A graphical systems model and tissue-specific functional gene sets to aid transcriptomic analysis of chemical impacts on the female teleost reproductive axis

Abstract

Oligonucleotide microarrays and other ‘omics’ approaches are powerful tools for unsupervised analysis of chemical impacts on biological systems. However, the lack of well annotated biological pathways for many aquatic organisms, including fish, and the limited power of microarray-based analyses to detect low level differential expression of individual genes can hinder the ability to infer and understand chemical effects based on transcriptomic data. Here we report on the supervised assembly of a series of tissue-specific functional gene sets intended to aid transcriptomic analysis of chemical impacts on the female teleost reproductive axis. Gene sets were defined based on an updated graphical systems model of the teleost brain–pituitary–gonadal-hepatic axis. Features depicted in the model were organized into gene sets and mapped to specific probes on three zebrafish (Danio rerio) and two fathead minnow (Pimephales promelas) microarray platforms. Coverage of target genes on the microarrays ranged from 48% for the fathead minnow arrays to 88% for the most current zebrafish platform. Additionally, extended fathead minnow gene sets, incorporating first degree neighbors identified from a Spearman correlation network derived from a large compendium of fathead minnow microarray data, were constructed. Overall, only 14% of the 78 genes queried were connected in the network. Among those, over half had less than five neighbors, while two genes, cyclin b1 and zona pellucida glycoprotein 3, had over 100 first degree neighbors, and were neighbors to one another. Gene set enrichment analyses were conducted using microarray data from a zebrafish hypoxia experiment and fathead minnow time-course experiments conducted with three different endocrine-active chemicals. Results of these analyses demonstrate the utility of the approach for supporting biological inference from ecotoxicogenomic data and comparisons across multiple toxicogenomic experiments. The graphical model, gene mapping, and gene sets described are now available to the scientific community as tools to support ecotoxicogenomic research.