Assessing the data quality in predictive toxicology using a panel of cell lines and cytotoxicity assays

In vitro cell viability assays have a central role in predictive toxicology, both in assessing acute toxicity of chemicals and as a source of experimental data for in silico methods. However, the quality of in vitro toxicity databanks fluctuates dramatically because information they contain is obtained under varying conditions and in different laboratories. The aim of this study was to identify the factors responsible for these deviations and thus the quality of the data extracted for predictive toxicology. Three cell viability assays measuring LDH leakage, WST-1 reduction, and intracellular ATP were compared in an automated environment using four mammalian cell lines: Caco-2, Calu-3, Huh-7, and BHK. Using four standard compounds--polymyxin B, gramicidin, 5-fluorouracil, and camptothecin--a significant lack of sensitivity in LDH assay compared with the other assays was observed. Because the viability IC(50) values for the standards were similar among the cell lines, the biochemical characteristics of different cell lines seem to play only a minor role, with an exception being the hepatocellular Huh-7 cell line. Toxicity assessment of new 1,2,4-triazoles revealed significant differences in their toxic potential, and the results indicate the same sensitivity profile among the assays as observed with the standard compounds. Overall, it can be argued that the assay selection is the most important factor governing the uniform quality of the data obtained from in vitro cell viability assays.