Main

In the yeasts Saccharomyces cerevisiae2,3 and Schizosaccharomyces pombe4,5, both exit from mitosis and coordination of cytokinesis require the Cdc14 (Flp1) phosphatase. In mammals, distinct genes encode two members of the Cdc14 family, Cdc14A and Cdc14B (ref. 1). To investigate how human Cdc14 phosphatases function in cell-cycle progression, we generated derivatives of the U-2-OS cell line6 that conditionally express Cdc14A and Cdc14B in a tetracycline-repressible manner (Fig. 1a). Although inducing Cdc14B did not have a major effect on the distribution of cells in different parts of the cell cycle (data not shown), conditional overproduction of both Myc- and green fluorescent protein (GFP)-tagged Cdc14A led to progressive cell death, accompanied by accumulation of pre-G1 DNA fragments and gradual elimination of cells at the G2–M transition (Fig. 1b, and data not shown). The cell-death peak was preceded by abnormalities in almost 50% of mitotic cells, including multipolar mitotic spindles and lagging and missegregated chromosomes (Fig. 1c). When the transgenes were repressed, there was no measurable alteration in cell-cycle progression in any of the clones when compared to parental cells (less than 10% had abnormal cell divisions).

Figure 1: Human cells exposed to ectopic Cdc14A phosphatase undergo aberrant mitosis.
figure 1

a, Conditional expression of the indicated forms of Cdc14 was detected by immunoblotting with the 9E10 anti-Myc antibody 24 h after removing tetracycline (Tet) from the medium. b, Distribution of DNA in the cells expressing Myc–Cdc14A was monitored by flow cytometry. The numbers indicate the increase of pre-G1 DNA fragments (marked by arrows) and decrease of cells in G2–M, respectively. c, Representative chromosomal (green) and mitotic spindle (red) aberrations induced in cells exposed to Myc–Cdc14A for 48 h. Chromatin and mitotic spindles were visualized by antibodies to phosphorylated histone H3 (H3P) and the spindle motor protein (Eg5), respectively. DIC, differential interference contrast. Scale bar, 10 μm.

Full size image

The ability of Cdc14A to deregulate cell division in stages before cytokinesis and/or mitotic exit suggested that at least some features of the human phosphatase are distinct from its yeast counterpart. The appearance of multipolar mitotic spindles led us to investigate a potential link between Cdc14A and the cellular microtubule-organizing centres (MTOCs)7,8,9. An antiserum to the amino terminus of human Cdc14A specifically recognized a single band of the expected size on immunoblots of total cell extracts (Fig. 2a). Apart from a weak cytoplasmic signal, immunofluorescence analysis revealed a pronounced centrosomal staining in interphase U-2-OS cells (Fig. 2b). That these Cdc14A-decorated organelles are centrosomes was confirmed by co-localization with γ-tubulin (Fig. 2b). Although the cytosolic Cdc14A was readily detectable in dividing cells, the bulk of the centrosome-associated, Cdc14A-specific signal disappeared after entry into mitosis (Fig. 2b), indicating that a significant fraction of Cdc14A dissociates from the centrosomes at the G2–M transition. Both Myc-tagged Cdc14A analysed in fixed cells, and its GFP-tagged counterpart (GFP constructs also contain the Myc tag)in live cells, confirmed that the protein is found in the cytoplasm and on centrosomes of interphase cells (Fig. 2c) and is reduced (but not entirely absent) on spindle poles during mitosis (data not shown). Because a mouse antiserum against full-length human Cdc14A also confirmed the results obtained with the rabbit antibody (data not shown), we conclude that both endogenous and ectopic Cdc14A proteins localize to the cytosol and interact with interphase MTOCs, a pattern reminiscent of established centrosome regulators10.

Figure 2: Cdc14A localizes on centrosomes during interphase and undergoes dynamic exchange independent of microtubules.
figure 2

a, Protein extracts from exponentially growing HeLa cells were subjected to immunoblotting with the affinity-purified rabbit antibody against human Cdc14A. Where indicated, the antibody was pre-adsorbed with the antigen. b, U-2-OS cells captured in interphase or various stages of mitosis were co-stained with the affinity-purified rabbit antibody to Cdc14A (red) and mouse monoclonal antibody to γ-tubulin (green) and analysed by confocal microscopy. Yellow indicates co-localization of the two proteins. Inserts show large magnification of selected centrosomes (interphase, prometaphase) and a centrosome together with a midbody structure (telophase). c, Cell lines conditionally expressing Myc- or GFP-tagged wild-type Cdc14A were induced for 24 h. Cells expressing the GFP-tagged protein were recorded live; the Myc-tagged proteins were immunostained with the anti-Myc antibody (arrows, centrosome-associated Cdc14A). d, Cells expressing GFP-tagged phosphatase-dead (PD) Cdc14A were induced and analysed as in c. Representative interphase and metaphase cells are shown. e, FRAP analysis of the centrosome-associated fluorescence in the GFP–Cdc14A-inducible cell line 24 h after induction. Where indicated, the cells were pre-treated with nocodazole 2 h before image acquisition. The control unbleached centrosome came from the same exponentially growing cell. The data were reproduced in several other cells in independent experiments. Scale bars, 10 μm.

