In vitro assessment of hepatotoxicity by metabolomics: a review

Omics technologies, and in particular metabolomics, have received an increasing attention during the assessment of hepatotoxicity in vitro. However, at present, a consensus on good metabolomics practices has yet to be reached. Therefore, in this review, a range of experimental approaches, applied methodologies, and data processing workflows are compared and critically evaluated. Experimental designs among the studies are similar, reporting the use of primary hepatocytes or hepatic cell lines as the most frequently used cell sources. Experiments are usually conducted in short time-frames (< 48 h) at sub-toxic dosages. Applied sample preparations are protein precipitation or Bligh-and-Dyer extraction. Most analytical platforms rely on chromatographic separations with mass spectrometric detection using high-resolution instruments. Untargeted metabolomics was typically used to allow the simultaneous detection of several classes of the metabolome, including endogenous metabolites that are not initially linked to toxicity. This non-biased detection platform is a valuable tool for generating hypothesis-based mechanistic research. The most frequently reported metabolites that are altered under toxicological impulses are alanine, lactate, and proline, which are often correlated. Other unspecific biomarkers of hepatotoxicity in vitro are the down-regulation of choline, glutathione, and 3-phospho-glycerate. Disruptions on the Krebs cycle are associated with increased glutamate, tryptophan, and valine. Phospholipid alterations are described in steatosis, lipo-apoptosis, and oxidative stress. Although there is a growing trend towards quality control, data analysis procedures do often not follow good contemporary metabolomics practices, which include feature filtering, false-discovery rate correction, and reporting the confidence of metabolite annotation. The currently annotated biomarkers can be used to identify hepatotoxicity in general and provide, to a certain extent, a tool for mechanistic distinction.