Drug-Induced Liver Injury (DILI) is responsible for the post-marketing withdrawal of many drugs. The limitations of routine safety biomarkers (e.g. standard clinical pathology parameters) in predicting DILI have long been recognised both by the industry and regulators. There are a number of common mechanisms of liver toxicity, and this article reviews the biomarkers in development and their applicability to different mechanisms of toxicity.
DILI has been the most frequent single cause of safety-related drug marketing withdrawals for the past 50 years and continues to be a significant cause of drug failure. Significant effort has been expended to try to reduce the failure rate of drugs due to hepatotoxicity and overtly hepatotoxic agents are typically removed during screening or in pre-clinical development. However, it has proved to be more difficult to eliminate DILI for all drugs as only the most potent hepatotoxins show severe DILI in the 1,000-3,000 subjects typically studied in clinical trials included in a New Drug Application (NDA) or Marketing Authorisation Application (MAA).
The liver is a multifunctional organ and drugs cause liver injury by many different mechanisms so a single biomarker is unlikely to be sufficient to detect all possible routes of liver damage. Traditional clinical chemistry biomarkers used to assess drug induced hepatotoxicity have often been insensitive and not always specific to liver injury. For example, ALT increases with Cytochrome P450 induction and is also present in muscle, meaning that it is difficult to use as a definitive marker of liver damage. ALP is inducible (in dogs) and is also found in a range of tissues, making it a non-specific marker of liver damage and potentially inaccurate for predicting DILI, especially in dog studies.
Due to limitations of traditional biomarkers, pre-clinical testing has not been able to eliminate all human hepatotoxins from progressing into clinical development (Greaves et al., 2004). Without good clinical biomarkers of DILI, clinical development has also had limited success in preventing drugs which cause liver damage from reaching market. Consequently, there is a significant need for improved pre-clinical and clinical biomarkers for DILI and it is likely, due to the multiple possible mechanisms of liver toxicity, that a panel of such biomarkers will be required to ensure that the potential for DILI is identified early in drug development.
In response to the Critical Path Initiative the Predictive Safety Testing Consortium (PSTC) has developed a legal framework to share the cost of qualification and to protect intellectual property associated with biomarker qualification. The PSTC is collaborating closely with the US, European and Japanese regulators and the European IMI SAFE-T Consortium is to identify and qualify safety biomarkers for a range of drug-induced toxicities including: cardiac hypertrophy, kidney, liver, skeletal muscle, testicular toxicity, and vascular injury.
A number of techniques are being used by members of the PSTC and IMI SAFE-T Consortium to measure biomarkers, including RNA expression profiling (e.g. albumin mRNA and miRNA 122), LC-MS (e.g. conjugated and unconjugated bile acids), Immunoassay (e.g. Keratin 18, PON-1) and activity assays (e.g. GLDH). Whilst many biomarkers are still undergoing qualification, progress is being made and some are starting to be used clinically to measure DILI (Harrill et al., 2012). Below I discuss some of the more promising biomarkers of DILI and some of the evidence supporting their use in the assessment of DILI.
Paracetamol (acetaminophen) is a widely used drug, but is also the leading cause of liver failure both in Europe and in the USA. In 2004 Heinloth et al. (2004) used paracetamol as a model hepatotoxicant to show the potential of liver gene expression profiling to provide sensitive markers of hepatotoxicity. Bushel et al. (2007) demonstrated that peripheral blood gene expression signatures correlated well with exposure level and could act as early biomarkers of paracetamol toxicity. Miyamoto et al. (2008) selected a1-microglobulin / bikunin precursor (Ambp) and albumin mRNAs as tentative liver-specific biomarkers and demonstrated their presence in the plasma of rats following initiation of liver damage by administration of hepatotoxicants paracetamol or D-galactosamine HCl (D-gal), but not following bupivacaine HCl–induced skeletal muscle damage.
In a study performed in our labs we took at similar approach to that taken by Bushel et al. (2007) to identify suitable biomarkers of hepatotoxicity. Rats were administered paracetamol orally at dose levels of 100, 600 and 1200 mg / kg / d for 1, 3 or 14 days. In addition to typical toxicology endpoints blood samples were collected for gene expression profiling.
Treatment with intermediate and high doses of paracetamol showed typical signs of hepatotoxicity, both in clinical chemistry and in histopathology from Day 1. Histopathology showed a clear time and dose-related response, with 100 mg / kg / d representing the No Adverse Effect Level (NOAEL). Centrilobular necrosis and inflammation were seen in animals at the 24 h, 3 day and 14 day time points, with the changes progressing and increasing in severity over time at the high dose. After three days of dosing at the highest dose of 1200 mg / kg / d, livers showed centrilobular necrosis with inflammation and centrilobular alteration (cytoplasmic condensing, slight rounding and enlargement of cells around the central vein). By Day 14, livers of all rats showed acute centrilobular necrosis and inflammation with centrilobular alteration and notable cytoplasmic rarefaction. An adaptive response was seen in the liver of animals administered 600 mg / kg / d, indicating some recovery over the 14 day dosing period, which was not seen at the high dose.
