Human Cdc14B Promotes Progression through Mitosis by Dephosphorylating Cdc25 and Regulating Cdk1/Cyclin B Activity

Entry into and progression through mitosis depends on phosphorylation and dephosphorylation of key substrates. In yeast, the nucleolar phosphatase Cdc14 is pivotal for exit from mitosis counteracting Cdk1-dependent phosphorylations. Whether hCdc14B, the human homolog of yeast Cdc14, plays a similar function in mitosis is not yet known. Here we show that hCdc14B serves a critical role in regulating progression through mitosis, which is distinct from hCdc14A. Unscheduled overexpression of hCdc14B delays activation of two master regulators of mitosis, Cdc25 and Cdk1, and slows down entry into mitosis. Depletion of hCdc14B by RNAi prevents timely inactivation of Cdk1/cyclin B and dephosphorylation of Cdc25, leading to severe mitotic defects, such as delay of metaphase/anaphase transition, lagging chromosomes, multipolar spindles and binucleation. The results demonstrate that hCdc14B-dependent modulation of Cdc25 phosphatase and Cdk1/cyclin B activity is tightly linked to correct chromosome segregation and bipolar spindle formation, processes that are required for proper progression through mitosis and maintenance of genomic stability.


Introduction
Cdk1/cyclin B is the central kinase that promotes entry into and progression through early stages of mitosis by triggering a variety of mitotic events, such as breakdown of the nuclear envelope, condensation of chromosomes, and assembly of the mitotic spindle [1][2][3]. During interphase, Cdk1 activity is downregulated by Wee1-and Myt1-dependent phosphorylation of conserved threonine T14 and tyrosine Y15 residues in the ATP-binding domain of Cdk1 [1]. Activation of Cdk1/cyclin B is achieved by a complex mitotic entry network, which consists of several feedback loops. Through a central feedback loop Cdk1 is activated by Cdc25-dependent dephosphorylation at pT14/pY15. Once activated, Cdk1/cyclin B phosphorylates Cdc25 phosphatases as well as Myt1 and Wee1 kinases, augmenting Cdc25 activity and repressing Myt1 and Wee1. Cdk1/cyclin B activity and mitotic entry are further controlled by additional superimposed feedback mechanisms upregulating the mitotic kinases Plk1 and Aurora A, and the Aurora/Plk1 activator Bora, coupling mitotic entry to centrosome maturation, and stimulating expression of proteins of the mitotic entry network, such as cyclin B [4].
Cdc25 phosphatases are highly conserved among eukaryotes. The three isoforms of mammalian Cdc25, A, B and C, are all regulated by reversible phosphorylation, phosphorylation affecting their enzymatic activity, intracellular localisation and stability [5][6][7][8][9]. Consistent with the notion that phosphorylation-dependent activation of Cdc25s is part of the central positive feedback amplification loop that increases Cdk1/cyclin B activity and promotes entry into mitosis, ablation of Cdc25A and B delays G 2 / M transition [8][9][10], whereas overexpression induces premature activation of Cdk1 and accelerates entry into mitosis [8,[10][11]. In contrast to Cdc25A and B, Cdc25C alone is not sufficient for mitotic entry [10]. During M/G 1 transition down-regulation of Cdk1 activates Wee1 and Myt1 kinases and inhibits Cdc25 phosphatases [12]. Thus, completion of mitosis depends on dephosphorylation and inactivation of Cdc25 and Cdk1/cyclin B as well as reversal of Cdk1/cyclin B-dependent phosphorylations.
Release from the nucleolus is also triggered by the G 2 -DNA damage checkpoint leading to hCdc14B-induced activation of APC/C Cdh1 [29]. Moreover, both hCdc14A and hCdc14B have been implicated in centriole amplification [30] and DNA repair [31]. Given the evolutionary conservation of basic biological mechanisms, one would anticipate that hCdc14B, like yCdc14, controls processes that trigger progression through mitosis. In support of this view, hCdc14B has been shown to modulate the assembly and disassembly of the mitotic spindle by bundling and stabilizing microtubules, yet apparently independent of its catalytic activity [28]. In addition, hCdc14B reverses mitotic phosphorylations on SIRT2 and Skp2, thereby triggering proteasomedependent degradation of these proteins, and promoting mitotic exit and entry into G 1 -phase [25,32].
