Dynactin p27 is phosphorylated by Cdk1 and binds Plk1

Plk1 binding commonly requires priming phosphorylation by Cdk1. In search of regulatory mechanisms that might involve p27 or p25 phosphorylation, we searched their amino-acid sequences for predicted phosphorylation motifs. The human p27 sequence contains a single consensus Cdk1 phosphorylation site, Thr186, that is conserved among vertebrates (Figure 3A). Phosphorylation at this site creates a consensus Plk1 binding site (Elia et al, 2003). Although p25 is somewhat related to p27, no p25 species we examined contained a Cdk1 consensus sequence or a predicted polo-binding motif. To verify that p27 could serve as a substrate for Cdk1, we performed in vitro phosphorylation using purified bovine and embryonic chick brain dynactin (Figure 3B). p27 was found to be a major target, along with p150^Glued. We then used point mutagenesis to verify that p27 Thr186 was the Cdk1 phosphorylation site. For this work, we utilized purified, recombinant p27/p25 heterodimers to mimic p27 in its native context. Wild-type p27, along with two variants, T186A and S184A (a MAPKKK consensus site), prepared as dimers with p25, were evaluated as Cdk1 substrates in vitro. Both wild-type p27 and the S184A variant were phosphorylated, but the T186A variant was not (Figure 3C). The p25 component of the p27/p25 heterodimer was not phosphorylated.

Open in a separate window

Figure 3

p27 phosphorylation and Plk1 binding. (A) Location in mammalian p27 of the predicted Cdk1 phosphorylation site (T186, red; Group-based Prediction System; GPS Version 2.1, http://gps.biocuckoo.org/) and the Plk1 binding motif (red underline). (B) In vitro phosphorylation of native bovine and chick embryo brain dynactin. The autoradiograms (left four lanes) show samples incubated in the absence (−) or presence (+) of purified Cdk1. Arrowheads indicate p27; asterisks mark cyclin B, which is autophosphorylated. The lane on the right shows bovine dynactin (Coomassie blue stain). (C) In vitro phosphorylation by Cdk1 of bacterially expressed GST-p27/p25His₆ complexes (wild-type p27 and the T186A and S184A variants) and a GST control. Left: Coomassie blue-stained gels, right: the corresponding autoradiograms. (D) In vivo phosphorylation of endogenous p27. Lysates prepared from Cos7 cells synchronized in G₁ (17h after double thymidine block) or M (via nocodazole arrest and shake off) were subjected to two-dimensional immunoblotting for p27. (E) In vivo phosphorylation of exogenously expressed wild-type and T186A EGFP-p27 was evaluated as in (D). (F) In vitro binding of purified p27/p25 heterodimers to GST-tagged wild type PBD (wt) and a non-binding variant (mut) (Elia et al, 2003). Proteins immobilized on glutathione beads were incubated with purified p27/p25 heterodimers that had been phosphorylated in vitro using purified Cdk1. Phosphorylation was verified by ProQ staining. +: phosphorylated p27; −: unphosphorylated p27.

Source data for this figure is available on the online supplementary information page.

Source Data for Figure 3B^{(149K, jpg)}

Source Data for Figure 3C^{(3.3M, zip)}

Source Data for Figure 3D-G1^{(328K, jpg)}

Source Data for Figure 3D - mitosis^{(330K, jpg)}

Source Data for Figure 3E^{(25M, zip)}

Source Data for Figure 3F^{(5.0M, zip)}

To evaluate p27 phosphorylation in vivo, we performed two-dimensional immunoblotting on lysates of Cos7 cells synchronized in G₁ or M phase (Figure 3D). p27 was seen to be phosphorylated in M phase, but not in G₁. When expressed in cells, E-GFP-tagged wild-type p27 also underwent mitotic phosphorylation, whereas the T186A mutant did not (Figure 3E). Taken together, these results verify that p27 is mitotically phosphorylated at Thr186 in vivo.

