VPA Read across: Development of predictive biomarkers by using toxicity data of structurally similar compounds

The common model for safety evaluation/human health risk assessment is currently repeated dose toxicity (RDT) testing in rodents. RDT studies require numerous animals and the capacity for this conventional testing is limited. The development of alternative toxicity tests is, thus, of great interest. However, this goal remains challenging as complex in vivo processes like absorption, distribution, metabolism, excretion (ADME) and different mechanisms of toxicity need to be addressed by a network of reliable test systems. In this project we evaluated -omics readouts to identify predictive biomarkers by using a read across approach. Main goal is the prediction of the type of toxicity, and the toxicological potency, the concentrations of a compound at which adverse effects in vivo are observed. Preliminary experiments with data base selected biomarkers in HepG2 cells and primary human hepatocytes demonstrate that, with a few exceptions, for most of the compounds it was possible to predict blood concentrations in humans which are associated with hepatotoxic effects. The current study focusses on the identification of biomarkers of toxicity in rat hepatocytes and on their capacity to predict toxicity for structurally similar compounds. Model compound for the study is valproic acid (VPA), a drug which induces microvesicular steatosis in liver in human patients at therapeutic doses. VPA also induced lipidosis, lipid accumulation or vacuolization of hepatocytes in in vivo RDT studies in rodents. Gene array data from in vitro experiments in rat and human hepatocytes and from rat in vivo studies demonstrate its strong effect on expression changes especially on genes involved in energy and lipid metabolism. Based on the most frequently deregulated genes we selected 11 candidate biomarker genes representing biological functions such as metabolism of xenobiotics, deregulation of energy and lipid metabolism, cellular stress response, transport, cell cycle, structure dynamics and migration and development and differentiation. These candidate biomarker genes were tested in the context of a read across approach. Six branched and unbranched carboxylic acids were identified showing a high structural similarity to the lead compound VPR. RDT studies were available for all six structural analogues e.g. from the RepDose database (http://www.fraunhofer-repdose.de). Three compounds, namely 2-ethyl hexanoic aid, 2-ethylhexanol and dioctyl adipate cause steatotic changes in vivo and are therefore classified as ""active"". Hexanoic acid, propionic acid as well as 2-ethylbutyric acid were found to be negative in vivo up to the highest tested doses and are classified as ""inactive"". In in vitro experiments, five different concentrations and untreated controls were tested for all seven compounds and the expression of the selected biomarkers was analyzed. Their ability to qualitatively/quantitatively predict steatosis and/or liver toxicity is discussed. The comparison of biomarker expression after exposure of target cells to structurally analogues provides insight in how far toxicity data from well characterized chemicals might help to predict toxicity for structurally similar data poor compounds. Simultaneously, comparing the concentration at which biomarker expression is induced with concentrations where histopathological effects are observed in vivo can be used as a tool to predict the concentration at which toxicity occurs in vivo.