Binding Partner Switching on Microtubules and Aurora-B in the Mitosis to Cytokinesis Transition*

The cytoskeleton globally reorganizes between mitosis (M phase) and cytokinesis (C phase), which presumably requires extensive regulatory changes. To reveal these changes, we undertook a comparative proteomics analysis of cells tightly drug-synchronized in each phase. We identified 25 proteins that bind selectively to microtubules in C phase and identified several novel binding partners including nucleolar and spindle-associated protein. C phase-selective microtubule binding of many of these proteins depended on activity of Aurora kinases as assayed by treatment with the drug VX680. Aurora-B binding partners switched dramatically between M phase to C phase, and we identified several novel C phase-selective Aurora-B binding partners including PRC1, KIF4, and anaphase-promoting complex/cyclosome. Our approach can be extended to other cellular compartments and cell states, and our data provide the first broad biochemical framework for understanding C phase. Concretely, we report a central role for Aurora-B in regulating the C phase cytoskeleton.

Cytokinesis requires coordinated reorganization of microtubules, actin, and membranes, which implies global regulation of cellular biochemistry. Animal cells are competent to undergo cytokinesis during a brief window in the cell cycle, called C phase, which starts shortly after cells exit mitosis (M phase) and lasts ϳ30 -60 min (1,2). The cytoplasm is globally regulated in M phase by the activity of one master kinase, Cyclin B-CDC2 (CDK1), that is strongly activated as cells enter M phase and controls the organization and function of 100s to 1000s of substrates (3,4). C phase regulation, in contrast, is poorly understood. Given the global cytoplasmic changes that characterize C phase, global regulation by kinases seems likely. CDK1 activity decreases abruptly at anaphase onset, and the phosphatases that oppose it increase in activity (5). Thus, one global change in the transition from M to C phase is likely to be the loss of CDK1 phosphorylation, although the kinetics by which different substrates are dephosphorylated is unclear. Two other kinases, Aurora-B and PLK1, are broadly implicated in the regulation of C phase biochemistry and might function to some extent as global regulators (6). However, both these kinases are also active in M phase, and it is not clear how their activity and substrate specificities change between M and C phase.
Many cytoplasmic systems and compartments are likely to be controlled by C phase-regulating kinases. Here, we focus on the microtubule cytoskeleton as one specific compartment, although our methods could easily be adapted to any other compartment that could be isolated. Microtubules change dramatically in organization from the highly dynamic mitotic spindle in M phase to the less dynamic midzone and astral arrays in C phase. Several conserved proteins, including PRC1 and MKLP1, have been identified by cytology and genetics as C phase-specific microtubule-binding proteins that are required for midzone assembly and cytokinesis (7), but given the complexities of cytokinesis, others are likely to exist.
A major barrier to biochemical analysis of C phase in mammalian cells has been the lack of a method to synchronize cells in this phase. Synchronization in M phase with spindledamaging drugs followed by washout allows only partial synchronization as C phase is short and cells take variable time to assemble spindles and exit M phase. Therefore, it is not possible to separate mitosis and C phase this way. We used a recently developed pharmacological approach; cells were arrested in monopolar mitosis where microtubule dynamics are relatively normal with a Kinesin-5 inhibitor and then forced into monopolar cytokinesis using a CDK1 inhibitor, giving excellent C phase synchrony (8,9). This method works in part by global inhibition of CDK1, in part by activation of phosphatases (5), and in part by activating Cdh1 (through loss of inhibitory CDK1 phosphorylation (10)), which activates the anaphase-promoting complex/cyclosome (APC/C). 1 The net effect is to mimic most or all of the known regulation that occurs after normal anaphase onset. Cytological characterization of C phase induced in this way revealed that essentially all of the changes characteristic of C phase occur with relatively normal kinetics (11), so this method is a good starting point for investigating the biochemistry of cytokinesis.
We used LC-MS to characterize large scale changes in protein biochemistry in C phase. Although MS was previously used to identify midbody proteins (12), it has not been applied before to the problem of elucidating global C phase regulatory mechanisms. To allow quantitative comparison between phases, we used stable isotope labeling by amino acids in cell culture (SILAC) (13). Our approach can be likened to performing thousands of Western blots to compare microtubule-and Aurora-B-binding proteins between M phase, C phase, and interphase except that it was unbiased in terms of the proteins analyzed. Thus, our proteomic measurements provide a thorough evaluation of the biochemistry of each cell state as several measurements were made for each protein across two or three biological replicates, thus providing quantitative data with better accuracy than Western blots for thousands of proteins in multiple analyses (14). Furthermore, as we performed this analysis on a large scale, several known C phase proteins were identified and quantitated, thus generating several internal standards to validate the data. Using these approaches, we provide the first large scale description of C phase biochemistry and use chemical genetics to understand C phase.

