Structure Analysis of the Flagellar Cap Filament Complex by Electron Cryomicroscopy and Single-Particle Image Analysis

Transcription

1 Journal of Structural Biology 133, (2001) doi: /jsbi , available online at on Structure Analysis of the Flagellar Cap Filament Complex by Electron Cryomicroscopy and Single-Particle Image Analysis Koji Yonekura,* Saori Maki-Yonekura,* and Keiichi Namba*, *Protonic NanoMachine Project, ERATO, JST, 3-4 Hikaridai, Seika Japan; and Advanced Technology Research Laboratories, Matsushita Electric Industrial Co., Ltd., 3-4 Hikaridai, Seika Japan Received December 27, 2000, and in revised form February 22, 2001; published online June 8, 2001 The cap of the bacterial flagellum plays an essential role in the growth of the long helical filament by promoting the efficient self-assembly of flagellin transported to the distal end through the narrow central channel of the flagellum. The structure of the cap filament complex was analyzed by electron cryomicroscopy and single-particle image analysis to understand how the cap stays attached while allowing the flagellin insertion between the cap and the filament end and also allowing the HAP proteins to pass through. In the images of the complex, the projection pattern of the helical subunit array in the filament portion occupied the major fraction but was variable depending on the azimuthal orientation of the filament; therefore the images showed a strong tendency to be misaligned. Various methods had to be newly developed to correctly align the images by overcoming this misalignment problem. The structure thus obtained clearly demonstrated the pentameric structure of the cap and how the cap operates. The new methods of analysis presented here would be generally applicable to cap structures of various filaments that play biologically important roles in cellular activities Academic Press Key Words: electron cryomicroscopy; single-particle image analysis; bacterial flagellum; flagellin; selfassembly; flagellar cap; HAP2; FliD; Salmonella. INTRODUCTION Bacteria move in liquid environments by rotating the flagellar filaments of a helical form. The rotary motor at the base of the filament drives the rotation of the filament (Berg and Anderson, 1973; Silverman and Simon, 1974). In E. coli and Salmonella, the motor structure called the flagellar basal body crosses both the cytoplasmic and outer membranes and continues as an extracellular structure called the hook and the filament. The assembly process starts with FliF ring formation in the cytoplasmic membrane (Suzuki et al., 1978). A dedicated export apparatus, homologous to the type III protein export system (Kubori et al., 1998), is believed to be integrated at the cytoplasmic opening of the FliF ring channel and to export selectively a set of flagellar proteins into the channel of the flagellum using the energy of ATP hydrolysis (Fan et al., 1997; Minamino and Macnab, 1999). The flagellar proteins travel through the channel of the growing structure to the distal end, where the assembly occurs (Iino, 1969; Emerson et al., 1970). The filament is only about 200 Å in diameter but grows to a length of around 15 m by self-assembly of as many as flagellin subunits. The cap at the distal end of the flagellum is essential for its growth, remaining stably attached while permitting the flagellin insertion and binding between the cap and the filament, thereby preventing flagellin monomers from simply diffusing away. The cap is made of a single protein, HAP2 (hook-associated protein 2, also called FliD). HAP2 from Salmonella typhimurium is composed of 466 amino acid residues, and its molecular mass is 49.8 kda. The cap was thought to be a pentameric complex of HAP2, because HAP2 was found to form a bipolar pair of pentamers in solution. However, there was neither structural evidence that the cap is the pentamer nor any clues to the mechanism of the cap binding and promotion of the flagellin self-assembly. For the filament structure, which has a well-defined helical symmetry, electron cryomicroscopy and helical image reconstruction were used to produce density maps at around 10-Å resolution (Mimori et al., 1995; Morgan et al., 1995; Mimori-Kiyosue et al., 1996, 1997, 1998). However, because the structure of the cap filament complex has no symmetry, singleparticle image analysis was the only means capable of visualizing its structure. What made this structure analysis difficult was twofold. First, there is an obvious limitation in collecting a sufficiently large number of images to produce meaningful three-di /01 $35.00 Copyright 2001 by Academic Press All rights of reproduction in any form reserved. 246

