Classifying Chemical Bioactivity by Coupling High-throughput Phenotypic Anchoring and Transcriptome Profiling in Zebrafish

Graduation date: 2017There are more than 87,000 chemicals in current use with little to no toxicity information available. Assessing such a large number of chemicals using traditional methods would take an unreasonable amount of time and money, and require the use a large number of animals. The incorporation of high-throughput in vivo model systems that can overcome limitations of in vitro screens while providing a rapid genome-wide view of chemical bioactivity are needed to help fully realize the vision of 21st century toxicity testing. In this dissertation, I employed transcriptomics using the high-throughput in vivo developmental zebrafish model to generate hypotheses regarding the mechanism of action of specific chemicals and to identify unique transcriptional signatures for a set of endocrine disrupting chemicals that can be used to classify the activity of known and unknown chemicals. We employed comparative transcriptomics in wild-type and ahr2-null zebrafish exposed to mono-substituted isopropylated triaryl phosphate (mITP), a component of the Firemaster 550 flame retardant mixture, and demonstrated that the cardiotoxic effects of mITP are likely mediated through inhibition of retinoic acid receptor signaling and not the aryl hydrocarbon receptor (AhR). Although mITP induced a transcriptional profile indicative of AhR activation in wild-type zebrafish, these signatures were not related to cardiotoxicity. I found that mITP exposure, independent of ahr2 status, decreased the expression of many Hox genes as well as enzymes responsible for retinoic acid metabolism, which are known to be regulated by retinoic acid receptors. Several of the dysregulated Hox genes are involved in cardiac cell lineage determination in early heart development, and in heart tube elongation and looping via cell recruitment in the second heart field. I also employed transcriptomics to investigate the mechanism of developmental toxicity of the antimicrobial agent, triclosan (TCS). Exposure to TCS resulted in robust transcriptome changes with a large number of transcripts being significantly decreased. Downstream functional analyses showed that many of the transcripts significantly affected by TCS are involved in liver functioning and development, suggesting that TCS is hepatotoxic in embryonic zebrafish. I also compared our transcriptomic analysis with the comprehensive in vitro bioactivity profile of TCS from the Environmental Protection Agency’s Toxicity Forecaster (ToxCast) program, and observed low concordance between these two screening strategies. Lastly, I conducted transcriptome profiling of 25 endocrine disrupting chemicals in order to identify discriminatory transcriptional signatures that would help classify chemicals with estrogen, androgen, or thyroid hormone activity. Clustered correlation analysis of the top 1000 significantly differentially expressed transcripts revealed four chemicals with highly similar and unique transcriptional profiles compared to the other 21 chemicals. These four chemicals included three known thyroid hormone receptor agonists, and one unknown chemical, which was later identified as a failed pharmaceutical with thyroid receptor agonist activity. I identified a panel of 27 transcripts as well as a unique pigmentation phenotype for these four chemicals which can be used in future chemical screens to detect other thyroid receptor agonist chemicals in zebrafish. Overall, the work presented here demonstrates the utility of phenotypically anchored whole genome transcriptomics in zebrafish to identify putative mechanisms of action for many chemicals. Furthermore, the information obtained from these types of analyses can be used to develop predictive models to classify the bioactivity of known and unknown chemicals, or can identify biomarkers that can be used in future high-throughput screens