Cdc14B has been shown to associate with distinctive cellular structures such as the nucleolus, nuclear filament, centrosome, and the spindle midzone and midbody (Kaiser et al., 2002; Mailand et al., 2002; Nalepa and Harper, 2004; Cho et al., 2005), and to undergo active nucleocytoplasmic translocation (Bembenek et al., 2005). To explore the functional significance of centrosomal Cdc14B, we first examined whether centrosomal localization of Cdc14B was cell cycle–dependent using a chicken anti-Cdc14B antibody. After preextraction and fixation of U2OS cells, immunofluorescence analysis revealed that Cdc14B partially colocalized with centrin, a marker that stains at the distal lumen of centrioles, in addition to the nucleolar, intracellular bridge (Fig. 1 A) and midzone (not depicted) localization previously characterized using other anti-Cdc14B antibodies (Kaiser et al., 2002; Mailand et al., 2002; Nalepa and Harper, 2004; Cho et al., 2005). Depending on the centriole configuration during different stages of the cell cycle (Tsou and Stearns, 2006a; Nigg, 2007), Cdc14B associates with each of the disengaged centrioles in G1 and late M phase but only one of the engaged centrioles in S, G2, and early M phase (Fig. 1 A). Cdc14B-GFP also showed a similar cell cycle–dependent centriole localization pattern (Fig. S1, available at http://www.jcb.org/cgi/content/full/jcb.200710127/DC1). We believe that the reason we and others previously did not observe centriole-associated Cdc14B with the rabbit anti-Cdc14B antibodies (Kaiser et al., 2002; Cho et al., 2005) was that these antibodies might react with epitopes that were not exposed at centrioles under the extraction/fixation conditions. In fact, Nalepa and Harper (2004) found that the centrosomal Cdc14B only became visible after Triton X-100 extraction before fixation using another rabbit anti-Cdc14B antibody. To independently evaluate the centrosomal association of Cdc14B, we purified centrosome fractions from asynchronized HeLa cells by discontinuous sucrose gradient centrifugation according to a published protocol (Moudjou and Bornens, 1998). Proteins in each fraction were resolved by SDS-PAGE and analyzed by Western blotting. At the expected ∼60% sucrose density, fractions 14–17 contained the most abundant amounts of γ-tubulin, an archetypical centrosomal marker. In parallel, Cdc14B was highly enriched in those γ-tubulin–containing centrosomal fractions (Fig. 1 B, top). Moreover, when a purified centrosomal fraction (fraction 15) was subject to immunostaining, we found that Cdc14B but not nucleolin, a noncentrosomal protein, stained positive at the purified centrosomes (Fig. 1 B, bottom). Together, these data suggest that a fraction of Cdc14B phosphatase specifically associates with centrioles.

To investigate whether Cdc14B catalytic activity is required for Cdc14B centrosome localization, we assessed the centrosomal localization of a catalytic “dead” Cdc14B^C314S mutant (Cho et al., 2005). As shown in Fig. 2 (A and B), Cdc14B^C314S-GFP localized in nuclei/nucleoli (Cho et al., 2005) and was absent from the centrosomes in virtually all the interphase cells examined. In contrast, wild-type Cdc14B-GFP (Cdc14B^WT-GFP) was found to associate with centrosomes in ∼75% of GFP-positive interphase cells (Fig. 2 B). Although these findings suggest that catalytic activity is essential for Cdc14B centrosomal localization, its perturbation may also inhibit Cdc14B nucleocytoplasmic shuttling ability, which indirectly prevents Cdc14B from reaching centrosomes in interphase cells where the nuclear envelope is still intact. To exclude this possibility, we studied the localization of Cdc14B^C314S-GFP at the centrosomes of mitotic cells where the nuclear envelopes have disassembled. As shown in Fig. 2, no Cdc14B^C314S-GFP was found to associate with mitotic centrosomes (0%, n = 113), whereas Cdc14B^WT-GFP associated with a majority (61%, n = 374) of mitotic centrosomes examined. This finding suggests that the failure of Cdc14B^C314S centrosomal retention is not because of the lack of its nucleocytoplasmic shuttling activity and thus supports the possibility that Cdc14B centrosomal localization requires its catalytic activity. We previously demonstrated that the cytoplasmic translocation/microtubule bundling ability of Cdc14B^C314S was impaired but restored when an additional mutation (KKIR29-32AAIA, Cdc14B^KKIR) of a potential NLS was introduced to the Cdc14B^C314S mutant (KKIR29-32AAIA+C314S, Cdc14B^K&C; Cho et al., 2005). Interestingly, introduction of this mutation also restored Cdc14B^C314S localization to centrosomes (Fig. 2). Because breakdown of the nuclear/cytoplasmic barrier in mitosis did not restore Cdc14B^C314S to centrosomes (Fig. 2), this finding indicates that the additional KKIR mutation not only helps Cdc14B^C314S regain cytoplasmic translocation but also acts as a gain-of-function mutant that facilitates Cdc14B^C314S to reassociate with centrosomes.

