NDEL1 Phosphorylation by Aurora-A Kinase Is Essential for Centrosomal Maturation, Separation, and TACC3 Recruitment^{▿ †}

In vitro phosphorylation of NDEL1 by brain extracts.

Cortices were dissected from wild-type mice or Cdk5^−/− mice (28) at embryonic day 15.5 (E15.5), and soluble protein was extracted with 20 mM K-HEPES (pH 7.4), 50 mM KCl, 1 mM dithiothreitol (DTT), 0.2% NP-40, and protease inhibitor cocktail (Roche). The protein concentration was adjusted to 100 μg/ml and used for in vitro kinase reactions. In vitro kinase reactions were carried out using 1 μg of recombinant NDEL1 and 0.5 μg of brain extract protein expressed in bacteria under the condition of 20 mM Tris-HCl (pH 7.4), 50 mM NaCl, 10 mM MgCl₂, and 1 mM dithiothreitol with 10 μCi of [γ-³²P]ATP. Phosphorylation was detected by radioactivity or by Western blot analysis using an anti-phospho-T219 monoclonal antibody (38).

Generation of Ndel1 stable transformants and replacement experiments.

We established mouse embryonic fibroblasts (MEFs) from Ndel1^cko/cko mice (33). We cotransfected the Cre expression vector and the glutathione S-transferase (GST)-NDEL1 expression vector, which carried a Neo^r gene (Clontech), and followed this by selection with G418. Ndel1-null MEFs are nonviable, whereas GST-NDEL1-transformed MEFs are viable and grow normally. We also examined the expression amounts and subcellular localization by Western blot analysis using the anti-NDEL1 antibody (32) and an anti-GST antibody (Upstate). GST-NDEL1 was purified by GST-Sepharose 4B (Amersham) based on the manufacturer's recommended conditions.

For replacement experiments, we transfected expression vectors that carry green fluorescent protein (GFP)-tagged wild-type NDEL1 and mutated NDEL1s (see below) to MEF cells in which endogenous Ndel1 was removed by CRE-mediated recombination before synchronization. MEFs were cultured in Dulbecco's modified Eagle's medium (DMEM) (Sigma) containing 10% fetal bovine serum. For synchronization, G₁ cells were obtained by culturing cells in DMEM with 0.1% newborn calf serum for 2 days. After release, MEFs were fixed or harvested at a given time.

Identification of phosphorylation site of NDEL1 by Aurora-A.

To determine native phosphorylation sites of NDEL1, GST-NDEL1 was extracted from 8 × 10⁷ MEF cells in which endogenous Ndel1 was replaced by GST-Ndel1 (see above) using lysis buffer (20 mM K-HEPES, pH 7.4, 50 mM KCl, 1 mM DTT, 0.2% NP-40, protease inhibitor cocktail [Roche], 1 mM EDTA, 1 mM sodium orthovanadate, and 1 mM NaF) and purified using GST-Sepharose (Amersham). Purified GST-NDEL1 (0.5 μg) was digested with trypsin (Roche), and this was followed by the determination of phosphorylation sites using liquid chromatography/atmospheric pressure ionization tandem mass spectrometry with a hybrid triple-quadruple linear ion trap mass spectrometer (QTrap) (ABI) based on the manufacturer's manual. We performed a series of experiments with three independent cell lysates and obtained the same results in each case.

Generation of recombinant protein and in vitro kinase reactions.

