From meiosis to mitosis – the sperm centrosome defines the kinetics of spindle assembly after fertilization in Xenopus

Once every cell cycle, the bipolar spindle assembles to faithfully segregate the genetic material to the daughter cells. In most animal cells, the establishment of spindle bipolarity is aided by efficient mechanisms driving centrosome separation, thereby defining the two spindle poles and their positions. Spindle bipolarization, however, also occurs in the absence of centrosomes. A chromosome-dependent pathway triggers microtubule assembly in a centrosome-independent manner and drives microtubule organization into a bipolar spindle (Carazo-Salas et al., 1999; Heald et al., 1996). This pathway is essential in cells naturally devoid of centrosomes, like vertebrate oocytes (Dumont et al., 2007), but also in most other cell types that do contain centrosomes (Gruss et al., 2002; Kalab et al., 2006).

A crucial transition in the dynamics of spindle assembly occurs after fertilization of the vertebrate oocyte. Upon fertilization, the acentrosomal egg receives from the sperm cell the male chromosomes and the single centrosome that acted as the basal body of its flagella. Therefore, the first embryonic division occurs under unique conditions that most probably require integration of the strong and dominant chromosome-dependent pathway that is supported by the egg cytoplasm with the presence of two new microtubule organizing centres (MTOCs) – the duplicated male centrosome. This integration must drive the formation of a bipolar spindle with the centrosomes correctly positioned at the spindle poles to ensure the faithful segregation of the chromosomes and the centrosomes to the first two daughter cells of the embryo. It is now clear that the correct segregation of the centrosomes is essential for the development of an adult organism because they are required for multiple functions and, in particular, for the assembly of cilia (Basto et al., 2006).

Although previous work has shown that the components and the regulation of the spindle assembly machinery transits gradually from the meiotic to the mitotic state (Courtois et al., 2012; Wilbur and Heald, 2013), it has also been proposed that a crucial transition from meiotic to early embryo spindle assembly occurs after fertilization (Kubiak et al., 2008). The dynamics of spindle assembly during this crucial transition has mostly been addressed in mice embryos. However, because in that system all the early cell divisions occur in the absence of centrosomes, it cannot provide insights into their integration in the mechanism of spindle assembly after fertilization (Courtois et al., 2012).

Here, we investigated the transition from the assembly of the meiotic acentrosomal spindle to the assembly of the embryonic spindle that occurs in the presence of centrosomes in the fertilized egg. To investigate this process further, we took advantage of the Xenopus laevis egg extract system to mimic in vitro the assembly of the first embryonic spindle and to dissect the relative contributions of the centrosomes and the chromosome-dependent pathway in the establishment of spindle bipolarity (Karsenti and Vernos, 2001).

We found that the centrosomes dictate the kinetics of spindle bipolarization over the dominant chromosome-dependent pathway. This mechanism is linked to the correct positioning of the centrosomes to the two spindle poles, thereby ensuring their faithful segregation and inheritance.

Xenopus laevis females lay eggs that are naturally arrested in metaphase of meiosis II by the cytostatic factor (CSF) (Masui and Markert, 1971). Cytoplasmic extracts prepared from these eggs (CSF extracts) maintain the cell cycle arrest and provide a good experimental system for studying various aspects of spindle assembly in vitro (Karsenti and Vernos, 2001). Here, we used this system to study the transition from a chromosome-dependent spindle assembly mechanism occurring in the oocyte to a combined mechanism including the male duplicated centrosomes for the assembly of the first embryonic spindle after fertilization (Fig. 1A). To mimic this transition, Xenopus sperm nuclei were added to a CSF egg extract that was released into interphase through addition of Ca²⁺ (Fig. 1A). Like in the fertilized egg, the sperm-associated centrosome duplicates and the DNA replicates during interphase, and upon mitotic entry after addition of CSF egg extract, the bipolar spindle assembles, recapitulating in vitro the assembly of the first embryonic spindle (hereafter referred to as ‘cycled extract’; Fig. 1A). In agreement with previous reports (Boleti et al., 1996), spindle assembly in the cycled egg extract transited through a monopolar configuration that peaked after 30 min (69.5±2.5% of spindles were monopolar; mean±s.e.m.) and were mostly organized in a bipolar manner at 60 min (71.5±3.5% of spindles were bipolar) (Fig. 1B), defining a time-dependent organization process that reaches a steady state with a given rate that we define as the kinetics of spindle bipolarization.

