Toxicogenomics-based discrimination of toxic mechanism in HepG2 human hepatoma cells

The rapid discovery of sequence information from the Human Genome Project h as exponentially increased the amount of data that can be retrieved from bi omedical experiments. Gene expression profiling, through the use of microar ray technology, is rapidly contributing to an improved understanding of glo bal, coordinated cellular events in a variety of paradigms. In the field of toxicology, the potential application of toxicogenomics to indicate the to xicity of unknown compounds has been suggested but remains largely unsubsta ntiated to date. A major supposition of toxicogenomics is that global chang es in the expression of individual mRNAs (i.e., the transcriptional respons es of cells to toxicants) will be sufficiently distinct, robust, and reprod ucible to allow discrimination of toxicants from different classes. Definit ive demonstration is still lacking for such specific "genetic fingerprints, " as opposed to nonspecific general stress responses that may be indistingu ishable between compounds and therefore not suitable as probes of toxic mec hanisms. The present studies demonstrate a general application of toxicogen omics that distinguishes two mechanistically unrelated classes of toxicants (cytotoxic anti-inflammatory drugs and DNA-damaging agents) based solely u pon a cluster-type analysis of genes differentially induced or repressed in cultured cells during exposure to these compounds. Initial comparisons of the expression patterns for 100 toxic compounds, using all similar to 250 g enes on a DNA microarray(similar to 2.5 million data points), failed to dis criminate between toxicant classes. A major obstacle encountered in these s tudies was the lack of reproducible gene responses, presumably due to biolo gical variability and technological limitations. Thus multiple replicate ob servations for the prototypical DNA damaging agent, cisplatin, and the non- steroidal anti-inflammatory drugs (NSAIDs) diflunisal and flufenamic acid w ere made, and a subset of genes yielding reproducible inductions/repression s was selected for comparison. Many of the "fingerprint genes" identified i n these studies were consistent with previous observations reported in the literature (e.g., the well-characterized induction by cisplatin of p53-regu lated transcripts such as p21(waf1/cip1) and PCNA [proliferating cell nucle ar antigen]). These gene subsets not only discriminated among the three com pounds in the learning set but also showed predictive value for the rest of the database (similar to 100 compounds of various toxic mechanisms). Furth er refinement of the clustering strategy, using a computer-based optimizati on algorithm, yielded even better results and demonstrated that genes that ultimately best discriminated between DNA damage and NSAIDs were involved i n such diverse processes as DNA repair, xenobiotic metabolism, transcriptio nal activation, structural maintenance, cell cycle control, signal transduc tion, and apoptosis. The determination of genes whose responses appropriate ly group and dissociate anti-inflammatory versus DNA-damaging agents provid es an initial paradigm upon which to build for future, higher throughput-ba sed identification of toxic compounds using gene expression patterns alone.