Evaluation of in vitro models for detection of hepatocyte toxicity induced by anti-cancer drugs

Abstract

Drug-induced hepatotoxicity is a major cause of failure of drug candidates during development and of drug withdrawal post-marketing. Exposure of individuals to many different classes of drugs, including several anti-cancer agents, can result in liver toxicity. In oncology, the aim of treatment with cytotoxic drugs is to selectively target and kill all cancerous cells, whilst leaving the healthy tissue unharmed. In practice, anti-cancer drugs induce cytotoxic effects in all proliferating cells and, to a certain extent, non-dividing cells. For some anti-cancer drugs this manifests as liverrelated dose-limiting toxicity. If this occurs in early phase clinical trials, it may stop clinical development of that drug. Unfortunately, current pre-clinical models poorly predict a potentially hepatotoxic anti-cancer agent.Therefore, the aim of the studies included in this thesis is to evaluate in vitro models in which anti-cancer drug-induced hepatotoxicity may be detected, to decrease the likelihood of hepatotoxic anticancer agents entering clinical development. In vitro models that represent hepatocytes, the major cell type in the liver, were selected, and were all of human origin. Murine models were not considered as there are significant inter-species differences that may impact on drug metabolism and therefore on drug-induced liver toxicity. The selected models comprised fresh and cryopreserved hepatocytes (from 24 and 3 individuals respectively), an immortalised hepatocyte cell line (THLE-2) and hepatoma cell lines (Flep 3B2.1-7, Hep G2, Huh7D12, PLC/PRF/5, and SK-HEP-1). These cells were characterised, and then assessed in terms of their sensitivity to a known hepatotoxic drug (mithramycin), drugs used in clinical practice in colorectal cancer (oxaliplatin, 5-FU) and novel agents in preclinical development (ruthenium compounds).In order to characterise the cell models, morphology, growth parameters, and viability were considered. Fresh and cryopreserved hepatocytes did not proliferate in culture, whereas the cell lines had doubling times ranging from 33 to 54 hours. Hepatocyte viability varied between donors at isolation and after transport. The impact of transporting fresh hepatocytes in different cold-storage media, and the duration between isolation and plating, on hepatocyte viability was investigated.Gene expression was also studied by quantitative reverse-transcription polymerase chain reaction (qRT-PCR). The genes evaluated were involved in drug transport (ABCB1), drug metabolism (CYP450s, UGTs), or were liver specific genes (albumin, transthyretin, a-1-antitrypsin). mRNA expression of all liver-specific genes except CYP1A1 was higher in fresh hepatocytes than in any of the other models investigated. None of the in vitro models mimic the gene expression profile of hepatocytes but the hepatoma cell lines express some liver-specific genes. Stressinduced genes (p21, heat-shock protein 70 and hypoxia-inducible factor-la) were evaluated to identify the impact of hepatocyte plating on stress gene expression. mRNA expression of all stress-induced genes increased following plating in the majority of donors. The impact on gene expression of the duration of transport, the transport media, and plating, was assessed to improve the use of fresh human cells in hepatotoxicity assessment.To identify the most appropriate in vitro model on the basis of the identification of hepatocyte-toxic drugs, cytotoxicity assays were carried out following exposure of each of the cell models to increasing concentrations of the selected drugs. DNA damaging agents induce cytotoxic effects; therefore cell death was considered as the endpoint for evaluating drug-induced hepatotoxicity. The sulforhodamine B and MTT cytotoxicity assays were carried out, followed by determination of IC50 values from the data obtained. Instead of aiming for an absence of toxicity, it is more interesting to determine the difference between an active concentration killing tumour cells and achievable in the clinic, and a toxic concentration, killing hepatocytes. This "in vitro therapeutic index" may be able to discriminate which drug is likely to induce significant liver toxicity, at predicted therapeutic doses. Colon cancer lines were used to determine the active concentrations for mithramycin, oxaliplatin and 5-FU. The effect of these drugs was then determined in the hepatocyte models. The "in vitro therapeutic index" was able to discriminate between known hepatotoxin mithramycin. and oxaliplatin and 5-FU which both had a higher therapeutic index in fresh hepatocytes. Flepatoma cell lines gave lower therapeutic indices but the ranking of the compounds was similar as with the fresh hepatocytes. Finally, using this approach, the hepatotoxicity of two ruthenium compounds in pre-clinical development (RM175 & HC11) was identified, using fresh hepatocytes and hepatoma cell lines.A subset of in vitro models (fresh and cryopreserved hepatocytes, Hep3B2.1-7, HepG2, Huh-7D12 and PLC/PRF/5), which appeared to be the most promising based on gene expression and cytotoxicity data, were then selected to study the contribution of oncosis and/or apoptosis to cytotoxicity, and to see whether a hepatotoxic drug could be detected earlier during drug exposure. Assays to measure ATP levels and capsase-3/7 activation in drug-treated cells were optimised and used for mithramycin-, oxaliplatin- and 5-FU-treated cells. ATP levels and caspase-3/7 activation varied between cell lines and hepatocytes in response to all three drugs investigated, and was detectable immediately following drug treatment.These data suggest that fresh hepatocytes are a suitable in vitro model in which to accurately identify hepatotoxicity associated with anti-cancer drugs. They express liver-specific genes, and are more sensitive to the hepatotoxic drug mithramycin, than to oxaliplatin and 5-FU. A cytotoxicity assay using hepatoma cell lines (Hep3B2.1-7 and Huh-7D12) could also be used, despite their low expression of liver-related genes. These results should help select the most appropriate in vitro models to detect hepatotoxic anticancer agents in the future.