- Email: [email protected]
Three-Dimensional Reconstruction of the Mammalian Centriole from Cryoelectron Micrographs: The Use of Common Lines for Orientation and Alignment
JOURNAL OF STRUCTURAL BIOLOGY ARTICLE NO. SB973922
120, 320–328 (1997)
Three-Dimensional Reconstruction of the Mammalian Centriole from Cryoelectron Micrographs: The Use of Common Lines for Orientation and Alignment John Kenney1 Structural Biology Programme, European Molecular Biology Laboratory, Meyerhofstrasse 1, Postfach 10.22.09, 69017 Heidelberg Federal Republic of Germany
Eric Karsenti Cell Biology Programme, European Molecular Biology Laboratory, Meyerhofstrasse 1, Postfach 10.22.09, 69017 Heidelberg, Federal Republic of Germany
and Brent Gowen and Stephen D. Fuller2 Structural Biology Programme, European Molecular Biology Laboratory, Meyerhofstrasse 1, Postfach 10.22.09, 69017 Heidelberg, Federal Republic of Germany Received September 17, 1997
intriguingly complex and includes an internal cylindrical feature which is a site of g tubulin localization. r 1997 Academic Press Key Words: centriole; centrosome; cryoelectron microscopy; microtubule nucleation; g-tubulin; three-dimensional reconstruction.
The microtubule organizing center of the animal cell (S. D. Fuller et al., 1992, Curr. Opin. Struct. Biol. 2, 264–274; D. M. Glover et al., 1993, Sci. Am. 268, 62–68; E. B. Wilson, 1925), (The Cell in Development and Heredity) comprises two centrioles and the pericentriolar material. We have completed several three-dimensional reconstructions of individual centrioles from tilt series of cryoelectron micrographs. The reconstruction procedure uses minimization of the common lines residual to define the orientation of the centriolar ninefold symmetry axis and then uses this symmetry to generate a structure by weighted backprojection to 28-nm resolution. Many of the features of these reconstructions agree with previous, conventional transmission electron microscopy studies (M. Paintrand et al., 1992, J. Struct. Biol. 108, 107–128). The microtubule barrel of the centriole is roughly 500 nm long and 300 nm in diameter and the microtubule bundles appear to taper toward the distal end. In addition, we see a handedness to the pericentriolar material at the base (distal end) of the centriole which is opposite to the skew of the microtubule triplets. The region at which the microtubule barrel joins this base is
INTRODUCTION
The centrosome serves as the organizing center for the microtubule network in most animal cells (Fuller et al., 1992; Glover et al., 1993; Wilson, 1925). Two centrioles are found within this enigmatic structure. Surrounding these ninefold microtubule barrels is the pericentriolar material from which the network of cellular microtubules grow. A large number of studies have described the morphology of the centriole on the basis of conventional electron microscopy of sectioned, fixed, and stained material. These studies inform a consensus model for the structure. The canonical centriole in animal cells contains a barrel formed by nine triplets of interlocking microtubules. The length of the barrel varies in different cell types from 0.1 to 0.7 µm. The centriole has a polarity which is defined by comparison with the related structure, the basal body (Anderson, 1972). The distal end of the basal body is defined as the one from which a cilia will grow. The distal end of the centriole is the site of a number of elaborations and
1 Present address: Department of Crystallography, Birkbeck College, University of London, Malet Street, London, WC1E 7HX, UK. 2 To whom correspondence should be addressed. Fax: 149-6221387-306. E-mail: [email protected]
1047-8477/97 $25.00 Copyright r 1997 by Academic Press All rights of reproduction in any form reserved.
320
THREE-DIMENSIONAL RECONSTRUCTION OF THE CENTRIOLE
satellites (Lange and Gull, 1996), as well as the site the first growth of microtubules. It also bears a fibrous structure which connects the two centrioles of a centrosome and is believed necessary for centriolar replication (Paintrand et al., 1992; Tournier et al., 1991a, b). The microtubule triplets of the mammalian centriole taper to two microtubules at the distal end (Paintrand et al., 1992). Centrosome division is linked to the cell cycle. The two centrioles of the centrosome separate and each forms a new centriole from the side of its centriolar barrel. The pair of centrioles in the centrosome are often at right angles with the axis of the daughter intersecting that of the older, mother centriole (Kochanski and Borisy, 1990; Rieder and Borisy, 1982). Growth of the daughter begins from the proximal end which usually remains closest to the mother (Kochanski and Borisy, 1990). Here we complement the existing body of structural work on the centriole with cryoelectron microscopy and three-dimensional reconstruction. Cryoelectron microscopy has important advantages for the study of a fragile structure such as the centriole. By preserving the structure in a layer of vitrified water and avoiding the use of stains for contrast, cryoelectron microscopy eliminates the artifacts of dehydration and the staining which afflict conventional approaches (Adrian et al., 1984). This is particularly important for the centriole since staining and dehydration are well known to destroy the structure and antigenicity of the microtubule, a major component of the centriole (Gowen et al., 1995; Wade and Chre´tien, 1993). These advantages of cryoelectron microscopy come with a significant cost. The microscopy itself is more difficult since the vitrified specimen is extremely radiation sensitive and must be imaged under low dose conditions. Further, the lack of contrasting agents such as stains results in a complex image which is difficult to interpret due the superposition of structural features. Three-dimensional reconstruction is required to extract an interpretable structure from these complex images. We have performed several reconstructions of centrioles from tilt series of low dose cryoelectron micrographs. Our reconstruction procedure uses the ninefold symmetry of the centriole to overcome the problems of dose sensitivity and limitted tilt angle. We first identify the orientation of the ninefold symmetry axis using a common lines approach and then use this axis to extend a tilt series of limited range to the full circle and allow complete reconstruction by weighted backprojection. This use of symmetry allows a ninefold reduction in the dose used to image the structure and hence make the reconstruction possible without significant radiation damage. Despite the successful use of symmetry, the resolu-
321
tion of the resulting reconstructions was limited by technical problems to 28 nm. Our reconstructions agree with the consensus model of the centriole derived from conventional methods in overall features. The reconstruction also reveals organization in the pericentriolar material and a complementary handedness to the distal base of the structure and the microtubule barrel. Details of the structure of the region of contact between the barrel and the base are also revealed. We have previously shown that this region is a site of gtubulin localization in the core of the centriole (Fuller et al., 1995; Gowen et al., 1994, 1995). MATERIALS AND METHODS Isolation of Centrioles Centrioles were isolated from KE37 lymphocytes as described previously (Bornens et al., 1987; Buendia, 1992, No. 322; Gowen, 1995, No. 4; Verde, 1992, No. 447) and concentrated by centrifugation onto a sucrose cushion (70%) (Chre´tien et al., in preparation) and fractionation in 60-µl fractions. The peak fraction which contained roughly half of the starting material with a fourfold higher concentration was used for microscopy. Centrioles were briefly incubated with tubulin as described previously (Chre´tien et al., 1995) before vitrification. Microscopy Specimens were prepared on holey carbon films supported on 400-mesh copper grids. A 5-µl droplet of the concentrated centriolar solution was mixed with 1 µl of 1 nm colloidal gold and applied to the grid for 1 min. The sucrose was then washed from the sample by floating the grid for 20 s on a 1-ml drop of 10 mM K Pipes, pH 7.2. This washing procedure (Stewart et al., 1991) was repeated twice and the grid was blotted and vitrified as described previously (Adrian et al., 1984; Stewart et al., 1991). Microscopy was performed using a GATAN cryoholder operated at 2168°C using a Philips EM400 microscope operated at 80 kV or a JEOL2000EX microscope operated at 200 kV. The sample was allowed to equilibrate in the microscope for 60 min before images were taken to allow the anticontaminator