The Structure of Escherichia coli Signal Recognition Particle Revealed by Scanning Transmission Electron Microscopy
Mol. Biol. Cell Mainprize et al. 17: 5063 Supplemental Material
This article contains the following supporting material:
- Supplementary Figure 1 - Angular plot and resolution determination of SRP reconstruction. A) The angular distribution of the SRP particle projections in the micrographs was not completely isotropic. Each dot represents the angular orientation assigned to a micrograph of a single particle. Clustered dots indicate the orientations that were more favourable for SRP lying on the carbon film. Areas without dots represent orientations in which SRP did not lay on the carbon film or that were not recognized as SRP structures. B) The orientations corresponding to clustered and sparse orientations of the E. coli SRP molecules were identified by forward projection of our SRP reconstruction. The regions with the highest occurrence coincide with different side-on views of SRP (projections 1, 5, 6, 9, 10 and 11). The fewest projections originated from a "back" view, or the view of SRP which resembles the mirror image of the numeral 9 (as in projections 2 and 4). However the a "front" view (projection 7) was well represented. C) The region of sparse orientations in A is equivalent to a missing cone of about 25º. The point spread function indicates that this missing cone would result in a worsening of the resolution by about 10% in the direction at right angles to the cone axis. For a 12Å resolution in the reconstruction, this could result in a smearing on the order of 1.2Å on the edge of the reconstruction. This factor alone can not account for the unexpected increased width of the tail of the SRP reconstruction (~42Å) compared to the modeled RNA (~25Å). An empty region of 120º (half-angle of 60º) would be required to effect such a degree of smearing. Conformational variability in the RNA tail due either to motion in this region or alignment of different conformers, likely contributes significantly to the increased tail width of the reconstruction. D) The Fourier shell correlation (FSC) method was used to estimate the resolution of the E. coli SRP reconstruction. To calculate the resolution of a reconstruction calculated from electron micrographs the population of images is randomly divided into two groups and a reconstruction is generated for each subset of images by back-projection using the previously determined angular orientations. These two Ahalf-reconstructions@ are compared in Fourier space as a function of their spatial frequencies. Essentially this corresponds to analyzing how similar two reconstructions are, each calculated using half of the images. The more images analyzed the better this approach defines the resolution obtained. The resolution limit of the reconstruction has been defined in the literature as either the resolution that corresponds to 0.5 FSC on a FSC versus spatial frequency plot or the resolution at which the FSC plot crosses a 3-sigma threshold line (i.e. the 3-sigma threshold is the FSC obtained from the comparison of two 3-D densities composed solely of three standard deviations of background noise). We obtained resolutions of 12Å for our SRP reconstruction calculated from 3,236 particle images using the 3-sigma criterion and 19Å using the more conservative 0.5 FSC criterion (which is the spatial frequency that results in an FSC value of 0.5). The implications of this estimation of resolution for the interpretation of details in the resulting reconstruction can be hard to assess. For this reason it is common to also compare the reconstruction to a model displayed after filtering to a variety of different resolutions (when using filtering, resolution is defined as the minimum distance two points must be separated such that the intensity between them drops to half the maximum value) (see Supplementary Figures 3 and 5).
- Supplementary Figure 2 - The effect of the reconstruction process on the reconstruction obtained. During the refinement process (projection matching), the particle images are aligned by the dominant/major feature(s). In the initial stage of the refinement, the images were randomly divided into 16 subsets of 201images. A reconstruction was calculated for each subset of images. These subset reconstructions were then combined into a single reconstruction and used to seed further refinement. Due to the noise in the images and the limited number of images in each subset, these initial reconstructions (generated from angular alignment via IQAD) result from coarse alignments of the images. In this case, certain features in the image subsets dominate the starting structures. In the case of E. coli SRP, the dominant features are that it is a flat, extended particle. Compounding this effect is that the population of STEM images of SRP had a higher abundance of side-on views of SRP that appear as rods, similar to the views seen in Figure 1B, right, and views 1, 5, 6, 9, 10 and 11 of Supplementary Figure 1B. When all of the particle images are initially aligned, conformational variations that cause a deviation from a flat particle will result in repositioning of the image to maximize the rod shape. This results in misalignment of images in which a domain has moved out of the flat plane of the molecule based on the rod shape. As a result minor populations of particles are averaged out and disappear if misalignment of different views are inconsistent (as expected based on the noise in the images). Thus minor populations of molecules resulting from motion provide only a minor contribution to the overall reconstruction. For SRP, the process of the reconstruction can be deduced by examining intermediates in the reconstruction process. A) Representative reconstructions determined from some of the initial 16 subsets of particle images were aligned in 3-D by EMAN (Ludtke et al. , 1999). Top row, 2-D projections of the aligned reconstructions from a common angle which represent the side-on view of SRP. Many of the individual STEM images resemble these side-on views (see Supplementary Figure 1). Bottom row, hard surface representations (volume corresponding to approx. 