In this study, we have examined the role of hCdc14B during mitosis. We show that hCdc14B is associated with nucleolar chromatin during interphase, released at prometaphase and rebound in early G 1 . RNAi-induced depletion of hCdc14B causes errors in chromosome segregation, metaphase delay, multipolar spindles, and cell death due to accumulation of multiple mitotic defects. hCdc14B dephosphorylates and inactivates the mitotic inducer Cdc25, enabling efficient inactivation of Cdk1 at late Mphase. Together, our results show for the first time that hCdc14B serves an important role in mitotic progression, regulating the activity of Cdc25s and Cdk1/cyclin B.

Unscheduled Expression of hCdc14B disturbs Timing of Mitosis
Immunofluorescence studies have demonstrated that hCdc14B localizes within nucleoli in interphase cells [23,28]. Fractionation of extracts from synchronized HeLa Kyoto cells that stably express GFP-tagged histone H2B (H2B-GFP) showed that hCdc14B was associated with chromatin during interphase and was released into the soluble fraction at prometaphase (Fig. 1A). Re-association with chromatin started around telophase and was completed after decondensation of chromatin in early G 1 (Fig. 1A, right panels and bottom images). In contrast, virtually all hCdc14A was contained in the soluble cytoplasmic fraction throughout the cell cycle and was not detectable in the nucleoplasmic or chromatin fraction (Fig. 1A). The cell cycle-dependent, dynamic interaction of hCdc14B with nucleolar chromatin suggests that nucleolar sequestration serves a regulatory function in cycling cells similar to yeast Cdc14, restricting hCdc14B activity from early mitosis to telophase.
To examine a potential role of hCdc14B in mitosis, we generated U2OS cell lines that conditionally express Flag-tagged hCdc14B (U2OS-fl-hCdc14B) or the catalytically inactive mutant (U2OS-fl-hCdc14B SL ). Clonal cell lines were synchronized by double thymidine block, hCdc14B expression was induced concomitant with release from the G 1 /S arrest (Supp. Fig. S1) and cell cycle progression was determined by flow cytometry analyses (FACS) (Fig. 1B). The FACS profiles were similar in parental and fl-hCdc14B cell lines until the cells reached the G 2 / M-phase (12 h after release), indicating that overexpression of wildtype or mutant hCdc14B had not affected progression through S-and G 2 -phase. However, re-entry into the next G 1 -phase was delayed in cells expressing fl-hCdc14B. While in parental cells progression through G 2 /M-phase and entry into the next G 1phase was completed after 16-18 h, a significant fraction of cells expressing fl-hCdc14B (38%) and fl-hCdc14B SL (21%) were still in G 2 /M 16 h after release indicating aberrant timing of mitosis.
To investigate whether delayed re-entry into G 1 was due to delay in mitotic entry or prolonged mitosis, we monitored the phosphorylation state of pY15-Cdk1 on immunoblots. In parental and U2OS-fl-hCdc14B SL cells, pY15-Cdk1 phosphorylation was low at 12 h after release from G 1 /S, consistent with a timely entry into mitosis. In contrast, in cells overexpressing wildtype fl-hCdc14B, pY15 levels remained elevated up to 13 h (Fig. 1C). This suggests that increased levels of hCdc14B prevented timely dephosphorylation and activation of Cdk1, thereby delaying entry into mitosis. Consistent with the FACS data showing prolonged G 2 /M and delayed entry into G 1 , there was no de novo phosphorylation of Cdk1 up to 18 h after release from G 1 /S in fl-hCdc14B cells. Moreover, cell proliferation was reduced in clonal cell lines expressing fl-hCdc14B, proliferation inversely correlating with the level of ectopic fl-hCdc14B (Supp. Fig. S2A and B). In contrast, overexpression of mutant fl-hCdc14B SL did not impair dephosphorylation of Cdk1 on Y15, but mitotic phosphorylation of H3-pSer10 was extended over a longer period of time (3 h compared to parental cells) (Fig. 1C), indicating that either progression through mitosis or mitotic exit were delayed.