Full size image

Next, we determined what Cdc14A requires for its normal subcellular distribution. Interaction with centrosomes is independent of the intrinsic catalytic activity of Cdc14A, as GFP-tagged phosphatase-dead Cdc14A localized to interphase centrosomes and became reduced on mitotic spindle poles in a manner indistinguishable from the wild-type protein (Fig. 2d). Fluorescence recovery after photobleaching (FRAP)11 performed on wild-type centrosome-associated GFP–Cdc14A showed that the majority of the GFP signal on the pulse-bleached centrosome recovered quickly; this recovery occurred with similar kinetics in cells pretreated with the microtubule-depolymerizing drug nocodazole (Fig. 2e). The latter findings indicate that Cdc14A is a genuine centrosomal protein, the majority of which undergoes a dynamic, microtubule-independent turnover.

To test whether the accumulation of Cdc14A in cytosolic compartments, including centrosomes, involves regulation of protein transport across the nuclear envelope, we subjected the cells expressing GFP–Cdc14A to leptomycin B (LMB), an inhibitor of chromosomal region maintenance (CRM1)-dependent nuclear export12. Indeed, a short treatment with LMB caused the majority of Cdc14A to be redistributed from the cytosol to nucleoli (Fig. 3a). Analyses of a series of truncated versions of Cdc14A showed that deleting the region between amino acids 343 and 393 was necessary and sufficient to mimic the effect of LMB (data not shown). This region contains a small domain rich in large, hydrophobic amino acids typical of an NES13,14 (Fig. 3b). To test the function of this putative NES, we generated a cell line that conditionally expressed a NES-deficient variant of Cdc14A, in which methionine and isoleucine residues (Met362 and Ile364) were changed to alanine residues (Fig. 3c). Indeed, the induced Cdc14A(ΔNES) accumulated in nucleoli, colocalized with nucleolin (Fig. 3c, upper panels), and lost its ability to associate with centrosomes (Fig. 3c, lower panels). Identical results were obtained with another mutant of Cdc14A, in which isoleucine (Ile355) and leucines residues (Leu356 and Leu359) in the NES were substituted with alanine residues (data not shown). Although the NES-containing region of Cdc14A is required to keep Cdc14A out of the nucleus, it is insufficient to localize Cdc14A to the centrosomes (Fig. 3d). Analysis of a series of Cdc14A deletion mutants revealed that sequences both N- and carboxy-terminal of the central NES might be involved in centrosome localization. Deleting amino acids 1–343 and 393–623, respectively, reduced the ability of the truncated proteins to interact with interphase centrosomes (Fig. 3d). Therefore, the NES is a prerequisite for proper localization of Cdc14A, but the specific centrosome localization signal is complex and involves regions both N- and C-terminal of the NES.

Figure 3: A nuclear export signal (NES) prevents sequestration of Cdc14A in the nucleoli and facilitates its localization with centrosomes.
figure 3

a, GFP-associated fluorescence was recorded in cells induced to express GFP–Cdc14A with or without 1-h pre-treatment with LMB. b, Identification of the NES. Positions and identities (in bold) of the large hydrophobic amino acids essential for the Cdc14A NES (shown as φ in the NES consensus sequence), as well as the alterations made to create the NES-deficient mutant Cdc14A(ΔNES). c, Cells conditionally expressing Myc–Cdc14A(ΔNES) were induced for 24 h and immunostained with anti-Myc, anti-nucleolin or anti-pericentriolar marker CTR-453 (CTR). Yellow in merged images indicates co-localization of Cdc14A(ΔNES) with nucleolin, whereas red (arrows) indicates the absence of this mutant on centrosomes. d, Schematic representation of the Cdc14A truncation mutants (all tagged with GFP at their N termini) and their preferential subcellular localization. e, Cells conditionally expressing Myc–Cdc14B were induced for 24 h and stained with antibodies to Myc. DIC facilitates the morphological identification of the nucleoli. Scale bars, 10 μm.

