The chromosomal passenger complex controls the function of endosomal sorting complex required for transport-III Snf7 proteins during cytokinesis

Summary Cytokinesis controls the proper segregation of nuclear and cytoplasmic materials at the end of cell division. The chromosomal passenger complex (CPC) has been proposed to monitor the final separation of the two daughter cells at the end of cytokinesis in order to prevent cell abscission in the presence of DNA at the cleavage site, but the precise molecular basis for this is unclear. Recent studies indicate that abscission could be mediated by the assembly of filaments comprising components of the endosomal sorting complex required for transport-III (ESCRT-III). Here, we show that the CPC subunit Borealin interacts directly with the Snf7 components of ESCRT-III in both Drosophila and human cells. Moreover, we find that the CPC's catalytic subunit, Aurora B kinase, phosphorylates one of the three human Snf7 paralogues—CHMP4C—in its C-terminal tail, a region known to regulate its ability to form polymers and associate with membranes. Phosphorylation at these sites appears essential for CHMP4C function because their mutation leads to cytokinesis defects. We propose that CPC controls abscission timing through inhibition of ESCRT-III Snf7 polymerization and membrane association using two concurrent mechanisms: interaction of its Borealin component with Snf7 proteins and phosphorylation of CHMP4C by Aurora B.


Introduction
Cytokinesis, the final separation of two daughter cells at the end of cell division, requires the coordinated action of several proteins that promote the formation and ingression of a cleavage furrow that bisects the dividing cell. Many key cytokinesis proteins localize to an array of anti-parallel and interdigitating microtubules known as the central spindle, which forms between the segregating anaphase chromosomes during furrow ingression [1]. The central spindle and its associated proteins control furrow formation and ingression, the transport of membrane vesicles to the cleavage site and the final abscission of the two daughter cells [2][3][4]. In the final stages of cytokinesis, the central spindle forms a compacted structure known as the midbody, which contains at its centre a phase-dense structure-the midbody ring or Flemming body-important for cell abscission. Our knowledge of the mechanisms that control furrow formation and ingression has advanced considerably in the last few years [5]. However, the mechanics and regulation of abscission are still unclear. Recent studies indicate that the endosomal sorting complexes required for transport (ESCRTs) play a key role during cell abscission [6][7][8][9][10][11]. These evolutionarily conserved complexes are best known for catalysing the sorting of receptors at endosomal membranes into vesicles that bud off into the lumen of the endosome, creating the multivesicular bodies (MVBs) [12]. The cargo of MVBs is then routed into lysosomes for degradation. Four distinct ESCRTsknown as ESCRT-0, -I, -II and -III-are sequentially recruited to endosomes through both protein and lipid interactions [12]. ESCRT-III, the final complex in the pathway, provides the core machinery that mediates membrane deformation and fission events during MVB biogenesis [13]. This complex comprises 11 subunits in humans (also known as charged multivesicular body proteins, CHMPs), which correspond to six counterpart proteins in yeast and seven in Drosophila (table 1) [14]. ESCRT proteins have also been implicated in other membrane-budding processes, including retroviral budding and, importantly, cytokinesis that are topologically similar to the budding required for MVB formation [12,14,15]. ESCRT-I and -III components have been found to localize to the midbody, and ESCRT-III components have been proposed to mediate membrane fission at the end of cytokinesis [6][7][8][9][10][11].
The spatial and temporal signals that control abscission in metazoans are poorly understood. One signalling component that has been implicated in these late stages of cytokinesis is the chromosomal passenger complex (CPC). This evolutionarily conserved complex comprises four components: the inner centromeric protein (INCENP), the Aurora B serine/threonine kinase, Survivin and Borealin [16,17]. CPC is essential for multiple processes during cell division, including condensation of chromosomes, their alignment and proper attachment to spindle microtubules, and cytokinesis [17]. Consistent with these functions, the CPC follows a dynamic distribution during mitosis: it localizes to centromeres at mitotic entry, translocates to the central spindle and cleavage furrow after anaphase onset and concentrates at the midbody in late telophase [16]. Studies in many organisms have revealed that CPC components are crucial for completion of cytokinesis [17] and evidence in both yeast and mammals have indicated that Aurora B activity is necessary to prevent cell abscission in the presence of chromosome bridges at the cleavage site, functioning as a 'NoCut' checkpoint [18 -20].
Here, we show that Borealin interacts directly with Snf7/ Shrb/CHMP4 components in both Drosophila and human cells and that the two proteins colocalize at the midbody in late cytokinesis. Moreover, we found that Aurora B phosphorylates CHMP4C at three serine residues located in its C-terminal linker region, a part of the protein known to regulate its ability to form polymers and interact with the membrane. Finally, over-expression of CHMP4C variants mutated in these three residues caused cytokinesis failure, suggesting that Aurora B inhibits CHMP4C activity during cytokinesis. We propose that CPC controls abscission timing in both flies and human cells by regulating the function of ESCRT-III Snf7 proteins during cytokinesis through the interaction of its Borealin component with the N-terminus of Shrb/CHMP4 proteins and Aurora B-mediated phosphorylation of the CHMP4C regulatory linker tail.