We first evaluated the respective contributions of the centrosomes and the chromosomes in spindle assembly by monitoring independently the kinetics of microtubule assembly and organization induced by each pathway (Fig. 1A,C). In agreement with previous reports (Verde et al., 1990), centrosomes that had been purified from human cells formed microtubule asters in CSF extracts, reaching a steady state after 5 min (Fig. 1C). However, these asters did not interact with each other, precluding the formation of more complex structures. We then supplemented a CSF extract with RanQ69L-GTP (a mutant form of the small Ran GTPase that cannot hydrolyse GTP) to mimic the chromatin-dependent pathway, as previously described (Carazo-Salas et al., 1999). The kinetics of microtubule assembly under these conditions were more complex than in the presence of centrosomes. The first microtubule asters appeared after 10 min of incubation and evolved within 5 min into asters that had a higher density of shorter microtubules. After 20 min, the asters started to interact and organize bipolar structures (mini-spindles) that increased in proportion over time (Fig. 1C and data not shown) (Carazo-Salas et al., 1999; Heald et al., 1996; Sardon et al., 2008). Therefore, although both the Ran-GTP-dependent pathway and the centrosomes trigger the formation of microtubule asters, only those generated by Ran-GTP organize into bipolar structures within 20–25 min.

Altogether, these results indicate that the centrosomes and the chromatin Ran-GTP-dependent pathway individually or in combination promote different kinetics of microtubule assembly and organization. Although the centrosomes do not promote bipolarity on their own, the chromatin Ran-GTP-dependent pathway drives microtubule organization into a bipolar configuration faster than sperm nuclei, which combine both pathways (Fig. 1B,C).

To define more specifically the integration and/or contribution of the centrosome in the chromosome-driven spindle assembly pathway, which also includes the activity of the chromosome passenger complex (CPC) (Maresca et al., 2009; Sampath et al., 2004), we incubated Xenopus sperm nuclei in CSF extract (Fig. 1A,D) to follow microtubule organization in the presence of chromosomes and the single sperm centrosome. The sperm-associated centrosome is atypical in structure and protein composition in many animals and insects (Avidor-Reiss et al., 2015), although it has two centrioles and it functions as an MTOC upon fertilization. Indeed, the Xenopus sperm centrosomes nucleated microtubules efficiently soon after incubation in CSF extract, like the human centrosomes (Albee and Wiese, 2008). After 30 min, a majority of monopolar spindles had formed around the sperm nuclei with only few bipolar spindles appearing after relatively long incubation times (Fig. 1D). Live-imaging experiments have previously reported that these bipolar spindles arise through spontaneous pole splitting or new pole self-organization (Mitchison et al., 2004). Those observations together with our results indicate that the chromosome-dependent pathway can efficiently drive bipolar spindle assembly, but the presence of a single centrosome imposes a constraint that limits this organization (Heald et al., 1997). Alternatively, or in addition, other factors – including the type of chromatin (non-replicated versus replicated) and the absence or presence of sister kinetochores – might also be at play.

Altogether, we conclude that the chromosome-dependent pathway activates one or several factors that drive the bipolar organization of microtubules, whereas the centrosomes change the kinetics of spindle bipolarization in a dominant manner. This suggests that an important adaptation occurs during the transition from meiotic to mitotic spindle assembly after fertilization, when the egg proceeds to the first embryonic division. These results also point to the importance of understanding the respective roles of the centrosomes and the chromosome-dependent pathway in spindle bipolarization in early embryos and in somatic cells.

To further explore spindle bipolarization during the first embryonic cycle, we decided to focus on the Xenopus TACC3-family member TACC3 (also known as Maskin; hereafter referred to as XTACC3). We and others have previously shown that XTACC3 promotes microtubule assembly at the centrosome in Xenopus egg extracts. Indeed, depletion of XTACC3 results in a significant reduction of microtubules by 40% in the centrosome asters, but it does not impair bipolar spindle formation in cycled egg extracts (Kinoshita et al., 2005; Mortuza et al., 2014; Peset et al., 2005; Sardon et al., 2008). We first incubated Xenopus sperm nuclei carrying one centrosome in mock- (control) or XTACC3-depleted CSF extracts (Fig. 2A; Fig. S1A). Strikingly, after 60 min of incubation, a significantly higher proportion of bipolar spindles assembled in XTACC3-depleted extracts than in control extracts (respectively 49.0±5.9% and 5.6±2.3%; mean±s.e.m.; Fig. 2B). To test whether XTACC3 depletion also affects spindle bipolarization in the presence of two centrosomes, we used cycled extracts.