2 STRUCTURE ANALYSIS OF FLAGELLAR CAP FILAMENT COMPLEX 247 mensional structures because the target objects are localized at the ends of long filaments. The probability of finding good images of the cap attached to the distal end of the filament in electron micrographs was low, because the long filament portions occupy a large area of the micrographs, making the number density of the distal ends very small. Second, because the cap occupies a relatively small volume of the cap filament complex, the strong feature of the helical symmetry in the filament portion caused misalignment in the conventional two-dimensional alignment procedure for the three-dimensional image reconstruction. The image of the cap filament complex was dominated by the projection pattern of the helical array of flagellin subunits, but the pattern was variable depending on the azimuthal orientation of the filament. Therefore, twodimensional cross-correlation used for the conventional alignment method failed to produce properly aligned particle images. Here we describe new methods that we developed for the structure analysis of the cap filament complex. These include a sample preparation procedure for its efficient data collection and a set of particle alignment procedures for reliable three-dimensional image reconstruction of a complex structure composed of a small nonhelical particle and a large helical structure. MATERIALS AND METHODS Sample preparation. The reconstituted cap filament complex was prepared as previously described (Maki et al., 1998) with some improvements in the efficiency of the reconstitution. First, relatively short flagellar filaments were reconstituted using flagellin from S. typhimurium, strain SJW1655, which forms the R-type straight filament. The filaments at a protein concentration of 1 mg/ml were incubated with 0.1 mg/ml HAP2 in 10 mm sodium phosphate (ph 7.0) overnight at 37 C. Then, the filaments were spun down by centrifugation at g for 10 min and resuspended in 100 mm NaCl, 2 mm MgCl 2, and 20 mm Tris HCl (ph 7.8). The centrifugation and resuspension were repeated twice to remove excess HAP2. The efficiency of the reconstitution was checked by electron microscopy of negatively stained samples. The fraction of the capped filament reached up to 60% of the total filaments. Electron cryomicroscopy. The solution of the capped filaments was applied to a holey carbon grid and rapidly frozen in liquid ethane. Then, they were examined using a JEM3000SFF electron microscope (Jeol) with a field emission gun operated at 300 kv and a specimen temperature of 4 40 K. Images were recorded on SO-163 film (Eastman Kodak Co.) at a magnification of The electron dose was 25 e/å 2. Image analysis. Images of the capped filaments in electron micrographs were selected first by eye, and then, preservation of the helical symmetry in the filament portion was checked by optical diffraction. Selected images were digitized with a Leaf- Scan 45 linear CCD densitometer (Leaf Systems) at a step size of 10 m and the image data were reduced by a factor of 2. In total, 610 images were digitized and processed. Most of the image analysis was carried out using the SPIDER/WEB package (Frank et al., 1996), a modified version of the MRC helical package (Toyoshima and Unwin, 1990; Toyoshima, 2000), and a set of programs that we developed for this work. The Brandeis helical package (Owen et al., 1996) was used for straightening the filament images and extracting layer-line data. The overall procedure of the data processing and structure analysis is shown in Fig. 1, and more detailed descriptions are given below, where the names of newly developed programs and scripts are written in boldface capital letters. The amounts of defocus were determined from the Fourier transform of carbon films near the cap filament images (Tani et al., 1996) and used to correct for the reversal of the phases caused by the contrast transfer function. Only images recorded at defocus levels of Å were used for analysis because those recorded at smaller defocus levels showed too low a contrast to process. The filament images were straightened using a spline function. The positions of the caps were marked by eye and the images of the capped filament were cut out so as to contain the filament portion with a length of 440 Å, which corresponds to approximately one repeat of the R-type straight flagellar filament (Mimori et al., 1995). After high-contrast noises, such as contaminated ice that was occasionally observed, were masked out, the image densities were scaled and adjusted to a reference image. Two-dimensional alignment of the images was carried out in a stepwise manner. First, the helical lattice pattern of the filament portion was smeared out by running an average along an approximate filament axis ( z axis) (MRCPRJ). The in-plane rotation angle was determined by cross-correlation between this runningaveraged image and a reference image. A manually averaged image was used as the initial reference. The angle obtained was used to align the image to the reference. Then, the image corrected for the in-plane rotation was projected along the filament axis (projection x) by MRCPRJ. The reference for the x-shift adjustment was made by averaging the rotation-corrected images and centering them in the x-direction (PRJCEN). By calculating the one-dimensional cross-correlation, the amount of x-shift was determined (PRJCOR) and applied to align the filament