Borealin-related protein interacts with the ESCRT-III Shrb component in Drosophila cells
In a proteomic survey of complexes involved in cell division in Drosophila, we tagged the CPC Borealin-related (Borr) component with two IgG-binding domains of Protein A (PtA) at either its N-or C-terminus. The presence of a tag at either end of Borr did not affect the localization of the protein (see electronic supplementary material, figure S1, and data not shown). Independent Drosophila cell lines stably expressing PtA-tagged Borr proteins were generated, and interacting partners isolated by affinity purification and identified by mass spectrometry (MS). In both purifications, we identified the ESCRT-III component Shrb with a score comparable to or even higher than INCENP (table 2). A reciprocal affinity purification using cells expressing PtA::Shrb also identified all the CPC components (table 2), confirming the association in vivo. Because Borr and INCENP interact directly, these MS data suggested that Borr could also bind to Shrb directly.  To confirm this, we used an in vitro glutathione S-transferase (GST) pull-down assay (figure 1a). We generated several GST-tagged Borr fragments in bacteria and tested their ability to pull down a radio-labelled Shrb polypeptide synthesized by in vitro translation (figure 1a). Full-length Borr 1 -315 and the truncated Borr 1 -249 fragment pulled down Shrb very efficiently, whereas Borr 1 -117 did not (figure 1a). Borr 118 -315 could also efficiently pull down Shrb, whereas both Borr 118 -249 and the most C-terminal Borr 250 -315 fragment did not bind to Shrb (figure 1a). We conclude that the Borr 118 -249 central region is necessary, but not sufficient, for the interaction with Shrb. No direct interaction was observed with two other components of the CPC, INCENP (data not shown) and Aurora B (figure 1b), using a similar assay. To assess if and when CPC colocalized with Shrb during mitosis, we generated a stable cell line stably expressing Shrb tagged with green fluorescent protein (GFP). In late telophase, both Shrb::GFP and Aurora B accumulated at the midbody where the two proteins partially colocalized (figure 1c). However, Shrb::GFP appeared to localize to the midbody ring, whereas Aurora B was more broadly distributed. Finally, during abscission, only Shrb was visible at the midbody (figure 1c). Altogether, these results indicate that Shrb and Borealin interact in late cytokinesis just before abscission.