to be withdrawn for tilting. Images were taken with a fluence of 0.3 electrons/Å2 per image. To allow comparison of the earlier and later images of the series, the tilt angle was alternated between high and low tilt on successive images (i.e., 0°, 220°, 22°, 24°, 6°, . . . rather than 225°, 223°, 21°, . . .) so that the final separation between the images of the series was 2°–4°. Images were taken on Kodak SO-163 film and processed for 12 min in full strength D19. Processing Images were screened for the absence of drift and astigmatism and for a similar orientation of the centriole and tilt axes. Selected series were scanned on a Perkin Elmer MicroD 1010GM at a pixel size of 50 µm. The positions of the colloidal gold particles was found in each image using SPIDER and then used to define the position of the tilt axis and the absolute value of the tilt angle by using a least squares fitting program (S.D.F., unpublished). Series which contained images for which the gold positions could not be refined consistently were discarded since this inconsistency reflects movement or distortion. The position of the symmetry axis of the centriole was first found by eye. It was then refined by minimizing the cross common lines residual for the entire series using a modified version of the program SIMPLEX (Fuller et al., 1996) in which the tilt angles of the members of the series were
322
KENNEY ET AL
held fixed and the origin (X, Y ), the declination (Q), and the rotation relative in the plane (V) were optimized. The refined orientation angles were used to calculate the necessary a, b, g angles for the SPIDER backprojection operator with generalized weighting (BP GW) (Radermacher, 1988) for the original image orientations and the eight symmetry related orientations for each view. The reliability of the reconstruction was determined by comparison of two reconstructions made from half data sets using Fourier shell correlation (FSC) (van Heel, 1987) and the final reconstruction was filtered to this resolution, typically 28 nm. Reconstructions were viewed with the application visualization system using the standard modules as well as some self-written ones (Sheehan et al., 1996) for normalization and cropping of the map. Immunocytochemistry Labeling with antibodies to a and to g tubulin was performed by the freeze substitution and postsectioning fixation technique described previously (Fuller et al., 1995; Gowen et al., 1994, 1995) using 10 nm colloidal gold coupled to protein A to visualize the positions of the rabbit IgGs. Control experiments showed that the labeling was specific (Fuller et al., 1995) and that both g and a tubulin could be labeled throughout the centrosome (Fuller et al., 1995; Gowen et al., 1995). Microtubule trajectories were traced across the serial sections to identify the section in which a microtubule terminated. RESULTS AND DISCUSSION
Quality of Data The centrosome preparation used for this work contained a large fraction (,40%) of paired centrioles surrounded by pericentriolar material. This preparation is active for both microtubule nucleation (Chre´tien et al., 1995) and for centriolar duplication (Bornens et al., 1987; Buendia et al., 1992; Tournier et al., 1991a). A number of technical problems confronted us during the data collection. These included the need to increase the centrosome concentration above that of our standard preparation for convenient microscopy. Simple centrifugation and resuspension caused distortion of the centrioles and hence the use of a sucrose cushion was essential. Washing the grid atop a droplet of buffer was then necessary to remove the sucrose (60%, w/v) of the cushion which causes problems for the microscopy of the vitrified sample. Apparently, the centrosomes adhere to the air water interface and are retained through the washing procedure which removes the majority of the sucrose as described previously for viruses (Stewart et al., 1991). The imaging conditions were a compromise between a desire to avoid beam damage and the need to record a usable image of the 250-nm-thick specimen. A number of tests were used to define the limits of beam dose for this specimen. Two tests were particularly helpful. The first was the consistency of the positions of the gold particles throughout the tilt series as an assay for distortion of the specimen. The second was the occurrence of a characteristic doublet of reflections in the optical diffraction pattern of the
centriole at 130 Å. Both of these showed signs of beam damage at doses above 20 e/Å2. We chose therefore to work at 0.3 e/Å2, which allowed us to take up to 20 images with a total dose of 6 e/Å2. This dose allowed us to take images at a magnification of 36250 and achieve an OD near 1.0 in the centriole itself although the surrounding features were heavily overexposed. A dose of 6 e/Å2 has been used for centriole grown microtubules imaged at higher magnification and was shown to preserve the 40-Å meridional reflection (Chre´tien et al., 1996). The images shown in Fig. 1 are typical of the data used for reconstruction. The image shows a mother centriole (vertical) and a budding daughter (to the right of the mother) in the centrosome with associated pericentriolar material. Microtubules were present on the pericentriolar material but are not seen in this image because they are heavily overexposed. Since our centriolar preparation was derived from logarithmically growing cells, all of the stages of the cell cycle should be represented. A consequence of this is that the morphology of the daughter centriole was quite variable and hence we focused our attention on the maternal member of the pair. The success rate for tilt series was extremely low; however 48 series (typically 620°) were obtained which met the criteria of completeness, lack of movement, appropriate defocus (0.8–1.0 µm underfocus), alignment of the centriole within 20° of the tilt axis, and the absence of interfering material in any of the images. Essentially all of the usable series were of centrosomes immediately adjacent to the edge of the hole of the carbon film. Inconsistencies in the first sets of reconstructions led us to consider the possibility that the centrioles were distorted by the thickness of the water layer. Our images were of centrioles embedded in ice of thickness between 3000 and 6000 Å. The clearest images were always obtained from the thinner layers of water; however, further analysis revealed distortion. Figure 2 shows a plot of the optical density of the image of the centriole relative to that of the background (which is inversely proportional to logarithm of the water layer thickness) and the variation in centriole barrel diameter throughout the tilt series. Figure 2 reveals that embedding in thinner layers of water distorted the centriolar barrel. This distortion, which rose to nearly 45% for the thinnest layers, invalidated our central assumption of ninefold symmetry for the reconstruction and rendered such high contrast tilt series from thin specimens useless for reconstruction by the common lines approach or indeed any approach based on the ninefold symmetry of the structure. The final reconstructions were made from the lower contrast im-
THREE-DIMENSIONAL RECONSTRUCTION OF THE CENTRIOLE
323
FIG. 1. Images of a tilt series. Eight members of the 12-member tilt series which was used to generate the reconstruction depicted in Figs. 3 and 4 are shown. The tilt angles and the order of the image in the series are: (a) 0°, image 1; (b) 235°, image 9; (c) 220°, image 8; (d) 212.5°, image 2; (e) 112.5°, image 11; (f ) 120°, image 5; (g) 135°, image 6; (h) 0°, Image 12. The pair of centrioles is arranged so that the older mature centriole is vertical with its distal end is toward the top of the figure and its proximal end at the bottom. The distal end of this centriole shows appendages with a budding procentriole to its right near the proximal end. The diameter of the mature centriole is ,300 nm. The scale bar represents 500 nm.
ages of centrioles embedded in thicker (.5000 Å) water layers. Model Calculations Our reconstruction procedure is based on the identity of common lines between the members of the
FIG. 2. Flattening effect. A plot is shown of the ratio of percentage flatness (1003 microtubule barrel diameter/diameter of the barrel in the most highly tilted image) versus log of the exposure difference (OD of film for a hole in the specimen-OD of the film for an area immediately adjacent to centriole) for the fifteen tilt series which showed the least distortion. The biphasic nature of the relationship indicates that flattening is more pronounced for thinner water layers.