To determine whether Cdc14B plays a role in centrosome cycle regulation, we took a Cdc14B knockdown approach. We first attempted to establish stable cell lines in which endogenous Cdc14B could be conditionally knocked down. In search of siRNA oligos to deplete the endogenous Cdc14B, we found that an oligo corresponding to nucleotide position 1,234–1,254 relative to the Cdc14B start codon was most effective in ablating Cdc14B expression. We cloned this siRNA oligo into a pSuperior-neo-GFP vector and the resulting plasmid was transfected into TREx HeLa cells to obtain Cdc14B knockdown stable clones. Western blot analysis showed a reduction of endogenous Cdc14B expression in two of the representative clones (clones No. 2 and 3) in the presence of tetracycline (Fig. 3 A). Consistently, real-time quantitative RT-PCR showed that the level of Cdc14B mRNA level was also decreased in these two clones (unpublished data). The same siRNA could also knock down Cdc14B-GFP expression in U2OS Tet-On cells, confirming the target specificity of the Cdc14B siRNA (Fig. 3 A). Moreover, immunostaining with the anti-Cdc14B antibody revealed that centrosomal Cdc14B expression was reduced in Cdc14BsiRNA cells (Fig. S2 A, available at http://www.jcb.org/cgi/content/full/jcb.200710127/DC1), though the efficiency of knockdown varied in each cell. Careful examination of those Cdc14B knockdown clones revealed a centriole amplification phenotype (Fig. 3, B and C). In comparison with the parental TREx HeLa and clone No. 1, which did not exhibit any reduction of Cdc14B expression (Fig. 3 A), clones No. 2 and 3 showed two- to threefold increases in the number of cells with more than four centrioles (Fig. 3, B and C) judged by immunostaining with an anti-centrin antibody. The centriole number usually ranged from 5 to 20 with the majority around six to eight centrioles per cell. Similar results were obtained (Fig. S2 B) when γ-tubulin was used as a centrosome marker (Fig. S2 C). To ascertain the specificity of Cdc14B knockdown phenotype, we tested a Cdc14B siRNA oligo pool (SMARTpool containing four different Cdc14B siRNA oligos) and a small hairpin RNA (shRNA) designated as pSuperCdc14BshRNA-E9, each targeting different regions of Cdc14B mRNA sequences in transient and stable transfection studies, respectively. Western blot analysis confirmed that both the siRNA pool and the shRNA specifically knocked down Cdc14B but not its Cdc14A paralogue in HeLa cells (Fig. 3 D, top). Moreover, knockdown of Cdc14B by both the siRNA pool and shRNA led to a significant accumulation of cells with more than four centrioles in HeLa cells (Fig. 3 D, bottom). Although HeLa cells have been widely used for centrosome studies, it is a tumor cell line. To examine the effect of Cdc14B on centrosome cycle control in normal cells, we transfected normal human fibroblast BJ and MRC-5 cells with the Cdc14B siRNA SMARTpool. As shown in Fig. 4, the centriole amplification phenotype was faithfully reproduced in the normal fibroblast cells depleted of Cdc14B. These results demonstrate that depletion of Cdc14B leads to centrosome amplification in both normal and transformed cells.

To determine if the observed supernumerary centrioles may derive from centrosome fragmentation, we carefully evaluated the γ-tubulin–decorated centrosomes in HeLa cells transiently transfected with the Cdc14BsiRNA SMARTpool. We found that a majority of the supernumerary centrosomes were clustered together and similar in size (Fig. 3 E). All the supernumerary centrosomes (n = 105) contained centrin-labeled centrioles and many of the centrioles were in pairs. These data suggest that the supernumerary centrosomes may not arise from centrosome fragmentation.

