Gene expression profiling in primary rat hepatocytes for the prediction of hepatotoxicity

New human pharmaceuticals are required by law to be tested in pre-clinical studies in order to predict any potential drug side effects. However, during the last decade the number of new drug approvals has markedly decreased while the cost of drug development has increased. The reasons for this are twofold: firstly, due to adverse effects in humans which are not predicted by animal studies, leading to the compound’s failure in late phases of the development process, and secondly because some drug candidates never reach clinical trials due to intolerable toxic effects in animals. Thus, a need exists for new in vitro assays to be developed which enable the detection of a compound’s toxicity prior to animal studies. A better pre-selection of drug candidates could increase the success rate during preclinical trials since very toxic compounds could be excluded from animal tests. Furthermore in vitro tests also provide mechanistic information which supports the interpretation of in vivo observations, as well as play a role in human safety assessment, since these tests could be performed in both animal and human cells. The present study describes one part of the research activities associated with the project Predict-IV financed by the European Commission. Predict-IV’s aim is the development of a non-animal based prediction system for the toxicity of new drugs related to the organs kidney and liver as well as the central nervous system. In the liver work package primary human and rat hepatocytes as well as the human hepatoma cell line HepaRG were treated for 14 days with toxic and non-toxic doses of eleven pharmaceutical reference substances of known in vivo toxicity. After one, three and 14 days, respectively, samples for proteomic, metabolomic, genomic and kinetic analyses were collected. The genomic endpoint was investigated by performing a whole-genome gene expression analysis with Illumina BeadChips. In the present study the global gene expression profiles of primary rat hepatocytes were interpreted biologically after treatment with seven of the eleven reference compounds to investigate the potential of drug-treated rat primary hepatocytes to reflect the in vivo effects noted in the literature. In this study the pharmaceutical mode of action was distinguished from off-target mechanisms which were discussed in relation to hepatotoxic effects in vivo. In accordance with its pharmaceutical mode of action, the PPARα-agonist Fenofibrate increased the expression of genes involved in lipid metabolism. In addition, genes were also induced which indicated the formation of oxidative stress and the depletion of glutathione which was considered to be a basic mechanism of Fenofibrate’s hepatotoxicity. Similar results were found for EMD335823, supporting the assumption that this withdrawn drug candidate is also a PPARα-agonist. However, the small number of genes deregulated by Valproic acid and Acetaminophen was insufficient to reflect any in vivo effects probably related to cellular treatment with too small doses. The PPARγ-agonists Troglitazone and Rosiglitazone induced genes which coded for drug-metabolizing enzymes known to oxidize these substances, especially Troglitazone, into reactive potentially cytotoxic metabolites. Additional genes involved in the metabolism of glutathione and the response to oxidative stress, a major toxic mechanism of Troglitazone and Rosiglitazone, were upregulated. Since both compounds pharmacologically act on muscle cells their mode of action could not be reconstructed. Metformin which acts on the liver without causing severe adverse effects was used as negative control. It deregulated a large number of genes but its gene expression profile clearly differed from that of the hepatotoxic compounds. In the last chapter of this study the genes commonly deregulated by Fenofibrate, EMD335823, Troglitazone, Rosiglitazone and Metformin were discussed. The major part of these genes was involved in lipid metabolism which seemed to be related to the mode of action of the tested compounds since PPARα- and PPARγ-agonists regulate lipid and glucose homeostasis. In conclusion the whole-genome gene expression profile of drug treated primary rat hepatocytes reflected cellular mechanisms which could explain hepatotoxic effects in vivo. During the next phases of Predict-IV the gene expression profiles of rat and human primary hepatocytes and HepaRG cells treated with all of the eleven reference compounds will be compared. The gene expression profiles of the four reference compounds not discussed in this study will be compared with their protein expression profiles. Additionally, the real cellular concentration of the test compounds and the kinetic of their metabolism will be calculated. Furthermore, species-specific effects as well as the responsiveness of the cell line compared to primary cells will be investigated in order to define the cell system best suited for an early predictive screening system. Finally, genomic and proteomic markers should be defined and validated which could enable the early prediction of new drugs’ hepatotoxic potential in future