Protein phosphatase 4 catalytic subunit regulates Cdk1 activity and microtubule organization via NDEL1 dephosphorylation

The vertebrate centrosome is a highly organized organelle that serves as the cell microtubule (MT) organizing center, among other functions (Doxsey, 2001; Bornens, 2002). During interphase, MTs are organized in astral arrays that radiate from the centrosome and function as a scaffold to direct organelle and vesicle trafficking (Thyberg and Moskalewski, 1999). One particular dramatic change is loss of the extensive interphase MT array and the subsequent assembly of a bipolar mitotic spindle (Compton, 2000). Progress in understanding how centrosomal MT arrays are regulated has revealed that many proteins participate in the nucleation (γ-tubulin and pericentrin), anchoring (ninein, centriolin, dynactin, XMAP215, and TACCs), and release (katanin and XKCM1/mitotic centromere-associated kinesin) of MTs from the centrosome (Walczak et al., 1996; Doxsey, 2001; Bornens, 2002; Kinoshita et al., 2002; Blagden and Glover, 2003). MTs and these accessory components are also critically regulated by mitotic kinases, including Cdk1, the Polo family, the NIMA (never in mitosis A) family, and the Aurora family, upon entrance into mitosis (Nigg, 2001; Blagden and Glover, 2003). Whereas protein kinases regulate protein activities by phosphorylating key residues on the molecules, protein phosphatases counteract kinase activities by dephosphorylating those residues. Protein phosphatases, including Cdc14A, Cdc25C, PP1, and PP4, have been shown to associate with mitotic centrosomes (Ou and Rattner, 2004). Both PP1 and PP4 are members of the PPP family of protein serine/threonine phosphatases, which associate with the centrosome during mitosis (Brewis et al., 1993; Andreassen et al., 1998; Helps et al., 1998). The orchestrated modulation of centrosomal components by kinases and phosphatases plays an essential role for the maintenance of MT organization and spindle formation (Meraldi and Nigg, 2001).

Accumulating evidence suggests that centrosomal components and their kinases play critical roles in neurogenesis and neuronal migration (Wynshaw-Boris and Gambello, 2001; Tsai and Gleeson, 2005). LIS1 was identified as a gene mutated in isolated lissencephaly sequence (Reiner et al., 1993), which is a cerebral cortical malformation characterized by a smooth cerebral surface and a disorganized cortex caused by incomplete neuronal migration (Dobyns, 1989; Dobyns et al., 1993). LIS1 and its binding partner, NDEL1, are preferentially distributed at the centrosome (Sasaki et al., 2000) and regulate the cytoplasmic dynein heavy chain (Vallee, 1991; Vallee et al., 2001). Lis1- and Ndel1-disrupted mice displayed similar defects in neuronal migration (Hirotsune et al., 1998; Sasaki et al., 2005). Interestingly, NDEL1 is a known substrate of several kinases, including Cdk5/Cdk1, which are essential for regulation of a proper MT organization (Toyo-Oka et al., 2005). In addition, Cdk5/Cdk1-mediated phosphorylation of NDEL1 recruits katanin p60, which controls MT dynamics (McNally and Vale, 1993; McNally, 2000) at the centrosome and facilitates MT remodeling (Toyo-Oka et al., 2005). Recently, we also demonstrated that NDEL1 is phosphorylated by Aurora A kinase, which is essential for centrosomal maturation and separation (Mori et al., 2007). These observations led us to clarify the precise functions of NDEL1 in MT dynamics.

In this study, we report that NDEL1 is a substrate of the centrosomal phosphatase protein phosphatase 4 catalytic subunit (PP4c; Helps et al., 1998; Hu et al., 1998). PP4c efficiently dephosphorylates Cdk1 sites of NDEL1 but does not dephosphorylate the Aurora A site. We also found that PP4c negatively regulates Cdk1 activity in interphase. To understand the physiological role of PP4c in vivo, we generated PP4c-disrupted mice by Cre-loxP recombination. Mouse embryonic fibroblast (MEF) cells in which PP4c was deleted by Cre exhibited severe impairments of MT organization. Surprisingly, loss of PP4c led to an unscheduled activation of Cdk1 at interphase and an up-regulation of the T219 phosphorylation of NDEL1 in interphase, which is associated with an excessive accumulation of katanin p60 to the centrosome. These findings suggest that PP4c is required for proper organization of MTs at the centrosome through regulation of the phosphorylation of NDEL1 and recruitment of katanin p60.

To identify proteins interacting with NDEL1, we performed a yeast two-hybrid analysis using NDEL1 as bait and identified PP4c (Helps et al., 1998; Hu et al., 1998). We next examined the ability of PP4c to dephosphorylate a Cdk1 phosphorylation site, phospho-T219 (Toyo-Oka et al., 2005), and an Aurora A phosphorylation site, phospho-S251 (Mori et al., 2007), of NDEL1 using recombinant proteins as a substrate. NDEL1 was initially subjected to phosphorylation by GST-Cdk1 or GST–Aurora A (Mori et al., 2007), and phosphoproteins were purified before the dephosphorylation experiments. PP4c efficiently removed the phosphate from Cdk1 phosphorylation sites but not from the Aurora A phosphorylation site (Fig. 1 A). The dephosphorylation activity of PP4c was completely suppressed by okadaic acid. In addition, the PP4c inactive mutant, PP4c-RL, in which Arg236 was replaced with Leu (Zhou et al., 2002), did not display any dephosphorylation activity (Fig. 1 A). We also confirmed dephosphorylation by PP4c by Western blotting. PP4c treatment selectively diminished the signal of Western blotting by an antiphospho-T219 antibody (Fig. 1 A). These results suggested that at least one of the Cdk1 phosphorylation sites of NDEL1 is a specific substrate of PP4c.

