The core of the mammalian centriole contains y-tubulin Stephen D. Fuller*, Brent E. Gowen*, Sigrid Reinsch t, Alan Sawyer t, Brigitte Buendiat, Roger Wepf* and Eric Karsentit *The Structural Biology Programme, tthe Cell Biology Programme and the Cell Biophysics Programme, European Molecular Biology Laboratory, Postfach 10.2209, Meyerhofstrasse 1, D69012 Heidelberg, Germany. Background: The microtubule network, upon which transport occurs in higher cells, is formed by the polymerization of a and 3 tubulin. The third major tubulin isoform, y tubulin, is believed to serve a role in organizing this network by nucleating microtubule growth on microtubule-organizing centers, such as the centrosome. Research in vitro has shown that y tubulin must be restored to stripped centrioles to regenerate the centrosomal functions of duplication and microtubule nucleation. Results: We have re-examined the localization of y tubulin in isolated and in situ mammalian centrosomes using a novel immunocytochemical technique that preserves antigenicity and morphology while allowing increased accessibility. As expected, a tubulin was localized in cytoplasmic and centriolar barrel microtubules and in the associated pericentriolar material. Foci of y tubulin were observed at the periphery of the organized pericentriolar material, as reported previously, often near the termini of microtubules. A further and major location of y tubulin was a structure within the proximal end of the centriolar barrel. The distributions were complementary, in that ao tubulin was excluded from the core of the centriole, and y tubulin was excluded from the microtubule barrel. Conclusions: We have shown that y tubulin is localized both in the pericentriolar material and in the core of the mammalian centriole. This result suggests that y tubulin has a role in the centriolar duplication process, perhaps as a template for growth of the centriolar microtubules, in addition to its established role in the nucleation of astral microtubules. Current Biology 1995, 5:1384-1393 Background The centrosome has fascinated cell biologists since it was first recognised to be the region of the cell from which asters arise during mitosis and the microtubule network originates in interphase [1,2]. Although the morphology of centrosomes varies with cell type and species, the typical mammalian centrosome is comprised of two perpendicularly arrayed centrioles and their associated pericentriolar material. The canonical centriolar structure is a barrel of nine triplets of microtubules, which is elaborated by appendages, satellites and a variety of internal structures. Despite extensive studies of centrosomal morphology, a complete correlation of its intricate structure with its essential molecular components is not yet available; dozens of centrosomal antigens have been described [3,4], but only a few have been localized within the structure. A major step toward an understanding of the molecular basis of the centrosomal control of microtubule nucleation was the discovery by Oakley and colleagues [5] of y tubulin. This is the product of a gene which complemented temperature-sensitive mutations in the gene encoding 3 tubulin. The sequence of -y tubulin is conserved from yeast to mammals [6], but is as distinct in sequence from ao and 3 tubulin as they are from each other. Although the genetic evidence indicated that y tubulin must interact with 3 tubulin [5], biochemical and immunocytochemical studies have shown that it does not enter microtubules formed from a and 3 tubulin [5,7]. Immunofluorescence studies in a number of species have shown that y tubulin is localized to the centrosome in animal cells, and to the functionally equivalent structures in plant and fungal cells. In each case, the localization of y tubulin matches the focus of microtubule-organizing activity in the cell. Deletion of the y tubulin gene in fungal cells [8], or interference with y tubulin activity using specific antibodies in animal cells [9], causes the loss of microtubule nucleation during mitosis and interphase. The presence of y tubulin in centrosomes is believed to be the molecular explanation for their role in organizing the microtubule network of the cell. A recent, elegant study of the role of y tubulin examined the effects of its overexpression in mammalian cells [10]. Microtubule nucleation was increased, demonstrating a direct link between y tubulin and microtubule nucleation in vivo. The authors also showed that overexpression resulted in the formation of 50 nm diameter tubes ('macrotubules'), which were rich in y tubulin but appeared to exclude a( and 3 tubulin. The localization of y tubulin by electron microscopy would be an important step in elucidating the function of this intriguing molecule. Unfortunately, the definitive localization of antigens within the centrosome has been hindered by the contending requirements of preserving antigenicity and maximizing accessibility. The use of a Correspondence to: Stephen Fuller. E-mail address: fuller@EMBL-Heidelberg.DE 1384 © Current Biology 1995, Vol 5 No 12 The centriole core contains y tubulin Fuller et al. pre-embedding technique to preserve antigenicity has demonstrated that y tubulin is localized to the pericentriolar material [7]; this supports the view that it is involved in microtubule nucleation as the pericentriolar material