Science Inventory

Evaluating two analysis methods for the microelectrode array network formation assay

Citation:

Mayil Vahanan, M., T. Shafer, K. Friedman, AND K. Carstens. Evaluating two analysis methods for the microelectrode array network formation assay. SOT Conference 2024: A New Approach Method (NAM) to Screen for the Impact of Endogenous Stress on Chemical Toxicity, Salt Lake City, UT, March 10 - 14, 2024. https://doi.org/10.23645/epacomptox.25565175

Impact/Purpose:

Presentation presented to SOT Conference 2024: A New Approach Method (NAM) to Screen for the Impact of Endogenous Stress on Chemical Toxicity

Description:

Background: In vitro new approach methods (NAMs) have been developed for evaluating developmental neurotoxicity (DNT) potential. One such DNT NAM, the Network Formation Assay (NFA), uses a twelve-day exposure model to measure neuronal network activity changes in primary rat cortical cell cultures grown on multi-well microelectrode arrays (MEAs). Network activity parameters are measured on days in vitro (DIV) 5, 7, 9, and 12, and efficacy over time at each chemical concentration tested is expressed as area under the curve (AUC) for concentration-response modeling. The aim of this project was to compare the AUC analysis approach to observed changes on only a single recording day (DIV12) to determine the impact of a data reduction approach on bioactivity results. It was hypothesized that between the two analysis methods, the AUC method would detect more changes in bioactivity than the DIV12 method given that the AUC method expressly uses activity changes from each DIV, whereas the DIV12 method relies on the assumption that the network changes observed are cumulative and lasting. Methods: Concentration-response modeling of a set of 396 chemicals screened in the NFA was performed using the ToxCast Pipeline (version 3.0) to determine changes in activity (positive and negative hit calls) and potency estimates (concentration at 50% maximal activity) for the AUC data and the DIV12 data. Results: Out of 8041 bioactivity curve fits (473 samples x 17 endpoints), 82.3% were concordant in terms of hit call between the two methods. On a chemical level, 159 chemicals were determined to have positive hits in at least three endpoints in both methods while 6 chemicals and 29 chemicals were a hit in only the AUC or DIV12 method respectively. Next the distribution of potency values for active curves between the two pre-processing methods was determined and it was found that the potency values estimated from the DIV12 data were left-shifted compared to the AUC data (Wilcoxon Rank Sum p-value = 4.461e-5). The minimum potency values estimated by each method differed by a mean of 0.60 log10-µM (± standard deviation of 0.75) indicating that the minimum potency values estimated by each method may vary by more than 0.5 log10 fold difference. These results suggest that the DIV12 approach estimated more potent values overall. Conclusions: These findings suggest that evaluating effects on the final DIV (DIV12) in the NFA may capture more sensitive potency values compared to using an AUC approach (DIV5, 7, 9, 12), while also capturing a subset of active chemicals that were not detected by the AUC approach alone. Future work focusing on comparisons between additional DIVs may elucidate network formation changes not captured by the AUC approach and ultimately may improve the sensitivity of the NFA bioactivity results for evaluating DNT. This abstract does not reflect U.S. EPA policy.