Haspin kinase regulates microtubule-organizing center clustering and stability through Aurora kinase C in mouse oocytes

ABSTRACT

Meiotic oocytes lack classic centrosomes and, therefore, bipolar spindle assembly depends on clustering of acentriolar microtubule-organizing centers (MTOCs) into two poles. However, the molecular mechanism regulating MTOC assembly into two poles is not fully understood. The kinase haspin (also known as GSG2) is required to regulate Aurora kinase C (AURKC) localization at chromosomes during meiosis I. Here, we show that inhibition of haspin perturbed MTOC clustering into two poles and the stability of the clustered MTOCs. Furthermore, we show that AURKC localizes to MTOCs in mouse oocytes. Inhibition of haspin perturbed the localization of AURKC at MTOCs, and overexpression of AURKC rescued the MTOC-clustering defects in haspin-inhibited oocytes. Taken together, our data uncover a role for haspin as a regulator of bipolar spindle assembly by regulating AURKC function at acentriolar MTOCs in oocytes.

Summary: Haspin kinase is a regulator of meiotic bipolar spindle assembly, which is mediated by acentriolar microtubule-organizing centers (MTOCs), and acts by regulating the localization of AURKC to the MTOCs in mouse oocytes.

INTRODUCTION

A bipolar spindle is an essential cellular component that ensures chromosome alignment at the metaphase plate and accurate chromosome segregation at anaphase. However, cells with multipolar spindles can divide, resulting in aneuploidy or cell death (Ganem et al., 2009; Kwon et al., 2008; Quintyne et al., 2005; Wang et al., 2004). In mitosis, spindle microtubules are assembled at poles called centrosomes. The classic composition of a centrosome is two centrioles that maintain the pole structure. A cloud of pericentriolar material (PCM) that supports microtubule nucleation and anchors the spindle within the cell surrounds the centrioles. In many animal species, oocytes that undergo meiosis lack this classic centrosome structure and instead contain microtubule-organizing centers (MTOCs) (Dumont et al., 2007; Schatten and Sun, 2009). For instance, in mammals, oocytes from mice lack centrioles but contain PCM, whereas oocytes from humans appear to lack both (Holubcova et al., 2015). Because oocytes have an increased likelihood of making mistakes in chromosome segregation giving rise to aneuploid conception (Hassold and Hunt, 2001), these unusual properties of the meiotic spindle have received substantial attention.

The mechanisms used to generate bipolar spindles in centriolar versus acentriolar spindle poles differ. In mitosis, a centrosome cell cycle exists to ensure spindle bipolarity. During S phase a centrosome duplicates and it is subsequently separated in M phase (Fu et al., 2015; Gonczy, 2015). However, there is a substantial difference in how the acentriolar meiotic spindle in mouse oocytes is regulated compared with the spindle in mitosis. Oocytes are arrested in prophase of meiosis I (akin to a mitotic G2 phase) with multiple MTOCs (Schuh and Ellenberg, 2007). Before nuclear envelope breakdown (NEBD), the MTOCs first undergo decondensation, followed by dynein- and microtubule-dependent stretching resulting in MTOC fragmentation (Łuksza et al., 2013). Finally, after NEBD, the kinesin KIF11 (also known as Eg5) drives the second round of MTOC fragmentation. During prometaphase I, the multiple MTOCs (upwards of 80) cluster into an apolar microtubule ball in the center of the oocyte that then takes several hours to resolve and organize into two poles in metaphase I. During this time the spindle often goes through multipolar intermediates (Schuh and Ellenberg, 2007). Therefore, oocytes have regulatory mechanisms distinct from those used in mitotic cells to control building a spindle. One mechanism used involves Ran-GTP gradients that emanate from chromatin-localized RCC1, and in Drosophila oocytes another mechanism involves the chromosomal passenger complex (CPC) (Dumont et al., 2007; Radford et al., 2012; Schuh and Ellenberg, 2007). Although centriole-containing cells also utilize alternative pathways such as Ran-GTP to build spindles, whether there are differences between cell types is unknown. The intricacies of the regulation and use of the classic spindle pole and alternative chromatin-based pathways in oocytes are still under exploration and might vary from species to species.

One protein that localizes with the PCM in somatic cells and oocytes is Aurora kinase A (AURKA) (Saskova et al., 2008; Shuda et al., 2009; Solc et al., 2012). In somatic cells (Ma et al., 2011), upon binding a microtubule-binding protein called TPX2, AURKA auto-activates and regulates the building of the bipolar spindle by recruiting other PCM proteins, promoting microtubule nucleation and anti-parallel sliding forces, and maintaining the integrity of astral microtubules with the cortex (Dodson and Bayliss, 2012; Giubettini et al., 2011; Kufer et al., 2002; Li et al., 2008; Scrofani et al., 2015). In mouse oocytes, overexpression of AURKA increases MTOC numbers prematurely (Saskova et al., 2008; Solc et al., 2012), and its depletion causes disorganized meiosis I spindles (Saskova et al., 2008). Overexpression of TPX2 in mouse oocytes results in multipolar spindles whereas its depletion ablates the spindle (Brunet et al., 2008). These phenotypes are likely due to changes in the phosphorylation of the AURKA substrate TACC3. In mammals, oocytes express two other Aurora kinase homologs, AURKB and AURKC, which function within the CPC (Balboula and Schindler, 2014; Schindler et al., 2012; Sharif et al., 2010; Shuda et al., 2009). Simultaneous inhibition of these kinases perturbs meiosis I spindles (Shuda et al., 2009), but whether they participate directly in building and maintaining a meiotic spindle is unknown.

