Spatiotemporal organization of Aurora-B by APC/C^Cdh1 after mitosis coordinates cell spreading through FHOD1

Cells preparing to divide undergo extensive rearrangement of the cytoskeleton that include loss of integrin-mediated adhesion, disassembly of actin structures and cell rounding (Kunda and Baum, 2009). As cells exit mitosis to return to interphase these processes are reversed. Very little is known about how such events are regulated, but in common with other steps of mitotic exit, they are assumed to require reversals of specific mitotic phosphorylation events that depend on upregulation of phosphatase activity and/or inactivation of mitotic kinases (Wurzenberger and Gerlich, 2011).

Ubiquitin-mediated proteolysis is one route to inactivation of mitotic kinases. The anaphase-promoting complex (APC/C) ubiquitin ligase targets a large number of substrates at mitotic exit (Pines, 2011) including mitotic kinases that play essential roles during mitotic exit, most notably Plk1 and AurB (Lindon and Pines, 2004; Stewart and Fang, 2005). Perturbing degradation of these kinases perturbs mitotic exit (Floyd et al., 2008; Lindon and Pines, 2004; Stewart and Fang, 2005), consistent with the idea that ubiquitin-mediated proteolysis can contribute to dynamic spatial redistribution of kinase activity through targeting of specific subpopulations (Lindon, 2008).

The multiple essential roles of AurB require repeated relocalisation of the kinase during mitosis (Carmena and Earnshaw, 2003). AurB is the enzymatic core of the CPC and depends on other members of the complex (notably INCENP) for localisation and activity. At anaphase CPC relocalises to the spindle midzone, with AurB playing critical roles in promoting assembly of the midzone and in signalling to the equatorial cortex to trigger cytokinesis (Douglas et al., 2010; Hümmer and Mayer, 2009). After anaphase the spindle midzone is compacted into the midbody matrix and AurB activity at the midbody regulates timing of abscission (Steigemann et al., 2009). AurB has specific target interactors and substrates during mitotic exit, including known regulators of cytokinesis and abscission (Bastos et al., 2012; Capalbo et al., 2012; Douglas et al., 2010; Neef et al., 2006) and others identified via proteomic studies whose function in mitotic exit is unknown (Ozlü et al., 2010). One such interactor is the diaphanous-related formin (DRF), formin homology 2 (FH2) domain-containing protein 1 (FHOD1). Formins nucleate actin filaments by stabilizing actin dimers through the conserved FH2 domain and promote their elongation by capping the barbed end of growing filaments. DRFs have also been found to regulate microtubule (MT) dynamics independent of their actin nucleation function (Bartolini et al., 2008; Cheng et al., 2011) and are increasingly recognized as mediators of interactions between MT and actin cytoskeletons (Bartolini and Gundersen, 2010; Chesarone et al., 2010).

In addition to specific roles for AurB in promoting mitotic exit, pathways attenuating AurB activity are also required for different steps in mitotic exit (Ramadan et al., 2007; Sumara et al., 2007; Vázquez-Novelle and Petronczki, 2010; Wurzenberger et al., 2012). However, the targeted destruction of AurB by the APC/C has not yet been shown to play any specific role in controlling AurB activity during mitotic exit. Here we have tested the idea that APC/C contributes to dynamic spatial organization of AurB activity as cells return to interphase. We uncover a role for Cdh1-dependent APC/C activity in controlling a subpopulation of AurB that is recruited to the cell cortex in a FHOD1-dependent manner in early G1 phase to play a role in promoting post-mitotic spreading via FHOD1.