is the initiation site of astral and cytoplasmic microtubule growth [11,12]. The y tubulin in the pericentriolar material may therefore act as a template for astral microtubules, promoting the growth of the 13 protofilament microtubules over the kinetically favoured 14 protofilament microtubules that dominate under selfassembly conditions. A recent result has shown that y tubulin binds to the minus ends of microtubules, as would be required for this role [13]. Other centrosomal components, such as pericentrin, appear to aid y tubulin in organizing microtubules [14], but an understanding of the role of such higher-order interactions of y tubulin [15] must await a characterization of the organization of the components of the centrosome. A role for y tubulin in the replication of centrioles has been suggested by studies in vitro [16,17]. This has been most convincingly demonstrated by the use of Xenopus egg extracts, which donate y tubulin to centrioles in vitro. Pre-embedding immunocytochemistry did not detect y tubulin in the centrioles prior to incubation in the egg extract. After incubation, y tubulin was found in high molecular weight complexes, and was localized to the sites of microtubule nucleation after centrosome formation [17]. This demonstration of complementation gave rise to a hierarchical model for centrosomal organization [17]. The available immunolocalization data distinguished three functional classes of centrosomal antigens: those strictly associated with centrioles or basal bodies, such as centrin and ot tubulin; those associated with centrioles and the pericentriolar material, such as pericentrin; and those associated only with the pericentriolar microtubule nucleation sites, such as y tubulin. The immunolocalization of centrosomal antigens is complicated by the need to preserve of antigenicity and morphology, which often conflicts with the requirements for accessibility throughout the structure and high contrast in the image. Pre-embedding techniques address only the first of these needs. We have developed a technique for performing immunocytochemistry on centrosomes that combines the antigenic preservation afforded by cryo-fixation with the accessibility that is typical of post-embedding immunogold labelling methods [18]. A key feature of our method is the use of a post-sectioning fixation step before the final staining procedures. This overcomes the problem that cryo-fixation followed by freeze substitution and Lowicryl embedding results in material that is sensitive to contrasting agents [18]. In this study, we have applied this technique to isolated centrosomes and to centrioles within intact cells to establish the distribution of tubulin isoforms. We confirm that -y tubulin is localized to the pericentriolar material [7,17], and show this localization corresponds to the sites of microtubule contact with the pericentriolar material - as RESEARCH PAPER would be expected for the role of y tubulin in microtubule nucleation. Furthermore, we show that y tubulin is a major component of the core of the centriole, as might have been expected from its demonstrated role in centrosome nucleation [7,16]. As the core and distal portions of the centriole are formed earliest during replication, y tubulin is properly positioned to serve as a template during centriolar duplication, in addition to its previously recognized role as a template for astral microtubule nucleation. Results Localization of y tubulin in isolated centrosomes Using the post-sectioning fixation technique, the internal features of isolated centrosomes were well preserved because the size of the structure allowed complete vitrification of the sample and the avoidance of ice damage. When the serial sections were cut nearly perpendicular to the barrel axis, the triplet microtubules of the centrosome barrel were well resolved (Fig. la-t), as were the other internal features such as the symmetric structure seen at the base of the barrel (Fig. c,d). In order to determine the distribution of y tubulin within the centrosome, we used an affinity-purified rabbit antiserum raised against a peptide within the carboxyl terminus of the Xenopus y tubulin. The antibody was then visualized using 10 nm protein A gold (PAG), and y tubulin seen to be localized throughout the proximal and medial portion of the pericentriolar material (as expected, [7]); stereo views of sections showed that this localization was often near the periphery of the structures formed by the pericentriolar material, in the satellites and other extensions that project from the proximal end of the centriole, and in the sheath of pericentriolar material that surrounds the centriolar barrel. Unexpectedly, this peripheral localization was not the only site of y tubulin localization. A 'core' structure that was entirely within the centrosome barrel was also seen to be consistently labelled with antibody, indicating the presence of y tubulin. Examination of dozens of stereo views of serially sectioned centrosomes showed that the core labelling extended from the proximal end of the centriole to its middle (Fig. lg-i). The stereo views provided no convincing evidence for the localization of y tubulin in the microtubule triplets themselves. However, y tubulin was consistently labelled in the connection [19] that joins the two centrioles of the centrosome (Fig. lj,k). Three