ヒト末梢血単核細胞及びヒト肝がん由来細胞の共培養系を用いた薬剤性肝障害リスク予測

抄録

[Purpose] Prediction of drug-induced liver injury (DILI) potential remains a major detection challenge due to multiple mechanisms. While immune and inflammation reactions are involved in pathogenesis of DILI, in vitro systems using immune cells for predicting DILI have not been studied extensively. We aimed to find potential biomarkers for predicting DILI in in vitro co-culture systems of peripheral blood mononuclear cells (PBMCs) with human liver cell lines.

[Methods] Human PBMCs were co-cultured with HepG2 or HepaRG using Transwell^® inserts and treated with 5 DILI-positive drugs (amodiaquine, ketoconazole, tienilic acid, diclofenac, and trovafloxacin) or 5 DILI-negative drugs (chloroquine, mebendazole, ethacrinic acid, ketorolac, and levofloxacin). Total RNA extracted from PBMCs was subjected to microarray analysis to explore biomarkers that can distinguish between DILI-positive and DILI-negative drugs. Successfully validated genes were evaluated for their predictability of DILI potential of 21 drugs by ROC curves.

[Results and Discussion] Out of 14 genes validated by qPCR in HepG2/PBMC, 5 genes (PID1, MET, PTGS2, EREG, and IL24) showed relatively high AUC values. Compared with AUC values of each individual gene, higher AUC values (0.882 and 0.855) were obtained by multiple logistic regression analysis and total score method, respectively, of the 5 genes, with the former having the highest sensitivity (80.0%) and specificity (90.9%). The established PBMC co-culture system using the 5 biomarkers would be of utility to detect DILI potential in nonclinical drug development.