Depletion of hCdc14B Impairs Progression through Mitosis
Given that overexpression of hCdc14B impaired progression through mitosis, we next analyzed the mitotic phenotype caused by ablation of hCdc14B. Treatment of HeLa Kyoto cells with hCdc14A-or hCdc14B-specific siRNAs for 72 h reduced the level of hCdc14A-or hCdc14B-mRNAs to 35-60% (Fig. 2D). Immunofluorescence analysis of cycling cells treated with Cdc14A-or Cdc14B si-RNAs revealed that pSer10-H3-positive, mitotic cells were enriched up to 3-fold upon depletion of hCdc14B compared to cells treated with control siRNA or hCdc14A-specific siRNA ( Fig. 2A and C). hCdc14A-and hCdc14B-specific siRNAs caused a drop in the overall cell number (data not shown), suggesting that both isoforms are required for proper cell proliferation. Consistent with results of Mailand et al. [24], the number of bi-and multinucleated cells was significantly increased in hCdc14Adeficient cells ( Fig. 2B and C). Notably, binucleation was also observed upon knockdown of hCdc14B, albeit at lower frequency compared to hCdc14A-depleted cells ( Fig. 2B and C). This suggests that knockdown of hCdc14A primarily perturbed cytokinesis without affecting progression through mitosis, whereas depletion of hCdc14B also impaired events required for progression through mitosis depletion increasing the amount of mitotic cells.
We next examined mitotic progression upon depletion of hCdc14B in synchronized HeLa Kyoto cells (Fig. 3A). After release from the second G 1 /S-arrest, progression into mitosis was monitored by live microscopy visualizing chromosome condensation via H2B-GFP-labeled chromatin. Representative images of the phenotypic responses from RNAi with an hCdc14B-specific siRNA are shown in Figure 3B. Cells entered mitosis 12 h after release, regardless whether or not hCdc14B was depleted. While most of control cells re-entered a new division cycle at 16 h after release from G 1 /S, a significant proportion (48%) of hCdc14Bdepleted cells was enriched in metaphase ( Fig. 3B and Table 1), suggesting that hCdc14B activity is required for successful completion of mitosis. Consistent with the delay of metaphase to anaphase transition, prolonged mitotic hyperphosphorylation of Cdc27/APC3 [33] was observed (Fig. 3C). Moreover, 24 h after release a large number of hCdc14B-siRNA treated cells stained positive in the TUNEL assay with cleavage of PARP-1 ( Fig. 3D and Table 1), indicating that the strong delay in mid-/late-mitosis has triggered apoptosis. Again and consistent with knockdown of hCdc14B in non-synchronized cells (Fig. 2), the amount of bi-and multinucleated cells was increased (Supp. Fig. S3 and Table 1). Together, these results strongly suggest that downregulation of hCdc14B impairs metaphase-anaphase transition mitotic delay leading to cell death.

Silencing of hCdc14B Causes Chromosome Mis-segregation and Multipolar Spindles
To investigate whether silencing of hCdc14B had perturbed spindle assembly or chromosome segregation, immunofluorescence microscopy was performed with CREST and anti-alphatubulin antibodies. Immunostaining of mitotic spindles in U2OS cells depleted of hCdc14B by hCdc14B-shRNA-2 or hCdc14B-shRNA-3 ( Fig. 4E) revealed multi-polar spindles at frequencies that were significantly higher (24-28%) than in cells transfected with an unrelated Ctrl-shRNA (2.5%). Most of the aberrant spindles were tri-or tetrapolar ( Fig. 4A and D), although some cells with even more than 5 spindle poles were observed. In addition, hCdc14B-depletion increased the number of bipolar anaphase cells with lagging chromosomes (18-25%), indicative of chromosome missegregation ( Fig. 4B and D). Again, while frequency of binucleation was low in control cells, it was significantly elevated in hCdc14B-depleted cells (Fig. 4C). These analyses indicate that silencing of hCdc14B caused problems in proper segregation of chromosomes, which in turn may explain the delay in mid-mitosis observed in HeLa Kyoto cells depleted of hCdc14B (see Fig. 3), and caused defects in bipolar mitotic spindle assembly, which is consistent with previous studies demonstrating a role of hCdc14B in bundling of microtubules [28] and in centriole amplification [30].

