A data mining approach for the elucidation of the action of putative etiological agents: application to the non-genotoxic carcinogenicity of genistein

Identification of potential biomarkers of genotoxicity and carcinogenicity in L5178Y mouse lymphoma cells by cDNA microarray analysis

In the present study, cDNA microarray analyses were performed with mouse cDNA chips in order to evaluate similarities and differences in the gene expression profiles for compounds differing in their genotoxic and carcinogenic potential. Eight test substances were evaluated, two each from four classes of compounds: genotoxic carcinogens (1,2-dibromoethane and glycidol), genotoxic noncarcinogens (8-hydroxyquinoline and emodin), nongenotoxic carcinogens (methyl carbamate and o-nitrotoluene), and nongenotoxic noncarcinogens (D-mannitol and 1,2-dichlorobenzene). Quadruplicate hybridization experiments were performed in order to identify a set of genes with significant expression changes for these four classes of substances. Twelve genes were consistently altered more than twofold by the genotoxic noncarcinogens while four genes were consistently regulated by the nongenotoxic carcinogens. One gene (Trp63) was identified whose expression was upregulated by all four genotoxic substances regardless of the presence or absence of carcinogenicity; this finding, however, was not confirmed by quantitative real-time RT-PCR. RT-PCR did confirm the change in expression of 9 of 15 genes (60%) identified by microarray analysis. Interestingly, the downregulated genes were least likely to be validated by real-time RT-PCR. Those genes showing more than a twofold change in expression level in response to at least one substance were further analyzed with hierarchical clustering after category assignment of each gene according to its main cellular function. Clustering revealed differences in the gene expression profiles between the genotoxic and nongenotoxic substances for genes involved in cell cycle control, the stress response, and the immune response. However, no clustering specific to all four carcinogenic substances was observed in any of the functional categories. Taken together, these results suggest that gene expression profiling in mouse lymphoma cells can provide valuable information for the evaluation of potential genotoxicity but may have limitations in predicting carcinogenicity.