II: Single particle cryoem - an averaging technique Radiation damage limits the total electron dose that can be used to image biological sample. Thus, images of frozen hydrated macromolecules are very noisy, with extremely low signal-to-noise ratio (SNR). Single particle EM (both negative stain and cryo) is to extract structural information (both 2D and 3D) of macromolecules by averaging a large number of molecules without crystals. An image is the projection of a 3D object Fourier Central section theorem Central Section Theorem : Fourier transform of a 2D projection equals the central section through its 3D Fourier transform perpendicular to the direction of projection. DeRosier, D. and Klug, A. (1968) Reconstruction of three dimensional structures from electron micrographs Nature 217 130-134 Hart, R.G. (1968) Electron microscopy of unstained biological material: the polytropic montage Science 159 1464-1467 DeRoiser and Klug (1968)
Cryo-EM techniques Electron crystallography: for membrane protein resolution achieved: 2.5Å bacteriorhodopsin, Aquaporin-0 (recently: 1.9Å aquaporin-0) averaging of molecules which form 2D crystal; Single particle cryo-electron microscopy: protein in soluble form; resolution achieved: intermediate 20 Å ~ 3 Å averaging of many single molecules with the same structure; cryo-electron tomography: large cellular organelle, whole cell etc. Resolution achieved: > 3nm no averaging, 3D reconstruction of single biological object; Frozen hydrated specimen preparation Adrian M, Dubochet J, Lepault J & McDowall AW (1984) Cryo-electron microscopy of viruses. Nature 308, 32-36. sample solution forceps Liquid N 2 holey carbon grid Filter paper Liquid ethane carbon matrix vitrified ice Quantifoil grid Plunge freezing Single particle cryo-electron microscopy of human transferrin receptor-transferrin complex
Protein molecules embedded in vitrified ice as single particles Carbon matrix vitrified ice " # y! x The geometry of each particles are determined by 5 parameters: three Euler angles and two in plane shifts. Single particles are randomly oriented in vitreous ice Single particle 3D reconstruction Single particle 3D reconstruction is a technique based on averaging. Many images of the same molecule at random orientations are needed. Every individual image is very noisy with unknown orientation. * to provide sufficient views covering entire Fourier space; * to improve signal-to-noisy ration (SNR); * all images needed to be aligned with each other; Therefore: the resolution of a 3D reconstruction is dependent on: * total number of images; * accuracy of alignment; * complete coverage of Fourier space; Image averaging Cryo-EM images are very noisy; have extremely low signal-tonoise ratio. Averaging of a large number of images are necessary to improve the SNR.
Averaging in darkroom Photographic image superposition (averaging) by Roy Markham, who shifted image and added to the original in darkroom. The trick is to know decide much and which direction to shift the image for superposition. Averaging in computer. David DeRosier used Markham s lattice to determine how much to shift, and performed averaging by using Adobe Photoshop. Averaging in 2D crystals How much and which direction to ship the image can be determined easily from FT of the image of a 2D crystal.
Image averaging in 2D crystal In 2D crystal, one can extract amplitudes and phases from peaks of FT (contributed by the identical repeats of structural motif) and ignore everything in between peaks (contributed by the random noise). A reverse Fourier Transform using extracted amplitude and phases will give us an averaged features. This is equivalent to the averaging. It is easy to perform averaging in 2D crystal. The molecules in the 2D crystal are identical in composition and orientation. Fourier Filter A noisy image of 2D crystal can be filtered to enhance SNR. Images of identical molecules in different orientations 50 out of 36,266 particles
Iterative refinement procedure Iterative refinement procedure, using reference model based projection matching: Generate a set of projections 3D model A better 3D model Projection matching with class averages 3D reconstruction Resolution estimation In single particle cryoem the resolution is often estimated by Fourier Shell Correlation. F n G n Reconstruction 1 Reconstruction 2 FSC(R) = $ %# & # F n G n n "R 1/ 2 ' 2 2 F n # G n ( n "R n "R ) FSC 1/A 0.5 An example: single particle cryoem of T. acidophilum 20S proteasome Cryo-electron microscopy: * Tecnai F20 microscope * Magnification: 50K * Pixel size: 1.37 Å * Defocus: -1.5 ~ 3.5µm * > 40,000 particle for each complexes * FSC = 0.143 as nominal resolution criteria
Initial model used for refinement * Filtered to 30A from atomic structure of 20S proteasome FSC curve of final 3D reconstruction Side view Structure of 20S proteasome α Top view β β α * Density map filtered to 6.8 Å;
Structure of 20S proteasome Side view α Top view β β α * Density map filtered to 6.8 Å; Structure of 20S proteasome Side view α Top view β β α * Density map filtered to 6.8 Å; Structure of 20S proteasome Side view α Top view β β α * Density map filtered to 6.8 Å;
CryoEM map of 20S at 4.5 Å resolution CryoEM map of 20S at 4.5 Å resolution CryoEM map of 20S at 4.5 Å resolution
An example, human TfR-Tf complex Crystal structures of diferric and apo-transferrin Rabbit serum transferrin Duck apo-ovotransferrin Hall et al. (2002) Acta Crystallogr. D 58: 70-80 Rawas et al. (1989) J. Mol. Biol. 208: 213-214 Crystal structure of the ectodomain of human TfR Proposed binding of Tf to TfR Helical Domain Apical Domain Proteaselike Domain Stalk (~3 nm) Plasma Membrane Lawrence et al. (1999) Science 286: 779-782
The TfR-Tf complex in vitrified ice 100nm Individual TfR-Tf Complexes in Vitrified Ice 50 out of 36,266 particles Class Averages of Vitrified TfR-Tf Complexes 50 out of 200 classes
3D Reconstruction using IMAGIC 90º 90º 180º Angular Distribution Fourier Shell Correlation 7.5Å
Refined 3D map obtained by FREAGLIN Fit of TfR and Tf Crystal Structures Comparison between Map and Model Density Map Model
Single particle v.s crystallography Crystallography is also an averaging technique: * all molecules are aligned by forming a crystal; * crystallinity guaranteed homogeneity, i.e. crystal is formed by both compositional and conformational identical molecules; * diffraction patterns can assess the quality of a crystal; Single particle v.s crystallography Single particle: * image of every individual molecule has has to be aligned computationally; * homogeneity is not guaranteed; * R. Henderson pointed out in 1995 that atomic resolution can be reached as far as one can align image accurate enough. He also predicted how many images are needed to achieve this level of resolution. Considering that low-does images of frozen hydrated sample are always extremely noisy, reference model induced bias is very strong. It is therefore possible to produce a wrong structure.
Alarming example: Structure of IP3 receptor Structure of the type 1 onsoitol 1,4,5- trisphosphate receptor revealed by electron cryomicroscopy. JBC 2003, 278, 21319-22. Insoitol 1,4,5-trisphosphate receptor contains multiple cavities and L-shaped Ligand-binding domains. JMB 2004, 336, 155-64. Alarming example: Structure of IP3 receptor da Fonseca P C A et al. PNAS 2003; 100:3936-3941 Alarming example: Structure of IP3 receptor IP3 receptor (Ludtke, et. al. 2011, Structure)
Sample requirement for structural analysis (X-ray crystallography): * high purity; - clean SDS PAGE gel (better with silver stain) * homogeneous conformation; - nice size exclusion chromatography profile * stable complex; - does not fall apart during crystallization Same are required for high-resolution single particle cryoem structural studies.