Inhibition of ectopic microtubule assembly by the kinesin-13 KLP-7 prevents chromosome segregation and cytokinesis defects in oocytes

Sexual reproduction relies on meiosis, a specialized type of cell division, which generates haploid germ cells or gametes. The genome size reduction that occurs during gametogenesis involves two successive cell divisions, termed meiosis I and II, preceded by a single round of genome replication (Dumont and Brunet, 2010). Chromosome gain or loss during meiosis generates aneuploid embryos after fertilization. Aneuploidy is hence a major obstacle in achieving reproductive success, as the vast majority of embryos formed from aneuploid oocytes are non-viable, leading to miscarriage (Nagaoka et al., 2012).

Accurate chromosome segregation is driven by the microtubule-based spindle. In somatic cells and spermatocytes, spindle microtubules are primarily assembled from the centrosomes, which duplicate once per cell cycle to form the two spindle poles (Walczak and Heald, 2008; Heald and Khodjakov, 2015). Chromosome alignment on the spindle and segregation in anaphase then occur through the interaction between spindle microtubules and kinetochores (Cheeseman, 2014). Chromosome spatial segregation is followed by their definitive physical separation during cytokinesis of the newly formed daughter cells (Green et al., 2012).

The generation of oocytes involves three major adaptations to the classical mechanism of cell division (Ohkura, 2015). First oocytes of most species lack conventional centriole-containing centrosomes (Szollosi et al., 1972). Spindle assembly in oocytes involves specific mechanisms such as chromatin-dependent microtubule assembly (Heald et al., 1996; Dumont and Desai, 2012). The second important adaptation of the meiotic cell division process, as yet only observed in C. elegans oocytes, is kinetochore-independent chromosome segregation (Dumont et al., 2010). The precise mechanism of this atypical segregation is still unclear but involves microtubule-dependent forces exerted on chromosomes (Dumont et al., 2010; Muscat et al., 2015; McNally et al., 2016). The third specific adaptation of oocyte meiosis is polar body extrusion (PBE), which corresponds to an extremely asymmetric partitioning of the oocyte creating a tiny polar body with most of the cytoplasm maintained in the oocyte (Zhang et al., 2008; Dorn et al., 2010; Fabritius et al., 2011; Maddox et al., 2012).

All three of these major adaptations to cell division essential for successful oocyte meiosis involve microtubules. Consequently, spindle microtubule dynamics in oocytes must be tightly regulated both temporally and spatially to successfully execute the meiotic cell division program. This is achieved primarily through the combined activities of microtubule-associated proteins (MAPs) and microtubule motors (Alfaro-Aco and Petry, 2015). Crucial among these are the microtubule-depolymerizing kinesin-13 family members (Walczak et al., 2013). Kinesin-13 proteins use the energy from ATP hydrolysis to depolymerize microtubules, and play essential roles in spindle assembly and chromosome segregation during mitosis (Wordeman and Mitchison, 1995; Walczak et al., 1996; Desai et al., 1999). Human mitotic cells depleted of the kinesin-13 MCAK (KIF2C) assemble spindles with abnormally long and stable microtubules that correlate with a high frequency of chromosome misattachments (Maney et al., 1998; Kline-Smith et al., 2004; Rogers et al., 2004; Domnitz et al., 2012). In mitotic C. elegans embryos, the unique kinesin-13 family member KLP-7 prevents assembly of an abnormally high number of astral microtubules and thus protects against an excessive increase in astral cortical pulling forces (Srayko et al., 2005). Accordingly, in KLP-7-depleted embryos the mitotic spindles break apart during anaphase and sister chromatids separate prematurely (Grill et al., 2001).

During oocyte meiosis kinesin-13 proteins are involved in multiple aspects of cell division, including the control of meiotic spindle length and proper chromosome alignment (Zou et al., 2008; Illingworth et al., 2010; Radford et al., 2012; Do et al., 2014). In the C. elegans oocyte, KLP-7 has been proposed to limit metaphase spindle pole numbers by correcting improper kinetochore-microtubule attachments, but its precise function throughout the two meiotic divisions remains elusive (Connolly et al., 2015; Han et al., 2015). Here, we show the crucial meiotic function of KLP-7 in preventing ectopic microtubule assembly that otherwise leads to spindle disorganization and chromosome segregation defects. We provide the first high-resolution, time-resolved comprehensive view of meiotic divisions in the C. elegans oocyte. Specifically, we show that KLP-7 acts by globally destabilizing microtubules within the meiotic spindle in metaphase and the central spindle in anaphase, as well as throughout the oocyte cell cortex. We also demonstrate that KLP-7 or MCAK activities prevent ectopic cytoplasmic aster formation during mitosis in the C. elegans zygote or in HeLa cells, respectively, when centrosome function is impaired. Our data suggest that the function of kinesin-13 is essential to delimit the proper local assembly of microtubules in dividing cells when centrosome activity is reduced or absent, and thus for accurate spindle assembly.