Cell Growth and Arrests
HeLa S3 cells were grown in Dulbecco's modified Eagle's medium (Caisson Laboratories) without lysine and arginine with 10% dialyzed fetal bovine serum (F0392; Sigma). Either light ( 12 C, 14 N) or heavy ( 13 C, 15 N) lysine and arginine amino acids (Sigma) were added in 0.146 and 0.08 4g/liter concentration, respectively. To arrest at mitosis, cells were first treated with 2 mM thymidine (Sigma) for 16 h; after 3 h of release, cells were treated with 10 M S-trityl-Lcysteine (STC) for 12 h. For C phase, after 12 h of STC treatment, cells were treated with 100 M purvalanol-A (Tocris Bioscience) for 15-20 min. Cells were washed extensively with PBS containing purvalanol-A and frozen. A double thymidine block protocol was followed for interphase synchronization. Subsequently, fluorescence-activated cell sorting (FACS) analysis and immunostaining were performed to confirm the cell stages.

Data Acquisition
Peptides derived from in-gel digested proteins were analyzed by on-line C 18 nanoflow reversed phase HPLC (Eksigent nanoLC⅐2D TM ) hyphenated to a linear ion trap/orbitrap mass spectrometer (LTQ-Orbitrap, Thermo Scientific). Samples were loaded onto an in-house packed 100-m-inner diameter ϫ 15-cm C 18 column (Magic C 18 , 5 m, 200 Å, Michrom Bioresources) and separated at 400 nl/min with 60-min linear gradients from 5 to 35% acetonitrile in 0.4% formic acid. Survey spectra were acquired on the Orbitrap with a resolution of 30,000. Up to six of the most intense ions per cycle were fragmented and analyzed in the linear trap.
The Thermo .raw files were converted into complete peak lists and entered into a relational database. The data conversion was performed using in-house written software utilizing the XRawfile2.dll library (version 2.1.1.0), which was provided by Thermo Scientific as part of the Xcalibur software package. The library is also used by MSQuant, Trans-Proteomic Pipeline, and the Sashimi project to convert the proprietary .raw file format into generic peak lists. For each MS/MS spectrum, the 200 most intense fragment ions were converted into an .mgf file without any further data processing such as smoothing, deisotoping, and filtering. Comparisons between the .mgf files generated by our approach and those generated by DTASuper-Charge showed that our approach performed favorably.

Database Searches
Database searches were performed using Mascot (v.2.104) with 10 -20-ppm precursor ion and 0.8-Da product ion search tolerances for MS2 data. Cysteine carbamidomethylation was used as a fixed modification, and pyroglutamic acid (N-terminal Gln), deamidation (Asn and Gln), oxidation (Met), and phosphorylation (Ser, Thr, and Tyr) were set as variable modifications. One missed tryptic cleavage was allowed.
Only the phosphopeptide hits with an ion score above the identity threshold as provided by the Mascot search were reported in this study with a false positive rate of 1%. For the phosphorylation analysis of non-labeled cells, ProteinPilot was used to analyze the data, and phosphopeptides with 95% and higher confidence were reported (15).
All peptides with a score above 33 were used for quantification for the microtubule binding assay and above 28 for Aurora-B immunoprecipitations. The data sets were searched against the International Protein Index human database (v.3.36) with 69,012 protein entries for the target database and 69,009 protein entries for the decoy database. A threshold of 33 resulted in an averaged false discovery rate of ϳ1%, and a score of 28 resulted in a false discovery rate of ϳ2%, estimated through the use of the decoy database searches. The SILAC ratios of all quantifiable peptides with good peaks were calculated using the MSQuant program (16). In microtubule pelleting assays and Aurora-B immunoprecipitations, phosphorylated peptides were excluded in the quantifications.

Microtubule Pelleting and Immunoprecipitation
M phase or interphase and C phase cells were lysed in a buffer (100 mM Pipes, 1 mM MgCl 2 , 2 mM EGTA, 0.5% Nonidet P-40, 1 mM DTT) containing protease inhibitor (Roche Applied Science EDTA-free mixture tablet) and phosphatase inhibitors (1 M okadaic acid, 1 M microcystin, 10 mM NaF, 1 mM sodium orthovanadate, 1 mM ␤-glycerol phosphate, 1 mM sodium pyrophosphate). Lysates were precleared at 14,000 rpm for 5 min. Cell lysates were mixed in an equal concentration, determined using Bradford assay. The mixed supernatant was further clarified by centrifugation at 43,100 rpm (100,000 ϫ g) in a TLA-100.3 rotor for 25 min at 4°C. The supernatants were incubated either with 25 M Taxol and 0.5 mM GTP for 20 min at room temperature or with Taxol-stabilized pure microtubules plus Taxol. Microtubules were pelleted through a 40% glycerol cushion in BRB80 (with protease inhibitors, phosphatase inhibitor, 1 mM DTT, 15 M Taxol, 1 mM GTP) for 25 min at 43,000 rpm at 20°C. The pellet was resuspended, loaded onto a second cushion, and spun as above (17). The pellet was resuspended in sample buffer and analyzed by SDS-PAGE. The Coomassie-stained gel was cut into 7-10 pieces; each gel piece was digested with trypsin (Promega).