Full size image

These data suggested that, in principle, human Cdc14 phosphatases could be regulated by periodic sequestration in nucleoli, similar to their yeast counterparts1,2,3,4,15. But, as demonstrated above, endogenous (as well as the ectopic) wild-type Cdc14A in fixed cells and GFP–Cdc14A in live cells always localized to centrosomes and never to nucleoli. Instead, we found that the nucleolus may harbour a distinct Cdc14 isoform that can avoid nuclear export. Conditionally expressed wild-type Cdc14B (ref. 1) accumulated in the nucleoli immediately after the induced protein became detectable (Fig. 3e). Cdc14B-associated fluorescence in the cytoplasm was negligible (Fig. 3e), Cdc14B did not colocalize with centrosomal markers, and its localization was unaltered by LMB treatment (data not shown). Whether this indicates that Cdc14B is a true functional homologue of the yeast counterparts requires further investigation. Mammalian Cdc14A, on the other hand, probably fulfils other, previously unrecognized function(s) related to the periodic activities of the major cellular MTOCs. As such, Cdc14A seems to have evolved an active, nuclear-export mechanism that prevents its nucleolar sequestration, which would otherwise lead to a functional inactivation of Cdc14A. This idea is consistent with what happens when wild-type Cdc14A is transiently overexpressed — this resulted in a ∼sevenfold increase in the amount of centrosome-bound Cdc14A, followed by a rapidly deregulated progression through mitosis in 47 ± 3% of all dividing cells (assessed by the number of multiple mitotic spindles and chromatin abnormalities; Fig. 1c). In contrast, neither NES-deficient Cdc14A nor wild-type Cdc14B (both localized in nucleoli) increased the frequency of mitotic abnormalities above the background level (7 ± 2%) observed in parental cells (data not shown). Thus, the compartmentalization of distinct mammalian Cdc14 phosphatases through selective nuclear export might ensure that Cdc14A fulfils its centrosome-associated function(s) and does not disrupt that of the nucleolar Cdc14B.

To test whether the Cdc14A phosphatase might directly regulate the centrosome-duplication cycle, we examined cell lines conditionally expressing wild-type or phosphatase-dead Cdc14A, synchronized in early S phase before the transgenes were induced. Inducing wild-type Cdc14A after release from the synchronization block and before the first wave of cell division was sufficient to generate mitotic cells with more than two centrosomes composed of both pericentriolar material and centrioles (Fig. 4a). Generation of multiple poles that can produce spindles, observed in almost 75% of all dividing cells, was indeed accompanied by a corresponding increase in the number of functional mitotic spindles (see Supplementary Information, Fig. S1), which triggered unequal partitioning of chromosomes into multiple daughter cells (see Supplementary Information, Movie 1). Therefore, the mitotic defects induced by elevated Cdc14A are not just a consequence of abnormal mitotic exit and/or cytokinesis from the preceding cell cycles, and the excess of Cdc14A could indeed interfere with physiological mechanisms restricting centrosome duplication to one round within the given cell-division cycle.

Figure 4: Cdc14A regulates centrosome splitting.
figure 4

a, U-2-OS cells expressing wild-type Cdc14A were synchronized by double-thymidine treatment, released by replating to drug-free medium and induced to express the transgene. After 14 h, the cells were fixed and immunostained with antibodies to the pericentriolar marker (CTR) and the centriole-associated protein (centrin). A representative mitotic cell with multiple centrosomes (each containing at least one centriole) induced by the ectopic wild-type Cdc14A is shown. b, U-2-OS cells expressing ectopic wild-type Cdc14A were synchronized by double-thymidine treatment and induced to express the transgene without release from the block. The cells were fixed and stained as in a 48 h after induction. Split centrosomes are indicated by arrows, paired centrosome doublets by arrowheads. c, HeLa cells were left untreated (NT), or transfected with the Cdc14A-directed siRNA duplex (14A) and control (Con) sense RNA, respectively. The endogenous Cdc14A protein was analysed by immunoblotting 48 h after transfection. Cdk7 was used as a loading control. d, A representative example of a bi-nucleated cell containing one unseparated centrosome pair (insert) induced by the Cdc14A-directed siRNA. Centrosomes were revealed by immunostaining against the pericentriolar antigen (CTR). contrast. Arrows indicate the position of the unseparated centrosome in all three images. Scale bars, 10 μm.

