Biomarkers for metabolic drug activation : towards an integrated risk assessment for drug-induced liver injury (DILI)

The term drug-induced liver injury (DILI) describes adverse effects upon therapeutic drug treatment. They are relatively rare, affecting only 1 of 10000 - 1000000 patients, and remain mostly unpredictable. Due to development of severe hepatotoxicity or death, drugs causing DILI display a high risk for patients and have been withdrawn from the market or severely restricted in use. For the pharmaceutical industry late stage attrition due to DILI represents a big burden stretching development time and effort and generating potential risk at high costs. A better characterization of the disease pattern and its contributing factors is needed. Currently experimental tools to build preclinical mitigation strategies are sparse, but urgently required to help establish an improved risk assessment. One possible mechanism of toxicity involves the formation of chemically reactive metabolites (RM) which interact with cellular macromolecules or signaling pathways. A direct link between RM formation and DILI remains speculative in most cases. Numerous studies of affected drugs demonstrate the plausible involvement of RM formation and subsequent covalent binding to proteins. Still, RMs are not detected for all DILI drugs and RMs do not lead to DILI in every case. Thus, a synergistic effect of multiple (unknown) mechanisms is supposed to result in DILI.
The aim of this work was to review mechanisms leading to DILI, consisting of RM formation and other potentially contributing risk factors such as oxidative stress, cyto- or mitochondrial toxicity. Results were critically evaluated in light of the predictivity for DILI and comprise a gap analysis of current approaches. Biomarkers are proposed as complementary endpoints. Development and validation of analytical methods were conducted for in vitro experiments followed by application of tool compounds to demonstrate the correlation to in vivo studies.
For the in-depth analysis of bioactivation data and its correlation to DILI, a validation set of drugs was selected. These included three groups of compounds, namely those with severe manifestation of DILI, drugs with reported DILI cases and drugs with a history of safe use. Different models were drafted to evaluate quantitative covalent binding as predictive parameter for DILI. The hypothesis was that the intrinsic property of in vitro covalent binding is not a descriptive parameter, as exposure of a toxic drug or metabolite in the body is determined by pharmacokinetic factors. E.g., low clearance drugs might result in experimental false negative results when they are not significantly activated in vitro. Thus, pharmacokinetic properties such as plasma clearance or hepatic inlet concentration were incorporated into the correlation analysis. A quantitative description of the models was established by sensitivity, specificity, precision and negative predictive value. As previously reported, a correlation between covalent binding, the daily dose and DILI was evident. This correlation further improved when adjusted for intrinsic clearance and substituting dose with the theoretical liver inlet concentration. It is further suggested to use glutathione adduct formation as surrogate for covalent binding. This approach was able to separate safe and high risk DILI drugs when evaluated in context of dose and clearance. The correlation did not hold true for medium risk drugs where a big overlap to safe drugs was noticeable. This may be due to equivocal drug classification or the fact that additional factors contribute to the development of DILI.
One of the risk factors contributing to DILI is the excessive overproduction of reactive oxygen species (ROS), i.e. oxidative stress. Oxidative stress can be measured e.g. by cellular damage, biomarkers of lipid peroxidation or secondary signals like gene expression. Isoprostanes were chosen as biomarkers for further investigation. They derive from radical-catalyzed peroxidation of arachidonic acid. Selected isomers of this heterogeneous group were reported as biomarkers of ROS in the past. An online separation chromatography coupled mass spectrometry method was developed to simultaneously detect various isoprostanes and prostaglandins with a low limit of quantification. Analytical method validation allowed application of these biomarkers to a proof of concept study in primary rat and human hepatocytes. Results indicate a significant time and dose dependent cellular response for different isoprostane isomers by treatment with ferric nitrilotriacetic acid, a chemical known to cause oxidative stress. Furthermore, the value of isoprostanes as biomarkers of cellular oxidative stress was shown for DILI model compounds. The anticancer agent flutamide is known to cause hepatotoxicity, most likely by formation of reactive metabolites and impairment of mitochondrial function. Formation of imino-quinone intermediates may initiate redox cycling and cause excessive generation of ROS. In order to attenuate drug-induced ROS, hepatocyte cell culture was supplemented with pro-oxidant substrates for the in situ generation of hydrogen peroxide. Treatment of rat and human hepatocytes with flutamide induced oxidative stress as indicated by a time and dose dependent increase of isoprostane concentration. Other lipid peroxidation products, namely the hydroxynonenal (HNE) derived conjugates, HNE mercapturic acid (MA) and its reduced form dihydroxynonene MA, were found to be augmented upon treatment with flutamide as well. These were included into the biomarker panel. Under the test conditions no cytotoxicity was present, emphasizing the potential of lipid peroxidation products to early detect upcoming liver damage in in vitro systems. The described biomarkers could be translated between species from rat to human in hepatocytes. Further, results in Fischer F344 rats revealed their applicability to in vivo and enabled their classification relative to other cellular oxidative stress markers. In rats, the antioxidant response pathway was investigated via quantitative determination of mRNA for cytoprotective enzymes. In rat hepatocytes and rat liver increased RNA expression levels for glutathione-S-transferase, heme oxygenase, and NADPH:quinone oxidoreductase were detected. This suggests adaptation of cell homeostasis upon oxidative stress induced damage prior to overt cellular or organ damage. It can be assumed that pro-oxidant processes result in pathophysiological changes contributing to manifestation of DILI. Thus, the characterization of bioactivation potentials and oxidative stress conditions as contributing factor to DILI may be appropriate to characterize DILI risk. The development of new analytical tools using state of the art mass spectrometry enabled quantitative biomarker analysis and glutathione adduct screening from the same sample.
In conclusion, this work describes the advances and limitations of RM characterization as risk for DILI. It highlights the value of characterizing danger signals, e.g. induced by oxidative stress. Specifically, biomarkers derived from lipid peroxidation and cell signal analysis may support preclinical risk assessment. It further stresses the importance of integrated risk mitigation strategies that are able to capture a variety of relevant drug properties and the mechanism by which they modulate toxicity. It must be also taken into account that patient related risk factors are likely to play a major role in development of DILI. Therefore, it is necessary to judge elucidated pathways on their potential to cause inter-individual differences. To minimize the general risk of adverse effects including DILI, the predominant goal in drug discovery must be the optimization of pharmacokinetic drug properties to yield low dose and selective drugs.