Borealin interacts with ESCRT-III CHMP4 proteins in human cells
We then investigated whether the interaction between Borealin and ESCRT-III Snf7 components had been conserved in human cells. There are three Snf7 paralogues in humans: CHMP4A, CHMP4B and CHMP4C (table 1). The primary sequence of the three CHMP4 proteins and their Drosophila homologue Shrb is very well conserved, although all four proteins diverge considerably in their C-terminal regions (figure 2). To assess whether Borealin could interact with CHMP4 proteins in vivo, we transfected HeLa cells with either PtA-tagged Borealin or PtA alone as control, and each of the three CHMP4 proteins tagged with GFP. PtA::Borealin pulled down GFP::CHMP4A and GFP::CHMP4B very efficiently, but not GFP::CHMP4C or GFP alone (figure 3a).
To investigate whether these interactions were direct, we carried out an in vitro GST pull-down assay as described earlier for Borr and Shrb. GST-tagged Borealin purified from bacteria could efficiently pull down all three CHMP4 proteins synthesized by in vitro translation (figure 3b). Our inability to detect an interaction between Borealin and CHMP4C in vivo, in contrast with the positive outcome of the in vitro assay, suggests that the interaction might be regulated by post-translation modifications and/or may occur only in a specific phase of the cell cycle and therefore, not be easily detectable in our unsynchronized cell populations.
To identify the Borealin domain responsible for the interaction with CHMP4 proteins, we tested the ability of different GST -Borealin fragments to pull down one of the three CHMP4 paralogues. Although all three CHMP4 proteins could strongly bind to Borealin in vitro, CHMP4C appeared to interact slightly more efficiently (figure 3b) and thus we decided to use this paralogue for the next pull-down assays. Borealin contains an N-terminal region (amino acid 1-109) that interacts with INCENP, and a C-terminal rsob.royalsocietypublishing.org Open Biol 2: 120070 region (amino acid 208-280) required for dimerization (figure 3c) [21,22]. We found that the Borealin central region encompassing residues 110 -207 was both necessary and sufficient to bind to CHMP4C (figure 3c). However, the interaction was lost if the fragment also contained the dimerization domain (figure 3c). We conclude that Borealin interacts with CHMP4 proteins via its central region, similar to its Drosophila counterpart Borr, but for reasons that are not clear the dimerization domain seems to interfere with the binding ability of the isolated central region but not of the full-length protein (figure 3c). To pinpoint the CHMP4C domain able to bind to Borealin, we tested whether the N-and C-terminal halves of the protein were able to pull down radio-labelled Borealin synthesized by in vitro translation (figure 3d). These two regions are quite different in their structure and charge. The N-terminal half (amino acid 1-121) is basic and contains two alpha-helices (a12), whereas the C-terminal half (amino acid 122 -233) is acidic and contains three alpha-helices (a345; figure 3d). We found that GST::CHMP4C 1 -121 could bind to Borealin very efficiently, even better than the fulllength protein, whereas GST::CHMP4C 122 -233 could only weakly pull down Borealin to a level similar to that of GST alone (figure 3d). Together, these results indicate that the central region of Borealin interacts with the N-terminal basic region of CHMP4C.
We then asked whether Borealin colocalized with CHMP4 proteins during cytokinesis in HeLa cells. We found that all three GFP-tagged CHMP4 proteins accumulated at the midbody in late cytokinesis (figure 4), in agreement with published observations [9,10]. However, both GFP::CHMP4A and GFP::CHMP4B were also found in the nucleus, whereas GFP::CHMP4C was predominantly cytoplasmic (figure 4a,b). Both GFP::CHMP4B and GFP::CHMP4C partially colocalized with Borealin during late cytokinesis (figure 4b) but during abscission only CHMP4 proteins were found at the midbody (figure 4b), a localization pattern similar to that observed in Drosophila cells (figure 1c). To analyse the localization of CHMP4 proteins in more detail, we stained midbodies purified from HeLa cells with antibodies directed against CHMP4B and either tubulin or Borealin (figure 4c). These experiments confirmed that CHMP4B accumulated at the midbody and also showed that the ESCRT-III component localized as discrete dots around the Flemming body to form a structure resembling a pearl necklace (figure 4c). Some of the CHMP4B dots appeared to colocalize with Borealin (figure 4c, arrows). In conclusion, CPC and ESCRT-III Snf7 components associate at the midbody before abscission in both Drosophila and human cells.

