Metabolomics and Cytokine Analysis for Identification of Severe Drug-Induced Liver Injury

Aim: To evaluate the levels of metabolites and cytokines in the serum of patients with severe and non-severe idiosyncratic drug-induced liver injury (DILI) and to identify biomarkers of DILI severity. Methods: Gas chromatography–mass spectrometry (GC–MS) and ultraperformance liquid chromatography–mass spectrometry (UPLC–MS) based metabolomic approaches were used to evaluate the metabolome of serum samples from 29 DILI patients of severity grade 3 (non-severe), 27 of severity grade 4 (severe), and 36 healthy control (HC). The levels of total keratin-18 (K18), fragment K18, and 27 cytokines were determined by enzyme-linked immunosorbent assay. Results: The alkaline phosphatase activity (p = 0.021) and international normalized ratio (INR) (p < 0.001) differed significantly between the severe and non-severe groups. The severe group had a higher serum fragment K18 level than the non-severe group. A multivariate analysis showed good separation between all pairs of the HC, non-severe, and severe groups. According to the orthogonal partial least-squares–discriminant analysis (OPLS-DA) model, 14 metabolites were selected by GC–MS and 17 by UPLC–MS. Among these metabolites, the levels of 16 were increased and of 15 were decreased in the severe group. A pathway analysis revealed major changes in the primary bile acid biosynthesis and alpha-linolenic acid metabolic pathways. The levels of PDGF-bb, IP-10, IL-1Rα, MIP-1β, and TNF-α differed significantly between the severe and non-severe groups, and the levels of most of the metabolites were negatively correlated with those of these cytokines. An OPLS-DA model that included the detected metabolites and cytokines revealed clear separation of the severe and non-severe groups. Conclusion: We identified 31 metabolites and 5 cytokines related to the severity of idiosyncratic DILI. The primary bile acid biosynthesis and alpha-linolenic acid metabolism pathways were also related to the severity of DILI. A model that incorporated the metabolites and cytokines showed clear separation between patients with severe and non-severe DILI, suggesting that these biomarkers have potential as indicators of DILI severity.