γ-Tubulin is present in acentriolar MTOCs during early mouse development

Almost all animal cells possess a centrosome, a structure generally composed of two centrioles surrounded by a cloud of electron-dense material called the peri-centriolar mater-ial (PCM). The PCM nucleates the cytoplasmic network of microtubules during interphase, and the spindle micro-tubules during mitosis. It is thus considered as the micro-tubule organizing center (MTOC; Brinkley, 1985; Karsenti and Maro, 1986). In many differentiated cell types, the rel-ative positions of the nucleus and the centrosome define both an axis of division and a cytoplasmic polarity. Asym-metric patterns of microtubules organized with respect to the position of the MTOCs are implicated in the setting up and maintenance of cell polarity (De Brabander, 1982; Solomon, 1981). Not all animal cells require centrioles to undergo normal bipolar mitosis. These include certain Drosophila and PtK₂ cell lines (Brenner et al., 1977; Debec et al., 1982) and many types of oocytes undergoing meio-sis, including mouse (Carpenter, 1975; Gerhart, 1980; Szöl-lösi et al., 1972). The first stages of mouse development are particularly interesting since cleavage-stage embryos (Szöllösi et al., 1972) are also devoid of centrioles. Cen-trioles have later been found in blastocysts (Magnuson and Epstein, 1984; Szöllösi et al., 1972), but it is not clear when and where they first arise.

The distribution of the microtubules during early mouse development is characteristic of acentriolar cells (Maro et al., 1985; Schatten et al., 1985): the meiotic spindle appears barrel-shaped and non-astral; during the following mitosis, the spindle appears more and more pointed, but is still non-astral. During interphase, the cytoplasmic microtubule (MT) network is homogeneously distributed, and does not seem to be centered on discrete foci of the PCM (Houlis-ton et al., 1987).

Even in the absence of centrioles, aggregates of compo-nents recognized by anti-PCM antibodies are present and electron dense material can be observed at the electron microscope level (Calarco-Gillam et al., 1983; Hiraoka et al., 1989; Houliston et al., 1987; Maro et al., 1985; Szöl-lösi et al., 1972). These are thus commonly considered as MTOCs. Various antibodies against PCM have been used to analyze the distribution of PCM in oocytes and mouse early embryos: the human autoantiserum 5051 (Calarco-Gillam et al., 1983); the monoclonal antibodies MPM-1 and MPM-2, which recognize mitotic phosphoproteins (Davis et al., 1983; Vandre et al., 1984); a serum raised against the mammalian homolog of centrin (Salisbury, 1991); and the NRS-01 serum, a nonimmune rabbit serum found to specifically recognize the centrosome (Hiraoka et al., 1989). All of these stain spindle poles in oocytes and blastocysts (Calarco-Gillam et al., 1983; Hiraoka et al., 1989; Maro et al., 1985). In addition to staining the spindle poles, MPM-2 and 5051 detect cytoplasmic MTOCs in metaphase II arrested oocytes (Maro et al., 1988; Maro et al., 1985; Schatten et al., 1986) and also stain spindle poles in mito-sis during cleavage stages. Moreover, 5051 is the only anti-serum able to detect foci during interphase, which are con-sidered to be aggregates of PCM material. They can be seen in the 1-cell embryo but then are no longer detectable until the 8-cell stage (Maro et al., 1985). At that stage, small foci of positive material found around the nucleus (Houliston et al., 1987) tend to form one or few apical aggregates. Then, usually one focus per cell is seen again during the follow-ing interphase, similar to centrosome staining observed in cell lines.

These data show that some important changes in the structure of MTOCs take place during early mouse devel-opment: firstly, there are changes in the organization of PCM foci, reflected by the fact that many foci are present in oocytes, whereas a single focus is found at each pole during cleavage. Secondly, the composition of the PCM foci changes. The various PCM components are expressed differently during the cell cycle (e.g. those detected by MPM-2 and 5051) or early development (e.g. those detected by centrin and 5051). Cell cycle modulation of the com-position of centrosomes is well known (Kallajoki et al., 1991; Todorov et al., 1992), but changes in PCM structure during development have not been described in other sys-tems.