Previous studies have indicated that, at the end of mitosis, AurB disappears from the cell later than other APC/C substrates (Floyd et al., 2008; Lindon and Pines, 2004; Sigl et al., 2009), and indeed in vivo time-lapse analysis of AurB–GFP degradation reveals Cdh1-dependent proteolysis of AurB continuing over a window of time that extends well into G1 phase (C.M., M.M. and C.L., unpublished data). We wanted to test the idea that ongoing AurB proteolysis contributes to the organization of mitotic exit. Therefore we examined the distribution of AurB at early G1 phase in synchronized, fixed populations of human HeLa, hTERT-RPE1 (RPE) and U2OS cells after brief treatment with the proteasome inhibitor MG132 or after siRNA-mediated silencing of Cdh1 expression (Fig. 1A–E; supplementary material Fig. S1 and data not shown). As expected we found most cellular AurB at the midbody, and in siRNA-treated (Cdh1-i) cells there was also some accumulation of AurB in the nucleus. In addition, we noticed in approximately half of MG132-treated or Cdh1-i cells a small population of AurB localised at the edge of the cell at sites distal to the midbody (Fig. 1A,B; supplementary material Fig. S1). We confirmed that other CPC components (INCENP, survivin) colocalised with AurB at these sites (supplementary material Fig. S1). In some cells these sites appeared to correspond to the cortical extremities of MTs (Fig. 1A,E). In other cells AurB colocalised with actin-rich structures (supplementary material Fig. S1), as previously reported during monopolar cytokinesis (Hu et al., 2008) or in cells overexpressing AurB–GFP (Abdullah et al., 2005), indicating that AurB might be able to interact with either MTs or F-actin at different times or under different conditions.

Next, we tested if this cortical pool of AurB contained active kinase, using a phospho-specific antibody raised against AurB phospho-T232 (pAur). The midbody stained strongly with pAur antibody in most control and Cdh1-i cells. In addition we found that the population of AurB at the cell cortex, but not that in the nucleus, also stained with the pAur antibody (Fig. 1C). We measured the intensity of staining with pAur and AurB antibodies at different locations in the cell to obtain an estimate of relative state of activity of AurB (Fig. 1D; supplementary material Fig. S1). We found that whereas chromatin-associated AurB did not stain with pAur, consistent with phosphatase-mediated inactivation of this pool (Murnion et al., 2001; Vagnarelli et al., 2011), relative AurB kinase activity at the cell edge in either control or Cdh1-i cells appeared almost as high as in the midbody. In conclusion, active AurB is present at the cell cortex in early G1 phase and is readily detectable under conditions where it is not degraded efficiently during mitotic exit.

We also found AurB at the cell cortex in a fraction of early G1 cells that had not been treated with Cdh1-i or MG132 (Fig. 1B). This early G1 window (during which daughter cells remain connected via a midbody-containing intercellular bridge) typically lasts an hour or more. Since even brief treatment with MG132 significantly increased the fraction of cells showing cortical AurB we hypothesized that AurB localisation to the cell edge was a transient event occurring in all early G1 phase cells. Indeed, approximate staging of our populations of fixed cells indicated that cortical AurB was systematically associated with ‘young’ G1 cells but absent from those with a more mature intercellular bridge (Fig. 1E). To test our hypothesis, we created a cell line expressing Venus-tagged AurB under tight tetracycline control, suitable for dynamic study of AurB localisation in living cells (Fig. 1F–H). AurB–Venus mimicked faithfully the enrichment of endogenous AurB to the cell edge seen in fixed populations of G1 phase cells (Fig. 1G–H). Moreover, a version of AurB–Venus lacking the N-terminal KEN motif (Nguyen et al., 2005) that is partially stabilized during mitotic exit (C.M., M.M. and C.L., unpublished data) was more frequently found in this localisation, confirming that proteolysis of AurB–Venus defines the fraction of cells showing cortical localisation of the kinase (Fig. 1H). In time-lapse analyses of living cells we found that AurB–Venus did indeed localise to the edge of the cell in a spatially and temporally restricted manner. Cortical enrichments of AurB–Venus appeared in almost every cell analysed (89.7%, n = 39), during a time window beginning 15±3 minutes after anaphase onset and lasting for 33±9 minutes (Fig. 1F; Fig. 2A; Table 1), and systematically associated with protrusions of the cell cortex that occur during post-mitotic spreading (supplementary material Movie 1). Therefore AurB recruitment to the cell edge is an event that occurs during a specific window of time during mitotic exit.