Examination of TREx-HeLa Cdc14BsiRNA stable cells revealed a slight increase in the number of polyploid cells, such as multinucleate and macronuclear cells (Fig. S3 A, available at http://www.jcb.org/cgi/content/full/jcb.200710127/DC1). Macronuclear and multinucleate cells were determined by visual judgment of the sizes (nuclear diameter and area) of DAPI-stained nuclei and the number of nuclei per cell, respectively. It is important to note that among the TREx-HeLa Cdc14BsiRNA stable cells with centriole amplifications, only ∼0.2 (n = 300) were polyploid. Using the same criteria, we were unable to detect any significant increase of polyploid cells in HeLa cells transiently transfected with Cdc14BsiRNA SMARTpool (Fig. S3 B). In line with this finding, flow cytometry analysis did not reveal any significant increase of polyploidy in those Cdc14B knockdown cells (Fig. S3 C). Moreover, none of the Cdc14B-depleted BJ (0%; n = 115) or MRC-5 (0%; n = 74) cells with more than four centrioles were polyploid. Thus, the centriole amplification phenotype in Cdc14B knockdown cells may not be the product of aborted cell division, and the slight increase in the number of polyploid cells in the TREx-HeLa Cdc14BsiRNA stable clones may have evolved during the course of stable clone selection.

In certain transformed cells, such as U2OS cells, prolonged S phase arrest by hydroxyurea (HU) causes multiple rounds of centriole duplication in the absence of DNA replication and mitotic division (Balczon et al., 1995; Chang et al., 2003). We therefore used this well-established centriole overduplication system to directly evaluate whether Cdc14B plays a role in the regulation of centriole duplication. For this purpose, we first tested if Cdc14B overexpression could inhibit HU-induced centriole overduplication in the stable doxycycline (DOX)-inducible Cdc14B^WT-GFP U2OS Tet-On cells. Based on the localization of Cdc14B^WT-GFP at centrioles (Fig. 2), the Cdc14B^WT-GFP–positive cells can be divided into two groups: Cdc14B^WT-GFP at centrioles and not at centrioles. This provides an ideal system to directly examine whether localization of Cdc14B at centrioles conveys a critical function in centriole duplication. The number of centrin-labeled centrioles was counted in the inducible Cdc14B^WT-GFP cells after cultivation in the presence of HU without DOX induction, and HU + DOX for 72 h. As expected, HU treatment led to centriole amplification in uninduced Cdc14B^WT-GFP cells (Fig. 5 A, bottom) or mock-transfected U2OS Tet-On cells (Fig. 5 B). Remarkably, this centriole amplification phenotype was significantly attenuated in cells where Cdc14B^WT-GFP was at centrioles but not in cells where Cdc14B^WT-GFP was not at centrioles (Fig. 5 A). This finding was confirmed in the same sets of cells treated with HU + DOX for 72 h using γ-tubulin as a centrosome marker (Fig. S4, A and B, available at http://www.jcb.org/cgi/content/full/jcb.200710127/DC1). Similar and yet more dramatic results were obtained in transient transfection experiments in which cells transfected with DOX-inducible Cdc14B^WT-GFP exhibited an even more significant reduction of HU-induced centrosome amplification in comparison with the mock-transfected cells (Fig. 5 B). Together, these results suggest that Cdc14B overexpression suppresses abnormal centriole amplification, and this inhibitory function requires the presence of Cdc14B-GFP at centrioles in the HU-induced centriole overduplication system. Fluorescence-activated cell sorting analysis revealed that induction of Cdc14B-GFP did not perturb cell cycle progression at the expression level set by the DOX-inducible promoter (Fig. 5 C), which suggests that the inhibition of centriole amplification in HU-arrested cells was not the result of a potential G1 cell cycle arrest.

To investigate whether inhibition of centriole overduplication in HU-arrested U2OS cells requires the presence of active Cdc14B at centrioles, we took advantage of and tested the centriole-bound catalytic dead Cdc14B^K&C-GFP mutant as described in Fig. 2. When the Cdc14B^K&C-GFP cells were exposed to HU, the DOX-induced Cdc14B^K&C-GFP was unable to block centriole overduplication, whereas the centriole-associated catalytic “active” Cdc14B^KKIR mutant (Fig. 2) was as potent as its wild-type counterpart in inhibition of centriole overduplication (Fig. 5 A). Similar results were obtained when the HU-treated cells were examined by γ-tubulin staining (Fig. S4, A and B). Western blot revealed that all the Cdc14B-GFPs were expressed at the comparable level after Dox induction (Fig. S4 C), indicating that the failure of Cdc14B^K&C-GFP to inhibit centriole overduplication was not caused by its expression level. This finding strongly argues that Cdc14B phosphatase activity is indispensable for the inhibition of centriole duplication and that docking a catalytically inactive Cdc14B-GFP to centrioles is not sufficient to prevent centriole overduplication.