We generated GST-tagged, full-length recombinant Plk1 by using a Bac-to-Bac baculo system (Invitrogen) with SF-9 or High Five insect cells (Pharmingen). To obtain active Plk1, insect cells were cultured in the presence of 100 nM okadaic acid (OA) (Sigma) in dimethyl sulfoxide 3 h prior to harvesting. Forty-eight hours postinfection, Plk1 proteins were purified as described previously (32) and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) or used for kinase assays. Active Aurora-A was expressed in insect cells and purified as reported previously (23). CDK1 was purchased from New England Biolabs. In vitro kinase reactions were performed in the following conditions with 10 μCi of [γ-³²P]ATP: for Cdk1 (New England Biolabs), 50 mM Tris-HCl (pH 7.4), 10 mM MgCl₂, 1 mM EGTA, 2 mM dithiothreitol, and 0.01% Brij 35; for Plk1, 20 mM HEPES (pH 7.7), 15 mM KCl, 10 mM MgCl₂, 1 mM EGTA, and 5 mM NaF at 37°C for 30 min; and for Aurora-A (23), 50 mM Tris-HCl (pH 7.4), 50 mM NaCl, 10 mM MgCl₂, and 1 mM dithiothreitol. A total of 0.4 μg of each purified kinase and 1 μg of histone H3 were used in each 50-μl reaction mixture. The reaction was stopped by the addition of Laemmli sample buffer, and phosphorylated substrates were detected by SDS-PAGE and autoradiography. Staurosporin (100 nM; CycLex) was used as an inhibitor of Aurora-A.

As a substrate, we generated recombinant NDEL1 using the pGEX (Amersham) vector to express in the bacteria. To generate structurally more natural recombinant proteins for the sucrose gradient sedimentation assay and the Biacore assay, we generated GST-tagged, full-length recombinant NDEL1, TACC3, and mutant proteins by using the Bac-to-Bac baculo system (Invitrogen) with SF-9 or High Five insect cells (Pharmingen). Protein purification was performed using GST-Sepharose (Amersham) based on the manufacturer's recommendations.

Generation of NDEL1 expression vectors.

A PCR fragment of full-length Ndel1 was cloned into Bluescript II (Stratagene) and sequenced. The XhoI-XbaI fragment excised from the original vector was cloned into the Bac-to-Bac baculo system or pEGFP (Clontech) for expression in insect cells and mammalian cells, respectively. For the expression in mammalian cells, we replaced the AseI-AgeI fragment carrying the cytomegalovirus promoter of pEGFP with the HindIII-HindIII fragment carrying the elongation factor promoter (EF-1α) from pEF6/Myc-His (Invitrogen) for gradual and mild expression. Mutated NDEL1 constructs in which phosphorylation sites were changed to alanine or glutamic acid were generated by QuikChange (Stratagene). These constructs contain the following mutations: NDEL1(S251A), S251A; NDEL1(S251E), S251E; NDEL1(tripleS/T→A), Ser198A, Thr219A, and Ser231A; NDEL1(tripleS/T→E), Ser198E, Thr219E, and Ser231E; NDEL1(tripleS/T→A;S251E), Ser198A, Thr219A, Ser231A and S251E; and NDEL1(tripleS/T→E;S251E), Ser198E, Thr219E, Ser231E and S251E. Transformation was performed with a Gene Pulser (Bio-Rad) based on the manufacturer's recommended conditions.

Generation of an anti-Ser251 monoclonal antibody.

Six-week-old BALB/c mice were immunized with 50 μg of keyhole limpet hemocyanin-conjugated oligopeptide [CTPSARIS(PO₃H₂)ALN] in Freund's complete adjuvant (Difco), with a second injection in Freund's incomplete adjuvant (Difco) 14 days later. The splenocytes of the immunized mouse were fused to SP2/0 mouse myeloma cells by using a standard polyethylene glycol protocol. The fused cell population was resuspended in hypoxanthine aminotropterin thymidine selection medium (Life) and plated into 96-well, flat-bottomed culture plates. Positive clones that specifically recognized phosphorylated NDEL1 were subcloned to obtain clonal hybridomas. Subtyping of our anti-phosphorylated NDEL1 monoclonal antibody revealed immunoglobulin M, which makes it possible to perform double staining with another anti-phosphorylated NDEL1 monoclonal antibody (38).

Ubiquitination assay.