axis position. Finally, the images corrected for the in-plane rotation and x-shift were projected perpendicular to the filament axis (projection z) (MRCPRJ). The amount of z-shift was obtained from the minimum density position in projection z, which corresponds to the outer edge of the cap, and the image position was adjusted for the z-shift (PRJYMIN). In each step, images that showed a relatively large translation or angle of rotation were excluded from the average to produce the reference in the next step. Using the average thus obtained as the new reference, the two-dimensional alignment procedure was iterated five times. After the fifth cycle, the number of included images was 589. A script program, 2DALIGN.CSH, carried out all the procedures needed in the two-dimensional alignment (Fig. 1). The azimuthal orientation was determined using the helical symmetry of the filament portion. From the Fourier transform of the images aligned in two dimensions as described above, the structure factors of the four strongest layer lines (n 11, n 5, n 6, and n 1) were extracted. The axial spacing of the n 1 layer line is 1/25.6 Å 1. To determine the out-of-plane tilt, the three highest peaks were automatically picked from each layer line by SRCHAIDLL. The out-of-plane tilt and additional x-shift were determined by SRCH (DeRosier and Moore, 1970) and the layer-line structure factors were corrected for these values (HOMEGAX). The amounts of these corrections were mostly within 10 and less than a few angstroms. Data sets that had the out-of-plane tilt and additional x-shift larger than these values were discarded at this point. Then, the azimuthal orientation of each particle was determined by finding the minimum phase residual of the layer-line structure factors in the interparticle fitting procedure using HLXFIT (Toyoshima and Unwin, 1990).

3 248 YONEKURA, MAKI-YONEKURA, AND NAMBA RESULTS FIG. 1. Protocol for the data processing and structure analysis of the cap filament complex. Names of newly developed programs and scripts are written in boldface capital letters. In this step, the z-shift was always set to 0. First, the azimuthal orientation was roughly estimated using the layer line of n 1 alone, and then, it was refined using all four layer lines. At the same time, the phase residual for the opposite polarity of the filament was calculated to check the image quality. The phase residuals for the wrong polarity were worse than the correct one from 20 to 40, and if the difference was smaller than 5, the image was excluded from the average. Using the averaged layerline data set as the new reference, the procedure to determine the azimuthal orientation was iterated five times (HLXFITLL.CSH). Finally, by using the azimuthal orientation thus determined, a three-dimensional density map of the cap filament complex was reconstructed by back-projection with a low-pass filter. The number of images included was 425. Sample Preparation for Efficient Data Collection The collection of many good images of the cap filament complex was extremely time-consuming because the number density of the cap attached to the filament end in each film was limited by the long filaments, to each of which only one cap is attached. There were also other factors that further reduced the probability of finding good images. For some unknown reason, the efficiency of the cap filament reconstitution was originally 30 40%. Also, the distal ends of the filaments could not always be found in ice regions within holes of holey carbon grids. In order to increase the probability of finding good images as much as possible, the cap filament complex was reconstituted using a relatively short filament, the length of which was as long as a few thousand angstroms. Also by carefully searching for the optimal conditions for the reconstitution, such as the salt concentration, protein concentration, and temperature, the fraction of capped filaments reached as high as 60%. In particular, the reconstitution at 37 C had a prominent effect. At this temperature, uncapped filaments slowly depolymerize into flagellin monomers, suggesting that frequent binding and unbinding of flagellin to the distal end occur. The binding of HAP2 was greatly facilitated under this condition compared to that at room temperature where the filament structure is stable, suggesting that some specific subunit arrangement at the distal end of the filament is necessary to reconstitute the complete cap structure. Most of the uncapped filaments must have depolymerized after overnight incubation at 37 C. The uncapped filaments we find in electron micrographs are probably partially capped filaments with one or two HAP2 molecules bound, which may not be visible but would make the filament structure more stable. Two-Dimensional Alignment of the Cap Filament Images A few typical examples of raw images are presented in Fig. 2a, which shows the low contrast and high level of noise typical of cryomicrographs. By conventional two-dimensional alignment, the cap position and filament axis were misaligned, because the projection pattern of the helical lattice of the filament dominates the two-dimensional cross-correlation, and the lattice pattern relative to the filament axis is variable depending on the azimuthal orientation of the filament. In Figs. 2b and 2c, averaged images with image alignment carried out by eye and the conventional alignment method are shown, respectively. A strong feature of the checker flag, obviously coming from the helical lattice of