To define parameters involved in spatiotemporal regulation of AurB in early G1 we examined the dynamics of AurB–Venus localisation when proteolysis was inhibited, by adding MG132 to cells during filming. We found that the timing of appearance of AurB–Venus at the cell cortex was unchanged, but that the kinase accumulated all along the spreading edge of early G1 cells (Fig. 2B; supplementary material Movie 2) and persisted there, frequently until the end of filming (≧87±21 minutes in MG132, Fig. 2A; Table 1). We concluded that removal of the kinase from the cell cortex is proteolysis-dependent. Next we examined the role of cytoskeletal components in localising AurB to the cell cortex. Treatment of cells with the MT depolymerizing drug nocodazole shortly after anaphase onset prevented AurB–Venus localisation to the cell cortex (Fig. 2A,B; supplementary material Movie 3). Consistently, treatment with the MT stabilizing agent taxol induced strong accumulation of AurB–Venus on cortical MT bundles (Fig. 2A,B). In contrast, disruption of F-actin by treatment with cytochalasin D increased the persistence of AurB–Venus spots visible near the cell edge (Fig. 2A,B). Inhibition of Aurora B kinase activity by addition of the inhibitor ZM447439 (ZM) (Ditchfield et al., 2003) reduced recruitment of AurB–Venus to the cell cortex (Fig. 2A), which could explain why AurB is not seen at the cortex in MG132-treated interphase cells, when it is not fully active. Together we interpreted these data to mean that active AurB travels to the early G1 cell cortex on MTs but is then removed from them in a manner that depends on F-actin, and retained at the cortex in a manner limited by ubiquitin-mediated proteolysis.

We investigated the role of the newly characterized cortical pool of AurB in post-mitotic spreading of daughter cells, a process that requires coordination of MT and actin networks in early G1 phase. We used RPE cells for this study, since these spread rapidly after division, showing marked polarization in early G1 phase. For time-lapse observation of actin organization in living cells we used Venus-tagged α-actinin to label the actin cytoskeleton, allowing visualization of stress fibres (SFs), focal adhesion complexes and other actin-rich structures (Edlund et al., 2001). We generated a stable cell line, RPE-α-actinin–Venus, in which daughter cells show strongly polarized spreading after division, such that protrusive activity occurs at the furthest point from the site of division (supplementary material Movie 4). This protrusive activity usually establishes a single leading edge per cell at a site distal to the intercellular bridge. Daughter cells then migrate away from each other, typically with a persistent directionality determined by the position of cleavage (Albrecht-Buehler, 1977) that maximizes the distance between them during early G1 (supplementary material Fig. S2). SFs become visible, either as dorsal fibres behind the leading edge or as lateral peripheral fibres, ∼45 minutes after anaphase onset (Fig. 2C,D).

We treated RPE-α-actinin–Venus cells with MG132 shortly after anaphase onset and found that proteolysis inhibition during mitotic exit prevented polarized cell spreading and significantly delayed re-assembly of SFs at the end of mitosis (Fig. 2C,D). At least some of this effect was the result of stabilizing substrates of APC/C^Cdh1 because we saw a similar effect in Cdh1-i cells. Cdh1-i cells displayed continuous ruffling and short-lived protrusions all around the edge of daughter cells, were slow to adopt a polarized shape (Fig. 2C,E; supplementary material Movie 5), failed to establish a clear leading edge and showed reduced directionality in migration (supplementary material Fig. S2). The timing of α-actinin–Venus recruitment to sites of cortical activity and in focal adhesions appeared normal (17±5 minutes and 18±5 minutes respectively, n = 16), while the appearance of SFs was delayed (63.3±23 versus 44.9±11 for control cells, P<0.01 by Student's t-test; Fig. 2D). In particular we observed in Cdh1-i cells an absence of SFs along the cell periphery. We concluded that proteolysis of one or more Cdh1 targets contribute to reassembly of SFs and to polarized cell shape at the start of interphase.

We then tested if one of these targets was AurB, by treating control and Cdh1-i cells with ZM shortly after anaphase entry to test if we could accelerate assembly of SFs through inhibition of AurB. We found that ZM treatment did significantly accelerate the appearance of SFs in control cells (Fig. 2C,D). In Cdh1-i cells, however, this effect was less consistent and moreover loss of cell shape was only partially rescued by AurB inhibition (Fig. 2C–E). Instead, ZM treatment produced immediate pronounced ruffling and extension of existing lamellipodia-like protrusions (Fig. 2C, supplementary material Movie 6). It is possible that ZM is added too late in these experiments to influence cell polarity (it cannot be added earlier than late telophase without causing cytokinesis to fail) and also likely that the effects of Cdh1-i depend on additional targets, such as Rho regulators p190rhoGAP and Ect2 (Liot et al., 2011; Naoe et al., 2010). We also reasoned that cell-wide inhibition of AurB might not be able to rescue effects on cell spreading that depend on spatial organization of AurB activity.