We sought to test the role of KLP-7 during meiotic spindle assembly. We first verified that our RNAi-mediated depletion strategy led to a strong embryonic lethality phenotype comparable to the klp-7 deletion mutant (hereafter klp-7Δ), and completely removed KLP-7 from oocytes (Fig. S1A,B). We also verified that expressing a functional GFP-tagged RNAi-resistant KLP-7 protein rescued this embryonic lethality (Fig. S1C,D). We then analyzed nuclear envelope breakdown (NEBD) and meiotic spindle assembly in control and KLP-7-depleted oocytes during both meiotic divisions in utero in immobilized worms expressing GFP-tagged β-tubulin or the microtubule minus-end and spindle pole marker protein Abnormal spindle protein 1 (ASPM-1) and mCherry-tagged histone 2B (H2B) (van der Voet et al., 2009). Spindle assembly can be separated into four distinct phases (Wolff et al., 2016). In both control and KLP-7-depleted oocytes before NEBD, microtubules were excluded from the nucleus (Fig. 1A,B, Fig. S2A,B). In controls, after NEBD (evidenced by the diffusion of the soluble pool of fluorescent H2B away from the nucleoplasm), microtubules progressively invaded the nucleus to form a diffuse nuclear cloud around chromosomes (Fig. S2A,B, Movie 1). In KLP-7-depleted oocytes and consistent with the cytoplasmic localization of GFP::KLP-7 in control oocytes at this stage, more microtubules were visible around the nucleus before NEBD relative to controls (Fig. 1A, Fig. S2A,B, Movies 1 and 2). Strikingly, the first phase of spindle assembly never took place in KLP-7-depleted oocytes and the diffuse microtubule cloud that formed at NEBD around chromosomes in controls was completely absent (Fig. 1C,D). Instead, ectopic microtubules persisted after NEBD around the breaking nuclear envelope. KLP-7 is therefore essential for the formation of a microtubule cloud around chromosomes after NEBD.

During the second phase, GFP::KLP-7 progressively accumulated on chromosomes and the assembled diffuse network of microtubules became bundled and coalesced around chromosomes in control oocytes (Fig. 1A,B,E, Movie 2). In parallel, the chromosomes became clustered together. In controls, this second phase correlated with the appearance of ASPM-1 foci around chromosomes (Fig. 1F, Fig. S1C, Movie 3). Progressive microtubule bundling and cross-linking led to the formation of a multipolar spindle with several ASPM-1 foci around chromosomes (Connolly et al., 2015). GFP::KLP-7 also concentrated on these multiple poles (Fig. 1E). In controls, this phase was accompanied by the dispersal of chromosomes on the forming spindle and their subsequent congression and alignment. In KLP-7-depleted oocytes, the ectopic perinuclear microtubules seen in phase 1 coalesced around chromosomes during phase 2 to form a multipolar spindle that displayed multiple ASPM-1 foci as in controls (Fig. 1F, Fig. S2C, Movie 3). Thus, although the oocytes started phase 2 in the absence of a diffuse microtubule cloud around chromosomes and with ectopic perinuclear microtubules, KLP-7-depleted oocytes formed a seemingly normal multipolar spindle early in meiosis I.

During phase 3 in controls, the multipolar microtubule structure was slowly shaped into a bipolar spindle and ASPM-1 and GFP::KLP-7 concentrated at the spindle poles. In KLP-7-depleted oocytes, the multiple microtubule foci persisted and a bipolar spindle was almost never observed until after anaphase onset (see below). Similar imaging experiments in a strain expressing the nuclear envelope component Lamin 1 (LMN-1) tagged with GFP and mCherry-tagged β-tubulin revealed that the entire process of initial meiotic spindle assembly took place within the limits of the rupturing nuclear envelope (Fig. 1G, Fig. S1D, Movie 4). In both controls and KLP-7-depleted oocytes, nuclear envelope remnants were visible around the spindle up until the end of the bipolarization phase. KLP-7 therefore plays a crucial role in phase 3, and is required for bipolar spindle formation at this stage.