Phospho-Thr186 is part of a consensus binding site for the master regulator, Plk1, that acts broadly during the process of cell division, from mitotic entry through cytokinesis. It is found at multiple sites in the spindle, including spindle poles, kinetochores, the spindle midzone and the intracellular cytokinetic bridge, all sites where dynactin and dynein also accumulate. Plk1 is recruited to these sites via a number of different binding partners (Petronczki et al, 2008). Dynactin, more specifically its p27 subunit, may be a previously unidentified Plk1 targeting factor. To demonstrate that p27 can bind Plk1 directly, we performed in vitro experiments using GST-tagged polo-binding domain (PBD; AA 365–603 of human Plk1) and purified recombinant p27/p25 heterodimers that had been phosphorylated by purified Cdk1 in vitro (Figure 3F). Phospho-p27 bound to immobilized wild-type PBD but not a PBD variant that is attenuated for binding (Elia et al, 2003). This interaction was dependent on Cdk1 phosphorylation of p27 because minimal binding was observed to the non-phosphorylated control.

The crystal structure of dynactin p27

To gain further insight into structure–function relationships in the p27/p25 heterodimer and to elucidate the structural context of the p27 phosphorylation site, we embarked on a crystallographic study of these proteins. Amino-acid sequence-based analysis of p27 and p25 using 3D-Jury (Ginalski et al, 2003; Kajan and Rychlewski, 2007) suggested that both proteins contain an unusual left-handed parallel β-helix (LβH) domain, in agreement with earlier hypotheses (Parisi et al, 2004). Co-expression of the two proteins in E. coli using a bi-cistronic vector resulted in a stable, soluble complex that was used in the in vitro phosphorylation and PBD binding experiments described above. Unfortunately, the complex did not yield crystals suitable for structural characterization despite extensive effort. Interestingly, recombinant isolated human p27, but not p25 (which is unstable in the absence of p27), could also be expressed in soluble form in E. coli. Starting with this protein, we obtained a good-quality atomic model using diffraction data to 2.2Å resolution (see Supplementary data, Materials and methods; Table I). The asymmetric unit contains two independent monomers, each with interpretable electron density for residues Ser8 to Thr70 and Lys80 to Gln159, accounting for 74% of the polypeptide chain.

Table 1

Crystallographic data

Open in a separate window

^aThe numbers in parentheses describe the relevant value for the last resolution shell.

^bR_sym=∑|I_i−I|/∑I, where I_i is the intensity of the ith observation, and I is the mean intensity of the reflections.

^cR_cryst=∑||F_obs|−|F_calc||/∑|F_obs|, crystallographic R factor, and R_free=∑||F_obs|−|F_calc||/∑|F_obs|, where all reflections belong to a test set of randomly selected data.

The ordered domain of p27 visualized by the crystal structure (Figure 4) shows the predicted LβH tertiary fold, making it the first vertebrate protein experimentally shown to display this architecture. Like its right-handed counterpart, this fold contains a β-helical coil formed by repeating units of three β-strands (‘rungs’) separated by turns or loops where the polypeptide chain takes a 300° turn (Choi et al, 2008). The rungs pack against each other by main-chain to main-chain hydrogen bonds parallel to the helical axis. The p27 LβH domain is a type I β-helix, containing seven rungs, with six residues per strand (i.e., 18 residues per rung). The first four amino acids in each rung display the canonical β-sheet secondary structure, with the side chains of the first and third pointing into the interior of the protein, and the side chains of the second and fourth pointing into the solvent. The fifth and sixth residues in each rung make up the turn. Thus, the molecule has a distinctly triangular cross section, with three faces (A, B and C) separated by edges (Figure 4A, referred to as ‘Turns’ in Figure 4C).

Open in a separate window

Figure 4

The structure of the LβH domain of human p27 (PDB accession number: 3TV0). (A) A view of the crystallographic model of the LβH domain looking down the main axis of the β-helix, from the N-terminus towards the C-terminus. The unique features of the β-helical fold, and residues in the key position 2, are shown in detail using different colours. (B) A view looking directly at Face B of the LβH. The molecular surface is coloured according to element (C: white, N: blue, O: red, and S: yellow). (C) Structure-based amino-acid sequence alignment of the seven ‘rungs’ of the p27 β-helix, coloured to match the structural model in (A). Amino acids that face the interior of the β-helix and thus contribute to the hydrophobic core are indicated in bold. The grey shading indicates the hydrophobic amino acids in Face B (the central ridge in B) that may contribute to the dimer interface.