For the microtubule pelleting experiment of VX680-treated C phase cells, cells were labeled either with the heavy or with the light isotope and arrested in mitosis as described above. One portion was treated Quantitative Analysis of C Phase Cytoskeleton with purvalanol-A and 400 nM VX680 simultaneously, and the other was treated with purvalanol-A alone. Lysates were prepared and clarified. To deconvolute the role of Aurora-B in spindle assembly from its role in regulating the C phase cytoskeleton, Taxol-stabilized pure microtubules and Taxol were added to each lysate separately. The microtubules were then pelleted by centrifugation through glycerol cushions. Finally, the microtubule pellets were pooled in SDS sample buffer immediately before separation by SDS-PAGE. Because the microtubule pelleting experiments require more than 2 h to perform, we used this late pooling strategy to minimize the biochemical equilibration between the two samples, which would lead to an underestimation of the differences between the VX680 treatment and the control.
For Aurora-B immunoprecipitation, cells were lysed in 50 mM Hepes, pH 7.4, 150 mM KCl, 1 mM MgCl 2 , 10% glycerol, 1 mM EGTA, 0.5% Nonidet P-40, 1 mM DTT including protease and phosphatase inhibitors as above. Lysates were pooled in an equal protein concentration following Bradford assays. The extract was incubated with protein A beads (Bio-Rad) alone followed by incubation with anti-Aurora-B antibody coupled with protein A to immunoprecipitate Aurora-B at 4°C. Subsequently beads were washed extensively with lysis buffer, and the immunoprecipitates were eluted by boiling in 1ϫ SDS sample buffer. The gel was cut into eight slices and in-gel trypsin-digested as described above.

Phosphoenrichment
The samples were separated using SDS-PAGE, the gel was cut into 10 slices, and each gel slice was trypsin-digested. TiO 2 was used to enrich for phosphopeptides. Briefly, lyophilized samples were dissolved in binding buffer (5% TFA, 40% acetonitrile). After loading samples onto the TiO 2 columns (TopTip, Glygen), the column was washed with the binding buffer. Phosphopeptides were eluted first with 100 mM ammonium bicarbonate and then with 0.4 M NH4OH, 40% acetonitrile.

Statistical Methodology
Transformation of Ratio Data-As the SILAC ratio data display an asymmetric grid with departures to the one side being exaggerated and to the other side being diminished, the data were transformed. Similar in its nature to the logarithmic transformation (20), we transformed a ratio value by o ϭ /( ϩ 1).
Peptide Level Analysis-For the phosphorylation analysis of total cell lysate, significance bounds were calculated using non-phosphorylated peptides from the flow-through of the phosphoenrichment analysis.
In the phosphorylation analysis of the microtubule binding assay, for the C phase selectivity, non-phosphorylated peptides of cytoki-nesis-selective microtubule-binding proteins were selected, whereas for the M phase selectivity, nonphosphorylated peptides of mitosisselective microtubule-binding proteins were selected. For the nonphosphorylated peptides, we computed the respective transformed ratios for each repeat and then computed empirical confidence intervals for the ratios. For all phosphorylated peptides in the analysis, we identified those peptides showing significantly different ratios, outside the empirical confidence intervals (90%), over the various repeats.
When assuming no systematic departure of a peptide (so under the null hypothesis, H 0 ), the distribution of peptides within the interval is binomial. Considering the 90% confidence interval used, the probability of not being within this interval is p ϭ 0.1. Thus, considering a peptide observed in n repeats and k times that is not within the interval, the p value for the departure from the majority of all peptides is given as p(minimum of k of the n observed peptides not in the Data from the phosphorylation analysis have been deposited in the Proteomics Identifications (PRIDE) database (57) with accession numbers 9971-9983.
Protein Level Analysis-To compare protein behavior against a reference protein, a peptide number-dependent approach was applied. For microtubule pelleting assays from different cell stages, the reference protein was tubulin. For microtubule pelleting with vs. without VX680 treatment, prepolymerized tubulin was added, so abundant housekeeper proteins, such as heat shock proteins and glyceraldehyde-3-phosphate dehydrogenase, were used as references. For the Aurora-B immunoprecipitation experiments, Aurora-B, tubulin, and actin were selected to have sufficient quantified reference peptides.
To discover whether a protein p 1 departed significantly from the reference protein p 1 r , the mean m 1 of the n transformed ratios of the peptides of p 1 was compared with a robust, nonparametric, bootstrap confidence interval for the mean of n peptides. Therefore, 1000 sets of n transformed peptide ratios were resampled with replacement from p 1 r , and for each of these sets, the mean was computed. Among these values, a 99% confidence interval was calculated. If the observed mean m 1 fell within this interval, it could not be concluded that p 1 departed from the reference protein.