Figure 2-A) Control rat liver (x40 magnification) showing normal cellular pathology. B) Liver of rat administered paracetamol at 1200 mg / kg / d for 3 days (x40 magnification) showing centrilobular necrosis with inflammation and centrilobular alteration seen in some animals. C) Liver of rat administered paracetamol at 1200 mg / kg / d for 14 days (x40 magnification) showing centrilobular necrosis and inflammation with centrilobular alteration and notable cytoplasmic rarefaction.
Clinical chemistry changes seen on day one and day three were not seen on day 14, although there was a clear development in histopathological changes in the liver over this time. In addition to the histopathological and clinical chemistry changes the study also showed statistically significant, dose and time dependent changes in gene expression profiles of whole blood from rats dosed at 600 mg / kg / d and 1200 mg / kg / d. At a dose of 600 mg / kg / d subtle changes in gene expression were first observed at 24 h, which were more pronounced at 3 days and seemed to diminish in magnitude by 14 days, mirroring the adaptive response seen histopathologically. A similar pattern was observed at 1200 mg / kg / d although the magnitude of change was greater both in the number of genes reaching significance at all time points and the degree (fold-change) of differential expression. A particularly marked innate immune signalling and T-cell marker response was observed at days three and 14 was indicative of an early inflammatory event, which predicted the immunocyte infiltrate confirmed histopathologically.
The recent discovery that small non-coding, regulatory microRNAs (miRNAs) are stable and can be detected in circulating blood has led to their assessment as possible safety biomarkers. Wang et al. (2009) assessed expression levels of miRNAs in liver and plasma of BALB / C mice following a single intraperitoneal injection of 75,150 or 300 mg / kg paracetamol. Following treatment with paracetamol, levels of liver specific miRNAs, including miR-122 and miR-192, were increased (proportionately to dose and exposure duration) in both liver and in plasma. Furthermore, the expression levels of miR-122 and miR-192 also mirrored ALT levels and histopathological changes, although changes in miRNA levels were seen as early as one hour after exposure indicating that they act as very sensitive and tissue specific biomarkers of DILI.
The work of Bushel, Miyamoto, Wang, Laterza, ourselves and others suggests that both messenger RNA and miRNA expression profiling in the peripheral circulation can provide surrogate biomarkers of liver damage. The added value of using RNA expression is that it has the potential to provide a suite of biomarkers suitable for different mechanisms of DILI.
Levels of lithocholic acid (LCA) and it’s metabolites are sensed by PXR which subsequently regulates the expression of a suite of enzymes and transporter molecules to increase drug metabolism and reduce bile acid synthesis to protect the liver against choleostasis (Staudinger et al., 2001). Unconjugated and conjugated bile acids, including LCA, have been identified by the PSTC and others as potential biomarkers of cholestatic hepatoxicity.
Like ALT, glutamate dehydrogenase (GLDH) can be easily measured in plasma and is more liver specific than ALT or AST. A GSK investigation of plasma GLDH as a potential biomarker of DILI in rats was conducted following treatment with methapyrilene, dexamethasone, cyproterone, isoniazid, lead nitrate, and Wyeth-14643 (O’Brien
et al., 2002). GLDH activity was increased up to 10-fold more, and up to 3 fold more persistently than ALT, and occurred when hepatocellular injury was present, but when plasma ALT was not increased. GLDH activity was not inhibited by either isoniazid or lead nitrate, and GLDH activity was unaffected by induction. GLDH was found to be a sensitive and effective biomarker of acute DILI which performed better in the rat than ALT, AST, SDH or ALP.
Antoine et al. (2010) have used biomarkers Keratin-18 and high mobility group box-1 protein (HMGB1) to further elucidate the mechanisms associated with paracetamol DILI. In mice given a single intraperitoneal injection of 530 mg / kg paracetamol, necrosis was the main mechanism of hepatocyte cell death. However, Keratin-18 cleavage, DNA laddering and pro-caspase 3 processing indicated that whilst necrosis was the dominant mechanism, that apoptosis was responsible for some paracetamol induced hepatocyte cell death as part of an endogenous protective response to paracetamol toxicity. Use of an HMGB1-neutralizing antibody confirmed a significant role of HMGB1 in the induction of inflammation following paracetamol intoxication, which was postulated to be associated with the regenerative response following initial damage. In fasted animals the inhibition of HMGB1 oxidation due to ATP depletion was demonstrated to reduce apoptosis and further promote the inflammatory response.
An initial assessment of Paraoxonase-1 (PON-1) by the IMI SAFE-T Consortium has indicated that due to marked variability and decreased activity with liver dysfunction that it is not a suitable candidate biomarker for DILI (Adler et al. 2010 and SAFE-T presentation March 2012). Following an initial assessment ALT isoenzymes have also been removed from the priority list of candidates for development due to poor performance.
I have, out of necessity, focussed primarily on blood biomarkers in this editorial, but it should be noted that there is also work ongoing to identify suitable biomarkers for DILI in urine using proteomics (Smyth et al. 2009) and metabonomics (Nicholson et al. 2002); the attraction being the non-invasive manner of sampling.
Whilst post-marketing withdrawal of drugs due to drug-induced liver injury continues, much progress has been made since the FDA’s Critical Path Initiative. A number of promising biomarkers have shown significant improvement over traditional clinical chemistry endpoints and are progressing into qualification. These hold out the hope of reducing the number of drugs removed from the market due to safety concerns and of reducing drug attrition rates, particularly when used together in a multiplexed assay.