Resolution and Probabilistic Models of Components in CryoEM Maps of Mature P22 Bacteriophage

Abstract

CryoEM continues to produce density maps of larger and more complex assemblies with multiple protein components of mixed symmetries. Resolution is not always uniform throughout a cryoEM map, and it can be useful to estimate the resolution in specific molecular components of a large assembly. In this study, we present procedures to 1) estimate the resolution in subcomponents by gold-standard Fourier shell correlation (FSC); 2) validate modeling procedures, particularly at medium resolutions, which can include loop modeling and flexible fitting; and 3) build probabilistic models that combine high-accuracy priors (such as crystallographic structures) with medium-resolution cryoEM densities. As an example, we apply these methods to new cryoEM maps of the mature bacteriophage P22, reconstructed without imposing icosahedral symmetry. Resolution estimates based on gold-standard FSC show the highest resolution in the coat region (7.6 Å), whereas other components are at slightly lower resolutions: portal (9.2 Å), hub (8.5 Å), tailspike (10.9 Å), and needle (10.5 Å). These differences are indicative of inherent structural heterogeneity and/or reconstruction accuracy in different subcomponents of the map. Probabilistic models for these subcomponents provide new insights, to our knowledge, and structural information when taking into account uncertainty given the limitations of the observed density.

Figure 1

Entire P22 density map viewed as (A) isosurface and (B) isosurface that is sliced through the middle. (C and D) Segmented components including the coat (brown), portal (green), hub (purple), tailspikes (blue), needle (gold), and DNA (red). (E) Gold-standard FSC plots of the entire map and individual components. This figure shows the complex arrangement of various components in the mature P22 virion, and how resolution can vary significantly from component to component.

Figure 2

(A) Gold-standard FSC plots for entire components (coat, portal, tailspikes, and hub) and one protein from each component. Although the map is asymmetrically reconstructed, the components are approximately symmetric (coat-icosahedral, portal-C12, hub-C12, tailspike trimers-C6). Hence the FSC plots for entire components and for single proteins from each component are very similar. (B) The effect of mask smoothness and whether the same or different masks are used in gold-standard FSC plots. Using the same mask or a mask with a small smoothing width (w) in both independent maps can introduce artificial correlations at high frequencies. Using independent masks and larger smoothing widths removes this effect. To see this figure in color, go online.

Figure 3

(A) Coat, portal, and tailspike densities with (on the left) docked models and (on the right) resmap resolution coloring (blue to red). (B, D, F, and H) Rigidly docked models of PDB structures and segmented densities of coat, portal, hub, tailspike, and needle proteins. Segmented densities are shown with a transparent gray surface and docked models are shown using a colored ribbon. PDB codes for docked models are as follows: coat, brown (PDB: 2XYZ); portal, green (PDB: 3LJ5); hub, purple (PDB: 3LJ4), tailspike N-terminal head-binding domain, cyan (PDB: 1LKT); tailspike C-terminal adhesin domain, orange (PDB: 1TYU); needle, yellow (PDB: 2POH). (C, E, G, and I) Surfaces of segmented densities of single proteins from each component, colored with per-voxel resolutions computed with resmap.

Figure 4

Flexible fitting validation protocol using two independent maps (A and B). On the far left, entire portals extracted from the two independent maps are shown, with a single protein colored in green. Map B is aligned to map A, though in the images the densities are shown separately. The protocol involves docking of a crystal structure and modeling based on map A (top row), then evaluating the resulting model by FSC plot to both maps A (top row) and B (bottom row). To see this figure in color, go online.

Figure 5

The top row shows segmented densities for a single protein from the portal as a transparent surface, the initial, rigidly docked model (A) with green ribbon, and the model after flexible fitting with MDFF with two different gradient weights of 0.3 (B) and 500 (C). On the bottom row, the FSC between simulated maps for each model and maps A and B is plotted, along with FSC between map A and map B (a single protein was smooth-masked in both maps). After flexible fitting with gradient scale of 0.3, the FSC between the model and map A and map B improves but stays below the signal plot between map A and map B, hence no overfitting took place. Using an extremely large gradient scale of 500, the FSC between the fitted model and map A (in which it is fitted) increases more dramatically, but the FSC between the fitted model and map B stays below the signal plot, indicating overfitting took place. The fitted model in this case also looks extremely distorted. To see this figure in color, go online.

Figure 6

(A and C) Segmented protein densities are shown with a transparent surface, whereas probabilistic models are shown as a worm model. Ribbon thickness and color correspond to standard deviations at each backbone residue. (B and D) The segmented protein densities for the proteins are also shown color-coded with resmap resolutions for comparison. Although higher standard deviations appear to correspond to lower resolutions for the portal protein, in the case of the hub and tailspike proteins, standard deviations are lower even at lower resolutions. Thus, standard deviations are not only influenced by resolution alone, but also by structural composition and the fitting method.