It has been documented that treatment of U2OS cells with a proteasome inhibitor Z-L₃VS causes multiple daughter centriole growth from a single mother centriole template (Duensing et al., 2007). This aberrant daughter centriole overduplication requires cyclin E/Cdk2 and Plk4, the known positive regulators of centriole duplication (Nigg, 2007). We thus used the Z-L₃VS–induced centriole duplication system to evaluate the possibility of Cdc14B as a counterbalancing enzyme of cyclin E/Cdk2 and/or Plk4 in centriole duplication control. In the absence of Cdc14B^WT-GFP induction (−DOX), treatment of U2OS Tet-On–Cdc14B^WT-GFP cells with Z-L₃VS evoked an aberrant centriole overgrowth, whereas in the presence of Cdc14B^WT-GFP induction (+DOX), centriole-bound Cdc14B^WT-GFP significantly attenuated the centriole overduplication phenotype (Fig. 6). Moreover, similar to our observation in the HU experiment (Figs. 5 and S4, A and B), the centriole-bound catalytic dead Cdc14B^K&C mutant failed to prevent Z-L₃VS–induced centriole overduplication (Fig. 6), indicating that this inhibition requires Cdc14B catalytic activity. Although additional experiments are required, our study supports the possibility that Cdc14B may counterbalance centrosomal kinases required for centriole overduplication in the Z-L₃VS induction system.

U2OS Tet-On (BD Biosciences), HeLa, and TREx HeLa (Invitrogen) cells were cultured as described previously (Cho et al., 2005). Normal human fibroblast BJ and MRC-5 cells were obtained from the American Type Culture Collection and maintained in Dulbecco's modified Eagle's medium supplied with 10% fetal bovine serum, 50 U/ml penicillin, and 50 μg/ml streptomycin. To establish DOX-inducible Cdc14B-GFP stable clones, the pBI-tet-Cdc14B^WT, pBI-tet-Cdc14B^KKIR, pBI-tet-Cdc14B^C314S, or pBI-tet-Cdc14B^K&C plasmid (Cho et al., 2005) was cotransfected with pBabe-puro into U2OS Tet-On cells harboring endogenous Cdc14B by FuGene6 (Roche). Stable clones were obtained after selection with 500 μg/ml G418 and 2 μg/ml puromycin in the absence of DOX. For the prolonged S phase arrest experiment, the Cdc14B-GFP stable clones were treated with or without 4 μg/ml DOX, 2 mM HU (Sigma-Aldrich) alone, or 2 mM HU and 4 μg/ml DOX together for 72 h. For the Z-L₃VS experiment, Cdc14B-GFP stable clones were treated with 1 μM of Z-L₃VS (BIOMOL International, L.P.) in the presence or absence of 4 μg/ml DOX for 48 h. For cell cycle analysis, cells were stained with a buffer containing 1 μg/ml propidium iodide, 0.6% NP-40, 0.1% sodium citrate, and 20 μg/ml RNase A and analyzed by a fluorescence-activated cell sorter (Guava Technologies, Inc.). The percentage of cells in difference phases of the cell cycle was determined using ModFit LT software (Verity Software House).

Cdc14B protein was depleted using the following three approaches. First, siRNA oligos that target the sequences 5′-GAACCCGAACCGTACAGTG-3′ (1,234–1,254, relative to the start codon of Cdc14B) were annealed and cloned into the HindIII and BglII sites of pSuperior-neo-GFP vector (Oligoengine). Tetracycline-inducible cell lines expressing pSuperiorCdc14B siRNA were generated by cotransfection with pBabe-puro into TREx-HeLa cells that contain Tet repressors. Stable clones were established by selection in a growth medium containing 800 μg/ml G418 and 1 μg/ml puromycin in the absence of tetracycline. Second, an shRNA expression vector carrying a Cdc14B shRNA oligo (989–1,017, relative to Cdc14B start codon) was obtained from Open Biosystems and subcloned into the pSuper-neo-GFP vector (Oligoengine), which was designated as pSuper-Cdc14BshRNA-E9. Stable transfectants were obtained by cotransfection with pBabe-puro into HeLa cells followed by selection with 800 μg/ml G418 and 1 μg/ml puromycin. Third, Cdc14B siRNA SMARTpool (targeting four different regions of Cdc14B mRNA sequences) and an siGLO control oligo were obtained from the siGENOME collection (Thermo Fisher Scientific) and transfected into HeLa, BJ, and MRC-5 cells using siPORT NeoFX reagent (Ambion) according to the manufacturer's instructions followed by a 66-h incubation. Cell lysates were then prepared and subjected to SDS-PAGE. Western blots were performed using anti-Cdc14A (Invitrogen), -Cdc14B (Invitrogen), -GFP (BD Biosciences), and –β-actin (Cytoskeleton, Inc.) antibodies.