Haspin (also known as GSG2) is a protein kinase that regulates localization of the AURKB–CPC in mitosis and AURKC–CPC in meiosis (De Antoni et al., 2012; Nguyen et al., 2014; Wang et al., 2012). In mitosis, depletion of haspin also causes multipolar spindles (Dai et al., 2009). This phenotype has been attributed to a failure to integrate a chromatin-initiated microtubule pathway with that of the spindle pole microtubule nucleation pathway. We, and others, have previously demonstrated differences in the requirement for haspin during meiosis I compared to in mitotic cells (Kang et al., 2015; Nguyen et al., 2014; Wang et al., 2016). These differences are that haspin regulates localization of AURKC–CPC along the interchromatid axis of the meiosis I bivalents, and not at centromeres, and that haspin inhibition alters meiotic chromosome condensation and not cohesin, as it does in mitosis. In our prior study, we also observed an increased frequency of MTOCs that failed to cluster at poles, which we describe in more detail here. Because of the differential functions of haspin in meiosis compared to mitosis, and because of the gross differences in building a spindle, we asked how this MTOC clustering defect arises in mouse oocytes. Here, we describe a new role for haspin in regulating MTOC clustering through MTOC-localized AURKC.

RESULTS

Haspin activity is required for bipolar spindle assembly during meiosis I

In mitotic cells, the kinase haspin is essential for maintenance of spindle pole integrity by regulating chromosome cohesion (Dai et al., 2009). Previous studies have shown that haspin regulates chromosome condensation and kinetochore microtubule attachments in mouse oocytes, but its inhibition does not alter cohesin protein levels (Kang et al., 2015; Nguyen et al., 2014). While conducting our analyses, we observed a second phenotype. This phenotype was a frequent occurrence of spindles with MTOC-clustering defects in metaphase I. To quantify this phenotype, we matured oocytes to metaphase I in 5-iodotubercidin (5-Itu), a small-molecule inhibitor of haspin (De Antoni et al., 2012; Wang et al., 2012) whose efficiency to inhibit haspin in mouse oocytes has been described previously (Nguyen et al., 2014). By performing α-tubulin and γ-tubulin immunostaining, we analyzed the MTOC clustering phenotype in more detail (Gueth-Hallonet et al., 1993). Control-treated oocytes rarely had spindles with unclustered MTOCs at the poles (7.35±2%, mean±s.e.m.) whereas inhibition of haspin activity significantly increased the percentage of oocytes with unclustered metaphase I MTOCs (60.2±5.4%) (Fig. 1A,B). On average, 5-Itu-treated oocytes contained three (3.13±0.19%) dominant MTOC clusters compared to two (2.17±0.07%) in controls (Fig. 1C). We also confirmed this result by evaluating pericentrin immunolocalization, and found unclustered foci at metaphase I when haspin was inhibited with 5-Itu (Fig. S1). Moreover, we observed a similar spindle pole phenotype when haspin was depleted using an injection of a cocktail containing morpholino oligonucleotides with small interfering RNAs (siRNAs) (Fig. S2) indicating that the phenotype upon 5-Itu treatment is specific to loss of haspin function.

Fig. 1.

Haspin inhibition causes unfocused metaphase I spindle poles. (A–E) Fully grown, prophase-I-arrested oocytes were in vitro matured in the presence of ethanol (EtOH) or 500 nM of 5-Iodotubercidin (5-Itu) to metaphase I (7 h) prior to fixation and detection of MTOCs (γ-tubulin; red in merge), spindle microtubules (α-tubulin; green in merge) and DNA (DAPI; blue). (B) Quantification of the percentage of oocytes analyzed in A with unfocused spindle poles. (C) Quantification of the number of γ-tubulin-positive poles analyzed per oocyte in A. These experiments were repeated five times and the total numbers of oocytes examined are indicated above the graph bars. (D) Prior to fixation detection of MTOCs (γ-tubulin; green in merge), spindle microtubules (α-tubulin; gray in merge), kinetochores (CREST; red in merge) and DNA (DAPI; blue), oocytes were exposed to ice-cold medium to allow visualization of kinetochore-attached microtubules. The optical zooms were taken from the indicated boxes and show two functional poles at the same end of the spindle. Arrows indicate microtubules making a connection to kinetochores. These experiments were repeated twice, and 25 and 30 oocytes were examined in the control and 5-Itu-treated groups, respectively. (E,F) Half of the oocytes were injected with non-degradable GFP-tagged Cyclin B cRNA (Ccnb1-Δ90-Gfp) whereas control oocytes were injected with Gfp control cRNA, and all the oocytes were in vitro matured in the presence of EtOH or 500 nM of 5-Itu to metaphase I (7 h) prior to fixation and confocal microscopy. Arrows highlight unclustered MTOCs. (F) Quantification of the percentage of oocytes analyzed in E with unfocused spindle poles. These experiments were repeated twice and the total numbers of oocytes examined are indicated above the graph bars. The data are expressed as mean±s.e.m. Scale bars: 10 µm. *P<0.05; **P<0.01 [Student's t-test (B,C) or one-way ANOVA (F)].

We then asked whether each of the foci made stable connections with kinetochores. We exposed control and 5-Itu-treated oocytes to cold medium to depolymerize unattached microtubules, thereby enhancing visualization of the stably attached microtubules. Control oocytes showed stable bipolar attachments; that is, sister kinetochore pairs attached to opposite poles (Fig. 1D). In contrast, when haspin was inhibited, stable attachments emanated from multiple MTOCs, indicating that each dominant MTOC foci in these oocytes could be functional upon chromosome segregation.

We have previously shown that oocytes treated with 5-Itu have an ∼2 h delay in prometaphase I (Nguyen et al., 2014). Therefore, it is possible that this defect in MTOC clustering reflects a cell cycle delay in meiotic progression. To exclude this possibility, we blocked metaphase I exit by injecting oocytes with non-degradable cyclin B1 (Ccnb1-Δ90) (Schindler et al., 2012; Schindler and Schultz, 2009). Vehicle-treated oocytes arrested with a bipolar spindle (Fig. 1E,F). Importantly, 5-Itu-treated oocytes failed to cluster MTOCs even after spending more than 8 h at metaphase I (15 or more hours total maturation time). Taken together, these results suggest an additional function of haspin in regulating bipolar spindle formation in meiosis I of mouse oocytes.

