Phenotypically Anchored mRNA and miRNA Expression Profiling in Zebrafish Reveals Flame Retardant Chemical Toxicity Networks

Abstract

The ubiquitous use of flame retardant chemicals (FRCs) in the manufacture of many consumer products leads to inevitable environmental releases and human exposures. Studying toxic effects of FRCs as a group is challenging since they widely differ in physicochemical properties. We previously used zebrafish as a model to screen 61 representative FRCs and showed that many induced behavioral and teratogenic effects, with aryl phosphates identified as the most active. In this study, we selected 10 FRCs belonging to diverse physicochemical classes and zebrafish toxicity profiles to identify the gene expression responses following exposures. For each FRC, we executed paired mRNA-micro-RNA (miR) sequencing, which enabled us to study mRNA expression patterns and investigate the role of miRs as posttranscriptional regulators of gene expression. We found widespread disruption of mRNA and miR expression across several FRCs. Neurodevelopment was a key disrupted biological process across multiple FRCs and was corroborated by behavioral deficits. Several mRNAs (e.g., osbpl2a) and miRs (e.g., mir-125b-5p), showed differential expression common to multiple FRCs (10 and 7 respectively). These common miRs were also predicted to regulate a network of differentially expressed genes with diverse functions, including apoptosis, neurodevelopment, lipid regulation and inflammation. Commonly disrupted transcription factors (TFs) such as retinoic acid receptor, retinoid X receptor, and vitamin D regulator were predicted to regulate a wide network of differentially expressed mRNAs across a majority of the FRCs. Many of the differential mRNA-TF and mRNA-miR pairs were predicted to play important roles in development as well as cancer signaling. Specific comparisons between TBBPA and its derivative TBBPA-DBPE showed contrasting gene expression patterns that corroborated with their phenotypic profiles. The newer generation FRCs such as IPP and TCEP produced distinct gene expression changes compared to the legacy FRC BDE-47. Our study is the first to establish a mRNA-miR-TF regulatory network across a large group of structurally diverse FRCs and diverse phenotypic responses. The purpose was to discover common and unique biological targets that will help us understand mechanisms of action for these important chemicals and establish this approach as an important tool for better understanding toxic effects of environmental contaminants.

Gene co-expression network analysis in zebrafish reveals chemical class specific modules

“…Clusters 1 (orange) and 2 (red) represented all the flame retardant chemicals (FRCs) examined and were also more similar to each other than to other clusters in the dataset, showing that the overall transcriptomic response to FRCs is distinct from PAH and TCDD responses. In addition to differences in transcriptomic response based on chemical type, we also found that when we compared our clustering to our lab’s previous studies investigating the morphology and behavior effects caused by FRC exposure [ 6 , 41 ], there were differences between transcriptomic response and developmental toxicity phenotype. For example, IPP exposure caused both morphology and behavior malformations at 120 hpf and yet, it clustered here with the relatively benign FRCs.…”

“…This was done using a large array of transcriptomic data combined with the first chemical-focused gene co-expression network inferred using zebrafish. Transcriptomic data from 48-h post fertilization (hpf) zebrafish included several previously published studies [ 7 , 39 – 41 ] in addition to new unpublished datasets. Details on these datasets including the exposure concentrations and the number of differentially expressed genes (DEGs) are included in Supplementary Table S 1 .…”

“…In total, we collected RNA sequencing data from 48-hpf zebrafish exposed to 33 unique chemical treatments (Supplementary Table S 1 ). For 29 of these treatments (18 PAHs, TCDD, and 10 FRCs), the data was collected from previously published previous studies [ 7 , 39 – 41 ], with the remaining 4 treatments (4 PAHs) initially analyzed in this study. Each treatment was examined with 3–8 replicates for a total of 170 RNA sequencing samples included in the study.…”

“…Detailed methods for published datasets can be found in our previous studies [ 7 , 39 – 41 ]. Methods for chemicals initially analyzed in this study, and TCDD exposures for RT-qPCR analysis (See below) are described here.…”

“…Detailed exposure methodology and phenotypic endpoints identified for each chemical for the previously published datasets can be found in their respective studies [ 7 , 39 – 41 ]. Briefly, for the PAHs (with the exception of 7,12-B [a] AQ and BEZO) and FRCs, an EC80 exposure concentration was selected based on a concentration-response zebrafish phenotypic screen [ 6 , 38 ].…”