For immunoprecipitation, wild-type GFP-NDEL1 and mutant NDEL1s were transfected to synchronized HeLa cells. Cells were lysed at given times by lysis buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% NP-40, and protease inhibitor cocktail; Roche) and incubated with an anti-GFP rabbit polyclonal antibody (Clontech) in 100 μl, followed by immunoprecipitation. Immunoprecipitates were subjected to SDS-PAGE analysis, followed by immunoblotting using an anti-Ub mouse monoclonal antibody (Santa Cruz; catalog no. sc-8017).

Synchronization of HeLa Cells, siRNA, and immunohistochemistry.

HeLa cells were cultured in modified Eagle's medium (MEM) (Sigma Chemical Co.) supplemented with 10% fetal calf serum. For mitosis synchronization, HeLa cells were exposed to 2 mM thymidine for 16 h and then resuspended in fresh medium supplemented with 24 μM 2′-deoxycytidine and allowed to grow for 9 h. Thymidine (2 mM) was added again for 16 h, causing cells to accumulate near the G₁/S boundary (23). After release from the double-thymidine block, cells were harvested at given times. The sequences of the short interfering RNAs (siRNAs) were as follows: for human Aurora-A, 5′-AUUCUUCCCAGCGCGUUCC-3′(corresponding to nucleotides 155 to 173 relative to the start codon) or 5′-AUGCCCUGUCUUACUGUCA-3′(nucleotides 725 to 743) and, for human TACC3, 5′-GUU ACC GGA AGA UCG UCU GTT-3′(nucleotides 85 to 103) or 5′-CAG ACG AUC UUC CGG UAA CTT-3′ (nucleotides 103 to 85) (19). For the inhibition of phosphatase and proteasome, we treated synchronized HeLa cells with okadaic acid and MG132, respectively. To minimize the secondary effect of these inhibitors, we optimized the dosage and duration of the treatment. Briefly, 1 nM of okadaic acid and 50 μM of MG132 were added at a given time for 30 min before harvesting for Western blotting or immunocytochemistry. Western blot analysis was performed for three independent cell lysates, and the standard error was calculated.

For rescue experiments, siRNA and each NDEL1 construct were introduced into HeLa cells simultaneously using Oligofectamine (Invitrogen). HeLa cells were grown in 35-mm petri dishes, fixed with 4% paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.4) for 20 min at room temperature, and permeabilized with 0.5% Triton X-100 in phosphate-buffered saline. After incubation for 60 min with 5% goat serum in phosphate-buffered saline, the cells were incubated at 4°C overnight with primary antibodies and then for 45 min at room temperature with fluorescein isothiocyanate-, Cy5-, or rhodamine-conjugated secondary antibodies (Bio-source, Molecular Probes, or Pharmacia).

Measurement of mitotic index.

Synchronization and transfection of HeLa is described above. Cells were observed 8.5 h after release by phase-contrast microscopy. The time course of the mitotic index of cells was measured at the indicated time, and the number of mitotic cells out of 1,000 cells examined in total was determined after fixation and staining with DAPI (4′,6′-diamidino-2-phenylindole).

Surface plasmon resonance.

Real-time biospecific interaction analysis was performed using a Biacore 2000 (Pharmacia Biosensor). Each protein was immobilized to CM5 biosensor microchips using N-hydroxy-succinimide and N-ethyl-N'-dimethylaminopropyl carbodiimide. In standard surface plasmon resonance experiments, kinetic binding experiments were performed under previously reported conditions (38). The same running buffer was used (50 mM KCl, 20 mM HEPES, pH 7.4, 1 mM DTT, and 0.1% NP-40), and the instrument was equilibrated at 24°C. Kinetic data were collected by injecting various concentrations of TACC3, CH-TOG/XMAP215, NDEL1, and mutated NDEL1s. Kinetic rate constants were determined by fitting the corrected response data to a simple bimolecular interaction model, A + B = AB. Kinetic constants K_A and K_D are the association constant and dissociation rate constant, respectively.

Identification of a novel phosphorylation site of NDEL1 by Aurora-A kinase.