In controls, phase 4, as previously described, corresponded to an extensive spindle pole disassembly (Yang et al., 2003). KLP-7-depleted oocytes did not usually reach bipolar spindle assembly, but klp-7(RNAi) did not affect the overall timing of nuclear envelope disassembly or anaphase onset (Fig. 1B,G, Fig. S1D, Movie 4). After anaphase onset, spindle bipolarity in KLP-7-depleted oocytes was almost always rescued, due largely to ʻpolar clustering'. Thus, although disruption of KLP-7 activity leads to multipolar spindle assembly prior to anaphase, spindle bipolarity appears to be established prior to or during meiotic anaphase and cytokinesis (see also below).

To further investigate the contribution of KLP-7 to acentrosomal spindle assembly and function, we filmed ex utero fertilized oocytes expressing GFP-tagged β-tubulin and mCherry-tagged H2B, which allows for higher spatial resolution (Fig. 2A, Movie 5). Control oocytes had all completed meiosis I spindle bipolarization at the time of dissection. In KLP-7-depleted or -deleted oocytes, spindle microtubule density (measured by average GFP::β-tubulin intensity) was increased as compared with controls at every step of meiosis I and II (Fig. 2B, Movie 5). In controls, the barrel-shaped meiosis I spindle displayed a few extremely short microtubules extending outward (Fig. 2C). By contrast, KLP-7-depleted oocytes displayed disorganized spindles with numerous long microtubules or microtubule bundles extending out toward the cytoplasm and the oocyte cortex (Fig. 2C). In agreement with a previous study and with our in utero analysis, we found that spindles assembled in KLP-7-depleted oocytes were multipolar (Connolly et al., 2015) (Fig. 2D). However, we found that the supernumerary poles were always resolved and incorporated into one of the two dominant spindle poles at or just after anaphase onset.

We next monitored the progression and accuracy of meiotic chromosome segregation in the presence and absence of KLP-7 in embryos (Fig. 2E). The overall timing of divisions was not significantly different between control and KLP-7-depleted or -deleted oocytes (Fig. 2H). In control oocytes, chromosomes aligned on tight metaphase plates during metaphase I and II. Segregating chromosomes remained tightly clustered during both anaphase I and II, which usually ended with successful PBE. By contrast and consistent with the spindle disorganization that we described above, KLP-7 depletion or deletion led to visible chromosome alignment and segregation defects during both meiotic divisions (Fig. 2E). Chromosome masses always separated after anaphase onset and lagging chromosomes were evident in most oocytes. Chromosome counting in fixed meiosis II oocytes that succeeded in first PBE revealed significant aneuploidy (control, 6 chromosomes in 14/14 oocytes; klp-7(RNAi), 4 chromosomes in 2/15 oocytes, 5 in 5/15, 6 in 7/15 and 7 in 1/15; klp-7Δ, 4 chromosomes in 2/16 oocytes, 5 in 6/16, 6 in 6/16 and 7 in 2/16). As expected, KLP-7 depletion or deletion did not affect chromosome number during meiosis I (control, 6 chromosomes in 24/24 oocytes; klp-7(RNAi), 6 chromosomes in 20/20 oocytes; klp-7Δ, 6 chromosomes in 20/20 oocytes).

We also found that KLP-7 is required for full chromosome segregation in meiosis I. Kymographs of anaphase I revealed that chromosome masses in KLP-7-depleted or -deleted oocytes separated at a rate comparable to controls during the first 2 min following anaphase onset (Fig. 2F,G). This timing corresponds approximately to the duration of anaphase A during meiosis I in C. elegans oocytes (McNally et al., 2016). Chromosomes in controls continued to separate during anaphase B for the following 3 min and reached a maximal distance of 5.5 μm. In striking contrast, chromosome masses abruptly slowed down in KLP-7-depleted or -deleted oocytes 1.5 min after anaphase onset and chromosome separation paused at a distance of ∼2.5 μm (Fig. 2G). Abnormal chromosome segregation was frequently followed by unsuccessful PBE and the formation of a multi-pronucleate polyploid embryo (Fig. 2E). KLP-7 is therefore essential for anaphase B chromosome movements and for the overall accuracy and success of meiotic chromosome segregation.

The defects that we observed in chromosome segregation and PBE in KLP-7-depleted or -deleted oocytes prompted us to analyze anaphase I central spindle organization and function, this spindle being essential for both chromosome segregation and PBE in the C. elegans oocyte (Dumont et al., 2010; Fabritius et al., 2011). Deconvolution microscopy on fixed oocytes in anaphase I showed obvious central spindle microtubule organization defects in KLP-7-depleted oocytes (Fig. 3A).

