Development of computational methods for exploration of cryo-electron tomograms
Biophysics
Final Report Abstract
Macromolecular complexes are traditionally studied by approaches that require biochemical purification of the molecule of interest, such as X-ray crystallography. However, for many cellular assemblies, most notably those associated to cellular membranes, purification is not feasible and the purification process, in particular solubilization, may alter the molecular structures. Cryo-electron tomography (CET) is a three-dimensional (3D] imaging modality that allows structural studies of complexes in their physiological settings, i.e., in situ. While the resolution of a cryo-tomogram per se is typically limited to 5-10 nm statistical analysis of large amounts of 3D images of different instances of a specific macromolecule can reveal much higher resolution insights into its structure. In this project we have developed image processing methods for obtaining detailed insights into the structures of complexes based on CET. We developed a workflow that (i) enables 3D reconstruction of tomographic volumes from transmission electron microscope micrographs, (ii) detects macromolecular complexes in tomograms, (iii) coherently aligns subtomograms depicting copies of a macromolecule, and (iv) classifies subtomograms according to different conformational states. Evaluations using simulations and experimental tomograms indicate that the developed algorithms for all these tasks bear significant advantages over state-of-the art methods: (i) An iterative Fourier space based reconstruction algorithm improves reconstruction accuracy (highlighted in Faculty of 1000, http://f1000.com/prime/718205092); (ii) machine learning classification using different subtomogram features as input improves identification specificity; (iii) spherical harmonics based subtomogram alignment increases alignment speed by several orders without loss of accuracy; (iv) an algorithm for simultaneous alignment and classification focusing on the most variable features of subtomograms improve classification accuracy. The developed methods have been used for different in situ structure studies during the funding period. Specifically, the native co-translational translocation machinery of the mammalian Endoplasmic Reticulum and the yeast mitochondrion was characterized as well as the mammalian 26S proteasome in intact cells, revealing unexpected features of their associated machineries.
Publications
- Fast and accurate reference-free alignment of subtomograms. J. Struct. Biol. 182:235-45 (2013)
Chen Y, Pfeffer S, Hrabe T, Schuller JM & Förster F
(See online at https://doi.org/10.1016/j.jsb.2013.03.002) - Autofocused 3D Classification of Cryoelectron Subtomograms. Structure 22:1528-37 (2014)
Chen Y, Pfeffer S, Fernandez JJ, Sorzano CO & Förster F
(See online at https://doi.org/10.1016/j.str.2014.08.007) - Iterative reconstruction of cryo-electron tomograms using nonuniform fast Fourier transforms. J. Struct. Biol. 185:309-316 (2014)
Chen Y & Förster F
(See online at https://doi.org/10.1016/j.jsb.2013.12.001) - Structure of the mammalian oligosaccharyl- transferase complex in the native ER protein translocon. Nat. Commun. 5:3072 (2014)
Pfeffer S, Dudek J, Gogala M, Schorr S, Linxweiler J, Lang S, Becker T, Beckmann R, Zimmermann R & Förster F
(See online at https://doi.org/10.1038/ncomms4072)