We and other groups found that NDEL1 is a substrate of Cdk5/p35 (26, 32), and Cdk5/p35-mediated phosphorylation was lost in Cdk5-null mice (26). We extracted brain lysates from E15.5 Cdk5^−/− mice or wild-type mice and performed in vitro kinase assays to examine whether the Cdk5^−/− lysates lost competence to phosphorylate NDEL1. As expected, lysates derived from Cdk5^−/− mice lost the ability to phosphorylate consensus sites (at Ser198, Thr219, and Ser231) of CDK5/CDK1 on GST-NDEL1, but GST-NDEL1 was still efficiently phosphorylated, suggesting that NDEL1 was a substrate of another kinase (Fig. 1A). To hunt for novel kinases of NDEL1, we established a stable MEF line in which endogenous NDEL1 was replaced by GST-tagged exogenous NDEL1 (33). In this MEF line, GST-NDEL1 was expressed in normal amounts and distributed normally compared to endogenous NDEL1 (see Fig. S1A and S1B in the supplemental material), suggesting that the protein could be phosphorylated by the same kinase as endogenous NDEL1. Next, we purified GST-NDEL1 (see Fig. S1C in the supplemental material) and performed a systematic analysis of phosphorylation sites using LC-MS/MS. We found several phosphorylation sites in MEFs, including T219 and S231, which are thought to be phosphorylation sites for CDK1/CDK5 (26, 32) (see Fig. S1D in the supplemental material). We also found an additional phosphorylation site, S251 (see Fig. S1D in the supplemental material). Interestingly, S251 phosphorylation was observed only when GST-NDEL1 was extracted from dividing MEFs and was undetectable when GST-NDEL1 was isolated from stationary MEFs (data not shown), suggesting that this novel phosphorylation of S251 may be attributed to a mitotic kinase.

NDEL1 is a centrosomal protein. Therefore, we considered several mitotic kinases that are preferentially localized at the centrosome as candidates to phosphorylate recombinant NDEL1. As expected, CDK1 efficiently phosphorylated NDEL1 (26, 32) (Fig. 1B). We also confirmed by LC-MS/MS that CDK1 phosphorylated T219 and S231, whereas the phosphorylation of S198 was less efficient (data not shown). Plk1 is another kinase that functions during mitosis (27), but even when using okadaic acid for activation of Plk1 (18, 20), recombinant NDEL1 was not phosphorylated (Fig. 1B). Finally, we examined Aurora-A kinase and found that Aurora-A efficiently phosphorylated recombinant NDEL1 expressed from insect cells (23) (Fig. 1B). S251 and the surrounding amino acid sequence were weakly compatible with the consensus sequence of budding yeast protein kinase Ipl1/Aurora {KR}X{TS}{ILV} (3). We subjected NDEL1 phosphorylated by Aurora-A to LC-MS/MS and confirmed that Aurora-A phosphorylated S251 (data not shown). Next, we examined the effect of the mutation of S251 or an inhibitor of Aurora-A on phosphorylation. The replacement of serine by alanine at S251 or the presence of Aurora-A inhibitor staurosporin (39) significantly reduced the phosphorylation by Aurora-A (Fig. 1C). In addition, an introduction of the mutation at the catalytic site of Aurora-A significantly reduced the phosphorylation of NDEL1 (Fig. 1C). Thus, we concluded that Aurora-A specifically phosphorylates S251 of NDEL1. Direct interaction with Aurora-A appears to be essential for phosphorylation (23). We examined whether NDEL1 directly binds Aurora-A and whether the phosphorylation of NDEL1 by Aurora-A influences the interaction between the two proteins using wild-type Ndel1 and mutated Ndel1s. Immunoprecipitation revealed that GFP-NDEL1(S251E) was more efficiently precipitated with Aurora-A than was GFP-NDEL1(S251A) (see Fig. S1E in the supplemental material), suggesting that phosphorylated NDEL1 binds more tightly with Aurora-A and phosphorylated NDEL1 may preferentially distribute at the centrosome (see below).

