Aurora A kinase regulates proper spindle positioning in C. elegans and in human cells

Proper positioning of the mitotic spindle is fundamental for accurate placement of the cleavage furrow in metazoan organisms and, therefore, for the faithful distribution of fate determinants to daughter cells during development and in stem cell lineages (reviewed in Gönczy, 2008; Knoblich, 2008; Siller and Doe, 2009; Morin and Bellaïche, 2011). Errors in spindle positioning can result in aberrant proliferation and are thought to contribute to cancer (reviewed in Knoblich, 2010). Although it has been demonstrated that spindle mispositioning can lead to tumor formation in Drosophila (Caussinus and Gonzalez, 2005; Bowman et al., 2006), whether bona fide oncogenes or tumor suppressors impact on this process in human cells is incompletely understood.

The one-cell stage Caenorhabditis elegans embryo is an attractive model for dissecting the mechanisms underlying spindle positioning (reviewed in Kotak and Gönczy, 2013; Rose and Gönczy, 2014). In this system, the spindle assembles in the cell center before being displaced towards the posterior during metaphase and anaphase; during anaphase, this displacement is accompanied by vigorous oscillatory movements of the posterior spindle pole, transversely to the anterior–posterior embryonic axis. Asymmetric spindle positioning results from an imbalance of net pulling forces acting on the two spindle poles, with a larger net force pulling on the posterior side, which explains the oscillatory spindle pole movements on that side (Grill et al., 2001).

Pulling forces acting on the two spindle poles during mitosis of one-cell stage C. elegans embryos reflect the action of individual force generators located at the cell cortex, which exert forces on the plus end of astral microtubules abutting the confines of the cell (reviewed in Kotak and Gönczy, 2013; Rose and Gönczy, 2014). These cortical forces rely on an evolutionary conserved ternary complex consisting of two partially redundant heterotrimeric G protein α-subunits, GOA-1 and GPA-16, the essentially identical GoLoco Proteins GPR-1 and GPR-2, as well as the coiled-coil protein LIN-5 (Gotta and Ahringer, 2001; Colombo et al., 2003; Gotta et al., 2003; Srinivasan et al., 2003). The available evidence suggests that this ternary complex promotes anchoring of the minus-end-directed microtubule-dependent motor protein complex dynein (hereafter referred to as dynein) at the cell cortex (Nguyen-Ngoc et al., 2007; Couwenbergs et al., 2007; Kotak et al., 2012). Such cortically anchored dynein is thought to mediate spindle positioning by exerting pulling forces on astral microtubules (reviewed in Kotak and Gönczy, 2013; Rose and Gönczy, 2014).

Several components, including the Gβ and Gγ proteins GPB-1 and GPC-2, RIC-8, LET-99, CSNK-1 and PKC-3, have been reported to regulate the levels of ternary complex components at the cell cortex, and thereby modulate spindle positioning in one-cell C. elegans embryos (Tsou et al., 2002; Afshar et al., 2004; Afshar et al., 2005; Panbianco et al., 2008; Park and Rose, 2008; Thyagarajan et al., 2011; Galli et al., 2011). Another such component of particular relevance in the context of this study is a complex consisting of the protein phosphatase 6 (PP6) catalytic subunit PPH-6 and its associated subunit SAPS-1 (Afshar et al., 2010). Depletion of PPH-6 or SAPS-1 leads to an absence of the characteristic oscillatory movements of the posterior spindle pole during anaphase (Afshar et al., 2010). Accordingly, spindle-severing experiments, in which the spindle midzone is targeted using a laser micro-beam, and which thus reveal the extent of net pulling force acting on each liberated spindle pole (Grill et al., 2001), have established that pulling forces are drastically diminished in embryos where pph-6 or saps-1 have been knocked down by RNA interference (RNAi) (Afshar et al., 2010). Interestingly, this coincides with, and is probably caused by, substantially reduced levels of GPR-1 and GPR-2 (hereafter GPR-1/2) and of LIN-5 at the cell cortex during mitosis (Afshar et al., 2010). How PPH-6 or SAPS-1 depletion causes decreased cortical levels of the ternary complex is not known.

Aurora A is a serine/threonine kinase that is essential for centrosome separation, centrosome maturation and spindle assembly across metazoan evolution, including in C. elegans (Hannak et al., 2001; Giet et al., 2002; Toji et al., 2004; Tsai and Zheng, 2005; Wong et al., 2008). In human cells, Aurora A activity peaks at the G2/M transition, when the protein is enriched at centrosomes; thereafter, the protein is also enriched on spindle microtubules (reviewed in Hochegger et al., 2013). Aurora A protein levels then diminish abruptly in late M and early G1 owing to proteasomal degradation directed by the anaphase-promoting complex and its associated Cdh1 subunit (Littlepage and Ruderman, 2002; Crane et al., 2004).