Early mouse development thus seems to be an excellent model to study the organization of PCM components impli-cated in the nucleation of microtubules in the absence of centrioles. The recently discovered γ-tubulin is an excellent candidate to be true minus-end nucleator of microtubules (Oakley, 1992): genetic arguments show that the γ-tubulin present in MTOCs is likely to interact directly with the β-tubulin found in microtubules (Oakley and Oakley, 1989). Furthermore, γ-tubulin seems to be present in the MTOCs of all eukaryotic cells (Horio et al., 1991; Oakley et al., 1990; Stearns et al., 1991; Zheng et al., 1991) and the inten-sity of γ-tubulin staining during the cell cycle seems to cor-relate with the number of microtubules ending at the cen-trosome (Zheng et al., 1991). It has also been shown that γ-tubulin gene disruption cause aberrant spindle structures (Oakley et al., 1990) and that microinjection with anti-γ-tubulin antibodies prevents nucleation of both interphase and mitotic microtubules (Joshi et al., 1992). Therefore, we have analyzed the localization of γ-tubulin in mouse oocytes and early embryos and, in parallel, we have examined the presence of the centrioles using electron microscopy.

Anti-γ-tubulin serum was raised against the peptide (Y)EYHAA-TRPDYISWGTQEQ, corresponding to the 18 amino acid car-boxyl-terminal sequence of human γ-tubulin (Zheng et al., 1991). This peptide shows no homology with α- and β-tubulins. The pep-tide was synthesized by the standard method of t-Boc amino acids with an automatic synthesizer (Applied Biosystem 430A), puri-fied by HPLC chromatography (Applied Biosystem 151), and its purity checked by analytical HPLC and mass spectrometry. A tyrosine residue has been added to the amino-terminal of the pep-tide, allowing its coupling to the carrier protein keyhole limpet hemocyanin (KLH; Sigma). A rabbit received six injections of 250 mg of the peptide in 50% complete Freund’s adjuvant (Cal-biochem) and the immune serum was collected 99 days later.

Five-to six-week-old Swiss female mice (Animalerie Spécialisée de Villejuif, Centre National de la Recherche Scientifique, France) were superovulated by intraperitoneal injections of 6.25 i.u. of pregnant mare’s serum gonadotrophin (PMSG; Intervet) and human chorionic gonadotrophin (hCG; Intervet) 48 hours apart.

Oocytes and embryos were recovered and manipulated in M2 medium containing 4 mg/ml bovine serum albumin (M2+BSA; Sigma; Fulton and Whittingham, 1978).

Unfertilized eggs were recovered at 12-14 hours post-hCG, by puncturing the ampullae of oviducts. Cumulus cells were dispersed by brief exposure to 0.1 M hyaluronidase (Sigma).

To obtain 1-cell-stage embryos blocked in first mitosis, unfer-tilized oocytes, recovered at 18 hours post-hCG were activated by a 6.5 min exposure to M2+BSA containing 8% ethanol (Cuth-berson, 1983). They were then cultured for 3 hours in M2+BSA containing 1 mg/ml cytochalasin D (Sigma), and the 1-cell-stage embryos exhibiting pronuclei were selected 5 hours after activa-tion. After 7 hours of culture in M2+BSA, they were transferred to M2+BSA containing 10 mM nocodazole (Sigma). Four hours later, they were selected for the absence of pronuclei (corre-sponding to passage and subsequent block in mitosis due to the nocodazole).

