Key Concepts:

Figure 1. Workflow of single particle processing in CET displayed in 2D for simplicity. First, tomograms are reconstructed from projections along different directions (‘tilt series’). The membrane-associated ‘particles’ of interest (‘A's) first need to be localised. A higher resolution structure of the complex of interest is determined by aligning and averaging the subtomograms. However, the resulting average remains somewhat fuzzy because the subtomograms image the complex of interest in different conformations (different fonts). Classification procedures group the subtomograms according to the different conformations. The corresponding class averages exceed the single average in resolution.

Figure 2. Subtomogram alignment and classification. (a) ‘Multireference alignment’ and averaging of subtomograms sketched schematically. Positions, orientations and class assignments of subtomograms are iteratively refined by ‘expectation maximisation’: in each iteration, subtomograms are aligned and classified according to the maximal score with a class-reference. In single-reference alignment, only one reference is used. (b) Variance of subtomograms explained by eigenvectors determined by PCA and CPCA for simulations of the chaperone thermosome (SNR=0.1) in two different conformations (Förster et al., 2008). Ideally, two eigenvectors should explain the data (mean value and difference of the conformations). With increasing size of the missing wedge, the third and fourth eigenvectors get more prominent – however, this effect is substantially smaller for CPCA. (c) CPCA-based classification of experimental tomograms of GroEL and GroEL/GroES (Förster et al., 2008). Left: Obtained class averages. The orange density is reminiscent of GroEL/GroES (ribbon diagram: GroEL/GroES crystal structure) and the green one to GroEL without GroES. Right: Subtomogram contributions to the second eigenvector. The second eigenvector correlates to the origin of the subtomograms (either tomograms of vitrified GroEL solution or GroEL/GroES solution). Subtomograms are not perfectly separated, either due to limited specificity of the method or due to sample heterogeneity.

Figure 3. Membranes of subtomogram averages from retroviruses imaged in toto. (a) Isosurface of a cryo-tomogram of a Moloney murine leukaemia retrovirus (membrane: yellow, Env complexes: magenta) positioned above a projection of the virion (Förster et al., 2005). (b) Envs particles were aligned and averaged to a resolution of 1/27 Å^–1, which allowed tentative fitting of the receptor binding domain into the density map. (c) Average of subtomogram depicting small regions of membrane and underlying density of HIV particles (Briggs et al., 2009). The surface cut perpendicular to the membrane reveals two membrane leaflets and the layers formed by the capsid (CA) module of Gag and the ribonucleoprotein–RNA complex (RNP). (d) Orthogonal view (plane: dashed black line in c) onto the N-terminal CA domains revealing the hexagonal lattice enclosing the HIV capsid. (e) Cut through the C-terminal CA domains (plane: white dashed line in (c)). Hexagon, triangle and rhombus indicate examples of six-, three- and two-fold symmetry axes, respectively, (c)–(e). Reproduced with permission from Briggs et al. (2009).

Figure 4. Nuclear pore complex (NPC) in situ. (a) Isosurface rendering of the Dictyostelium discoideum NPC (blue) in the nuclear membrane (yellow) at a resolution of approximately 1/6 nm^–1 (Beck et al., 2007). (b) View from the cytoplasm through the central transport channel formed by NPC.

Figure 5. Ribosomes imaged in vivo and in situ. (a) The background displays a slice through a tomogram of an intact E. coli cell, overlaid by an isosurface of the segmented cell membranes. Isosurface representations of the 100S ribosome are positioned according to their localisation. (b) Magnified isosurface of the in vivo hibernating ribosome (50S subunit marked blue, 30S marked gold). (c) Isosurface of the hibernating ribosome determined from in vitro samples. The orange isosurface corresponds to ‘unidentified densities’, that is, densities that are not present in the crystal structures of the 70S ribosome, close to the exit of the mRNA tunnel (Ortiz et al., 2010).