Towards Multi-Resolution Imaging of Dynamic Processes in Biological Systems

Abstract

Dynamic processes in complex systems, such as viral-cell interactions, cargo transport along the cytoskeleton and processive enzyme catalysis, span several orders of magnitude in space and time. To study these events, one must visualize, in real time and three dimensions, both the fast and small target of interest---the virus, the cargo or the enzyme---as well as the larger-scale, nearly stationary counterpart---the cell, the cytoskeleton or the substrate. Designed for this purpose, the 3D multi-resolution microscope (3DMRM) can simultaneously track a nanoscale target and image its surrounding environment. Although this tool provided the possibility to conduct such studies in real biological systems, the actual experiments that start with sample preparation to result interpretation can be complicated. This contribution addresses challenges that arise during the application of 3DMRM on different types of biological systems, in aspects of labeling strategies, trajectory analysis and resolution matching. To probe viral-cell interactions using more biologically-relevant virions instead of abiotic nanoparticles as in the proof-of-concept 3DMRM experiment, a labeling strategy that encapsulates nonblinking giant quantum dots inside pseudotyped HIV-1 virions was designed, which enabled the multi-resolution observation of more realistic viral-cell interactions. When studying a nanoscale object moving next to a curvilinear structure, such as in the cases of cargo transport along microtubules or processive cellulase catalysis on cellulose fibers, challenges remain in the lack of rigorous tools to extract basic physical parameters and detect their changes, as well as in super localization of the curvilinear features in images to bridge up the resolution gap between the tracking and imaging modalities. For trajectory analysis, a two-variable change point method was developed to detect velocity and/or diffusivity changes in single-particle trajectories. For resolution matching, an unsupervised learning framework was established to infer positions of line-shaped features in low signal-to-noise images, using only the descriptive prior information of the feature shape. These efforts help to resolve the uncertainty in detectability when moving from artificial nanomachinery to more biologically-relevant systems and allow for quantitative analysis of multi-scale datasets, thus facilitate the application of 3DMRM to direct observation and understanding of dynamics in complex biological systems.