Quantitative Super-Resolution Microscopy
Date of Award
Doctor of Philosophy (PhD)
Engineering and Applied Science
The diffraction limit fundamentally restricts the achievable resolution of conventional microscopes, however it can be circumvented if only a subset of fluorophores within a given diffraction-limited volume are simultaneously emitting. Such super-resolution methods routinely achieve tens of nanometer resolution in whole (if not living) cells. However, being relatively young, these techniques still struggle with providing robust, quantitative, and statistically powerful data in a routine manner. Accurate quantification of subcellular features requires an estimate of the achieved spatial resolution, but traditional methods of measuring it do not cope well with these techniques. Effects subject to biological variability additionally require increased population sampling and controls, however achieving spatial super-resolution comes at the cost of throughput. This is particularly limiting for single-molecule localization microscopy (SMLM), which can only image a handful of cells a day. In this thesis, I developed methods and instruments to make super-resolution microscopy a more quantitative tool. Measuring resolution by imaging beads and fitting their widths is adequate when the beads are substantially smaller than the effective point-spread function (PSF) width and that width is not very dependent on fluorophore/local-environment properties, however these requirements are often unsatisfied for super-resolution microscopy, making beads poor proxies for biological resolution. I developed a method to simultaneously fit the PSF width and underlying feature size directly from images if the structure of the feature can be modeled. This enables direct resolution and feature measurement even in live-cell stimulated emission depletion images. Large SMLM studies can easily require weeks to months of camera integration time alone. Raw frame rates can be increased from the typical 50 Hz to 800 Hz, however this has been largely unexplored for multicolor and 3D imaging, and the resulting 35 - 70 TB/day data volumes and reconstruction processing have been intractable. I leverage advances in hardware control, data compression, and GPU-accelerated localization to enable multicolor 3D SMLM at rates of 10,000 cells a day. To tackle entire multi-condition studies in tens of hours, I extend our analysis architecture to dynamically queue new acquisition tasks, enabling both multi-well plate acquisitions and high-latency decision workflows such as optimizing settings per cell from intermediate widefield images. Finally, I propose an acquisition concept for SMLM which enables extended fields of view without nonlinearities from stitching or bleaching artifacts. Continuous movement of the 3D focal position adiabatically with respect to localization precision theoretically allows high-quality SMLM reconstructions limited only by the travel of stages, preserving unbounded spatial resolution across now-unbounded sample volumes.
Barentine, Andrew, "Quantitative Super-Resolution Microscopy" (2021). Yale Graduate School of Arts and Sciences Dissertations. 298.