A simple transcriptomic signature able to predict drug-induced hepatic steatosis.

It is estimated that only a few marketed drugs are able to directly induce liver steatosis. However, many other drugs may exacerbate or precipitate fatty liver in the presence of other risk factors or in patients prone to non-alcoholic fatty liver disease. On the other hand, current in vitro tests for drug-induced steatosis in preclinical research are scarce and not very sensitive or reproducible. In the present study, we have investigated the effect of well-characterized steatotic drugs on the expression profile of 47 transcription factors (TFs) in human hepatoma HepG2 cells and found that these drugs are able to up- and down-regulate a substantial number of these factors. Multivariate data analysis revealed a common TF signature for steatotic drugs, which consistently and significantly repressed FOXA1, HEX and SREBP1C in cultured cells. This signature was also observed in the livers of rats and in cultured human hepatocytes. Therefore, we selected these three TFs as predictive biomarkers for iatrogenic steatosis. With these biomarkers, a logistic regression analysis yielded a predictive model, which was able to correctly classify 92 % of drugs. The developed algorithm also predicted that ibuprofen, nifedipine and irinotecan are potential steatotic drugs, whereas troglitazone is not. In summary, this is a sensitive, specific and simple RT-PCR test that can be easily implemented in preclinical drug development to predict drug-induced steatosis. Our results also indicate that steatotic drugs affect expression of both common and specific subsets of TF and lipid metabolism genes, thus generating complex transcriptomic responses that cause or contribute to steatosis in hepatocytes.