Auto-phosphorylation of vertebrate Aurora A on threonine 288, a residue located in the activation loop, is needed for kinase activation (Littlepage et al., 2002). The same holds for the analogous residue threonine 201 in the C. elegans Aurora A protein AIR-1 (Toya et al., 2011). In human cells, biochemical and cell biological data indicate that the PP6 phosphatase catalytic subunit PPP6C acts as a T-loop phosphatase for threonine 288, thus negatively regulating Aurora A kinase activity (Zeng et al., 2010). Whether the same holds in C. elegans is not known. In addition to the above mode of activation, Aurora A is also allosterically activated through association with TPX2, a protein that is also needed for targeting Aurora A to spindle microtubules (Kufer et al., 2002; Zorba et al., 2014). Accordingly, TPX2 depletion in human cells leads to the assembly of shorter spindles and to chromosome instability (Bird and Hyman, 2008).

Aurora A is important for proliferation control and, in doing so, can act as an oncogene or instead as a tumor suppressor, depending on the system. For instance, Aurora A overexpression results in transformation of mammalian tissue culture cells and is amplified or otherwise overexpressed in several human tumors, with Aurora-A-containing amplicons correlating with poor prognosis (reviewed in Karthigeyan et al., 2011; D'assoro et al., 2015). By contrast, loss of Aurora A function in Drosophila larval neuroblasts results in tissue overgrowth, indicating that the kinase negatively regulates proliferation in this setting (Wang et al., 2006; Lee et al., 2006). Intriguingly, Aurora A has been implicated in regulating spindle orientation in Drosophila, as well as in murine mammary epithelial cells (Wang et al., 2006; Lee et al., 2006; Regan et al., 2013), and most recently in human cells (Gallini et al., 2016). Whether these requirements extend to C. elegans where spindle positioning can be investigated with high spatiotemporal resolution is not known.

We have reported previously that depletion of saps-1 by RNAi [denoted saps-1(RNAi)] in one-cell stage C. elegans embryos causes a spindle positioning phenotype characterized by a lack of oscillatory movements of the posterior spindle pole during anaphase (Afshar et al., 2010; Fig. 1B, compare with Fig. 1A; Movies 1 and 2). Moreover, spindle-severing experiments established that cortical pulling forces are dampened both on the anterior and posterior spindle poles in anaphase saps-1(RNAi) embryos (Afshar et al., 2010; Fig. 1E).

To investigate further the mechanisms by which PPH-6–SAPS-1 modulates spindle positioning, we used a proteomics approach to identify interacting components. We conducted immunoprecipitation experiments with anti-GFP antibodies in lysates prepared from embryos expressing GFP–SAPS-1, as well as from wild-type embryos that served as a negative control. This approach led to the identification of the Aurora A kinase AIR-1 as a candidate GFP–SAPS-1 interactor (Fig. S1A,B). Co-immunoprecipitation experiments further established that GFP–SAPS-1 interacts with endogenous AIR-1 (Fig. S1C). Moreover, additional proteomic analyses using antibodies against AIR-1 or SAPS-1 ascertained that endogenous SAPS-1 and AIR-1 also interact (data not shown). An interaction between these two proteins has also been reported in a extensive fragment-based screen using a heterologous yeast two-hybrid system (Boxem et al., 2008), suggesting that this interaction is direct. Taken together, these data indicate that AIR-1 and PPH-6–SAPS-1 associate with each other in C. elegans embryos.

Given the interaction between AIR-1 and SAPS-1, we sought to investigate the role of AIR-1 in spindle positioning. However, this could not be assessed using RNAi, because spindle assembly is defective in one-cell stage air-1(RNAi) embryos due to an earlier requirement in centrosome separation and spindle assembly (Schumacher et al., 1998; Hannak et al., 2001).

To circumvent this issue, we sought to utilize a chemical approach to acutely inactivate AIR-1 after centrosome separation and spindle assembly have taken place. VX-680 is a selective small-molecule inhibitor that targets the ATP-binding pocket of human Aurora A (Harrington et al., 2004). The kinase domain is conserved among Aurora family members, and structure prediction of AIR-1 suggests that VX-680 could also prevent the activity of C. elegans AIR-1 (Fig. S2A). To test whether this is the case indeed, we bathed early one-cell stage embryos in a solution containing 100 nM VX-680; owing to the presence of the eggshell, embryos were initially shielded from the drug. To expose embryos to the effect of VX-680, we used a laser micro-beam to pierce a hole in the eggshell early in the cell cycle. As shown in Fig. S2C, we found that this leads to a canonical air-1(RNAi) phenotype, with a lack of centrosome separation and spindle assembly (Schumacher et al., 1998; Hannak et al., 2001; compare with Fig. S2B). We conclude that VX-680 is a suitable reagent to acutely inactivate C. elegans AIR-1.