Aurora B phosphorylates three serine residues in the C-terminal linker region of CHMP4C
One obvious implication of the interaction between Borealin and CHMP4 proteins is that CPC could regulate ESCRT-III activity through phosphorylation of one or more of its components. To test this hypothesis, we investigated whether Aurora B could phosphorylate recombinant GST::CHMP4 proteins in an in vitro kinase assay. We found that all three GST::CHMP4 proteins were phosphorylated by Aurora B in vitro, although GST::CHMP4C seemed the best target (figure 5a). Therefore, we next focused on the identification of the CHMP4C residues phosphorylated by Aurora B using the same in vitro phosphorylation assay. Only the acidic C-terminal half of CHMP4C was strongly phosphorylated by Aurora B (figure 5b), and an in silico prediction analysis using the GPS 2.1 software [23] identified a serine located at position 210 as the best candidate for phosphorylation by Aurora B (table 3). The introduction of a serine (S) to alanine (A) mutation at this position (S210A) significantly reduced, but did not completely abolish, the level of Aurora B phosphorylation (figure 5c). To identify the other Aurora B phosphorylation sites, we mutagenized the majority of serines and threonines listed in table 3 to alanine and analysed the effect of these mutations, either alone or in combination with S210A, on Aurora B phosphorylation (see electronic supplementary material, figure S2).   figure S2) indicating that these three residues are the major Aurora B phosphorylation sites in the C-terminal region of CHMP4C. Quite interestingly, these residues lie in a 'linker' region between the last two alpha-helices that has been shown to be important for the regulation of CHMP4 protein activity [24].
To analyse CHMP4C phosphorylation in vivo, we generated an antibody against a CHMP4C peptide encompassing residues 206 -220 and containing phosphorylated serines at positions 210, 214 and 215 (TARRSRAASSQRAEEC). This antibody detected a band of approximately 30 kD and a band corresponding to GFP-tagged CHMP4C in Western blots of synchronized HeLa cell extracts. Both signals were absent in a twin blot that had been pre-incubated with l-phosphatase (see electronic supplementary material, figure S3), indicating that our antibody specifically recognizes a phosphorylated form of CHMP4C. We next used this antibody to analyse the profile of phosphorylated CHMP4C during mitotic exit and after treatment with the Aurora B inhibitor ZM447439 [25]. HeLa cells were transfected with GFP::CHMP4C, synchronized in metaphase by thymidine/ nocodazole block and then released into medium containing either ZM447439 or its solvent DMSO as control. In control cells, phosphorylated CHMP4C was strongly reduced 3 h after nocodazole release, immediately after cyclin B degradation and concomitantly with a reduction in the level of Borealin (figure 5d). These results indicate that CHMP4C is dephosphorylated during mitotic exit at a time when cells have already completed furrowing and CPC is being degraded. We rule out the possibility that this profile might reflect overall degradation of CHMP4C because the signal   Strikingly, the phospho-specific CHMP4C signal disappeared much more rapidly after the addition of the Aurora B inhibitor, even though both cyclin B and Borealin persisted (figure 5d). We therefore conclude that Aurora B phosphorylates CHMP4C in vivo and that the Aurora B target sites are dephosphorylated in cells that have completed furrow ingression.
3.4. CHMP4C phospho-mimic and nonphosphorylatable mutants accumulate normally at the midbody but their over-expression cause cytokinesis failure To investigate whether Aurora B phosphorylation could affect CHMP4C localization and/or function, we transfected HeLa cells with GFP::CHMP4C mutants containing either phosphomimic (serine to glutamate; S to E) or non-phosphorylatable (S to A) substitutions at position 210 alone or at positions 210, 214 and 215 (triple mutants). All the mutants accumulated at the midbody with a pattern similar to the wild-type protein ( figure 6a). This suggests that the phosphorylation status of CHMP4C does not affect its ability to accumulate at the midbody, with the caveat that phospho-mimic mutants might not exactly reflect the localization and behaviour of the endogenous pool of phosphorylated CHMP4C. We next asked if expression of these GFP::CHMP4C mutants could affect cytokinesis. Over-expression of GFP-tagged wild-type CHMP4C has already been reported to cause cytokinesis failure [6] and we found that transfection of all four GFP::CHMP4C phosphomutants caused an increase in the number of multi-nucleated cells compared with control cells transfected with GFP alone or cells transfected with wild-type GFP::CHMP4C ( figure 6b,c). That the presumptive phospho-mimic mutants behaved like the non-phosphorylatable variant might indicate either that the S to E substitutions cannot successfully mimic Aurora B phosphorylation or that the CHMP4C phosphorylation-dephosphorylation cycle must be tightly regulated for proper CHMP4C function. Taken together, our results indicate that phosphorylation by Aurora B controls the function of CHMP4C during late cytokinesis.