hCdc14B Regulates Cdk1/Cyclin B Activity
Reversal of mitotic Cdk1-dependent phosphorylations and inhibition of Cdk1 are hallmarks of Cdc14 function in yeast [18,34]. To decipher a functional link between human Cdc14B and Cdk1/cyclin B, we examined whether changes in hCdc14B levels would affect Cdk1 activity. For this, we compared activation of Cdk1/cyclin B in U2OS cell lines conditionally expressing fl-hCdc14B or fl-hCdc14B SL . After induction of wildtype or mutant fl-hCdc14B in G 1 /S arrested cells (0 h), kinase activity of immunopurified Cdk1/cyclin B was assayed at distinct times after release. In parental cells, Cdk1/cyclin B activity reached maximal   Table S1 for sequence information) and synchronized as shown in the scheme. 8 h after release from the second G 1 /S arrest, cells were lysed and the amount of hCdc14B, hCdc14A, H2B-GFP, and actin was analyzed on immunoblots. B. Phase contrast and H2B-GFP fluorescence images of HeLa Kyoto cells that were transfected with control-siRNA (siCtrl) or hCdc14B-siRNA (siCdc14B-3) and released from G 1 /S according to the scheme above. Before the second G 1 /S arrest, equal numbers of cells were seeded onto coverslips. The levels after 12-13 h, and then declined because of cyclin B degradation (Fig. 5A, left). In contrast and consistent with data shown in Figure 1B, activation of Cdk1 was delayed in cells expressing fl-hCdc14B, reaching maximal levels at 15 h, which declined more slowly than in parental cells (Fig. 5A, middle). The delay in Cdk1 activation correlated with elevated levels of pY15 (Fig. 5A, middle). Cdk1 activity was high over an extended period in cells expressing the catalytically inactive mutant fl-hCdc14B SL (Fig. 5A, right), suggesting that hCdc14B phosphatase is required for timely inactivation of Cdk1/cyclin B.
To further corroborate this result, Cdk1/cyclin B activity was measured in synchronized cells depleted of hCdc14B. Treatment with hCdc14B-siRNAs increased Cdk1/cyclin B activity almost 2fold compared to control (Fig. 5B). Consistent with hCdc14B promoting activation of APC Cdh1 [29], cyclin B levels were elevated upon depletion of hCdc14B correlating with increased Cdk1 activity ( Fig. 5B and C) implying that degradation of cyclin B was impaired upon hCdc14B depletion. Significantly, hCdc14B did not affect Cdk1/cyclin B activity in vitro (Supp. Fig. S4), suggesting that hCdc14B does not directly target Cdk1/cyclin B, but rather activates proteins and regulators that act upstream of Cdk1/cyclin B.

hCdc14B Dephosphorylates and Inactivates Cdc25
Phosphorylation of Cdc25 promotes activation of Cdk1, inactivation of Cdc25 correlating with repression of Cdk1 activity [35]. We therefore investigated whether hCdc14B targets Cdc25 phosphatases, dephosphorylation of Cdc25 proteins by hCdc14B promoting inactivation of Cdk1 and completion of mitosis. Indeed, wildtype but not mutant hCdc14B or alkaline phosphatase (CIAP) efficiently dephosphorylated all three isoforms of Cdc25 in vitro (Supp. Fig. S5A and B). To prove that hCdc14B also targeted Cdc25s in vivo, cells co-expressing fl-hCdc14B and the different isoforms of Cdc25 were arrested in mitosis, and phosphorylation of Cdc25A, B and C was analyzed on immunoblots. Consistent with the in vitro data, fl-hCdc14B dephosphorylated all three isoforms of Cdc25, shifting the hyperphosphorylated to the faster migrating hypophosphorylated form (Fig. 6A). Notably, siRNA-mediated depletion of hCdc14B increased the level of hyperphosphorylated, active Cdc25A, B and C (Fig. 6B) demonstrating a link between hCdc14B and the phosphorylation level of all three Cdc25 isoforms.