Therefore, we set out to subject embryos to VX-680 after spindle assembly, using either the laser micro-beam to pierce the eggshell or by adding the drug to perm-1(RNAi) embryos, which have a compromised eggshell (Carvalho et al., 2011). Interestingly, we found that the spindle undergoes exaggerated oscillations during anaphase in embryos acutely subjected to VX-680 using either method (Fig. 1C; Movie 3). Analogous results were obtained with another Aurora A inhibitor, MLN 8054 (data not shown). Despite such exaggerated oscillations, we found that the spindle eventually ends up in the embryo posterior, leading to unequal first division (Movie 3), as observed previously in other conditions characterized by excess oscillations, including following depletion of the negative force regulator Gβγ or upon YFP–GPR-1 overexpression (Afshar et al., 2005; Redemann et al., 2011). Cytokinesis proceeds normally in embryos acutely subjected to VX-680 or MLN 8054, indicating that these drugs do not target the Aurora B kinase AIR-2 under these experimental conditions (Movie 3). Importantly, spindle-severing experiments in embryos treated with VX-680 revealed excess pulling forces on the liberated anterior spindle pole (Fig. 1E); the absence of an apparent impact on the liberated posterior spindle pole might be due to geometric constraints, as is thought to be the case upon inactivation of Gβγ (Afshar et al., 2005).

We surmise that the increase in pulling forces following AIR-1 depletion is accompanied by excess cortical enrichment of ternary complex components and of dynein. Unfortunately, we could not verify this experimentally because monitoring levels of ternary complex components and of dynein during mitosis in embryos acutely treated with VX-680 has proven challenging.

We next addressed whether the role of AIR-1 during spindle positioning relies on TPXL-1, the C. elegans homolog of human TPX2, which in the worm acts as an AIR-1 activator in promoting centrosome separation and spindle assembly (Özlü et al., 2005). We found that tpxl-1(RNAi) embryos do not exhibit exaggerated oscillations during anaphase (Fig. 1D; Movie 4), indicating that the role of AIR-1 in spindle positioning is independent of TPXL-1. Taken together, we conclude that AIR-1 negatively regulates cortical pulling forces in a TPXL-1-independent manner.

We next used VX-680 to address the relationship between PPH-6–SAPS-1 and AIR-1 during spindle positioning. If PPH-6–SAPS-1 acts upstream of AIR-1, then the phenotype of embryos lacking both PPH-6–SAPS-1 and AIR-1 function should resemble that of embryos lacking solely AIR-1. Conversely, if PPH-6–SAPS-1 acts downstream of AIR-1, then the phenotype of embryos lacking both PPH-6–SAPS-1 and AIR-1 function should resemble that of embryos lacking solely PPH-6–SAPS-1. By contrast, if AIR-1 and PPH-6–SAPS-1 act in parallel pathways that independently modulate pulling forces, then embryos depleted of both components might exhibit an intermediate, wild-type-like, phenotype. Importantly, we found that saps-1(RNAi) embryos in which AIR-1 activity was inhibited using VX-680 exhibited exaggerated oscillations and cortical pulling forces, just like wild-type embryos treated with VX-680 (Fig. 1E; Fig. S2D,E). These experiments suggest that PPH-6–SAPS-1 acts upstream of AIR-1 to modulate spindle positioning, in line with the finding in human cells that PP6 can negatively regulate Aurora A by de-phosphorylating it on threonine 288 (Zeng et al., 2010).

Given the above result, we investigated whether PPH-6–SAPS-1 influences AIR-1 localization during mitosis. In wild-type embryos, AIR-1 localizes consistently to the two centrosomes and to the astral microtubules emanating from them (Fig. 2A; Hannak et al., 2001). As shown in Fig. 2B, we found that AIR-1 also localizes to centrosomes and astral microtubules in saps-1(RNAi) embryos. Strikingly, in addition, we discovered that AIR-1 is enriched at the cell cortex in embryos depleted of SAPS-1 (compare Fig. 2B with Fig. 2A). An analogous distribution of AIR-1 was observed upon PPH-6 depletion (data not shown). Close examination of wild-type embryos, in particular during subsequent stages of embryogenesis, revealed traces of AIR-1 cortical enrichment (Fig. S3A, arrow), suggesting that the excess accumulation observed upon PPH-6–SAPS-1 depletion likely reflects a physiologically relevant distribution. We conclude that PPH-6–SAPS-1 negatively regulates the localization of AIR-1 at the cell cortex.

We set out to explore how the presence of AIR-1 at the cell cortex might result in diminished pulling forces in saps-1(RNAi) embryos. As anticipated from the lack of defects in centrosome separation and in spindle assembly, we found that the distribution of centrosomal TBG-1 (γ-tubulin), TAC-1 and MCAK are indistinguishable from the wild-type in saps-1(RNAi) embryos (data not shown), as is that of TPXL-1 (compare Fig. S4B with Fig. S4A). Likewise, the nucleation and growth rates of astral microtubules, as well as their cortical residency time, were similar to those in the wild-type (data not shown), together indicating that cortical AIR-1 does not alter spindle positioning by modulating astral microtubule dynamics in a substantial way.

