Modeling and Analysis of the Spatio-Temporal Dynamics of the Mitochondrial Network

PROJECT SUMMARY/ABSTRACT Mitochondria provide 90% of our energy; defects in mitochondria lead to a wide range of diseases including seizures, stroke, heart disease, neurodegeneration, and cancer. Far from their static kidney-bean shaped depiction in many textbooks, mitochondria form a dynamic three-dimensional network that spans the entire volume of the cell. This network undergoes continuous remodeling through fission and fusion, motility, biogenesis and clearance. Under stress or disease conditions, the mitochondrial network fragments and changes its dynamic equilibrium. Understanding this equilibrium, and its changes and adjustments to disease, is an archetypical question in quantitative cellular organelle biology. The dynamic mitochondrial network has so far evaded experimental interrogation and modeling as mitochondria were too small and too fast for volumetric fluorescence microscopy. Fortunately, recent advances in imaging technology, namely lattice light-sheet microscopy (LLSM), have changed that. Substantial preliminary data in this application supports the working hypothesis that a combination of quantitative LLSM image processing, and particle based spatial modeling can succeed in creating the first four-dimensional (4D) spatiotemporal model of the mitochondrial network. The goal of the proposed work is to elucidate the fundamental biophysical principles of mitochondrial network homeostasis. We have outlined three aims that will enable us to close this knowledge gap. Aim 1 will test the hypothesis that deep learning-based mitochondria segmentation will demonstrate more accurate extraction of the 4D mitochondrial network from LLSM data as compared to traditional methods. New deep neural network architectures will be developed to test this hypothesis. It is expected that a tool will be delivered that generalizes across diverse imaging conditions and diverse mitochondrial form and function impaired conditions. Aim 2 will test the hypothesis that graph-based topological linking will demonstrate the first temporal tracking of the 4D mitochondrial network. New linear assignment problem-based algorithms will be developed to precisely track the mitochondrial network backbone as well as its fission/fusion events. It is expected that a tool will be delivered that can track the mitochondrial network in a variety of imaging conditions and mitochondrial form and function impaired conditions. Aim 3 will test the hypothesis that morphology, dynamics, and function of the mitochondrial network are linked and can be predicted. A new particle-based polymer simulation model will be developed based on 4D graph temporal analysis of experimental data. It is expected that the first 4D spatio-temporal model of the mitochondrial network will be developed that can predict form and function observables and their time evolution from first principles.