The Development Of Three-dimensional Super-resolution Microscopy System, Localization Algorithm And Fluorescent Proteins

Fluorescence microscopy has become an indispensable tool for modern cell biology. Because of the issue of diffraction, the resolution of traditional light microscopy is restricted approximately to～200nm in lateral and～500nm in axial direction, which is insufficient for subcelluar study. However, with the development of new fluorescent probes and imaging theories, the last decade has witnessed the emergence and development of various super-resolution imaging techniques which break the diffraction barrier and shed light on biological studies at subcellular level.Despite these breakthroughs, super-resolution microscopy also faces some challenges. This thesis introduces the improvement of super-resolution microscopy from the following three aspects:The first part mainly focuses on the establishment of a three-dimensional super-resolution microscopy system based on single molecule interference (chapter2).Several approaches have attempted to encode the axial position of fluorophores including astigmatism, biplane imaging and double-helix point spread function. Nevertheless, the axial resolution is still below the lateral resolution. To further improve the axial resolution, dual-objective interferometry methods such as iPALM (interferometric photoactivated localization microscopy) and4Pi-SMS have been developed. However, both of the two methods are too expensive and difficult to implement for non-experts. So far, no other laboriteshave established the same system. Hence, we designed and setup a simpler system based on single molecule interference, which is easy to adjust and affordable for ordinary lab. We described the basic structure and principle of the system as well as the setup and adjust process, and performed experimental imaging. The results indicated that the resolution of the system can reach at20～30nm in three dimensional.The second part is about the development of single molecule localization algorithm based on an artificial neural network(chapter3). The most commonly used method for single molecule localization is two-dimensional Gaussian fitting either by nonlinear least-squares (NLLS) or maximum likelihood estimation (MLE). However, Gaussian is not the true point spread function. Therefore, it is not appropriate for fluorophores with fixed dipole orientations and can introduce systematic errors of tens of nanometers. Besides, fluorophores in cells may not be fixed but have different rotational mobility. The majority of fluorophores are neither fixed nor free, but in an intermediate state called restricted motion. Hence, to maximize the resolving power of super-resolution microscopy, it is important to develop an unbiased estimator with high speed and robustness for dipoles with orientation and restricted motion.We derived the point spread function of tilt dipoles with restricted mobility and developed an algorithm based on an artificial neural network (ANN) for estimating the localization, orientation and mobility of individual dipoles. Compared with fitting-based methods, our algorithm demonstrated ultrafast speed and higher accuracy, reduced sensitivity to defocusing, strong robustness and adaptability, making it an optimal choice for both two-dimensional and three-dimensional super-resolution imaging analysis.The third part describes the rational design of a true monomeric and bright photoactivatable fluorescent protein mEos3(chapter4).The performance of super-resolution imaging relies heavily on the characteristics of photoactivatable fluorescent proteins (PA-FPs). Key properties that determine the practical use of PA-FPs include (but not limited to) size, brightness, maturation rate, oligomeric nature, pH stability, photon budget after switching. Recently, it has been realized that the often-neglected label density limits the effective spatial resolution and this limitation applies to all super-resolution fluorescence microscopy.Monomeric (m)Eos2is an engineered photoactivatable fluorescent protein widely used for super-resolution microscopy. However, we proved that mEos2tended to oligomerize at high concentrations and form aggregates when labeling membrane proteins, which suggests that it is not a true monomer and limits its application as a fusion partner. We solved the crystal structure of tetrameric mEos2and rationally designed two substitutes that are truly monomeric, brighter, faster in maturation, and exhibit higher photon budget and label density. The thesis described the improvement and optimation of super-resolution microscopy from the hardware system, localization algorithm and fluorecent proteins. These three parts are mutually independent, but complementary forming a compete system, which expands the application of super-resolution imaging.