The three faces of the p27 LβH domain are characterized by different types of exposed amino acids. In each rung, the second residue is positioned in the centre and thus contributes significantly to the physicochemical properties of that respective face. Faces A and C show a variety of amino acids in position 2, but face B has exclusively hydrophobic amino acids at this position (Figure 4C). This is likely to have functional significance, as discussed below. The edge between the A and B faces (Turn 1) is made up of regular, two-residue long turns, as is the edge between the B and C faces (Turn 2). The turns that make up the edge between the C and A faces (Turn 3) are irregular and include two longer loops, one comprising residues 43–47 and a longer one comprising residues 64–81, most of which are not resolved in the electron density map.

Beyond Gln159, the last residue visible in the electron density, secondary structure prediction suggests the presence of a short disordered sequence, a region with high propensity to form an α-helix (residues 170–185; a similar α-helix is predicted in p25), and a disordered C-terminus. The Cdk1 phosphorylation site, Thr186, is the first residue in the disordered segment, suggesting high accessibility to kinase.

A model for p25 structure and mechanism of heterodimerization

The shared predicted LβH tertiary fold of p27 and p25 allowed us to propose a model of the p25 LβH domain based on the p27 crystal structure. The model (Figure 5) shows that the C face of the p25 LβH domain also contains a ridge of exposed hydrophobic residues, making this face structurally similar to the B face of p27. This raises the possibility that the p27 and p25 proteins utilize their respective B and C faces to interact within the heterodimer. This hypothesis is corroborated by the fact that within the asymmetric unit of the p27 crystal, two independent monomers form an intimate homotypic interface between the two B faces, burying ~1300Ų of solvent exposed surface (Figure 5A). In this geometry (irrespective of orientation), the loops in Turn 3 of both proteins are near each other, raising the possibility that this part of the structure contributes to interactions with other dynactin components or extrinsic binding partners.

Open in a separate window

Figure 5

A proposed dimerization mode for p27 and p25. (A) The crystallographic p27 homodimer, visualized perpendicular to the dimerization axis with the residues of the hydrophobic ridge on the B face highlighted. (B) Structure-based amino-acid sequence alignment of the seven ‘rungs’ of p25. β-helix faces are coloured as in Figure 4. The grey shading indicates the hydrophobic amino acids in Face C that are predicted to form the dimer interface. Amino acids that face the interior of the β-helix and thus contribute to the hydrophobic core are indicated in bold. (C) A model of the physiological p27/p25 heterodimer. The hydrophobic amino acids exposed on the surfaces of Faces B and C of p27 and p25 (respectively) are depicted with red spheres. The locations of the two edges that contain extended loops (between faces A and C) are also indicated.

Cdk1-phosphorylated p27 targets Plk1 to kinetochores

Our imaging analysis indicates that loss of the p27/p25 heterodimer from kinetochore-associated dynactin allows cells to enter anaphase more quickly without altering steady-state levels of many kinetochore-associated molecules. To explore the possibility that the accelerated mitotic progression might be due to impaired Plk1 targeting to kinetochores, we examined Plk1 localization in mitotic cells. The amount of Plk1 present at spindle poles and in intracellular cytokinetic bridges was unchanged in cells depleted of any of the dynactin subunits (Figure 6A–C) indicating that Plk1 recruitment to these sites depends primarily upon binding partners other than the dynactin p27 component. However, cells from which kinetochore dynactin was removed completely (by p150^Glued or Arp11 depletion) or stripped of p27/p25 (by p27 depletion) showed a significant reduction in Plk1 at prometaphase kinetochores (Figure 6D and F). These results indicate that dynactin, specifically its p27 and p25 subunits, is required for full Plk1 targeting to kinetochores. The residual Plk1 present at kinetochores in p27/p25-depleted cells is most likely recruited via other known polo-binding proteins such as Bub1, BubR1, PBIP1 and/or NudC, none of which are reduced as a result of p27/p25 depletion (Figure 2C; Supplementary Figure 4).