Aurora-A phosphorylates S251 of NDEL1 at the beginning of prophase, followed by rapid downregulation.

To examine the phosphorylation of S251 by Aurora-A, we generated an anti-phosphorylated S251-specific monoclonal antibody (see Fig. S2 in the supplemental material) and monitored Aurora-A-mediated S251 phosphorylation through mitosis using HeLa cells. Given that the phosphorylation of Aurora-A on T288, which is located in the activation T loop of the kinase, results in a marked increase in enzymatic activity (40), we examined the phosphorylation state of T288 by immunohistochemistry with the anti-phospho-T288 antibody of Aurora-A. Consistent with previous observations (23, 40), the phosphorylation of Aurora-A on T288 occurred during the late G₂-prophase transition, coincident with enzymatic activation (Fig. 2A). As expected, S251 phosphorylation of NDEL1 appeared coincidently with Aurora-A activation. Interestingly, the activation of Aurora-A continued through mitosis and peaked at metaphase-anaphase, while S251 phosphorylation suddenly disappeared after prophase (Fig. 2A). TACC3 is another target of Aurora-A, which localizes specifically to mitotic centrosomes at the onset of prophase (9, 19) (Fig. 2B). Interestingly, the activation of Aurora-A and the phosphorylation of NDEL1 by Aurora-A preceded the phosphorylation of TACC3 by Aurora-A.

We compared Aurora-A-mediated S251 phosphorylation with CDK1-mediated T219 phosphorylation of NDEL1 (Fig. 2C). T219 phosphorylation began during prophase instead of the late G₂ prophase transition and continued throughout metaphase. To further clarify the temporal relationship between the enzymatic activity and S251 phosphorylation, we performed immunoblot analysis (Fig. 2D), which revealed that Aurora-A was activated at 8 h after the release of the double-thymidine block and continued to be active throughout metaphase. S251 phosphorylation appeared at 8 h after release but was downregulated soon afterwards, consistent with the immunohistochemistry data. In contrast, T219 phosphorylation appeared at 8.5 h after release and continued throughout mitosis. To address the mechanism of the sudden downregulation of S251, we examined S251 phosphorylation in the presence of MG132, which is a proteasome-specific inhibitor, or OA, which is a phosphatase inhibitor. MG132 significantly prolonged S251 phosphorylation. In contrast, OA was not effective in preventing the disappearance of S251 phosphorylation. These findings suggest that Ser251 phosphorylation may trigger proteasome-mediated rapid degradation and/or that the degradation cascade is activated during prophase (Fig. 2D).

Although Aurora-A was activated during mitosis, S251 phosphorylation disappeared at the beginning of prophase, and the sudden downregulation of S251 NDEL1 was coincident with the phosphorylation of NDEL1 by CDK1. This suggests that the phosphorylation by CDK1 may inhibit further phosphorylation by Aurora-A. We therefore examined whether mutant NDEL1 in which phosphorylation sites by CDK1 were replaced by glutamic acid [GST-NDEL1(tripleS/T→E)] or alanine [GST-NDEL1(tripleS/T→A)] was a better or a worse substrate of Aurora-A. Compared to the wild-type NDEL1 or GST-NDEL1(tripleS/T→A), which was also a good substrate of Aurora-A, GST-NDEL1(tripleS/T→E) was a much less efficient substrate (Fig. 2E). In contrast, the replacement of S251 by glutamic acid [GST-NDEL1(S251E)] or alanine [GST-NDEL1(S251A)] did not have an obvious effect on the phosphorylation by CDK1 (Fig. 2F). As expected, mutation of the phosphorylation sites of Aurora-A or CDK1 resulted in the lack of phosphorylation of these sites (Fig. 2E and F). These findings suggest that NDEL1 is initially phosphorylated by Aurora-A, followed by downregulation. Then, distinct pools of NDEL1 may be phosphorylated by CDK1.

