Principal component analysis Treated Vehicle Treated Vehicle

Heat Map & Cluster Analysis Ethinyl Estradiol 500mg/kg Fenbufen 250mg/kg Ibuprofen 500mg/kg Fenbufen 250mg/kg Ibuprofen 500mg/kg Fenbufen 250mg/kg Diflunisal 500mg/kg Benzbromarone 200mg/kg Diethy-hexyl-phthalate 1000mg/kg Diflunisal 750mg/kg Diethy-hexyl-phthalate 1000mg/kg Benzbromarone 200mg/kg Diflunisal 500mg/kg Benzafibrate 500mg/kg Ibuprofen 500mg/kg Diethy-hexyl-phthalate 1000mg/kg Clofibrate 600mg/kg WY mg/kg Perfluoro-n-heptanoic Acid 150mg/kg Perfluoro-n-octanoic Acid 150mg/kg Perfluoro-n-decanoic Acid 50mg/kg Perfluoro-n-heptanoic Acid 150mg/kg WY mg/kg Clofibrate 600mg/kg WY mg/kg Diiso-nonyl-phthalate 1000mg/kg Perfluoro-n-decanoic Acid 50mg/kg NM_ NM_ NM_ NM_ NM_ NM_ NM_ NM_ NM_ NM_ NM_ NM_ NM_ M11794 BE NM_030850

Applications of genomics in toxicology Mechanistic Toxicology Investigative toxicology –Hypothesis generation Risk assessment –Understanding the mechanism of toxicity at the molecular level Predictive toxicology Compound avoidance –Elimination of liabilities Compound selection –Select compound with least toxic liability from a series Compound management –Tailor conventional studies and perform timely investigational toxicology studies

Where Predictive & Mechanistic Toxicology Fit Drug Discovery PreClinical Testing Clinical Development Phase IV FDA Mechanism-based Mechanistic studies Pattern-based Predictive screens

Mechanistic Toxicology Using Genomics/Transcriptomics

Morphologic Analysis Correlates with Gene Expression Changes in Cultured F344 Rat Mesothelial Cells L. M. Crosby,* K. S. Hyder,† A. B. DeAngelo,‡ T. B. Kepler,§ B. Gaskill, G. R. Benavides, L. Yoon, and K. T. Morgan (Toxicol Appl Pharmacol Dec 15;169(3): ) The gene expression pattern of mesothelial cells in vitro was determined after 4 or 12 h exposure to the rat mesothelial, kidney, and thyroid carcinogen and oxidative stressor potassium bromate (KBrO3). Gene expression changes observed using cDNA arrays indicated oxidative stress, mitotic arrest, and apoptosis in treated immortalized rat peritoneal mesothelial cells.

Morphologic Analysis Correlates with Gene Expression Changes in Cultured F344 Rat Mesothelial Cells L. M. Crosby,* K. S. Hyder,† A. B. DeAngelo,‡ T. B. Kepler,§ B. Gaskill, G. R. Benavides, L. Yoon, and K. T. Morgan (Toxicol Appl Pharmacol Dec 15;169(3): )

Increases occurred in oxidative stress responsive genes; transcriptional regulators; protein repair components; DNA repair components; lipid peroxide excision enzyme PLA2; and apoptogenic components. Numerous signal transduction, cell membrane transport, membrane- associated receptor, and fatty acid biosynthesis and repair components were altered Propose a model for KBrO3-induced carcinogenicity in the F344 rat mesothelium is proposed, whereby KBrO3 generates a redox signal that activates p53 and results in transcriptional activation of oxidative stress and repair genes, dysregulation of growth control, and imperfect DNA repair leading to carcinogenesis.

Predictive Toxicology Prediction = Probability Best estimate from available information Does not provide definitive result or answer Provides alerts and/or guidance

Predictive Toxicology in Compound Management In Drug Development Selection/deselection of compounds Initiate a proactive investigative toxicology programme to be forewarned is to be forearmed Risk assessment Conventional toxicology studies test the hypotheses generated by predictive toxicology (hazard + dose response + risk = assessment) Decision making using both sets of data

Pattern-based Predictive Screens Using Genomics/Transcriptomics

Genomic Profiling - comparing toxins From Ulrich & Friend (2002) Nature Reviews, 1:84-88

Toxicogenomics-based Discrimination of Toxic Mechanism in HepG2 Human Hepatoma Cells ME Burczynski, M McMillian, J Ciervo, L Li, JB Parker, RT Dunn, S Hicken, S Farr & MD Johnson Toxicological Sciences 58, Initial comparisons of the expression patterns for 100 toxic compounds using a 250 gene microarray failed to discriminate between toxicant classes However, taking multiple replicate observations of gene expression for cisplatin, diflunisal & flufenamic acid yielded a reproducible discriminatory subsets of genes. The subsets not only discriminated between the three compounds but also showed predictive value for the other 100 toxic compounds tested. “Supervised learning” Based on statistics and understanding of mechanism

Application of genomics/transcriptomics in toxicology - What has been learned? Hypotheses can be generated Mechanisms can be unravelled Profiles can discriminate between compounds Understanding molecular mechanisms helps Profiles can classify compounds/mechanisms Not a standalone technology to identify toxicity (never an expectation)

Application of genomics/transcriptomics in toxicology - Current understanding Rapid hypothesis generation Rapid classification Additive not standalone Particularly for mechanistic investigations Questions of sensitivity/reproducibility Most gene expression changes at high doses Interlab variation Developing more realistic expectations through collaboration and open debate ILSI, MGED/EBI database standard

A Few References Review of Arrays and Data analysis Lockhart & Winzeler (2000) genomics, gene expression and DNA arrays. Nature 405: Hypothesis generation Crosby et al (2000) Morphologic analysis correlates with gene expression changes in cultured F344 rat mesothelial cells. Toxicol. & Applied Pharmacol. 169: Screening Burczynski et al (2000) Toxicogenomics-based discrimination of toxic mechanism in HepG2 human hepatoma cells. Toxicological Sciences 58: Waring et al (2001) Clustering of hepatotoxins based on mechanism of toxicity using gene expression profiles. Toxicology and Applied Pharmacology 175,