We next tested the subcellular distribution of NDEL1, PP4c, and another binding protein of PP4c, R1 (Kloeker and Wadzinski, 1999). In interphase, PP4c was predominantly distributed at the centrosome and inside the nucleus, whereas the signal of PP4c disappeared from the centrosome during mitosis (Fig. 1 B). In contrast to the cell cycle–dependent distribution of PP4c, NDEL1 and R1 were stably localized at the centrosome regardless of the cell cycle (Fig. 1 B). Western blotting on synchronized HeLa cells indicated that PP4c protein levels were equivalent throughout the cell cycle (unpublished data). We next sought to characterize whether the presence of PP4c is related to the phosphorylation of cyclin B1 and NDEL1. We examined S126 phosphorylation of human cyclin B1, the site of the mitosis-specific phosphorylation of cyclin B1 by activated Cdk1 (Jackman et al., 2003) and T219 phosphorylation of NDEL1 (Toyo-Oka et al., 2005). Although immunocytochemistry showed virtually no phosphorylation of human cyclin B1 and T219 of NDEL1 in interphase, robust phosphorylation was seen after the mitotic entry. Conversely, PP4c localized to the centrosome in interphase but disappeared from the centrosome after mitotic entry (Fig. 1 C). These findings suggest that that PP4c normally may prevent activation of Cdk1 at the centrosome.

We next asked whether the persistent expression of PP4c might inhibit the activation of Cdk1. To achieve restricted expression of PP4c at the centrosome, we conjugated PP4c to a PACT domain (Gillingham and Munro, 2000). Prolonged centrosomal expression of PACT-PP4c clearly abolished cyclin B1 S126 phosphorylation and suppressed the separation of centrosomes in prophase (Fig. 1 D), whereas activation of Aurora A was still observed by monitoring T288 phosphorylation (not depicted). In contrast, the expression of a control PACT domain or PACT-PP4c-RL did not display this effect (Fig. 1 D). Phosphorylation of NDEL1 was also clearly suppressed by the centrosomal expression of PP4c (Fig. 1 D).

To examine the dephosphorylation activity of NDEL1 by PP4c in vivo, we coexpressed PACT-PP4c and PACT–mitotic Cdk1 (Litvak et al., 2004) and examined the phosphorylation of NDEL1 and human cyclin B1 in G2. The expression of PACT–mitotic Cdk1 led to the phosphorylation of cyclin B1 and NDEL1 (Fig. 2 A), whereas the coexpression of PACT–mitotic cyclin and PACT-PP4c clearly suppressed the phosphorylation of both proteins, suggesting that PP4c is capable of dephosphorylating cyclin B1 and NDEL1 in vivo and that PP4c might be functionally dominant over Cdk1. We previously reported that phosphorylation of NDEL1 by Cdk1 facilitates recruitment of katanin p60 at the centrosome (Toyo-Oka et al., 2005). Therefore, we examined whether the expression of mitotic Cdk1 in G2 facilitates the recruitment of katanin p60 at the centrosome. Interestingly, the expression of PACT–mitotic Cdk1 in G2 HeLa cells increased the centrosomal localization of katanin p60 and resulted in fewer MTs (Fig. 2 B and see Fig. 5 B).

To explore the in vivo role of PP4c and its function in the organization of MTs, we generated a mutant line, PP4c^neo/+ (Fig. S1, A–C; and see Materials and methods). We further generated PP4c^cko/+ and PP4c^−/+ by a Cre-mediated partial and complete recombination of PP4c^neo/+ mice, respectively (Fig. S1, A–C; and see Materials and methods). These three lines of heterozygous mutant mice were viable and fertile. We generated homozygous mice by the mating of each heterozygote and found that PP4c^cko/cko mice were viable and fertile, whereas PP4c^neo/neo mice and PP4c^−/− mice died in the mid and early embryonic stages, respectively. This is consistent with a previous study (Fig. S1, D–F; Shui et al., 2007).

To analyze the role of PP4c in MT regulation and cell proliferation, we established an MEF line. PP4c was inactivated by RFP-Cre–mediated recombination through transfection of a plasmid or adenovirus-mediated gene transfer carrying RFP-Cre (see each figure legend). Cre-mediated recombination efficiently removed part of the PP4c gene (PP4c^−/− MEF cells), resulting in the disruption of PP4c (Fig. 3 A). Disruption of PP4c in MEF cells resulted in the serious disorganization of MTs (Fig. 3 B), whereas the expression of RFP-Cre in PP4c^+/+ MEF cells did not display any obvious effect on the organization of MTs. We categorized the MT patterns into four groups: normal, in which MTs appeared normal with fiberlike staining and centrosomal focusing; type A, in which MTs were abundant but no obvious centrosomal focusing of the MTs was observed; type B, in which overall MT levels were reduced and remaining MTs appeared to be bundled; and type C, in which only weak diffuse staining and a few fragmented MTs were observed within the cytoplasm. Type C cells were more numerous in cells transfected for longer times (Fig. 3 B). Interestingly, in these cells with profoundly defective MT arrays, the signal of γ-tubulin was comparable with wild type (Fig. 3 A). These observations prompted us to investigate the stability of MTs using an antiacetylated α-tubulin antibody to visualize a stable tubulin (Fig. 3 C). The intensity of antiacetylated tubulin antibody staining was clearly decreased in PP4c^−/− MEF cells compared with PP4c^+/+ MEF cells, suggesting that MTs became unstable in PP4c^−/− MEF cells.

In the mouse genome, PP4c is located on chromosome 7 adjacent to Tbx6 (Chapman and Papaioannou, 1998). Inadvertently, the PGK-neo gene bracketed by loxP was inserted into the last exon of Tbx6, and the last four amino acids were replaced by 15–32 different amino acids, which may influence the function of Tbx6 (Fig. S2, A and B). To exclude the possibility that MT defects might be attributed to impairment of the Tbx6 gene, we established MEF cells from a Tbx6^−/− embryo and found normal MT organization (Fig. S2 C). We also confirmed that the impaired organization of MTs was rescued by the exogenous expression of PP4c but not by the exogenous expression of Tbx6 (Fig. S2, D and E). Thus, we concluded that impairment of MT organization is attributable to the loss of PP4c rather than the mutation of Tbx6.