Open in a separate window

Figure 6

Analysis of Plk1 localization to mitotic structures in cells depleted of dynactin subunits. (A) Plk1 fluorescence intensities in cytokinetic bridges (B) and at spindle poles (C) of cells co-stained for tubulin (MT) were quantified, normalized to microtubule pixel values in the same area and expressed as per cent of controls (mean±s.e.m., n>150 cells). The loss of midbody density seen in (B) is commonly seen in cells depleted of dynactin subunits. Bar=5μm. (D through K) Kinetochore localization of Plk1 and 3F3/2. (D, E) Representative images from asynchronous Cos7 populations fixed and stained for Plk1 (green, left panels) or 3F3/2 (green, right panels) plus anti-centromere antigen (ACA; red). Representative prometaphase cells are shown here. (E) Cells stained after brief nocodazole treatment are shown. Bar=5μm. The insets (3 ×) show merged images of Plk1 or 3F3/2 with ACA. (F through K) Quantitative analysis of Plk1 (F, H, J) and 3F3/2 (G, I, K) fluorescence intensity at kinetochores, normalized to ACA pixel values (mean±s.e.m., n>600 kinetochores, 15 cells). (F, G) Cells transfected with different dynactin subunit siRNAs in the presence (+) or absence (−) of the wt p27 rescue vector. (H, I) Cells stained after brief nocodazole treatment. (J, K) Cells cotransfected with control or p27 siRNAs and the rescue vectors indicated (vector=pCAGIG lacking an insert).

Source data for this figure is available on the online supplementary information page.

Source Data for Figure 6A^{(85K, xls)}

Source Data for Figure 6F, 6G, 6H, 6I^{(58K, xls)}

Source Data for Figure 6J and 6K^{(35K, xls)}

To verify that dynactin-mediated recruitment of Plk1 to kinetochores was functionally relevant, we used the 3F3/2 antibody to assess the levels of phosphorylated Plk1 substrates. This antibody recognizes a phosphoepitope generated by Plk1 phosphorylation on a number of spindle-associated proteins (Ahonen et al, 2005; Wong and Fang, 2005). Kinetochore labelling with 3F3/2 has been shown to be reduced when Plk1 is depleted or inhibited (Ahonen et al, 2005; Wong and Fang, 2005). 3F3/2 labelling was significantly reduced at kinetochores of cells treated with dynactin subunit siRNAs, verifying that Plk1 activity is attenuated (Figure 6D and G). Depletion of Spindly, which causes both dynein and dynactin to be lost from kinetochores (Griffis et al, 2007; Chan et al, 2009; Barisic et al, 2010; Gassmann et al, 2010) and thus removes p27, also resulted in dramatically reduced Plk1 binding to prometaphase kinetochores (Supplementary Figure 5). The reduction in Plk1 and 3F3/2 staining seen in cells depleted of dynactin subunits reflected the intrinsic composition of the kinetochore because similar results were observed in nocodazole-treated cells in which MT-dependent tension has been eliminated (Figure 6E, H and I).

These findings provide strong evidence that dynactin targets Plk1 to mitotic kinetochores via phospho-Thr186 in the p27/p25 heterodimer. As a final test of this hypothesis, we evaluated whether the non-phosphorylatable p27 variant, T186A, expressed at roughly the same level as endogenous p27 so it incorporates fully into dynactin-like wild-type p27 (Supplementary Figure 1), could rescue the mitotic phenotypes seen in cells depleted of p27 and p25. Wild-type p27, but not the T186A variant, fully rescued the mitotic timing defects (Figure 1A, B and F; Supplementary Movies 4 and 10). It also restored Plk1 and 3F3/2 at kinetochores to normal levels (Figure 6F and G), whereas the T186A variant did not (Figure 6J and K; Supplementary Figure 6). In cells depleted of Arp11 or p150^Glued, expression of wild-type p27 did not restore kinetochore Plk1 or 3F3/2 to normal levels (Figure 6F and G), as expected given that p27/p25 is not recruited to kinetochores unless they are incorporated into dynactin (Supplementary Figure 3). Expression of neither wild-type p27 nor the T186A variant had a discernable impact on mitotic timing or Plk1 targeting in cells transfected with control siRNAs (Supplementary Figures 6 and 7), verifying that low level expression of these proteins in cells containing endogenous dynactin has no effect.