We also tested whether the ternary complex might somehow modulate cortical AIR-1. However, no accumulation of cortical AIR-1 was observed in double gpr-1 and gpr-2 RNAi [gpr-1/2(RNAi)] or lin-5(RNAi) embryos (data not shown). Moreover, the levels of cortical AIR-1 upon SAPS-1 depletion were not affected by the additional depletion of the Gα proteins GOA-1 and GPA-16, or GPR-1/2 (Fig. 2C,D). Although the mechanisms by which cortical AIR-1 modulates pulling forces remain to be deciphered, these experiments lead us to conclude that AIR-1 negatively regulates spindle positioning independently of an impact on astral microtubules. Moreover, we conclude that the ternary complex is not needed for AIR-1 cortical localization upon PPH-6–SAPS-1 depletion.

We were interested in investigating whether AIR-1 cortical enrichment upon PPH-6–SAPS-1 depletion needs the presence of AIR-1 on astral microtubules. As shown in Fig. S3B, we found that the depletion of TBG-1 does not impair cortical accumulation of AIR-1 in saps-1(RNAi) embryos, suggesting that such accumulation does not require a robust astral microtubule network. Furthermore, we analyzed embryos depleted of TPXL-1, which is required for the presence of AIR-1 on astral microtubules (Özlü et al., 2005). We found that AIR-1 cortical enrichment is still observed in saps-1(RNAi) tpxl-1(RNAi) embryos (Fig. 2E), albeit to a lesser extent in some embryos (data not shown).

Taken together, these results indicate that PPH-6–SAPS-1 negatively regulates the localization of AIR-1 at the cell cortex in a manner that does not depend on either TPXL-1 nor on the presence of AIR-1 on astral microtubules.

We set out to address whether the pool of cortical AIR-1 accumulating upon PPH-6–SAPS-1 depletion has hallmarks of an active kinase. To this end, we utilized phosphospecific threonine 201 antibodies to monitor the status of this residue, which is located in the activation loop and whose phosphorylation is required for AIR-1 kinase activity (Toya et al., 2011). We did not observe a signal at the cell cortex upon SAPS-1 depletion (compare Fig. S3D with Fig. S3C). These results raise the possibility that the pool of cortical AIR-1 that accumulates upon SAPS-1 depletion is not phosphorylated on this residue, and might thus not be active, although we cannot exclude that phosphorylated threonine 201 is present at the cell cortex below detection levels (see Discussion).

We set out to test whether AIR-1 kinase activity is needed for cortical accumulation of the protein upon PPH-6–SAPS-1 depletion. We found that, just like endogenous AIR-1, GFP–AIR-1 could be readily detected at centrosomes and astral microtubules in transgenic worms expressing GFP–AIR-1 (Fig. 3A) (Toya et al., 2011). We found furthermore that GFP–AIR-1 is enriched at the cell cortex of saps-1(RNAi) embryos (Fig. 3B), in line with the distribution of endogenous AIR-1. By contrast, we found that kinase-dead GFP-AIR-1^T201A was not enriched at the cell cortex in saps-1(RNAi) embryos (Fig. 3C). Moreover, spinning disc confocal microscopy of embryos expressing GFP–AIR-1 allowed the visualization of the cortical GFP signal in saps-1(RNAi) embryos, most readily in the second cell cycle (Fig. 3E, arrow and arrowhead, compare with Fig. 3D). Importantly, in addition, we found that acute treatment with VX-680 during mitosis results in the loss of the fusion protein from the cortex (compare Fig. 3F with Fig. 3E). These findings led us to conclude that AIR-1 kinase activity is essential for cortical localization of the protein in saps-1(RNAi) embryos. Given the result with the phosphospecific threonine 201 antibodies, whether this activity is needed at the cell cortex or elsewhere in the cell is an open question. Regardless, these findings are compatible with the notion that cortical AIR-1 causes the reduction of pulling forces observed upon PPH-6–SAPS-1 depletion.

Given the evolutionarily conservation of Aurora A kinases and PP6 phosphatases, we addressed whether the function of these two players in spindle positioning is conserved in human cells. To this end, we monitored spindle positioning by analyzing HeLa cells fixed on coverslips uniformly coated with fibronectin (Fig. 4A; Toyoshima and Nishida, 2007; Kotak et al., 2012). Interestingly, we found that partial depletion of Aurora A using siRNAs or a 60-min incubation with MLN 8054 causes striking spindle positioning defects (Fig. 4B–G), as reported also recently elsewhere (Gallini et al., 2016). Moreover, we found that siRNA-mediated depletion of PPP6C also causes spindle-positioning defects (Fig. 4J,K). However, we found that PPP6C depletion did not lead to detectable cortical enrichment of Aurora A in human cells (data not shown).