Discussion
Our findings identify a novel, evolutionarily conserved interaction between CPC and ESCRT-III Snf 7 components. In both Drosophila and human cells, the CPC Borealin subunit interacts directly with Shrb/CHMP4 proteins, and in human cells Aurora B phosphorylates the C-terminal tail of CHMP4C. The latter mechanism seems to have evolved specifically for the human CHMP4C protein because the three residues phosphorylated by Aurora B-S210, S214 and S215-are not present in the other two human paralogues and Shrb (figure 2). Consistent with this, Aurora B weakly phosphorylated CHMP4A, CHMP4B and Shrb in vitro (figure 5, and data not shown). The lack of conservation of phosphorylation sites between orthologues is not uncommon; for example, the Polo-like kinase-1 phosphorylation sites in PRC1, a protein required for bundling the central spindle, are not conserved in its Drosophila orthologue Fascetto [26,27]. Our results suggest that the CHMP4C phosphorylation status does not affect its ability to accumulate at the midbody but it does impair its function during cytokinesis ( figure 6). Interestingly, the serines phosphorylated by Aurora B are located in a region linking the C-terminal-most two alpha-helices of CHMP4C ( figure 7). This region has been found to be essential for regulating the activity of many ESCRT-III proteins in various biological processes, including viral budding and cytokinesis [8,24]. Current models propose that ESCRT-III proteins cycle between a default auto-inhibitory or 'closed' status and an 'open', active form (figure 7) [24,28]. In the 'closed' status, the basic and acidic halves of the ESCRT-III protein are tightly associated, and the protein is soluble and does not interact with other components of the ESCRT machinery and the membrane. In the 'open' form, the two halves of the ESCRT-III protein dissociate, and the protein forms polymers that associate with the membrane and other ESCRT-III components (figure 7) [24]. The C-terminal tail of ESCRT-III proteins, including the last a-helix and the linker region, has been proposed to play a key role in the transition from the closed to the open status, possibly through its interaction with other regulatory factors [24,28]. Our results suggest that Aurora B phosphorylation could prevent the conversion of The interaction of Borealin with the N-terminal basic region of CHMP4 components could simply serve to bring Aurora B in close proximity of its CHMP4C target or could have itself a regulatory function. The evidence that this interaction is conserved in Drosophila (figure 1), where Shrb does not contain the Aurora B phosphorylation sites identified in CHMP4C (figure 2) and does not seem to be a target of this kinase (data not shown), strongly supports this second hypothesis. Thus, we would like to propose a model where CPC could preclude the activation of CHMP4 components at the cleavage site by two concurrent mechanisms: (i) the binding of Borealin to CHMP4 proteins could directly interfere with their ability to form polymers and associate with the membrane, and (ii) Aurora B phosphorylation could prevent the interaction of CHMP4C with regulatory factors necessary for its transition from an auto-inhibited to an active state (figure 7). These two mechanisms could inhibit the formation of CHMP4 arrays or filaments that have been described at the abscission site in human cells [10] and thus abscission could proceed only when CPC is no longer present at the midbody (figure 7). This model would explain how CPC could control abscission at the end of cytokinesis to prevent cell separation in the presence of DNA at the cleavage site and the consequent formation of aneuploid and/or polyploid cells [18,19].

Cell culture, DNA transfection and siRNA treatments
The Dmel strain of S2 cells (Invitrogen) was grown in a serumfree medium supplemented with antibiotics. Generation of blasticidin-resistant stable cell lines was performed as described [29] with the only exception that Fugene HD transfection reagent (Roche) was used following the supplier's instructions. Human HeLa (Kyoto and S3) cells were maintained in Dulbecco's modified Eagle medium (Invitrogen) supplemented with 10 per cent foetal calf serum and antibiotics. To synchronize cells in metaphase, HeLa cells were arrested in S phase by the addition of 2 mM thymidine (Sigma) for 18 h, washed twice with phosphate-buffered solution (PBS) and released for 5 h in a fresh complete medium. After release, cells were cultured for a further 16 h in a fresh complete medium in the presence of 50 ng ml 21 nocodazole (Sigma) and then harvested by mitotic shake-off. Mitotic cells were washed three times with PBS and resuspended in a fresh complete medium with or without the addition of the Aurora B inhibitor ZM447439 (Tocris Bioscience) at the final concentration of 2 mM.