Aurora-A-mediated phosphorylation of NDEL1 facilitates the centrosomal targeting of the phosphorylated NDEL1.

To address the fate and subcellular localization of wild-type NDEL1 or phosphorylated NDEL1 in vivo, we generated MEF cells in which endogenous Ndel1 was replaced by wild-type GFP-Ndel1s or mutated GFP-Ndel1s. Wild-type GFP-NDEL1 or mutated GFP-NDEL1s were expressed at fairly normal levels, and endogenous NDEL1 was not detected (see Fig. S3 in the supplemental material). MEF cells were synchronized and were examined at a given cell cycle phase. Wild-type GFP-NDEL1, GFP-NDEL1(S251E), appeared centrosomal at G₂ prophase compared to GFP-NDEL1(S251A), suggesting that S251 phosphorylation facilitated the limited distribution of NDEL1 to centrosomes (Fig. 3A). GFP-NDEL1(S251E) was stable during interphase, but it was more rapidly downregulated at prophase than GFP-NDEL1. This rapid decay was efficiently blocked by MG132 (Fig. 3A). These observations suggest that the phosphorylation of NDEL1 by Aurora-A triggers the ubiquitin-mediated protein degradation and/or that the degradation machinery is activated during prophase. Interestingly, the overexpression of GFP-NDEL1(S251E) resulted in the appearance of multiple centrosomes at prophase (Fig. 3A) (see below). GFP-NDEL1(tripleS/T→E) also displayed centrosomal localization that was less restricted at the centrosome (Fig. 3B). In addition, a signal was not detected by a phospho-S251-specific monoclonal antibody, which is consistent with biochemical data, suggesting that the phosphorylation by CDK1 may inhibit further phosphorylation by Aurora-A. Although GFP-NDEL1(tripleS/T→A;S251E) and GFP-NDEL1(tripleS/T→E;S251E) localized at centrosomes during prophase, GFP-NDEL1(tripleS/T→E;S251E) appeared to be less stable than other mutants (Fig. 3C). The treatment of MEF cells with MG132 improved the stability in either case, suggesting that the phosphorylation of S251 is critical for proteasome-mediated protein degradation (data not shown). While the expression of GFP-NDEL1(tripleS/T→E;S251E) does not have an obvious effect on interphase cells, expression during mitosis results in a cell with a multiple-centrosome pattern at prophase, suggesting that double phosphorylations by Aurora-A and CDK1 are not physiological and compromise centrosomal integrity. These multicentrosomal phenotypes associated with the expression of GFP-NDEL1(tripleS/T→E;S251E) were also observed in wild-type cells (Fig. 3D), suggesting that the effect of GFP-NDEL1(tripleS/T→E;S251E) expression is dominant.

We next documented the role of Aurora-A-mediated phosphorylation in the ubiquitination of NDEL1 (Fig. 3E). We extracted proteins from synchronized MEF cells, and this was followed by immunoprecipitation and immunoblotting. The ubiquitination of wild-type NDEL1 appeared at prophase, which was coincident with Aurora-A-mediated phosphorylation. In contrast, GFP-NDEL1(S251A) appeared to be more poorly ubiquitinated, whereas GFP-NDEL1(S251E) displayed higher ubiquitination (Fig. 3D), which was enhanced by MG132 (Fig. 3E). Our findings suggest that NDEL1 phosphorylation by Aurora-A triggers recruitment and restricted distribution of NDEL1 to the centrosome at G₂ prophase, and leads to rapid degradation by the ubiquitination-proteasome pathway. This order of the phosphorylation and the degradation appears to be critical for the maintenance of centrosome integrity since a multiple-centrosome pattern was observed after the overexpression of GFP-NDEL1(S251E) or GFP-NDEL1(tripleS/T→E;S251E).

Biochemical and genetic interactions between Aurora-A, NDEL1 and TACC3.