control experiments were carried out to determine the specificity of y tubulin labelling in these experiments. Firstly, the antibody specific to the carboxy-terminal peptide of y tubulin was omitted from the procedure; secondly, before labelling, this antibody was incubated with the peptide antigen used to raise the antiserum. No labelling of pericentriolar material or internal structures with 10 nm PAG was observed under these conditions (data not shown). Thirdly, a rabbit 1385 1386 Current Biology 1995, Vol 5 No 12 Fig. 1. Serial sections of isolated centrosomes. The location of anti-y tubulin antibody binding was visualized with 10 nm PAG in these and all subsequent pictures, unless specified. (a-f) A series of sections which are nearly perpendicular to the centriolar barrel axis and proceed from the (a) proximal to the (f) distal end of the structure. (c-f) The sites of y tubulin labelling are at the periphery of the pericentriolar material and in the organized structure within the proximal end of the barrel. Scale bar, 0.25 pum. (g-i) Sections at a slight angle to the barrel axis; y tubulin is found in the proximal end of the barrel and in the pericentriolar material. The densely staining material seen in (g) is the connection between two centrioles and is also a site of y tubulin labelling. No labelling of the centriolar microtubules is observed. (j,k) Sections that show the labelling of the connection between two centrioles more clearly. (g-k) Scale bar, 0.25 ptm. polyclonal antibody (Xgam) raised against a fusion protein of Xenopus y tubulin [7] was used. This antibody gave labelling patterns that were indistinguishable from those obtained with the peptide antibody (data not shown). Complementary distributions of y tubulin and a tubulin within the centriole The pre-embedding technique used for previous y tubulin immunolocalizations was not designed to visualize antigens buried within the structure of the centrosome. Using this technique, oa tubulin was only localized to its previously described distribution at the edge of the pericentriolar material (data not shown). However, the postsectioning fixation technique has also shown extensive labelling of ao tubulin through the microtubule barrel of the centriole [18]. The greater accessibility of antigens afforded by the postsectioning fixation technique therefore allowed us to determine the relative locations of y and o tubulin in the centriole by double labelling experiments. Both ot tubulin (labelled with 5 nm PAG) and y tubulin (labelled with 10 nm PAG) were seen in the pericentriolar material (Fig. 2). The labelling of a( tubulin was peripheral to that of y tubulin, consistent with the established role of y tubulin in astral microtubule nucleation (Fig. 2a) [10]. The labelling patterns of ot tubulin and y tubulin in the centriole itself were complementary. No ot tubulin labelling of the core was ever seen, although the triplet microtubules were always labelled. In addition to its localization in the core, y tubulin labelling was found peripheral to the triplet microtubules when pericentriolar material could be observed (Fig. 2c,d). A comparison The centriole core contains yytubulin Fuller eta/.. The centriole core contains -ytubulin Fuller eta!. RESEARCH PAPER RESEARCH PAPER sections showed that the y tubulin was found throughout the length of the core and within a radius of 200 A of the centre (Fig. 3c). This measurement is consistent with the diameter of the macrotubules formed by the overexpression of y tubulin [10]. The distribution shows that y tubulin is not significantly enriched at the base, and cannot be the only component of the morphologically defined core. Fig. 2. Doubly labelled centrioles. (a) Cross-section through the base of a centriole which has been labelled first with antibody to y tubulin (10 nm PAG) and then with antibody to a tubulin (5 nm PAG). (b) Oblique section through the centriole core, labelled as in (a). (c) Longitudinal section of a centriole which has been labelled first with antibody to a tubulin (5 nm PAG) and then with antibody to y tubulin (10 nm PAG). (d) Cross section through the proximal region of the centriole barrel, labelled as in (c). Scale bar, 0.25 Lrnm. of Figure 2a,b with Figure 2c,d demonstrates that the localization of the two tubulin isoforms was independent of the order of application of the reagents. Quantitative distribution of tubulin isoform labelling The images shown in Figures 1 and 2 are representative of the several hundred images from several scores' of sets of serial sections used in this study. We found that it was important to examine the distribution of labelling in serial sections and often in stereo images to untangle the threedimensional nature of the labelling pattern in individual images. Quantitation also overcomes this effect because the average distribution of label is a more robust measure of localization than any individual measurement. We therefore characterized the distribution of ot and y tubulin in sections of centrosomes with respect to the morphological landmarks of the centriole visible in the section - the microtubule barrel, the internal core and the base (Fig. 3a). The quantitation of the distribution of t and y tubulin in sections perpendicular to the barrel showed that two populations of y tubulin were sharply demarcated by the position