Given that Aurora A is important for centrosome maturation and thus efficient microtubule nucleation during mitosis (reviewed in Hochegger et al., 2013), we wondered whether the impact of Aurora A on spindle positioning in human cells might be due to an impact on astral microtubules. However, we found that astral microtubules are not noticeably affected and reach out normally towards the cortex in MLN-8054-treated cells (Fig. 4H,I, arrowheads). Therefore, the impact of Aurora A inactivation on spindle positioning in human cells is unlikely to be mediated through an impairment of astral microtubules.

How does Aurora A modulate spindle positioning in human cells? A large-scale quantitative phosphoproteomics approach has identified the LIN-5 ortholog NuMA (also known as NUMA1) as an Aurora A substrate (Kettenbach et al., 2011). Since cortical NuMA and dynein are necessary for proper spindle positioning in human cells (Woodard et al., 2010; Kotak et al., 2012; Kiyomitsu and Cheeseman, 2012), we investigated the consequence of MLN-8054-mediated Aurora A inactivation on the distribution of NuMA, as well as on the dynein-associated dynactin subunit p150^Glued (also known as DCTN1). Interestingly, we found that treatment with MLN 8054 results in depletion of NuMA and p150^Glued from the cell cortex of metaphase cells and concomitant substantial accumulation at spindle poles (compare Fig. 5B with Fig. 5A). Similar findings have been recently reported elsewhere (Gallini et al., 2016).

We note that given the increased cortical pulling forces observed upon AIR-1 inactivation in worms (see Fig. 1E), an increase of cortical NuMA and p150^Glued might have been anticipated upon Aurora A inactivation in human cells if the findings in C. elegans were strictly transferable. Likewise, depletion of PPP6C might have been anticipated to cause an augmentation of cortical NuMA and p150^Glued; however, no such alteration has been found (Kotak et al., 2013). Regardless of the root of these apparent differences between systems, we conclude that the activity of Aurora A kinase and PP6 phosphatases orchestrates spindle positioning from worms to men.

We were interested in investigating the relationship between Aurora A and the master cell cycle regulator CDK1 in modulating spindle positioning in human cells. It has been shown previously that acute CDK1 inactivation during metaphase using RO-3306 results in a substantial increase in cortical localization of NuMA and p150^Glued, and a concomitant decrease at spindle poles, thus mimicking the situation normally observed during anaphase (compare Fig. 5D and Fig. 5C; Kotak et al., 2013). We addressed whether CDK1 inactivation is sufficient to ensure such higher cortical levels, or whether the presence of Aurora A kinase activity is needed in addition. As shown in Fig. 5E, we uncovered that the latter is the case, because treatment of cells with both RO-3306 and MLN 8054 led to a loss of cortical NuMA and p150^Glued during metaphase, as observed upon Aurora A inactivation alone. Furthermore, we found a substantial decrease in cortical NuMA in MLN-8054-treated anaphase cells (compare Fig. 5G with Fig. 5F). Interestingly, in addition, we found that RO-3306 treatment of cells partially depleted of TPX2 does not impact upon cortical NuMA and p150^Glued levels (Fig. 5H–K). This result indicates that the function of Aurora A in mediating excess NuMA and p150^Glued cortical localization is independent of its co-activator TPX2, thus mirroring the observations with TPXL-1 in C. elegans.

Overall, these findings lead us to propose that CDK1 inactivation at the metaphase to anaphase transition is not sufficient to ensure substantial cortical levels of NuMA at the cell cortex, but that Aurora A kinase activity must be maintained in addition, and this occurs in a TPX2-independent manner.

Accurate positioning of the mitotic spindle is essential during development and in stem cell lineages (reviewed in Knoblich, 2008; Kotak and Gönczy, 2013). Here, we demonstrate that Aurora A kinases regulate spindle positioning from worms to men, which might have bearing on proliferation control.

In a search for PPH-6–SAPS-1 partners in C. elegans, we uncovered an interaction with the kinase AIR-1. We discovered that loss of AIR-1 function causes excess spindle-pulling forces during anaphase, a phenotype that is opposite from that incurred upon PPH-6–SAPS-1 depletion (Fig. 6; Afshar et al., 2010). Furthermore, our work indicates that AIR-1 acts downstream of PPH-6–SAPS-1 in regulating spindle positioning. How does PPH-6–SAPS-1 control AIR-1 function? In vertebrate cells, autophosphorylation of threonine 288 is required for Aurora A catalytic activity (Littlepage et al., 2002), and PP6 phosphatase dephosphorylates this residue, thus dampening kinase activity (Zeng et al., 2010). Somewhat unexpectedly, we did not observe enrichment of phospho-threonine 201 at centrosomes or at the cortex in saps-1(RNAi) embryos. It is conceivable that PPH-6–SAPS-1 negatively regulates AIR-1 kinase function by acting on another residue than threonine 201. However, based on the observation that kinase activity is needed for cortical enrichment, we favor the view that the lack of cortical phospho-threonine 201 reflects a detection limit in these experiments.