Midbody purification
HeLa S3 cells were synchronized in metaphase as described earlier, but after nocodazole treatment, they were released in a fresh medium supplemented with 25 mM MG132 (Sigma). After 2 h, cells were washed three times in PBS, released in a normal medium and allowed to progress throughout mitosis. After 1.5-2 h, when the vast majority of cells had completed furrowing, cells were centrifuged at 200g for 3 min, washed once with warm H 2 O and gently resuspended in a hypotonic swelling solution containing 1 mM PIPES pH 7, 1 mM MgCl 2 , 5 mg ml 21 Taxol (Sigma) and a cocktail of protease inhibitors (Roche). Cells were immediately centrifuged at 200g for 3 min, resuspended in a solution containing 1 mM PIPES pH 7, 1 mM EGTA, 1 per cent NP40, 5 mg ml 21 Taxol and protease inhibitors (Roche), and vortexed vigorously. After the addition of 0.3 volumes of cold 50 mM 2-(N-morpholino) ethane sulphonic acid (MES) pH 6.3, cells were incubated on ice for 20 min and then centrifuged at 250g for 10 min. The supernatant was then transferred to a new tube and centrifuged at 740g for 20 min to pellet midbodies. The pellet was resuspended in 50 mM MES pH 6.3, layered over a cushion of 40 per cent glycerol (w/v) in 50 mM MES pH 6.3 and centrifuged at 2800g for 45 min. After a final wash in 50 mM MES pH 6.3, midbodies were plated on poly-lysine-coated coverslips and processed for immunofluorescence as described in §5.6.

Molecular biology
Gateway technology (Invitrogen) was used in all cloning procedures as described [30]. The destination vectors used for the expression in Drosophila cultured cells were described [29 -31]. The pDEST15 vector (Invitrogen) was used for bacterial expression of GST-tagged proteins. pDEST53 and pDEST47 (Invitrogen) were used to express GFP-tagged proteins in human cells. pNPE1 (a gift from G. Tzolovsky, University of Cambridge, UK) was used to express PtAtagged Borealin in HeLa cells.

Affinity purification and in vivo pull-down assays
PtA purifications were performed essentially as described previously [29,31]. Approximately, 10 9 cells were harvested by centrifugation and frozen in liquid nitrogen. The cell pellet was then resuspended in 5 ml of extraction buffer (50 mM HEPES pH 7.5, 100 mM KAc, 50 mM KCl, 2 mM MgCl 2 , 2 mM EGTA, 0.1% NP-40, 5 mM DTT, 5% glycerol and Roche Complete Protease Inhibitors) and homogenized using a high-performance disperser (Fisher). Homogenates were agitated at 48C for 30 min and clarified by centrifugation at 6000 r.p.m. in a SS34 rotor. Dynabeads (200 ml; Invitrogen) conjugated to rabbit IgG (MP Biochemicals) were added to the supernatants and incubated for 4 h under continuous agitation at 48C. Beads were then washed five times for 10 min in 10 ml of extraction buffer. Proteins were then eluted from beads with 0.5 M NH 4 OH and 0.5 mM EDTA, concentrated, and analysed by LC-MS/MS. The MS/MS fragmentation data that we achieved were used to search the National Center for Biotechnology Information and Flybase databases using the MASCOT search engine (http://www. matrixscience.com). Probability-based MASCOT scores were used to evaluate identifications. Only matches with p , 0.05 for random occurrence were considered significant.
For PtA pull-down assays in HeLa cells, cells were collected and processed as explained earlier. Eluates were then fractioned on an 8-16 per cent SDS-PAGE, transferred onto a polyvinylidene fluoride (PVDF) membrane and probed with the antibodies indicated in figure.