We previously reported that NDEL1 phosphorylation by Cdk1 recruits katanin p60 to the centrosome (Toyo-Oka et al., 2005). We hypothesized that the defect of MT organization in PP4c^−/− MEF cells might be attributed to excessive recruitment of katanin p60 to the centrosome, which is associated with the abnormal phosphorylation of NDEL1. To address this possibility, we first examined the phosphorylation of endogenous NDEL1 using known antiphospho-NDEL1 antibodies (Toyo-Oka et al., 2005; Mori et al., 2007). Interestingly, PP4c^−/− MEF cells displayed abnormal phosphorylation of NDEL1 at the T219 Cdk1 site (Fig. 4 A), whereas S251, the Aurora A site, was not phosphorylated (not depicted). Therefore, we sought to determine whether Cdk1 activity was increased in PP4c^−/− MEF cells. We raised an antibody against phospho-S123 of mouse cyclin B1, which corresponds to S126 of human cyclin B (Jackman et al., 2003), as an indicator of activated Cdk1 (Fig. S3). Surprisingly, Cdk1 was frequently activated in PP4c^−/− MEF cells as determined by monitoring the mitotic phosphorylation of S123, whereas none of these phosphorylations were observed in interphase of control cells (Fig. 4 B). The activation of Cdk1 in PP4c^−/− MEF cells was also supported by the enzymatic activity of Cdk1 (Fig. 4 C). GFP-PACT-Cdk1 extracted from PP4c^−/− MEF cells revealed higher phosphorylation activity compared with control MEF cells, suggesting that Cdk1 is activated in PP4c^−/− MEF cells. Collectively, our data support the notion that PP4c is a negative regulator of Cdk1.

Phospho-S123 cyclin B1–positive cells in PP4c^−/− MEF cells rarely displayed chromosome condensation. Although the activation of Cdk1 could occur in interphase, mitotic arrest by activation of the G2/M checkpoint instead of the unscheduled activation of Cdk1 is another possibility. To address this issue, we first examined the cell cycle pattern of PP4c^−/− MEF cells by flow cytometry. A higher 4c population appeared in PP4c^−/− MEF cells, suggesting that PP4c^−/− MEF cells were arrested in G2/M (Fig. 4 D). We next asked whether histone H3, a marker of prophase, was phosphorylated in PP4c^−/− MEF cells (Fig. 4 D). In PP4c^−/− MEF cells, phosphorylation of histone H3 was not observed, indicating that PP4c^−/− MEF cells did not enter into prophase. Furthermore, activation of Cdk1 in PP4c^−/− MEF cells was observed in MEF cells in which cell cycle was arrested by mitomycin C treatment (Fig. S4, A–C) or contact inhibition and serum starvation (Fig. S4 D). Thus, we conclude that Cdk1 is abnormally activated at interphase in PP4c^−/− MEF cells.

The aberrant MT organization in PP4c^−/− MEF cells is reminiscent of the disruption of MTs caused by the overexpression of katanin p60 (Fig. 5 A). Therefore, we investigated whether katanin p60 was excessively recruited to the centrosome. Immunostaining revealed that higher amounts of katanin p60 were accumulated at the centrosome in PP4c^−/− MEF cells (Fig. 5 B). Quantitative analysis of the fluorescence intensity demonstrated a twofold increase of katanin p60 at the centrosome in PP4c^−/− MEF cells compared with PP4c^cko/cko MEF cells (Fig. 5 B). As an independent approach to confirm the phosphorylation of NDEL1 and restricted distribution of katanin p60 with centrosomes, we isolated centrosomes from MEF cells and fractionated the protein lysate. Immunoblot analysis of fractions from PP4c^−/− MEF cells revealed that phosphorylated NDEL1 was indeed present with γ-tubulin, which was cofractionated with katanin p60 (Fig. 5 C; Moudjou and Bornens, 1998). The total amount of katanin p60 was not changed in PP4c^−/− MEF cells; however, katanin p60 displayed a more restricted distribution compared with PP4c^cko/cko MEF cells. Our finding suggests that PP4c regulates the organization of MTs partly through regulation of the phosphorylation of NDEL1 and the distribution of katanin p60.

We demonstrated that the disruption of PP4c results in abnormal T219 phosphorylation of NDEL1 and centrosomal recruitment of katanin p60. Oligomerization of katanin p60 increased the affinity of katanin for MTs and stimulated its ATPase activity (McNally and Vale, 1993; Hartman et al., 1998). Therefore, excessive concentration of katanin p60 at the centrosome might enhance the severing of MTs. To evaluate the nucleation of MTs at the centrosome, we performed the MT regrowth assay after total depolymerization of MTs by nocodazole treatment (Abal et al., 2002). After 5 min of recovery from the nocodazole treatment, newly synthesized MTs radiated from the centrosome in PP4c^cko/cko MEF cells (Fig. 6 A). In contrast, PP4c^−/− MEF cells revealed severe reduction in the regrowth of MTs from the centrosome. In addition, MTs did not create long fibers. These observations suggest that loss of PP4c might cause a nucleation defect of MTs and/or loss of stability of MT array from the centrosome.