How does AIR-1 negatively regulate cortical pulling forces in C. elegans? Given that AIR-1 kinase activity is essential for this role, we postulate that AIR-1 phosphorylates a ternary complex protein or a dynein component, thus negatively regulating its function in spindle positioning. Although the molecular nature of the relevant target in C. elegans remains to be identified, analogies with other systems raise the possibility that GPR-1/2 or LIN-5 could be important. In Drosophila larval neuroblasts, Aurora A ensures proper spindle positioning by restricting the cortical localization of the LIN-5 ortholog Mud to the apical side of the cell (Wang et al., 2006; Lee et al., 2006). Whether Mud is directly regulated by Aurora A kinase is not known, whereas it has been established that Drosophila Aurora A phosphorylates the GPR-1/2 homolog Pins on serine 436 (Johnston et al., 2009). This residue does not appear to be conserved in C. elegans GPR-1/2 (Johnston et al., 2009), but perhaps a divergent phosphorylation site in these proteins is the target of AIR-1 in the worm. In human cells, NuMA appears to be the crucial target of Aurora A in the context of spindle positioning. Both NuMA and the dynein-associated dynactin component p150^Glued are phosphorylated by Aurora A (Kattenbach et al., 2011; Rome et al., 2010; Reboutier et al., 2013), and quantitative phosphoproteomics has uncovered that NuMA can be directly phosphorylated by Aurora A at serine 1969 (Kattenbach et al., 2011). We found that this site appears to be conserved in C. elegans LIN-5 (Fig. S4C), suggesting that it could likewise be directly phosphorylated by AIR-1. Moreover, in human cells, mutation of serine 1969 into a non-phosphorylatable alanine residue impairs cortical NuMA localization, as is the case upon Aurora A inactivation (Kattenbach et al., 2011). Interestingly, recent work has revealed that phosphorylation of serine 1969 by Aurora A regulates the mobility of NuMA at spindle poles; upon Aurora A knockdown, NuMA is depleted from the cell cortex and accumulates at the spindle poles (Gallini et al., 2016), as was found also in our study.

Intriguingly, the molecular phenotype upon Aurora A depletion in human cells, namely loss of cortical NuMA, is not what would have been expected based on the results obtained in C. elegans, where AIR-1 inactivation causes excessive pulling forces, likely due to cortical enrichment of the ternary complex. Regardless of this apparent difference in mode of action, our data unequivocally establishes that Aurora A plays an evolutionarily conserved role in spindle positioning from worms to humans.

The molecular signature of Aurora A inactivation in human cells is the opposite from that observed upon CDK1 inactivation (Kotak et al., 2013). Prompted by these opposite signatures, we characterized the interplay between CDK1 and Aurora A in ensuring proper levels of cortical NuMA and p150^Glued. Our findings, taken together, demonstrate that Aurora A must be active to enable cortical accumulation of NuMA and dynein during metaphase and anaphase. Given that robust cortical NuMA and dynein is needed for proper spindle elongation during anaphase (Kotak et al., 2013), this likely explains why acute Aurora A inactivation correlates with spindle elongation defects (Reboutier et al., 2013) (Fig. 6).

Aurora A is essential for the initial activation of CDK1 at centrosomes that is required for timely entry into mitosis (Hirota et al., 2003). In the present work, we reveal that during metaphase, Aurora A is involved in a tug-of-war with CDK1 to modulate cortical NuMA and dynein levels, whereby CDK1 negatively and Aurora A positively regulates these levels. Acute inactivation of CDK1 causes a substantial cortical NuMA enrichment (Kotak et al., 2013), and one possibility is that this is due to the loss of a negative regulation of Aurora A by CDK1 that normally occurs during metaphase. Intriguingly, our data further reveal that TPX2 is not needed for NuMA and p150^Glued cortical enrichment upon CDK1 depletion, suggesting that Aurora A functions independently of this co-activator in regulating accurate cortical levels of NuMA and p150^Glued, and thus proper spindle positioning. During anaphase, CDK1 inactivation would allow the remaining Aurora A activity to promote cortical NuMA and dynein enrichment. Aurora A is not alone in this endeavor, however, because the activity of PP2A phosphatase is also important to ensure a substantial cortical localization of NuMA and dynein during anaphase (Kotak et al., 2013). In the future, it will be interesting to investigate whether PP2A exerts this function through modulation of Aurora A activity. Furthermore, because Aurora A is commonly amplified in a range of human cancers of epithelial origin (reviewed in Marumoto et al., 2005), it will also be interesting to determine whether altered Aurora A levels in such settings impact upon cortical NuMA and dynein and thus alter spindle positioning, potentially fueling tumorigenesis.