To understand KLP-7 function in meiotic central spindle assembly and in polar body cytokinesis, we analyzed microtubule organization and density over time by live imaging during anaphase I. We filmed oocytes expressing GFP-tagged β-tubulin and mCherry-tagged H2B (Fig. 3B, Fig. S3A). We noticed that the microtubule structures were more dense during anaphase in KLP-7-depleted oocytes. Specifically, the segregating chromosome masses were devoid of microtubules in controls but remained embedded in a microtubule mesh throughout anaphase in KLP-7-depleted oocytes. Consistent with this, the microtubule minus-end marker ASPM-1 was abnormally concentrated around the segregating chromosomes throughout anaphase (Fig. 3C, Table S1). Thus, ectopic microtubules assembled in the vicinity of chromosomes persisted throughout anaphase in KLP-7-depleted oocytes leading to central spindle defects.

To test whether these defects directly affect central spindle component localization, we analyzed the dynamic recruitment of the central spindle microtubule bundling protein SPD-1 (ortholog of PRC1) and of the Centralspindlin complex subunit CYK-4 (ortholog of MgcRacGAP or RACGAP1). Both proteins are normally specifically recruited on central spindle microtubules during anaphase, where they are essential for microtubule organization and cytokinesis (Mishima et al., 2002; Verbrugghe and White, 2004; Glotzer, 2005; Maton et al., 2015). The dense and disorganized microtubules of the central spindle in KLP-7-depleted oocytes correlated with a reduction in the recruitment of these two proteins to the meiotic central spindle and with ectopic CYK-4 on chromosomes (Fig. 3D,E, Movie 6, Table S1). The lack of KLP-7 therefore leads to the mislocalization of essential central spindle components, which is likely to contribute to the observed central spindle defects.

In control oocytes, recruitment of central spindle components ultimately leads to the formation of a contractile actomyosin ring that promotes plasma membrane furrowing and cytokinesis (Maddox et al., 2012). To test if the improper central spindle component localization in KLP-7-depleted oocytes correlated with defects in actomyosin organization, we analyzed oocytes expressing GFP-tagged myosin II (NMY-2) and mCherry-tagged H2B during anaphase I. Consistent with previous findings, in control oocytes NMY-2 formed a disc above the segregating chromosomes that progressively evolved into a cylinder, which ultimately formed the meiotic midbody between the segregated chromosomes (Dorn et al., 2010) (Fig. 3F, Movie 6, Table S1). In KLP-7-depleted oocytes, a normal disc of NMY-2 was initially visible above chromosomes but it strikingly almost never evolved into a cylinder. Instead, the set of chromosomes that would normally end up in the first polar body re-entered the oocyte cytoplasm and was surrounded by a thick layer of cortical NMY-2. Altogether, these results show that KLP-7 is essential for meiotic cytokinesis and PBE through its function in central spindle organization.

A recent study analyzing feeding RNAi-mediated depletion of KLP-7 or a KLP-7 temperature-sensitive loss-of-function mutant showed that the multipolar spindle phenotype observed when KLP-7 activity is decreased could be rescued upon co-depletion of the NDC-80 kinetochore component (Connolly et al., 2015). That study concluded that KLP-7 is involved in destabilizing improper kinetochore-microtubule attachments established during early prometaphase, similar to the function of its vertebrate ortholog MCAK during mitosis. This in turn would release tension within meiotic spindles that would otherwise lead to extra spindle pole formation. Although we found a similar rescuing effect of the ndc-80(RNAi) on early prometaphase spindles assembled in klp-7-deleted oocytes, we noticed that these spindles were still disorganized during most of prometaphase/metaphase (Fig. S4A,B). Furthermore, depleting the core kinetochore scaffold protein KNL-1 did not rescue spindle bipolarity when KLP-7 is absent (Fig. S4A,B). Altogether, these results suggest that destabilizing kinetochore-microtubule attachments in oocytes is not sufficient to stably rescue bipolar spindle formation when KLP-7 is absent.

To understand the origin of the extra spindle poles that assemble in KLP-7-depleted or -deleted oocytes, we performed ex utero live imaging of the spindle assembly process in oocytes. In line with a previous study, we noticed the presence of numerous ectopic microtubule asters near the cell cortex of KLP-7-depleted or -deleted oocytes (Fig. 4A,B) (Han et al., 2015). Ectopic cortical asters persisted throughout meiosis but disappeared at anaphase II onset (Fig. 4C). These asters displayed rapid movements at the cortex and tended to cluster together. Importantly, we observed a significant number of asters that aggregated at the metaphase spindle (Fig. 4D). Cortical asters were positive for ASPM-1, and when they were localized near the meiotic spindle they contributed to generating supernumerary spindle poles (Fig. 4E,F). Therefore, KLP-7 is required to prevent ectopic microtubule assembly at the cell cortex, which otherwise leads to the formation of extra spindle poles and contributes to the observed multipolar spindle phenotype.

