Metabolomic profiling for drug-induced liver injury with autoantibodies

Drug induced liver injury (DILI) is sometimes similar to autoimmune hepatitis (AIH) in serology and histology. Clinicians empirically screened DILI with significant autoimmune characteristics to implement clinical intervention. We tried to characterize DILI with autoantibodies by metabolomics.

Untargeted metabolomics coupled with pattern recognition approaches were performed on sera samples including AIH (n = 59), DILI with autoantibodies (DILI^Ab+, n = 68), and DILI without autoantibodies (DILI^Ab-, n = 75). The differential metabolites and fingerprint metabolites between AIH and DILI^Ab- were screened by orthogonal partial least squares-discriminant analysis and hierarchical clustering respectively.

Of the 388 annotated differential metabolites between AIH and DILI^Ab-, 74 fingerprint metabolites were screened. The eigenmetabolite compressed from the fingerprint possessed high discrimination efficacy (AUC:0.891; 95 %CI, 0.838-0.944). In the fingerprint-based PCA model, AIH and DILI^Ab- were separated into three regions: the "pure region" of AIH (Region 1), the "pure region" of DILI^Ab- (Region 3), the mixture region of AIH and DILI^Ab- (Region 2). After incorporated into the PCA model, DILI^Ab+ samples were distributed into the three regions, indicating that DILI^Ab+ samples had different etiological tendencies. Moreover, the fingerprint-based radar model verified the results of PCA model characterizing DILI^Ab+. Notably, the antibody titers of DILI^Ab+ in the three regions did not differ significantly, while the response rates for glucocorticoids were obviously different. The metabolic difference among DILI^Ab+ in different regions mainly lies in energy metabolism.

In terms of metabolic signature, DILI^Ab+ may not be a community of same pathogenesis, including AIH-inclined parts. Which deserves further study.