After immunoprecipitations, the beads were washed three times with PBS and resuspended in SDS gel loading buffer to extract bound proteins. Samples were migrated on a 10% mini polyacrylamide gel for about 2.0 cm, and stained with Coomassie Blue. Entire gel lanes were excised into five equal regions from top to bottom and digested with trypsin (Promega) as previously described (Shevchenko et al., 1996; Wilm et al., 1996). Data-dependent liquid-chromatography tandem mass spectrometry (LC-MS/MS) analysis of extracted peptide mixtures after digestion was carried out on a hybrid linear trap LTQ-Orbitrap XL mass spectrometer (Thermo Fisher Scientific) interfaced to a nanocapillary HPLC equipped with a C18 reversed-phase column. Collections of tandem mass spectra for database searching were generated from raw data and searched using Mascot (Matrix Science, London, UK; version 2.1.0). The software Scaffold (version Scaffold-01_06_03, Proteome Software Inc.) was used to validate MS/MS-based peptide (minimum 95% probability) and protein (min 99% probability) identifications, perform dataset alignment and subtraction, as well as parsimony analysis to discriminate homologous hits. Differentially detected proteins (absent in one sample or with varying numbers of spectra) were manually validated.

C. elegans wild-type (N2) as well as transgenic lines expressing GFP–SAPS-1 (Afshar et al., 2010), GFP-AIR-1 (Hannak et al., 2001), as well as recoded and RNAi-resistant GFP-AIR-1^R or a kinase-dead mutant equivalent (GFP-AIR-1^R-T201A) (Toya et al., 2011) were maintained at 24°C. Bacterial RNAi feeding strains for saps-1, air-1, tpxl-1 and perm-1 were obtained from the C. elegans ORFeome RNAi library (gift from Jean- François Rual and Marc Vidal, Harvard Medical School, Boston, MA) or from Source BioScience (Kamath et al., 2003). gpr-1/2(RNAi); goa-1/gpa-16(RNAi) constructs are as described previously (Colombo et al., 2003), as is the air-1^N(RNAi) feeding strain depleting only the endogenous pool of AIR-1 (Toya et al., 2011). RNAi for saps-1, air-1, air-1^N(RNAi), saps-1 air-1 and saps-1 tpxl-1 was performed by feeding animals starting at the L3 stage with bacteria expressing the corresponding dsRNAs at 20°C for 30-36 h before analysis.

Time-lapse DIC microscopy and dual DIC and fluorescence microscopy were performed essentially as described (Bellanger and Gönczy, 2003). Movies were subsequently processed using ImageJ and QuickTime. Spindle severing was performed using a Leica LMD microscope equipped with a pulsed N2 laser (337 nM) (Afshar et al., 2004). The central region of the spindle was cut either at metaphase, just before the slight posterior shift of the spindle or at the onset of anaphase. Nuclear envelope breakdown (NEBD) was used as an additional reference point for timing the cuts. Subsequent tracking of spindle poles and measurements of average peak velocities were conducted essentially as described (Grill et al., 2001). In the experiments with VX-680 or MLN 8054, embryos were first bathed in M9 containing 100 nM of VX-680 or 500 nM of MLN 8054 in 0.5% DMSO. The eggshell protects the embryo from the drug until permeabilization, which was performed with the laser microbeam 3–5 min prior to NEBD. 0.5% DMSO alone did not induce any phenotype (data not shown).

HeLa cells expressing GFP–Centrin 1 (Piel et al., 2000) were cultured in high-glucose DMEM with GlutaMAX (Invitrogen) medium supplemented with 10% fetal calf serum (FCS) in a humidified 5% CO2 incubator at 37°C. For monitoring spindle positioning in fixed specimens, cells were grown on coverslips uniformly coated with fibronectin (BD Bioscience, 354088) and synchronized using a double-thymidine block. In brief, cells were incubated with 2 mM thymidine for 17 h, released for 8 h and again incubated with 2 mM thymidine for 17 h. Cells were then released by washing with PBS and fixed 10 h thereafter, when a maximum number of cells were in mitosis.

For small interfering RNA (siRNA) experiments, ∼100,000 cells were seeded either on fibronectin-coated coverslips or regular glass coverslips in six-well plates. 6 μl of 20 μM siRNA in 100 μl OptiMEM medium (Invitrogen, Carlsbad, CA) and 4 μl of Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA) in 100 μl OptiMEM were incubated in parallel for 5 min, mixed for 20 min and then added to 2.5 ml of medium per well. Double-stranded siRNA oligonucleotides against human Aurora A kinase were synthesized with the sequence 5′-UAGGAAAUCAUGAUCAAGCAA-3′ (Aurora A siRNAs, Qiagen), against TPX2 with the sequence 5′-CCCACCGAGCCTATTGGCTTTGATT-3′ (TPX2 siRNAs, Stealth Life Technology) and against PPP6C with the sequence 5′-GACTACGTTTGTGACCTCCTCTTAG-3′ (PPP6C siRNAs, Stealth Life Technology). For acute Aurora A inactivation, cells were incubated for 60 min with 250 nM MLN 8054 before analysis. For CDK1 inactivation, cells were treated with 10 μM RO-3306 for 10 min before fixation. In all the experiments where an inhibitor was used, control cells were treated with DMSO at a concentration of <0.01% (written as DMSO for simplicity in the figure legends).