Thus, for proteins with fewer quantified peptides, the confidence intervals are less stringent than for proteins with more quantified peptides. This allowed the elimination of proteins with fewer peptides unless they displayed a dramatic change between the two conditions.

RESULTS
Drug-based Synchronization in C Phase-Our synchronization methods are outlined in Fig. 1. HeLa S3 cells grown in suspension were arrested in monopolar mitosis using the Kinesin-5 inhibitor S-trityl-L-cysteine, and C phase was subsequently induced by adding the CDK1 inhibitor purvalanol-A for 15 min. Flow cytometry and microscopy data showed that synchronies of Ͼ90% were typically achieved for mitosis. Immunofluorescence was used to estimate C phase synchrony: cells in which phospho-Histone H3 staining was lost and microtubules were arranged in a monopolar midzone were scored as being in monopolar cytokinesis (11). By this criterion, Ͼ85% of the cells (n ϭ 64) were in C phase following a 15-min treatment with purvalanol-A (Fig. 1). To determine whether C phase synchronization induced with a CDK1 inhibitor was a good biochemical surrogate for normal C phase, we collated a list of proteins known to be involved in C phase from the literature and interrogated our data to determine whether these proteins were C phase-selective under our conditions (supplemental Table VI) within the experiments discussed below. We found an excellent correlation, suggesting that the drug synchronization method is valid for probing C phase biochemistry.
C Phase-selective Microtubule-binding Proteins-To identify C phase-selective microtubule-binding proteins, we combined clarified lysates from cells synchronized in C phase with either M phase or interphase lysates that were differentially labeled with heavy isotopes. We then polymerized endogenous tubulin with Taxol, sedimented the microtubules, and analyzed bound proteins ( Fig. 2A). All the analyses in this study were performed on FTMS instruments. Fig. 2B shows a typical mass spectrum for a sample where C phase cells labeled with light lysine and arginine (Lys: 12 C 6 , 14 N 2 ; Arg: 12 C 6 , 14 N 4 ) were pooled with M phase cells labeled with heavy lysine and arginine (Lys: 13 C 6 , 15 N 2 ; Arg: 13 C 6 , 15 N 4 ). A false discovery rate calculation was performed for the peptides as described under "Materials and Methods." More than 1000 proteins were identified with two or more quantified peptides with Mascot scores of Ͼ33 (supplemental Table I). A significance value for the quantitative difference (ratio of M phase/C phase or C phase/interphase) for each protein was determined by comparison with tubulin as a reference protein (see "Materials and Methods"). Proteins that were marginally significant based on the bootstrap confidence interval were manually evaluated. Proteins that bound reproducibly to microtubules over three replicate experiments and that showed a statistically significant preference for binding in C phase or M phase lysates are shown in Table I. All microtubule-binding proteins observed and their isotope ratios are reported in supplemental Table I. To evaluate the reliability of our SILAC analysis, we used Western blotting and confirmed cell cycle selectivity of several putative M or C phase-selective microtubule binders (Fig. 2C). Different groups of proteins showed different selectivity of microtubule binding in the three cell stages studied (Table I).
Proteins enriched in C phase relative to both M phase and interphase include several known midzone-enriched proteins: KIF20A (also known as MLKP2) (21), RACGAP1 (22,23), PRC1 (24,25), and INCENP (26). The most striking novel C phase-selective binder is NUSAP. This is a potent microtubule-stabilizing and -bundling protein (27) and might be important for stabilizing the midzone. The fact that NUSAP is an importin substrate (19) and that the midzone protein TD60 is thought to have Ran guanine nucleotide exchange factor (GEF) activity (28) suggests the possibility that spatial regulation of the Ran system contributes to spatial organization of microtubules in C phase as it does in M phase (29). Another class of proteins that might contribute to C phase microtubule organization is the ϩtip proteins CLIP1, CLIP2, and CLASP1. Their M/C ratios were all ϳ0.3, implicating them in increasing microtubule stability as cells leave M phase. A novel C and interphase phase-selective binder was nucleophosmin (NPM1), a small, abundant phosphoprotein to which multiple functions have been ascribed (30). This protein was found in midbodies in a proteomics study (12) but had not otherwise been implicated in cytokinesis. Immunostaining showed NPM1 to be localized to the midbody late in cytokinesis (Fig.  2D). The C phase-selective subset also included several RNA processing enzymes whose significance is unknown. M phase-selective microtubule-binding proteins included several well known mitotic spindle proteins, including Kif22 (Kid), TACC3, and Aurora-A kinase. The most M phase-enriched was CKAP2L, a microtubule-associated protein that seems worthy of further study on the basis of this result.