Haspin activity is required to cluster MTOCs

Bipolar spindle formation in mouse oocytes goes through several steps that occur after NEBD (Schuh and Ellenberg, 2007). To assess which of these steps are perturbed when haspin is inhibited, we added 5-Itu to oocytes microinjected with cRNAs for an MTOC marker, mEGFP–CDK5RAP2, and a chromosome marker, H2B–mCherry. Live oocytes were cultured in the presence of the fluorogenic probe SiR–tubulin to visualize spindle formation (Lukinavicius et al., 2014). Meiotic maturation and spindle formation were followed by time-lapse confocal microscopy. In somatic cells, CDK5RAP2 is a centrosomal protein that associates with the γ-tubulin ring complex (Fong et al., 2008). In control oocytes, we observed a typical pattern of MTOC clustering and bipolar spindle assembly (n=16) (Fig. 2A, upper panel; Movie 1) as previously reported (Clift and Schuh, 2015; Schuh and Ellenberg, 2007). After NEBD, chromosomes individualized, and fragmented MTOCs nucleated microtubules without well-defined poles (Fig. 2A, upper panel, time 1:00). Next, the spindle elongated and MTOCs started clustering together to form two spindle poles (Fig. 2A, upper panel, time 3:10) although at this time point many MTOCs were still not sorted into two poles. We also analyzed the area of the MTOCs and chromosomes at a fixed time during prometaphase I (NEBD+3 h) and plotted their distribution along the spindle axis. Control oocytes were in the process of clustering MTOCs, as most MTOCs were at the ends of the spindle axis, and they were organizing chromosomes into the metaphase plate. In contrast, 5-Itu-treated oocytes had not yet clustered MTOCs, as some foci were found in the center of the spindle axis, and chromosomes were still scattered along the spindle axis (n= 12) (Fig. 2B). Therefore haspin inhibition causes MTOC-clustering defects in prometaphase I.

Fig. 2.

Haspin-inhibited oocytes fail to cluster MTOCs. (A) Time-lapse imaging of live mouse oocytes expressing mEGFP–CDK5RAP2 (green, MTOCs) and H2B–mCherry (red, chromosomes) stained with SiR–tubulin (microtubules, gray). Mouse oocytes were treated with either ethanol (control, upper panel) or 1 µM 5-Iodotubercidin (5-Itu) (middle and lower panels). A maximum intensity z-projection of 16 sections of selected times are presented. White and yellow arrowheads indicate MTOCs outside of the clustered spindle pole; yellow arrows indicate microtubules that originate at spindle poles but do not make contact with chromosomes or the metaphase plate. The upper panel corresponds to Movie 1, the middle panel corresponds to Movie 2 and the lower panel corresponds to Movie 3. Time points are indicated relative to the time after nuclear envelope breakdown (hh: min). Scale bar: 10 µm. (B) Distribution of MTOCs (green) and chromosomes (red) along the spindle axis at 3 h after NEBD and at metaphase I in control and 5-Itu-treated oocytes. Distribution is expressed relative to the total area, and the length of spindle axis is arbitrarily set to 1. Data for individual representative oocytes are shown. (C) Quantification of the percentage of filmed oocytes with defective MTOC clustering. (D) Quantification of the percentage of filmed oocytes with chromosome segregation errors (n=16 for EtOH, and 12 for 5-Itu). *P<0.05; **P<0.01 (Fisher's exact test).

At metaphase I, control oocytes contained a bipolar meiotic spindle with fully clustered MTOCs and aligned chromosomes (Fig. 2A, upper panel, time 06:10 and 07:40). Furthermore, when analyzing the distribution of the MTOCs and chromosomes along the axis of the spindle, the MTOCs were clustered at the two ends and the chromosomes were at the metaphase plate in the center of the axis (Fig. 2B). In contrast to controls, 42% of 5-Itu-treated oocytes contained at least one MTOC distinct from the dominant spindle poles during metaphase I or anaphase I (Fig. 2A,C, arrowheads in middle panel, time 07:40, and lower panel, time 11:30; Movies 2,3). These defects were also evident when analyzing the distribution along the spindle axis because chromosomes were not centrally aligned at the metaphase plate, and MTOCs were not exclusively at the ends of the axis and were often at or near the metaphase plate (Fig. 2B). Moreover, in some treated oocytes spindle formation was defective because microtubules emanated from spindle poles to an area outside of chromosome plate (Fig. 2A, lower panel, time 11:30, yellow arrow) that persisted at anaphase I onset. These defects in MTOC clustering and spindle formation were accompanied by segregation errors during anaphase I (Fig. 2D).

We also observed that 5-Itu treatment significantly delayed onset of spindle formation by ∼16 min (Fig. S3A) without affecting the onset of spindle elongation (bipolarization) (Fig. S3B). Moreover, the microtubule density (Fig. S3C) and spindle volume (Fig. S3D), which are quantitative markers of microtubule polymerization, were significantly decreased early in prometaphase I (30 min after NEBD) in 5-Itu-treated oocytes compared to controls. However, these differences resolved by the time oocytes reached metaphase I. These analyses suggest that spindle formation is less efficient when haspin is inhibited, consistent with our previous report that there are delays in prometaphase I (Nguyen et al., 2014).

Taken together, haspin activity is required for clustering MTOCs into two functional spindle poles and influences the initial phase of spindle formation by regulating microtubule polymerization.

Haspin inhibition causes MTOC instability

Once two functional poles form, there is still ∼2–4 h before anaphase I onset. Therefore, the poles must be maintained to prevent aberrant chromosome segregation. To evaluate a role for haspin in spindle maintenance, we allowed bipolar spindles to form prior to inhibiting haspin with 5-Itu (Fig. 3A). As a control, we fixed oocytes at late prometaphase I (5 h) to confirm bipolarity. In the same experiment, subsets of oocytes were either treated with ethanol (EtOH) or 5-Itu at late prometaphase I for 2.5 h prior to fixation and confocal microscopy to examine spindle pole number. EtOH-treated control oocytes maintained bipolar spindles, whereas oocytes treated with the haspin inhibitor were unable to maintain bipolarity (Fig. 3B). Nearly one-half of these oocytes underwent pole fragmentation resulting in some MTOCs not being clustered into two main spindle poles (Fig. 3C).

Fig. 3.

