LAUSR

Global gene expression profiling is useful for elucidating a drug's mechanism of action on the liver; however, such profiling in rats is not very sensitive for predicting human drug-induced liver injury, while dedifferentiated monolayers of primary human hepatocytes (PHHs) do not permit chronic drug treatment. In contrast, micropatterned cocultures (MPCCs) containing PHH colonies and 3T3-J2 fibroblasts maintain a stable liver phenotype for 4-6 weeks. Here, we used MPCCs to test the hypothesis that global gene expression patterns in stable PHHs can be used to distinguish clinical hepatotoxic drugs from their non-liver-toxic analogs and understand the mechanism of action prior to the onset of overt hepatotoxicity. We found that MPCCs treated with the clinical hepatotoxic/non-liver-toxic pair, troglitazone/rosiglitazone, at each drug's reported and non-toxic Cmax (maximum concentration in human plasma) for 1, 7, and 14 days displayed a total of 12, 269, and 628 differentially expressed genes, respectively, relative to the vehicle-treated control. Troglitazone modulated >75% of transcripts across pathways such as fatty acid and drug metabolism, oxidative stress, inflammatory response, and complement/coagulation cascades. Escalating rosiglitazone's dose to that of troglitazone's Cmax increased modulated transcripts relative to the lower dose; however, over half the identified transcripts were still exclusively modulated by troglitazone. Last, other hepatotoxins (nefazodone, ibufenac, and tolcapone) also induced a greater number of differentially expressed genes in MPCCs than their non-liver-toxic analogs (buspirone, ibuprofen, and entacapone) following 7 days of treatment. In conclusion, MPCCs allow evaluation of time- and dose-dependent gene expression patterns in PHHs treated chronically with analog drugs.