Schematic diagram of the workflow for localized reconstruction. First the structure of the macromolecular complex is solved using conventional 3D refinement (1), after which the locations of the sub-structures (red spheres) are calculated based on the particle orientation, a symmetry operator and a vector defining one sub-structure relative to the particle model (red stick; 2). After extracting the sub-particles (red dots) from the particle images, a localized 3D reconstruction is calculated (3). This reconstruction can be used as a starting model for further classification (4) and 3D refinement (5) of sub-particles to improve the structure. Finally two independent sets of data (6) are compared by Fourier shell correlation (FSC) to assess the resolution of the reconstruction (7). See text for a detailed description.

(a) Three representative 3D sub-particle class averages (1, 2 and 3) were filtered to 24-Å resolution for comparison and are shown (red) from the side (top row) and from the top (bottom row). The original particle 3D reconstruction (at 35-Å resolution; transparent blue) is shown at a lower threshold to provide a frame of reference. The vertices exhibited two modes of flexibility, in the vertex height and edge angle, as indicated by dashed lines in top and bottom row, respectively. (b) A comparison of the three classes (coloured in different shades of red, 1 being the darkest, 2 being medium and 3 being the lightest) is shown as cross-sections of the vertices. The difference between the heights of the vertex in classes 1 and 3 is 15 Å (top row) and the angular difference is two degrees (bottom row). (c) The earlier published structure of the fixed COPII cage at 12-Å resolution (EMD-5524) and (d) the localized reconstruction of the native COPII vertex solved in this study to 14-Å resolution are both shown along the twofold axis of symmetry for comparison.

Subtraction of densities corresponding to the bulk of the particle is shown in a schematic diagram illustrating a rotavirus triple-layered particle (in blue with external parts of the VP4 spikes in pink). (a) First, subunits of interest are removed from the reconstruction of the complex. A mask (green), defining the subunits of interest, is created and then applied on the particle reconstruction to remove densities within the mask. (b) The masked reconstruction is projected in the orientations of the particle images (top) to create a stack of masked particle projections (middle). These are then subtracted from the original particle images, to create a new stack of particle images (bottom) that contain density corresponding mainly to the subunits of interest.

(a) Three-dimensional classification of VP4 spikes. Class averages (pink) are shown rendered at the same isosurface threshold and resolution (12 Å). The number of sub-particles in each class is indicated. The VP7 density (transparent blue), averaged from all of the sub-particles, is shown as a frame of reference. Sub-particles in the full data set (top row) were first classified into four classes (second row). The first two classes revealed a clear spike density. The sub-particles in these classes were further classified into four classes (bottom row). These classes showed minor differences in the spike conformation. (b) Localized reconstructions of the VP4 spike are shown before (∼260,000; left) and after (∼113,000; right) classification and selection of the best sub-particles. Both of the maps have been low-pass filtered to 7.7-Å resolution and rendered at the same isosurface threshold. Most importantly, classification of the sub-particles before reconstruction improved the density for the tip of the spike (rectangle). (c) Close-up of the spike tip is shown from the side (left) and the top (right). The structure that forms the two tips of the spike (PDB:4DRR) has been fitted to the density.

(a) Icosahedral reconstruction of the polymerase complex particle at 4.8-Å resolution is shown along the icosahedral threefold axis of symmetry. The P1 monomers around fivefolds are coloured in blue and the P1 monomers around twofolds and threefolds in red. (b) Four three-dimensional sub-particle class averages (pink, resolution 12 Å) of the densities under the threefolds and the P1 shell, averaged from all of the sub-particles to provide a frame of reference (blue, resolution 12 Å), are shown. The view is from the inside of the particle. Three classes revealed clear P2 density in the three possible orientations relative to the symmetry axis (0°, 120° and 240°). For another class no P2 density was evident (asterisk). Incorrectly averaged P2 density in the original threefold symmetrized reconstruction is indicated with an arrowhead. (c) Localized reconstruction of the P2 density at 7.9-Å resolution with a fitted X-ray structure of P2 (PDB:1HHS) is shown from three different orthogonal orientations. Colouring of the X-ray structure follows the canonical polymerase domain architecture (fingers, red; palm, green; thumb, blue; C terminus, yellow; connecting chains, mauve). Two small densities, not accounted by the P2 X-ray crystallographic structures, are indicated with arrowheads. (d) P2 density, reconstructed together with the P1 shell, is shown in the same orientation as in c. Densities connecting the P2 polymerase to the surrounding P1 shell are marked with arrowheads, and correspond to the densities marked in c. (e–g) Possible contact sites between P2 (the coloured ribbon) and P1 (grey ribbon) are shown and the corresponding amino acid residues labelled.