Both NDEL1 and TACC3 are specific substrates of Aurora-A kinase (9, 19), and TACC3 is recruited to the centrosome in an Aurora-A-dependent fashion, suggesting that NDEL1 can also bind TACC3 directly. We first examined the physical interaction between NDEL1 and TACC3 by coimmunoprecipitation. HeLa cells were transfected with GFP-Ndel1 or GFP-Tacc3. An anti-GFP antibody was used for immunoprecipitation, and an anti-TACC3 antibody and an anti-NDEL1 antibody were used for detection. GFP-NDEL1 was coimmunoprecipitated with endogenous TACC3, but not with GFP used as a negative control (Fig. 4A). Conversely, when GFP-TACC3 was immunoprecipitated with an anti-GFP antibody and blotted with an anti-NDEL1 antibody, similar results were achieved (Fig. 4A). The interaction of endogenous NDEL1 and TACC3 was confirmed by immunoprecipitation using an anti-NDEL1 antibody (Fig. 4B). These results suggested that NDEL1 specifically recognizes TACC3. The direct physical interaction between NDEL1 and TACC3 in vitro was verified by GST-pull-down analysis. HIS-TACC3 was pulled down by association with GST-NDEL1 (Fig. 4C), and HIS-NDEL1 was pulled down by association with GST-TACC3 (Fig. 4C). No interaction was detected between the GST control and HIS-NDEL1 or HIS-TACC3. These results further suggested that NDEL1 bound directly with TACC3. We quantitated the affinity between NDEL1 and TACC3 by Biacore using recombinant proteins (see Fig. S4A in the supplemental material). Specific interaction between NDEL1 and TACC3 was detected at an extraordinarily strong affinity [K_A (1/M), 9.45 × 10⁹ ± 3.52 × 10⁹], an effect that was completely reproduced when the protein combinations were reversed [K_A (1/M), 1.33 × 10¹⁰ ± 0.34 × 10¹⁰] (see Fig. S4B in the supplemental material). Wild-type NDEL1 and NDEL1 mutated at sites of phosphorylation by Aurora-A and CDK1 displayed similar binding affinities to TACC3, suggesting that the phosphorylation of NDEL1 does not influence the binding between NDEL1 and TACC3 [K_A (1/M), 9.82 × 10⁹ ± 3.92 × 10⁹, GST-NDEL1(S251A)] [K_A (1/M), 8.98 × 10⁹ ± 4.01 × 10⁹, GST-NDEL1(S251E)] (see Fig. S4C in the supplemental material).

To characterize how these three proteins are related, we examined whether the inactivation of each protein influences the localization and the phosphorylation of other proteins. To evaluate the role of Aurora-A in the phosphorylation and distribution of target proteins, we used siRNA to reduce the levels of Aurora-A in HeLa cells (16, 23) and examined the phosphorylation of each protein and the centrosome morphology. In the series of experiments, we defined early prophase as the beginning of chromosome condensation with a single centrosome and prophase as the completion of chromosome condensation with separated centrosomes. Aurora-A depletion by siRNA significantly reduced the phosphorylation level and centrosomal accumulation of NDEL1 and TACC3, as in the cases of lower organisms (12) (Fig. 4D). We previously generated conditional knockout mice of Ndel1 by CRE-mediated homologous recombination (33). Therefore, we disrupted Ndel1 by the expression of the Cre gene in established MEF cells. Ndel1-disrupted MEF cells displayed various impairments of cell cycle progression, including centrosome maturation defects, separation defects, and spindle aberrations, such as weak or split spindles (data not shown). T288 phosphorylation of Aurora-A was normal, suggesting that the activation of Aurora-A does not require the presence of NDEL1. However, centrosomal targeting and S347 phosphorylation of TACC3 were severely impaired despite the normal activation of Aurora-A (Fig. 4E). Finally, we knocked down TACC3 expression by siRNA (19). The reduction of TACC3 levels resulted in misaligned chromosomes that are indicative of the TACC3 RNA interference phenotype as previously reported (19). Most interestingly, TACC3 depletion resulted in no gross abnormalities in the phosphorylation of Aurora-A or NDEL1 and normal centrosomal targeting of NDEL1 (Fig. 4F), suggesting that TACC3 likely does not substantially contribute to the activation of Aurora-A and NDEL1 at mitotic entry. Our findings suggest that NDEL1 is located at the middle of the pathway between Aurora-A and TACC3.