In vitro binding assay
DNA fragments coding for Borr, Borealin and CHMP4C were generated by PCR and cloned into pDEST15 (Invitrogen) to express N-terminal GST-tagged polypeptides in Escherichia rsob.royalsocietypublishing.org Open Biol 2: 120070 coli. The GST-tagged products were then purified using Glutathione Sepharose 4B according to the manufacturer's instruction (GE Healthcare). [ 35 S] Methionine-labelled Borealin, CHMP4A, CHMP4B, CHMP4C and Shrb fragments were prepared from corresponding PCR products amplified using primers harbouring a T7 promoter and then transcribed and translated in vitro using the TnT T7 quick-coupled transcription/translation system (Promega) in the presence of [ 35 S] methionine (Perkin Elmer). Generally, 25 ml of glutathione sepharose beads containing purified GST-proteins were mixed with 10 ml of [ 35 S] methionine-labelled polypeptides and 300 ml of NET-N þ buffer (50 mM Tris -HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 0.5% NP-40 and a cocktail of Roche complete protease inhibitors) and incubated in ice for 30 min with periodic agitation. The mixture was then washed five times by adding 500 ml of NET-Nþ buffer followed by centrifugation at 500g for 1 min. Beads were resuspended in 25 ml of Laemmli SDS-PAGE sample buffer and typically one-fifth of the mixture (10 ml) was loaded on 8-16 per cent Tris-Glycine gel. Proteins were then transferred onto a nitrocellulose membrane using the iBlot dry blotting system (Invitrogen) and exposed to X-ray films at 2808C.
For immunofluorescence, Dmel and HeLa cells were grown on 22 Â 22 mm coverslips (Menzel-Gläser) and fixed in PHEM buffer (60 mM Pipes, 25 mM Hepes pH 7, 10 mM EGTA, 4 mM MgCl 2 , 3.7% formaldehyde) for 12 min. They were then washed three times for 10 min with PBS and incubated in blocking buffer (PBS, 0.5% Triton X-100, 3-5% BSA) for 1 h at room temperature. Incubation with primary antibodies was performed over night at 48C in PBT (PBS, 0.1% Triton X-100, 1% BSA). After two washes in PBT, coverslips were incubated with Alexa-fluor-conjugated secondary antibodies (Invitrogen) diluted in PBT for 2 h at room temperature and then washed twice with PBT and once with PBS. Coverslips were finally mounted using Vectashield mounting medium with DAPI (4 0 ,6-diamidino-2-phenylindole; Vector Laboratories) and visualized on a Zeiss Axiovert 200 fluorescence microscope equipped with a 100Â objective (NA 1.4). Images were acquired using a Photometrics CoolSNAP HQ2 camera with METAMORPH software (Molecular Devices) and assembled and adjusted for contrast using ADOBE PHOTOSHOP CS3.

In vitro phosphorylation assay
Purified GST-tagged CHMP4A, CHMP4B and CHMP4C proteins, including CHMP4Ca12, CHMP4Ca345 and various CHMP4Ca345 variants containing S to A mutations, were incubated with or without 100 ng recombinant Aurora B (Cell Signaling) in a kinase buffer (20 mM Hepes pH 7.5, 5 mM MgCl2, 1 mM DTT, 0.1 mM ATP) containing 2 mCi [g-32 P]ATP (Perkin Elmer). Reactions were incubated at 308C for 30 min, quenched by the addition of 2Â Laemmli sample buffer, boiled and separated by electrophoresis on a 4-20 per cent SDS-PAGE. Gels were stained with Bio-Safe Coomassie stain (Bio-Rad), and then dried and exposed at 2808C.

Phosphatase treatment
Proteins were separated by SDS-PAGE and transferred to PVDF Immobilon-P membrane (Millipore). The blots were then washed twice with water, blocked in Tris-buffered saline (TBS) containing 0.1 per cent BSA and 0.1 per cent Triton X-100 for 1 h at room temperature and then incubated in phosphatase buffer (TBS 1X, 0.1% BSA, 0.1% Triton X-100, 2 mM MnCl 2 ) containing 400 units ml 21 of l-phosphatase (NEB) for 4 h at room temperature. After incubation, blots were washed for 5 min in PBS þ 0.1 per cent Tween 20 and processed for antibody detection.