4 STRUCTURE ANALYSIS OF FLAGELLAR CAP FILAMENT COMPLEX 249 FIG. 2. Electron micrographs of the cap filament complex and images produced by different methods of alignment and averaging. (a) Electron cryomicrographs of the cap filament complex. The thin, flat plate of the cap is barely seen at the top of the filament, which shows the helical array of flagellin subunits. (b) Averaged image produced by manual alignment of individual images. The images of the cap and the tubular structure of the filament are enhanced, but the helical lattice of flagellin subunits is still visible, indicating that the accuracy of the image alignment is not sufficient. (c) Averaged image produced by the conventional method. The helical lattice of flagellin subunits was strongly enhanced and there is no sign of the cap image, indicating that the images were completely misaligned. (d) Average of 589 images using our new alignment method. The cap plate and the tubular structures of the filament are clearly visible, without any sign of the helical lattice of the filament. The tubular structures within the filament are labeled according to the domain label of flagellin: D0 (the inner tube); D1 (the outer tube); and D2 D3 (the outer domains). The density of the filament portion is axially continuous because the image is equivalent to the result of cylindrical averaging. The weak density of the outer domains is caused by the cylindrical average of well-separated domains. Bar represents the diameter of the filament, 230 Å. flagellin subunits, appeared by the conventional method (Fig. 2c). Ideally, however, what we expect to see in this averaged image is the projection of the cylindrically averaged density distribution, which is supposed to show continuous cylindrical tubes with a flat plate at its end. Therefore, the result obtained by the conventional alignment procedure is the consequence of total misalignment. To avoid such a strong influence of the helical lattice on the alignment process, we devised a new alignment method (Fig. 1). For the in-plane rotation alignment, a running average along the approximate filament axis was used to smear out the helical lattice feature. A one-dimensional projection in each axis was used for the translation alignment in the direction perpendicular to the projection axis to avoid the influence of the helical lattice feature and to enhance the boundary of the cap filament complex by increasing the signal-to-noise ratio. Examples of one-dimensional projections are shown in Fig. 3: along the z-axis (projection x, Fig. 3a) and along the x-axis (projection z, Fig. 3b). In projection x of an averaged image, the density profile shows the mirror symmetry as expected from the axial projection of the cylindrically averaged filament structure. There is a deep trough at the center and the density profile on each side has a pair of relatively high peaks, together forming a wide peak with a small dent at the top of the peak and a small peak in the outer region. The central trough corresponds to the central channel of the filament. The two high peaks correspond to the inner tube (domain D0) and the outer tube (D1), which form the filament core, and the small peak is the outer domains (D2 and D3) of the filament (Mimori et al., 1995). These features were identified in projection x of individual images relatively clearly, which made possible the reliable alignment of the filament axis by cross-correlation of these profiles. In projection z, there is one deep trough in the density profile, which corresponds to just outside of the edge of the cap plate at the distal end of the filament. This feature can also be identified in individual profiles of projection z as well as in the averaged profile. By adjusting the position of the sharp trough, the z-position of the cap filament image was accurately aligned. In total, 589 images were aligned and averaged in two dimensions, which produced an image that clearly showed density features expected for the two-dimensional projection of the cylindrically averaged filament structure. The features expected in the averaged images are the continuous density distributions along the filament axis in the inner- and outer-tube region and the outer domain regions, and these features were all clearly reproduced (Fig. 2d). The inner- and outer-tube regions extended over radial ranges from 15 to 30 Å (D0) and from 35 to 60 Å (D1), respectively, and the outer domain extended from 60 to 115 Å (D2 and D3), values that are quantitatively consistent with those obtained previously by helical image reconstruction (Mimori et al., 1995; Morgan et al., 1995). The averaged image also showed the thin plate of the cap and the cavity underneath clearly. Determination of the Azimuthal Orientation For three-dimensional reconstruction, the azimuthal orientation of the cap filament complex