Kinetochore–MT attachment, but not kinetochore spacing, is perturbed when p27 and p25 are removed from dynactin

Kinetochore-associated Plk1 has been implicated in the establishment of MT–kinetochore attachments (Sumara et al, 2004; Lenart et al, 2007; Petronczki et al, 2008). To determine whether MT–kinetochore interactions were perturbed in cells depleted of dynactin subunits, we treated cells with monastrol to generate monopolar spindles. These structures allow intimate end-on attachments to be distinguished from lateral interactions and unattached kinetochores (Kapoor et al, 2000; Lenart et al, 2007). Under normal conditions, this assay reveals that >90% of kinetochores sit at the ends of bundles of MTs that emanate from the pole in an orientation where one sister faces the pole the other sister faces away and no MTs appear to run between them (i.e., mono-oriented k-fibres; Figure 7A and B). Mono-oriented k-fibres are seen in only about 60% of cells depleted of p150^Glued and we see an increased number of cases where tubulin bundles run between sisters (lateral k-fibres), consistent with an increase in merotelic attachments. In cells depleted of p27/p25 we also observed fewer mono-oriented attachments and more lateral associations. This is as expected given that these cells show an increased number of uncongressed and lagging chromosomes (Figure 1E) and supports the idea that p27/p25-based recruitment of Plk1 contributes to proper establishment of kinetochore–MT interactions.

Open in a separate window

Figure 7

Effect of dynactin subunit depletion on kinetochore behaviour. (A) Representative images of monopolar spindles generated by monastrol treatment of asynchronous Cos7 cells stained with DAPI plus Abs to tubulin and ACA. Structures were observed using deconvolution microscopy. These images are maximal projections of three z-planes but kinetochore–microtubule associations were assessed by examining the entire deconvolved image. The insets provide images of selected kinetochore pairs at higher magnification to illustrate the different interactions seen. Bar=5μm. (B) Images as in (A) were scored for the types of kinetochore–microtubule interaction observed (end-on=mono-oriented k-fibres; lateral=lateral k-fibres; mean±s.d., n>100 kinetochore pairs per condition). (C) The distances between sister kinetochores on bi-oriented chromosomes were measured in asynchronous Cos7 cells stained with DAPI plus Abs to tubulin and ACA (mean±s.e.m.; *P<0.01 compared to control; N=number of kinetochores). Taxol treatment (1μM, 12min) was used to provide a control for the amount of tension expected at unattached kinetochores. (D, E) Analysis of kinetochore Mad1 in cells treated with dynactin subunit siRNAs. (D) Representative images of Mad1 localization at prometaphase kinetochores in asynchronous cell populations. Bar=5μm. (E) Quantitation of Mad1 levels at kinetochores of unaligned (i.e., prometaphase) and aligned (i.e., metaphase) chromosomes (judged by DAPI staining). Mad1 fluorescence intensity was normalized to ACA pixel values (mean±s.e.m., n>600 kinetochores, 15 cells) and expressed as per cent of the intensity at kinetochores of unaligned chromosomes in control cells. (F, G) As in (D) and (E), but in cells subjected to nocodazole treatment. Bar=5μm.

Source data for this figure is available on the online supplementary information page.

Source Data for Figure 7B^{(31K, xls)}

Source Data for Figure 7C^{(66K, xls)}

Source Data for Figure 7E and 7G^{(35K, xls)}

The spindle assembly checkpoint has been proposed to monitor tension between and within kinetochores to ensure that chromosomes are properly bi-oriented before anaphase is initiated. Kinetochores from which dynactin or dynein has been removed exhibit decreased tension between sister kinetochore pairs, as evidenced by closer interkinetochore spacing (Varma et al, 2008; Maresca and Salmon, 2009). Depletion of dynactin from kinetochores via Arp11 or p150^Glued siRNA treatment also reduced interkinetochore spacing in bi-oriented pairs (Figure 7C). This is most likely due to the absence of dynein, which is believed to exert poleward pulling forces at kinetochores (Varma et al, 2008). Depletion of p27/p25, by contrast, does not result in a loss of tension, as expected given that kinetochores in these cells still have normal steady-state levels of dynein. This suggests that inhibition of Plk1 binding to dynactin at kinetochores does not alter the tension generating mechanism, but instead interferes with the generation of and/or prematurely terminates the ‘wait-anaphase’ signal.