Interestingly, spindle bipolarization was also favoured in XTACC3-depleted cycled extracts (Fig. 2C). Indeed, 30 min after cycling the extract into mitosis in the presence of Xenopus sperm nuclei, a significantly higher proportion of bipolar spindles were present in XTACC3-depleted extract (42.6±6.8%) compared to in the control extract (19.8±5.0%). This difference was still significant after 45 min with 55.7±3.0% of bipolar spindles in XTACC3-depleted extracts and 39.9±5.1% in control extracts (Fig. 2C; Fig. S1B). However, by 60 min, the proportion of bipolar spindles in control and XTACC3-depleted extracts was similar (64.7±7.2% and 63.0±6.1%, respectively) (Fig. 2C). These phenotypes were specific for the XTACC3 depletion as adding back recombinant XTACC3 at endogenous concentrations (≈100 nM) to the depleted extracts fully rescued the kinetics of microtubule organization both in CSF and in cycled egg extracts (Figs S2A,B and S1C,D). These results suggest that the kinetics of bipolar spindle assembly can vary and that the time required for this process is an adjustable parameter. We then investigated how the depletion of XTACC3 might facilitate bipolar spindle assembly.

Because XTACC3 has a role in microtubule stabilization (Peset et al., 2005), we first explored whether a mild destabilization of the microtubules could change the kinetics of spindle bipolarization. We therefore monitored microtubule destabilization and spindle assembly upon addition of different concentrations of nocodazole to the egg extract. The addition of 0.1 µM nocodazole did not prevent the assembly of bipolar spindles, but the density of tubulin within them was reduced by 45.4±0.48% and their size by 31.7±0.3%, indicating that microtubules were indeed partially destabilized (Fig. S3). We therefore selected this concentration for further studies. The proportion of bipolar spindles that assembled in either CSF or cycled extracts in the presence of 0.1 µM nocodazole was similar to that under control conditions (Fig. 2B,C).

We conclude that the kinetics of spindle bipolarization are adjustable, but they are not directly dependent on microtubule dynamics. To investigate which parameters might define the kinetics of spindle bipolarization, we took advantage of our experimental setup to address how XTACC3 influences these kinetics.

Kinetochore fibres (K-fibres) have been shown to play a role in spindle bipolarization in somatic cells (Toso et al., 2009). Because TACC3 has been shown to be involved in K-fibre assembly in somatic cells (Booth et al., 2011), we first examined whether K-fibres play a role in spindle bipolarization in egg extracts. Spindle assembly was monitored in cycled extracts that had been supplemented with control IgGs (control) or antibody against Nuf2 (Fig. S1E) that have been previously shown to prevent K-fibre formation in Xenopus egg extracts (Loughlin et al., 2011) and that can be used to label the kinetochores with immunofluorescence (Fig. S4A). The quantification of the mitotic structures formed after 30, 45 and 60 min in cycled egg extracts showed no statistical difference in the proportion of bipolar spindles formed at any time point in control and in extracts containing antibodies against Nuf2 (Fig. 3A). These results indicate that, in contrast to that in human somatic cells (Toso et al., 2009), preventing K-fibre formation in egg extracts has no influence on the spindle bipolarization kinetics. Because K-fibres account for only 5% of the total microtubules of the egg extract spindles (Ohi et al., 2007), this result is not unexpected. However, it strongly indicates that the change in bipolarization kinetics observed in XTACC3-depleted extracts is unrelated to K-fibre assembly.