5 250 YONEKURA, MAKI-YONEKURA, AND NAMBA FIG. 3. One-dimensional projection profiles used for image alignment. (a) Projection x and (b) projection z of averaged image (top) and individual images (middle and bottom) of the cap filament complex. Arrows in (b) indicate the distal end of the cap filament complex. around its filament axis must be determined for individual images. The helical symmetry of the filament portion was used for this purpose. First, the structure factors along the four strongest layer lines were extracted from the Fourier transform of each image. Correction for the out-of-plane tilt was needed to obtain reliable azimuthal orientations of individual particles by fitting to a reference. In the conventional procedure, the strong layer lines had to be selected and picked up manually to determine the out-of-plane tilt (Toyoshima, 2000). The processing of about 600 image data sets by the conventional method would take a long time. To facilitate the process of this step, a new program, SRCHAIDLL, was developed to pick up strong peaks automatically. After each data set was corrected for the outof-plane tilt, the azimuthal orientations were determined by finding the minimum in the interparticle phase residual. Because the cap filament complex itself does not have the helical symmetry and its z-position was already fixed, we are supposed to find only one minimum in the phase residual profile around the 360 azimuthal rotation. However, for each of layer lines having the helical start number n with n 1, the phase residual profile repeats n times and there are n equivalent answers to the minimum search. If the two strongest layer lines, e.g., n 5 and n 6, were used for the phase residual calculation, the second deepest minimum would be nearly as deep as the true minimum, because the minimum from the layer line of n 5 and that from n 6 found near this second deepest minimum are only 12 (72 60 ) apart, making the minimum in the joint phase residual profile nearly as deep as the true one. Amplitude-weighted phase residuals calculated from all four strongest layer lines would be strongly influenced by these two layer lines as well. Therefore, only the layer line of n 1 was used first for a rough estimate of the azimuthal orientation, and then it was refined by using all four layer lines. In Fig. 4a the distribution of the azimuthal orientations thus obtained for all the images

6 STRUCTURE ANALYSIS OF FLAGELLAR CAP FILAMENT COMPLEX 251 Three-Dimensional Image Reconstruction Before proceeding to the three-dimensional image reconstruction step, the quality of each image was further checked by the difference of the two phase residuals obtained by fitting the image to a reference image oriented in the two opposite polarities. If the difference in the phase residual was smaller than a certain preset value (i.e., 5 ), the image quality was judged insufficient and the image was rejected. By this screening, 164 images were rejected and the remaining 425 images were used for three-dimensional image reconstruction by back-projection. The three-dimensional density map obtained is shown in Fig. 5 in a solid surface representation (adapted from Yonekura et al., 2000). The density map shows the expected helical array of the outer domains of flagellin subunits in the filament portion, as in the previous maps produced by helical image reconstruction (Mimori et al., 1995; Morgan et al., 1995), indicating the quality and reliability of the density distribution. The array of these outer domains could be traced along the one-start helix of the filament, a very shallow right-handed helix with a pitch of about 26 Å containing approximately 11 subunits per two turns. The lid-like plate or cap at the top was FIG. 4. Distribution of the azimuthal orientation and Fourier shell correlation. (a) Distribution of the azimuthal orientation of 425 images used to reconstruct the three-dimensional density map shown in Fig. 5. (b) FSC coefficient showing pairwise phase comparisons obtained in the image reconstruction. The FSC coefficient decayed to 0.5 at a resolution of 27 Å (dotted lines). is presented. This homogeneous distribution without any particular segregation ensured a reliable three-dimensional image reconstruction by backprojection. FIG. 5. Three-dimensional density map of the cap filament complex reconstructed from 425 images. (a) End-on view from the top, showing a pentagonal shape of the plate domain of the cap. (b) Side view, showing a regular helical array of flagellin subunits and the plate domain of the cap at the top end of the filament. Arrowhead shows an inverted L-shaped gap that is likely to be the site of assembly for a newly arriving flagellin subunit (Yonekura et al., 2000). Bar represents the diameter of the filament, 230 Å.

