DEVELOPMENT OF ANALYTIC TOOLS FOR HIGH-THROUGHPUT LABEL-FREE MICROSCOPY OF CELL MOTILITY AND CELL SPREADING

Abstract

High-throughput microscopy is an approach that emerged a decade ago. It is an efficient tool to solve numerous tasks in cell biology and drug discovery in combination with automated microscopy. Automated microscopy is suited to perform direct observation of living cells and allows to obtain consistent and straightforward results from dynamic processes. The aim of various images processing tools is to study cell migration. However, there is a deficiency in the tools to perform accurate and efficient image processing of unlabeled cells, compared to fluorescently labeled ones. Labelfree microscopic techniques like Phase contrast and DIC result in low contrast images, therefore the tools designed for fluorescent image segmentation are inefficient. To overcome this limitation, many methods to automatically analyze images of wound healing were developed within the last decade, but these tools require manual tuning of parameters and lack of automation to process stacks of thousands of images. This work focuses on building an efficient image processing pipeline for brightfield image segmentation. The solution proposed herein is a filtering sequence to modify the bright-field image intensity histogram so that it resembles that of fluorescent images. In addition, a conditional operator was introduced, which is a logic loop to test the quality of wound gap segmentation in unsupervised mode. This allowed achieving >95% accuracy of segmentation in the un-supervised mode. The processing pipeline for cell spreading is further enhanced by the transformation of image coordinates to reduce the dimensionality of the image, and to simplify the image edge to a single direction gradient. The tools developed in this Ph.D. project were applied to evaluate the effect of inhibiting microtubule dynamics on cell motility and spreading. It was confirmed that wound healing closure occurs in a non-linear manner for the majority of cell lines studied. A piecewise regression analysis was performed to specify the period when the wound closure occurs at constant velocity. Further, it was found that the observations made in the first 12 hours after scratching the cell layer were optimal for obtaining precise measurements of the wound edge for both normal and cancer cells. A reduction in cell motility in response to microtubule inhibitors' action occurred in a dose-dependent manner at concentrations below the apparent cytotoxic doses. It was demonstrated that the speed of wound closure observed in 24 hours did not depend on a change in cell proliferation induced by the microtubule drugs. Therefore, the image processing pipeline developed in this Ph.D. research project can significantly reduce time consumption for RAW data processing and at the same time greatly increase the precision in analyzing motility-based assays.