To understand more about the spatial organization of AurB in mitotic exit we tested known AurB interactors that might contribute to its distinctive localisation during mitotic exit. These included the C-(cytokinesis)-phase-specific AurB interactor FHOD1 (Ozlü et al., 2010), which seemed a particularly promising candidate for an interactor defining a distinct pool of AurB, since in our experiments – unlike other documented interactors – it appeared to be absent from the midbody. Moreover, FHOD1 had previously been implicated in driving cell elongation in human tissue culture cells (Gasteier et al., 2005).

The available antibody against FHOD1 did not recognize endogenous FHOD1 by immunofluorescence but we found some colocalisation of HA-tagged FHOD1 with AurB–Venus and pAurB at sites at the cell periphery in early G1 phase (Fig. 3A). After siRNA-mediated knockdown of FHOD1 expression (FHOD1-i), we found that AurB–Venus localisation to protrusions of spreading cells was either abolished (52/90 cells from three separate experiments) or attenuated (≤14 minutes in 16/90 cells; Fig. 3B; supplementary material Movies 7, 8). Cortical localisation of AurB was partially rescued by co-transfection of siRNA-resistant RFP-FHOD1 (supplementary material Fig. S3). We also found that localisation of endogenous AurB to the cell periphery in early G1 phase was abolished in FHOD1-i, and rescued by siRNA-resistant FHOD1 (supplementary material Fig. S3). We concluded that FHOD1 is required to maintain a pool of AurB at the cell cortex in early G1 phase.

We next examined whether FHOD1 was required for cell spreading in RPE-α-actinin–Venus cells and found pronounced effects on cell shape in FHOD1-i cells after mitosis. Cells spread rapidly but did not do so in correctly polarized fashion, displaying multiple sites of sustained cortical ruffling in early G1 phase (Fig. 3C; supplementary material Movie 9) and, as in Cdh1-i cells, failure to develop a leading edge or to adopt the elongated shape typical of RPE cells (Fig. 3D). We also examined the distribution of F-actin in paraformaldehyde (PFA)-fixed populations of U2OS cells by phalloidin staining, observing in FHOD1-i-treated cells an absence of the robust peripheral actin cables characteristic of control post-mitotic cells (Fig. 3E, see also Fig. 4D). The absence of these cortical actin cables was directly related to loss of cell shape (Fig. 3F). Since cell shape and polarity are generally defined by cues delivered by MTs, we examined the MT network in methanol-fixed populations of U2OS cells and found that the density and organization of the MT network in early G1 cells was reduced after FHOD1-i (Fig. 3G). We noted the absence of MTs oriented along the long axis of the cell and that are required to drive cell elongation (Picone et al., 2010). This population of MTs could be rescued by co-expression of siRNA-resistant HA–FHOD1wt, along with rescue of cell shape (Fig. 3G,H). A version of HA–FHOD1 lacking the FH2 domain (shown to mediate MT stabilization in related formin mDia2) (Bartolini et al., 2008) was not able to rescue (Fig. 3G,H). We then stained U2OS and RPE cells with MT plus-end (+TIP) marker EB3, to examine the presence of +TIPs at the cell cortex. We found that the number of EB3 comets close to the cell cortex was reduced in FHOD1-i (Fig. 3I; supplementary material Fig. S3), consistent with a role for FHOD1 in promoting MT growth to the cell cortex in early G1 cells. In the same experiment we found the density of EB3 comets close to the cortex was increased after treatment of cells with ZM, indicating that AurB activity at this time would generally act to suppress MT growth at the cortex. AurB is known to selectively destabilize MTs at incorrectly attached kinetochores, earlier in mitosis, through recruitment of the kinesin-13 MCAK (Cimini et al., 2006; Knowlton et al., 2006) and MCAK localised to +TIPs close to the cell cortex in early G1 cells (data not shown). We thus propose that AurB and FHOD1 contribute respectively to the selective destabilization and stabilization of MTs required to establish a polarised network in early G1 phase.