Antibodies

Commercial antibodies are α-tubulin (DM1A), phospho-histone 3 (serine 10): Sigma; p150^Glued (clone 1): BD, DIC clone 74.1: Millipore; p25 and p27: ProteinTech Group; Plk1: Santa Cruz; GFP: Invitrogen; PBIP1: Rockland; NudC: Abcam; GST: GE, Alexa488-, Alexa568-, and Alexa633-conjugated secondary antibodies: Invitrogen; alkaline phosphatase-conjugated secondary antibodies: Tropix. Monoclonal antibodies against p27 (27A) and Arp1 (45A) were produced and purified as described previously (Schafer et al, 1994; Melkonian et al, 2007). Other antibodies were obtained from our colleagues: mouse 3F3/2 ascites (Campbell and Gorbsky, 1995): Gary Gorbsky; rat anti-ZW10 3.2 (Chan et al, 2000): Gordon Chan, sheep anti-Bub1 (SB1.3), BubR1 (SBR1.1), Bub3 (SB3.2), MPS1 (SMP1.1) (Taylor et al, 2001; Tighe et al, 2008; Gurden et al, 2010): Stephen Taylor; rabbit anti-Spindly (Griffis et al, 2007): Eric Griffis; mouse anti-Mad1 (BB3–8) (De Antoni et al, 2005): Andrea Musacchio; human anti-centromere serum (ACA; Martin et al, 1990): JB Rattner.

Dynactin purification

Chicken embryo and bovine brain dynactin were purified as described previously (Schroer and Sheetz, 1991; Bingham et al, 1998).

Expression and purification of p25/p27 heterodimers

Full-length human p27 and full-length mouse p25 cDNA (ATCC) were amplified and cloned into pDEST15 to create the bi-cistronic expression vector pDEST-GST-p27/p25-His₆, with an rTEV cleavage site downstream of GST and a ribosomal binding site (RBS) between the p27 and p25 coding sequences. The non-cleavable C-terminal His₆ tag was inserted for purification purposes. The T186A (ACT to GCT) and S184A (AGC to GCC) p27 variants were created in pDEST-GST-p27/p25-His₆ via QuikChange. BL21-CodonPlus (DE3)-RIL cells (Agilent) were transformed with the plasmid and grown to an OD₆₀₀ of 0.6 in 0.5l LB at 37°C, then induced with 0.1mM IPTG for 16h at 18°C. For purification of GST-tagged heterodimers, the bacterial pellets were resuspended in 40ml PBS (pH 7.4) containing 1mM DTT and 1mg/ml lysozyme, rotated for 20min at 25°C and sonicated. The clarified bacterial supernatants were applied to a GSTtrap FPLC column (GE) and eluted with 50mM Tris–HCl, 100mM NaCl, 10mM L-glutathione and 1mM DTT, pH 8.0.

GST-PBD binding assays

Preparation of GST-PBD beads: The human Plk1 PBD (AA 365–603; accession #BC01486) from a pCMV-SPORT6 clone (ATCC) was cloned into pGEX4T-2. The PBD binding-null variant, GST-PBD H538A/K540M (Elia et al, 2003), was generated using engineered PCR primers that also contained a unique restriction site (BsaBI) for insertion into the parent vector. The constructs were transformed into BL21-CodonPlus (DE3)-RIL, grown to an OD₆₀₀ of 0.6 at 37°C in 0.5l LB and induced with 0.1mM IPTG for 3h at 30°C. The bacterial pellet was resuspended in 40ml binding buffer (20mM Tris–HCl, 250mM NaCl, 1mM EDTA, 1mM NaVO₄, 20mM NaF, 5mM β-ME, pH 8.0) containing 1mg/ml lysozyme, 5U/ml Benzonase and protease inhibitors, subjected to a freeze/thaw cycle, and sonicated. Two milliliters of the post-sonication supernatant was incubated with Glutathione-Sepharose (for each 10μl aliquot; GE) pre-blocked with 5% non-fat milk in binding buffer, washed 5 × with 500μl binding buffer, then test proteins were added.

To prepare p27/p25 heterodimers for the binding assay, the GST tag was cleaved via TEV protease (Invitrogen) for 16h at 4°C from samples that had been phosphorylated in vitro (see below). The p27/p25 heterodimers were separated from GST and TEV protease using a Superose12 FPLC column attached to a GSTrap column (GE) in 20mM Tris–HCl, 250mM NaCl, 1mM EDTA, 1mM NaVO₄, 20mM NaF, 5mM β-ME, pH 8.0. The peak fractions containing p27/p25 heterodimers were identified by Coomassie staining. The purified p27/p25 heterodimers (25μg) were incubated with 10μl Glutathione-Sepharose carrying GST-PBD for 1h at 4°C. The beads were washed with 3 × 500μl binding buffer and proteins were eluted in 500μl binding buffer plus 30mM L-glutathione for 1h at 4°C prior to analysis of 10μl by immunoblotting.