Full size image

Next, we assayed the consequences of inducing wild-type and phosphatase-dead Cdc14A in cells stably arrested in S-phase by an extended double-thymidine treatment. Virtually all centrosomes in control and phosphatase-dead Cdc14A-expressing cells appeared as paired doublets, each containing two duplicated centrioles (Fig. 4b), whereas inducing wild-type Cdc14A triggered centrosome splitting in 50% of these S-phase cells (Fig. 4b). Again, the separated centrosomes contained both pericentriolar material and centrioles (Fig. 4b). Several pieces of evidence indicate that the premature centrosome splitting in cells overproducing Cdc14A occurred within a genuine S phase and not as a consequence of an unscheduled exit into G2 and M. Thus, cells with separated centrosomes did not have morphological signs of mitosis, and early mitotic targets, such as histone H3, had not been phosphorylated. Moreover, DNA synthesis did not occur — regardless of the amount of ectopic Cdc14A — as long as the cells remained arrested, but resumed rapidly after the block (excess thymidine in the culture medium) was removed (data not shown). We conclude that the excess Cdc14A can induce premature centrosome splitting in the absence of mitosis.

Finally, we studied the consequences of downregulating endogenous Cdc14A. Transient transfection of siRNA oligonucleotides that target Cdc14A mRNA in HeLa cells led to a quantitative decrease in the amount of endogenous Cdc14A protein (Fig. 4c). Detailed analysis of cells exposed to Cdc14A siRNA duplexes (but not those transfected with control, sense RNA) had a complex phenotype which indicates that the cell-division and centrosome-duplication processes can be uncoupled. Flow cytometry measurements between 48 and 72 hours after treatment with Cdc14A-directed siRNA showed an increase in cells with a 4N DNA content (from 16 ± 3% to 28 ± 2%). Time-lapse videomicroscopy revealed that this was not a result of cells accumulating in G2 phase but a consequence of failure to complete cell division. Of the cells that attempted to divide in the observation period, 43 ± 6% failed to generate two separated daughter cells, whereas similar defects were detected in only 11% of the mock-treated cells. Many of the observed mitotic abnormalities culminated in a formation of bi-nucleated cells and included: an abortive anaphase attempt (see Supplementary Information, Movie 2); and a failure to undergo the cytoplasmic abscission (see Supplementary Information, Movie 3). Because the outcome of such mitotic abnormalities (cells with two sets of chromosomes) was opposite to what we observed after Cdc14A overexpression (partitioning of the chromosomal pool into three or more smaller nuclei), and because our previous data showed that the excess of Cdc14A induced premature centrosome splitting, we reasoned that the inability to execute productive mitosis in the Cdc14A-deprived cells might have been associated with the lack of centrosome separation. Indeed, immunofluorescence analysis with antibodies to both pericentriolar and centriole-specific markers revealed a decreased proportion of cells with separated centrosomes after treatment with Cdc14A-directed siRNA (6 ± 2% compared with 17 ± 4% in mock-treated cells). The difference between Cdc14A siRNA-treated and control cells was probably even more pronounced because the lack of productive cell division in the Cdc14A-deficient cells was lethal, resulting in a detachment and loss of many productively transfected cells (data not shown). Despite this, virtually all the remaining bi-nucleated cells formed as a consequence of Cdc14A deprivation (representing 10 ± 2% of the entire siRNA-treated cell population compared with less than 2% in mock-treated cells) shared one unseparated centrosome pair (Fig. 4d). The appearance of such cells is consistent with a defect in the centrosome-duplication cycle. A recent study16 showed that separation and migration of a mother centriole, and its transient interaction with the microtubule network within the midbody connecting two daughter cells, was required for cytoplasmic abscission, an event that marks the final step of cytokinesis. In our experiments, the inability to separate centrosomes might have impaired release of the mother centriole and could explain both the fusion of nearly separated cells (see Movie 3) and the significant increase in bi-nucleated cells with one shared centrosome after the siRNA-mediated downregulation of Cdc14A (Fig. 4d).

Collectively, our data provide the first evidence that a mammalian Cdc14 phosphatase is involved in regulating the centrosome-duplication cycle. The dynamic interaction of Cdc14A with centrosomes reported here may help generate a balance between centrosomal phosphorylation and dephosphorylation events that control essential regulatory processes, such as the switch between centrosome cohesion and centrosome separation. Interestingly, antibody-mediated neutralization or overexpression of the truncated form of C-Nap1, an established regulator of centrosome cohesion, resulted in a premature centrosome splitting10 similar to that observed in our cells with deregulated Cdc14A. It has been proposed that aneuploidy in at least some human cancers may be caused by centrosome malfunction7. Our findings that deregulated Cdc14A abolishes the centrosome-division cycle and leads to chromosome missegregation indicate that this phosphatase might be involved in multistep tumorigenesis17, promoting chromosomal instability.