M and C Phase-selective Protein Phosphorylation-Microtubule-binding proteins are often regulated by phosphorylation, and phosphorylation probably causes global cytoplas- mic changes during cell cycle progression. To identify proteins that are differentially phosphorylated between M and C phase, we performed phosphopeptide enrichment using TiO 2 on total lysate proteins. 50,000 MS2/MS3 spectra were acquired that identified ϳ4000 unique phosphopeptides in mitosis and ϳ3000 in cytokinesis (supplemental Table II). To obtain a more selective set of C phase-selective phosphorylation sites, phosphopeptide enrichment was combined with SILAC. Fig. 3A shows the distribution of all phosphopeptides from total clarified lysate in a typical SILAC comparison with M/C ratios displayed on a log scale (M was heavy in this experiment). Similar results were obtained in a biological replicate where the labels were switched (Fig. 3A). This curve is notably skewed to the M phase side, implying that there are more total phosphopeptides in M phase than C phase. A statistical analysis (see "Materials and Methods") revealed ϳ700 phosphorylation sites that were significantly enriched in M phase and ϳ170 in C phase (Fig. 3B). Table II summarizes some of the identified M phase-and C phase-selective phosphorylation sites (full data are in supplemental Table III). C phase-specific phosphorylations have been characterized for few proteins with the notable exception of intermediate filaments (31). We reidentified a C phase-specific Vimentin Ser-72 phosphorylation site (32) as a doubly phosphorylated Ser-71, Ser-72 peptide that was extremely C phase-specific (isotope ratio of 0.08 with M phase heavy (Table II)), validating our phosphopeptide analysis.
To more specifically probe the role of C phase phosphorylation in controlling microtubule organization, we analyzed phosphopeptides from microtubule-binding proteins by combining SILAC, microtubule isolation, and phosphopeptide enrichment. C phase selectivity in this experiment can arise either from selective binding of the protein to microtubules and/or C phase-selective phosphorylation. Most of the phosphorylation sites identified were common to both samples. 122 M phase and 72 C phase highly selective phosphorylation sites on microtubule-binding proteins were identified ( Fig. 3C and supplemental Table III). This analysis confirmed several previously identified phosphorylation sites, including GLAP-NpTPGK (where pT is phosphothreonine) from PRC1 (a known CDK1 site (33)) and ASKSDATSGILNSpTNIQS from PRC1 (a known PLK1 site (34)). It also revealed many novel M  Proteins that bind to microtubules preferentially in C phase or M phase are listed. M/C is the mean ratio of M phase (heavy) versus C phase (light) of quantified peptides; C/I is the mean ratio of C phase (heavy) versus interphase; STD is the standard deviation of the mean values; # Pep. represents the number of quantified peptides; U. # Pep. is the number of quantified unique peptides for M/C. The microtubule-binding proteins are grouped and color-coded by their SILAC ratios: yellow, M phase-selective; light yellow, both M and C phase-selective compared with interphase; blue, C phase-selective; light blue, C phase-and interphase-selective compared with M phase. The gray panel is the group of C phase-selective proteins that were not detected in C phase versus interphase experiments. MT, microtubule. and C phase sites in microtubule-binding proteins ( Fig. 3D and Table II). We used Sequence Logo (35,36) to identify numerically dominant consensus phosphopeptide sequences for phosphopeptides from microtubule-binding proteins. In M phase, the dominant motif, (T/S)PXX(K/R) (Fig. 3E), was similar to the known CDK1 consensus motif. In C phase, (S/T)P still dominates, but it is not clear whether this is a CDK1 motif because the second position is less enriched in Pro and the fourth position shows less of a bias for basic residues. Interestingly, Aurora-B was the second notable consensus sequence listed. Further analyses showed that the Aurora kinase consensus sequences ((R/K)X(S/T)) was found in 14% of highly C phase-selective phosphopeptides (n ϭ 72, M/C Ͻ 0.3) and 17% of moderately C phase-selective phosphopeptides (n ϭ 142, M/C Ͻ 0.6), but only 7% of M phase selective phosphopeptides (n ϭ 123, M/C Ͼ 1.6), suggesting that Aurora-B is an important regulator of C phase microtubuleassociated proteins.
Aurora-B Kinase Broadly Regulates the C Phase Cytoskeleton-To dissect the role of Aurora-B kinase in regulating the C phase microtubule cytoskeleton, we took a chemical genetics approach using a small molecule kinase inhibitor. VX680 is a potent, selective inhibitor of Aurora kinases (37), and we previously showed that it blocked cell polarization and midzone assembly during monopolar cytokinesis (11). Here, we used SILAC to test how inhibiting Aurora kinases perturbed C phase microtubule-binding proteins. Microtubules were sedimented from VX680-treated and control C phase cells, and the samples were analyzed using mass spectrometry. Based on two biological replicates (heavy VX680/light mock or heavy mock/heavy VX680), we identified ϳ250 candidate microtubule-binding proteins that were perturbed by the VX680 treatment (supplemental Table IV). Fig. 4A shows a subset, focusing on annotated cytoskeleton proteins. Most cytoskeleton proteins were reduced in the microtubule pellet from VX680-treated cells, implying that phosphorylation by Aurora kinases promotes microtubule binding of many proteins in C phase, contrary to the general notion that phosphorylation tends to reduce microtubule binding affinity of microtubule-associated proteins (38). A notable exception was a group of kinetochore proteins (shaded yellow, together with NuMA) whose microtubule binding was increased by VX680. Aurora-B phosphorylation is known to decrease the affinity of the Ndc80 kinetochore complex for microtubules (39), suggesting that either other kinetochore proteins are similarly regulated or they tend to co-sediment with Ndc80.
