hepatotoxicity are poorly understood. Here, the 31 KIs were tested in isolated rat liver mitochondria, an in vitro system recently proposed to be a useful tool to predict drug-induced hepatotoxicity in humans. The KIs were incubated with mitochondria or submitochondrial particles at concentrations ranging from therapeutic maximal blood concentrations (Cmax) levels to 100-fold Cmax levels. Ten endpoints were measured, including oxygen consumption rate, inner membrane potential, cytochrome c release, swelling, reactive oxygen species, and individual respiratory chain complex (I–V) activities. Of the 31 KIs examined only three including sorafenib, regorafenib and pazopanib, all of which are hepatotoxic, caused significant mitochondrial toxicity at concentrations equal to the Cmax, indicating that mitochondrial toxicity likely contributes to the pathogenesis of hepatotoxicity associated with these KIs. At concentrations equal to 100-fold Cmax, 18 KIs were found to be toxic to mitochondria, and among six KIs with BBW-H, mitochondrial injury was induced by regorafenib, lapatinib, idelalisib, and pazopanib, but not ponatinib, or sunitinib. Mitochondrial liability at 100-fold Cmax had a positive predictive power (PPV) of 72% and negative predictive power (NPV) of 33% in predicting human KI hepatotoxicity as defined by product labeling, with the sensitivity and specificity being 62% and 44%, respectively. Similar predictive power was obtained using the criterion of Cmax ≥1.1 µM or daily dose ≥100 mg. Mitochondrial liability at 1–2.5-fold Cmax showed a 100% PPV and specificity, though the NPV and sensitivity were 32% and 14%, respectively. These data provide novel mechanistic insights into KI hepatotoxicity and indicate that mitochondrial toxicity at therapeutic levels can help identify hepatotoxic KIs.Electronic supplementary materialThe online version of this article (doi:10.1007/s00204-016-1918-1) contains supplementary material, which is available to authorized users.

Developing biomarkers for detecting acetaminophen (APAP) toxicity has been widely investigated. Recent studies of adults with APAP-induced liver injury have reported human serum microRNA-122 (miR-122) as a novel biomarker of APAP-induced liver injury. The goal of this study was to examine extracellular microRNAs (miRNAs) as potential biomarkers for APAP liver injury in children. Global levels of serum and urine miRNAs were examined in three pediatric subgroups: 1) healthy children (n=10), 2) hospitalized children receiving therapeutic doses of APAP (n=10) and 3) children hospitalized for APAP overdose (n=8). Out of 147 miRNAs detected in the APAP overdose group, eight showed significantly increased median levels in serum (miR-122, −375, −423-5p, −30d-5p, −125b-5p, −4732-5p, −204-5p, and −574-3p), compared to the other groups. Analysis of urine samples from the same patients had significantly increased median levels of four miRNAs (miR-375, −940, −9-3p and −302a) compared to the other groups. Importantly, correlation of peak serum APAP protein adduct levels (an indicator of the oxidation of APAP to the reactive metabolite N-acetylpara-quinone imine) with peak miRNA levels showed that the highest correlation was observed for serum miR-122 (R=0.94; p<0.01) followed by miR-375 (R=0.70; p=0.05). Conclusion: Our findings demonstrate that miRNAs are increased in children with APAP toxicity and correlate with APAP protein adducts, suggesting a potential role as biomarkers of APAP toxicity.