tilt series. Common lines are a well established method for defining the orientation of the symmetry axes in projections of highly symmetric structures (Crowther, 1971; Crowther et al., 1970a, b; Fuller et al., 1996). The Fourier transform of a single projection of an icosahedral virus contains 37 pairs of common lines due to the high (532) symmetry and hence provides sufficient constraints to define the orientation from a single projection (Fuller et al., 1996). While a single projection of a ninefold object such as the centriole contains only four pairs of common lines, there are nine pairs of cross common lines between any two members of a tilt series and hence 9 3 (N(N 1 1)/2) pairs between the members of an N member tilt series. This large number of common lines, in the series as a whole, allows the position of the tilt axis to be determined with high accuracy. Once the position of the symmetry axis has been determined, it is used to generate the tilt parameters of the nine equivalent positions for each image of the tilt series which allows a reconstruction to be performed without problems such as a missing cone of data. We simulated images of ninefold symmetric barrel structures with signal to noise ratios lower than those expected in our actual images and found that the reconstruction method was reliable and effective for such model data. The novelty of a reconstruction method which
324
KENNEY ET AL
combines symmetry and back projection led us to perform model calculations to understand the effects of inaccuracies and distortions on the reconstruction process. The difference in defocus between the top and bottom of a 0.25-µm object would impose a 3-nm limit on the resolution (Mancini et al., 1996). A Monte Carlo analysis using the measured errors in the positions of the colloidal gold markers was performed to determine their effect on the determination of the tilt axis position. This showed that the axis position could be determined with an accuracy of 1.5–2.0% depending on the number of gold particles used for determination. This limits the final resolution of the reconstruction to ,8 nm. Model studies showed that the symmetry axis position could be determined by ‘‘eye’’ to an accuracy of 3°–5° and that this was improved to 1° when the ninefold symmetric common lines were used to refine its orientation. This inaccuracy would limit the resolution of the reconstruction to 15 nm. The distortion of the barrel was particularly deleterious since it affects not only the averaging of the data in the reconstruction but also our ability to find the orienta-
tion of the symmetry axis using the common lines procedures. Distortions of as small as 10% limited the resolution of the final reconstruction to 20 nm. Greater distortions caused the use of common lines to be untenable as seen by the inability to refine the axis to an orientation near to the one seen by eye. Together, these factors limit the resolution that we could expect from our reconstruction to worse than 20 nm. We attempted to correct the distortion due to flattening by modeling it as a nonisotropic shrinking and reconstructing with different sampling in the directions perpendicular and parallel to the water layer. This resulted in no improvement and supports the impression gained from individual images that the distortion may be accompanied by twisting of the microtubule barrel. Reconstructions from Cryoelectron Microscopic Data Three reconstructions were completed of maternal centrioles using 12- to 16-member tilt series of centrioles which showed less than 10% distortion. The cross common lines residual could be refined to
FIG. 3. Reconstruction Surface views. Two stereo triplets showing surface views of one of the reconstructions at a density level corresponding to two standard deviations above background. The three domains of the centriole are clearly visible as is the characteristic opposite skew of the microtubule triplets and the structures of the base. The distal end is to the right in both sets.
THREE-DIMENSIONAL RECONSTRUCTION OF THE CENTRIOLE
1° precision. The reconstructions were initially calculated to a resolution of 20 nm. They were evaluated for reliability by FSC (van Heel, 1987) between reconstructions calculated from the two halves of the tilt series and shown to be reliable to ,28 nm. The
325
correspondence between the three reconstructions was much poorer than their individual reliability as would be expected from differences in the extent of pericentriolar material and barrel length between the three centrioles used for reconstruction.
FIG. 4. Cross sections. Cross-sections perpendicular to the microtubule barrel axis of the centriole reconstruction shown in Fig. 2. The position of each section is shown on the surface representation. The numbers denote the position of each section in the entire (128) 3 volume and go from 1 (proximal) to 128 (distal) in steps of 6 nm. The surface view is tilted 1° from the vertical. The sections shown are 128 by 6 nm across.
326
KENNEY ET AL
Surface views of the reconstructions revealed common features which are represented at two standard deviations above the background level. The centriole seen in Fig. 3 is 480 nm in length from its bulbous distal base to the proximal tip of the microtubule barrel. The structure divides logically into three domains: the microtubule barrel, the junction between the barrel, and the base and the base itself. The staves of the centriolar barrel always showed a comma-like cross section of dimensions 30 by 90 nm as expected for three adjacent microtubules of the 28 nm diameter seen in cryoelectron micrographs. The staves always tapered to 30 nm at the proximal end. Sections (Fig. 4) show the changes in the structure as one proceeds from proximal to distal. The microtubule barrel has an outer diameter of ,300 nm (s(section) 234) and can be resolved for a length of 300 nm (s75, s80). The distal base of the structure displayed the same ninefold symmetry as the barrel but consistently showed a twist opposite to that of the comma-like bundles of microtubules. The base itself has a maximum width of 480 nm (s75, s80, s87) which is somewhat less than the extent of the pericentriolar material seen in individual images. We believe that this reflects the retention of ninefold symmetry only within the region of the pericentriolar material closest to the base. Difference images between the projections of the reconstruction and the original images supported this interpretation. The junction between the base and the barrel occurs at 228 nm (s68) from the proximal end and is a very complex structure. A central, smooth, cylinder crosses this region (s50, s68, s75) and closes to a solid structure within the base (s80, s87). The effect of distortions on these features was examined. The three centriole reconstructions did
not show a helical twist to the barrel (,5° in the most twisted example). Reconstructions of centrioles with distortions greater than 10% showed pronounced twist to the barrel; however, we know that our reconstruction method is not reliable for such distorted structures. In reconstructions of distorted centrioles, the internal cylinder is also present although its appearance changes. Several other features, such as the fenestrations in the periphery of the base, are visible in all of the reconstructions; however, their placement is not consistent between reconstructions. Labeling of the Centriole The link between low resolution structure and function requires localization of biochemically defined components. We have done this for two of the tubulin isoforms: a tubulin, which forms microtubules with b tubulin, and g tubulin, which is believed to be involved in microtubule nucleation (Raff et al., 1993; Stearns et al., 1991). As expected, a tubulin was found in the pericentriolar material and the staves of the microtubule barrel (Fuller et al., 1995; Gowen et al., 1995). It is not a component of the internal structures. The centrosome contains two sites of g tubulin localization. The first is the pericentriolar material where previous work by preembedding techniques had shown its presence (Stearns et al., 1991; Stearns and Kirschner, 1994). The second localization could be seen only by the use of a freeze substitution method in combination with postsectioning fixation (Fuller et al., 1995) which we have developed (Gowen et al., 1995) and corresponds to the internal cylindrical structure (Fuller et al., 1995) described above (Fig. 5). The four sections shown all cross the centriolar barrel. The three (Figs. 5a, 5b,
FIG. 5. g-Tubulin labeling of putative microtubule nucleation sites. Serial sectioning was performed on MDCK cells which had been freeze substituted and labeled with anti-g tubulin following the postsectioning fixation procedure described previously (Fuller et al., 1995; Gowen et al., 1995). The labeling of the central cylinder can also be seen in (a), (b), and (d) which include the distal end of the centriole. Notice that the microtubule barrel shows a smaller diameter (,250 nm) than is seen in the cryoelectron micrographs (Fig. 1) or the reconstruction (Fig. 4). The positions of microtubule ends near the centriole were determined from the series of sections which spanned each centriole. The sections shown are representative of those which show putative microtubule nucleation sites. The arrowheads mark sites of g-tubulin labeling visualized with protein A conjugated to 10 nm gold. The arrows mark the path of a microtubule which comes closest to the g tubulin label and which terminates within the section shown. All of the microtubule ends are near to sites of g tubulin labeling. Further, the trajectory was tangential to the cluster of g tubulin label. The scale bar represents 500 nm.