To obtain embryos, the females were paired overnight with Swiss males and checked for vaginal plugs the next morning. Under our conditions, fertilization occurred at 12 hours post-hCG and the zygotes were recovered at 15 hours post-hCG. Two-cell-stage embryos were recovered by flushing from the oviduct at 40 hours post-hCG to obtain 2-cell-stage embryos in interphase, or at 48 hours post-hCG for further cultivation. Embryos were cul-tured in medium T6 containing BSA (T6+BSA; Howlett et al., 1987), under paraffin oil at 37°C with 5% CO₂, in plastic dishes (Falcon). 8-cell-stage embryos were selected after 20 hours of cul-ture (68 hours post-hCG), 16-cell-stage embryos after 32 hours of culture (80 hours post-hCG), early cavitating embryos after 44 hours of culture (92 hours post-hCG) and fully expanded blasto-cysts after 64 hours (112 hours post-hCG).

When necessary, zonae pellucidae were removed by a brief incubation in acid Tyrode’s solution (Nicolson et al., 1975).

Zona-free late 4-cell or 8-cell embryos were placed in Ca2+-free M2 containing 6 mg/ml BSA for 15-45 min, during which time they were disaggregated to single blastomeres using a flame pol-ished micropipette. Isolated cells were cultured in Falcon poly-styrene culture dishes in drops of T6+BSA under oil at 37°C in 5% CO₂. Each hour, all newly formed pairs (2/8 or 2/16 pairs, respectively) were removed. Pairs were then cultured in T6+BSA. Quartets of 32-cell blastomeres were derived from 2 out of 16 pairs.

Zona-free embryos were placed in specially designed chambers (Maro and Pickering, 1984) previously coated with 0.1 mg/ml Concanavalin A (Sigma) in phosphate-buffered saline (PBS). The chambers containing samples were centrifuged at 450 g for 10 min at 37°C and then rapidly rinsed in PHEM (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, 2 mM MgCl₂ and 0.6 μM taxol, pH 6.9). In some experiments, the chambers were then treated with M2 supplemented with 6 mg/ml polyvinyl-pyrrolidone (PVP) con-taining either 1.5 μM taxol, for 5 min, or 10 μM nocodazole, for 20 min. Cells were then extracted in PHEM containing 0.25% Triton X-100 for 5 min, then rinsed and fixed in PHEM contain-ing 3.7% formaldehyde for 30 min. For staining with the mono-clonal anti-α-tubulin antibody YL1/2 (Kilmartin et al., 1982), samples were fixed in PBS containing 0.2% glutaraldehyde and 1% Triton X-100 for 10 min at 32°C and then transferred to PBS containing 2% Triton X-100 for 30 min. Immunological staining was performed using the anti-γ-tubulin serum (1/1500 dilution), which had been pre-incubated for 1 hour at 37°C with 10 μg/ml KLH and, occasionally, the immunogenic peptide at 2 μg/ml, or the pre-immune serum (dilution 1/1500) followed by FITC-con-jugated anti-rabbit antibodies (KPL). Chromatin was visualized with propidium iodide (5 μg/ml; Molecular Probes).

Coverslips were removed from the chambers, mounted in Citifluor (City University, London) and examined with a BioRad MRC 600 Confocal Laser Scanning Microscope, mounted on an Optiphot II Nikon microscope (×60 objective Nikon plan apo; NA, 1.4). An argon ion laser adjusted at 488 nm wave length was used for flu-orescein and a helium-neon ion laser adjusted at 543 nm for pro-pidium iodide. The signal was averaged using a Kalman filter (on 10 images). Photographs were taken on a VM 1710 Lucius and Baer black and white high-resolution monitor using Kodak T-Max 100 films.