7 252 YONEKURA, MAKI-YONEKURA, AND NAMBA 120 Å wide and 25 Å thick; the diameter was roughly the same as that of the outer tube of the filament (domain D1). Its pentagonal shape in the end-on view (Fig. 5a) indicates that the cap is a pentamer of HAP2, half of the decamer complex formed in solution (Maki et al., 1998). To estimate the resolution of the map, we calculated the Fourier shell correlation (FSC) coefficients between two maps obtained by image reconstruction from two independent groups of images that were randomly chosen (Frank, 1996). The resolution determined with the FSC crossing a level of 0.5 was 27 Å (Fig. 4b). The detailed descriptions of the cap filament structure and its implications for the rotary cap mechanism by which the self-assembly of flagellin is efficiently promoted are given by Yonekura et al. (2000). DISCUSSION New Developments To visualize the site of the flagellin assembly between the cap and the filament at the distal end of the bacterial flagellum, we developed a set of new methods for three-dimensional image reconstruction of the cap filament complex. Many difficulties arose in the structure analysis, mostly because the cap is a relatively small object compared to the filament to which it is intimately attached, and also the cap and filament have completely different structural symmetries. We developed new methods to overcome these difficulties. The newly developed methods are as follows: a specimen preparation procedure that allows efficient reconstitution of the complexes with short filaments to facilitate rapid data collection; a set of two-dimensional alignment procedures that completely suppresses the strong but undesirable influence of the helical lattice of the filament portion; and procedures for automatic layer-line peak search and accurate determination of the azimuthal orientation of the whole particle for rapid processing of a number of images. In particular, the problem in the two-dimensional alignment step required a lot of effort to solve. In terms of the structure analysis, the flagellar hook basal body complex presents similar problems. In this case, the position of the basal body in projection can be aligned by itself because the basal body has a relatively large diameter compared to the hook and its axial extension is large enough to treat it as an independent particle (Francis et al., 1994). After conventional two-dimensional image alignment of the basal body, the helical symmetry of the straight hook portion can be used to determine its azimuthal orientation (Thomas, 1999). In contrast, the distinctly visible portion of the flagellar cap attached to the distal end of the filament is a thin plate having a thickness of about 25 Å and a diameter of about 120 Å, which is only half of the filament diameter. This was too small to align by conventional two-dimensional alignment methods. Therefore, we used the running average along the filament axis for in-plane rotation alignment and one-dimensional projections along or perpendicular to the filament axis for the translation alignment in each of the two orthogonal directions. This new method worked very effectively and produced an averaged image with characteristics theoretically predicted for the result (Fig. 2d). In single-particle image analysis, one-dimensional projection is sometimes used to determine the relative orientation between two-dimensional class averages, which is known as the Sinogram method (Frank, 1996). Our method uses a few elements of the Sinogram method. Additional devices, such as the automatic search for the layer-line peaks and the quick search for the phase residual minimum to determine approximate azimuthal orientation followed by its refinement, efficiently facilitated the whole process of data processing and analysis. Together with a help of highquality image collection carried out by using a JEM- 3000SFF electron microscope with its field emission gun and a specimen temperature