The longer and denser meiotic spindles and ectopic cortical microtubule asters observed suggested that microtubules are, overall, more stable after KLP-7 depletion. To test this hypothesis, we performed fluorescence recovery after photobleaching (FRAP) experiments of the entire metaphase I spindle in control and KLP-7-depleted oocytes that were genetically arrested in metaphase I (see Materials and Methods; Fig. 5A, Movie 7). Control metaphase I spindles were highly dynamic and recovered 86% of their initial fluorescence with a halftime of recovery (t_1/2) of 22.2 s. Meiotic spindles assembled in KLP-7-depleted oocytes also recovered almost completely (87% of initial fluorescence); however, recovery was delayed compared with controls with a t_1/2 of 33.3 s (Fig. 5B). Thus, although the proportion of fully stable spindle microtubules that did not recover fluorescence over the course of the quantification period was not significantly different (P=0.9765) in control or KLP-7-depleted oocytes, microtubules were on average more stable in the latter (P=0.0338).

To determine the origin and dynamics of the ectopic cortical microtubule asters, we performed spinning disc cortical live imaging in control and klp-7-deleted oocytes expressing GFP-tagged β-tubulin (Fig. 5E,F). At the cell cortex, a dynamic cortical microtubule meshwork that slid rapidly in parallel with the cortex was visible in both control and klp-7-deleted oocytes (Fig. 5C-E). However, the microtubule meshwork was more dense in the absence of KLP-7, suggesting that global microtubule stabilization leads to the observed ectopic cortical microtubule asters (Fig. 5C). Consistent with this, cortical microtubules were overall less dynamic in klp-7-deleted oocytes, as evident by the increased time spent in pause (not growing or shrinking, t_control=28.1±29.4 s, t_klp-7Δ=68.9±35.11 s, Fig. 5F) and the overall reduction of all microtubule dynamics parameters (Fig. 5G-I) (Lacroix et al., 2014). By applying a simple model of microtubule dynamics to our data, we calculated that the average theoretical length of microtubules in the absence of KLP-7 was higher than in control oocytes at steady state (L_control=6 μm, L_klp-7Δ=8.2 μm; see Materials and Methods), which is consistent with our live observations (Verde et al., 1992). Thus, KLP-7 increases cortical microtubule dynamics in the C. elegans oocyte, preventing ectopic microtubule aster formation.

Strikingly, the ectopic asters that formed in oocytes in the absence of KLP-7 were never observed in mitotic embryos. Instead, during mitosis, the absence of KLP-7 leads to increased astral microtubule density and a corresponding increase in astral microtubule pulling forces at centrosomes (Fig. S3B) (Srayko et al., 2005). To test if this difference between oocytes and zygotes could be linked to the presence of functional centrosomes in the zygote, we analyzed the effect of reducing centrosomal activity following depletion of the scaffold component SPD-5 in klp-7-deleted zygotes. In the absence of a functional centrosome in spd-5(RNAi) zygotes when KLP-7 is present, cytoplasmic asters were never observed and the few microtubules that assembled following NEBD always radiated from the condensed chromosomes (Fig. 6A) (Hamill et al., 2002). Strikingly, however, when SPD-5 was depleted in klp-7-deleted zygotes, numerous cytoplasmic ectopic asters assembled at NEBD at a distance from the chromosomes (Fig. 6A,B). These asters subsequently coalesced around the condensed chromosomes to form a single larger microtubule structure (not shown). Thus, in the absence of functional centrosomes in C. elegans zygotes, KLP-7 activity is essential to prevent ectopic microtubule assembly.

To test if this function of KLP-7 is a general feature of kinesin-13, we tested the effect of MCAK depletion in HeLa cells when centrosome activity is reduced during microtubule regrowth after nocodazole washout (Cavazza et al., 2016). Forty-five minutes after removing nocodazole, microtubules reassembled from two microtubule-organizing centers (MTOCs) on average in control cells (2.16±0.06; Fig. 6C,D). By contrast, three MTOCs (3.13±0.12) could be detected in most MCAK-depleted cells. Importantly, the effect of depleting MCAK was specific to cells in which centrosome activity was reduced by the nocodazole treatment. Thus, ectopic microtubule nucleation centers are activated in human cells with reduced centrosome activity when MCAK levels are decreased. Altogether, these results suggest that preventing ectopic microtubule assembly in cells with reduced or absent centrosome activity is a previously uncharacterized general and conserved function of kinesin-13 depolymerases (Fig. 6E).