90kDa) of the aligned reconstructions from a view related to the projections in the top row by a 90º rotation about the long axis of the reconstructions. In this view, one can appreciate that the refinement process is positioning the protein domain of SRP at various locations, sometimes more than one location in a single reconstruction (e.g. #1). These reconstructions generated from the initial subsets of particle images show a higher degree of structural heterogeneity since there are fewer images available for the refinement and reconstruction calculation. Significantly, some of the initial images closely resemble the final reconstruction (e.g. #4). From this figure, it is evident that the initial reconstructions were aligned primarily based on the flat, rod shaped features, as the side-on views of the aligned structures match quite well but the perpendicular views are not accurately matched. For instance, the globular head region of reconstruction #3 has been aligned such that it is on the opposite end of the particle compared to reconstruction #2. Therefore, for these smaller populations of images, the rod-shaped quality of the particle is the dominant feature. In subsequent refinement steps, all of the images will independently match to projections calculated from the previous reconstruction. An early result of this process is shown in B. B) A hard surface representation of a reconstruction obtained from an early stage of refinement, just after the small subsets of particle images have been merged to produce a single reconstruction. This reconstruction is shown from three views: right, the number 9' view; middle, side-on view (90º rotation about long axis); and, right, a slight rotation from the side-on view about the horizontal axis to produce an oblique, rod shaped view. After merging the subsets of images, the resultant reconstruction has an obvious globular region and an extended tail which persist in the final SRP reconstruction (Figure 1B). The increased abundance of the rod shaped particle images in the STEM image population result in an improved resolution from the corresponding view of the SRP reconstruction (middle) as evident from the repetitive RNA helical features visible even at this early stage in the refinement process from the side-on view (black arrows), similar to those seen in Figure 1B. One consequence of the refinement algorithm forcing/favoring the flat, rod shaped view is that any deviation in the RNA tail in the direction indicated by the black, double-headed arrow would result in a decrease in the effective resolution of the tail in this direction but the image would still fit well with the dominant features. Therefore, misalignments in other regions of SRP that may result from bona fide populations of different structures would be refined to a single structure of the head domain by allowing misalignment of the RNA tail in the direction indicated by the double-headed arrow. The unexpected increase in the width of the RNA tail is likely caused by this effect. The effect of convergence on a dominant structure from populations of molecules will be described in more detail elsewhere (Mainprize et al. , MS in preparation).
- Supplementary Figure 3. - Reconstruction of the RNA of SRP compared to RNA models - A) Repeating ridges seen in the tail of the E. coli SRP reconstruction are comparable to repetitive features of the RNA helix predicted for SRP RNA. The image in A is similar to the image shown in the main text Figure 1B. The lack of a clearly discernible continuous ribbon-like structure is indicative of an RNA molecule that deviates from a perfect double helix (see C). B) The RNA molecule in the ec_SRP model (right, blue), which was created from a known structure (top 1/3 of model, from PDB: 1DUL) combined with a modeled structure (model of SRP RNA obtained from the SRP Database), also deviates significantly from a perfect double helix. This in agreement with sequence and experimental data suggesting that the secondary structure of E. coli SRP RNA contains several bulges and loops. To compare this model to the reconstruction in A, the ec_SRP RNA structure was filtered to 12Å (left, orange) and shown as a hard surface of the mass of the structure. This representation is similar to that used for the reconstruction in A. The filtered structure has ridges that correspond well with the repetitive features of the tail of the SRP reconstruction. C) The structure of a perfect helix of RNA (right, purple, PDB: 1RNA) was repeated five times to generate an RNA molecule of comparable size to the RNA in ec_SRP. When this structure is filtered to 12Å (left, orange) a very distinctive continuous, ribbon-like structure is evident which is not seen in the SRP reconstruction nor in the RNA model. Thus, the hard shell image of the reconstruction has features of SRP RNA rather than features of a continuous ribbon of helical RNA, as expected.
