Predicting Modifiers of Genotype-Phenotype Correlations in Craniofacial Development

Abstract

Most human birth defects are phenotypically variable even when they share a common genetic basis. Our understanding of the mechanisms of this variation is limited, but they are thought to be due to complex gene-environment interactions. Loss of the transcription factor Gata3 associates with the highly variable human birth defects HDR syndrome and microsomia, and can lead to disruption of the neural crest-derived facial skeleton. We have demonstrated that zebrafish gata3 mutants model the variability seen in humans, with genetic background and candidate pathways modifying the resulting phenotype. In this study, we sought to use an unbiased bioinformatic approach to identify environmental modifiers of gata3 mutant craniofacial phenotypes. The LINCs L1000 dataset identifies chemicals that generate differential gene expression that either positively or negatively correlates with an input gene list. These chemicals are predicted to worsen or lessen the mutant phenotype, respectively. We performed RNA-seq on neural crest cells isolated from zebrafish across control, Gata3 loss-of-function, and Gata3 rescue groups. Differential expression analyses revealed 551 potential targets of gata3. We queried the LINCs database with the 100 most upregulated and 100 most downregulated genes. We tested the top eight available chemicals predicted to worsen the mutant phenotype and the top eight predicted to lessen the phenotype. Of these, we found that vinblastine, a microtubule inhibitor, and clofibric acid, a PPAR-alpha agonist, did indeed worsen the gata3 phenotype. The Topoisomerase II and RNA-pol II inhibitors daunorubicin and triptolide, respectively, lessened the phenotype. GO analysis identified Wnt signaling and RNA polymerase function as being enriched in our RNA-seq data, consistent with the mechanism of action of some of the chemicals. Our study illustrates multiple potential pathways for Gata3 function, and demonstrates a systematic, unbiased process to identify modifiers of genotype-phenotype correlations.