UCLA

Moving over X-ray crystallography, single particle cryogenic electron microscopy (cryoEM) has emerged as a method for determining atomic structures of purified proteins in recent years, kicking up a “cryoEM revolution”. However, obtaining high-resolution, in-situ cellular structures that faithfully represent the native environment remains a formidable challenge. To address this need, cryogenic electron tomography (cryoET) has emerged as an invaluable tool for biomedical research. CryoET has the unique capability to provide detailed insights into the molecular mechanisms and biological functions of macromolecular assemblies within their native cellular context. This dissertation embarks on a journey through the world of cryoET combined with subtomogram averaging (STA) to unravel the mysteries of cellular organelles. The first part of this work showcases the power of cryoET and STA in studying the entire flagellar structure of Trypanosoma brucei. This includes the elucidation of the conserved "9+2" axonemal 96 nm subunit, paraflagellar rod (PFR), PFR and Axoneme connectors (PACs), and center pair complex (CPC) structure. This comprehensive exploration sheds light on the motility unit of T. brucei, crucial for the parasite's transmission and pathogenesis. Leveraging the RNAi knockdown technique, we delve into the functions of critical proteins like DRC11, unraveling their structural role in parasite motility. The dissertation further addresses major challenges in the cryoET and STA pipeline through the introduction of two innovative software packages. Firstly, IsoNet tackles the persistent issue of "missing wedge" artifacts. In cryoET data collection, obtaining high-tilt angle images becomes impractical due to the increasing effective sample thickness. This missing information results in wedge-shaped artifacts in Fourier space, causing elongation artifacts along the Z-axis. IsoNet employs deep neural network techniques to fill meaningful signals into the Fourier space's missing wedge area, significantly mitigating or eliminating these artifacts. Secondly, TomoNet focuses on automating particle picking functions through a blend of optimized template matching and deep learning methods. This powerful combination aids in the identification and localization of particles, particularly within lattice-like structures. TomoNet not only simplifies the organization of complete tomography projects but also efficiently manages extensive tomogram datasets through its user-friendly graphical interface. To illustrate TomoNet's particle picking capability, we implemented it in the study of the sheath structure of Methanospirillum hungatei. Applying cryoET with STA, we attained a remarkable 7.9 Å resolution, unveiling the topological features and subunit organization responsible for the native cylindrical sheath tube. Furthermore, a near-atomic level structure prediction of the monomeric sheath protein provides intricate insights into the architecture of individual M. hungatei sheath hoops and their assembly into the sheath layer. We propose a novel route for the synthesis and assembly of sheath fibrils, specifically their insertion into growing sheath sites at the junctions between individual cells. This body of work exemplifies the immense potential of cryoET, STA, and innovative software solutions like IsoNet and TomoNet in elucidating complex biological structures and mechanisms.