Poster R5 - Developing image analysis processes for high throughput drug and gene screening of cell chirality

Introduction: Chirality has recently been determined as an intrinsic property of cells that is heavily implicated in the development of left-right asymmetry organisms. While there is mounting interest in this field, there is still much left to be learned as to the molecular mechanisms which govern this behavior and the tissue-wide consequences of failures in proper formation. We have recently developed a platform that allows for high throughput testing of drugs and genetic markers and how they may alter the chirality of different cell types. Screening of chirality-associated drugs and molecular markers provides us with more information on disease states such as congenital heart defects. Our microgroove screening platform thus requires an analysis method that is both robust and can handle up to thousands of images. Here, we discuss the development of two approaches for chirality image analysis of our platform. One focuses on an automated cell direction estimator based on phase contrast image gradient intensity. The other utilizes a Fast Fourier Transform (FFT) to rapidly detect the global alignment bias of a given image. This allows us to further assess the alignment of several cell types on varying topologies.

Materials and

Methods: Microgrooves were prepared by photolithography then etching of glass to achieve widths and depths from 20 µm to 200 µm and 0.5 µm to 2 µm respectively. Human foreskin fibroblasts (HFF), Human umbilical vein endothelial cells (HUVEC), and AC16 cardiomyocytes cell lines were seeded onto the fibronectin coated grooves and were cultured at 37℃ and 5% CO2 in corresponding cell growth media. After roughly 24 hours of culture, images were taken at 10X magnification in the Keyence BZX microscope. A custom MATLAB code was written for FFT and gradient based analyses. The user is asked to confirm the direction of the grooves in the image before analysis to ensure proper processing for both methods. The phase contrast image subregion gradient is obtained by convolution with gradient filters acting in the x and y directions. The cell alignment direction is taken to be orthogonal to the gradient intensity direction. For the Fourier transform based approach, the built in 2D FFT function in Matlab is used to obtain the power spectrum on which a radial summation is performed. Both analyses are tested individually on all images.

Results, Conclusions, and Discussions: Results and Discussions: Phase contrast images were obtained in bulk from the substrates for large sample size (fig 1A). From these we can apply the two analysis methods. The gradient approach uses the phase contrast pixel intensities to infer the local alignment direction. This can generate information such as the average angle of alignment, trends of alignment, and statistical significance for every image. An FFT based approach shown here (fig. 1B-D) instead generates information on the global alignment of the image. From the raw image, we generate a power spectrum (fig. 1B) and perform a radial summation from angles -90° to 90° (fig. 1C). This allows for the generation of a chiral factor based on the area under the curve about the direction of the groove for all surface dimensions (fig. 1D). Using both methods, we then assessed the alignment of multiple cell lines on differing groove topologies. Results of FFT and gradient based analyses showed agreement in the overall biased alignment of cells on the microgrooves. The FFT based approach only outputs the chiral factor per image but is significantly faster than the gradient based approach.

Figure 1: (a) Phase contrast image of human foreskin fibroblasts (HFFs) on 100µm width grooves 2µm deep. (b) FFT power spectrum of image. (c) FFT Radial summation. (d) Average chiral Factor for all HFF microgrooves.

Conclusions: Overall, we find that these methods of chirality analysis are sufficient for our purposes of high throughput drug screening. Both methodologies generate results that agree with each other as well as previous findings of cell alignment biases. These preliminary studies start to show that different cell types show differences in alignment with changes in the topology. Altogether, we believe that we can reliably move forward with the testing of many drugs and different molecular markers with various cell types and surface topologies. The screening of drugs and molecular markers associated with chirality and disease states may provide potential new routes for therapeutics and tissue engineering applications.

Acknowledgements (Optional):