Embryos were prepared in chambers as above, and pre-extracted for 3 s in PHEM buffer containing 0.5% Triton X-100, then pre-fixed in PHEM containing 0.3% glutaraldehyde and 0.3% Triton X-100 for 5 min. The embryos were then fixed with PHEM con-taining 0.3% glutaraldehyde for 15 min and extracted in PHEM containing 0.5% Triton X-100 for 30 min. Immunocytochemistry was performed as described above, using 10 nm gold-labelled anti-rabbit antibodies. Fixation was then continued in Sørensen buffer (0.1 M Nα₂HPO₄, 0.1 M NaH₂PO₄, pH 7.2) containing 2% glu-taraldehyde and 0.2% tannic acid for 20 min, post-fixed with 0.5% OsO4 in Sørensen buffer on ice and stained ‘en bloc’ in 2% ethanolic uranyl acetate for 15 min. After dehydratation in an ethanol series, samples were flat embedded in Epon. Thin sections (0.12 to 0.2 mm) were cut on a Reichert Ultracut, contrasted with uranyl acetate and lead citrate, and observed on a Philips EM 410 electron microscope at 80 kV.

Eggs and embryos were washed three times in M2+PVP, collected in a small drop of M2+PVP, mixed with the same volume of double-strength SDS sample buffer (Laemmli, 1970) and boiled for 2 min. Proteins were separated using a 10% SDS-polyacry-lamide gel on a microgel apparatus (Hoeffer) and transferred elec-trophoretically (BioLyon) onto nitrocellulose. Reactive proteins were detected after incubation with the anti-γ-tubulin serum (dilu-tion 1/1500), previously preincubated for 1 hour at 37°C with KLH (10 μg/ml) and sometimes with the immunogenic peptide (2 μg/ml). Immunolabelling was revealed either by alkaline phos-phatase-labelled anti-rabbit antibodies (1/7500; Promega; Fig. 1) or by enhanced chemoluminescence (ECL Western Blotting Detection System, Amersham; Fig. 4). Incubation and washing buffers contained 3% milk powder and 0.1% Tween-20. In some experiments, detection with an anti-γ-tubulin serum was followed by detection with anti-α-tubulin antibodies (diluted 1/500; Amer-sham).

The antiserum we used contained antibodies specifically recognizing γ-tubulin in PtK₂, human and Chinese hamster cells (Julian et al., 1993). Several observations indicate that this serum was also specific for γ-tubulin in mouse cells.

Immunofluorescence performed on blastocysts with this serum gave two types of staining; an homogeneous surface staining, and cytoplasmic dots (Fig. 1A). These cytoplas-mic dots were absent when the serum was preincubated with the immunogenic peptide at 2 μg/ml (Fig. 1B), but the surface staining persisted, indicating that it was non-specific. In contrast, the cytoplasmic staining persisted when the serum was preincubated with purified mouse brain tubulin, containing mainly α/β-tubulin dimers and some microtubule-associated proteins. Immunofluorescence per-formed with the pre-immune serum gave no cytoplasmic staining and a much fainter homogeneous surface staining, indicating that the serum had no anti-PCM antibodies prior to immunization. We also observed the absence of cyto-plasmic signal in oocytes and cleavage-stage embryos when the serum was preincubated with the immunogenic peptide at 2 μg/ml, or when the pre-immune serum was used instead of the immune-serum.

Immunogold-electron microscopy with the anti-γ-tubulin serum on fully expanded blastocysts gave a staining around the two centrioles of the centrosome in trophectoderm cells (Fig. 1C,D). This centrosomal localization corresponds to that observed in previous studies in other cells. A few addi-tional gold particles were sometimes found dispersed around the nucleus in the close vicinity of microtubules. These may represent γ-tubulin able to nucleate micro-tubules, not associated with the centrioles.

Immunoblotting performed on 440 fully-expanded blas-tocysts revealed a band at 46 kDa (Fig. 1E). The 46 kDa band was not detected if the immunogenic peptide was included during the incubation with the anti-γ-tubulin serum (2 μg/ml) or if it was incubated with the pre-immune serum instead. A minor band at 54 kDa, persisting when the immunogenic peptide was included during the incubation with the anti-γ-tubulin serum, is also detectable. These results are consistent with those obtained in PtK₂ cells in immunoblotting experiments, where a single band at 46 kDa was also specifically recognized (Julian et al., 1993).

