You are here:
Computational Toxicology and Computational Modeling of Embryonic Limb Development using ToxCast High-throughput Screening Data (Teratology Society Annual Meeting)
Citation:
Ahir, B., R. Spencer, N. Baker, R. Judson, M. Martin, AND T. Knudsen. Computational Toxicology and Computational Modeling of Embryonic Limb Development using ToxCast High-throughput Screening Data (Teratology Society Annual Meeting). Presented at Teratology Society Annual Meeting, Denver, CO, June 24 - 28, 2017.
Impact/Purpose:
Poster presentation at the Teratology Society Annual meeting
Description:
Skeletal defects are one of the major adverse outcomes observed across many prenatal developmental toxicity studies. We mined the ToxCast database (ToxCastDB) for in vitro chemical-bioactivity profiles that significantly correlated with skeletal defects in the Toxicity Reference Database (ToxRefDB). Using high-throughput screening (HTS) data from ˜1060 chemicals tested in 1192 assay-endpoints, 734 chemicals had in vivo developmental toxicity data in ToxRefDB and 44 (4-5%) produced fetal limb defects. In addition, we identified 112 ToxCast chemicals not in ToxRefDB that according to published articles produced limb defects. Significant univariate associations (e.g., assay-endpoint) were used to filter HTS assays based on statistical correlation (p<0.05) with distinct in vivo developmental limb defects. Retinoic acid receptors (RARs) assays emerged in the top target assay list and were activated by 23 ToxCast chemicals. Retinoic acid (RA) signaling plays an essential role in limb patterning as a ligand for nuclear RA receptors. We incorporated RA signaling into a multicellular agent-based model (ABM) of early limb development in CompuCell3D. The ABM stimulates complex cellular interactions (adhesion, apoptosis, chemotaxis, migration, mitosis, secretion) through formation of apical epidermal ridge, zone of polarizing activity, and expansion of mesenchyme driven by morphogenetic signals (BMPs, FGFs, SHH, RA). To evaluate the model we selected two known prototypes: Dieldrin (AC50 = 1.679 µM on RARγ, and 0.770 µM on RARα) and Aldrin (AC50 = 0.911 µM on RARγ, and inactive on RARα) based on ToxCast data. Simulating the impact of RAR disruption via down-regulating GREM1 expression by these compounds on other pathways (subsequently impact on other signaling e.g. SHH, FGFs) in the limb model can provide insight into how the spatial-temporal dynamics of altered limb development as a tool for predictive toxicology. [This abstract does not necessarily reflect EPA policy.]