We have previously shown that XTACC3 depletion results in a delay in Ran-GTP-induced microtubule assembly in egg extracts (Sardon et al., 2008). To test whether this could explain the change of the spindle bipolarization kinetics that we observed in XTACC3-depleted extracts, we monitored spindle assembly that was exclusively driven by the chromatin-dependent pathway (in the absence of centrosomes and kinetochores) by incubating chromatin-coated beads (Heald et al., 1996) in mock- (control) and XTACC3-depleted extracts (Figs S1F and S4B). Bipolar spindles assembled around chromatin-coated beads in control and XTACC3-depleted extracts with a similar efficiency (Fig. S4C), although consistent with previous data, the tubulin density of spindles that assembled in XTACC3-depleted extracts was significantly decreased by 25.6±2.5% (mean±s.e.m.) with respect to that in control spindles, without change in the spindle length (Fig. S4D). However, there was no statistical difference in the proportion of bipolar spindles present at any of the time points in control and XTACC3-depleted extracts (Fig. 3B). These results indicate that the kinetics of bipolar spindle assembly are not dictated by the kinetics of the chromosome-dependent microtubule assembly pathway and strongly suggest that another mechanism governs the kinetics of spindle bipolarity establishment. This mechanism could involve the centrosomes and could be particularly relevant during the assembly of the first embryonic spindle.

Because XTACC3 plays a major role in promoting microtubule assembly at the mitotic centrosomes (Albee and Wiese, 2008; Kinoshita et al., 2005; O'Brien et al., 2005; Peset et al., 2005), we then focused on the putative role of the centrosome in spindle bipolarization in egg extracts.

We first decided to follow a complementary approach that was independent of XTACC3 to monitor spindle bipolarization around Xenopus sperm nuclei in the absence of centrosomes. Dynein inhibition in egg extracts has been previously shown to promote the dissociation of the centrosomes from the spindle poles as well as spindle bipolarization in CSF extracts (Heald et al., 1997). We therefore inhibited Dynein by supplementing mock- (control) and XTACC3-depleted extracts with Dynamitin (also known as p50 and Dctn2) (Wittmann and Hyman, 1999).

Addition of p50 to CSF extracts promoted the formation of a higher proportion of bipolar spindles (76.3±4.2%; mean±s.e.m.) than in control conditions. As reported above, XTACC3-depleted extracts contained more bipolar spindles (41.4±7.7%) than control extracts (13.6±8.6%) (Fig. 4B; Fig. S1G). Interestingly, the addition of p50 to XTACC3-depleted extracts did not significantly change the proportion of bipolar spindles (86.1±2.0%) when compared to control extracts containing p50 (76.3±4.2%) (Fig. 4B), suggesting no additive effects. Consistently, because p50 promotes the release of the centrosome from the spindle pole, the effects of p50 on spindle bipolarization were stronger than in the absence of XTACC3 because, under these conditions, the centrosome retains some activity.

These data indicate that ‘removing’ the centrosome (through addition of p50) or reducing its microtubule assembly activity in CSF extracts (by depleting XTACC3) favours the self-organization of the chromosome-dependent microtubules into a bipolar configuration. These data support the idea that the male centrosome provided by the Xenopus sperm nuclei has a dominant microtubule-focusing activity (Heald et al., 1997), interfering with the bipolarization mechanism driven by the chromosome-dependent pathway in CSF extracts.

We next monitored spindle bipolarization in control and XTACC3-depleted cycled egg extracts that had been supplemented with p50. Control extracts had significantly fewer bipolar spindles than p50-containing extracts at 30 min after cycling (21.6±5.1% in control extracts and 47.7±7.0% in extracts containing p50). This difference persisted at 45 min with 60.3±4.9% of bipolar spindles in control extracts and 82.3±5.2% in extracts containing p50. However, at 60 min, the proportion of bipolar spindles was similar in control and p50-containing extracts (82.7±6.7% and 83.1±5.6%, respectively) (Fig. 4C). No significant differences were observed in control and XTACC3-depleted extracts containing p50 (Fig. 4C), suggesting no additive effects on the kinetics of bipolarization in cycled extracts.

Altogether, these data indicate that spindle bipolarization occurs more rapidly when centrosomes are removed from the spindle, suggesting that the centrosomes impose a constraint on the chromosome-dependent self-organization mechanism that drives spontaneous bipolarization. Under natural conditions, this negative constraint might be relieved upon centrosome separation.