Haspin inhibition at metaphase I causes MTOC instability. (A) Schematic to illustrate the timing of haspin inhibition as it correlates with the cell cycle stages. Meiotic resumption occurs at the start of the experiment when milrinone is washed from the maturation medium. Nuclear envelope breakdown, NEBD; prometaphase I, Pro-Met I; metaphase I, Met I. (B,C) Fully grown prophase-I-arrested oocytes were in vitro matured for 5 h prior to fixation (Pro-Met I) or followed by an additional maturation for 2.5 h in the presence of ethanol (EtOH) or 1 µM of 5-Iodotubercidin (5-Itu) prior to fixation. MTOCs (γ-tubulin; red in merge), spindle microtubules (α-tubulin; green in merge) and DNA (DAPI; blue in merge) were detected by performing confocal microscopy. Shown are representative z-projections and the arrows illustrate the multiple poles. (C) Quantification of the percentage of oocytes with multipolar spindles. The data are expressed as mean±s.e.m. The experiment was repeated three times and the total numbers of oocytes examined are indicated above the graph bars. **P<0.01 (one-way ANOVA). (D) Time-lapse imaging of live mouse oocytes expressing mEGFP–CDK5RAP2 (green, MTOCs) and H2B–mCherry (red, chromosomes) stained with SiR–tubulin (microtubules, gray). Mouse oocytes were treated with either ethanol (control, upper panel; n=15) or 1 µM 5-Iodotubercidin (5-Itu; n=15) (lower panel). Maximum intensity z-projection of 16 sections of selected times are presented. The white arrowheads show MTOCs that are no longer clustered and the white arrows indicate microtubules that emanate from a MTOC but do not make contact with a chromosome. Time points are indicated relative to the time after addition of 5-Itu (hh:min), which was added in late prometaphase I (5 h). The upper panel corresponds to Movie 4 and the lower panel corresponds to Movie 5. Scale bars: 10 µm.

To visualize the effect of haspin inhibition on the stability of spindle poles, we added EtOH or 5-Itu to oocytes expressing mEGFP–CDK5RAP2, H2B–mCherry and stained with SiR–tubulin in late prometaphase I (5 h, time 0) followed by time-lapse imaging (Fig. 3D; Movies 4 and 5). For technical reasons, time-lapse imaging began 45 min after drug treatment. As expected, control oocytes showed normal MTOC clustering and had spindles with two focused spindle poles at 0:45 (n=15). However, 50% of 5-Itu-treated oocytes already had defects where MTOCs were unclustered despite having a bipolar spindle (n=15). These data suggest that haspin inhibition compromised spindle pole integrity shortly after addition of 5-Itu and before the initiation of imaging. Control oocytes maintained clustered MTOCs and segregated homologous chromosomes in anaphase I. However, the 5-Itu-treatment group exhibited more oocytes with chromosome misalignment and segregation errors (40% of oocytes had errors in 5-Itu versus 13% with errors in control, P=0.0455). In addition, 5-Itu treated oocytes had significantly accelerated anaphase entry (188 min in control versus 98 min in 5-Itu, P=0.0011) as previously reported (Wang et al., 2016). Therefore, haspin activity is also required to maintain spindle pole integrity after MTOCs have coalesced.

Haspin does not regulate AURKA localization at MTOCs

Because AURKA localizes to MTOCs (Saskova et al., 2008; Shuda et al., 2009; Solc et al., 2012), and because its inhibition can cause spindle pole instability in mitotic cells (Asteriti et al., 2011), we wondered whether haspin regulates AURKA localization at MTOCs. First, we compared AURKA levels in oocytes fixed at metaphase I by performing immunocytochemistry. Compared to EtOH-treated controls, AURKA immunoreactivity was not significantly different in haspin-inhibited oocytes (Fig. 4A,B). Next, we asked whether the localized activated form of AURKA (phosphorylated AURKA, pAURKA) differs. Similar to our findings with AURKA, as assessed with a phospho-specific antibody that recognizes AURKA phosphorylated at T279 (T288 in human), the amount of active AURKA at MTOCs did not differ between control and inhibited groups (Fig. 4C,D). Therefore, haspin activity is not required to regulate AURKA MTOC localization during meiosis I in mouse oocytes.

Fig. 4.

AURKA localization and activation are not perturbed at MTOCs when haspin is inhibited. (A–D) Fully grown prophase-I-arrested oocytes were in vitro matured in the presence of ethanol (EtOH) or 500 nM of 5-Iodotubercidin (5-Itu) to metaphase I (7 h) prior to fixation and detection of MTOCs (γ-tubulin; red in merge), spindle microtubules (α-tubulin; green in merge) and DNA (DAPI; blue). (B,D) Quantification of the average pixel intensities of AURKA (B) or pAURKA (D) at MTOCs. The values in EtOH-treated oocytes were set to 1. The data are expressed as mean±s.e.m. These experiments were conducted two times and the total numbers of oocytes examined are indicated above the graph bars. There was no significant difference between treated and control oocytes (Student's t-test). Scale bars: 10µm.

Haspin activity is required for AURKC localization at MTOCs

In some cancer cell lines, AURKC localizes to centrosomes (Dutertre et al., 2005; Khan et al., 2011), and overexpression of AURKC can rescue multipolar spindle formation in RBBP4 and RBBP7-depleted oocytes (Balboula et al., 2014, 2015). Because we know that haspin regulates AURKC chromosomal localization (Nguyen et al., 2014), we re-evaluated AURKC and haspin localization. In our previous work with AURKA antibodies, we found that immunoreactivity at MTOCs required methanol fixation (Shuda et al., 2009). We hypothesized that AURKC had not been detected at MTOCs with standard paraformaldehyde fixation methods. Consistent with our hypothesis, AURKC colocalized with γ-tubulin at all stages of meiosis when oocytes were fixed in 100% methanol (Fig. 5A). To support our conclusion that the colocalization is specific, we treated oocytes with nocodazole to depolymerize the microtubules but keep MTOCs intact or we treated oocytes with taxol to stimulate synthesis of new MTOCs. In both instances, we observed colocalization of AURKC with γ-tubulin, indicating that AURKC localizes with meiotic MTOCs in addition to its traditional chromosome localization.

Fig. 5.