VX680 treatment of C phase cells reduced the recovery of many known microtubule-binding proteins, including several we previously identified as C phase-selective in their binding in Table I (ECT2, KIF23 (MKLP1), KIF20 (MKLP2), RACGAP1, CLASP1, CLIP1, and Aurora-B itself). VX680 inhibits Aurora-A as well as Aurora-B, and some of its effects could be a result of Aurora-A inhibition, including microtubule binding of the known Aurora-A substrates TPX2 (40), TACC3 (41), DLG7 (42), HMMR (43), and Aurora-A itself (Fig. 4A). We also observed that myosin II and intermediate filaments (and some of their binding proteins) were reduced in the microtubule pellet from VX680-treated cells. Myosin II in particular was strongly perturbed; both its heavy and regulatory light chains were reduced in the pellet. It is not clear whether this important cleavage furrow motor is sedimenting by affinity to microtubules or actin filaments or because of its own polymerization, but it is clear that VX680 reduces its recruitment to the cytoskeleton, implying positive regulation of myosin II activity by Aurora kinases during cytokinesis. Co-sedimentation of intermediate filament proteins was also reduced by VX680 (Fig.  4A, shaded in green), suggesting that Aurora kinase activity is required in C phase for organization of all three parts of the cytoskeleton.
We retested the VX680 modulation of microtubule affinity of several proteins by Western blotting (Fig. 4B) and assayed inactivation of Aurora-B kinase activity in parallel using phospho-Histone H3 (Ser-10) (44). We found strong inhibition in both lysates and microtubule pellets. Phospho Myosin regulatory light chain (MRLC) (Ser(P)-19), which stimulates myosin II activity, was unaffected in lysates, demonstrating VX680 specificity, and reduced in the pellet, consistent with the MS results showing reduced recovery of myosin II protein in the pellet (Fig. 4A). Western blotting also confirmed several of our SILAC data points, such as Aurora-B itself and MKLP1.
Reduction of intermediate filament subunit recovery in the microtubule pellet by VX680 treatment was further analyzed to see whether it correlated with loss of phosphopeptides with likely Aurora-B consensus sequences. We performed a quantitative phosphopeptide enrichment with VX680 and mocktreated C phase cell extracts using SILAC as described above. Fig. 4C lists the Vimentin and Keratin-8 phosphorylation sites that are strongly diminished after VX680 treatment.

FIG. 3. M phase and C phase phosphorylation analysis.
A, isotope ratio distribution of all phosphopeptides from clarified lysates. The x axis shows the M phase/C phase ratio on a log scale (red) and a replicate inverse isotope-labeled experiment in blue. Note that more phosphopeptides were recovered from mitotic lysates. B, number of M or C phase-selective phosphopeptides identified in clarified total lysate. This is the same data as in A, replotted to show the number of peptides for which the isotope ratio was significantly higher or lower than 1.
The cutoff values for statistical significance are shown as lines parallel to the x axis. C, number of M or C phase-selective phosphopeptides identified in microtubule-binding proteins, plotted as in B. D, quantification of the known M phase-selective Thr-481 and C phase-selective Thr-602 phosphorylation sites on PRC1 (top) and the newly identified M phase-selective Ser-628 and C phase-selective Thr-580 phosphorylation sites on RACGAP1 (bottom) in M phase (heavy) versus C phase (light). E, the consensus motif of M phase phosphopeptides (right) and C phase phosphopeptides (left) from microtubule-binding proteins. Consensus was extracted using Sequence Logo.

TABLE II Examples of M phase-and C phase-selective phosphorylation sites
#Pep represents the number of times a peptide was quantified in multiple experiments. Average (avg) Mascot scores are shown in the last column (see also supplemental Table III). Mean, standard deviation (st dev), median (med.), and median absolute deviation (mad) are calculated from the ratio M phase (heavy)/C phase (light). For each protein, C phase-selective phosphorylation sites are marked with C, and M phaseselective sites are marked with M. Bold residues in the sequence are phosphorylated: pS, phosphoserine; pT, phosphothreonine.
Most of these phosphorylation sites are C phase-selective. The Vimentin Ser-72 (LRpSSVPGVR where pS is phosphoserine) site was previously reported to be phosphorylated by Aurora-B in the cleavage furrow during normal cytokinesis (32). Fig. 4D shows the isotope ratios of the Keratin-8 Ser-13 (SYKVpSTSGPR) in M phase to C phase (top) and mock to VX680-treated C phase cell lysates (bottom). Keratin-8 Ser-13 phosphopeptide is clearly C phase-selective (5-fold elevated in C phase in comparison with M phase) and is affected by the VX680 treatment (10-fold diminished in VX680-treated C phase extracts; Fig. 4D), consistent with the hypothesis that Keratin, like Vimentin, is phosphorylated by Aurora-B during cytokinesis, perhaps to allow constriction of the cleavage furrow.