To visualize the MT dynamics in living cells, we used EB1-GFP as a marker of growing distal tips of MTs (Mimori-Kiyosue et al., 2000) using optical scanning microscopy (see Materials and methods). In the PP4c^cko/cko MEF cells, the great majority of EB1-GFP dots emanated from the centrosome and spread throughout the cytoplasm (Video 1). In contrast, PP4c^−/− MEF cells displayed fewer EB1-GFP foci moving away from the centrosome, which is associated with an intense signal of EB1-GFP around the centrosome, suggesting that these cells carry unusually large amounts of MT plus ends close to the centrosome (Fig. 6, B and C; and Video 2). EB1-GFP foci emanating from the centrosome in PP4c^−/− MEF cells were growing at a relatively normal speed, implying that loss of PP4c does not perturb the assembly of tubulins. Quantitative analysis of the fluorescence intensity of EB1-GFP at the centrosome in PP4c^−/− MEF cells indicated that PP4c disruption caused a twofold increase in EB1-GFP at the centrosome compared with PP4c^cko/cko MEF cells (Fig. 6 D). In contrast, the number of MT plus ends in the cytoplasm, which was determined by counting EB1-GFP spots, was reduced by ∼45% in PP4c^−/− MEF cells compared with PP4c^cko/cko MEF cells (Fig. 6 E). The intensity of γ-tubulin was similar between normal and PP4c^−/− MEF cells as shown in Fig. 3 A. These observations suggest that MTs are able to nucleate at the centrosome, but they are abnormally severed thereafter in PP4c^−/− MEF cells. Free MT minus ends would likely be unstable in PP4c^−/− MEF cells; thus, fragmentation could lead to an overall reduction in the number of astral MTs (Rodionov et al., 1999). This would provide an explanation for the reduction of EB1 signal in the cytoplasm in PP4c^−/− MEF cells.

To examine the perturbations in MT dynamics caused by the loss of PP4c, we transiently expressed GFP-tubulin. Because of the high density and complex organization of MTs in the vicinity of the centrosome, monitoring individual dynamics of MTs in MEF cells is technically challenging. Therefore, we analyzed the time course of MT organization in the peripheral region of the cell (Table 1). The reduced number of MTs in these basal patches allowed us to identify MT distributions and activities that contribute to the organization of MT networks. In types B and C, PP4c^−/− MEF cells displayed a severe fragmentation of MTs, which made it impossible to define MTs connecting to the centrosome. Therefore, we selected PP4c^−/− MEF cells from type A, which allowed us to find MTs connected to the centrosome. In PP4c^cko/cko MEF cells, dynamic MTs grew steadily and radially toward the periphery. Derived from the instantaneous rates of dynamic instability, the growth rate and the shrink rate were 10.31 ± 0.65 μm/min and 17.93 ± 1.11 μm/min, respectively (Table I). As growing MTs reached cell margins, the majority initiated episodes of typically 10–20 s of pausing followed by rapid shortening (catastrophes) and regrowth (rescues), preserving the radial orientation of most MTs. In contrast, PP4c^−/− MEF cells failed to undergo growth, exhibiting 43% less time in growth and 34% more time in pausing. PP4c^−/− MEF cells exhibited 1.6-fold more episodes of catastrophic shortening, whereas the rescue frequency was similar. Despite the noticeable differences in MT dynamics, actual rates of MT growth and shrinkage were similar between PP4c^cko/cko MEF cells and PP4c^−/− MEF cells.

We found that PP4c disruption results in the increased phosphorylation of T219 of NDEL1 accompanied by the unscheduled activation of Cdk1 in interphase. PP4c^−/− MEF cells revealed fragmented MTs near the centrosome. Therefore, we reasoned that perturbations of MTs in PP4c^−/− MEF cells could, in part, be caused by an increase in katanin p60 levels at the centrosome, which could be relieved by inhibition of Cdk1, additional disruption of Ndel1, or suppression of katanin p60 activity.

First, to explore the causative relationships between the unscheduled activation of Cdk1 and disorganization of MTs, we used an siRNA knockdown approach to inactivate Cdk1 (Fig. 7 A and Fig. S5 A). Silencing of Cdk1 clearly suppressed the phosphorylation of NDEL1 at T219, which lead to the reduction of katanin p60 recruitment to the centrosome, resulting in an improvement of MT organization. As another method to inhibit Cdk1 activity, PP4c^−/− MEF cells were cultured in the presence of the Cdk inhibitor butyrolactone I (Fig. S5 B; Kitagawa et al., 1993). Inhibition of Cdk1 suppressed the phosphorylation of NDEL1 at T219 and facilitated the redistribution of katanin p60 from the centrosome to the cytoplasm, resulting in a reduction of katanin p60 concentration at the centrosome. As with Cdk1 siRNA, butyrolactone I resulted in an improvement of the MT organization in PP4c^−/− MEF cells (Fig. S5 B). These observations suggest that abnormal activation of Cdk1 may be a causative mechanism for abnormal katanin p60 accumulation at the centrosome.

Second, we addressed whether double disruption of PP4 and Ndel1 can rescue MT disorganization. We mated PP4c^cko/cko mice and Ndel1^cko/cko mice (Sasaki et al., 2005) and generated PP4c^cko/cko/Ndel1^cko/cko mice. PP4c^cko/cko/Ndel1^cko/cko mice are viable and fertile, and we established MEF cell lines from embryos of these mice. We previously reported that the disruption of Ndel1 resulted in a mislocalization of katanin p60 into the nucleus (Toyo-Oka et al., 2005). In double mutants, katanin p60 displayed diffuse distribution involving the nucleus rather than a centrosomal distribution despite the presence of activated Cdk1 (Fig. 7 B). Importantly, the centrosomal array of MTs was clearly improved. Our findings suggest that NDEL1 is located in the middle of the pathway between PP4c and katanin p60.