of the barrel (Fig. 3b). As suggested by the individual images, the two tubulin isoforms did not overlap within the barrel, but did overlap within the pericentriolar material. Of the 676 y tubulin labels measured, 16.7 % were within the barrel. The ratio of concentrations is higher as the frequency at different radii must be corrected by a factor of l/r for a comparison of concentrations in cross sections. The quantitation of longitudinal Localization of y tubulin in intact cells The use of isolated centrosomes allows good preservation of the structure for analysis, but does so at the expense of the contextual information that defines the relationship of this organelle to other structures in the cell. Furthermore, the preparation procedure causes some distortion of the centriolar barrel and pericentriolar features, presumably due to the pelleting used during isolation. The distribution of y tubulin might also alter during isolation, such that localization reflects adventitious binding of y tubulin to the centrosome. In order to determine whether such experimental artifacts were affecting our results with isolated material, and to obtain the contextual information about the centrosome, we applied the post-sectioning fixation technique to whole Madin-Darby canine kidney (MDCK) cells. Freeze substitution and post-sectioning maintained the morphology of interphase MDCK cells, and structures such as the mitochondria, Golgi apparatus and cytoplasmic microtubules were well preserved (Fig. 4). The centriolar barrel and pericentriolar material of the centrosomes in these sections were not distorted and, as in the isolated centrosomes, y tubulin was localized to the pericentriolar material and to the centriolar core (Fig. 4b). Serial sections of centrioles in these intact cells confirmed this localization. Figure 5 shows the distribution of y tubulin in serial sections of a pair of centrioles; the right member of the pair is sectioned nearly perpendicular to the barrel (Fig. 5e-i shows serial distal to proximal sections), and the left member is sectioned obliquely. The cores of both mother and daughter centrioles show y tubulin labelling, and labelling is also seen at the edge of the pericentriolar material. The context provided by the sections of the intact cell showed that the sites of y tubulin labelling in the pericentriolar material were often the sites of tangential contact with cytoplasmic microtubules (Figs 5 and 6). Although we have not defined these positions as the sites of initiation of microtubule growth, the similarity between our results and those obtained in the threedimensional reconstruction of Drosophila centrosomes [20] suggests that this is indeed the geometry of the y tubulin-microtubule interaction. Mitotic and interphase centrosomes show similar distributions of y tubulin The MDCK cells used for the labelling were subconfluent, and some mitotic cells could be examined. Low-magnification analysis of sections of metaphase cells showed that the microtubules connecting the kinetochores and 1387 1387 1388 Current Biology 1995, Vol 5 No 12 Fig. 3. Quantitation of y tubulin labelling. (a) Schematic of the morphology of a centriole. Pericentriolar material is shown as a grey cloud around the centriole; the black internal bars indicate the median radial and longitudinal limits of the core in which y tubulin labelling was observed. Quantitation of labelling was performed by using cross sections that were perpendicular to the centriolar barrel axis and longitudinal sections that contained the barrel axis. Measurements of the label (black square) were made relative to the morphological features of the centrioles in individual sections. For the entire population, the values of the minimum and maximum outer radii were 862 A (SE 9.4 A, median 900 A) and 965 A (SE 9.4 A, median 1000 A) for the barrel and 453 A (SE 7.7 A, median 440 A) and 513 A (SE 11.4 A, median 480 A) for the core. Distances in longitudinal sections, d, were measured from the base. (b) Histogram of the distribution of radii of a and y tubulin labelling relative to the barrel in cross sections revealing the two populations of y tubulin and the separation of the inner population from a tubulin. (c) Distribution of y tubulin within the core itself as seen in longitudinal sections with the median positions of the outer edge of the barrel and the core marked. Micrographs of 147 separate sections were analysed for a total of 601 a tubulin labels and 732 y tubulin labels to produce the distributions. centrosomes were easily visible (Fig. 6a). As in interphase sectioned pairs of centrioles in the metaphase cells. Figure cells, y tubulin was localized to the edge of the peri- 7 shows an example in which the sectioning is near to the centriolar material and to the core within the proximal common plane of the centriolar barrels; y tubulin was end of the centriole (Fig. 6b,c). In a few instances, we labelled in the cores of both the mother and the daughter Fig. 4. Distribution of y tubulin in an intact interphase MDCK cell. (a) Overview of an interphase cell in which the morphology of the cell and the preservation of structures, such as the mitochondria (m), Golgi complex (g) and microtubules (arrowheads), are well displayed. The nucleus (n) and centrioles (arrow) show some freeze damage. Scale bar, 2 im. (b) Detail view of one of the