THREE-DIMENSIONAL RECONSTRUCTION OF THE CENTRIOLE
and 5d), which include the distal end of the barrel, show g tubulin labeling within the core of the structure. The section which passes through the proximal end of the barrel shows no g tubulin labeling in the center although labeling is seen the pericentriolar material. The centriole appears smaller in these sections than it is in cryoelectron micrographs such as those of Fig. 1. This difference is not a consequence of flattening since we have shown that the centrioles in Fig. 1 are not appreciably flattened (Fig. 2). The centrosome imposes organization on the microtubule network of the cell. The handed and polar structure which we see in our reconstructions can transfer such an organization to the microtubules which it nucleates only if the sites of microtubule nucleation are also organized in a polar manner. Moritz et al. (1995a, b) have presented reconstructions of Drosophila centrosomes in which preembedding labeling with anti-g tubulin revealed that ringlike structures are the sites of clusters of anti-gtubulin labeling. Two-thirds of the g tubulin labeling was found at the minus ends of microtubules. We found that our postsectioning fixation and labeling technique produced a similar image of the relationship between g tubulin and microtubule ends. We explored this by examining 25 series of sections of g tubulin labeled MDCK cells. All of the microtubule ends which we could identify in these serial sections were near to sites of g tubulin labeling as expected from the work on Drosophila centrosomes (Moritz et al., 1995a). Our results contrast with those on the Drosophila system in that we see many fewer microtubules per centriole. This reflects the difference between centrioles in MDCK cells in situ and isolated Drosophila centrioles which had been mixed with tubulin and allowed to grow microtubules in vitro. As a consequence much of our g tubulin labeling is not microtubule associated. The trajectories of the microtubules near the centrosome could be defined in 12 cases and were tangential to the sites of g tubulin labeling (Fig. 5). A tangential interaction of the microtubules would allow the transfer of organization and polarity from the complex centriolar structure to the microtubule network. The authors are pleased to acknowledge our generous and patient colleagues who made the completion of this work possible. We thank Marek Cyrklaff (EMBL) for help with some of the early microscopy and Dr. Michael Radermacher (Max-Planck-Institut fu¨r Biophysik, Frankfurt, F. R. Germany) for many helpful discussions of generalized weighted backprojection. We are indebted to Dr. Denis Chre´tien (EMBL) for very helpful discussions on the nature of microtubule nucleation and the polarity of the centriole and for his comments on the manuscript. We are also grateful to Dr. Michelle Moritz and Dr. David Agard (UCSF) for their comments on our labeling studies.
327
REFERENCES Adrian, M., Dubochet, J., Lepault, J., and McDowall, A. W. (1984) Cryo-electron microscopy of viruses, Nature 308, 32–36. Anderson, R. G. W. (1972) The three-dimensional structure of the basal body from the rhesus monkey oviduct, J. Cell Biol 54, 246–265. Bornens, M., Paintrand, M., Berges, J., Marty, M. C., and Karsenti, E. (1987) Structural and chemical characterization of isolated centrosomes, Cell Motil. Cytoskel. 8, 238–249. Buendia, B., Draetta, G., and Karsenti, E. (1992) Regulation of the microtubule nucleating activity of centrosomes in Xenopus egg extracts: Role of cyclin A-associated protein kinase, J. Cell. Biol. 116, 1431–1442. Chre´tien, D., Buendia, B., Fuller, S., and Karsenti, E. (1997) Reconstruction of centrosome cycle from cryo-electron micrographs. J. Struct. Biol., in press. Chre´tien, D., Fuller, S. D., and Karsenti, E. (1995) Structure of growing microtubule ends: Two-dimensional sheets close into tubes at variable rates, J. Cell Biol. 129, 1311–1328. Chre´tien, D., Kenney, J. M., Fuller, S. D., and Wade, R. H. (1996) Determination of microtubule polarity by cryo-electron microscopy, Structure 4, 1031–1040. Crowther, R. A. (1971) Procedures for three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs, Phil. Trans. R. Soc. Lond. B 261, 221–230. Crowther, R. A., Amos, L. A., Finch, J. T., De Rosier, D. J., and Klug, A. (1970a) Three-dimensional reconstructions of spherical viruses by Fourier synthesis from electron micrographs, Nature 226, 421–423. Crowther, R. A., DeRosier, D. J., and Klug, A. (1970b) The reconstruction of a three-dimensional structure from projections and its application to electron microscopy, Proc. Roy. Soc. Lond. A 317, 319–340. Fuller, S. D., Butcher, S. J., Cheng, R. H., and Baker, T. S. (1996) Three-dimensional reconstruction of icosahedral particles: The uncommon line, J. Struct. Biol. 116, 48–55. Fuller, S. D., Gowen, B. E., Buendia, B., and Karsenti, E. (1995) The core of the mammalian centriole contains g-tubulin, Curr. Biol. 5, 1384–1393. Fuller, S. D., Kenney, J. M., and Karsenti, E. (1992) Centrosomes, Curr. Opin. Struct. Biol. 2, 264–274. Glover, D. M., Gonzalez, C., and Raff, J. W. (1993) The centrosome, Sci. Am. 268, 62–68. Gowen, B. E., Buendia, B., Karsenti, E., and Fuller, S. D. (1995) Postembedding a-tubulin labelling of isolated centrosomes, Histochem. J. 27, 240–246. Gowen, B. E., Buendia, B., Reinsch, S., Kenney, J. M., Karsenti, E., and Fuller, S. D. (1994) The centriole: Inside and out, in 13th International Congress of Electron Microscopy: Applications in Biological Sciences, Vol. 3A, pp. 265–266. Les Editions de Physique Les Ulis, Paris. Kochanski, R. S., and Borisy, G. G. (1990) Mode of centriole duplication and distribution, J. Cell. Biol. 110, 1599–1605. Lange, B. M. H., and Gull, K. (1996) Structure and function of the centriole in animal cells: Progress and questions, Trends Cell Biol. 6, 348–352. Mancini, E. J., de Haas, F., and Fuller, S. D. (1997) High resolution icosahedral reconstruction: Fulfilling the promise of cryo-electron microscopy, Structure, 5, 741–750. Moritz, M., Braunfeld, M., Sedat, J., Alberts, B., and Agard, D. (1995a) Microtubule nucleation by g-tubulin containing rings in the centrosome, Nature 378, 638–640. Moritz, M., Braunfeld, M. B., Fung, J. C., Sedat, J. W., Alberts, B.
328
KENNEY ET AL
M., and Agard, D. A. (1995b) Three-dimensional structural characterization of centrosomes from early Drosophila embryos, J. Cell Biol. 130, 1149–1159. Paintrand, M., Moudjou, M., Delacroix, H., and Bornens, M. (1992) Centrosome organization and centriole architecture: Their sensitivity to divalent cations, J. Struct. Biol. 108, 107–128. Radermacher, M. (1988) Three-dimensional reconstruction of single particles from random and non-random tilt series, J. Electron Microsc. Tech. 9, 359–394. Raff, J. W., Kellogg, D. R., and Alberts, B. M. (1993) Drosophila gamma-tubulin is part of a complex containing two previously identified centrosomal MAPs, J. Cell. Biol. 121, 823–835. Rieder, C. L., and Borisy, G. G. (1982) The centrosome cycle in PTK2 cells: Assymetric distribution and structural changes in the pericentriolar material, J. Cell. Biol. 44, 117–132. Sheehan, B., Fuller, S. D., Pique, M. E., and Yeager, M. (1996) AVS software for visualization in molecular microscopy, J. Struct. Biol. 116, 99–105. Stearns, T., Evans, L., and Kirschner, M. (1991) Gamma-tubulin is a highly conserved component of the centrosome, Cell 65, 825–836. Stearns, T., and Kirschner, M. (1994) In vitro reconstitution of
centrosome assembly and function: The central role of g tubulin, Cell 76, 623–637. Stewart, P. L., Burnett, R. M., Cyrklaff, M., and Fuller, S. D. (1991) Image reconstruction reveals the complex molecular organization of adenovirus, Cell 67, 145–154. Tournier, F., Cyrklaff, M., Karsenti, E., and Bornens, M. (1991a) Centrosomes competent for parthenogenesis in Xenopus eggs support procentriole budding in cell-free extracts, Proc. Natl. Acad. Sci. USA 88, 9929–9933. Tournier, F., Komesli, S., Paintrand, M., Job, D., and Bornens, M. (1991b) The intercentriolar linkage is critical for the ability of heterologous centrosomes to induce parthenogenesis in Xenopus, J. Cell. Biol. 113, 1361–1369. van Heel, M. (1987) Similarity measures between images, Ultramicroscopy 21, 95–100. Verde, F., Dogterom, M., Stelzer, E., Karsenti, E., and Leibler, S. (1992) Control of microtubule dynamics and length by cyclin Aand cyclin B-dependent kinases in Xenopus egg extracts. J. Cell. Biol. 118, 1097–1108. Wade, R. H., and Chre´tien, D. (1993) Cryo-electron microscopy of microtubules. J. Struct. Biol. 110, 1–27. Wilson, E. B. (1925) The Cell in Development and Heredity, MacMillan, New York.