A Switch in Aurora-B Interactions between M Phase and C Phase-The data above show that Aurora kinases (most likely Aurora-B) play a central role in regulating the M to C phase switch in microtubule-binding proteins. However, we know that Aurora-B is active in both phases, so how can it mediate such a switch? One possibility is that the recruitment of the kinase to the vicinity of its substrates changes from M to C phase. To investigate how Aurora-B targeting changes between M and C phase, we characterized its binding partners, again using SILAC for quantitative comparisons. Differentially labeled lysates from M and C phase cells were pooled, and Aurora-B-binding proteins were isolated by immunoprecipitation. To decrease the influence of outliers possibly introduced by posttranslational modifications induced by Aurora-B kinase, we scored statistical significance from median and median absolute deviation values of the C/M ratio. These values were measured for three replicate experiments. We did not analyze interphase because Aurora-B is degraded at the end of C phase (45,46). Fig. 5A shows a summary of proteins that selectively bind to Aurora-B in C phase or M phase. Full data are provided in supplemental Table V. Western blotting analysis was again used to confirm the SILAC data for selected proteins as shown in Fig. 5, B and C. A notable finding is that most of the C/M ratios are Ͼ1, and some are Ͼ Ͼ1. This indicates that Aurora-B gains many new binding partners as cells enter C phase. For several interactors, including PRC1, this change is essentially switchlike.
M phase-selective Aurora-B-binding proteins are shaded in orange in Fig. 5A. Aurora-B itself and other members of the Aurora-B complex (also known as the chromosomal passen-ger complex) display C/M ratios less than 1, indicating that less of the complex was recovered from the C phase lysate (see asterisks). This may reflect the onset of degradation of Aurora-B by the APC/C pathway (45). The C/M ratios of interacting proteins were not corrected for this loss of Aurora-B; thus, the C/M ratios in Fig. 5A under-report C phase selectivity of binding.
C phase-selective Aurora-B binders with known cell division roles are shaded in blue in Fig. 5A. Intriguingly, we found that several C phase-selective microtubule binders were also C phase-selective Aurora-B binders: PRC1, with a large C/M ratio, is essentially C phase-specific in its Aurora-B binding. PRC1 also binds selectively to microtubules in C phase (Table  I) and may play a major role in targeting the kinase complex to midzone microtubules. Interestingly, the motor KIF4 is a strong interactor of Aurora-B during C phase. Because KIF4 also binds PRC1 (47), it is possible that a complex containing Aurora-B, PRC1, and KIF4 selectively assembles on C phase microtubules. SPAG-5/astrin also bound selectively to both microtubules and Aurora-B in C phase, and INCENP, a member of the Aurora-B complex, was also identified as a C phase-selective microtubule binder. The phosphosite analysis and these data together suggest two non-exclusive possibilities: the change in Aurora-B partners causes a change in Aurora-B localization, and the change in Aurora-B partners reveals a change in substrates of Aurora-B.
The formin FHOD-1 (48) was identified as a new interaction partner of Aurora-B in cytokinesis. A different formin, Diaphanous, is thought to nucleate cleavage furrow actin (49). FHOD-1 might contribute to the furrow or nucleate the actin cables observed to be associated with Aurora-B in and around the midzone during monopolar cytokinesis (11). We found that FHOD-1 localizes to the cortex in general but also to the midzone during cytokinesis where it colocalizes with Aurora-B (Fig. 5D). We also identified a phosphorylation site on FHOD-1 at Ser-367, RRpSL, an Aurora-B consensus sequence (supplemental Fig. S1). A prominent C phase-selective Aurora-B binder was the APC/C, several subunits of which appear in Fig. 5A (highlighted in light orange) with very high C/M ratios. Immunofluorescence showed that these two complexes overlap in localization at some stages of cytokinesis, for example in the midzone (Fig. 5E). Association between Aurora-B and CDC-27 was reported previously (45), but the association of the APC/C in the context of cytokinesis and localization to the midzone is novel. We have confirmed this association by reciprocal immunoprecipitation (CDC-27 antibody in C phase lysate; data not shown). DISCUSSION The transition from M to C phase involves a major reorganization of microtubules and a significant change in their dynamics (50). They change again in the subsequent transition from C phase to interphase when loss of midzone proteins due to ubiquitination by CDH1/APC/C and proteolysis may be important (51). Our analysis of microtubule-binding proteins (Table I) showed that several mitotic spindle proteins lose affinity for microtubules as the cell enters C phase, and several known midzone and ϩtip proteins gain affinity. However, we observed much larger ratios for C phase-selective Aurora-B binding partners, suggesting that differential microtubule binding may be of a more graded nature, whereas Aurora-B partners change in a more switchlike manner.