Our data suggest that the faster kinetics of spindle bipolarization observed in XTACC3-depleted extracts are most certainly due to a relief of the negative constraint imposed by the centrosome(s). We next addressed the consequences of relieving this constraint on centrosome positioning at the spindle poles. To monitor the localization of the centrosomes in spindles that assembled in mock- (control) and XTACC3-depleted cycled extracts, we performed immunofluorescence analysis with antibodies against Centrin to stain centrioles (Fig. 5A,B). We found that in control extracts, 85.4±6.5% (mean±s.e.m.) of the spindles had one centrosome associated to each spindle pole. In contrast, 49.7±6.6% of the bipolar spindles that had assembled in XTACC3-depleted extracts had centrosome-positioning defects. These defects included the presence of only one centrosome at one spindle pole (29.0±2.3%), two centrosomes at one pole (9.9±2.1%) or no associated centrosome at all (10.9±2.4%) (Fig. 5B).

These data indicate that centrosomes generating few microtubules get either randomly associated to a spindle pole or are lost. This suggests that spindle bipolarization and centrosome positioning at the spindle poles are two closely intertwined processes – by imposing the kinetics of bipolar organization concomitantly with their dominant microtubule-focusing activity, the centrosomes ensure their correct positioning to each spindle pole. This mechanism thereby ensures the correct inheritance of the centrosomes in the first daughter cells and therefore the development of the fertilized egg into a healthy embryo.

We present here a series of experiments performed in Xenopus egg extracts to address the transition from an acentrosomal spindle assembly mechanism driven by the chromosome-dependent pathway to a combined mechanism that includes the centrosome provided by the male gamete upon fertilization (Fig. 1A).

Our data suggest that, in Xenopus laevis, the combination of the centrosome and the chromosome-dependent pathways that occurs upon fertilization impacts the kinetics of spindle bipolarization (Fig. 5C). By specifically targeting the centrosomes using two independent and complementary approaches (XTACC3 depletion and p50 addition), we provide data that strongly suggest that centrosomes have a clear role in the transition from meiotic to mitotic spindle assembly. This transition is not gradual for aspects of the mechanism concerning the bipolarization of the spindle, which is in contrast with what has been shown in mice (Courtois et al., 2012). In fact, we found that, as in somatic cells (Tanenbaum and Medema, 2010), the activity of the centrosomes in promoting microtubule assembly is crucial for establishing the kinetics of spindle bipolarization during this transition (Fig. 5C), ensuring their concomitant and correct association with the two spindle poles (Fig. 5B). In contrast, preventing K-fibre formation or altering the dynamics of the chromosome-dependent pathway in this early embryonic system does not influence the spindle bipolarization kinetics (Fig. 5C). This is somehow surprising because the chromosome-dependent pathway has a dominant role in microtubule assembly in the Xenopus egg extract system (Heald et al., 1996) and its activity significantly alters the kinetics of spindle bipolarization in oocytes (Dumont et al., 2007).

Our results indicate that the centrosomes need to retain a minimal microtubule assembly activity in order to block the spontaneous spindle bipolarization promoted by the chromosome-dependent pathway. This might provide the time for centrosome separation to occur, ensuring that the centrosomes get correctly positioned at each spindle pole, thereby ensuring their faithful segregation during the first embryonic cell division.

These data fit nicely also with the observations that, in the early divisions of the Xenopus embryos, the centrosomes separate at the end of the previous cell division, favouring fast bipolar spindle assembly (Wuhr et al., 2010). Thus, the early separation of the centrosomes might be necessary to avoid the transition through a monopolar spindle and to ensure both fast bipolar spindle assembly and the correct association of the centrosomes to the spindle poles. The early centrosome separation and the fast spindle assembly would be particularly important in the Xenopus early embryos, where the spindle assembly checkpoint is inactive and the cell cycle clock is determined by the oscillation of the maturation-promoting factor (MPF) activity (Newport and Kirschner, 1984).

In this respect, the Xenopus early embryonic cell cycle is different from that of mammals, which have an active spindle assembly checkpoint that is, at least, partially active (Kubiak et al., 2008). Most of the information concerning the dynamics of spindle assembly in early embryos has been obtained using mice as model system (Courtois et al., 2012). However, the early stages of development of rodents is peculiar because the fertilized egg does not contain any centrosome that is generated de novo after the 16-cell stage (Courtois et al., 2012). It will be important to test whether, in embryos of mammals that contain centrosomes, the kinetics of spindle assembly are determined by the centrosomes and whether this is relevant for the inheritance of centrosomes in all tissues.