For indirect immunofluorescence, cells were grown on glass coverslips and fixed with either paraformaldehyde or cold 100% methanol. The cells were then permeabilized with PBS/0.5% Triton X-100 for 10 min followed by blocking with PBS/1% BSA for 30 min. Centrosomes or centrioles were visualized by immunostaining with antibodies against γ-tubulin (GTU-88 [Sigma-Alrich] and C-20 [Santa Cruz Biotechnology, Inc.), pericentrin (Abcam), or centrin (20H5; a gift from J. Salisbury, Mayo Clinic Foundation, Rochester, MN). For visualization of centriole-associated endogenous Cdc14B, cells were treated with or without 10 μg/ml nocodazole (for better exposure of centrosomal Cdc14B) for 2 h, briefly extracted with 0.5% Triton X-100 on ice, fixed with cold 100% methanol, and immunostained with chicken anti-Cdc14B antibody (GenWay Biotech, Inc.). In our hands, methanol fixation preserved GFP signals and, thus, anti-GFP immunostaining was not used to visualize Cdc14B-GFP fusion proteins in the fixed cells. Secondary antibodies including Alexa Fluor 488 and 594 donkey anti–mouse and anti–goat and goat anti–rabbit, and anti–chicken IgY antibodies were obtained from Invitrogen. DNA was counterstained by DAPI. Cells were visualized with a 100× Plan Neofluar objective (1.30 oil; ∞/0.17; Carl Zeiss, Inc.) under an epifluorescence microscope (Axioskop 2; Carl Zeiss Inc). Images were acquired with a charge-coupled device camera (AxioCam HRC; Carl Zeiss, Inc.) controlled by Openlab software (version 3.5; PerkinElmer). For confocal microscopy, images were captured with an HCX-PL/APO 63 X 1.32 oil objective (Leica) under a SP2 laser scanning confocal microscope (Leica) equipped with confocal software (LCS version 2.0; Leica). The coverslips were mounted using PermaFlour Mountant media (Thermo Fisher Scientific) and imaging was performed at room temperature. Image processing was performed using Photoshop CS (8.0).

Centrosomes were prepared from exponentially growing HeLa cells according to a previously published procedure (Moudjou and Bornens, 1998). In brief, a total of 6 × 10⁸ cells were treated with 1 μg/ml cytochalasin D and 2.2 μM nocodazole. Cells were lysed in a buffer containing 1 M Hepes, pH 7.2, 0.5% NP-40, 0.5 mM MgCl₂, 0.1% β-mercaptoethanol, and protease inhibitor cocktail (Roche) and centrifuged at 2,500 g. The resulting supernatant was filtered, incubated with 2 U/ml DNase I, and loaded over a 60% sucrose cushion for centrifugation at 10,000 g with a SW28 rotor (Beckman Coulter). Concentrated centrosomes were centrifuged again over a discontinuous gradient containing 70, 50, and 40% sucrose solutions at 75,000 g. A total of 32 fractions were collected from the bottom of the tube. Each fraction was separated by SDS-PAGE. Centrosome-enriched fractions were determined by immunoblotting with anti–γ-tubulin antibody, and the presence of Cdc14B was judged by immunoblotting with anti-Cdc14B antibody (Invitrogen). For immunofluorescence analysis of isolated centrosomes, each 10 μl of fraction 15 was diluted into 4 ml of 10 mM Pipes buffer, pH 7.2, and transferred into a 38.5-ml ultracentrifuge tube (Beckman Coulter) with a specially designed adaptor to fit to a 15-mm round coverslip. The samples were then subjected to centrifugation at 20,000 g (10,000 rpm) for 20 min with a SW28 rotor followed by fixation in methanol at −20°C for 10 min and immunostaining with antibodies against centrin, Cdc14B (GenWay Biotech, Inc.), and nucleolin (4E2; Research Diagnostics, Inc.), respectively. Finally, coverslips were placed in pure ethanol for 2 min at room temperature, air-dried, and mounted with a drop of Permaflour (Thermo Fischer Scientific) on a microscope slide. Images were captured as described in the preceding paragraph.

Fig. S1 shows that a fraction of Cdc14B-GFP associates with centrioles. Fig. S2 shows the specificities of Cdc14B and γ-tubulin antibodies, and centrosome amplification in Cdc14B knockdown cells. Fig. S3 shows polyploidy cells in Cdc14BsiRNA stable clones. Fig. S4 shows that Cdc14B catalytic activity is required to prevent HU-induced centrosome overduplication. Online supplemental material is available at http://www.jcb.org/cgi/content/full/jcb.200710127/DC1.