Finally, we examined whether depletion of katanin p60 by RNA interference or expression of dominant-negative katanin p60 can rescue MT disorganization (Fig. S5 C). In katanin p60–depleted PP4c^cko/cko MEF cells, MTs exhibited twofold less time in pausing and longer periods of growth (Table 1). In addition, the growth rate was clearly augmented. The behavior of MTs also revealed that these MTs continued to grow, bending and curling at margins instead of tethering transiently to the membrane. Consistent with this feature, many MTs in katanin p60–depleted cells appeared to be longer than their counterparts. Interestingly, in these cells, MTs revealed more frequent episodes of catastrophic shortening. Most importantly, however, perturbation of MT nucleation in PP4c^−/− MEF cells was clearly rescued by the depletion of katanin p60 (Fig. 7 C). In addition, the extended pause of MTs and augmented episodes of catastrophe were also relieved (Table 1). Depletion of katanin p60 in PP4c^−/− MEF cells did not suppress the unscheduled activation of Cdk1 and did not inhibit the abnormal phosphorylation of T219 of NDEL1 (Fig. 7 C). Next, we tested whether the expression of dominant-negative katanin p60 is able to rescue MT defects in PP4c^−/− MEF cells (Fig. S5 D; Buster et al., 2002). The expression of dominant-negative p60 in wild-type cells had a similar effect on MT dynamics as the depletion of katanin p60 by siRNA (Table 1). Importantly, the expression of dominant-negative p60 in PP4c^−/− MEF cells restored MT behavior. These observations suggest that the disorganization of MT array in PP4c^−/− MEF cells could be partly attributed to the accumulation of excess katanin p60 associated with the unscheduled activation of Cdk1 and the phosphorylation of NDEL1 by Cdk1.

We have identified PP4c as an NDEL1-interacting protein. PP4c is a highly conserved PP2A-related serine/threonine phosphatase (Brewis et al., 1993) that efficiently dephosphorylates NDEL1 at Cdk1 phosphorylation sites. Interestingly, PP4c is preferentially distributed at the interphase centrosome and rapidly relocates from the centrosome upon entry into mitosis, which coincides with the activation of Cdk1 and the phosphorylation of NDEL1 by activated Cdk1. We further found that persistent expression of PP4c at the centrosome prevented the activation of Cdk1 and mitotic progression. In addition, the expression of PP4c prevented phosphorylation of NDEL1 by coexpressed active Cdk1, suggesting that PP4c is functionally dominant over Cdk1 and might be a checkpoint protein for the entrance of mitosis. Interestingly, the expression of mitotic Cdk1 in interphase facilitated the phosphorylation of NDEL1 and increased centrosomal levels of katanin p60, resulting in a reduction in MTs and suggesting that a coordination of PP4c and Cdk1 activities is essential for the proper regulation of MT organization.

To address how PP4c regulates NDEL1 at the centrosome, we have generated mice carrying a conditional KO allele of PP4c. PP4c^−/− MEF cells exhibited a disorganized MT array associated with the reduction of centrosomal connections. Transient expression of EB1-GFP and analysis of MT dynamics revealed the significant reduction of emanation of MTs from the centrosome. EB1 also accumulated at the centrosome and was reduced within the cytoplasm in PP4c^−/− MEF cells. We propose that EB1 behavior in PP4c^−/− MEF cells is attributable to the increased severing activity of MTs at the centrosome. This would explain the accumulation of EB1 at the centrosome. A failure of MT connection to the centrosome is known to cause MT instability (Rodionov et al., 1999), which would result in a reduction of EB1 signal in the cytoplasm. We also examined MT behaviors using GFP-tubulin. In PP4c^−/− MEF cells, MTs were prone to stay paused with reduced growth. Catastrophe frequency was markedly increased in PP4c^−/− MEF cells. In addition, MT dynamics in PP4c^−/− MEF cells were similar both at the cell periphery and around the centrosome. Taking this into account, we speculate that these MT behaviors would reflect an event at the centrosome, presumably the frequent loss of MT connection with the centrosome. Interestingly, PP4c disruption resulted in the unscheduled activation of Cdk1 and aberrant phosphorylation of NDEL1, which is associated with the excessive recruitment of katanin p60 at the centrosome. These findings suggest that PP4c may regulate the distribution of katanin p60 through regulation of Cdk1 activity and phosphorylation/dephosphorylation of NDEL1. Excessive katanin p60 at the centrosome would partly explain the frequent loss of MT connection with the centrosome in PP4c^−/− MEF cells. We found increased levels of centrosomal 14-3-3ε and decreased distribution of centrosomal dynactin 1 and differential interference contrast (unpublished data). Dynactin 1 binds to EB1, and these interactions are essential for MT anchoring at the centrosome (Berrueta et al., 1999; Askham et al., 2002). Although these additional factors would further modulate the phenotypes of MT disorganization in PP4c^−/− MEF cells, we believe that abnormal accumulation of katanin p60 is a primary event. Depletion or inhibition of Cdk1 activity by siRNA or an inhibitor, additional disruption of Ndel1, or inhibition of katanin p60 by siRNA or dominant-negative p60 suppressed the perturbation of MT dynamics in PP4c^−/− MEF cells, fully supporting our interpretation.

Recently, chromosome movement toward mitotic spindle poles by a Pacman-flux mechanism has been proposed (Zhang et al., 2007). In this model, katanin p60 appears to function primarily on anaphase chromosomes, where it stimulates MT plus end depolymerization. In our model, katanin p60 plays an essential role at the centrosome for MT remodeling. These differences might be attributable to the stage of cell cycle (mitosis or interphase) and/or the origin of cells (Drosophila and mammals). The prolonged growth rate of MTs by depletion of katanin p60 by siRNA or expression of dominant-negative p60 supports our interpretation.