Phosphorylation-mimetic mutants of NDEL1 rescued the mitotic entry defect by Aurora-A depletion.

Aurora-A has been implicated in the regulation of centrosome maturation and spindle formation (1, 15, 36). We demonstrated that NDEL1 is a substrate of Aurora-A and is specifically phosphorylated at mitotic entry. We also found that NDEL1 is required for TACC3 targeting to the centrosome. These observations suggest that the phosphorylation of NDEL1 by Aurora-A might be essential for mitotic entry. To address the role of S251 phosphorylation, we examined whether the expression of Ndel1 or mutant forms of Ndel1 rescues the phenotypes generated by siRNAs against Aurora-A. First, we examined the mitotic index. A significant reduction of the mitotic index suggests that the depletion of Aurora-A markedly impaired the ability of cells to enter mitosis (16). The expression of GFP-NDEL1(S251E), GFP-NDEL1(tripleS/T→A;S251E), and GFP-NDEL1(tripleS/T→E;S251E) in Aurora-A-depleted cells restored mitotic entry, whereas Aurora-A-depleted cells expressing other constructs failed to recover (Fig. 5A). These data suggest that the phosphorylation of NDEL1 by Aurora-A supports mitotic entry. γ-Tubulin is an indicator of centrosome maturation, which is required for MT nucleation (16, 23). We explored whether γ-tubulin accumulation was recovered by the expression of GFP-Ndel1s by the examination of intensity of immunocytochemistry (Fig. 5B). A reduction of the γ-tubulin signal by Aurora-A depletion was fairly well restored by the expression of GFP-NDEL1(S251E), GFP-NDEL1(tripleS/T→A;S251E), and GFP-NDEL1(tripleS/T→E;S251E), whereas cells expressing other constructs failed to recover the accumulation of γ-tubulin (Fig. 5B). We also examined whether the expression of NDEL1 and mutated NDEL1s can rescue a lack of centrosome targeting of TACC3 after the depletion of Aurora-A by siRNA. Importantly, the recovery of centrosomal targeting of TACC3 by the expression of GFP-NDEL1(S251E), but not by GFP-NDEL1(S251A) (Fig. 5C), was observed. These findings suggest that NDEL1 is a key molecule for the recruitment of proteins to the centrosome that are downstream of Aurora-A.

Centrosome separation and chromosome condensation are important indicators of the normal progression of prophase. These two events are independently but coordinately regulated. Aurora-A depletion by siRNA leads to separation defects of centrosomes, whereas chromosome condensation is maintained fairly intact (16, 23). We selected HeLa cells with chromosome condensation as prophase HeLa cells and explored whether centrosome separation was restored by the expression of phosphorylation-mimetic NDEL1s. Centrosome separation during prophase was observed after the transfection of GFP-NDEL1(S251E) and GFP-NDEL1(tripleS/T→A;S251E). We measured the distance between two centrosomes at prophase by using an anti-centrin-3 antibody as a centriole marker (16). The results of quantitation revealed that the transfection of GFP-NDEL1(S251E) and GFP-NDEL1(tripleS/T→A;S251E) resulted in separations of 3.9 μm ± 1.1 μM and 3.3 μm ± 0.9 μM, respectively (Fig. 5D), suggesting that the expression of GFP-NDEL1(S251E) leads to more normal centrosomal separation. In contrast, clear centrosome separation was not observed after the expression of GFP, GFP-NDEL1, GFP-NDEL1(S251A), and GFP-NDEL1(tripleS/T→E). These findings indicate that the phosphorylation of NDEL1 is essential to support centrosome maturation and separation during mitosis.