120, 320–328 (1997)
Three-Dimensional Reconstruction of the Mammalian Centriole from Cryoelectron Micrographs: The Use of Common Lines for Orientation and Alignment John Kenney1 Structural Biology Programme, European Molecular Biology Laboratory, Meyerhofstrasse 1, Postfach 10.22.09, 69017 Heidelberg Federal Republic of Germany
Eric Karsenti Cell Biology Programme, European Molecular Biology Laboratory, Meyerhofstrasse 1, Postfach 10.22.09, 69017 Heidelberg, Federal Republic of Germany
and Brent Gowen and Stephen D. Fuller2 Structural Biology Programme, European Molecular Biology Laboratory, Meyerhofstrasse 1, Postfach 10.22.09, 69017 Heidelberg, Federal Republic of Germany Received September 17, 1997
intriguingly complex and includes an internal cylindrical feature which is a site of g tubulin localization. r 1997 Academic Press Key Words: centriole; centrosome; cryoelectron microscopy; microtubule nucleation; g-tubulin; three-dimensional reconstruction.
The microtubule organizing center of the animal cell (S. D. Fuller et al., 1992, Curr. Opin. Struct. Biol. 2, 264–274; D. M. Glover et al., 1993, Sci. Am. 268, 62–68; E. B. Wilson, 1925), (The Cell in Development and Heredity) comprises two centrioles and the pericentriolar material. We have completed several three-dimensional reconstructions of individual centrioles from tilt series of cryoelectron micrographs. The reconstruction procedure uses minimization of the common lines residual to define the orientation of the centriolar ninefold symmetry axis and then uses this symmetry to generate a structure by weighted backprojection to 28-nm resolution. Many of the features of these reconstructions agree with previous, conventional transmission electron microscopy studies (M. Paintrand et al., 1992, J. Struct. Biol. 108, 107–128). The microtubule barrel of the centriole is roughly 500 nm long and 300 nm in diameter and the microtubule bundles appear to taper toward the distal end. In addition, we see a handedness to the pericentriolar material at the base (distal end) of the centriole which is opposite to the skew of the microtubule triplets. The region at which the microtubule barrel joins this base is
INTRODUCTION
The centrosome serves as the organizing center for the microtubule network in most animal cells (Fuller et al., 1992; Glover et al., 1993; Wilson, 1925). Two centrioles are found within this enigmatic structure. Surrounding these ninefold microtubule barrels is the pericentriolar material from which the network of cellular microtubules grow. A large number of studies have described the morphology of the centriole on the basis of conventional electron microscopy of sectioned, fixed, and stained material. These studies inform a consensus model for the structure. The canonical centriole in animal cells contains a barrel formed by nine triplets of interlocking microtubules. The length of the barrel varies in different cell types from 0.1 to 0.7 µm. The centriole has a polarity which is defined by comparison with the related structure, the basal body (Anderson, 1972). The distal end of the basal body is defined as the one from which a cilia will grow. The distal end of the centriole is the site of a number of elaborations and
1 Present address: Department of Crystallography, Birkbeck College, University of London, Malet Street, London, WC1E 7HX, UK. 2 To whom correspondence should be addressed. Fax: 149-6221387-306. E-mail: [email protected]
1047-8477/97 $25.00 Copyright r 1997 by Academic Press All rights of reproduction in any form reserved.
320
THREE-DIMENSIONAL RECONSTRUCTION OF THE CENTRIOLE
satellites (Lange and Gull, 1996), as well as the site the first growth of microtubules. It also bears a fibrous structure which connects the two centrioles of a centrosome and is believed necessary for centriolar replication (Paintrand et al., 1992; Tournier et al., 1991a, b). The microtubule triplets of the mammalian centriole taper to two microtubules at the distal end (Paintrand et al., 1992). Centrosome division is linked to the cell cycle. The two centrioles of the centrosome separate and each forms a new centriole from the side of its centriolar barrel. The pair of centrioles in the centrosome are often at right angles with the axis of the daughter intersecting that of the older, mother centriole (Kochanski and Borisy, 1990; Rieder and Borisy, 1982). Growth of the daughter begins from the proximal end which usually remains closest to the mother (Kochanski and Borisy, 1990). Here we complement the existing body of structural work on the centriole with cryoelectron microscopy and three-dimensional reconstruction. Cryoelectron microscopy has important advantages for the study of a fragile structure such as the centriole. By preserving the structure in a layer of vitrified water and avoiding the use of stains for contrast, cryoelectron microscopy eliminates the artifacts of dehydration and the staining which afflict conventional approaches (Adrian et al., 1984). This is particularly important for the centriole since staining and dehydration are well known to destroy the structure and antigenicity of the microtubule, a major component of the centriole (Gowen et al., 1995; Wade and Chre´tien, 1993). These advantages of cryoelectron microscopy come with a significant cost. The microscopy itself is more difficult since the vitrified specimen is extremely radiation sensitive and must be imaged under low dose conditions. Further, the lack of contrasting agents such as stains results in a complex image which is difficult to interpret due the superposition of structural features. Three-dimensional reconstruction is required to extract an interpretable structure from these complex images. We have performed several reconstructions of centrioles from tilt series of low dose cryoelectron micrographs. Our reconstruction procedure uses the ninefold symmetry of the centriole to overcome the problems of dose sensitivity and limitted tilt angle. We first identify the orientation of the ninefold symmetry axis using a common lines approach and then use this axis to extend a tilt series of limited range to the full circle and allow complete reconstruction by weighted backprojection. This use of symmetry allows a ninefold reduction in the dose used to image the structure and hence make the reconstruction possible without significant radiation damage. Despite the successful use of symmetry, the resolu-
321
tion of the resulting reconstructions was limited by technical problems to 28 nm. Our reconstructions agree with the consensus model of the centriole derived from conventional methods in overall features. The reconstruction also reveals organization in the pericentriolar material and a complementary handedness to the distal base of the structure and the microtubule barrel. Details of the structure of the region of contact between the barrel and the base are also revealed. We have previously shown that this region is a site of gtubulin localization in the core of the centriole (Fuller et al., 1995; Gowen et al., 1994, 1995). MATERIALS AND METHODS Isolation of Centrioles Centrioles were isolated from KE37 lymphocytes as described previously (Bornens et al., 1987; Buendia, 1992, No. 322; Gowen, 1995, No. 4; Verde, 1992, No. 447) and concentrated by centrifugation onto a sucrose cushion (70%) (Chre´tien et al., in preparation) and fractionation in 60-µl fractions. The peak fraction which contained roughly half of the starting material with a fourfold higher concentration was used for microscopy. Centrioles were briefly incubated with tubulin as described previously (Chre´tien et al., 1995) before vitrification. Microscopy Specimens were prepared on holey carbon films supported on 400-mesh copper grids. A 5-µl droplet of the concentrated centriolar solution was mixed with 1 µl of 1 nm colloidal gold and applied to the grid for 1 min. The sucrose was then washed from the sample by floating the grid for 20 s on a 1-ml drop of 10 mM K Pipes, pH 7.2. This washing procedure (Stewart et al., 1991) was repeated twice and the grid was blotted and vitrified as described previously (Adrian et al., 1984; Stewart et al., 1991). Microscopy was performed using a GATAN cryoholder operated at 2168°C using a Philips EM400 microscope operated at 80 kV or a JEOL2000EX microscope operated at 200 kV. The sample was allowed to equilibrate in the microscope for 60 min before images were taken to allow the anticontaminator to be withdrawn for tilting. Images were taken with a fluence of 0.3 electrons/Å2 per image. To allow comparison of the earlier and later images of the series, the tilt angle was alternated between high and low tilt on successive images (i.e., 0°, 220°, 22°, 24°, 6°, . . . rather than 225°, 223°, 21°, . . .) so that the final separation between the images of the series was 2°–4°. Images were taken on Kodak SO-163 film and processed for 12 min in full strength D19. Processing Images were screened for the absence of drift and astigmatism and for a similar orientation of the centriole and tilt axes. Selected series were scanned on a Perkin Elmer MicroD 1010GM at a pixel size of 50 µm. The positions of the colloidal gold particles was found in each image using SPIDER and then used to define the position of the tilt axis and the absolute value of the tilt angle by using a least squares fitting program (S.D.F., unpublished). Series which contained images for which the gold positions could not be refined consistently were discarded since this inconsistency reflects movement or distortion. The position of the symmetry axis of the centriole was first found by eye. It was then refined by minimizing the cross common lines residual for the entire series using a modified version of the program SIMPLEX (Fuller et al., 1996) in which the tilt angles of the members of the series were