Circulating microRNAs (miRNAs) have emerged as novel noninvasive biomarkers for several diseases and other types of tissue injury. This study tested the hypothesis that changes in the levels of urinary miRNAs correlate with liver injury induced by hepatotoxicants. Sprague-Dawley rats were administered acetaminophen (APAP) or carbon tetrachloride (CCl(4)) and one nonhepatotoxicant (penicillin/PCN). Urine samples were collected over a 24 h period after a single oral dose of APAP (1250 mg/kg), CCl(4) (2000 mg/kg), or PCN (2400 mg/kg). APAP and CCl(4) induced liver injury based upon increased serum alanine and aspartate aminotransferase levels and histopathological findings, including liver necrosis. APAP and CCl(4) both significantly increased the urinary levels of 44 and 28 miRNAs, respectively. In addition, 10 of the increased miRNAs were in common between APAP and CCl(4). In contrast, PCN caused a slight decrease of a different nonoverlapping set of urinary miRNAs. Cluster analysis revealed a distinct urinary miRNA pattern from the hepatotoxicant-treated groups when compared with vehicle controls and PCN. Analysis of hepatic miRNA levels suggested that the liver was the source of the increased urinary miRNAs after APAP exposure; however, the results from CCl(4) were equivocal. Computational analysis was used to predict target genes of the 10 shared hepatotoxicant-induced miRNAs. Liver gene expression profiling using whole genome microarrays identified eight putative miRNA target genes that were significantly altered in the liver of APAP- and CCl(4)-treated animals. In conclusion, the patterns of urinary miRNA may hold promise as biomarkers of hepatotoxicant-induced liver injury.

The tyrosine kinase inhibitor regorafenib was approved by regulatory agencies for cancer treatment, albeit with strong warnings of severe hepatotoxicity included in the product label. The basis of this toxicity is unknown; one possible mechanism, that of mitochondrial damage, was tested. In isolated rat liver mitochondria, regorafenib directly uncoupled oxidative phosphorylation (OXPHOS) and promoted calcium overload-induced swelling, which were respectively prevented by the recoupler 6-ketocholestanol (KC) and the mitochondrial permeability transition (MPT) pore blocker cyclosporine A (CsA). In primary hepatocytes, regorafenib uncoupled OXPHOS, disrupted mitochondrial inner membrane potential (MMP), and decreased cellular ATP at 1h, and triggered MPT at 3h, which was followed by necrosis but not apoptosis at 7h and 24h, all of which were abrogated by KC. The combination of the glycolysis enhancer fructose plus the mitochondrial ATPase synthase inhibitor oligomycin A abolished regorafenib induced necrosis at 7h. This effect was not seen at 24h nor with the fructose or oligomycin A separately. CsA in combination with trifluoperazine, both MPT blockers, showed similar effects. Two compensatory mechanisms, activation of AMP-activated protein kinase (AMPK) to ameliorate ATP shortage and induction of autophagy to remove dysfunctional mitochondria, were found to be mobilized. Hepatocyte necrosis was enhanced either by the AMPK inhibitor Compound C or the autophagy inhibitor chloroquine, while autophagy inducer rapamycin was strongly cytoprotective. Remarkably, all toxic effects were observed at clinically-relevant concentrations of 2.5-15μM. These data suggest that uncoupling of OXPHOS and the resulting ATP shortage and MPT induction are the key mechanisms for regorafenib induced hepatocyte injury, and AMPK activation and autophagy induction serve as pro-survival pathways against such toxicity.

While conventional parameters used to detect hepatotoxicity in drug safety assessment studies are generally informative, the need remains for parameters that can detect the potential for hepatotoxicity at lower doses and/or at earlier time points. Previous work has shown that metabolite profiling (metabonomics/metabolomics) can detect signals of potential hepatotoxicity in rats treated with doxorubicin at doses that do not elicit hepatotoxicity as monitored with conventional parameters. The current study extended this observation to the question of whether such signals could be detected in rats treated with compounds that can elicit hepatotoxicity in humans (i.e., drug-induced liver injury, DILI) but have not been reported to do so in rats. Nine compounds were selected on the basis of their known DILI potential, with six other compounds chosen as negative for DILI potential. A database of rat plasma metabolite profiles, MetaMap(®)Tox (developed by metanomics GmbH and BASF SE) was used for both metabolite profiles and mode of action (MoA) metabolite signatures for a number of known toxicities. Eight of the nine compounds with DILI potential elicited metabolite profiles that matched with MoA patterns of various rat liver toxicities, including cholestasis, oxidative stress, acetaminophen-type toxicity and peroxisome proliferation. By contrast, only one of the six non-DILI compounds showed a weak match with rat liver toxicity. These results suggest that metabolite profiling may indeed have promise to detect signals of hepatotoxicity in rats treated with compounds having DILI potential.