centrioles in (a), which has been sectioned near to its proximal end, showing the central and pericentriolar pattern of y tubulin labelling. Scale bar, 0.25 m. The centriole core contains y tubulin Fuller et al. RESEARCH PAPER Fig. 5. Cross sections of a pair of centrioles in an interphase MDCK cell. Labeliing of y tubulin is seen in both the mother and daughter of this pair of centrioles. The right member of the pair is sectioned perpendicular to the barrel axis with the sections running from (f) distal to (i) proximal. Scale bar, 0.5 pm. centrioles, and in a structure at their proximal ends. The region of the daughter (intersecting) centriole that was closest to the mother was labelled by y tubulin (Fig. 7b). Contacts between the positions of pericentriolar y tubulin localization and microtubules had the same geometry as in interphase cells (Fig. 7). We conclude that the general features of the distribution of y tubulin in the core nd pericentriolar material are similar in the centrosomes of both mitotic and interphase cells. Discussion The major result of this work is a more complete description of the distribution of y tubulin in the centrosome. We have analyzed this distribution in both isolated centrosomes from a lymphoid line, which allow very good preservation of internal features due to their ease of vitrification, and in intact MDCK cells, which provide the context for the centrosomal structure but are less efficiently vitrified. The freeze substitution, post-sectioning fixation technique allows access to internal structures, as demonstrated by the high degree of labelling of a tubulin in the barrel microtubules of isolated centrosomes [18]. Using this technique with an anti-y tubulin antibody, we observed y tubulin labelling at the periphery of the pericentriolar material, in agreement with a previous report that used pre-embedding immunocytochemistry [7]. The electron-dense connection between centrioles [19] is also a site of y tubulin localization. We also report the novel finding that an organized structure within the proximal end of the centriole contains y tubulin. The length of the structure is approximately 0.2 m and its radius is 200 A. Although the structure is found within the proximal end of the centriole, the proximal end itself was never labelled with the antibody, nor were the microtubules of the centriole barrel. In double labelling experiments using antibodies directed against ot and y tubulin, we determined that their distribution within the centriole was complementary - ot tubulin was found only in the microtubule barrel, whereas y tubulin was found only in the proximal core structure. The y tubulin distributions which we observe in the isolated centrosomes appear to accurately reflect the cellular distribution. The observation that the distribution seen in isolated centrosomes matches that seen in intact cells rules out the possibility that y tubulin has redistributed to bind selectively to the central core. Our observations confirm and extend those obtained by pre-embedding techniques [7]. Our information is more complete because the freeze substitution and post-sectioning fixation technique preserves morphology and antigenicity while allowing access to internal antigens. The use of intact cells confirmed the localization of the y tubulin in the centriole core, and allowed us to determine that the features of y tubulin distribution in mitotic and interphase cells are very similar. In the intact cells, however, the morphology of the microtubules within the 1389 1390 Current Biology 1995, Vol 5 No 12 Fig. 6. The distribution of y tubulin in an intact metaphase MDCK cell. (a) Overview of an MDCK cell in metaphase. The morphology of the cell and structures such as the mitochondria (m), microfilaments (f)and microtubules (arrowheads) are well preserved. The chromosomes (C) and centrioles (arrows) show some freeze damage. Scale bar, 2 am. (b,c) Detail views of the (b) proximal end and (c) distal end of a centrosomes which have been sectioned perpendicular to the barrel axis. Both centrosomes display the central and pericentriolar pattern of y tubulin labelling. Scale bar, 0.25 Iam. centriole was not as well preserved as in the isolated material. This reflects freeze damage affecting the relatively fragile centriolar structure as well as regions of the nucleus. The preparation of the sample using the high cooling rate of the propane jet mitigates ice damage within the cell, but does not prevent it completely. Highpressure freezing would abolish ice damage, but has been reported to depolymerize microtubules [21]. Nevertheless, the propane-jet method allowed us to visualize the context of the centriole as well as the distribution of y tubulin. The generally accepted role for y tubulin is the nucleation of cytoplasmic microtubules from the pericentriolar material. The sites of contact between y tubulin localization and microtubules have a characteristic geometry that would impose a handed and polar organization on the cytoplasmic microtubules [20], indicating that the pericentriolar material has a specific organization despite its nebulous appearance. In this context, the description of egg-extract y tubulin as part of a 25 S cytoplasmic complex is intriguing [17]. Such an extended, higher-order structure would be capable of the interactions that could impose geometry