AURKC colocalizes with MTOCs. (A–C) Fully grown prophase-I-arrested oocytes were in vitro matured for 0 (Pro I, prophase I), 2 (Pro Met I, prometaphase I) or 7 h (Met I, metaphase I) prior to fixation and immunocytochemistry. (A) In two groups, metaphase I oocytes were treated with Nocodozole (Noc) for 10 min or Taxol for 45 min to perturb the spindle. AURKC (green in merge), MTOCs (γ-tubulin; red in merge) and DNA (DAPI; blue) were detected by confocal microscopy; shown are representative z-projections. The arrows indicate co-localization of AURKC and γ-tubulin. 53 oocytes were examined. (B) Representative z-projection, obtained with confocal microscopy, of colocalization of AURKC (green in merge), MTOCs (γ-tubulin; red in merge) and DNA (DAPI; blue) in ethanol (EtOH) or 5-Iodotubercidin (5-Itu) treatment groups at metaphase I. (C) Quantification of the intensity of EtOH- or 5-Itu-treated AURKC at MTOCs in oocytes. The values were normalized to the DNA signal and the average values in EtOH-treated oocytes were set to 1. These experiments were conducted three times and the total numbers of oocytes examined are indicated above the graph bars. The data are expressed as mean±s.e.m. *P<0.05 (Student's t-test). (D) Fully grown prophase-I-arrested oocytes were injected with GFP–haspin cRNA and matured to metaphase I (7 h) prior to fixation and confocal imaging. The brightness of the GFP channel was increased so that the portion of protein at the spindle poles could be visualized, thereby causing saturation on the chromosomes. Twenty oocytes were examined. Scale bars: 10µm.

To determine whether haspin activity is required for localization of AURKC at MTOCs, we asked whether AURKC MTOC localization was perturbed in 5-Itu-treated oocytes. In EtOH-treated oocytes, AURKC localized to the chromosomes and MTOCs (Fig. 5B,C). When haspin was inhibited with 5-Itu, we observed a reduction of chromosome-localized AURKC, as previously reported (Nguyen et al., 2014). Importantly, we also observed a loss of MTOC-localized AURKC. Furthermore, we re-evaluated localization of exogenous GFP-tagged haspin and found that a proportion of the protein localizes to spindle poles (Fig. 5D). Taken together, these data suggest that haspin targets AURKC to MTOCs during meiosis I.

AURKC inhibition causes MTOC-clustering defects that are CPC independent

To determine whether haspin regulates AURKC at MTOCs, we compared the phenotypes between AURKC-inhibited and haspin-inhibited oocytes. First, we expressed a dominant-negative allele of AURKC in mouse oocytes. This allele of AURKC (Aurkc-L93A; Aurkc-LA) is catalytically inactive, and inhibits wild-type (WT) AURKC, but not WT AURKB (Balboula and Schindler, 2014). We matured these and control-mCherry-injected oocytes to metaphase I and assessed MTOC clustering (Fig. 6A,B). Similar to 5-Itu-treated oocytes (Fig. 1A–C), we observed a failure to cluster MTOCs in 49.03±7.3% (mean±s.e.m.) of Aurkc-LA-expressing oocytes; whereas 6.66±2.3% of control oocytes failed to cluster MTOCs (Fig. 6A,B).

Fig. 6.

AURKC perturbation causes unclustered metaphase I spindle poles in a CPC-independent manner. (A–F) Fully grown prophase-I-arrested oocytes were in vitro matured to metaphase I (7 h) prior to fixation and detection of MTOCs (γ-tubulin, red in merge) or pINCENP (pink in merge), spindle microtubules (α-tubulin; green in merge) and DNA (DAPI; blue). (A) Oocytes were microinjected with mCherry (control) or Aurkc-L93A-mCherry (Aurka-LA-mCh) cRNA prior to maturation. The arrows indicate multiple MTOCs. (B) Quantification of the percentage of oocytes analyzed in A with unclustered poles. These experiments were conducted four times and the total numbers of oocytes examined are indicated above the graph bars. (C–F) Oocytes were microinjected with a morpholino oligonucleotide of either a scrambled sequence (control) or one designed to target INCENP [INCENP knockdown (KD)]. (D) Quantification of the intensity of pINCENP on chromosomes as shown in C. These experiments were conducted three times and the total numbers of oocytes examined are indicated above the graph bars. (F) Quantification of the percentage of oocytes analyzed in E with clustered MTOCs. These experiments were conducted two times and the total numbers of oocytes examined are indicated above the graph bars. The data are expressed as mean±s.e.m. Scale bars: 10 µm. **P<0.01; ****P<0.0001 (Student's t-test).

In other cell types, perturbation of components on chromosomes (i.e. AURKB or CPC components) can cause a failure to cluster centrosomes (Leber et al., 2010). To determine whether this clustering activity is specific to MTOC-localized AURKC, we depleted the CPC scaffolding subunit INCENP by injection of oocytes with a morpholino oligonucleotide. As previously demonstrated, INCENP depletion significantly reduces immunoreactivity with the phospho-specific INCENP antibody on chromosomes compared to injection with a scrambled control (Fig. 6C,D). Similarly, other components of the CPC, AURKC and survivin, were reduced upon INCENP depletion (Fig. S4). Importantly, when we quantified the incidence of unclustered MTOC foci, there was no statistical difference between control and INCENP-morpholino-injected oocytes (Fig. 6E,F). Therefore, the data support a model where haspin regulates AURKC localization to MTOCs, where it controls MTOC clustering independently of the CPC.

Overexpression of AURKC rescues MTOC-clustering defects

To confirm our hypothesis that haspin regulates the localization of AURKC to MTOCs to regulate bipolar spindle formation, we asked whether the MTOC-clustering defects could be rescued by overexpression of AURKC. To this end, we expressed GFP-tagged versions of AURKA, AURKB or AURKC in either EtOH- or 5-Itu-treated oocytes. Expression of AURKA or AURKB did not rescue the phenotype in 5-Itu-treated oocytes (Fig. 7). Importantly, we found that overexpression of AURKC partially rescued the clustering defect in 5-Itu-treated oocytes, reducing the incidence by half. Notably, AURKC–GFP localized to spindle poles, supporting the data with the anti-AURKC antibody (Fig. 7). Therefore, the data support that haspin functions upstream of AURKC to regulate MTOC clustering during oocyte meiosis.

