Laboratory variability does not preclude identification of biological functions impacted by hydroxyurea

• Published in: Environmental and molecular mutagenesis 2005-12, Vol.46 (4), p.221-235
• Main Authors: Müller, Arne, Boitier, Eric, Hu, Ting, Carr, Gregory J., Le Fèvre, Anne-Céline, Marchandeau, Jean-Pierre, Flor, Manoli, Jefferson, Felicia, Aardema, Marilyn J., Thybaud, Véronique
• Format: Article
• Language: English
• Subjects: Animals; ANOVA; Antineoplastic Agents - toxicity; Biological and medical sciences; Cell Cycle Proteins - drug effects; Cell Cycle Proteins - genetics; Cell Cycle Proteins - metabolism; Cell Line, Tumor; Clinical Laboratory Techniques - standards; Clinical Laboratory Techniques - statistics & numerical data; dose-response; Dose-Response Relationship, Drug; Fundamental and applied biological sciences. Psychology; gene expression; Gene Expression Profiling; Gene Expression Regulation, Neoplastic - drug effects; Gene Ontology; Genetics of eukaryotes. Biological and molecular evolution; Hydroxyurea - toxicity; Leukemia L5178; Medical sciences; Mice; microarray; Models, Biological; Mutagenicity Tests; regression; Reproducibility of Results; Signal Transduction - drug effects; study comparison; toxicogenomics; Toxicology
• ISSN: 0893-6692; 1098-2280
• DOI: 10.1002/em.20164

The multi‐lab International Life Sciences Institute (ILSI) project on the application of genomics to risk assessment offered the unique opportunity to investigate the influence of variability within and between laboratories on identifying biologically relevant gene expression changes. We assessed the gene expression profiles of mouse lymphoma L5178Y cells treated with hydroxyurea (HU) in three independent studies from two different laboratories, sanofi aventis and Procter & Gamble. Cells were dosed for 4 hr and harvested immediately at the end of the treatment or after a 20‐hr recovery period. Cytotoxicity and genotoxicity were evaluated by standard assays. Statistical analysis of these data revealed that, while gene expression responses to HU treatment were markedly different at 4 hr vs. 24 hr, there was otherwise a consistent pattern of dose–response across the three studies. Therefore, the studies were merged and each time point was evaluated separately. At 4 hr, we identified 173 (P < 0.0001) dose‐responsive genes with a common trend in all three studies. These were mainly associated with the cell cycle, DNA repair and DNA metabolism, and in particular, the intra‐S and G2/M phase checkpoints. At 24 hr, we identified 434 dose‐responsive genes common across studies. These genes were involved in lymphocyte‐specific activities and the activation of apoptosis via the caspase cascade. Our results show that despite inter‐laboratory variability, combining the three studies in a single statistical analysis identifies more significantly‐modulated genes than in any of the individual studies, due to improved statistical sensitivity. The genes identified in our study provide information that is relevant to HU biology. Environ. Mol. Mutagen., 2005. © 2005 Wiley‐Liss, Inc.