of 4 K, we were able to determine the positions and orientations of individual images precisely. The three-dimensional density map of the cap filament complex thus obtained from a relatively small number of images showed a sufficiently good quality and presented interesting and important implications for the functions and movements of the flagellar cap at the growing end of the filament (Yonekura et al., 2000). General Applicability The methods developed here can be immediately applied for structural analysis of the following parts of the bacterial flagellum: the junction between the hook and the filament, where hook-associated protein 1 (HAP1) and 3 (HAP3) form a short segment that connects the two filamentous structures with distinct mechanical properties (Homma et al., 1984); the distal end of the hook with its cap formed by a protein, FlgD, during the growth of the hook (Ohnishi et al., 1994); the proximal end of the flagellum detached from the cell, where part of the rod is still attached (Okino et al., 1989). The methods can also be applied to any other filament cap complex, such as the actin filament with its various capping protein complexes, which play many important roles in the cellular activities of eukaryotes (Cooper and Schafer, 2000). Although the molecular mass of the pentameric HAP2 cap is slightly less than 250 kda, which is close to the lowest limit of usual single-particle image analysis, the density map of the cap filament complex showed clear structural

8 STRUCTURE ANALYSIS OF FLAGELLAR CAP FILAMENT COMPLEX 253 features that were interpretable in terms of the cap function. This implies that we would be able to visualize the structures of even smaller molecules when they are attached to the end of filamentous structures, by taking advantage of the large size of the filament portion as the base for the image data collection and alignment. This is similar to the visualization of antibody structures attached to coat proteins of icosahedral virus particles (Baker et al., 1999), except that we would not be able to take full advantage of the symmetry of the base structure in the structure analysis because the attached molecular complexes and the filaments would have different symmetries. Still, our new methods would greatly expand the power of single-particle image analysis to a wide variety of the filament cap complexes that play important roles in the formation of the cytoskeleton, in cellular morphogenesis, and even in signal transduction. We thank D. G. Morgan, T. R. Shaikh, N. Francis, and D. J. DeRosier for their help in the initial stage of the image analysis, D. Thomas for helpful comments in the development of the new image alignment procedures, F. Vonderviszt for helpful comments in sample preparation, K. Murata and Y. Fujiyoshi for technical advice in electron cryomicroscopy, and T. Nitta and F. Oosawa for continuous support and encouragement. REFERENCES Baker, T. S., Olson, N. H., and Fuller, S. D. (1999) Adding the third dimension to virus life cycles: Three-dimensional reconstruction of icosahedral viruses from cryo-electron micrographs, Microbiol. Mol. Biol. Rev. 63, Berg, H. C., and Anderson, R. A. (1973) Bacteria swim by rotating their flagellar filaments, Nature 245, Cooper, J. A., and Schafer, D. A. (2000) Control of actin assembly and disassembly at filament ends, Curr. Opin. Cell Biol. 12, DeRosier, D. J., and Moore, P. B. (1970) Reconstruction of threedimensional images from electron micrographs of structures with helical symmetry, J. Mol. Biol. 52, Emerson, S. U., Tokuyasu, K., and Simon, M. I. (1970) Bacterial flagella: Polarity of elongation, Science 169, Fan, F., Ohnishi, K., Francis, N. R., and Macnab, R. M. (1997) The FliP and FliR proteins of Salmonella typhimurium, putative components of the type III flagellar export apparatus, are located in the flagellar basal body, Mol. Microbiol. 