Finally, we investigated whether FHOD1 is a downstream target of AurB in regulating cell spreading. AurB-dependent phosphorylation (RphosSL, AurB consensus sequence) of FHOD1 had previously been reported (Ozlü et al., 2010). To identify AurB phosphorylation sites in FHOD1 we subjected in vitro-phosphorylated full-length FHOD1 to mass spectrometry (supplementary material Fig. S4). Multiple AurB phosphorylation targets were identified in FHOD1 including Ser367 as initially reported (Ozlü et al., 2010) and an additional cluster of four phospho-Ser/Thr residues (between 486 and 498) within the coiled-coil domain that lies just upstream of the conserved FH1 and FH2 domains (Fig. 4A). Since a recent proteomic study confirms that this phosphorylation cluster is present on FHOD1 in mitotic cells (Kettenbach et al., 2011) and these four sites together with Ser367 comprise the large majority of all AurB phospho-sites in FHOD1 (supplementary material Fig. S4), we mutated these five residues to generate phospho-mimetic (FHOD1-5D) or non-phosphorylatable (FHOD1-5A) versions of FHOD1. Expression and immuno-isolation of HA-tagged versions of FHOD1 from human cells confirmed that these five sites (S367, S486, T489, T495, S498) contribute significantly to phosphorylation of FHOD1 by AurB in vitro (Fig. 4B,C). To investigate the role of FHOD1 phosphorylation by AurB in cell spreading we carried out siRNA-rescue experiments using the 5D and 5A versions of HA–FHOD1. We used PFA-fixed populations of cells, to examine together actin and MT cytoskeletons and HA–FHOD1 localisation, and MEOH fixation to examine AurB localisation at the cell cortex. We found that AurB recruitment to the cortex was not affected by our phospho-site mutations in FHOD1 (supplementary material Fig. S3). However, we found clear differences in organization of the peripheral cytoskeleton between cells rescued with phospho-mimetic and non-phosphorylatable versions of HA–FHOD1 (Fig. 4D–G). Examination of cells rescued with wild-type HA–FHOD1 revealed, as expected, assembly of peripheral actin SFs in early G1 phase, closely aligned with MTs oriented along the cortex (Fig. 4D left-hand panels, compare HA-positive cell with adjacent non-rescued cell in early G1 phase). We also noted that HA–FHOD1wt itself showed colocalisation with peripheral actin SFs/MTs, and additionally localised to the tips of some MTs at the leading edges of the cell (Fig. 4D,F). The phospho-mimetic HA–FHOD1-5D was able to rescue the assembly of robust peripheral SFs in early G1 phase and showed similar localisation to FHOD1wt, predominantly at sites of colocalisation of MTs with SFs (Fig. 4D–F). HA–FHOD1-5A was less efficient in rescue of SFs and showed a different pattern of localisation. Peripheral SFs were significantly weaker, and polymerized actin was frequently seen extending beyond the MT bundles that usually define the cell edge (Fig. 4D,E). HA–FHOD1-5A, like HA–FHOD1ΔFH2, also showed a reduced tendency to localise with MTs but was observed to colocalise with F-actin along the cell boundaries (Fig. 4D,F). This suggests that AurB-mediated phosphorylation of FHOD1 is required to coordinate the assembly of F-actin fibres with stable oriented MTs at the cell periphery in early G1 phase.

We identified a pool of AurB recruited at the cell cortex in early G1 phase that, in contrast to other pools of AurB that undergo phosphatase-mediated inactivation (Steigemann et al., 2009; Wurzenberger et al., 2012) relies exclusively on APC/C-mediated proteolysis for its inactivation at mitotic exit. APC/C^Cdh1 is required to eliminate most of this pool of AurB, restricting it to protrusions associated with cell spreading and thereby exerting spatial control on the activity of the kinase. We have observed that the duration and size of this pool are sensitive to the surface on which AurB–Venus cells are cultured (N.W. and C.L., unpublished observations) leading us to speculate that AurB could modulate responses to external cues. We are currently working to establish an in vivo degradation assay for AurB that will allow us to test whether the rate or timing of AurB proteolysis by APC/C^Cdh1 is influenced by parameters affecting cell adhesion after division.