The functional interplay between Cdc25 and hCdc14B suggests that the two phosphatases may interact in vivo with each other. Indeed, co-immunoprecipitation experiments demonstrated that hCdc14B interacts with all three Cdc25 isoforms (Fig. 6C). Notably, these interactions are direct as shown by GST-pull down assays using GST-Cdc25 proteins and in vitro translated hCdc14B (Supp. Fig. S7). To demonstrate that hCdc14B inactivates Cdc25, we immunopurified Cdc25A from non-induced and induced U2OS-fl-hCdc14B cells, and assayed Cdc25A activity by measur- HeLa Kyoto cells transfected with hCdc14B-specific siRNA-3 (siCdc14B-3) or non-targeting siRNAs (siCtrl) were synchronized as shown in Fig. 3A  GFP images represent magnifications of regions marked by rectangles in the phase contrast images. C. Western blot analysis of Cdc27/APC3, hCdc14B, and b-tubulin in lysates from cells treated with GFP-specific (siGFP) or hCdc14B-specific (siCdc14B-3) siRNAs. The hyperphosphorylated form of Cdc27/APC3 is indicated. D. hCdc14B depletion induces apoptosis. siRNA transfections and cell synchronization was done as described above. Before the second thymidine block, equal numbers of cells were seeded on coverslips and apoptosis was monitored 24 h after release from the second G 1 /S-arrest in control (siGFP) or hCdc14B-depleted (siCdc14B-3) HeLa Kyoto cells by TUNEL staining (left) and on immunoblots monitoring caspase-induced proteolysis of PARP-1 (right ing its ability to activate Cdk1 that was isolated from S-phase cells. As shown in Figure 6D, overexpression of fl-hCdc14B reduced Cdk1-dependent phosphorylation of histone H1 to background levels, indicating that hCdc14B has inactivated Cdc25A (upper panel, lane 3). In contrast, Cdc25A isolated from non-induced cells was twice as active towards stimulating Cdk1 activity as shown by increased phosphorylation of histone H1 (upper panel, lane 2). Collectively, the results indicate that hCdc14B inactivates Cdc25 phosphatases, thereby counteracting unscheduled activation of Cdk1, which in turn promotes completion of mitosis.

Discussion
Though several studies have demonstrated the essential role of yeast Cdc14 for reversal of Cdk-mediated mitotic phosphorylations and mitotic exit [16], it remained to be established whether mammalian Cdc14 homologues serve a similar function. In this study, we present several lines of evidence showing that human Cdc14B plays a key role in regulation of Cdk1/cyclin B activity and progression through mitosis. First, activation of Cdk1/cyclin B and entry into mitosis is delayed in cells overexpressing hCdc14B, the delay of mitotic entry correlating with inhibitory phosphorylation of Y15/Cdk1. Second, hCdc14B dephosphorylated and inactivated mitotic Cdc25 proteins, the phosphatases that remove inhibitory phosphates from T14/Y15 and activate Cdk1 at G 2 /M transition. Third, depletion of hCdc14B by siRNA led to hyperphosphorylation of Cdc25, increased Cdk1 activity, mitotic defects such as delayed anaphase chromosomes and multipolar spindles, and finally to mitotic arrest and cell death.
Similar to yeast, hCdc14B is associated with nucleolar chromatin only during interphase, and is released into the cytoplasm from prometaphase on [19,23,28]. Thus, nucleolar sequestration represents an evolutionarily conserved mechanism that controls hCdc14B function enabling hCdc14B to target substrates at defined stages during mitosis and shielding target proteins from being dephosphorylated during interphase. In support of this view, elevated levels of hCdc14B perturbed nucleolar sequestration and delayed entry into mitosis. Notably, it was recently shown that perturbation of cell cycle progression by genotoxic stress also induces nucleolar release of hCdc14B correlating with hCdc14B-dependent activation of APC/C Cdh1 and G 2 -arrest [29]. G 2 /M transition is tightly controlled by a complex mitotic entry network activation of Cdk1 representing the pivotal event. To achieve full activity, Cdc25 phosphatases are upregulated at G 2 /M by Cdk1-and Plk-dependent phosphorylations, initiating an autoamplification loop that ensures rapid mitotic entry. Vice versa, completion of mitosis relies on inactivation of Cdk1/cyclin B as well as reversal of mitotic phosphorylations. In mammalian cells more than one thousand phosphosites are generated by proline-directed kinases during mitosis, mostly by Cdk1 [37]. The mitotic phosphatases that dephosphorylate Cdk1 substrates are poorly characterized. In yeast, Cdc14 is a component of two consecutive