Aims
This study aims to investigate the mechanistic link between energetically demanding mechanical processes, such as cell migration, and mitochondrial dynamics, elucidating how mitochondria adapt their morphology and motility to fluctuating energy requirements. A critical challenge in studying these inherently dynamic processes together is the need to capture data across sub-second to hour timescales. Therefore, a central aim of this work is to establish new capability for generating optimal datasets that encompass high spatiotemporal resolution imaging over extended durations. This includes developing photogentle imaging routines to preserve native mitochondrial dynamics. Furthermore, we aim to push the boundaries of these imaging routines through the development of novel deep-learning pipelines capable of segmenting noisy datasets, which are a by-product of photogentle imaging.
Methods
To enable high spatiotemporal imaging of mitochondria over several hours, we employed widefield microscopy equipped with optimized hardware and optical elements. For zebrafish, we used an Airy beam light-sheet microscope to achieve organelle-level resolution (500nm) within 0.5 mm tissue samples. A novel dual-output deep-learning approach was developed to segment and track individual mitochondrial lineages from the resulting image data. This segmentation and tracking enabled the unbiased and robust quantification of mitochondrial dynamics, including fission-fusion events.
Results
We demonstrate fast, long-term, and photogentle imaging capabilities in single cells, achieving sub-second time resolution and single-mitochondrion spatial resolution over several hours. Critically, we establish an optimal imaging regime balancing image quality and phototoxicity. We identify a lower threshold of image quality that still yields quantifiable data. While photogentle regimes are constrained by limited light dosages required to preserve native mitochondrial dynamics, we demonstrate that quantifiable data can still be obtained within these constraints. Furthermore, we present a comprehensive mitochondrial analysis pipeline that extends the applicability of such noisier, photogentle datasets. Combining this photogentle imaging approach with our deep-learning pipeline enables detailed analysis of mitochondrial dynamics, such as fission and fusion events.
Conclusion
By establishing a photogentle imaging routine and quantitative pipeline, this research anticipates uncovering novel insights into the link between energy-consuming processes and mitochondrial dynamics. This work has broad implications for understanding mitochondrial roles in diverse biological contexts, including drug screening, disease modelling, and the investigation of fundamental cellular processes. Ultimately, this work aims to contribute a powerful toolset for bridging experimental and computational domains in biological research, facilitating a deeper understanding of mitochondrial dynamics and their impact on cellular function as we move toward more physiologically relevant contexts and models.