We further identified the formin FHOD1, a known C-phase-specific binding partner (Ozlü et al., 2010), as an essential factor in retaining AurB at the cell cortex. FHOD1 has previously been shown to be an effector of Rac1, with expression of a constitutively active version shown to induce SF assembly, to promote alignment of MTs with SFs, and to enhance cell elongation and migration (Gasteier et al., 2003; Gasteier et al., 2005; Koka et al., 2003). Consistent with this we find phenotypes in FHOD1-knockdown daughter cells that include loss of cell shape and misaligned MTs and SFs, suggesting that FHOD1 plays a specific role in the dynamic interplay between MT and actin cytoskeletons that is required to remodel the shape and polarity of newly divided cells. Our work demonstrates that spatial control of AurB kinase activity by APC/C^Cdh1 is critical to this remodelling through regulation of targets that include FHOD1. Absence of AurB from the cortex (in FHOD1-i), or else over-recruitment of AurB at the cortex (in Cdh1-i) both lead to loss of cell shape. Therefore we propose that correct cell shape is achieved through stabilization of selected MTs in a manner depending on tightly regulated localisation of AurB. In this scenario, we envision that AurB triggers FHOD1 activity by at least two distinct mechanisms. First, FHOD1 subcellular localisation and activity are regulated by Rac1 (Gasteier et al., 2003; Koka et al., 2003; Westendorf, 2001) and since AurB is an upstream regulator of Rac1 (Bastos et al., 2012; Minoshima et al., 2003), the kinase may indirectly activate FHOD1 via Rac1. Second, we find that a cluster of phosphorylation sites upstream of the FH1 and FH2 domains is phosphorylated by AurB in vitro, and that a non-phosphorylatable version of this cluster interferes with the function of FHOD1 in coordinating SF assembly with oriented MTs. We propose that, as in regulation of FHOD1 by ROCK (Hannemann et al., 2008; Takeya et al., 2008), phosphorylation of the DRF by AurB may directly induce conformational changes and/or allow for additional protein interactions. Given the position of AurB phosphorylation sites in FHOD1 near the coiled-coil domain that governs multimerization of the DRF (Madrid et al., 2005), these modifications are likely to affect multimerization rather then autoinhibition of FHOD1. Based on preliminary results indicating decreased multimerization of FHOD1-5D (data not shown), it is tempting to speculate that AurB mediated phosphorylation results in monomeric FHOD1 that is particularly prone to stabilize MTs while FHOD1 without AurB phosphorylation persists as multimer and exclusively acts on actin organization. Further work is required to test this hypothesis and determine the precise properties of FHOD1 regulated by AurB in MT stabilization or actin nucleation.

In our model, active FHOD1 would stabilize a selected population of MTs in FH2 domain-dependent manner, and direct oriented SF assembly at the cell periphery, activities promoted by AurB delivered to the cortex on FHOD1-stabilized MTs (Fig. 5). Our finding that global inhibition of AurB by the kinase inhibitor ZM promoted the appearance of MT plus-ends close to the cell edge suggests, however, that AurB would play a role in destabilizing other MTs at this time in the cell cycle. We reconcile these findings by proposing that alternative binding partners of AurB control distinct populations of MTs. In this model AurB delivered at the cell cortex would either contribute to MT destabilization or, through recruitment by FHOD, to stabilization of selected MTs. APC/C^Cdh1 would act to restrict the pool of AurB activity close to or at the cell cortex to promote polarized cell shape (Fig. 5).

A central role of AurB activity in polarizing events involving both MTs and actin has previously been described during monopolar cytokinesis (Hu et al., 2008), events that we have also found to depend upon FHOD1 (C.M., M.M. and C.L., unpublished data). Coordination of MT and F-actin networks may thus emerge as a central biological property of FHOD1. Understanding further how these components interact with each other both physically and functionally will be critical to understanding how daughter cell identity is established at the beginning of interphase.

pTRE-AurB-Venus was made from Aurora B-GFP (kind gift of YL Wang). pEGFP-N1-α-actinin1 (Edlund et al., 2001) was obtained from Addgene and GFP swapped for Venus fluorescent protein to make pVenus-N1-α-actinin. Expression plasmids for HA-FHOD1, HA-FHOD1ΔFH2 were previously described (Gasteier et al., 2003). Expression plasmids for FHOD1-RFP, siRNA-resistant FHOD1 versions, and FHOD1-5D and -5A point mutants were made using standard cloning and site directed mutagenesis techniques and verified by sequencing. Full cloning details for all constructs available on request.

HeLa cells were cultured in high glucose DMEM (Invitrogen, Carlsbad, CA, USA). hTERT-RPE1 (RPE) cells were cultured in 50∶50 mix of Ham's F12:DMEM (Sigma-Aldrich, St. Louis, MO, USA). U2OS cells were cultured in 50∶50 mix of Leibovitz L15:DMEM. All cell culture media were supplemented with FBS (10%), penicillin–streptomycin and amphotericin B (all from PAA Laboratories, Pasching, Austria). U2OS tet-OFF cell culture medium was additionally supplemented with 500 µg/ml geneticin (PAA) and 1 µg/ml tetracycline hydrochloride (Calbiochem, San Diego, CA, USA).