mitotic exit systems, FEAR and MEN that activate Cdc14 leading to inhibition of mitotic Cdk1 activity and reversal of Cdk1dependent phosphorylations [13,38]. In this study we have shown that hCdc14B plays a pivotal role in dephosphorylating and inactivating Cdc25 proteins leading to decrease of Cdk1/cyclin B activity. Inhibition of Cdk1/cyclin B is largely accomplished by proteasomal degradation of cyclin B through APC Cdc20 and APC Cdh1 [2]. Besides this well-established pathway Cdk1/cyclin B is also inactivated in a non-proteolytic fashion at onset of anaphase by transient inhibitory Wee1-mediated phosphorylation of Y15 on Cdk1 [39]. Since inhibitory phosphorylation on Cdk1 is counteracted by Cdc25 proteins, phosphatases that target and repress Cdc25 proteins in mid-mitosis have been postulated [40]. Our data demonstrate that hCdc14B is a candidate phosphatase, Figure 5. hCdc14B regulates Cdk1/cyclin B activity. A. Overexpression of hCdc14B alters Cdk1/cyclin B activity. Cdk1/cyclin B was immunopurified from U2OS, U2OS-fl-hCdc14B and U2OS-fl-hCdc14B SL cells that were synchronized at G 1 /S (0 h) and released for the indicated times. Kinase activity was assayed in vitro using Cdk1-specific peptides and 32 P-ATP. Values were normalized to the activity of Cdk1/cyclin B in G 1 /S cells. The Western blots show the amount of immunoprecipitated cyclin B, Cdk1 and Cdk1/pY15. B. Depletion of hCdc14B increases Cdk1/cyclin B activity. Cells were transfected with GFP-(siGFP, light bars) or hCdc14B-specific siRNA-2 (siCdc14B-2, dark bars) and arrested by double thymidine block. Kinase activity of immunopurified Cdk1/cyclin B was assayed in vitro at the indicated times (h) after release from G 1 /S. The immunoblot on the top shows the amounts of cyclin B used in the assays. C. Cyclin B is stabilized upon knockdown of hCdc14B. Cells were transfected with siRNAs against GFP (siGFP), hCdc14B (siCdc14B-2 and siCdc14B-3) or left untransfected (mock), synchronized by double thymidine block, released from G 1 /S for 16 h and analyzed on immunoblots. doi:10.1371/journal.pone.0014711.g005 Cdc25s. Therefore, hCdc14B may act in two ways to inactivate Cdk1/cyclin B at late mitosis, e.g. by dephosphorylating and activating the APC co-activator Cdh1 [29] and by enabling Cdk1/ Y15 phosphorylation by Wee1 upon dephosphorylation and inactivation of Cdc25 phosphatases.
Our results are consistent with the model depicted in Figure 6E. Entry into mitosis depends on the establishment of a positive feedback loop involving Cdk1-dependent activation of Cdc25 phosphatases and Cdc25-dependent dephosphorylation of Cdk1 on pT14/pY15. At this stage of the cell cycle, nucleolar sequestration of hCdc14B prevents premature dephosphorylation and inactivation of Cdc25s. Transient release from nucleolar chromatin at prometaphase enables hCdc14B to interact with Cdc25 proteins. Dephosphorylation by hCdc14B inactivates Cdc25 phosphatases, allowing de novo phosphorylation of Cdk1 on Y15 and inactivation of Cdk1. Consistently depletion of hCdc14B prevents timely inactivation of Cdk1/cyclin B, leading to delay of metaphase and severe mitotic defects. As in human cells a large number of proteins are phosphorylated by Cdk1/cyclin B during mitosis [37], prolonged Cdk1 activity and failure of reversal of Cdk1 phosphorylations will affect a wide range of mitotic processes. Therefore, depletion of hCdc14B caused severe mitotic defects, including chromosome missegregation, erroneous mitotic spindle formation and binucleation.
So far, it was not known whether the hCdc14A and B isoforms contribute to site-specific dephosphorylation of Cdc25 proteins. According to our semi-quantitative mass spectrometric analysis substrate specificity differs between hCdc14A and hCdc14B at least to some extent. Some phosphorylations of Cdc25A, e.g. pS88 and pS261, were removed by hCdc14B, but were retained after incubation with hCdc14A. This suggests that the hCdc14 isoforms target different sites within substrates and are functionally not fully redundant. Functional analysis of mitotic regulation by phosphatases has been hampered by the complex phenotypes associated with inactivation of mitotic phosphatases, making it difficult to distinguish between primary and secondary effects [38]. Nevertheless, the phenotypes resulting from knockdown of hCdc14B or hCdc14A are distinct and consistent with hCdc14A playing primarily a role in centrosome splitting, and hCdc14B promoting primarily progression through mitosis and chromosome segregation.