Methods

Plasmids

Cdc14A and Cdc14B cDNAs (accession numbers AF064102.1 and AF064104.1, respectively) were obtained from H. Charbonneau and cloned into pcDNA3 (Invitrogen, Carlsbad, CA) containing an N-terminal Myc epitope tag. Myc-tagged Cdc14A and Cdc14B were moved into the tetracycline-repressible pBI vector (Clontech, Palo Alto, CA). To generate pBI-Myc-GFP–Cdc14A, enhanced GFP (EGFP) cDNA lacking a stop codon was amplified from pEGFP-N1 (Clontech) by polymerase chain reaction (PCR) with Pfu polymerase (Stratagene, La Jolla, CA) and cloned into pBI-Myc–Cdc14A. The Cdc14AΔNES (Met362→Ala/Ile364→Ala) and Cdc14APD (Cys278→Ser) mutants were generated using the QuikChange Site-Directed Mutagenesis kit (Stratagene).

Cell culture

Derivatives of the U-2-OS cell line6 conditionally expressing Cdc14 alleles were generated by transfection with the respective transgenes in pBI plasmids, and cultured as described previously6. Derepression of the transgenes was induced by culturing the cells in tetracycline-free medium for the duration specified in the figure legends. When indicated, nocodazole (Sigma) and leptomycin B (gift from M. Yoshida) was added to culture medium at a concentration of 4 μg ml−1 and 10 ng ml−1, respectively. Synchronization of cells in early S phase by double-thymidine treatment has been described previously18.

Antibodies

Bacterially purified glutathione-S-transferase (GST) protein fused to the N-terminal part of Cdc14A (residues 1–390) was used to raise an antiserum in rabbits. Detailed characterization of this and other Cdc14A reagents is provided elsewhere (B.K.K., Z.A. Zimmerman, H. Charbonneau, P.K.J., unpublished observations). Mouse antibody to nucleolin (MS-3) and goat antiserum to γ-tubulin (C-20) were from Santa Cruz (Santa Cruz, CA), and mouse antibody to Eg5 was from Transduction Laboratories (Lexington, KY). Phospho-specific rabbit antiserum to histone 3 was from Upstate Biotechnology (Charlottesville, VA). Mouse monoclonal antibodies to the Myc epitope (9E10) and the pericentriolar antigen (CTR 453) were gifts from G. Evan and M. Bornens, respectively. Rabbit serum against centrin was donated by E. Nigg.

Microscopy

Cells grown on glass coverslips were immunostained with the combinations of antibodies specified in figure legends. Fluorescence and differential interference contrast (DIC) images were captured and processed using a Zeiss 510 laser-scanning microscope mounted on Axiovert 100M. Matching confocal planes were analysed in all co-localization studies. For live imaging, cells induced to express GFP–Cdc14A were plated on Lab-Tek chambered coverglass (Nalge Nunc International, Rochester, NY), supplied with CO2-independent medium (Invitrogen), and kept on a heated stage (37 °C) throughout the image acquisition. For FRAP analysis11, a single bleach pulse was generated by 20 iterations of the 488-nm argon laser set to the maximal energy output over the entire centrosome. The total bleaching time did not exceed 500 ms. Subsequently, 20–30 single-section imaging scans were collected at a maximum rate (∼1.5 s intervals) with the laser power attenuated to 2% of the bleach intensity. FRAP curves were generated from background-subtracted fluorescence intensity values, quantified in Metamorph (Universal Imaging, Downington, PA), and expressed as relative values to pre-bleached signal (set to 100%). Each measurement was reproduced at least three times in different experiments. Time-lapse videomicroscopy was performed on cells cultured on a heated microscope stage, supplied with CO2-independent medium, and overlaid with mineral oil. DIC images were acquired every 3 min using a charge-coupled device (CCD) camera (CoolSnap, Roper Scientific, Tuscon, AZ) and subsequently processed by the Metamorph software package.

RNA interference (RNAi)

RNAi was carried out as described previously19. The sequence of the region targeted by the siRNA, 5′-GCA CAG TAA ATA CCC ACT ATT-3′, corresponded to bases 89–109 located downstream of the first nucleotide of the start codon of the Cdc14A cDNA. Transfection of HeLa cells was performed by Oligofectamine reagent (Invitrogen). Final concentration of the siRNA duplex in culture medium was 100 nM.