In conclusion, all these control experiments demonstrate that the anti-serum used is specifically directed against γ-tubulin.

Immunofluorescence staining using the anti-γ-tubulin serum revealed a fine band of positive material at each pole of the spindle. The spindle itself, and the chromosomes, were also faintly stained. Confocal microscopy revealed that this band consisted of a linear arrangement of several foci of reac-tive material (Fig. 2A). The anti-γ-tubulin antibodies also detected several dispersed cytoplasmic foci, presumed to correspond to the mouse oocyte cytoplasmic MTOCs (Fig. 2B) already described.

Incubation of the oocytes with 10 μM nocodazole for 20 min completely disassembled the microtubules. In these conditions, reactive foci were still observed (Fig. 2C,D). Some foci, near the chromosomes, presumably corre-sponded to a relocalization of the material found at the spin-dle poles in untreated oocytes. A similar number of dis-persed cytoplasmic foci, not in the proximity of the chromosomes, was found in nocodazole-treated and control oocytes (Table 1).

Incubation in taxol (1.5 μM for 5 min) induces the for-mation of many asters of microtubules, as previously described by Maro et al. (1985). Foci of material were again detected by the anti-γ-tubulin serum both at the spindle poles and in the cytoplasm (Fig. 2E,F). The cytoplasmic foci were larger and more abundant than in untreated oocytes (Table 1). Double staining revealed that the anti-γ-tubulin reactivity was found in the center of all the asters observed with the anti-α-tubulin antibody. The observed increase in the number of γ-tubulin-positive foci in the cyto-plasm of taxol-treated oocytes is likely to result from the aggregation of nucleating material (Verde et al., 1991), thus forming new MTOCs in addition to the pre-existing one.

These observations show that γ-tubulin is present both at the spindle poles and in cytoplasmic MTOCs. The cyto-plasmic MTOCs nucleate asters in the presence of taxol and remain associated with γ-tubulin even in the absence of microtubules.

2-, 4-, 8- and 16-cell-stage embryos were double-stained with propidium iodide to visualize the chromatin (and thus their cell cycle stage) and the anti-γ-tubulin serum. A bright reactivity was found at both poles of the spindle during mitosis (Fig. 3A,E). The spindle pole staining appeared as a single dot, rather than as a band of several foci as it was at the poles of the barrel-shaped meiotic spindle. This specific staining extended slightly along the spindle micro-tubules, close to the poles.

In interphase, no cytoplasmic staining was seen in the blastomeres. The absence of γ-tubulin staining during inter-phase could reflect a cyclic synthesis and degradation of γ-tubulin, cyclic post-translational modifications (preventing its recognition by the antibodies), or a cytoplasmic dispersion during interphase. To eliminate some of these possi-bilities, we performed immunoblots with the anti-γ-tubulin antiserum on samples collected during interphase or mito-sis. Because of the lack of synchrony of division between the different blastomeres within an embryo, which increases during development, we chose to collect samples during the first two cell cycles, in M-phase (meiotic metaphase II-arrested oocytes and 1-cell-stage embryos blocked in the first mitosis with nocodazole) and in interphase (1-cell and 2-cell-stage embryos). We observed γ-tubulin in all sam-ples (Fig. 4) and at equivalent levels with respect to the level of β-tubulin in the same blot (data not shown). Thus, the absence of γ-tubulin immunofluorescence staining in interphase does not seem to involve a post-translational modification that prevents its recognition by the antibodies, or a degradation of γ-tubulin. In contrast, dispersal of γ-tubulin in the cytoplasm could account for the absence of immunostaining in interphase.

Embryos at the 32-cell stage were selected at the beginning of cavitation (Table 1). In these embryos, spindle pole stain-ing was detected as in earlier stages. In addition, a cyto-plasmic dot-like staining could be observed in some blas-tomeres during interphase (Fig. 3B,F). The reactive dots appeared mainly in trophectoderm cells, although some were also found in the inner cell mass. This staining is sim-ilar to that observed in various cell lines containing centri-oles.

