Citation:

Nyffeler, J., C. Willis, G. Patlewicz, I. Shah, AND J. Harrill. High-throughput phenotypic profiling to inform putative mode of action for environmental chemicals (MCBIOS 2020). Annual Conference Of The Midsouth Computational Biology & Bioinformatics Society (MCBIOS), Cary, NC, April 06 - 08, 2020.

Impact/Purpose:

Abstract submitted to the 17th Annual Conference Of The Midsouth Computational Biology & Bioinformatics Society (MCBIOS) scheduled for April 2020 (conference cancelled). The Environmental Protection Agency is exploring the use of profiling methods for rapid bioactivity screening and hazard evaluation of environmental chemicals. Phenotypic profiling is an imaging-based method that measures morphological features of organelles at the single cell level. Here, we adapted an existing phenotypic profiling assay for use in high-throughput bioactivity screening of environmental chemicals with a focus on two applications: (1) Estimation of potency thresholds for chemical bioactivity; (2) Use of phenotypic profiles to discern putative mechanisms-of-action.

Description:

Method: For the use of phenotypic profiles, we screened a set of 120 reference chemicals and toxicological model compounds as well as 462 environmental chemicals at 8 concentrations. U2OS cells were exposed for 24 h, fixed and labeled with fluorescent dyes to visualize multiple organelles: nucleus, nucleoli, endoplasmic reticulum, golgi, actin cytoskeleton, plasma membrane and mitochondria. Confocal images were acquired, and 1300 features extracted per cell. Cell-level data was normalized to vehicle controls and aggregated to the well level. Well-level data was averaged across replicates and summarized across concentrations by retaining the largest effect size to derive a profile for each chemical. Biological similarity was computed by comparing profiles using Pearson correlation. Structural similarity was measured by comparing Morgan fingerprints using Tanimoto similarity. Results: Overall, several distinct profile clusters were observed for different classes of pesticides (organochlorine, strobins, dinitroaniline) Reference chemicals with various DNA damaging mechanisms (alkylators, topoisomerase inhibitors, antimetabolites) clustered together along with a subset of environmental chemicals. Another example included three microtubule stabilizers: paclitaxel and docetaxel share high structural similarity (0.68), while epothilone B is structurally unrelated to the former two (0.11). However, the three chemicals were biologically highly similar (>0.8). Conclusion: To summarize, phenotypic profiling identified groups of chemicals with similar cellular effects, both among structurally diverse and structurally related chemicals. This abstract does not reflect US EPA policy.

Record Details: