γ-Tubulin participates in the formation of the midbody during cytokinesis in mammalian cells

The tubulin superfamily comprises three types of evolu-tionarily related polypeptide chains: α, β and γ (Oakley and Oakley, 1989). The CYβ-tubulin heterodimer constitutes the main component of microtubules. In contrast, γ-tubu-lin seems to be restricted to the mitotic poles: spindle pole bodies in Aspergillus (Oakley et al., 1990) and yeast (Horio et al., 1991), centrosomes in animal cells (Stearns et al., 1991; Zheng et al., 1991). In the yeast Schizosaccha -romyces (Horio et al., 1991), γ-tubulin is also transiently present in a microtubule organizing centre responsible for the nucleation of cytoplasmic microtubules during the course of cytokinesis. In mammalian cells undergoing cytokinesis, the two daughter cells are linked by an inter-cellular cytoplasmic bridge for up to 4 hours after metaphase (Sanger et al., 1985). This organelle contains two microtubule bundles (McIntosh and Landis, 1971), which interdigitate by the microtubule plus ends, forming the midbody (Euteneuer and McIntosh, 1980). A faint staining of the midbody by rabbit polyclonal antibodies raised against human γ-tubulin has been reported, but the significance of this observation has been questioned (Zheng et al., 1991). Disruption of the gene coding for tubulin, in both Aspergillus (Oakley et al., 1990) and Schizosaccharomyces (Horio et al., 1991), or micro-injec-tion of γ-tubulin antibodies into mammalian cells (Joshi et al., 1992) prevented the nucleation of the mitotic spindle. These observations and the association of γ-tubulin with the minus end region of microtubules (Oakley, 1992) suggest that γ-tubulin plays a role in microtubule nucle-ation. Thus, the presence of γ-tubulin in the intercellular bridge linking daughter cells is potentially of interest in order to determine the origin of the microtubule skeleton characteristic of this organelle.

In mammalian cells, we show that γ-tubulin is associated with the microtubule bundles forming the midbody. In anaphase it is absent from the interzone located between the two separating sets of chromosomes and was observed to be associated with the minus ends of the microtubule bundles constituting the midbody in telophase. The maxi-mum amount of γ-tubulin at the ends of the midbody is reached during the early part of cytokinesis. Gamma-tubu-lin becomes undetectable when the intercellular bridge elongates between the two separating cells. In agreement with these observations, micro-injection of γ-tubulin anti-bodies during the anaphase stage prevents the subsequent formation of the midbody between the two separating cells.

Synthesized peptides (Fig. 1; Applied Biosystem 430 A, on 4-(oxy-methyl) phenylacetamido methyl resin) (Merrifield, 1963, 1964) have been purified by HPLC chromatography (Applied Biosystem 151, column Aquapore RP300, C8, 100 mm × 10 mm, 0.1% trifluoroacetic acid and elution by a linear gradient of 30% to 70% CH₃CN in 0.1% trifluoroacetic acid at 3.5 ml/min for 40 min). Peptide purity has been checked by analytical HPLC (Column Aquapore RP 300, C8, 220 mm × 4.6 mm), and the mass of their protonated molecular ions has been determined (Fast Atom Bombardment, ZAB-HS double focusing with a VG analytical mass spectrometer, Manchester, UK): (Y)38-50 MH⁺=1615.5, (Y)323-338 MH⁺=1899.7, and 435-452 MH⁺=2155.

Peptides (5 mg) were coupled to 7 mg keyhole limpet haemo-cyanin (peptide 435-452 in 5 ml of 0.1 M NaHCO₃, pH 8.4, by overnight incubation with 12 μl of 25 % glutaraldehyde; and pep-tides (Y)38-50 and (Y)323-338 in 3 ml 0.16 M borate buffer, pH 7.8, with 0.6 ml of 24 mM bis-diazo-O-tolidine for 4 h at 0°C), and dialysed for 24 hours against 0.15 M NaCl. Rabbits were repeatedly injected with 200-250 μg of peptide (≈ 0.5 mg of pro-tein) in 50% complete Freund’s adjuvent (Calbiochem) and the titers of the sera were followed by fluorescence-labeling of PtK₂ cells (Potorous tridactylis).

PtK₂ cells, resuspended in 10 mM PIPES (1,4-piperazine diethane sulfonic acid), pH 8.0, 50 mM EDTA (ethylenediamine tetraacetic acid), 0.1% Triton X-100, 0.25 M NaCl, 0.1 mM PMSF (phenyl-methylsulfonyl fluoride), 1 μg/ml leupeptin, 50 μg/ml TLCK (Na-p-tosyl-L-lysine chloro-methyl ketone), 100 μg/ml TPCK (N-tosyl-L-phenylalanine chloromethyl ketone), 1 μg/ml pepstatin, 1 μg/ml chymostatin, 1 mM dithiothreitol, were disrupted by soni-cation, and the extract was centrifuged for 5 min at 15,000 r.p.m. After denaturation, the proteins were submitted to SDS-PAGE (7.5% gel), electro-transfered to a nitrocellulose membrane (Hybond C, Amersham), and immunolabeled (Enhanced chemo-luminescence western blotting detection system, Amersham). Apparent molecular masses of the polypeptides labeled by the antibodies were determined by comparison with molecular mass markers: bovine carbonic anhydrase (31 kDa); ovalbumin (45 kDa); brain tubulin (50 kDa); bovine serum albumin (66 kDa) and phosphorylase b (97 kDa).

Cells fixed for 60 min in 2% formaldehyde in PBS, were washed (3×5 min in PBS), incubated for 60 min in 2% glycine in PBS and washed. After 6 min in acetone (−20°C) and washing, cells were reacted for 90 min at 37°C with primary antibodies (in PBS with 25% fetal calf serum), washed (2×10 min with 0.05% Tween-20 in PBS and once in PBS). The staining was revealed with goat anti-rabbit or anti-rat antibodies (Nordic) in PBS with 40% fetal calf serum. Both simple labeling of γ-tubulin (with fluorescein or rhodamin-labeled secondary antibodies) and double labeling of γ- and CY-tubulin (with fluorescein and rhodamin-labeled secondary antibodies, respectively, or the reverse combination) were per-formed. Nuclei and chromosomes were stained with 4′,6-diamidino-2-phenylindole (0.2 μg/ml; DAPI). Preparations were dried, mounted in Mowiol (Rodriguez and Deinhardt, 1960), sealed, and observed by epifluorescence with a Zeiss Axiophot equipped with 40× and 100× Plan-neofluar objectives (NA: 0.90 and 1.30, respectively), an Optovar varying from 1.25× to 2.5×, a 4× TV camera adaptator and a stabilized exciting beam. Images recorded in the linear dynamic range by a Nocticon camera (LH 4015 from Lhesa), were digitized (50-200 frames) with an image processing system (Sapphire from Quantel) and recalculated, with-out substracting the fluorescence background, using a linear func-tion (‘stretch’ program). Fluorescence intensities (‘luminance’ program) and organelle surfaces (‘threshold’ program) were determined in order to calculate integrated fluorescence values for each organelle. Cells grown on photoetched coverslips (Bellco Glass), fixed at different times after early anaphase, and processed for flu-orescence staining, were used to determine the morphogenetic evolution of the midbody.

PtK₂ cells were washed in 80 mM PIPES, pH 6.9, 5 mM EGTA, 1 mM MgCl₂ (PEM), permeabilized for 5-15 s in PEM with 0.1-0.5% Triton X-100, washed (3 min in PEM), fixed (PEM with 1% glutaraldehyde, 30 min), and incubated with 2% glycine (2×30 min). Then, cells were treated for immunocytochemistry (immune serum from rabbit 75, 1/500, 37°C, 3 hours; gold-labeled sec-ondary antibody, 37°C, 90 min), post-fixed with glutaraldehyde (2%, 14 hours) and processed for electron microscopy.

Immunoglobulins (14 mg/ml) from the immune serum 70 (sequence 323-338) were purified with Protein A immobilized on agarose beads (Pierce) and centrifuged before micro-injection (Airfuge Beckman, 4°C, 5 min, 18 lbf/in²; 1 lbf/in² ∼ 6.9 kPa, 75,000 r.p.m.). PtK₂ cells were micro-injected in early anaphase using a Zeiss inverted microscope (Axiovert with a 32× Achrostigmat phase-contrast objective, NA 0.40) equipped with a heating stage, a micromanipulator and a micro-injector Eppendorf (40-50 hPa for 0.3 s). After a further incubation of 30 min, cells were fixed and processed to label the microtubule cytoskeleton (YL1/2 rat monoclonal antibody; Kilmartin et al., 1982) and chro-matin (DAPI).

Polyclonal antibodies were prepared against three specific and distinct sequences of γ-tubulin (Fig. 1). All immune sera failed to detect pure sheep brain α-or β-tubulin after immunoblotting. Except for the immune serum raised against the sequence 38–50 (rabbit 38), all three other immune sera specifically revealed a polypeptide chain with an apparent molecular mass of 46 kDa in PtK₂ cell extracts (Fig. 2), in agreement with the reported value for human γ-tubulin (Zheng et al., 1991). However, the immune sera from rabbit 70 and rabbit 78 specifically labeled additional polypeptides (Fig. 2). The serum obtained from rabbit 75, against the carboxyl-terminal amino acid sequence of human γ-tubulin, showed the best specificity and stained a single polypeptide chain of 46 kDa.

PtK₂ cells were stained by the immune serum against the carboxyl amino acid sequence of human γ-tubulin (rabbit 75) and the corresponding antibodies purified on a peptide-Sepharose affinity column. Staining labeled the centrosome during interphase and the spindle poles during all stages of mitosis, in agreement with the previously published local-ization (Stearns et al., 1991; Zheng et al., 1991). Further-more both serum- and affinity-purified antibodies labeled some, but not all, midbodies (Fig. 3A,G), in contrast with previous reports. The staining of the centrosome, spindle poles and midbody was specific. No staining was observed with the pre-immune serum (Fig. 3B) or when the immune serum was pre-incubated with the peptide 435-452 (Fig. 3C), or with Xenopus γ-tubulin (compare Fig. 3D and E). In contrast, the staining was not modified when the immune serum was pre-incubated with the carrier protein or with purified sheep brain αβ-tubulin (Fig. 3F). Identical obser-vations have been made with the other immune sera (rab-bits 38, 70 and 78). Thus, four polyclonal antibodies raised against three distinct and specific regions of γ-tubulin specifically labeled the centrosome, the spindle poles and the microtubule bundles involved in the structure of the midbody. The location of γ-tubulin in the microtubule bun-dles of the midbody is not restricted to PtK₂ cells. It has also been observed in human breast adenocarcinoma (MCF7), Chinese hamster lung cells (CHL) (Fig. 3I and J), and J714.2 mouse cells. In these cells the labeling was fainter than in PtK₂ cells. It is likely that γ-tubulin inter-acts with other proteins (Oakley and Oakley, 1989; Stearns et al., 1991); however, our present knowledge is insuffi-cient to decide whether these variations of fluorescence intensity reflect different amounts of γ-tubulin, or a differ-ential accessibility of this protein to antibodies.

The comparative location of γ- and α-tubulin in the inter-cellular bridge has been followed in PtK₂ cells by double immunolabeling using the immune serum from rabbit 75 and a rat monoclonal antibody against α-tubulin (Kilmartin et al., 1982). In both cases, the fluorescent staining was symmetrical, in agreement with the ultrastructure of this organelle. However, the two staining patterns were differ-ent (Fig. 4). In early telophase, γ-tubulin was present over a large area located in each daughter cell. This area was coincident with most but not all the microtubule-rich region of the intercellular bridge: the zone devoid of detectable γ-tubulin was larger than the zone corresponding to the inter-digitated midbody microtubules (Fig. 4, upper panel). Later, in typical midbody, the restriction of γ-tubulin labeling to the two minus regions of the microtubule bundles was more striking (Fig. 4, middle panel). This location of γ-tubulin is further supported by the observation that the maximun intensity of each peak of γ-tubulin did not coincide with the maximum intensity of α-tubulin labeling. The maxi-mum intensity of γ-tubulin labeling corresponded to 76% and 52% of the maximum α-tubulin staining in early mid-body (Fig. 4, upper panel) and in typical midbody (Fig. 4, middle panel), respectively. A relocation of γ-tubulin and/or a sliding and growth of midbody microtubules could account for these observations. In tripolar dividing cells, the three extremities of the midbody were labeled (Fig. 3H), in agreement with the overall shape of the midbody in these cells (Keryer et al., 1984). Thus, in all cases, the location of γ-tubulin in the intercellular bridge was preferentially restricted to the minus regions of the microtubule bundles. These observations were confirmed by electron microscopic immunolocation of γ-tubulin using a secondary antibody labeled with gold particles (Fig. 5). No labeling was detected over most of the microtubule bundles (less than 1% of gold particles). Most of the gold particles (89%) were located in the two terminal minus regions of the micro-tubule bundles, both at and between the microtubule extremities. Moreover, 11% of gold particles were present in the outer region of the microtubule bundles, presumably revealing shorter microtubules diverging from the main bundles. This overall ultrastructural location of γ-tubulin in the minus region of microtubules is consistent with previ-ous reports (Oakley et al., 1990; Horio et al., 1991; Stearns et al., 1991; Zheng et al., 1991). However, it is not possi-ble to decide whether γ-tubulin is strictly located at the minus microtubule ends or in the minus region of each microtubule. In later stages (Fig. 4, lower panel), the stain-ing of γ-tubulin was restricted to two small dots. The peak of γ-tubulin staining corresponded to 11% of the maximum staining of α-tubulin, suggesting either that most of the microtubule minus regions were devoid of γ-tubulin or that γ-tubulin was no longer accessible to the antibodies.

In PtK₂ cells undergoing anaphase and early telophase, interzonal microtubule bundles labeled with α-tubulin anti-bodies were present between the two masses of chromo-somes (Fig. 6A and B, right), but the staining of γ-tubulin was negligible in the interzone (Fig. 6A and B, left, and Table 1). When cells progressed into telophase, interzonal microtubule bundles accumulated in the constriction plane (Fig. 6C, right), but did not seem to interdigitate, since there was no interruption of their staining by antibodies against α-tubulin. The diffuse fluorescent staining that was observed with γ-tubulin antibodies on each side of the cleavage furrow increased 15 min after the end of metaphase (Fig. 6C, left) and its overall fluorescence was 2-4 times higher than the fluorescence detected over the centrosome (Table 1C). But most microtubule bundles did not extend from the division plane to these γ-tubulin-rich granules and no interruption of the microtubule staining was observed in the cleavage furrow (Fig. 6C, right). Then, microtubules interdigitated between the two daughter cells, as shown by the presence in the division plane of an area unstained by α-tubulin antibodies (Fig. 6D, right). At this stage, they were apparently in contact with the granules stained by γ-tubulin antibodies, which increased in number (Fig. 6D, left), and constituted a large area of apparently filamentous structure (Fig. 6E-G, left). The spindle poles remained labeled by γ-tubulin antibodies and the fluores-cent staining of γ-tubulin on the extremities of the mid-body was twice that observed on the centrosomes (Table 1). When the intercellular bridge began to elongate between the two daughter cells, i.e. from 45-60 min after the end of metaphase, the areas stained by γ-tubulin antibodies were restricted to two small regions on each side of the midbody (Fig. 6H, left). Thus, γ-tubulin-rich material seemed to accumulate in the two daughter cells mostly independently from the interzonal microtubule bundles present in the cleavage furrow, and interacted with the microtubule bun-dles present in the intercellular bridge only when a typical midbody was assembled.

In later stages, the staining of γ-tubulin remained restricted to two small dots (Fig. 6I, left), although the mid-body itself was strongly labeled by α-tubulin antibodies (Fig. 6I, right). In the same cells, the intensity of the fluo-rescence labeling of γ-tubulin remained high at the centro-some level (Table 1I). Eventually, from 45–60 to 240 min after the end of metaphase, as midbody microtubules elongated in the intercellular bridge (Fig. 6J, right), γ-tubu-lin was no longer detectable at the extremities of this organelle (Fig. 6J, left) although it remained present in the centrosome (Fig. 6J, left, arrowhead, and Table 1). Thus, γ-tubulin was present in the microtubule bundles consti-tuting the midbody from 15 min to 45–60 min after the end of metaphase. But, as the intercellular bridge elongated between the two daughter cells, 45–60 min after the end of metaphase, γ-tubulin was no longer detectable, although a typical midbody was present for 3 hours after cytokinesis and γ-tubulin was detectable in the centriolar region.