The centrosome is a very versatile organelle that supports many cellular functions. Although centrosomes are in fact dispensable for spindle assembly, they are important for other aspects of cell division, such as defining spindle orientation and asymmetric cell divisions. Our data indicate that, when present, the centrosomes define the timing of spindle bipolarization, one of the crucial parameters that influence the duration of mitosis.

The centrosomes are also necessary in the developing embryo and the adult organism because they carry the centrioles that are required for cilia assembly. Although flies can develop without centrosomes (Basto et al., 2006), the adult males are sterile, and both males and females have severe problems owing to the lack of cilia in type-I mechanosensory neurons, the only ciliated Drosophila cell type (Basto et al., 2006). In mammals, centrioles are necessary in almost all cell types, and the correct inheritance of centrosomes is therefore essential for healthy development.

Altogether, our results indicate that the male centrosome is crucial in order to favour the transition from meiotic to mitotic spindle assembly, ensuring the correct development of a healthy organism.

Xenopus laevis female and male frogs were purchased from Nasco and were used at an age between 1 and 3 years. All experiments involving animals were performed according to standard protocols approved by the ethical committee of the Parc de Recerca Biomédica of Barcelona, Barcelona, Spain. CSF egg extracts were prepared as described previously (Murray, 1991). To visualize microtubules, Rhodamine-labelled tubulin was added to the egg extract to a final concentration of ≈0.2 mg/ml. Spindle assembly in CSF extract was performed by incubating Xenopus sperm nuclei (500 nuclei/µl) in CSF extract for 60 min at 20°C. For cycled spindle assembly, CSF extract supplemented with Xenopus sperm nuclei was released into interphase by addition of 0.4 mM Ca²⁺. After 90 min of incubation at 20°C, the interphase extract was cycled back into mitosis by adding one volume of CSF extract. After 60 min (or when indicated), spindle assembly was monitored by ‘squashing’ 1 µl of the reaction and 3 µl of fix solution [11% formaldehyde, 48% glycerol, 1 µg/ml Hoechst 33342 in CSF-XB buffer (10 mM Hepes, 100 mM KCl, 0.1 mM CaCl₂, 2 mM MgCl₂, 50 mM sucrose, 5 mM EGTA)] between a 18×18 mm coverslip and a glass slide. For immunofluorescence analysis, the spindle reactions were resuspended in spindle dilution buffer (30% glycerol, 1% Triton-X100, BRB80) and spun down (20 min, 3200 g, 20°C) on a coverslip through a 4-ml spindle cushion (40% glycerol, BRB80). Coverslips were recovered, washed in PBS and fixed for 10 min in freezing-cold methanol.

Chromatin-coated beads were prepared by cutting the MCP1 plasmid with HindIII. The purified DNA was biotinylated by extending the overhangs using the Klenow exo-fragment in the presence of biotinylated dATP and normal dCTP, dGTP and dUTP. The biotinylated DNA was added to washed Streptavidin Dynabeads M280 (Invitrogen) in the proportion of 2 µl of beads per µg of DNA in the presence of coupling buffer (PVA 2.5%, NaCl 1 M, EDTA 2 mM, Tris-HCl 50 mM, pH 8). After overnight incubation on a rotating wheel at room temperature, beads were retrieved. DNA loading was estimated by measuring the unbound DNA at 260 nm. Beads were resuspended in Bead Buffer (2 M NaCl, 1 mM EDTA, 10 mM Tris-HCl, pH 7.6) to give a final concentration of 0.2 µg DNA per µl of preparation. DNA-coated beads were chromatinized (histone assembly and/or modification) through incubation at 20°C in ten volumes of CSF extract for 4 h. 30-µl aliquots were fast-frozen in liquid nitrogen. To promote assembly of spindles, aliquots were rapidly thawed and the supernatants discarded, and the chromatin beads were resuspended in 20–50 µl of fresh CSF extract. After 10 min at 20°C, the extract was cycled into interphase and processed as described for cycled spindle assembly around Xenopus sperm nuclei.

Centrosomes purified from human KE37 cells, as previously described (Bornens and Moudjou, 1998), were incubated in CSF extracts at 20°C, and the resulting microtubule asters analyzed by squashing under a coverslip. KE-37 cells (human, leukaemia, acute lymphoblastic T cell) were originally obtained from Eric Karsenti's laboratory (EMBL, Heidelberg, Germany) and were routinely tested for contamination. Recombinant RanQ69L-GTP was added to CSF extracts at the saturating concentration of 15 µM (Caudron et al., 2005). The reactions were incubated at 20°C for the indicated time and analysed by squashing.