322
KENNEY ET AL
held fixed and the origin (X, Y ), the declination (Q), and the rotation relative in the plane (V) were optimized. The refined orientation angles were used to calculate the necessary a, b, g angles for the SPIDER backprojection operator with generalized weighting (BP GW) (Radermacher, 1988) for the original image orientations and the eight symmetry related orientations for each view. The reliability of the reconstruction was determined by comparison of two reconstructions made from half data sets using Fourier shell correlation (FSC) (van Heel, 1987) and the final reconstruction was filtered to this resolution, typically 28 nm. Reconstructions were viewed with the application visualization system using the standard modules as well as some self-written ones (Sheehan et al., 1996) for normalization and cropping of the map. Immunocytochemistry Labeling with antibodies to a and to g tubulin was performed by the freeze substitution and postsectioning fixation technique described previously (Fuller et al., 1995; Gowen et al., 1994, 1995) using 10 nm colloidal gold coupled to protein A to visualize the positions of the rabbit IgGs. Control experiments showed that the labeling was specific (Fuller et al., 1995) and that both g and a tubulin could be labeled throughout the centrosome (Fuller et al., 1995; Gowen et al., 1995). Microtubule trajectories were traced across the serial sections to identify the section in which a microtubule terminated. RESULTS AND DISCUSSION
Quality of Data The centrosome preparation used for this work contained a large fraction (,40%) of paired centrioles surrounded by pericentriolar material. This preparation is active for both microtubule nucleation (Chre´tien et al., 1995) and for centriolar duplication (Bornens et al., 1987; Buendia et al., 1992; Tournier et al., 1991a). A number of technical problems confronted us during the data collection. These included the need to increase the centrosome concentration above that of our standard preparation for convenient microscopy. Simple centrifugation and resuspension caused distortion of the centrioles and hence the use of a sucrose cushion was essential. Washing the grid atop a droplet of buffer was then necessary to remove the sucrose (60%, w/v) of the cushion which causes problems for the microscopy of the vitrified sample. Apparently, the centrosomes adhere to the air water interface and are retained through the washing procedure which removes the majority of the sucrose as described previously for viruses (Stewart et al., 1991). The imaging conditions were a compromise between a desire to avoid beam damage and the need to record a usable image of the 250-nm-thick specimen. A number of tests were used to define the limits of beam dose for this specimen. Two tests were particularly helpful. The first was the consistency of the positions of the gold particles throughout the tilt series as an assay for distortion of the specimen. The second was the occurrence of a characteristic doublet of reflections in the optical diffraction pattern of the
centriole at 130 Å. Both of these showed signs of beam damage at doses above 20 e/Å2. We chose therefore to work at 0.3 e/Å2, which allowed us to take up to 20 images with a total dose of 6 e/Å2. This dose allowed us to take images at a magnification of 36250 and achieve an OD near 1.0 in the centriole itself although the surrounding features were heavily overexposed. A dose of 6 e/Å2 has been used for centriole grown microtubules imaged at higher magnification and was shown to preserve the 40-Å meridional reflection (Chre´tien et al., 1996). The images shown in Fig. 1 are typical of the data used for reconstruction. The image shows a mother centriole (vertical) and a budding daughter (to the right of the mother) in the centrosome with associated pericentriolar material. Microtubules were present on the pericentriolar material but are not seen in this image because they are heavily overexposed. Since our centriolar preparation was derived from logarithmically growing cells, all of the stages of the cell cycle should be represented. A consequence of this is that the morphology of the daughter centriole was quite variable and hence we focused our attention on the maternal member of the pair. The success rate for tilt series was extremely low; however 48 series (typically 620°) were obtained which met the criteria of completeness, lack of movement, appropriate defocus (0.8–1.0 µm underfocus), alignment of the centriole within 20° of the tilt axis, and the absence of interfering material in any of the images. Essentially all of the usable series were of centrosomes immediately adjacent to the edge of the hole of the carbon film. Inconsistencies in the first sets of reconstructions led us to consider the possibility that the centrioles were distorted by the thickness of the water layer. Our images were of centrioles embedded in ice of thickness between 3000 and 6000 Å. The clearest images were always obtained from the thinner layers of water; however, further analysis revealed distortion. Figure 2 shows a plot of the optical density of the image of the centriole relative to that of the background (which is inversely proportional to logarithm of the water layer thickness) and the variation in centriole barrel diameter throughout the tilt series. Figure 2 reveals that embedding in thinner layers of water distorted the centriolar barrel. This distortion, which rose to nearly 45% for the thinnest layers, invalidated our central assumption of ninefold symmetry for the reconstruction and rendered such high contrast tilt series from thin specimens useless for reconstruction by the common lines approach or indeed any approach based on the ninefold symmetry of the structure. The final reconstructions were made from the lower contrast im-
THREE-DIMENSIONAL RECONSTRUCTION OF THE CENTRIOLE
323
FIG. 1. Images of a tilt series. Eight members of the 12-member tilt series which was used to generate the reconstruction depicted in Figs. 3 and 4 are shown. The tilt angles and the order of the image in the series are: (a) 0°, image 1; (b) 235°, image 9; (c) 220°, image 8; (d) 212.5°, image 2; (e) 112.5°, image 11; (f ) 120°, image 5; (g) 135°, image 6; (h) 0°, Image 12. The pair of centrioles is arranged so that the older mature centriole is vertical with its distal end is toward the top of the figure and its proximal end at the bottom. The distal end of this centriole shows appendages with a budding procentriole to its right near the proximal end. The diameter of the mature centriole is ,300 nm. The scale bar represents 500 nm.
ages of centrioles embedded in thicker (.5000 Å) water layers. Model Calculations Our reconstruction procedure is based on the identity of common lines between the members of the
FIG. 2. Flattening effect. A plot is shown of the ratio of percentage flatness (1003 microtubule barrel diameter/diameter of the barrel in the most highly tilted image) versus log of the exposure difference (OD of film for a hole in the specimen-OD of the film for an area immediately adjacent to centriole) for the fifteen tilt series which showed the least distortion. The biphasic nature of the relationship indicates that flattening is more pronounced for thinner water layers.