on the microtubule contact. Indeed, a hierarchical model of the centrosome in which central components such as pericentrin impart positional information to higher-order structures containing y tubulin would allow such organized interactions [14,15,17]. A hierarchy of organizing interactions is a common theme in complex virus structures - a variety of simple components combine in subassemblies, which then interact to organize the complete structure [22,23]. The observation that sperm centrioles lack y tubulin prior to incubation with egg extract refers only to the peripheral y tubulin accessible to pre-embedding techniques [17]. However, an increasing body of evidence indicates that y tubulin has a structural role in the centrosome Fig. 7. A series of sections (a-c) from a metaphase cell, showing a pair of centrioles that are sectioned near to the common plane of their barrels. The spindle pole (not shown) is to the lower right. The mother (the lower of the pair) shows both pericentriolar and core labelling of y tubulin. The daughter (above) shows a similar pattern in which the labelling at the proximal end is seen closest to the mother. Scale bar, 0.5 pm. The centriole core contains y tubulin uller et al. The centriole core contains y tubulin Fuller et al. RESEARCH PAPER RESEARCH PAPER Fig. 8. Template-mediated centriole duplication. The suggested distribution of y tubulin and other components during centriole duplication is shown using a low-resolution centriole reconstruction [35] to indicate positions; y tubulin is shown in red. The maternal centriole has y tubulin (a) in its core as well as (b) on the periphery of the pericentriolar material. The microtubules of the centriolar barrel are made up of a and tubulin, and are represented as straight black lines. (c,d) Duplication begins with the association of y tubulin and other components (blue) with a site at the side of the centriolar barrel. (e-g) These components form a template from which the microtubules of the daughter centriole grow. (h,i) Once the daughter's barrel has grown to full length, the centrioles separate but remain linked by a structure that contains y tubulin. [9,16,17,24]. We have determined that y tubulin is component of both the periphery and the core of the centriole, and that the y tubulin within the core fills a volume with similar dimensions to that reported for the macrotubules produced by overexpression [10]. The presence of y tubulin in a core structure within the centriolar barrel suggests that it has a specific role as a centriolar template for this population of the molecule. Two groups have shown that sperm centrioles acquire y tubulin when incubated in an egg extract [16,17] which supports centrosome duplication [25]. An attractive model is that the observed binding of y tubulin to the mother centriole represents a template for the growth of the microtubule barrel of the procentriole. Indeed images such as Figure 7, which show the motherdaughter intersection, evoke such a hypothesis. Figure 8 presents a schematic of this view of centriole duplication. The maternal centriole contains y tubulin at its core and at the periphery of the pericentriolar material (Fig. 8a,b). Duplication begins with the association of y tubulin and other components with a site at the side of the centriolar barrel (Fig. 8c,d), forming a template from which the microtubules of the daughter centriole grow (Fig. 8e-g). Once the daughter's barrel has grown to full length, the centrioles separate (Fig. 8h,i) but remain linked by a structure that contains y tubulin (Fig. 8i; see also Fig. 1). The relatively high stability of centriolar microtubules could be a consequence of the recently demonstrated stability of the y tubulin macrotubules [10]. The present work does not demonstrate this model. However, by recasting the question of the nucleation of daughter centrioles into one of the organization of the y tubulin template, clear paths are indicated for experimental tests by reconstitution and depletion experiments with in vitro systems [16,17,25]. The model leaves open the question of the control of the centriole length. The origin of the two populations of centrosomal y tubulin is also an important question. The complementation of sperm with egg extract indicates strongly that the peripheral population arises maternally [17], but the core population may have an different origin and may be chemically distinct. The demonstration of an antigenically distinct population of y tubulin that seems to be associated with centrioles but not with their peripheral structures [26], suggests that the two localizations correspond to different proteins. In Drosophila, the two y tubulin genes [27] are differentially expressed during development (C. Gonzalez, unpublished observations), as might be expected if they play different functional roles. It will also be interesting to apply our post-sectioning method to the fungal equivalent of the centrosome, the spindle pole body, which is known to contain y tubulin, to see whether it contains a core as well as a halo of y tubulin. Materials and methods Immunological reagents The antibody to ao tubulin (a kind gift from T.E. Kreis) was raised against the carboxy-terminal 13 amino acids of ot tubulin (KVEGEGEEEGEEY) [28]. The antibody to y tubulin was generated against the synthetic peptide, HAATRPDYISWGTQC, which