XTACC3 depletion experiments were performed using three successive rounds of XTACC3 immunoprecipitation, as described previously (Peset et al., 2005).

Spindle assembly in the presence of nocodazole (Sigma) was performed by adding the drug to a final concentration of 0.1 µM. Spindle assembly in the presence of anti-Nuf2 antibody was performed by addition of the antibody at the maximum possible concentration (1:10 dilution), as previously described (Loughlin et al., 2011). Rabbit generic IgGs were added to similar levels, calibrating the loading by performing western blot analyses. p50 was added to the extract at ≈1 µg/µl (Wittmann and Hyman, 1999).

Blocking and antibody dilution buffer was 2% BSA (Sigma) with 0.1% Triton-X100 (Sigma). Coverslips were mounted in 10% Mowiol (Calbiochem) in 0.1 M Tris-HCl at pH 8.2 with 25% glycerol (Merck). Samples were visualized with a 40× objective on an inverted DMI-6000 Leica wide-field fluorescent microscope. For Centrin staining, cycled extracts containing spindles (after 60 min of incubation at 20°C) were supplemented with 0.5 µM nocodazole and 15 min later spun down onto coverslips. The samples were fixed in methanol at −20°C for no longer than 5 min. The blocking and antibody incubation buffer contained 5% BSA. Samples were visualized with a 100× objective on an inverted DMI-6000 Leica wide-field fluorescent microscope. Pictures were acquired with the Leica Application Suite software. Images were processed and mounted using Adobe Photoshop CS5.1 (Adobe).

Blots were developed using Alexa-Fluor-680- (Invitrogen) and IRdye-800CW- (LI-COR) labelled antibodies and analysed using the Odyssey Infrared imaging system.

Polyclonal rabbit antibodies anti-XTACC3 and anti-GST were produced in-house against recombinant proteins His–XTACC3 and GST, respectively (Peset et al., 2005). The affinity-purified antibodies were used at 10 μg/ml for immunofluorescence analyses, and at 1 μg/ml for western blotting. The anti-Nuf2 antibody was a gift from Todd Stukenberg (University of Virginia, VA) (Loughlin et al., 2011) and was added to the extracts at 1:10 dilution. For immunofluorescence, it was used at 1:800. The 135 and 137 mouse monoclonal anti-XTACC3 antibodies are custom-made antibodies generated by injecting mice with recombinant His–XTACC3 and were used at 1:20 for immunofluorescences and at 1:200 for western blots (Mortuza et al., 2014). Generic rabbit IgGs were purchased from Sigma (I5006-50MG). The monoclonal anti-Centrin antibody was purchased from Merck Chemicals and Life Science (clone 20H5) and used for immunofluorescence at 1:1000.

Secondary anti-rabbit and anti-mouse antibodies conjugated to Alexa-Fluor-488, -568 or -680 (Invitrogen), or to IRdye 800 CW (LI-COR) were used at 1:1000 for immunofluorescence and 1:10,000 for western blot.

RanQ69L was purified and loaded with GTP as previously described (Brunet et al., 2004). p50 was purified as described previously (Wittmann and Hyman, 1999). GST–XTACC3 and His–TACC3 was purified as described previously (Peset et al., 2005).

The sample size quantified in each experiment and the number of experimental replicates are reported in the figure legends. For the quantifications of the spindle structures in CSF extracts and the centrosome distribution in mock- and XTACC3-depleted extracts, the investigator was blinded to the experiments. Mitotic structures forming around more than one Xenopus sperm nucleus were not considered for quantification. For the measurement of spindle length and microtubule density in spindles assembled in egg extracts, the pictures were taken using the 40× objective and identical camera settings. Images were opened using ImageJ, and spindle areas were drawn manually using the freehand selection tool. Spindle length was measured using the straight-line tool. The statistical analysis of the data was done by performing an independent two-sample t-test in Prism 6 (Graphpad) or Microsoft Excel (Microsoft).

We are thankful to Nuria Mallol and Jacobo Cela for their excellent technical help with protein purification, antibody purification and frog handling. We thank Todd Stukenberg (University of Virginia, USA) for the gift of the anti-Nuf2 antibody. We are grateful to the Vernos lab members for discussions during the development of the project.