In expanded blastocysts, the specific staining of the mitotic spindle poles was still observed, but staining of the spindle microtubules became restricted to a smaller portion of the spindle. A single-dot staining was found in inter-phase in almost all blastomeres and in the two blastocyst cell lineages (trophectoderm and inner cell mass; Fig. 3C,G). The intracellular localization of the staining in the trophectoderm was mainly baso-lateral, and particularly fre-quent in a basal cavity of the nucleus (Fig. 3D,H).

We detected γ-tubulin reactive interphasic dots in cavitat-ing embryos (32-cell stage) whereas centrioles have rou-tinely been observed only at the fully-expanded blastocyst stage (128-cell stage; Magnuson and Epstein, 1984; Szöl-lösi et al., 1972). Therefore, we tried to better characterize the time of appearance of the centrioles. Semi-thick sec-tions were prepared from pairs of 8- and 16-cell blastomeres or quartets of 32-cell blastomeres or embryos between the 32-cell stage and the fully expanded blastocyst stage (about 128 cells). Blastocysts are composed of two different tissues: an outer layer of epithelial cells, the trophectoderm; and the inner cell mass. The stage of blastocyst develop-ment was evaluated according to the size of the blastocoel cavity (Table 2). In a systematic examination of the sam-ples by transmission electron microscope, no centrioles were found before the 64-cell stage (expansion of the blas-tocoel cavity to about 75% of the size of the embryo). The centrioles were found preferentially in the basal part of the trophectoderm cells (Fig. 5), although we cannot exclude the possibility that rare centrioles are present in the inner cell mass at this stage, since exhaustive serial sectioning was not performed (the embryos are about 80 μm in diam-eter).

Immunofluorescence and immunoblotting experiments per-formed with the γ-tubulin antiserum revealed a specific 46 kDa polypeptide associated with spindle poles and cyto-plasmic MTOCs. Since a polypeptide with the same mole-cular mass and the same localization was detected by the same γ-tubulin antibodies in PtK₂ cells (Julian et al., 1993) and since its detection at the electron microscope level con-firmed a PCM localization, we considered that the observed cytoplasmic staining results from a specific labelling of γ-tubulin.

The pattern of immunofluorescence with γ-tubulin anti-serum observed during the three first cleavages is compa-rable to that previously described for PCM antigens recog-nized by the 5051 non-immune serum and by the monoclonal antibody MPM-2 (Calarco-Gillam et al., 1983; Hiraoka et al., 1989; Houliston et al., 1987; Maro et al., 1985). Aggregates of γ-tubulin were found both at the spin-dle poles and in several cytoplasmic foci in oocytes. In cleavage-stage embryos, they were found only at the spin-dle poles in mitotic blastomeres. The extent of the staining of the spindle itself differs with the various antibodies. Anti-γ-tubulin antibodies stained the extremities of the spindle whereas only the poles were stained with the 5051 serum, and the MPM-2 monoclonal antibody stained the whole spindle (Houliston et al., 1987; Maro et al., 1985; Maro et al., 1988). In addition, no interphase staining with anti-γ-tubulin serum was detected before the 32-cell stage, whereas staining with the 5051 antiserum appeared gradu-ally during interphase of the 8-cell stage (Houliston et al., 1987).

The detection of γ-tubulin at interphase by immunoblot-ting but not by immunofluorescence suggested that it prob-ably disperses in the cytoplasm during interphase, before the 32-cell stage. It is noteworthy that, at these stages, microtubules do not seem to be nucleated by discrete MTOCs in interphase (Houliston et al., 1987; Houliston et al., 1989). This detection of γ-tubulin at interphase by immunoblotting is consistent with the idea that γ-tubulin plays a role in the nucleation of the microtubules in all phases of the cell-cycle (Oakley, 1992). Cell cycle-depen-dent changes in the location of other PCM components have also been reported, for instance the SPN and NuMA anti-gens are centrosomal only during mitosis (Kallajoki et al., 1991; Lydersen and Pettijohn, 1980).