γ-Tubulin present at the extremities of the microtubule bundles of the midbody could either be linked to already assembled microtubules, nucleate new microtubules, or interact with previously or newly assembled microtubules. Antibodies directed against the sequence 323–338 of γ-tubulin were injected into PtK₂ cells undergoing anaphase. After a 30 min incubation, the microtubule cytoskeleton and chromatin were labeled in order to determine whether the injection prevented the formation of the midbody. All con-trol non-injected cells showed a typical midbody 30 min after early anaphase, although the timing of the morpho-genesis of this organelle differed somewhat from cell to cell (Schulze and Blose, 1984). The injected cells could be clas-sified in two groups. In the first group, cells failed to progress further into mitosis (Table 2): their chromosomes remained condensed and the microtubule cytoskeleton was either blocked in early telophase or highly disturbed (Fig. 7A and B). In the other group, cells progressed towards cytokinesis (Table 2) exhibiting various stages of recon-struction nuclei (Fig. 7C and D). These distinct actions could reflect both the injection traumatism (16 % of the cells were blocked by buffer injection) and the uncertainty concerning the exact time of micro-injection, since anaphase is the most dynamic and shortest period of mito-sis. Most of the cells progressing towards cytokinesis did not exhibit a midbody (Table 2; Fig. 7D), in contrast to cells micro-injected with buffer or γ-tubulin antibodies pre-incubated with the peptide 323-338 (10 mg/ml, 2 hours, 37°C). These observations suggest not only that γ-tubulin is transiently present at the extremities of the microtubule bundles of the midbody, but also that it plays a role in the formation of this organelle.

Using antibodies directed against γ-tubulin we demonstrate that this protein is not restricted to the centrosome and the spindle poles in mammalian cells, but is also transiently present at the extremities of the microtubule bundles of the midbody during cytokinesis. It has been emphasized that the staining of a subcellular structure by antibodies raised against a specific protein sequence may be misleading, due to cross-reactivity of the antibodies with different but sim-ilar sequences present in unrelated proteins (Nigg et al., 1982). In order to avoid this problem we used four distinct polyclonal antibodies prepared against three different and specific epitopes of γ-tubulin. In all cases the spindle poles, the centrosomes and the midbodies were specifically labeled. Previous studies (Stearns et al., 1991; Zheng et al., 1991; Joshi et al., 1992) failed to detect γ-tubulin in the interconnecting bridge even with antibodies raised against a peptide corresponding to the sequence 38-54 (Joshi et al., 1992), which overlaps one of the peptides (sequence 38-50) that we have used. In PtK₂ cells the intensity of the flu-orescent staining of γ-tubulin on the extremities of the mid-body was equal to, or even higher than, that observed on the centrosomes in the same cells, but was lower in the midbodies of human and Chinese hamster cells. The tran-sient location of γ-tubulin at the extremities of the mid-body, the lower staining intensity observed in some cell types, and the variation of this labeling with the different sera, are all possibilities that could account for the diffi-culties of observing the presence of γ-tubulin in this organelle.