Previous studies of meiotic spindle assembly in the C. elegans oocyte have been performed at relatively low spatial and temporal resolutions, did not provide temporal information and/or missed the very early steps of spindle assembly (Yang et al., 2003; Connolly et al., 2015; Wolff et al., 2016). Here, we provide a precise quantitative picture covering the full time window of interest and the first time-resolved analysis of the entire process of meiotic spindle formation in this system. Oocytes of most species lack centriole-containing centrosomes, and microtubules assemble through the chromatin-dependent pathway or from acentriolar MTOCs (Dumont and Desai, 2012). In these acentrosomal oocytes, microtubules can be seen originating locally from the chromatin itself or from discrete organizing centers (Huchon et al., 1981; Gard, 1992; Dumont et al., 2007; Schuh and Ellenberg, 2007; Colombie et al., 2008). By contrast, we found that in the C. elegans oocyte, microtubules, which are excluded from the nucleus before NEBD, assemble in the nuclear space after NEBD to form a diffuse cloud. This result is consistent with qualitative observations made in previous studies and with the lack of discrete MTOCs in this system (Yang et al., 2003). We found that spindle assembly is constrained within the space of the rupturing nuclear envelope. We thus propose that in C. elegans oocytes the space delimited by the nuclear envelope remnants acts as a diffuse MTOC. In KLP-7-depleted oocytes, NEBD occurred normally but the microtubule nuclear cloud did not form and an excess of microtubules persisted around the breaking nuclear envelope throughout phase 1. This suggests that the function of KLP-7, which is cytoplasmic before and at NEBD, is to destabilize these perinuclear microtubules in order to release free tubulin necessary for the formation of the nuclear cloud.

Following formation of the microtubule nuclear cloud, bundling and cross-linking activities led to microtubule coalescence around meiotic chromosomes. This second step is likely to be under the control of microtubule motors previously implicated in successful meiotic divisions, such as dynein, the two redundant kinesin-14 family members KLP-15/16 (orthologs of NCD) and the kinesin-12 family member KLP-18 (ortholog of XKLP2 or KIF15) (Dernburg et al., 2000; Piano et al., 2000; Colaiacovo et al., 2002; Segbert et al., 2003; Wolff et al., 2016). We found that in the absence of a nuclear cloud of microtubules in KLP-7-depleted oocytes, the ectopic perinuclear microtubules are instead bundled and coalesce around chromosomes to form a seemingly normal multipolar spindle. However, the subsequent organization of microtubules into a bipolar spindle was impaired. Instead, abnormally dense multipolar spindles with long, disorganized and stable microtubules were formed.

Following bipolar spindle formation and chromosome alignment on a tight metaphase plate, drastic microtubule reorganization occurs that ultimately leads to chromosome segregation and PBE. We previously showed that chromosome segregation in the C. elegans oocyte is driven by an atypical kinetochore-independent mechanism (Dumont et al., 2010). In this system, central spindle organization is essential for chromosome segregation (Muscat et al., 2015; McNally et al., 2016). In agreement, we show here that KLP-7 depletion leads to disorganized central spindles that correlate with impaired chromosome segregation. Specifically, anaphase B, which normally accounts for most of the segregation process, does not occur. In line with this result, central spindle elongation was proposed to be specifically important for anaphase B chromosome movements (McNally et al., 2016). Live imaging of the minus-end marker GFP::ASPM-1 at this stage showed that microtubule minus-ends are distributed all over the disorganized central spindle instead of being concentrated toward chromosomes and generate an antiparallel microtubule overlap. We suspect that KLP-7 is required to generate this overlap by preventing excessive and/or ectopic microtubule elongation from chromosomes, where it is concentrated during meiotic anaphase (Fig. 1E) (Han et al., 2015). SPD-1 and the Centralspindlin complex (including CYK-4) have been shown to preferentially interact with overlapping microtubule plus-ends, which might explain their delocalization in the denser and disorganized central spindle that is assembled following klp-7(RNAi) (Bieling et al., 2010; Davies et al., 2015).

In the absence of KLP-7, another striking defect in the organization of the microtubule network is the formation of multiple microtubule asters at the oocyte cortex. Although a cortical meshwork of microtubules is present in control oocytes, asters are normally not present at the cortex. A cytoplasmic pool of KLP-7 might be responsible for reducing the stability of this microtubule meshwork and to prevent ectopic aster formation. We observed that cortical asters located near the meiotic spindle often joined the spindles and contributed to the formation of the multipolar spindle. The minus-end-directed motor dynein, which is present throughout the cortex of the oocyte, is probably responsible for the aster aggregation that we observed (Crowder et al., 2015). Asters incorporated in the spindle could saturate the activity of microtubule motors and thus prevent normal spindle bipolarization.