In some of the outer cells at the 32-cell stage, a well-focused aggregate of γ-tubulin persisted throughout inter-phase. By the 64-cell stage, γ-tubulin aggregates were found in interphase in both cell lineages. This indicates that stable aggregates form in almost all cells during the course of a single cell cycle.

Immunoblotting experiments showed an almost identical amount of γ-tubulin in metaphase II oocytes and fully expanded blastocysts (not shown), suggesting that each blastomere contains about 100 to 200 times less γ-tubulin than the oocyte. Despite this greater amount of γ-tubulin in the oocyte, the immunofluorescence staining observed during metaphase is no stronger than that observed at later stages, indicating that the oocyte probably contains a stock of unaggregated γ-tubulin. There is also a pool of γ-tubu-lin in Xenopus eggs, since the sperm centrosome can recruit γ-tubulin in Xenopus egg extracts (M.-A. Félix, personal communication).

From our systematic study at the electron microscope level, we have determined that centrioles appear at the 64-cell stage in trophectoderm cells. At least two PCM compo-nents, γ-tubulin and 5051, organize into interphase com-plexes prior to the appearance of structural centrioles. An ultrastructural analysis of the γ-tubulin reactivity at the 32-cell stage may allow the identification of some centriole precursors in early mouse embryos, and thus shed some light on the mechanisms involved in de novo centriole appearance.

The γ-tubulin labelling of the PCM surrounding the centrioles observed by electron microscopy at the blastocyst stage suggests that centrioles are secondarily constructed within aggregates of PCM components. This is consistent with the idea that centriole location may be controlled directly or indirectly by the PCM itself (Mazia, 1984; Sluder et al., 1990). The centrioles have a microtubular skeleton. Thus, it is possible that they are nucleated de novo through the influence of PCM. An equivalent process could be involved in the assembly of pro-centrioles during each cell cycle (Kuriyama and Borisy, 1981).

In Xenopus eggs, components of the MTOCs are stored in the egg cytoplasm in sufficient amounts to allow numer-ous rapid cleavages (Buendia et al., 1992; Gard et al., 1990).

A centriole, either from the sperm or provided experimen-tally, is required to form normal mitotic spindles and cleav-age furrows (Klotz et al., 1990). Thus, Xenopus eggs differ from mouse eggs in that they do not have the capacity to organize the PCM components into MTOCs capable of organizing a spindle independently of the centrioles.

The presence of cytoplasmic MTOCs not associated with the meiotic spindle is a peculiarity of the mouse oocyte. γ-Tubulin is observed both at the spindle poles and in these cytoplasmic foci. Similarly, γ-tubulin has been observed in the cytoplasm without direct connection with the centro-some in the midbody of PtK₂ cells, a structure that can also be considered as an MTOC (Julian et al., 1993). In fission yeast, γ-tubulin has been observed in another putative MTOC, at the cell equator during cytokinesis (Horio et al., 1991). Thus γ-tubulin could be a universal component of MTOCs and may be responsible for their nucleating activity.

γ-Tubulin persists in cytoplasmic foci in nocodazole-treated oocytes, as previously shown for the 5051 antigens (Maro et al., 1985). This observation is in agreement with the presence of γ-tubulin in the centrosome of nocodazole treated cell lines, and in the spindle pole bodies in fission yeast during interphase (Horio et al., 1991; Stearns et al., 1991).