In vitro phosphorylation

To detect Cdk1 phosphorylation, purified dynactin (27nM) or recombinant GST-p27/p25-His₆ heterodimers (0.57μM) were incubated with 10 units of Cdk1 (0.01μg, NEB), 100μM ATP and 3.75 μCi γ-³²P ATP in Cdk1 buffer in a 15-μl reaction volume for 30min at 30°C. The entire sample was subjected to SDS–PAGE and autoradiography.

To generate phosphorylated proteins for GST-PBD binding, 350μg purified GST-p27/p25-His₆ heterodimers were incubated with 20 units of Cdk1 (NEB) plus 500μM ATP in a 1-ml volume for 16h at 30°C. A parallel sample incubated without Cdk1 was used as the non-phosphorylated control. Phosphorylation was verified using Pro-Q Diamond Stain (Invitrogen).

Live-cell imaging

To image mitosis in cells expressing histone 2B-mCherry, cells were harvested by trypsinization 24h after transfection with siRNAs, histone 2B-mCherry and rescue plasmids, then 1.5 × 10⁵ cells were plated into glass bottom dishes (MatTek) and cultured in HEPES-buffered DMEM for 24h prior to synchronization using a double thymidine block. Live-cell recordings were initiated at 9–10h after release from the block (G₁/S boundary). Cells were imaged on a Zeiss Axiovert 200M inverted microscope (40 × /0.85 NA Plan-Neofluor air objective) heated to 37°C using an In Vivo Scientific incubator. Images were collected using a Cooke SensiCam camera. Samples were illuminated at 25% intensity via a 120W X-Cite 120PC Xenon arc lamp (EXFO Life Science). To reduce intensity and prevent radiation-induced changes in the cell cycle (Khodjakov and Rieder, 1999), the light beam was passed through GG400 (UV) and KG5 (IR) filters (Newport Corporation) as well as a 12% neutral density filter (Zeiss). This filter combination allowed cells to proceed through mitosis without discernable delays. Images (100ms/frame) were collected at 120-s intervals. Three z planes (0.5μm step size) were collected at each time point and at the end of the imaging interval the best optical plane for each cell was chosen and used to generate movies in SlideBook.

For phase contrast imaging, RNAi-treated cells synchronized as above were illuminated using a 12V/100W halogen lamp at 30–35% intensity with heat absorbing (KG1) and green interference filters (546±20nm bandpass, Zeiss) (Wadsworth et al, 2005). Time-lapse movies were acquired at 90-s intervals on a Zeiss Axiovert 200 Microscope (LD Plan-Neuofluor 40 × , 0.6 NA Phase2 objective). Three z planes (0.5μm step size) were collected at each time point, and the best optical plane for each cell was used to generate movies.

We thank Ms Natalya Olekhnovich for technical assistance. We are grateful to Drs G Chan, G Gorbsky, E Griffis, JB Rattner, A Musacchio and S Taylor for generously providing antibodies, to Dr Q Vong and Y Zheng for the histone 2B-mCherry plasmid, and the two anonymous reviewers plus Dr T Stukenberg for input and comments on the manuscript. The use of the Advanced Photon Source was supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under contract No. W-31-109-Eng-38. We thank the SER-CAT staff for assistance with data collection. Supporting institutions of SER-CAT may be found at http://www.ser-cat.org/members.html. This research was supported by grants from the NIH to TAS (GM044589) and ZSD (NS036267) and training grant T32 GM07231 (Johns Hopkins University CMDB Program). Microscopy was performed at the Johns Hopkins Integrated Imaging Center.

Author contributions: TAS and ZSD supervised the experimental design and wrote the paper. TY performed the experiments in Figures 1, ​,2,2, 3D, E, ​,66 and ​and7,7, Supplementary Movies, and Supplementary Figures. BRS performed the experiments in Figure 3C and F. FKC performed the experiments in Figure 3B and C. AK, MYZ and UD performed the experiments in Figures 4 and ​and5.5. All the authors understand their responsibilities connected to authorship.