Fig. 7.

Haspin regulates bipolar spindle assembly through AURKC. (A,B) Fully grown prophase-I-arrested oocytes were microinjected with the indicated cRNA and matured in the presence EtOH or 500 nM 5-Itu. The GFP channel (green in merge) demonstrates the localization of the GFP-tagged proteins. The arrows indicate MTOCs. (B) Quantification of the percentage of oocytes analyzed in A with unclustered MTOCs. ‘+’ indicates the addition of 500 nM 5-Itu, and a, b, c indicates the isoform of Aurk cRNA used in microinjection. The data are expressed as mean±s.e.m. These experiments were conducted two times and the total numbers of oocytes examined are indicated above the graph bars. Scale bar: 10 µm. *P<0.05; ***P<0.001 (one-way ANOVA).

DISCUSSION

In many species, oocytes lack classically defined centrosomes yet they build bipolar spindles (Dumont et al., 2007; Schuh and Ellenberg, 2007). In fact, spindles can form in cell-free extracts that lack centrosomes because of a chromatin-dependent microtubule nucleation pathway (Heald et al., 1996; Khodjakov et al., 2000; Mahoney et al., 2006). Mouse oocytes contain acentriolar centrosomes (MTOCs) that fragment and cluster to form meiosis I spindle poles. The mechanism by which oocytes use these MTOCs to build spindles is of great interest because of the high frequency of aneuploidy in female gametes (Hassold and Hunt, 2001), and the similarities of this clustering mechanism to that found in some cancers (Breuer et al., 2010; Kleylein-Sohn et al., 2012).

While conducting studies on the role of haspin in regulating chromosome segregation in mouse oocytes (Nguyen et al., 2014), we observed an increased incidence of MTOC foci in metaphase I spindles. This phenotype is described for the first time and it is explored in more detail here. By observing spindle formation and MTOC dynamics using high-resolution confocal microscopy, we found that the multiple MTOC foci that should cluster into two poles, fail to do so when haspin is inhibited (Fig. 2; Movies 1–3). Importantly, these unfocused foci nucleate microtubules that make stable connections with kinetochores suggesting that they act as functional spindle poles (Fig. 1D). The result of this phenotype is aberrant chromosome segregation (Fig. 2) and aneuploidy (Nguyen et al., 2014). The activity of haspin is also required for spindle-pole maintenance because addition of the inhibitor after bipolar spindle formation causes the MTOC foci to split (Fig. 3). These abnormal behaviors could be due to altered activity of the Eg5 kinesin, a known AURKA substrate that slides anti-parallel microtubules apart.

A similar phenotype has been described when haspin is depleted in mitotic tissue culture cells (U2OS cells) (Dai et al., 2009). Moreover, our observations that MTOC-localized AURKA is not perturbed in haspin-inhibited oocytes (Fig. 4), is similar to the data in the previous depletion study. However, there are two interesting differences. First, upon anaphase onset in mitosis, mis-segregation was followed by cytokinesis failure, mitotic exit and micronuclei formation. In oocytes, anaphase I proceeds with chromosomes being mis-segregated, but cells complete cytokinesis and progress to metaphase II. This scenario would be problematic because the resulting egg is usually aneuploid (Nguyen et al., 2014), and, if fertilized, would not produce a euploid embryo. The second difference between the cell types is the proposed role of haspin. In mitosis, haspin is required to maintain centromeric cohesion between sister chromatids. Depletion of cohesin complex members results in a similar multipolar spindle phenotype. Furthermore inhibition of topoisomerase II to artificially maintain sister chromatid association partially rescues this phenotype. We found that during meiosis I, haspin does not regulate sister chromatid cohesion, but instead regulates chromatin condensation, suggesting different functions during meiosis (Nguyen et al., 2014). We found that overexpression of AURKC rescues the MTOC clustering defect. One possible mechanism is that chromosome-localized AURKC–CPC regulates spindle formation through a chromatin-driven pathway, as has been described in other systems (Carazo-Salas et al., 1999; Dumont et al., 2007; Gruss et al., 2001; Radford et al., 2012; Sampath et al., 2004; Wiese et al., 2001). However, because we observed that endogenous and ectopic AURKC colocalized at MTOCs with γ-tubulin, a direct function at MTOCs or a combination of both are possible. A direct function is also supported by experiments where we depleted the CPC by knocking down INCENP, but did not observe an increase in MTOC-clustering defects (Fig. 6D,E). Further evidence for a direct role is the data demonstrating that when AURKC is overexpressed and rescued the spindle defect of 5-Itu-treated oocytes, we observed restoration of pole-localized but not chromosome-localized AURKC (Fig. 7A). These differences further support the model that the functions of haspin in meiosis I are fundamentally different from those in mitosis. These differences might not be surprising given that oocytes lack classic centrosomes, that homologous pairs of chromosomes are connected through chiasmata left over from recombination giving rise to a bivalent structure and that the length of prometaphase I is extremely long (∼7 h).

The well-established haspin substrate is threonine 3 on histone H3 on chromatin. Because haspin activity is required to recruit AURKC to MTOCs (Fig. 5B), the data suggest that haspin is phosphorylating other substrates. In an attempt to identify additional haspin substrates on chromatin, one study has identified some proteins known to localize and function at centrosomes that are phosphorylated in a haspin-dependent manner (Maiolica et al., 2014). This identification further supports the idea that haspin localizes at MTOCs, as we show here with GFP–haspin (Fig. 5D). One of the identified proteins is NuMA (also known as NUMA1), a regulator of spindle placement, organization and maintenance. NuMA, along with dynein, are required to anchor minus-end microtubules to centrosomes in mitosis. Mouse oocytes lacking NuMA have long and twisted meiosis I spindles, with unfocused poles (Kolano et al., 2012). However, the defects in MTOC clustering that we observe when haspin is inhibited were not observed in these knockout oocytes. Another centrosome substrate identified in the phosphoproteomic study was Cep72. Little is known about this protein, but siRNA knockdown in HeLA cells reduces γ-tubulin incorporation in centrosomes and microtubule nucleation activity (Oshimori et al., 2009). Importantly, the spindle poles in these cells were not focused. Cep72 function in mouse oocyte MTOCs has not yet been explored.