corresponds to amino acids 433-437 of Xenopus y tubulin [7], with an additional carboxy-terminal cysteine residue for coupling. The peptide was conjugated with keyhole limpet haemacyanin [29], and used to immunize rabbits [30] with 50 g injected into lymph nodes and 25 Lg injected 1391 1391 1392 Current Biology 1995, Vol 5 No 12 subcutaneously thereafter. The antiserum was affinity-purified on the synthetic peptide coupled to CNBr-activated Sepharose [28]. The specificity of the antibody was controlled by preincubating it with a five-fold molar excess of peptide before using it for immunocytochemistry. For some experiments, the Xgam antibody [7] was used and gave identical results. Immunolabelling and sectioning Isolated centrosomes were prepared as described [31]. Strain 2 Madin-Darby canine kidney (MDCK) cells were grown on gold foil as described [32]. A detailed discussion of the immunocytochemical technique used with freeze-substituted samples has been presented [18]. Briefly, centrioles were concentrated by centrifugation onto carbon and Formvar coated 100 mesh electron microscopy grids. The samples of isolated centrosomes were vitrified on the grids by plunging into ethane slush [33]. MDCK cells were vitrified using a propane-jet freezing device (Balzers CryoJet type QFD-101) to compensate for the greater thickness of the sample. The samples were then placed in a Reichert-Jung CS Auto Cryo Substitution apparatus in the presence of 0.5 % (w/v) uranyl acetate in methanol at -90 C. After freeze substitution, the samples were embedded at low temperature (-45 C) in Lowicryl HM20 and polymerized with ultraviolet light [18]. The blocks were removed from the device and allowed to complete polymerization in ambient light for three days. Serial sections were cut from each sample and placed upon carbon and Formvar coated slot grids (Synaptek" Notch-Num. 2 x 1 mm slot grids; Electron Microscopy Sciences, Miinchen, Germany). Prior to immunolabelling, grids were blocked by placing them on drops of calcium- and magnesium-free phosphate buffered saline, pH 7.0 (PBS) containing 1% (w/v) ovalbumin for 5 min. Grids of a series were then placed on: PBS with 1 % (w/v) ovalbumin; or anti-ct tubulin antibody pre-adsorbed with antigen for 60 min at room temperature, diluted 1/50; or anti-y tubulin antibody pre-adsorbed with antigen for 60 min at room temperature, diluted 1/25; or anti-oa tubulin antibody diluted 1/50 in PBS with I % (w/v) ovalbumin; or a 1/25 dilution of anti-y tubulin antibody in PBS with 1% (w/v) ovalbumin. After 60 min incubation and three 5 min washes in PBS, the grids were placed on a drop of 10 nm PAG, diluted 1/40 in PBS with 1 % (w/v) ovalbumin, for 30 min. The grids were then washed three times in PBS for 5 min each. Double labelling was performed by successive incubation with the first antibody, 5 nm PAG (1/65 dilution in PBS, 1% (w/v) ovalbumin) or 10 nm PAG (1/40 dilution in PBS, 1% (w/v) ovalbumin), 5 ,g ml- I free protein A, the second antibody and 10 or 5 nm PAG [34]. All four possible combinations ( tubulin with 5 nm gold followed by y tubulin with 10 nm gold, or y tubulin with 5 nm gold followed by ot tubulin with 10 nm gold, and so on) were used and found to give equivalent results. After immunolabelling, the grids were rinsed four times (1 min each) in triple-distilled water and fixed by successive treatments with 2 % (w/v) tannic acid for 10 min, 3 washes for 5 sec each in 1 % (w/v) sodium sulfate, 2 % (w/v) glutaraldehyde for 10 min, and then 1 % (w/v) osmium tetroxide for 10 min. The sections were contrasted by staining in 5 % (w/v) uranyl acetate in 50 % ethanol for 10 min, followed by treatment with tannic acid and sodium sulfate as above, 5 % (w/v) lead citrate for 5 min, and then air dried [18]. Pre-embedding labelling of centrosomes with anti-a tubulin and Epon embedding was performed as described [7]. Sections were examined in a Phillips EM400 equipped with a goniometer stage, at an accelerating voltage of 80 kV and using magnifications of either x 11 700 or x 33 000. Stereo views were generated by bringing the grid to the eucentric position and taking successive images at 0 °, +9 and -9 tilts. Quantitation was performed by measuring the radius (r, see Fig. 3a) of individual gold labels and of morphological features on individual cross sections. Only sections that were close to perpendicular to the barrel axis (rmax - rmin < 100 A) were used, such that the morphological measurement was unambiguous. Longitudinal sections, parallel to and including the centriole axis, were used to characterize the distribution of y tubulin in the core in terms of radius (r) and distance (d, see Fig. 3a) from the centriole base. Acknowledgements: We are pleased to acknowledge critical readings of the manuscript by our colleagues J. Kenney and D. Chr&tien (EMBL). We also thank T. Stearns (UCSF, San Francisco, USA) for his kind gift of a rabbit antibody (Xgam) to a fusion protein of Xenopus y tubulin that was used as a control for the tubulin labelling with our anti-peptide antibody, and T.E. Kreis (University of Geneva) for the antibody to tubulin used in Figures 1 and 4. References 1. Wilson EB: The Cell in Development and Heredity, 3rd edn. New York: MacMillan Publishing; 1925. 2. Glover DM, Gonzalez C, Raft JW: The centrosome. Sci Am 1993, 268:62-68. 