Characterization and Computational Evaluation of a Genetically Encoded Fluorescent Probe for Reactive Oxygen Species

Abstract

Reactive Oxygen Species (ROS) are highly reactive chemical molecules that have essential roles in physiological processes. H2O2 is a relatively stable member of ROS family in aqueous environment and serves as an important signaling molecule in various mechanisms, such as the opioid overdose and myocardial ischemia. Among the various analytical tools, genetically encoded fluorescent indicator (GEFI) is a profound technology that demonstrates high specificity and spatiotemporal resolution in capturing the physiological production of H2O2 . We have developed HRM63, a H2O2 GEFI that demonstrates 2-times higher fluorescent magnitude and 10-times faster responding dynamic compared to HyPerRed, one of the most widely used H2O2 GEFI. For the scope of this thesis, my investigation includes the HRM63 sensor characterization, including its subcellular specificity, reversibility and photoactivation, and the overall computational pipeline of analyzing the fluorescent readout of HRM63. We develop a single-cell tracking algorithm Cell of Interest Nearest Neighbor (COINN) based on the result from Cellpose, a deep learning-based segmentation algorithm, to visualize the cell-to-cell difference in the fluorescent output of sensor-transfected cells. We then design a ratiometric sensor HRM63-mCherry to correct the effect of protein expression level. Combining the use of COINN and HRM63-mCherry, we present a new representation of fluorescent reading indicating the oxidation level that also accounts for the difference among individual cells. We also use the machine learning-based classification to predict whether the individual cells will respond to the agonist administration. Ultimately, these findings will allow more effective investigation on the pipeline of characterizing the H2O2 GEFIs using the fluorescent imaging.