To our knowledge, this is the first report to identify localization of AURKC at MTOCs in mouse oocytes, and show that it might have a function outside of binding the CPC. We currently do not understand how the haspin–AURKC pathway is regulating MTOC clustering. One possible mechanism is that AURKC or its unknown binding partner could negatively regulate the affinity of Eg5 for microtubules as AURKA and TPX2 have been shown to do (Balchand et al., 2015; Giet et al., 1999). Although it was surprising that AURKC colocalized with MTOCs, this localization is consistent with reports demonstrating localization of human AURKC at centrosomes in cancer cell lines (Khan et al., 2011; Tsou et al., 2011). Some cancer cell types contain multiple centrosomes. Instead of building multipolar spindles and becoming apoptotic, these cells cluster the centrosomes into two poles to continue proliferation (Kramer et al., 2011; Lingle and Salisbury, 1999; Ring et al., 1982). It is tempting to speculate that aberrant expression of a meiosis-specific Aurora kinase could drive this clustering pathway in cancer cells. Future studies to compare the function of AURKC at MTOCs in mouse oocytes to that at centrosomes in cancer cells could provide insight as to why AURKC is oncogenic.

AURKC is considered the catalytic subunit of CPC that localizes to chromosomes (Balboula and Schindler, 2014; Schindler et al., 2012). Sequence-based evolution studies suggest that Aurkc arose as a genome duplication of Aurkb (Brown et al., 2004). Therefore, it is not surprising that these protein kinases can compensate for loss of one another and both function in the CPC (Balboula and Schindler, 2014; Schindler et al., 2012). AURKA is the Aurora kinase family member that localizes to spindle poles in nearly all organisms (Kufer et al., 2002; Marumoto et al., 2003; Rannou et al., 2008; Saskova et al., 2008). The finding that AURKC localizes to both chromosomes and MTOCs suggests that AURKC might function as a hybrid of or is a competitor of both AURKA and AURKB. Previous work has identified a single amino acid in subdomain IV of the AURKA that defines its specificity for binding TPX2 over INCENP (Fu et al., 2009; Hans et al., 2009). This specificity dictates whether the kinase is at centrosomes or chromosomes, respectively. Subdomain IV is a region that contributes to the structure of the ATP-binding lobe and is adjacent to subdomain V, which links the two lobes of the kinase together and coordinates ATP and substrate binding (Hanks and Hunter, 1995). These sequences determine the substrate specificity by establishing the size and charge of the opening. A simple comparison of sequences in this domain reveals that some residues of AURKC are identical to AURKA whereas others are identical to AURKB. Future work to explore what amino acids and what binding partners are required for AURKC localization at MTOCs will likely reveal new insights into AURKC function. To our knowledge, our data are the first to show a new function for haspin and AURKC to regulate MTOCs clustering, and to shed light on the underlying mechanism regulating bipolar spindle assembly in acentriolar oocytes, which are notoriously known for high rates of aneuploidy.

MATERIALS AND METHODS

cRNA preparation

Generation of non-degradable cyclin B (Ccnb1-Δ90), Aurka, Aurkb, Aurkc, Aurkc-L93A-mCherry, Gfp-Haspin (Nguyen et al., 2014) and H2B-mCherry constructs was described previously (Schindler and Schultz, 2009; Shuda et al., 2009). mEGFP-Cdk5rap2 cRNA was produced from a pGEMHE vector containing mouse full-length Cdk5rap2 (NM_145990.4) N-terminally fused to EGFP (a gift from Tomoya Kitajima, RIKEN, Japan) (Kitajima et al., 2011). To prepare cRNAs, plasmids were linearized followed by purification (Qiagen, QIAquick PCR Purification) and in vitro transcription using an mMessage mMachine T7 kit (Ambion) according to the manufacturer's instructions. The synthesized cRNAs were then purified using an RNAeasy kit (Qiagen) and stored at −80°C.

Oocyte collection, microinjection and culture

All animals were maintained following the Rutgers Institutional Animal Use and Care Committee (#11-032) and the National Institutes of Health guidelines, and according to the Department Expert Committee for the Approval of Projects of Experiments on Animals of the Academy of Sciences of the Czech Republic. Full-grown, germinal-vesicle-intact oocytes were collected from 6-week-old CF-1 female mice previously primed (48 h) by pregnant mare serum gonadotropin (PMSG) (Calbiochem #367222) (Schultz et al., 1983). Oocyte collection and microinjection were performed in bicarbonate-free minimal essential medium (MEM) containing, 25 mM HEPES, pH 7.3, 3 mg/ml polyvinylpyrrolidone (MEM/PVP) and supplemented with 2.5 µM milrinone (Sigma #M4659) to prevent maturation resumption (Tsafriri et al., 1996). After denudation, oocytes were either microinjected with ∼10 pl of the indicated material or cultured directly in Chatot, Ziomek and Bavister (CZB) medium plus 2.5 µM milrinone for 14 h (overnight). After overnight incubation, the oocytes were washed in milrinone-free CZB medium prior to meiotic maturation in the indicated drug in an atmosphere of 5% CO₂ at 37°C for 7 h.

Morpholino oligonucleotides were designed and produced by Gene Tools and injected at a concentration of 100 µM. The morpholino sequences are: INCENP, 5′-CGTCTTCCCGGACCACTCCTTTTCC-3′; haspin, 5′-GCCCGCCCGGCGTTTCAAACAGAAA-3′; and control, 5′-GCGCCCCCGCCGTTTCAAAGACAAA-3′. The Haspin siRNA (Life Technologies #S67095) was injected at a concentration of 100 µM.