The ectopic asters observed in oocytes in the absence of KLP-7 disappeared abruptly at anaphase II and were never observed in mitotic embryos. During mitosis, KLP-7 depletion leads to increased astral microtubule density but does not lead to ectopic cortical aster formation (Srayko et al., 2005). We hypothesize that this difference is linked to the large size of the embryonic mitotic spindle as compared with the tiny oocyte spindles, and to the absence of functional centrosomes in oocytes, which are the dominant MTOCs during mitotic divisions (Hannak et al., 2002). Oocytes and the single-celled fertilized zygote share a common cytoplasmic composition, including the same concentration of tubulin heterodimers. The large astral spindle in zygotes contains a higher microtubule mass than the tiny meiotic spindles in oocytes. This leads to a lower cytoplasmic concentration of free tubulin heterodimer in zygotes compared with oocytes. The cytoplasmic tubulin concentration in oocytes is thus likely to be closer to the in vivo critical concentration at which microtubules can spontaneously nucleate and form microtubule asters. KLP-7 depolymerase activity must restrain this spontaneous microtubule assembly in oocytes. During mitosis, the centrosomes would thus act as a microtubule polymerization-buffering system and prevent overall spindle disorganization. By contrast, in oocytes depleted of KLP-7, free tubulin heterodimers are incorporated into all existing microtubule networks, including the perinuclear microtubules in unfertilized oocytes, and into the cortical meshwork and the spindle after fertilization, ultimately leading to its disorganization. Consistent with this interpretation we showed that, in the absence of KLP-7, microtubule asters spontaneously assembled during mitosis in the one-celled zygote only when the free tubulin heterodimer concentration was experimentally increased through reduction of centrosome activity. Similarly, in cultured human cells when centrosome activity was reduced (during microtubule regrowth after nocodazole washout), we observed a higher number of MTOCs after kinesin-13 MCAK depletion than in control cells. Similar ectopic asters have been observed in Drosophila oocytes depleted of the kinesin-13 KLP10A (Radford et al., 2012; Do et al., 2014). We propose that, when centrosome activity is reduced or absent, global microtubule destabilization by a kinesin-13 family member(s) is essential to prevent the formation of ectopic microtubule asters, which otherwise lead to spindle disorganization and chromosome mis-segregation (Fig. 6E). This previously uncharacterized function of kinesin-13 proteins defines a new level in the regulation of microtubule assembly in vivo, which is particularly important for the generation of euploid oocytes that lack centrosomes. As kinesin-13 motors are highly conserved across evolution, this new paradigm is likely to apply to other species and could further our understanding of human reproduction and the etiology of sterility.

C. elegans strains are listed in Table S2 and were maintained at 16°C or 23°C (Oegema et al., 2001). Primers for dsRNA production are listed in Table S3 (Oegema et al., 2001). L4 hermaphrodites were microinjected with dsRNA and incubated at 20°C for 48 h before processing.

HeLa cells, tested monthly for mycoplasma contamination using a luminometer detection method (Lonza), were maintained in DMEM (Lonza) supplemented with 10% FBS and penicillin/streptomycin (Gibco) at 37°C in a humidified atmosphere with 5% CO₂. Cells were plated on glass coverslips coated with poly-L-lysine (Sigma-Aldrich). RNAi experiments were conducted using RNAi MAX transfection reagent (Invitrogen) according to the manufacturer's guidelines. Previously published siRNA oligos were used to deplete MCAK (Domnitz et al., 2012). After 48 h of siRNA treatment, the cells were incubated for 2-3 h with 300 ng/ml nocodazole (Sigma, M1404). The nocodazole was then washed out five times with fresh DMEM and cells were left for 45 min in fresh DMEM. Cells were then briefly washed in PBS and fixed in PHEM (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, 2 mM MgCl₂, pH 6.9) containing 4% formaldehyde for 10 min. Immunofluorescence was conducted using antibodies against mouse anti-α-tubulin (Sigma, T9026; used at 1 μg/ml) and human anti-NDC80 antibody (kind gift from Iain Cheeseman; used at 1 μg/ml). DNA was then counterstained with 1 µg/ml Hoechst.