3. Fuller SD, Kenney M, Karsenti E: Centrosomes. Curr Opin Struct Biol 1992, 2:264-274. 4. Kalt A, Schliwa M: Molecular components of the centrosome. Trends Cell Biol 1993, 3:118-128. 5. Oakley BR, Oakley CE, Yoon Y, Jung MK: Gamma-tubulin is a component of the spindle pole body that is essential for microtubule function in Aspergillus nidulans. Cell 1990, 61:1289-1301. 6. Burns RG: -, P3-, and y-tubulins: sequence comparisons and structural constraints. Cell Motil Cytoskel 1991, 20:181-189. 7. Stearns T, Evans L, Kirschner M: Gamma-tubulin is a highly conserved component of the centrosome. Cell 1991, 65:825-836. 8. Horio T, Uzawa S, Jung MK, Oakley BR, Tanaka K, Yanagida M: The fission yeast gamma-tubulin is essential for mitosis and is localized at microtubule organizing centers. J Cell Sci 1991, 99:693-700. 9. Joshi HC, Palacios MJ, McNamara L, Cleveland DW: Gamma-tubulin is a centrosomal protein required for cell cycle-dependent microtubule nucleation. Nature 1992, 356:80-83. 10. Shu H-B, Joshi HC: 'y-tubulin can both nucleate microtubule assembly and self-assemble into novel tubular structures in mammalian cells. J Cell Biol 1995, 130:1137-1147. 11. Vorobjev IA, Chentsov YS: Centrioles in the cell cycle. . Epithelial cells. J Cell Biol 1992, 93:938-949. 12. Gould RR, Borisy GG: The pericentriolar material in Chinese hamster ovary cells nucleates microtubule formation. J Cell Biol 1977, 73:601-615. 13. Li Q, Joshi HC: y-tubulin is a minus end specific microtubule binding protein. J Cell Biol 1995, 131:207-214. 14. Doxsey SJ, Stein P, Evans L, Calarco PD, Kirschner M: Pericentrin, a highly conserved centrosome protein involved in microtubule orga- nization. Cell 1994, 76:639-650. 15. Raf JW, Kellogg DR, Alberts BM: Drosophila gamma-tubulin is part of a complex containing two previously identified centrosomal MAPs. J Cell Biol 1993, 121:823-835. 16. Felix M-A, Antony C, Wright M, Maro B: Centrosome assembly in vitro: role of -y-tubulin recruitment in Xenopus sperm aster formation. Cell Biol 1994, 124:19-31. 17. Stearns T, Kirschner M: In vitro reconstitution of centrosome assembly and function: the central role of y tubulin. Cell 1994, 76:623-637. 18. Gowen BE, Buendia B, Karsenti E, Fuller SD: Post-embedding et- tubulin labelling of isolated centrosomes. Histochem 1995, 27:240-246. 19. Tournier F, Komesli S, Paintrand M, Job D, Bornens M: The intercentriolar linkage is critical for the ability of heterologous centrosomes to induce parthenogenesis in Xenopus. J Cell Biol 1991, 113:1361-1369. The centriole core contains y tubulin Fuller et al. 20. Moritz M, Braunfeld MB, Fung JC, Sedat JW, Alberts BM, Agard DA: Three-dimensional structural characterization of centrosomes from early Drosophila embryos. J Cell Biol 1995, 130:1149-1159. 21. Michel M, Gnagi H, Muller M: Diamonds are a cryosectioner's best friend. J Microsc 1992, 166:43-46. 22. Fuller SD, Berriman JA, Butcher SJ, Gowen BE: Low pH induces the swivelling of the glycoprotein heterodimers in the Semliki Forest virus spike complex. Cell 1995, 81:715-725. 23. Stewart PL, Burnett RM, Cyrklaff M, Fuller SD: Image reconstruction reveals the complex molecular organization of adenovirus. Cell 1991, 67:145-154. 24. Sunkel CE, Comes R,Sampio P, Perdigao, Gonzalez C: Gamma tubulin is required for the structure and function of the microtubule organizing center in Drosophila neuroblasts. EMBOJ 1995, 14:28-36. 25. Tournier F,Cyrklaff M, Karsenti E, Bornens M: Centrosomes competent for parthenogenesis in Xenopus eggs support procentriole budding in cell-free extracts. Proc Natl Acad Sci USA 1991, 88: 9929-9933. 26. Dibbayawan TP, Harper JI,Elliott JE, Gunning BES, Marc J: A y-tubulin that associates specifically with centrioles in HeLa cells and the basal body complex in Chlamydomonas. Cell Biol Int 1995, 19:559-567. 27. Zheng Y, Oakley CE, Oakley BR: Identification of a second y-tubulin gene in Drosophila melanogaster. J Cell Biol 1991, 115:382. 28. Kreis TE: Microtubules containing detyrosinated tubulin are less dynamic. EMBOJ 1987, 6:2597-2606. RESEARCH PAPER 29. Kreis TE: Microinjected antibodies against the cytoplasmic domain of vesicular stomatitis virus glycoprotein block its transport to the cell surface. EMBOJ 1986, 5:931-941. 30. Howell KE, Kern H, Fuller SD, Tooze J: Sorting and coordinate induction of three endoplasmic reticulum proteins containing the KDEL retention signal. J Cell Biol 1988, 107:772a. 31. Bornens M, Paintrand M, Berges J, Marty MC, Karsenti E: Structural and chemical characterization of isolated centrosomes. Cell Motil Cytoskel 1987, 8:238-249. 32. Fuller SD, von Bonsdorff SC, Simons K: Vesicular stomatitis virus infects and matures only through the basolateral surface of the polarized epithelial cell line, MDCK. Cell 1984, 38:65-67. 33. Adrian M, Dubochet J, Lepault J, McDowall AW: Cryo-electron microscopy of viruses. Nature 1984, 308:32-36. 34. Slot JW, Geuze HJ: Gold markers for single and double immunolabelling of ultrathin cryosections. In Immunolabelling for Electron Microscopy. Edited by Polak JA and Varndell IM. Amsterdam: Elsevier; 1984: 129-142. 35. Gowen BE, Buendia B, Reinsch S, Kenney JM, Karsenti E, Fuller SD: The centriole: inside and out. In 13th International Congress of Electron Microscopy. Paris: Les Editions de Physique Les Ulis; 1994:265-266 Received: 18 July 1995; revised 16 October 1995. Accepted: 1 November 1995. 1393