- Supplementary Figure 4. - Details of the process used to construct the high-resolution model, ec_SRP. A) The SRP RNA molecule can be fit into the SRP reconstruction (orange wireframe) in, at least, three different ways. In these images the RNA is modeled as a blue rectangle of 130Å long consistent with the predicted length of the RNA based on the compaction observed for the crystallized fragment (SRP RNA, PDB: 1DUL). Left, the RNA could extend along the entire length of the tail; middle, the RNA could be compacted to fit within the head region; or, right, the RNA could occupy portions of both the head and tail. The latter model is not consistent with a particle composed of one RNA and one protein. Because STEM provides mass measurements we know that particles are monomers of SRP. Therefore, this model was rejected. This presentation highlights the remaining regions (non-RNA) that would need to be filled by Ffh, the protein component of E. coli SRP. B) To distinguish between the two reasonable models we used ESI detection of phosphorus atoms (blue) to localize the RNA molecule (dark blue on orange wireframe structure of SRP). The ESI data suggests that the RNA runs along the entire length of the SRP reconstruction. C) The low resolution of the 3-D phosphorus map allows for some latitude in the vertical position of the RNA component within the SRP reconstruction, which is displayed as error bars on the rectangular representation of the SRP RNA. D) The fragment of RNA and M domain were modeled into the reconstruction as outlined in the text. In addition, other positions were examined for the M-RNA domain structure in the SRP reconstruction. In the orientation shown, the particle is not thick enough for NG to overlap the RNA or M domain. Furthermore, the length of the NG domain structure (~76Å) means that it must lie in the plane of the reconstruction shown as other orientations are not compatible with the thickness of the particle. Thus for two docked positions of the M domain (green) and RNA (blue) the largest contiguous area of the SRP reconstruction that remains for the docking of the NG domain is indicated by a red ellipse (with approximate dimensions matching the NG structure from T. aquaticus). The placement of M-RNA, shown to the left, leaves a more visually satisfying fit for the NG region than the position shown to the right in which the M-RNA structure was translated vertically. E) The NG domain could be docked into the SRP reconstruction with: Left, the G region near the M domain and the N region above the RNA (referred to as G >N); or, right, the converse orientation, with the N region near the M domain (green) and the G region situated above the RNA (referred to as N>G). F) The distance between the carboxyl-terminus of the NG domain and the amino-terminus of the M domain was measured in relation to rotation around the long-axis of the NG domain to determine compatibility with an NG-M interdomain linker. Both N>G and G>N orientations from E were analyzed. Measurements were made from the alpha-carbons of the respective residues using InsightII (Accelrys Software Inc., San Diego, USA). The range of rotations for the N>G and G>N orientations, that are compatible with an NG-M interdomain linker (modeled from the S. solfataricus SRP54 structure (approximately 40Å)), are relatively constrained with about 60 degrees of rotational freedom in each case. G) A space-filling model of the NG domain (red) was fit into the remaining head volume of the SRP reconstruction for the NG rotations that were compatible with an interdomain linker, as determined in F. The resolution of our structure is not sufficient to distinguish the precise rotation of NG relative to M, for both the N>G and G>N orientations. Therefore, the rotation angle that gave a minimal linker distance was selected for further analysis. The N>G orientation was not compatible with the FRET data, due to many of the acceptor positions in Ffh being too close to the donor sites. The G>N orientation was compatible with the FRET data (see main text and Table 1). Left, the G>N orientation; and, right, the N>G orientation. The G>N orientation generated a more optimal visual fit. The NG domain is contained within the reconstruction and a more contiguous area is filled (arrowheads indicate regions of poor overlap for the N>G orientation). Also, the low density region in the head section (visualized as a hole, common to all later stage reconstructions) is preserved in this model (the display of a solid object on the wireframe makes it difficult to see in this figure therefore see Supplementary Figure 5). H) The final model (ec_SRP) of E. coli SRP, was constructed from available structures of the NG and M domains and SRP RNA fragments as described in the main text. The ec_SRP model has been docked into the more conservative 90 kDa SRP reconstruction (orange wireframe).
- Supplementary Figure 5. - To compare the ec_SRP model with the reconstruction, the model was filtered to different resolutions (7, 12, 19 and 26Å) and displayed as hard surface representations. These representations provide visual references to compare to our SRP reconstruction. Viewing thresholds were selected to represent volumes encompassing the approximate expected mass of E. coli SRP, 90 kDa (upper panel) and more cut-in volumes corresponding to 60 kDa (lower panel). Each volume is shown from two views, corresponding to the views of the reconstruction to the left of the panels: the number 9-like view; and, after a rotation of 90º about the long axis. The SRP reconstruction in has more fine structural detail than the 19Å reference from both views. The rotated view of the SRP reconstruction has features that are visible only in the 7Å reference but the number 9' view of the reconstruction appears to be of lower resolution. Even though there is noise in the reconstruction that is not present in the filtered models we conclude that the resolution of the SRP model that most closely approximates the reconstruction is the one filtered to 12Å resolution, a value consistent with the value obtained from the 3-sigma criteria by FSC analysis.
- Supplementary Figure 6. - Movement of G region relative to N upon FtsY binding - The structures solved for apo-Ffh NG from T. aquaticus (PDB: 1FFH) and T. aquaticus Ffh NG bound to the NG of FtsY and GMPCP (PDB: 1OKK) differ significantly in the relative orientations of the N and G regions. We aligned the Ffh NG information from the FtsY-bound structure (blue) onto our model, ec_SRP, to visualize what effect this movement may have. Our docking studies suggest that the tetraloop of the 4.5S RNA could be acting as an anchor point for the N region. Thus, as a result of the interdomain movement that occurs when FtsY binds Ffh the G region moves away from the M domain. This movement would further open the putative signal peptide binding site in SRP and we speculate that this movement may facilitate release of signal sequence from SRP upon binding of SRP to its receptor. We are grateful to D. M. Freymann for suggesting this type of analysis.