Images were acquired on a DeltaVision Core deconvolution microscope (Applied Precision) equipped with a CoolSNAP HQ2 CCD camera (Photometrics). Twenty z-sections were acquired at 0.3 µm steps using a 100×1.4 NA Olympus U-PlanApo objective without binning. Maximal projections of stacks of interest after image deconvolution (SoftWorks) are presented. Equivalent exposure conditions were used between controls and drug-treated cells. Experiments were repeated three times. The number of spindle poles or the presence of ectopic microtubule foci in the cytoplasm was visually assessed and quantified.

For in utero live imaging experiments, adult worms were anaesthetized using 100 mg tricaine (Sigma-Aldrich, E10521) and 10 mg tetramisol hydrochloride (Sigma-Aldrich, T1512) diluted in 1 ml M9 buffer. Immobilized worms were then mounted on a 2% agarose pad in M9 buffer between a slide and a coverslip. Live imaging was performed using a Nikon CFI APO LBDA S 40×/NA 1.25 water objective on a spinning disk confocal microscope (Roper Scientific) equipped with a CoolSNAP HQ2 camera, and acquisition parameters were controlled by MetaMorph 7 software (Molecular Devices). Four z-sections every 2 µm were acquired at 20 s intervals. Imaging on ex utero oocytes was performed as described (Dumont et al., 2010).

FRAP experiments were performed on ex utero oocytes using a Nikon CFI APO LBDA S 60×/NA 1.4 oil objective with 2×2 binning on a spinning disk confocal microscope equipped with the iLas Pulse FRAP/Photoactivation module (Roper Scientific). The extensive disassembly of microtubules observed during the spindle shrinkage phase could preclude measuring fluorescence recovery. To avoid this caveat and to measure fluorescence recovery in a steady state, we performed the FRAP experiments in the mat-2(ax76ts) temperature-sensitive (ts) strain that arrests in metaphase I when shifted to the restrictive temperature (26°C) (Golden et al., 2000). Stacks of four z-sections with a spacing of 2 µm were acquired every 3 s in the GFP channel before a single FRAP event of the entire surface of the metaphase spindle. After the FRAP event, images were acquired every 3 s for the first 120 s, then every 10 s for the following 100 s, and every 20 s for the last 500 s. A maximum projection of the four z-sections is presented for each time point. The average fluorescence was measured in a box around the metaphase spindle (Fspin) and in a box at a distance from the spindle in the cytoplasm (Fcyt). Normalization, correction and fitting of the measured fluorescence intensities were performed using Prism 6 software (GraphPad). Although we verified that the imaging conditions that we used did not lead to any significant photobleaching on embryos that did not undergo a FRAP event, the data were corrected (Fcor) for any potential photobleaching occurring during acquisition by multiplying each time point by Fcyt(0)/Fcyt(t). In order to be able to compare different experiments, the last prebleach and first postbleach time points were normalized to 1 and 0, respectively, by: FcorNormalized(t) = (Fcor(t)−FcorPost)/(FcorPre−FcorPost). The mean value of FcorNormalized was then calculated for individual embryos at each time point. The corresponding plot was fitted to a mono-exponential function and the half-time for recovery was extracted.

Image analyses and quantifications were performed using Fiji (Schindelin et al., 2012) and Icy (de Chaumont et al., 2012) software. Kymographs were generated using the Multi Kymograph tool in Fiji.

For estimating the average length of microtubules at steady state, we used a simple mathematical model that links microtubule length distribution to dynamics parameters (Verde et al., 1992). In this model, L=(Rshrink×Rgrowth)/[(Rshrink×Fcat)−(Rgrowth×Fres)], where L is the average microtubule length, Rshrink and Rgrowth are the rates of microtubule shrinkage and growth, respectively, and Fcat and Fres are the frequencies of microtubule catastrophe and rescue, respectively.

Immunofluorescent staining was performed as described (Dumont et al., 2010). The rabbit anti-KLP-7 antibody was custom produced, validated in this study (see Fig. S1) and used at 1 μg/μl.

Experiments were repeated at least twice and a minimum number of ten oocytes were quantified for each experimental condition. All graphs were produced and statistical analyses performed with Excel (Microsoft) and Prism 6. Statistical significance was evaluated using unpaired t-tests with Welch's correction or one-way ANOVA.

We thank Jeremy Cramer from Cherry Biotech (Rennes, France) for allowing us to use pre-commercial development versions of the CherryTemp system; Patricia Moussounda for providing technical support; the Caenorhabditis Genetics Center (University of Minnesota, USA) for worm strains; the National BioResource Project (National Institute of Genetics, Japan) for the tm2143 mutant strain; and Iain Cheeseman (Whitehead Institute for Biomedical Research, Cambridge, MA, USA) for the NDC80 bonsai antibody.