Embryo fixation and staining for indirect immunofluorescence was performed essentially as described previously (Gönczy et al., 1999), using 1:200 mouse anti-α-tubulin antibodies (DM1A, Sigma), in combination with one of the following antibodies: 1:200 rabbit anti-AIR-1 (Hannak et al., 2001), 1:200 rabbit anti-SAPS-1 (Afshar et al., 2010), 1:200 rabbit anti-TPXL-1 (Özlü et al., 2005), 1:300 mouse anti-α-tubulin (DM1a, Sigma-Aldrich), 1:2000 mouse anti-γ-tubulin (GTU88, Sigma-Aldrich), 1:300 mouse anti-GFP (MAB3580, Millipore) and 1:200 rabbit anti-GFP (gift from Viesturs Simanis; EPFL, Lausanne, Switzerland). Embryos were fixed in methanol at −20°C for 30 min and incubated with primary antibodies for 1 h at room temperature. Secondary antibodies were Alexa-Fluor-488-coupled anti-mouse-IgG and Alexa-Fluor-568-coupled anti-rabbit-IgG, both used at 1:500. Slides were counterstained with 1 mg/ml Hoechst 33258 (Sigma) to reveal DNA.

For immunofluorescence, HeLa cells were fixed in −20°C methanol for 7–10 min and washed in PBS with 0.05% Triton X-100 (PBST). After blocking in 1% bovine serum albumin (BSA) in PBST for 1 h, cells were incubated with primary antibodies overnight at 4°C. Following three washes in PBST for 5 min each, cells were incubated with secondary antibodies for 1 h at room temperature, counterstained with 1 μg/ml Hoechst 33342, washed three times for 5 min in PBST and mounted. Primary antibodies were 1:200 rabbit anti-NuMA (Santa Cruz Biotechnology, sc-48773), 1:200 rabbit anti-TPX2 (Santa Cruz Biotechnology, sc-32863), 1:200 mouse anti-p150^Glued (Transduction Laboratories, 612709) and 1:2000 mouse anti-γ-tubulin (GTU88, Sigma-Aldrich). Secondary antibodies were Alexa-Fluor-488-coupled anti-mouse-IgG (Life Technologies) and Alexa-Fluor-568-coupled anti-rabbit-IgG (Life Technologies), both used at 1:500.

Confocal images were acquired on a Zeiss LSM 710 confocal microscope equipped with a CCD Axiocam MRm camera (B/W) with a 63× NA 1.0 oil objective and processed in ImageJ and Adobe Photoshop, maintaining relative image intensities within a series.

The angle of the metaphase spindle with respect to the fibronectin substratum was determined as described previously (Toyoshima and Nishida, 2007). Briefly, cells were stained with γ-tubulin antibodies to mark spindle poles and counterstained with 1 μg/ml Hoechst 33342 (Sigma-Aldrich) to mark chromosomes. Stacks of confocal images 0.4 µm apart were acquired and the distance between the two spindle poles in z and in xy determined using Imaris (Bitplane Inc.); the spindle angle to the substratum was then calculated using inverse trigonometry.

Preparation of embryonic extracts, western blot analysis and immunoprecipitation (using ∼5 μg of PPH-6 antibodies and ∼3 μg of GFP antibodies) were performed as described (Afshar et al., 2004). For western blot analysis, anti-SAPS-1 antibodies were used at 1:1000 and GFP antibodies at 1:5000. The secondary antibodies, HRP-conjugated goat anti-rabbit-IgG or goat anti-mouse-IgG (Invitrogen) were used at 1:5000 and 1:1000, respectively, and the signal detected by performing ECL on the nitrocellulose blot followed by developing the X-ray film.

We thank Asako Sugimoto and Anthony Hyman for their gifts of reagents, Manfredo Quadroni from the Protein Analysis Facility, Center for Integrative Genomics, University of Lausanne, Switzerland, for protein identification, and George Hatzopoulos for help in generating Fig. S2A. We are also grateful to Alexandra Bezler, Sveta Chakrabarti and Virginie Hamel for their comments on the manuscript, as well as Marina Mapelli for discussion of unpublished work. We also thank the Bioimaging platform (BIOP) of the School of Life Sciences (EPFL) and the microscopy facility of the Department of Microbiology and Cell Biology (MCB) at Indian Institute of Science (IISc) for help in confocal image acquisition.