RPE-α-actinin–Venus cells were made by clonal selection with 500 µg/ml geneticin of cell populations transfected with pVenus-N1-α-actinin-Venus. U2OS-AurB–Venus cells were obtained by clonal selection with 170 µg/ml hygromycin B (Calbiochem) after transfection of U2OS tet-OFF cells with pTRE-AurB-Venus∶pCMVhygro in a ratio of 10∶1.

HeLa and U2OS cells were synchronized in early G1 phase by release for 13 hours from thymidine/aphidicolin double block. RPE cells were synchronized in early G1 phase by 16 hour release from single thymidine block.

siRNA duplexes targeting sequences in hsCdh1 and hsFHOD1 as previously described (Floyd et al., 2008; Hannemann et al., 2008), and control siRNA duplex against GL2, were obtained from Ambion (Austin, TX, USA).

Transfections were carried out using Lipofectamine® 2000 or Oligofectamine® (Invitrogen). siRNA-rescue experiments were achieved by electroporation using Invitrogen Neon® system to electroporate cells with mixtures of siRNA and siRNA-resistant plasmid, according to the manufacturers' instructions. Cells were analysed 44–48 hours following transfection/electroporation.

Cells were grown on 12 mm glass coverslips, synchronized and fixed using either 100% methanol at −20°C or 4% PFA treatment at ambient temperature. Full details of buffers, reagents and antibodies used are given in supplementary material Table S1. Widefield images were captured on an Olympus CellR platform (Fig. 1C; Fig. 3A; supplementary material Fig. S1; see description in next section) or an inverted Zeiss Axiovert 200M microscope with 63× 1.3 NA oil objective and CoolSNAP HQ camera (Photometrics, Tucson AZ, USA), running Metamorph® software (Molecular Devices, Sunnyvale, CA, USA; Fig. 1A,E; Fig. 3E,G; Fig. 4D). Confocal images were captured using a TCS SP5 Confocal Leica laser scanning microscope equipped with a DMI60000 microscope and a HCX PL APO Lambda blue 63× 1.4 NA oil objective (Fig. 3I; supplementary material Fig. S3). Intensity level, contrast and brightness of images were adjusted using Adobe Photoshop where indicated.

Pixel quantifications (Fig. 1G; Fig. 3F; Fig. 4E) were carried out in ImageJ, using 16-bit images acquired on widefield microscopes. Circularity calculations (Fig. 3H; Fig. 4G) were carried out in ImageJ, using the equation: circularity = 4π(area/perimeter²).

Cells were seeded onto 8-well plastic-bottom slides (Ibidi GmbH, Martinsried, Germany) for time-lapse analyses. For both AurB–Venus and α-actinin–Venus experiments cells were seeded at 8×10⁴ cm⁻². Imaging medium was L-15 supplemented with FBS and antibiotics as described above. Expression of AurB–Venus was induced 24 hours prior to filming, and cells chosen for time-lapse recording according to the presence of a Venus-positive metaphase plate. Dishes were used up to 30 hours after induction, after which AurB–Venus levels were deemed too high (and mitotic aberrations were frequently observed). All time-lapse imaging was carried out on an Olympus CellR imaging platform comprised of Olympus IX81 motorized inverted microscope, Orca CCD camera (Hamamatsu Photonics, Japan), motorized stage (Prior Scientific, Cambridge, UK) and 37°C incubation chamber (Solent Scientific, Segensworth, UK). Cells were imaged by epifluorescence and DIC microscopy, using appropriate filter sets to record AurB–Venus localisation and α-actinin distribution, using 40× 1.3 NA oil and 60× 1.42 NA oil objectives, respectively. Movies were recorded using Olympus CellR software, as 1 µm stacks of up to 12 fields per experiment, acquired at 2 minute intervals and with 2×2 bin to minimize bleaching. All movies and movie frames are displayed as maximal projections of stacks. Image sequences were analysed using CellR software or else exported as 12-bit tiff files for analysis in ImageJ.