tilt series. Common lines are a well established method for defining the orientation of the symmetry axes in projections of highly symmetric structures (Crowther, 1971; Crowther et al., 1970a, b; Fuller et al., 1996). The Fourier transform of a single projection of an icosahedral virus contains 37 pairs of common lines due to the high (532) symmetry and hence provides sufficient constraints to define the orientation from a single projection (Fuller et al., 1996). While a single projection of a ninefold object such as the centriole contains only four pairs of common lines, there are nine pairs of cross common lines between any two members of a tilt series and hence 9 3 (N(N 1 1)/2) pairs between the members of an N member tilt series. This large number of common lines, in the series as a whole, allows the position of the tilt axis to be determined with high accuracy. Once the position of the symmetry axis has been determined, it is used to generate the tilt parameters of the nine equivalent positions for each image of the tilt series which allows a reconstruction to be performed without problems such as a missing cone of data. We simulated images of ninefold symmetric barrel structures with signal to noise ratios lower than those expected in our actual images and found that the reconstruction method was reliable and effective for such model data. The novelty of a reconstruction method which
324
KENNEY ET AL
combines symmetry and back projection led us to perform model calculations to understand the effects of inaccuracies and distortions on the reconstruction process. The difference in defocus between the top and bottom of a 0.25-µm object would impose a 3-nm limit on the resolution (Mancini et al., 1996). A Monte Carlo analysis using the measured errors in the positions of the colloidal gold markers was performed to determine their effect on the determination of the tilt axis position. This showed that the axis position could be determined with an accuracy of 1.5–2.0% depending on the number of gold particles used for determination. This limits the final resolution of the reconstruction to ,8 nm. Model studies showed that the symmetry axis position could be determined by ‘‘eye’’ to an accuracy of 3°–5° and that this was improved to 1° when the ninefold symmetric common lines were used to refine its orientation. This inaccuracy would limit the resolution of the reconstruction to 15 nm. The distortion of the barrel was particularly deleterious since it affects not only the averaging of the data in the reconstruction but also our ability to find the orienta-
tion of the symmetry axis using the common lines procedures. Distortions of as small as 10% limited the resolution of the final reconstruction to 20 nm. Greater distortions caused the use of common lines to be untenable as seen by the inability to refine the axis to an orientation near to the one seen by eye. Together, these factors limit the resolution that we could expect from our reconstruction to worse than 20 nm. We attempted to correct the distortion due to flattening by modeling it as a nonisotropic shrinking and reconstructing with different sampling in the directions perpendicular and parallel to the water layer. This resulted in no improvement and supports the impression gained from individual images that the distortion may be accompanied by twisting of the microtubule barrel. Reconstructions from Cryoelectron Microscopic Data Three reconstructions were completed of maternal centrioles using 12- to 16-member tilt series of centrioles which showed less than 10% distortion. The cross common lines residual could be refined to
FIG. 3. Reconstruction Surface views. Two stereo triplets showing surface views of one of the reconstructions at a density level corresponding to two standard deviations above background. The three domains of the centriole are clearly visible as is the characteristic opposite skew of the microtubule triplets and the structures of the base. The distal end is to the right in both sets.
THREE-DIMENSIONAL RECONSTRUCTION OF THE CENTRIOLE
1° precision. The reconstructions were initially calculated to a resolution of 20 nm. They were evaluated for reliability by FSC (van Heel, 1987) between reconstructions calculated from the two halves of the tilt series and shown to be reliable to ,28 nm. The
325
correspondence between the three reconstructions was much poorer than their individual reliability as would be expected from differences in the extent of pericentriolar material and barrel length between the three centrioles used for reconstruction.
FIG. 4. Cross sections. Cross-sections perpendicular to the microtubule barrel axis of the centriole reconstruction shown in Fig. 2. The position of each section is shown on the surface representation. The numbers denote the position of each section in the entire (128) 3 volume and go from 1 (proximal) to 128 (distal) in steps of 6 nm. The surface view is tilted 1° from the vertical. The sections shown are 128 by 6 nm across.
326
KENNEY ET AL
Surface views of the reconstructions revealed common features which are represented at two standard deviations above the background level. The centriole seen in Fig. 3 is 480 nm in length from its bulbous distal base to the proximal tip of the microtubule barrel. The structure divides logically into three domains: the microtubule barrel, the junction between the barrel, and the base and the base itself. The staves of the centriolar barrel always showed a comma-like cross section of dimensions 30 by 90 nm as expected for three adjacent microtubules of the 28 nm diameter seen in cryoelectron micrographs. The staves always tapered to 30 nm at the proximal end. Sections (Fig. 4) show the changes in the structure as one proceeds from proximal to distal. The microtubule barrel has an outer diameter of ,300 nm (s(section) 234) and can be resolved for a length of 300 nm (s75, s80). The distal base of the structure displayed the same ninefold symmetry as the barrel but consistently showed a twist opposite to that of the comma-like bundles of microtubules. The base itself has a maximum width of 480 nm (s75, s80, s87) which is somewhat less than the extent of the pericentriolar material seen in individual images. We believe that this reflects the retention of ninefold symmetry only within the region of the pericentriolar material closest to the base. Difference images between the projections of the reconstruction and the original images supported this interpretation. The junction between the base and the barrel occurs at 228 nm (s68) from the proximal end and is a very complex structure. A central, smooth, cylinder crosses this region (s50, s68, s75) and closes to a solid structure within the base (s80, s87). The effect of distortions on these features was examined. The three centriole reconstructions did
not show a helical twist to the barrel (,5° in the most twisted example). Reconstructions of centrioles with distortions greater than 10% showed pronounced twist to the barrel; however, we know that our reconstruction method is not reliable for such distorted structures. In reconstructions of distorted centrioles, the internal cylinder is also present although its appearance changes. Several other features, such as the fenestrations in the periphery of the base, are visible in all of the reconstructions; however, their placement is not consistent between reconstructions. Labeling of the Centriole The link between low resolution structure and function requires localization of biochemically defined components. We have done this for two of the tubulin isoforms: a tubulin, which forms microtubules with b tubulin, and g tubulin, which is believed to be involved in microtubule nucleation (Raff et al., 1993; Stearns et al., 1991). As expected, a tubulin was found in the pericentriolar material and the staves of the microtubule barrel (Fuller et al., 1995; Gowen et al., 1995). It is not a component of the internal structures. The centrosome contains two sites of g tubulin localization. The first is the pericentriolar material where previous work by preembedding techniques had shown its presence (Stearns et al., 1991; Stearns and Kirschner, 1994). The second localization could be seen only by the use of a freeze substitution method in combination with postsectioning fixation (Fuller et al., 1995) which we have developed (Gowen et al., 1995) and corresponds to the internal cylindrical structure (Fuller et al., 1995) described above (Fig. 5). The four sections shown all cross the centriolar barrel. The three (Figs. 5a, 5b,
FIG. 5. g-Tubulin labeling of putative microtubule nucleation sites. Serial sectioning was performed on MDCK cells which had been freeze substituted and labeled with anti-g tubulin following the postsectioning fixation procedure described previously (Fuller et al., 1995; Gowen et al., 1995). The labeling of the central cylinder can also be seen in (a), (b), and (d) which include the distal end of the centriole. Notice that the microtubule barrel shows a smaller diameter (,250 nm) than is seen in the cryoelectron micrographs (Fig. 1) or the reconstruction (Fig. 4). The positions of microtubule ends near the centriole were determined from the series of sections which spanned each centriole. The sections shown are representative of those which show putative microtubule nucleation sites. The arrowheads mark sites of g-tubulin labeling visualized with protein A conjugated to 10 nm gold. The arrows mark the path of a microtubule which comes closest to the g tubulin label and which terminates within the section shown. All of the microtubule ends are near to sites of g tubulin labeling. Further, the trajectory was tangential to the cluster of g tubulin label. The scale bar represents 500 nm.
