Modeling kinase networks involved in peroxisome biogenesis via single-cell colocalization and morphology metrics
Therese M. Pacio, Fred D. Mast, and John D. Aitchison
- Abstract
- Peroxisomes play an integral role in human health. Dysfunction in peroxisome biogenesis can compromise overall peroxisome and cellular function and can lead to severe neuropathologies and metabolic disorders. While a spatiotemporal model of peroxisome biogenesis has been characterized in yeast cells, the extent of the conservation of this model in humans is unknown. Here we perform Kinase Regression Analysis (KiR) using single-cell features extracted from microscopy datasets to deconvolve kinase networks involved in peroxisome biogenesis. Huh7 cells with a KiR panel were fixed and stained for Pex3, Pmp70 and Sec61 to capture Pex3 colocalization at both early and late stages of peroxisome biogenesis. Primary pediatric hepatocyte cells donated by two patients were also stained for Pex3, Pmp70, and actin. CLARITY, a python-based image analysis pipeline, was used for single-cell feature extraction from the image datasets. Images were stacked into multi-channel 3D arrays and restored via deconvolution. Images were then cropped to center the in-focus z-plane and an average array of the 3 image channels was generated in preparation for image segmentation. The Allen Institute’s deep learning based cell segmenter and structure segmenter were employed to binarize and identify single cells and peroxisomes in the images. Finally, colocalization and peroxisome morphology features were extracted from the images and inputted into the KiR algorithm to predict kinases involved in peroxisome biogenesis. Morphometric analysis of peroxisomes show high variation in peroxisome volume, intensity and number per cell, demonstrating multiple states of peroxisome biogenesis present in a single treatment. Population distributions of Pearson’s Correlation Coefficients demonstrate varying distributions of colocalization between Pex3 and Pmp70 as well as Pex3 and Sec61 in response to different kinase inhibitors. Multiple wells showed multimodal correlation distributions, demonstrating the presence of multiple states of colocalization within a single treatment. Colocalization metrics were binned and inputted into the KiR algorithm. For each bin, 35-45 kinases were identified and hypothesized to alter Pex3 colocalization. The pairing of single-cell image data extraction with KiR analysis demonstrates a novel method of modeling kinase networks. Future work includes performing validation experiments to verify the predictions of kinases involved in peroxisome biogenesis.
- Presented by
- Therese Pacio
- Institution
- Center for Global Infectious Disease Research
- Keywords
- CGIDR, 1:00-1:30