The origin of the microtubule bundles constituting the midbody is unclear. Numerous proteins co-localize at the level of this organelle. Some proteins have been reported to be present both in the centrosome and in the midbody (Bastmeyer and Russell, 1987; Tousson et al., 1991). In most cases, their role, if any, is still unknown (Tooze and Burke, 1987; Schollmeyer, 1988; He et al., 1991; Zhou et al., 1991). Microtubules constitute the major skeleton of the midbody (McIntosh and Landis, 1971). It has been suggested (McIntosh and Landis, 1971; McIntosh et al., 1975) that they could be remnants of interpolar micro-tubules (Fig. 8, left). Alternatively (Salmon et al., 1976), they could result from the transient nucleating activity of special microtubule organizing centres distinct from the centrosomes (Fig. 8, right). Both possibilities account for the orientation of the microtubules in the midbody, since their minus ends are directed towards the extremities of the microtubule bundles forming this organelle (Euteneuer and McIntosh, 1980). In the first hypothesis, spindle micro-tubules, nucleated at the two poles of the mitotic appara-tus, are expected to leave the polar regions, then aggregate and interdigitate to constitute the midbody (Fig. 8, left). According to microtubule polarity (Euteneuer and McIn-tosh, 1981), γ-tubulin would be present at the minus ends of stable migrating microtubules during anaphase and telophase stages. We failed to detect γ-tubulin in the inter-chromosomal region in late anaphase and early telophase cells. The asynchronous migration of interzonal micro-tubules could perhaps prevent the detection of significant fluorescence over the background. However, no γ-tubulin was present on the extremities of the interzonal micro-tubules located in the cleavage furrow. Our observations support the second possibility (Fig. 8, right). In this model, the interzonal microtubule bundles would not be responsi-ble for the formation of the midbody. The extremities of the interzonal microtubule bundles localized at the level of the division furrow in telophase are apparently not linked to the area stained by γ-tubulin antibodies; they would dis-assemble in agreement with their reported unstability (Salmon et al., 1976), while γ-tubulin would participate in the nucleation of the new microtubules forming the mid-body (Alfa and Hyams, 1991; Oakley, 1992). This view accounts for several observations. First, there is a gradual accumulation and organization of γ-tubulin-containing material on each side of the division plane. Second, the interzonal microtubule bundles initially present between the two daughter cells do not interact with γ-tubulin-rich mate-rial. Third, γ-tubulin is present in the extremities of the microtubule bundles only when they interdigitate to con-stitute the midbody. Fourth, micro-injection of γ-tubulin antibodies perturbs the formation of the midbody. Thus, we suggest the assembly in each daughter cell of transient microtubule organizing centres distinct from the centro-some and playing a role in cytokinesis.

The location of γ-tubulin at the extremities of the mid-body is transient. No γ-tubulin was detected with any of the antibodies in the late stages of cytokinesis, i.e. in the elongated intercellular bridge. However, in the same daugh-ter cells, γ-tubulin was present in the centrosomes. Thus, it is unlikely that the absence of γ-tubulin on the extremi-ties of the elongated midbody results simply from the deple-tion of the pool of active γ-tubulin in G₁ cells. These obser-vations suggest that the activity of the microtubule organizing centres of midbody microtubules are transient. The microtubules constituting the midbody are stable to cold (Brinkley, 1975; Margolis et al., 1990), high pressure (Salmon et al., 1976), and colchicine (Oppenheim et al., 1973), and exhibit a slow turn over (Gorbsky et al., 1990). This stability accounts for their maintenance even without microtubule nucleation as suggested by the absence of γ-tubulin.

In the yeast Schizosaccharomyces (Horio et al., 1991) and in mammalian cells, transient microtubule organizing centres, distinct from the spindle pole bodies and from the centrosomes, could play a role during cytokinesis. In Schizosaccharomyces, cytoplasmic microtubules are nucle-ated from microtubule organizing centres located in the division plane. Their plus ends are directed towards the two daughter cells (Horio et al., 1991). In mammalian cells, the microtubules constituting the midbody are nucleated by microtubule organizing centres located in each daughter cell. Their plus ends are directed towards the intercellular bridge (Euteneuer and McIntosh, 1980). But in both cases γ-tubulin is present in organelles distinct from the spindle poles and is located at the minus end region of micro-tubules. These observations demonstrate that the subcellu-lar location of γ-tubulin varies during the cell cycle, and suggest that γ-tubulin could be a universal element of all microtubule organizing centres.

We have greatly appreciated the support from ‘L’Association pour la Recherche sur le Cancer’, the sequence comparisons per-formed by Dr P. Rouget, the gift of antibodies from Dr J. Kil-martin, of E. coli strain RR1 from Dr T. Stearns, of mammalian cells from Dr J.Y. Charcosset and Dr A. Valette. The photo-graphical help of L. Donnat, the comments and corrections of Dr M. Adelman were helpful in bringing the manuscript to its final form.