Idiosyncratic, drug-induced liver injury (IDILI) causes morbidity and mortality in affected patients and necessitates the withdrawal of otherwise efficacious drugs from the market. There is a need for more effective preclinical screening of drug candidates to avoid developing and marketing drugs with IDILI liability. Although mechanisms underlying IDILI reactions are not fully understood, available evidence suggests that adaptive or innate immune responses are involved. A role for cytokines such as tumor necrosis factor-alpha (TNF) has been proposed. Published studies as well as our preliminary results suggest that drugs that cause IDILI in humans sensitize hepatocytes to killing by this cytokine in vitro. Accordingly, the overall hypothesis to be tested is that the abilities of drugs to sensitize hepatocytes to injurious effects of TNF can classify them according to their potential to cause IDILI. In preliminary studies, we exposed HepG2 cells to noncytotoxic concentrations of 24 drugs having either high or no/low IDILI liability in humans. The cells were coexposed to TNF at a concentration that was by itself noncytotoxic. Almost all of the drugs with IDILI liability became cytotoxic in the presence of TNF. In contrast, drugs without IDILI liability showed no such response. Detailed drug concentration-response curves were generated in the absence and presence of TNF, and curve characteristics (EC50s, maximal responses, etc.) were used singly or in combination as covariates to develop statistical classification models. Several models showed superb ability to classify the 24 drugs according to their IDILI liabilities, some yielding areas under receiver operating characteristic (ROC) curves of > 0.95, indicating remarkably high sensitivity and specificity. This approach has numerous qualities that make it attractive for screening new drug candidates; however, it requires validation and refinement before it would be acceptable as an assay for IDILI prediction in a preclinical, drug development setting. Accordingly, the aim of this proposal is strictly focused to validate and refine promising statistical classification models for development as assays that predict IDILI potential of drug candidates. We will begin by adapting this in vitro approach to a high throughput format. Next, using a larger set of drugs, detailed concentration-response curves obtained in the absence and presence of TNF will provide covariates for evaluating (ie, validating) the most promising of the statistical classification models that resulted from our preliminary study. Finally, the results will be used to refine the models and to select one or more with extraordinarily high performance. Results from the proposed studies have the potential to lead to a robust yet simple, high-throughput, in vitro assay capable of markedly improving preclinical strategies for identifying drug candidates with IDILI potential.

Public Health Relevance

Some pharmaceuticals cause rare but severe liver damage in patients. Methods that are relatively easy to use and effective in detecting which drugs and drug candidates have the potential to cause these types of adverse responses would be immensely beneficial in developing safer drugs and thereby preventing such reactions. The overall goal of the proposed studies is to evaluate the utility of a cell culture-based approach to identify drugs with the propensity to cause these adverse responses.

National Institute of Health (NIH)
National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)
Research Project (R01)
Project #
Application #
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Serrano, Jose
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Michigan State University
Schools of Medicine
East Lansing
United States
Zip Code