26, Francis, N. R., Sosinsky, G. E., Thomas, D., and DeRosier, D. J. (1994) Isolation, characterization and structure of bacterial flagellar motors containing the switch complex, J. Mol. Biol. 235, Frank, J. (1996) Three-Dimensional Electron Microscopy of Macromolecular Assemblies, Academic Press, San Diego. Frank, J., Radermacher, M., Penczek, P., Zhu, J., Li, Y., Ladjadj, M., and Leith, A. (1996) SPIDER and WEB: Processing and visualization of images in 3D electron microscopy and related fields, J. Struct. Biol. 116, Homma, M., Kutsukake, K., Iino, T., and Yamaguchi, S. (1984) Hook-associated proteins essential for flagellar filament formation in Salmonella typhimurium, J. Bacteriol. 157, Iino, T. (1969) Polarity of flagellar growth in Salmonella, J. Gen. Microbiol. 56, Kubori, T., Matsushima, Y., Nakamura, D., Uralil, J., Lara-Tajero, M., Sukhan, A., Galán, J. E., and Aizawa, S.-I. (1998) Supramolecular structure of the Salmonella typhimurium type III protein secretion system, Science 280, Maki, S., Vonderviszt, F., Furukawa, Y., Imada, K., and Namba, K. (1998) Plugging interactions of HAP2 pentamer into the distal end of flagellar filament revealed by electron microscopy, J. Mol. Biol. 277, Mimori, Y., Yamashita, I., Murata, K., Fujiyoshi, Y., Yonekura, K., Toyoshima, C., and Namba, K. (1995) The structure of the R-type straight flagellar filament of Salmonella at 9 Å resolution by electron cryomicroscopy, J. Mol. Biol. 249, Mimori-Kiyosue, Y., Vonderviszt, F., Yamashita, I., Fujiyoshi, Y., and Namba, K. (1996) Direct interaction of flagellin termini essential for polymorphic ability of flagellar filament, Proc. Natl. Acad. Sci. USA 93, Mimori-Kiyosue, Y., Vonderviszt, F., and Namba, K. (1997) Locations of terminal segments of flagellin in the filament structure and their roles in polymerization and polymorphism, J. Mol. Biol. 270, Mimori-Kiyosue, Y., Yamashita, I., Fujiyoshi, Y., Yamaguchi, S., and Namba, K. (1998) Role of the outermost subdomain of Salmonella flagellin in the filament structure revealed by electron cryomicroscopy, J. Mol. Biol. 284, Minamino, T., and Macnab, R. M. (1999) Components of the Salmonella flagellar export apparatus and classification of export substrates, J. Bacteriol. 181, Morgan, D. G., Owen, C., Melanson, L. A., and DeRosier, D. J. (1995) Structure of bacterial flagellar filaments at 11 Å resolution: Packing of the a-helices, J. Mol. Biol. 249, Ohnishi, K., Ohta, Y., Aizawa, S.-I., Macnab, R. M., and Iino, T. (1994) FlgD is a scaffolding protein needed for flagellar hook assembly in Salmonella typhimurium, J. Bacteriol. 176, Okino, H., Isomura, M., Yamaguchi, S., Magariyama, Y., Kudo, S., and Aizawa, S.-I. (1989) Release of flagellar filament hook rod complex by a Salmonella typhimurium mutant defective in the M-ring of the basal body, J. Bacteriol. 171, Owen, C. H., Morgan, D. G., and DeRosier, D. J. (1996) Image analysis of helical objects: The Brandeis Helical Package, J. Struct. Biol. 116, Silverman, M., and Simon, M. (1974) Flagellar rotation and the mechanism of bacterial motility, Nature 249, Suzuki, T., Iino, T., Horiguchi, T., and Yamaguchi, S. (1978) Incomplete flagellar structures in nonflagellate mutants of Salmonella typhimurium, J. Bacteriol. 33, Tani, K., Sasabe, H., and Toyoshima, C. (1996) A set of programs for determining defocus and astigmatism in electron images, Ultramicroscopy 65, Thomas, D. (1999) Structural analysis of the bacterial flagellar rotary motor by electron cryomicroscopy, Ph.D. thesis, Brandeis University. Toyoshima, C. (2000) Structure determination of tubular crystals of membrane proteins. I. Indexing of diffraction patterns, Ultramicroscopy 84, Toyoshima, C., and Unwin, N. (1990) Three dimensional structure of the acetylcholine receptor by cryoelectron microscopy and helical image reconstruction, J. Cell Biol. 111, Yonekura, K., Maki, S., Morgan, D. G., DeRosier, D. J., Vonderviszt, F., Imada, K., and Namba, K. (2000) The bacterial flagellar cap as the rotary promoter of flagellin self-assembly, Science 290,