Centrosomes are the dominant sites of MT assembly. In particular, as cells enter mitosis, centrioles recruit pericentriolar material in the process of centrosome maturation, which increases the MT nucleating capacity at the centrosomes (Doxsey et al., 2005a,b). The increased dynamics of MTs in mitosis is an essential prerequisite for spindle formation. It is driven by coordination of the activities of MT-stabilizing and -destabilizing factors. For example, the combination of XMAP215 and XKCM1/mitotic centromere-associated kinesin is essential to promote the dynamic properties of mitotic MT assembly (Tournebize et al., 2000; Kinoshita et al., 2001). The coordination of these components is further modulated by the activities of kinases and phosphatases. We previously reported that NDEL1 is essential for centrosomal targeting of katanin p60 (Toyo-Oka et al., 2005). Targeted disruption of Ndel1 resulted in an abnormal nuclear distribution of katanin p60 and perturbation of MT organization. In addition, phosphorylation of NDEL1 by Cdk1 enhances the affinity between NDEL1 and katanin p60. We recently reported that NDEL1 is phosphorylated by Aurora A and is required for recruitment of TACC3 to the centrosome (Mori et al., 2007). Phosphorylation of NDEL1 by Cdk1 may contribute to the destabilization of MT connections at the centrosome by recruitment of katanin p60. Conversely, phosphorylation of NDEL1 by Aurora A may influence the stabilizing machinery by accentuating the localization of TACC3 at the centrosome. PP4c contributes to regulation of the proper activation of Cdk1 and recruitment of katanin p60 at the centrosome. The MT behavior in PP4c^−/− MEF cells leads us to speculate that PP4c may be a part of the regulatory machinery for the MT-stabilizing pathway, acting directly/indirectly through the regulation of centrosomal components, including Cdk1, NDEL1, and katanin p60. Our observations underscore the importance of NDEL1 in the recruitment and coordination of centrosomal components. In addition, maintenance of the proper phosphorylation of NDEL1 is essential for the regulation of MT dynamics by recruitment of proper target molecules.

Full-length Ndel1 cDNA was conjugated to pLexA (Clontech Laboratories, Inc.) and used to screen a fetal mouse brain cDNA yeast two-hybrid library as described previously (Sasaki et al., 2000). Although protein interaction was confirmed by a yeast two-hybrid analysis, direct interaction was not detected by an immunoprecipitation assay, suggesting that the enzyme–substrate interaction might not be strong enough for the detection of protein interaction by nonequilibrium binding assays. We generated GST-tagged full-length recombinant NDEL1 (Toyo-Oka et al., 2005), cyclin B1, Cdk1, Aurora A (Marumoto et al., 2003), and PP4c by Bac-to-Bac baculo-system (Invitrogen) using SF-9 or High Five insect cells (BD Biosciences). To obtain active Cdk1, baculovirus carrying GST–cyclin B1 and GST-Cdk1 were simultaneously inoculated to High Five insect cells. Recombinant proteins were purified using GST-Sepharose (GE Healthcare), and the GST tag of NDEL1 was removed by thrombin digestion (GE Healthcare) based on the manufacturer's manual. A mutated GST-PP4c construct in which arginine 236 was replaced with leucine, resulting in a loss of phosphatase activity (Zhou et al., 2002), was generated by QuikChange (Stratagene). NDEL1 was phosphorylated by GST-Cdk1 or GST–Aurora A as reported previously (Mori et al., 2007). Kinases and unincorporated ATP were removed by GST column and gel filtration column (GE Healthcare). Dephosphorylation assays of NDEL1 by GST-PP4c were performed in a 50-μl reaction mixture containing 0.1 μg of phosphorylated NDEL1, 20 mM Tris-HCl, pH 7.4, 1 mM DTT, and 1 mM EDTA. After incubation at 30°C for 20 min, the reaction was terminated and subjected to SDS-PAGE followed by autoradiography.

We generated a conditional knockout mouse to inactivate PP4c by Cre-mediated recombination. We assembled a targeting construct in which a loxP site and a PGK-neo gene flanked by two loxP sites were inserted into intron IV and downstream of PP4c, respectively. The linearized targeting construct was introduced into TC1 embryonic stem cells (Deng et al., 1996) from a 129S6 background by electroporation. The targeted embryonic stem clones were screened by Southern blotting and injected into blastocysts to create chimeric mice. Highly agouti chimeric males were mated to wild-type females to give rise to heterozygotes for PP4c^neo/+, which were identified by Southern blot analysis and PCR. These heterozygous mice were mated with EIIa-Cre germline deleter transgenic mice (Lakso et al., 1996). Offsprings from the matings between a PP4c^neo/+ line and an EIIa-Cre transgenic line were genotyped by Southern blot analysis and PCR. The Southern blot analysis and PCR examination indicated deletion of the fragment carrying exons V–VIII by Cre-mediated recombination in vivo (PP4c^−/+). PP4c^neo/neo mice revealed embryonic lethality, suggesting that the presence of the neo gene affects the expression of PP4c. To generate PP4c^cko/+, we mated a PP4c^neo/+ line with an EIIa-Cre transgenic line and isolated offsprings in which only loxp-flanked PGK-neo was removed by Cre-mediated partial recombination. PP4c^cko/cko mice were generated by the mating of PP4c^cko/+ and turned out to be viable, fertile, and able to be maintained as homozygotes. PP4c^cko/cko MEF cells were established from embryonic day 12.5 embryos from matings of PP4c^cko/cko. We also established PP4c^+/+ MEF cells from embryonic day 12.5 embryos from matings of wild-type mice. These MEF cell lines were transfected with each plasmid or infected with adenovirus carrying a Cre gene (adeno-Cre) and were subjected to live imaging or immunostaining at given times after transfection. PP4c^+/+ MEF cells were used for controls as the same condition.

To quantify fluorescence intensity, centrosomes were surrounded by a 0.5-μm-diameter circle, and fluorescence intensity of the circle was measured. We also measured background intensity close to the centrosome with the same diameter circle and subtracted the background intensity from the centrosomal intensity in each case.