Cells were subjected to different drug treatments during time-lapse experiments, by addition of drug in a volume not less than 1/100 of the existing dish volume and gentle mixing of medium on the dish using a 1 ml Gilson pipette. Drugs were used at the following final concentrations: MG132 (Tocris Bioscience, Bristol, UK) 4 µM; ZM447439 (Tocris Bioscience) 10 µM; nocodazole (Sigma-Aldrich) 2 µM; taxol (Tocris Bioscience) 5 µM; cytochalasin D (Sigma-Aldrich) 1 µM.

To assay for phosphorylation by AurB, wt or mutant HA-FHOD1 variants were expressed in HEK293T cells. Following lysis in ROCK lysis buffer [10 mM Tris-HCl, pH 7.5; 40 mM Na₄P₂O₇•10H₂O; 5 mM EDTA; 150 mM NaCl; 1% (v/v) NP-40; 0.5% (v/v) sodium deoxycholate; 0.025% (v/v) SDS; 1∶1000 protease inhibitor cocktail (Roche)], FHOD1 proteins were immuno-isolated with anti-HA antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA) coupled to protein-G–Sepharose (GE Healthcare, Little Chalfont, UK). After intensive washing in lysis buffer, immunoprecipitates were subjected to kinase reactions in 30 µl kinase assay buffer (5 mM MOPS, pH 7.2; 2.5 mM β-glycerophosphate; 1 mM EGTA; 0.4 mM EDTA; 5 mM MgCl₂; 0.5 mM DTT; 1 mM ATP), supplemented with 4 µCi [γ-³²P]ATP and with 1 µl recombinant active GST-AurB (Sigma Aldrich) and incubated at 30°C for 90 minutes. The kinase reaction was inhibited with specific AurB inhibitor ZM447439 or Staurosporine (Sigma Aldrich) where indicated.

For mass spectrometry, the assay was conducted with wild-type full length recombinant FHOD1 [kind gift from Matthias Geyer (Hannemann et al., 2008)] with non-radioactive ATP and 30 minutes incubation at 30°C. Samples were run on an SDS-polyacrylamide gel and stained with Coomassie Brilliant Blue G-250.

For sample preparation, protein bands were manually excised from the gel and tryptic in-gel-digest performed based on the protocol from Rosenfeld (http://www.ncbi.nlm.nih.gov/pubmed/1524213) using modified porcine trypsin, sequencing grade, from Promega (Madison, WI, USA). The collected peptide samples were dried and dissolved in 3% acetonitrile and 0.1% formic acid.

For LC-MS/MS, peptides were separated using a nanoAcquity ultra performance liquid chromatography (UPLC) system (Waters Corporation, Milford, MA, USA) fitted with a trapping (nanoAcquity Symmetry C18.5 µm, 180 µm×20 mm) and an analytical column (nanoAcquity BEH C18, 1.7 µm, 75 µm×200 mm). The outlet of the analytical column was coupled directly to a linear trap quadrupole (LTQ) Orbitrap Velos (Thermo Fisher Scientific, Waltham, MA, USA). Peptides were eluted through the analytical column at a constant flow of 0.3 µl/minute and total runtime 45 minutes. Full scan mass spectrometry spectra with mass range 200–2.000 mass-to-charge ratio (m/z) were acquired in profile mode in the Orbitrap with resolution of 30.000. The most intense ions (up to 15) from the full-scan mass spectrometry were selected for fragmentation in the LTQ. Neutral loss masses for phosphorylations were screened and selected for further fragmentation if present and the multistage activation was enabled.

The raw data were processed using MaxQuant (version 1.0.13.13) (Cox and Mann, 2008), and MS/MS spectra were searched using the Mascot search engine (version 2.2.07, Matrix Science) against a Swissprot database (download 11.07.2012) restricted to human entries, with a maximum of one missed cleavage allowed. Carbamidomethylation was set as fixed modification while methionine oxidation and phosphorylation on S and T were included as variable modifications. The search was performed with an initial mass tolerance of 5 ppm for the precursor ion and 0.5 Da for the MS/MS spectra and a minimum ion score of 20 was applied.

We thank Anna Akhmanova for EB3 antibody, Yu-Li Wang for plasmids, Jeroen Krijgsveld and Stefan Leicht (Proteomics Core Facility, EMBL Heidelberg) for MS analysis and Rosa M. Rios for supporting the contribution of M.P.G.