THREE-DIMENSIONAL RECONSTRUCTION OF THE CENTRIOLE
and 5d), which include the distal end of the barrel, show g tubulin labeling within the core of the structure. The section which passes through the proximal end of the barrel shows no g tubulin labeling in the center although labeling is seen the pericentriolar material. The centriole appears smaller in these sections than it is in cryoelectron micrographs such as those of Fig. 1. This difference is not a consequence of flattening since we have shown that the centrioles in Fig. 1 are not appreciably flattened (Fig. 2). The centrosome imposes organization on the microtubule network of the cell. The handed and polar structure which we see in our reconstructions can transfer such an organization to the microtubules which it nucleates only if the sites of microtubule nucleation are also organized in a polar manner. Moritz et al. (1995a, b) have presented reconstructions of Drosophila centrosomes in which preembedding labeling with anti-g tubulin revealed that ringlike structures are the sites of clusters of anti-gtubulin labeling. Two-thirds of the g tubulin labeling was found at the minus ends of microtubules. We found that our postsectioning fixation and labeling technique produced a similar image of the relationship between g tubulin and microtubule ends. We explored this by examining 25 series of sections of g tubulin labeled MDCK cells. All of the microtubule ends which we could identify in these serial sections were near to sites of g tubulin labeling as expected from the work on Drosophila centrosomes (Moritz et al., 1995a). Our results contrast with those on the Drosophila system in that we see many fewer microtubules per centriole. This reflects the difference between centrioles in MDCK cells in situ and isolated Drosophila centrioles which had been mixed with tubulin and allowed to grow microtubules in vitro. As a consequence much of our g tubulin labeling is not microtubule associated. The trajectories of the microtubules near the centrosome could be defined in 12 cases and were tangential to the sites of g tubulin labeling (Fig. 5). A tangential interaction of the microtubules would allow the transfer of organization and polarity from the complex centriolar structure to the microtubule network. The authors are pleased to acknowledge our generous and patient colleagues who made the completion of this work possible. We thank Marek Cyrklaff (EMBL) for help with some of the early microscopy and Dr. Michael Radermacher (Max-Planck-Institut fu¨r Biophysik, Frankfurt, F. R. Germany) for many helpful discussions of generalized weighted backprojection. We are indebted to Dr. Denis Chre´tien (EMBL) for very helpful discussions on the nature of microtubule nucleation and the polarity of the centriole and for his comments on the manuscript. We are also grateful to Dr. Michelle Moritz and Dr. David Agard (UCSF) for their comments on our labeling studies.
327
REFERENCES Adrian, M., Dubochet, J., Lepault, J., and McDowall, A. W. (1984) Cryo-electron microscopy of viruses, Nature 308, 32–36. Anderson, R. G. W. (1972) The three-dimensional structure of the basal body from the rhesus monkey oviduct, J. Cell Biol 54, 246–265. Bornens, M., Paintrand, M., Berges, J., Marty, M. C., and Karsenti, E. (1987) Structural and chemical characterization of isolated centrosomes, Cell Motil. Cytoskel. 8, 238–249. Buendia, B., Draetta, G., and Karsenti, E. (1992) Regulation of the microtubule nucleating activity of centrosomes in Xenopus egg extracts: Role of cyclin A-associated protein kinase, J. Cell. Biol. 116, 1431–1442. Chre´tien, D., Buendia, B., Fuller, S., and Karsenti, E. (1997) Reconstruction of centrosome cycle from cryo-electron micrographs. J. Struct. Biol., in press. Chre´tien, D., Fuller, S. D., and Karsenti, E. (1995) Structure of growing microtubule ends: Two-dimensional sheets close into tubes at variable rates, J. Cell Biol. 129, 1311–1328. Chre´tien, D., Kenney, J. M., Fuller, S. D., and Wade, R. H. (1996) Determination of microtubule polarity by cryo-electron microscopy, Structure 4, 1031–1040. Crowther, R. A. (1971) Procedures for three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs, Phil. Trans. R. Soc. Lond. B 261, 221–230. Crowther, R. A., Amos, L. A., Finch, J. T., De Rosier, D. J., and Klug, A. (1970a) Three-dimensional reconstructions of spherical viruses by Fourier synthesis from electron micrographs, Nature 226, 421–423. Crowther, R. A., DeRosier, D. J., and Klug, A. (1970b) The reconstruction of a three-dimensional structure from projections and its application to electron microscopy, Proc. Roy. Soc. Lond. A 317, 319–340. Fuller, S. D., Butcher, S. J., Cheng, R. H., and Baker, T. S. (1996) Three-dimensional reconstruction of icosahedral particles: The uncommon line, J. Struct. Biol. 116, 48–55. Fuller, S. D., Gowen, B. E., Buendia, B., and Karsenti, E. (1995) The core of the mammalian centriole contains g-tubulin, Curr. Biol. 5, 1384–1393. Fuller, S. D., Kenney, J. M., and Karsenti, E. (1992) Centrosomes, Curr. Opin. Struct. Biol. 2, 264–274. Glover, D. M., Gonzalez, C., and Raff, J. W. (1993) The centrosome, Sci. Am. 268, 62–68. Gowen, B. E., Buendia, B., Karsenti, E., and Fuller, S. D. (1995) Postembedding a-tubulin labelling of isolated centrosomes, Histochem. J. 27, 240–246. Gowen, B. E., Buendia, B., Reinsch, S., Kenney, J. M., Karsenti, E., and Fuller, S. D. (1994) The centriole: Inside and out, in 13th International Congress of Electron Microscopy: Applications in Biological Sciences, Vol. 3A, pp. 265–266. Les Editions de Physique Les Ulis, Paris. Kochanski, R. S., and Borisy, G. G. (1990) Mode of centriole duplication and distribution, J. Cell. Biol. 110, 1599–1605. Lange, B. M. H., and Gull, K. (1996) Structure and function of the centriole in animal cells: Progress and questions, Trends Cell Biol. 6, 348–352. Mancini, E. J., de Haas, F., and Fuller, S. D. (1997) High resolution icosahedral reconstruction: Fulfilling the promise of cryo-electron microscopy, Structure, 5, 741–750. Moritz, M., Braunfeld, M., Sedat, J., Alberts, B., and Agard, D. (1995a) Microtubule nucleation by g-tubulin containing rings in the centrosome, Nature 378, 638–640. Moritz, M., Braunfeld, M. B., Fung, J. C., Sedat, J. W., Alberts, B.
328
KENNEY ET AL
M., and Agard, D. A. (1995b) Three-dimensional structural characterization of centrosomes from early Drosophila embryos, J. Cell Biol. 130, 1149–1159. Paintrand, M., Moudjou, M., Delacroix, H., and Bornens, M. (1992) Centrosome organization and centriole architecture: Their sensitivity to divalent cations, J. Struct. Biol. 108, 107–128. Radermacher, M. (1988) Three-dimensional reconstruction of single particles from random and non-random tilt series, J. Electron Microsc. Tech. 9, 359–394. Raff, J. W., Kellogg, D. R., and Alberts, B. M. (1993) Drosophila gamma-tubulin is part of a complex containing two previously identified centrosomal MAPs, J. Cell. Biol. 121, 823–835. Rieder, C. L., and Borisy, G. G. (1982) The centrosome cycle in PTK2 cells: Assymetric distribution and structural changes in the pericentriolar material, J. Cell. Biol. 44, 117–132. Sheehan, B., Fuller, S. D., Pique, M. E., and Yeager, M. (1996) AVS software for visualization in molecular microscopy, J. Struct. Biol. 116, 99–105. Stearns, T., Evans, L., and Kirschner, M. (1991) Gamma-tubulin is a highly conserved component of the centrosome, Cell 65, 825–836. Stearns, T., and Kirschner, M. (1994) In vitro reconstitution of
centrosome assembly and function: The central role of g tubulin, Cell 76, 623–637. Stewart, P. L., Burnett, R. M., Cyrklaff, M., and Fuller, S. D. (1991) Image reconstruction reveals the complex molecular organization of adenovirus, Cell 67, 145–154. Tournier, F., Cyrklaff, M., Karsenti, E., and Bornens, M. (1991a) Centrosomes competent for parthenogenesis in Xenopus eggs support procentriole budding in cell-free extracts, Proc. Natl. Acad. Sci. USA 88, 9929–9933. Tournier, F., Komesli, S., Paintrand, M., Job, D., and Bornens, M. (1991b) The intercentriolar linkage is critical for the ability of heterologous centrosomes to induce parthenogenesis in Xenopus, J. Cell. Biol. 113, 1361–1369. van Heel, M. (1987) Similarity measures between images, Ultramicroscopy 21, 95–100. Verde, F., Dogterom, M., Stelzer, E., Karsenti, E., and Leibler, S. (1992) Control of microtubule dynamics and length by cyclin Aand cyclin B-dependent kinases in Xenopus egg extracts. J. Cell. Biol. 118, 1097–1108. Wade, R. H., and Chre´tien, D. (1993) Cryo-electron microscopy of microtubules. J. Struct. Biol. 110, 1–27. Wilson, E. B. (1925) The Cell in Development and Heredity, MacMillan, New York.