To make expression vectors, we cloned each cDNA into pEGFP (Clontech Laboratories, Inc.). Phosphatase-inactive PP4c and mitotic Cdk1 were generated by QuikChange (Zhou et al., 2002; Litvak et al., 2004). For the Cre expression vector, the Cre gene was conjugated to a Cherry vector. To achieve centrosome-restricted expression of PP4c or Cdk1, we conjugated each protein to a PACT domain derived from pericentrin (Gillingham and Munro, 2000). Dominant-negative katanin p60 used in the experiments was reported previously (Toyo-Oka et al., 2005).

To make anti-PP4c or R1 antibodies, we immunized New Zealand white rabbits with a GST-conjugated recombinant PP4c (1–55 amino acids) or R1 (452–622 amino acids) expressed in bacteria and purified by GST-Sepharose according to standard procedures. The antisera against PP4c- (1–55 amino acids) or R1 (452–622 amino acids)-peptide were collected and purified by HiTrap columns (GE Healthcare) coupled with the antigens used (PP4c [1–55 amino acids] or R1 [452–622 amino acids]). To produce antiphosphorylated mouse cyclin B1 (phospho-S123) antibody, keyhole limpet hemocyanin (KLH)–conjugated mouse cyclin B1 phosphopeptides (CILVDNP[pS]PSPME) were injected into New Zealand white rabbits according to standard procedures. The antisera were first applied to KLH-conjugated nonphosphorylated peptides (CILVDNPSPSPME) to remove anti-KLH antibodies and antibodies that bind to nonphosphorylated peptides and were purified by using phosphopeptide (CILVDNP[pS]PSPME)-coupled HiTrap columns.

HeLa cells were cultured in MEM (Sigma-Aldrich) supplemented with 10% FCS. For mitosis synchronization, HeLa cells were exposed to 2 mM thymidine for 16 h and were resuspended in fresh medium supplemented with 24 μM 2′-deoxycytidine and allowed to grow for 9 h. 2 mM thymidine was added again for 16 h, causing cells to accumulate near the G1/S boundary. MEF cells were prepared from various conditional knockout mouse embryos according to standard procedure. MEF cells were grown in DME supplemented with 10% FBS and penicillin/streptomycin. For transfection, cells were collected by using 0.25% trypsin/1 mM EDTA in HBSS and transfected with the Nucleofector transfection system (Amaxa) with a MEF1 kit according to the manufacturer's instructions. For immunostaining, cells were plated onto eight-well chamber slides (BD Biosciences) at a density of 1–2 × 10⁴ cells/well after transfection. For time-lapse imaging, cells were plated onto four-well LabTekII chamber slides (coverslip bottom; Thermo Fisher Scientific) at a density of 1–1.5 × 10⁵ cells/well after transfection. Cells were used for experiments 24–48 h after transfection. For the analysis of MT dynamics, cells were treated with 0.3 U/ml Oxyrase oxygen scavenging system (EC Oxyrase) to reduce photodamage and photobleaching (Mikhailov and Gundersen, 1995). The Oxyrase-treated cells were covered with mineral oil to prevent gas exchange.

Recombinant adenoviruses (Ad-FRP-Cre) were constructed according to the manufacturer's instructions (BD Biosciences). In brief, cDNAs encoding an RFP-Cre were inserted into the pShuttle2 vector. The expression cassette was excised from the recombinant pShuttle2 plasmid DNA by digesting with I-CeuI and PI-SceI. Afterward, the expression cassette ligated to Adeno-X viral DNA (BD Biosciences). These adenoviral vectors were transfected into human embryonic kidney 293 cells, which provide the E1A gene product necessary for viral replication. The resultant recombinant viruses were propagated in 293 cells, and titers were determined by the Adeno-X Rapid titer kit (BD Biosciences).

Cells were fixed in cold methanol for 3 min at −20°C or 4% PFA/PBS for 3 min at room temperature. Fixed cells were incubated with 0.2% Triton X-100/TBS to be permeabilized for 10 min at room temperature followed by blocking treatment. 5% BSA/0.2% Tween 20/TBS or milk-based blocking reagent Blockace (Dainippon Sumitomo Pharma) supplemented with 2% FCS was used for blocking an antiphosphorylated or anti-unphosphorylated protein antibody, respectively. After blocking, we incubated the cells with various antibodies at 4°C overnight or for 1 h at room temperature. We used the antibodies at the following dilutions in blocking solution: rabbit anti-NDEL1 at 1:200; 7 μg/ml affinity-purified rabbit anti-p60; 7 μg/ml affinity-purified rabbit anti-p80; 2 μg/ml affinity-purified rabbit anti-PP4c; 2 μg/ml affinity-purified rabbit anti-PP4R1; 2 μg/ml mouse monoclonal anti-T219 phosphorylated NDEL1 (clone N219TP; MBL International); rabbit anti–γ-tubulin (BioLegend) at 1:400; mouse monoclonal anti-T219 phosphorylated NDEL1 at 1:25 (clone 5; BD Biosciences); mouse monoclonal anti–β-tubulin at 1:300 (clone TUB21; Sigma-Aldrich); mouse monoclonal anti–γ-tubulin at 1:200 (clone GTU88; Sigma-Aldrich); mouse monoclonal antiacetylated tubulin at 1:100 (clone 6-11B-1; Sigma-Aldrich); rabbit anti-NDEL1 at 1:200; and 2 μg/ml affinity-purified rabbit antiphosphorylated mouse cyclin B1 (phospho-S123) antibody. After washing three times with 0.1% Tween 20/TBS for 10 min each, cells were treated with various secondary antibodies in blocking solution for 30 min at room temperature. We used the following secondary antibodies: Cy5-labeled donkey anti–mouse IgG at 1:500 or anti–rabbit IgG at 1:400 (Jackson ImmunoResearch Laboratories) and AlexaFluor488-labeled donkey anti–mouse IgG at 1:200 or anti–rabbit IgG at 1:400 (Invitrogen). We visualized nuclei with 300 nM DAPI and mounted the cells with 90% glycerol/PBS.

