Interior_Banner_Events

Morphometric Feature Selection for the High-throughput Image-based Chemical Phenotyping of Per- and Polyfluoroalkyl Substances

Nicholas Cemalovic, Anagha Tapaswi, Chanese Forte, Justin Colacino, Department of Environmental Health Sciences, University of Michigan School of Public Health, 1415 Washington Heights, Ann Arbor 48109

High content imaging represents an emerging set of methods to assess total cell response to pharmaceuticals and environmental toxicants in a high-throughput manner. Here, our goal was to apply these methods to derive a chemical’s morphometric phenotypic “fingerprint”, classifying mode of action of per- and polyfluoroalkyl substances (PFAS) in a dose-dependent fashion. MCF10A breast epithelial cells were first exposed to 4 doses of histone deacetylase inhibitors of known molecular structure and function and analyzed via CellPainting, a high-content imaging assay, to optimize and validate our computational methods. We also exposed MCF10A cells to 4 doses, ranging from 25nM to 25µM, of PFDA, PFNA, PFOA, and the novel PFOA derivative “GenX”. Given the prospect of PFAS epigenetically reprogramming mammary development and promoting breast cancer, we analyzed features from the Hoechst 33442 stained nuclear channel. Following microscope image processing using CellProfiler, 240 nuclear features were calculated for over 42,000 cells. We used generalized linear models to rank each feature for its chemical specific significance and dose-dependent directionality, visualized through heatmaps and matrix plots. Given our method’s selectivity in identifying structure-bioactivity relationships, PFOA and its branched ether derivative GenX had low feature correlation (r = 0.43). This indicates contrasting bioactivities with novel and distinct molecular targets of action for GenX, a poorly characterized toxicant recently introduced into the consumer and ecological environment. Further, this method effectively differentiated branched and linear PFAS, clustering the 8 and 9 carbon PFOA and PFNA with high feature rank correlation (r = 0.70). Heatmaps and unbiased clustering of feature rank and directionality demonstrated distinct morphometric fingerprints for each toxicant, identifying feature with distinguishable structure-activity relationships, modes of action, and dose-dependent behaviors. We identified cellular morphological targets of PFAS previously unexplored with traditional single-cell and high-throughput techniques, potentially identifying new hallmarks of adverse cell response to these environmental perturbations.




Additional Abstract Information

Presenter: Nicholas Cemalovic

Institution: University of Michigan - Ann Arbor

Type: Poster

Subject: Nursing & Public Health

Status: Approved


Time and Location

Session: Poster 9
Date/Time: Wed 12:00pm-1:00pm
Session Number: 6038