A rapid and fully automated in vitro micronucleus assay using imaging flow cytometry and convolutional neural network analysis

Abstract Micronuclei (MN) originate from whole chromosomes or chromosome fragments that lag behind during cell division and fail to be incorporated into one of the two main nuclei. As a result, scoring MN using the well-established in vitro micronucleus assay evaluates the ability of chemicals or other agents to induce DNA damage. This technique is typically performed by manual microscopy, which can be time-consuming and prone to variability. Additionally, automated methods lack cytoplasmic visualization when using slide-scanning microscopy, and conventional flow cytometry doesn’t provide visual confirmation of MN. The ImageStream®X Mk II (ISX) imaging flow cytometer combines the high-resolution imagery of microscopy with conventional flow cytometry’s speed and statistical robustness in a single system. Previously, we developed a rapid and automated MN assay based on high-throughput image capture and feature-based image analysis using IDEAS® Software. However, the feature-based analysis was not readily applicable to multiple cell lines and chemicals, so we developed a deep learning method based on convolutional neural networks to score imaging flow cytometry data in both the cytokinesis-blocked and unblocked versions of the MN assay using Amnis® AI Software. Our current study validates our previously established assay and analyses using three different chemicals (Mitomycin C, Cyclophosphamide, and Eugenol) and three different cell lines (TK6, L5178Y, CHO-K1). Here, we demonstrate how using Amnis AI to score imagery acquired on the ISX provides a rapid and fully automated in vitro MN assay with improved accuracy, reproducibility, and time-to-results in toxicity and biodosimetry applications across multiple cell lines.