Gene expression profiles of hepatotoxin-treated human hepatocytes can be used to cluster unknown compounds according to their mode of action.

Abstract

We treated cryopreserved human hepatocytes from three donors with a range of model hepatotoxic and non-hepatotoxic compounds and monitored their gene expression profiles using Affymetrix GeneChip technology. In order to create a “training list” with which to analyse unknown drug classes, a range of well studied compounds were chosen for their ability to elicit different pathological responses in the liver.Plated hepatocytes were treated with a single dose of acetaminophen (necrosis), chlorpromazine (cholestasis), diclofenac (hepatitis), gemfibrozil (positive control), isoniazid (hepatitis), nitrosodimethylamine (genotoxic carcinogen), phenobarbital (non-genotoxic carcinogen) and tetracycline (steatosis). Total RNA was isolated after 1, 4 and 24 h and used to synthesize biotin-labelled cRNA. Labelled cRNA from the three donors was pooled and hybridized to an Affymetrix HG U133A chip. Two replicates for each compound at each time point were performed and the expression values averaged. Changes in gene expression values compared to vehicle treated hepatocytes were used to identify genes capable of distinguishing the compounds.A second set of chemicals belonging to different drug classes (thiazolidinediones, anti-estrogens and HDAC-inhibitors) was tested in the same manner and the “training list” used to cluster all samples. Principal component analysis of the 24 h-treated samples showed distinct groupings according to chemical class.