In taxol-treated oocytes, γ-tubulin was found in the center of the many cytoplasmic asters induced by taxol. Taxol is known to lower the critical tubulin concentration for tubu-lin assembly (Schiff et al., 1979). In interphase cells it induces the formation of microtubule bundles (De Braban-der et al., 1981), whereas in mitotic or meiotic cells it induces microtubule asters (Buendia et al., 1990; Heide-mann and Kirschner, 1975; Maro et al., 1985). Taxol also induces the formation of many asters in cytoplasmic extracts derived from Xenopus eggs (Verde et al., 1991). These asters do not grow from preformed nucleation cen-ters but microtubules aggregate and induce the reorganiza-tion of PCM material through the action of microtubule-associated motor proteins. We observed more asters (centered on γ-tubulin-positive foci) in taxol treated-oocytes than γ-tubulin-positive foci in control oocytes. This obser-vation suggests that, here too, taxol induces the formation of asters independently of preformed centers as well as inducing nucleation by pre-existing MTOCs. PCM compo-nents such as 5051 (Maro et al., 1985) and γ-tubulin (this study) are aggregated at the center of these asters. It is likely that at least some of the material was probably recruited from a soluble stock of molecules (see above).

The cytoplasmic foci observed during interphase in blas-tocysts may be involved in microtubule nucleation. How-ever, the cytoplasmic microtubule network in interphase blastomeres is very dense and is not obviously nucleated by a single MTOC (Houliston et al., 1987; Houliston et al., 1989). A similar conclusion was reached when blastocysts were double-labelled for γ-tubulin and α-tubulin and observed under the confocal microscope (data not shown).

We consistently obtained a specific staining associated with the minus-end regions of the spindle microtubules in mitotic cleavage-stage blastomeres with the γ-tubulin antiserum. This staining disappeared when the serum was pre-incu-bated with the peptide but not with α/β-tubulin. No other kind of dispersion of the fluorescent signal is observed around the spindle pole, so we think that this staining rep-resents the association of γ-tubulin with the spindle micro-tubules and that the amount of γ-tubulin quantitatively decreases with distance from the spindle pole. γ-tubulin was not detected in Xenopus microtubules in vivo or in vitro (Stearns et al., 1991). One model for the function of γ-tubu-lin is that aggregates of γ-tubulin attached to the MTOCs nucleate the assembly of microtubules by direct interactions with the β-tubulin subunit of tubulin dimers at their minus end (Oakley, 1992). The staining of the minus ends of the spindle in cleavage-stage embryos could correspond to a more diffuse distribution of the microtubules minus ends at the spindle poles in the absence of centrioles.

Alternatively, the expanded distribution of γ-tubulin in comparison with its restrictive centrosomal localization in other systems may be due to the presence of a large pool of soluble γ-tubulin in the early embryos. The observation that staining of the minus ends of the spindle becomes less pronounced as development progresses is consistent with this hypothesis.

Some characteristics of mouse early development could be linked to the absence of centrioles (for a review see Gueth-Hallonet and Maro, 1992): (1) the absence of regular cleav-age planes during early development; (2) the absence of any detectable axis of polarity within the blastomeres before the 8-cell stage (Houliston et al., 1989; Johnson and Ziomek, 1981; Ziomek and Johnson, 1980); and (3) the random position of the spindle relative to the axis of polar-ity within the blastomere during the fourth cleavage (Pick-ering et al., 1988). Centrioles may act to keep PCM com-ponents in a precise position throughout the cell cycle and so be useful in the control of the position of the axis of polarity and division. This may become more important in differentiated cells, such as those found in the outer layer of the blastocyst.

We thank Marie-Anne Félix, Evelyn Houliston and Nicola Win-ston for critical reading of the manuscript, and Richard Schwartz-mann and Gérard Géraud for their expert photographic work. We are grateful to Dr Guénard for the gift of taxol and Dr Kilmartin for the gift of the YL1/2 antibody. This work was supported by grants from the Institut National pour la Santé et la Recherche Médicale, the Ligue Nationale contre le Cancer, the Association pour la Recherche contre le Cancer and the Fondation pour la Recherche Médicale to B.M.