Our analysis of protein phosphorylation in total clarified lysate and in microtubule-binding proteins revealed many more phosphopeptides that were M phase-selective than C phase-selective. This observation is consistent with two nonexclusive possibilities. (i) C phase is regulated by lack of CDK1 activity, and/or (ii) new phosphorylation sites are less important for C phase than for M phase. This possibility also emphasizes the dramatic activity of phosphatases during mitotic exit. Phosphopeptide analysis of microtubule-binding proteins nevertheless revealed many new sites in C phase, and ϳ17% of these had sequences consistent with their being Aurora-B substrates. Because several microtubulebinding proteins also interact with Aurora-B and VX680 inhibits microtubule binding of several proteins, it is likely that some of these new C phase-selective sites are phosphorylated by Aurora-B. Overall, our data point to less global phosphorylation in C phase than M phase but suggest an important role for Aurora-B phosphorylation in positively regulating microtubule binding of many proteins, including several important midzone proteins. These data together with previous studies are illustrated in Fig. 6 (3-5, 7, 8).
Aurora-B is known to regulate various aspects of C phase, but the extent of its regulatory influence has been previously uncharted. This is the first time, to our knowledge, that chemical genetics and quantitative mass spectrometry have been used together to demonstrate the importance of kinase activities on mitotic exit and cytokinesis. Large scale proteomics analyses are crucial in revealing this broad effect of Aurora kinase activity. Blocking Aurora kinase activity with the potent inhibitor VX680 strongly reduced the microtubule binding efficiency of many microtubule-binding proteins. Only a small group of kinetochore proteins strongly gained affinity. The phosphorylation site analysis of C phase microtubule-binding proteins and the gain of Aurora-B interactors in C phase implicate Aurora-B as a central regulator of the microtubule organization in C phase. Association of activated myosin with the cytoskeleton was also strongly reduced by VX680, suggesting that Aurora-B might directly regulate this central component of the contractile ring. The best characterized myosin regulatory phosphorylation, Ser-19 on the regulatory light chain, was not affected, suggesting that some other site(s) may govern myosin II regulation by Aurora-B. Intermediate filaments also undergo dramatic rearrangements during cytokinesis, and local phosphorylation in the cleavage furrow by Aurora-B was already implicated in rearrangement of Vimentin (32). Our data extend this idea to Keratin-8/18, which is a dominant intermediate filament class in dividing epithelial cancer cells (52,53). Taken together, our data show that Aurora-B influences the intermediate filaments, microtubule, and actin cytoskeleton elements during C phase.
A challenge in understanding the role of Aurora-B is that the kinase itself is active in both M phase and C phase, so changes in its function are likely mediated by changes in interaction partners and/or subcellular localization. Dramatic spatial regulation of Aurora-B activity in cytokinesis was shown using fluorescence resonance energy transfer report- ers for its activity (54). Our SILAC analysis of Aurora-B-interacting proteins reveals a similarly dramatic acquisition of new binding partners in C phase. In M phase, we mainly observed DNA-binding proteins binding to Aurora-B complex, whereas in C phase, microtubule-binding proteins and other cytoskeletal proteins were very prominent in the interaction list. C phase-selective Aurora-B binding partners include several proteins with known roles in cytokinesis as well as novel interactions, notably KIF4, APC/C, and FHOD-1. Overall, our data suggest that a switch in multiple Aurora-B partners is a major factor in regulating the biochemical transition from M to C phase.
The APC/C is known to play an important, but still poorly understood, role in cytokinesis (55), and proteosome-mediated proteolysis is thought to govern the orderly progression from mitosis through C phase to G 1 (51). We observed C phase-selective interaction between Aurora-B and APC/C, and both Aurora-B and CDC-27 localized to the midzone during the late cytokinesis stage. Consistent with these data, a recent study showed that CDH1/APC/C helps target Aurora-B kinase to the midzone (56). All these data support the idea that APC/C interactions with Aurora-B play a major regulatory role in C phase, perhaps especially in the midzone. We speculate that this interaction helps to target APC/C to relevant substrates as cytokinesis progresses, including the Aurora-B complex itself.
In summary, our analysis combines quantitative mass spectrometry with chemical genetics to provide the first broad overview of C phase biochemistry and reveals a large scale change of binding partners for microtubules and Aurora-B as cells enter C phase. It also provides evidence that Aurora-B should be considered one of the master regulatory kinases for C phase, at least for the microtubule cytoskeleton, because its activity is required to promote microtubule binding of many proteins during C phase. The mass spectrometry-based analyses described here provide a unique view of C phase that is complementary to using a single gene/protein perspective that is more traditionally used. Detailed analysis of any one of these interactions will yield more insights into local biochemistry, but proteomics methods are best suited to provide a global overview, and thus conceptual understanding, of how the cytoplasm is rearranged in the last step of the cell cycle.