Nocodazole (Sigma #M1404), taxol (Paclitaxel, Sigma #T1912), and ZM447439 (Tocris #2458) were dissolved in dimethyl sulfoxide (DMSO) and added to the CZB culture medium at a final concentration of 10 µM, 1.5 µM, 25 µM and 8 µM, respectively. 5-Iodotubercidin (5-Itu; Cayman Chemical, Ann Arbor, MI; #10010375) was dissolved in 100% ethanol and diluted in CZB medium to a final concentration of 500 nM or 1 µM in live imaging and prometaphase I experiments. We note that we found that oocytes from C57BL/6 mice require a higher dose of 5-Itu to see similar phenotypes to those in CF-1 mice. In vitro maturation was performed in organ culture dishes under humidified conditions (Becton Dickinson #353037).

Immunocytochemistry and imaging

For analysis of stable kinetochore–microtubule attachments, cold treatment of oocytes prior to fixation was performed as previously described (Balboula and Schindler, 2014). Oocytes were fixed using 3.7% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 1 h (to detect pericentrin, γ-tubulin or α-tubulin) or 2% PFA for 20 min at room temperature (to detect CREST, AURKA, pAURKA, AURKC, pINCENP or α-tubulin), 4% PFA overnight at 4°C for detection of survivin, or pre-chilled 100% cold methanol for 10 min at −20°C (to detect AURKC at poles) followed by permeabilization [PBS with 0.1% (v/v) Triton X-100 and 0.3% (w/v) bovine serum albumin (BSA)] for 20 min. The oocytes were then incubated in blocking solution (PBS, 0.3% BSA and 0.01% Tween-20) for 15 min. After blocking, fixed oocytes were incubated in primary antibody diluted in blocking solution for 1 h in a humidified chamber. The oocytes were then washed in blocking buffer followed by incubating for an additional hour in secondary antibody. Oocytes were then washed and mounted in 4–5 µl of Vectashield (Vector Laboratories, Burlingame, CA) containing 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI; 1:177, Sigma, St Louis, MO).

Fluorescence was detected using a Zeiss 510 Meta laser-scanning confocal microscope with a 40× objective with a zoom of two. The images were captured at the same laser power when experimental analysis required intensity measurements. The intensity of fluorescence was quantified using NIH image J software and processed under the same parameters for all oocytes in the same experiment. Regions of interest (ROIs) corresponding to the spindle pole area were selected. The relative values were obtained by dividing the mean intensity by the mean intensity measured in an equivalent background ROI. Metaphase I spindles were scored as unclustered if they had more than two dominant poles. Dominant poles were characterized by detecting MTOC protein (γ-tubulin) with obvious microtubule nucleation (α-tubulin).

Antibodies

The following primary antibodies were used in immunofluorescence: CREST autoimmune serum (Antibodies Incorporated; #15-234; 1:30), anti-AURKC (Bethyl #A400-023A- BL1217; 1:30) antibody, anti-α-tubulin–Alexa-Fluor-488 conjugate (Life Technologies #322588; 1:100), and anti-pericentrin (BD Biosciences #611814; 1:100), anti-γ-tubulin (Sigma #T6557; 1:100), anti-AURKA (Bethyl A300-072A; 1:30), anti-pAURKA (Cell Signaling #3079S; 1:30), anti-pINCENP (gift of Michael Lampson, University of Pennsylvania, Philadelphia, PA; 1:1000; Salimian et al., 2011) and anti-survivin (Cell Signaling Technology #2808S; 1:500) antibodies.

Live-cell imaging and image analysis

12–16-week-old C57BL/6 mice used for oocyte isolation were stimulated by 5 IU of PMSG 46 h prior to being killed. Oocytes were collected in M2 medium (Sigma-Aldrich #M7167) and cultured in Opti-MEM medium (ThermoFisher Scientific #31985062) supplemented with 4 mg/ml BSA (Sigma-Aldrich #A3311) in a 5% CO₂ atmosphere. Meiotic maturation was prevented by 2.5 μM milrinone (Sigma-Aldrich #M4659).

Oocytes were microinjected in M2 medium with 5–10 pl of 125 ng/µl mEGFP-Cdk5rap2 and 50 ng/µl of H2B-mCherry cRNAs (Kitajima et al., 2011). Microinjected oocytes were arrested in the germinal vesicle stage for 2 h to allow exogenous protein expression. Oocytes were stained with 100 nM SiR–tubulin for microtubule visualization (Lukinavicius et al., 2014). SiR–tubulin was added to both M2 and Opti-MEM medium.

Time-lapse image acquisition was performed using a Leica TCS SP5 microscope with an HCX PL Apo Lambda Blue 40×1.25 NA oil objective. Oocytes were scanned using sequential scan in between line mode at a 12-bit image depth with 7.75× or 9.6× zoom on the chromosome area. 3D scanning was performed using 2.5-µm optical sections through spindle volume. In the first sequential scan, oocytes were irradiated simultaneously with 4% of 488-nm and 3% of 633-nm laser lines for EGFP and SiR excitations. EGFP emission was detected at 500–570 nm and SiR emission was detected at 650–700 nm by two independent hybrid detectors in the standard mode. In the second sequential scan, mCherry was excited by 1.5% of 561-nm laser line and detected at 580–630 nm by the PMT detector.

Image analysis was performed using FiJi software (Schindelin et al., 2012). The measurements of MTOCs and chromosome distribution (Fig. 2B) were performed similarly to described previously (Breuer et al., 2010; Kolano et al., 2012). 3D stacks were processed by maximum z-projection and the spindle area was manually cropped and processed by using the straighten function along the spindle axis. Areas of MTOCs and chromosomes were segmented by intensity thresholding. MTOCs and chromosome distribution along spindle axis were calculated relative to the total areas of MTOCs and chromosomes, where length of spindle axis was set to 1.

Microtubule density (Fig. S3) was measured on the single confocal section situated in the center of the spindle. SiR–tubulin fluorescence intensity was quantified in a circle with a fixed defined diameter. For spindle volume measurements (Fig. S3), 3D volumes were reconstructed from a confocal section and spindles were segmented by intensity thresholding.

Statistical analysis

Statistical analysis was carried out using Prism GraphPad Software (La Jolla, CA). One-way ANOVA, Fisher's Exact and Student's t-test, as indicated in figure legends, were used to calculate the differences between groups with Tukey post-tests and 95% confidence intervals. A difference of P<0.05 was considered significant.