Dose Response Modeling of In Vitro High Content Screening Identifies Genetic Variants Modulating Sensitivity to Monomethylarsonous Acid Exposure
Authors: Callan O'Connor 1,2; Gregory Keele 1; Whitney Martin 1; Daniel Gatti 1; Ron Korstanje 1; Gary Churchill 1; Laura Reinholdt 1
Affiliations: 1) The Jackson Laboratory, Bar Harbor, ME; 2) Tufts University, Graduate School of Biomedical Sciences, Boston, MA
Keywords: Complex traits
The mechanisms underlying variation in susceptibility to environmental exposures are generally not well understood which limits our ability to accurately predict risk. Our goal is to establish a model for unbiased gene-by-environment exposure studies that can be used to both identify and validate genetic modifiers of response to toxicant exposure. We exposed 215 primary fibroblast cell lines derived from a genetically diverse mouse population to varying concentrations of monomethylarsonous acid (MMA). We assayed the cells using image-based high content cellular screening (HCS) to capture cellular morphometric features with the goal of identifying heritable, dose-response features for genetic mapping. We found that several experimental factors exacerbate non-linear modeling for these types of data including cellular heterogeneity, imaging artifacts, and cell culture conditions. To mitigate these problems, we introduced changes to our experimental design and used a two-stage estimation approach to capture MMA-exposure responses. First, we fit the cellular features data to a four-parameter log-logistic dose-response model (DRM) for replicates from the same individual (n = 4). Second, we extracted DRM parameters, which we then modeled using a linear mixed effects model to correct for technical sources of variation. This resulted in 1431 corrected model fit parameter summaries representing slope, asymptotes, and critical effect size estimates (i.e. EC 50) from >300 cellular features. We mapped 10 suggestive quantitative trait loci (QTL) (genome-wide p < 0.05). The slope parameter of the fibroblast cell area response to MMA had the strongest QTL (LOD score > 9) at chromosome 10 (82.9 Mbp) demonstrating that 1) changes in cell size occur in response to MMA exposure, and 2) that genetic variation has an impact on the rate of this change during exposure. Within the QTL region, we identified Txnrd1 as a candidate gene which is highly expressed in fibroblasts and has known interactions with arsenic. We identified genetic variants (SNPs) in the 3’-UTR of Txnrd1, notably with alleles specific to the NZO/H1LtJ (NZO) founder strain, associated with higher slopes for cell area. This region of Txnrd1 regulates the incorporation of the amino acid selenocysteine (Sec) during TXNRD1 synthesis which is critical for the reduction potential of this enzyme. Our findings suggest that the NZO haplotype in the 3’-UTR of Txnrd1 increases susceptibility of fibroblasts to MMA exposure. To test this hypothesis, we plan to perform parallel in vitro and in vivo assessments of MMA sensitivity in genetically diverse inbred mouse strains (Collaborative Cross) that possess the NZO allele at this locus. In conclusion, our current data demonstrates the utility of using cellular reference panels derived from genetically diverse populations to identify genetic modifiers of environmental toxicant exposure.