Development of an in vitro liver toxicity prediction model based on longer term primary rat hepatocyte culture.

The assumption that compounds with similar toxic endpoints generate unique gene expression signatures has led to attempts to classify unknown compounds according to their genomic profile. However, studies reported so far have mostly been conducted in vivo. In this proof of concept study, we assessed the use of rat hepatocyte sandwich cultures in combination with a toxicogenomic classification method to generate a predictive in vitro toxicity classification model. After pre-incubation for 3 days, primary rat hepatocytes were treated for up to 9 days with 13 well known model compounds, changes in the global gene expression profile were measured and subsequently used for the establishment of a predictive classification model. A subset of 724 genes was capable of discriminating compounds with a misclassification rate of 7.5%. As a preliminary verification, the resulting classifier was applied to two blinded control compounds. The classification of compounds according to transient changes in global gene expression allowed the correct prediction independently from the knowledge of their underlying toxic mechanisms. The results of this first pilot study demonstrate the possibility of in vitro gene expression models to contribute to candidate selection early in drug discovery by improving the predictivity of toxicological studies and thereby reducing animal usage in toxicology.