Interpreting in vitro micronucleus positive results: Simple biomarker matrix discriminates clastogens, aneugens, and misleading positive agents

The specificity of in vitro mammalian cell genotoxicity assays is low, as they yield a high incidence of positive results that are not observed in animal genotoxicity and carcinogenicity tests, that is, “misleading” or “irrelevant” positives. We set out to develop a rapid and effective follow‐up testing strategy that would predict whether apparent in vitro micronucleus‐inducing effects are due to a clastogenic, aneugenic, or secondary irrelevant mode(s) of action. Priority was given to biomarkers that could be multiplexed onto flow cytometric acquisition of micronucleus frequencies, or that could be accomplished in parallel using a homogeneous‐type assay. A training set of 30 chemicals comprised of clastogens, aneugens, and misleading positive chemicals was studied. These experiments were conducted with human TK6 cells over a range of closely spaced concentrations in a continuous exposure design. In addition to micronucleus frequency, the following endpoints were investigated, most often at time of harvest: cleaved Parp‐positive chromatin, cleaved caspase 3‐positive chromatin, ethidium monoazide bromide‐positive chromatin, polyploid nuclei, phospho‐histone H3‐positive (metaphase) cells, tetramethylrhodamine ethyl ester‐negative cells, cellular ATP levels, cell cycle perturbation, and shift in γ‐H2AX fluorescence relative to solvent control. Logistic regression was used to identify endpoints that effectively predict chemicals' a priori classification. Cross validation using a leave‐one‐out approach indicated that a promising base model includes γ‐H2AX shift and change in phospho‐histone H3‐positive events (25/30 correct calls). Improvements were realized when one or two additional endpoints were included (26—30/30 correct calls). These models were further evaluated with a test set of 10 chemicals, and also by evaluating 3 chemicals at a collaborating laboratory. The resulting data support the hypothesis that a matrix of high throughput‐compatible biomarkers can effectively delineate two important modes of genotoxic action as well as identify cytotoxicity that can lead to irrelevant positive results. Environ. Mol. Mutagen. 55:542–555, 2014. © 2014 Wiley Periodicals, Inc.