Images of cells were acquired with a laser-scanning confocal microscope (LSM510 version 2.3; Carl Zeiss, Inc.) or an optical sectioning microscope (DeltaVision version 2.50; Applied Precision) equipped with Axiovert plan Apochromat (63× 1.40 NA or 100× 1.40 NA; Carl Zeiss, Inc.) oil immersion or UPlanApo (20× 0.70 NA dry, PlanApo 60×1.40 NA oil, or PlanApo 100× 1.40 NA oil immersion; Olympus) objectives, respectively. Using the confocal microscope, images were recorded using confocal software (Carl Zeiss, Inc.) and were exported as TIFF. Figures were then generated using Photoshop 7.0 and Illustrator CS2 (Adobe). Using the DeltaVision system, images were acquired through a cooled CCD camera (series 300 CH350; Photometrics) with appropriate neutral density filters, binning of pixels, exposure times, and time intervals. Fluorescent signals were visualized using an Endow GFP bandpass emission filter set (model 41 017) or a Sedat Quad filter set (model 86 000; Chroma Technology Corp.). Pixel positions, distances, and areas were measured on the digital images using the analysis function of the DeltaVision Aquacosmos software (Hamamatsu) or MetaMorph software (MDS Analytical Technologies). Cell areas were quantified using the particle analysis function of the Aquacosmos software.

Obtained images were exported as TIFF files and processed with ImageJ version 1.37h (National Institutes of Health). Black and white raw data were converted to color images to visualize EB1-GFP and the end of MTs easily and were processed with the bandpass filter function of ImageJ software to remove haze. EB1-GFP and the end of MTs were traced by a mouse-driven cursor under ImageJ software plugged in the manual track function (developed and programmed by Fabrice Cordelières, Institut Curie, Orsay, France). Changes >0.5 μm between two points were considered as growth or shortening events. Changes <0.5 μm were considered as a pause. More than 100 MTs from 8–14 cells were analyzed. The results are indicated as the mean and SEM, and at least three independent experiments were performed. The frequency of catastrophe or rescue was determined as previously described (Rusan et al., 2001). In brief, the frequency of catastrophe was determined by dividing the number of transitions from growth to shortening and from pause to shortening by the time spent in growth or pause. The frequency of rescue was calculated by dividing the number of transitions from shortening to growth and from shortening to pause by the time spent shortening. MTs that could be traced for >1 min were included in the analysis. Statistical analysis was performed by using Prism software (GraphPad).

21-nucleotide RNAs were chemically synthesized by Dharmacon Research and transfected with Lipofectamine 2000 reagents (Invitrogen) or the Nucleofector transfection system. The target sequences used were GGACAUAGAUGAGGCUUUATT and CAGAGAACAUGGAAGGCUATT against mouse katanin p60 and GGAGUGCCCAGUACUGCAATT against mouse Cdk1. Certified control siRNA (UUCUUCGAACGUGUCACGUTT) was used as a negative control (QIAGEN). Cell lysates were prepared and analyzed for the silencing. In addition, COS7 cells expressing GFP-tagged mouse katanin p60 (GFP-p60) were treated with siRNA and verified that two siRNAs down-regulated the expression of GFP-p60 (Fig. S5 C).

The MT regrowth assay was performed as previously reported (Fry et al., 1998). In brief, RFP-Cre–transfected cells were seeded in chamber slides and treated with 1.5 μg/ml nocodazole in prewarmed culture medium for 20 min at 37°C under 5% CO₂. After washing three times with prewarmed PBS, cells were cultured with DME supplemented with 10% FBS at a given time to observe the regrowth of MTs at various points. At each time point, cells were fixed with cold methanol for 3 min followed by immunostaining with anti–β-tubulin antibody.

Fig. S1 shows the generation of a gene-disrupted mouse of PP4c. Fig. S2 shows the genomic organization of PP4c and Tbx6. Fig. S3 shows a characterization of an antibody against phospho-S123 of mouse cyclin B1. Fig. S4 shows a characterization of cell cycle arrest after treatment with mitomycin C or contact inhibition. Fig. S5 shows rescue experiments in PP4c^−/− MEF cells. Video 1 and Video 2 show a PP4c^+/+ MEF cell and PP4c^−/− MEF cell, respectively, undergoing emanations of EB1 from the centrosome.

We thank Virginia E. Papaioannou for providing a Tbx6-disrupted mouse and Mark Jackman and Jonathon Pines for providing an antiphospho–cyclin B1 antibody. We also thank Shinsuke Matsuzaki for providing a vector carrying a PACT domain, Kozo Kaibuchi for providing a GFP-EB1 construct, Takao Kenko, Emi Donoue, and Keisuke Inoue for technical support, and Hiromichi Nishimura and Keiko Fujimoto for mouse breeding. We thank Yoshihiko Funae, Toshio Yamauch, and Yoshitaka Nagai for generous support and encouragement and Virginia E. Papaioannou, Mark Jackman, Jonathon Pines, Kazuhisa Kinoshita, and Yuko Kiyosue for valuable discussion and comments.

This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture of Japan to K. Toyo-oka and S. Hirotsune, by a Grant-in-Aid for Young Scientists (B) to K. Toyo-oka, and a Grant-in-Aid for Scientific Research (B) from the Ministry of Education, Science, Sports and Culture of Japan to S. Hirotsune. This work was also supported by the Yasuda Medical Research Foundation, the Cell Science Research Foundation, the Japan Spina Bifida and Hydrocephalus Research Foundation, and the Hoh-ansha Foundation (grants to S. Hirotsune) as well as by the Senri Life Science Foundation (grant to K. Toyo-oka).