Studies on the role of hCdc14B in cell cycle regulation have been controversial, depending on the experimental systems used, e.g. depletion by RNAi, overexpression or gene knockout in cultured cell lines. hCdc14B has been implicated in centriole amplification [30], mitotic exit [25], M/G 1 transition [32], G 2 -DNA damage check point activation [29], DNA repair [31], bundling and stabilizing microtubules, and assembly of the mitotic spindle [28]. In contrast to our results, a recent study by Berdougo et al [40] showed no obvious mitotic defects in human colorectal carcinoma HCT116 cells in which the hCdc14B gene has been genetically inactivated. This obvious discrepancy may be due to mechanisms that established during selection of the somatic knockout cells compensating loss of hCdc14B. Requirement for hCdc14B may also vary between different cell lines. We observed that hCdc14B levels are high in HeLa cells but low in HCT116 cells (unpublished data). A recent RNAi screen identified PP2A-B55a as a phosphatase that promotes late mitosis and mitotic exit suggesting that other phosphatases may compensate loss of hCdc14B in a cell-type specific manner [41]. In addition, hominoids contain another hCdc14B-like gene, known as hCdc14C [42]. Therefore it cannot be ruled out, that other phosphatases functionally replace hCdc14B in different cell lines to trigger inactivation of Cdk1 in mid mitosis. Notably, expression of hCdc14C is inhibited by the siRNAs used in our study but not by the hCdc14B knockout strategy used by Berdougo et al [40]. Therefore, the differences in mitotic phenotypes may be attributed to hCdc14C replacing the function of hCdc14B. This would also explain the observation that Cdc14B-specific antibodies detected the hCdc14 protein in wildtype as well as in Cdc14B 2/2 cells [40]. Clearly, further systemic approaches are required to identify additional physiological substrates of hCdc14B and to elucidate the processes that link hCdc14B activity to cell division.
Immunoblots showing co-precipitation of Cdc25 isoforms with fl-hCdc14B SL from 293T cells co-expressing HA-Cdc25A, HA-Cdc25B, GFP-Cdc25C, and fl-hCdc14B SL . D. Overexpression of hCdc14B leads to inactivation of Cdc25A. Cdc25A was immunoprecipitated from doxycyclin-treated (dox) or untreated U2OS-fl-hCdc14B cells, incubated with Cdk1/cyclin B from S-phase cells, and kinase activity of Cdk1/cyclin B was monitored by in vitro phosphorylation of histone H1 (top). As a control, activation of Cdk1/cyclin B was monitored using recombinant GST-Cdc25A (bottom). Quantification of histone H1 phosphorylation was performed with a PhosphorImager and is shown below. E. Model depicting the role of hCdc14B during mitosis. Entry into mitosis requires a positive feedback loop, involving Cdk1-dependent activation of Cdc25 phosphatases and Cdc25-dependent dephosphorylation of Cdk1 on pT14/Y15. Before onset of mitosis, hCdc14B is sequestered within the nucleolus, preventing premature dephosphorylation and inactivation of Cdc25 proteins. From prometaphase until late mitosis, hCdc14B is released from nucleolar chromatin, leading to inactivation of Cdc25s, which triggers inhibitory phosphorylation of Cdk1 and inactivation of Cdk1/cyclin B. doi:10.1371/journal.pone.0014711.g006 hCdc14B cell lines, U2OS cells were co-transfected with pcDNA/ 6TR (Clontech) and pcDNA/TO-fl-hCdc14B or pcDNA/TO-fl-hCdc14B SL . Clones were selected in tetracycline-free medium supplemented with zeocin (250 mg/ml) and blasticidin (5 mg/ml). For fl-hCdc14B expression, cells were treated with doxycyclin (1 mg/ml). Cells were synchronized by a double thymidine-block, and fl-hCdc14B expression was induced by addition of doxycyclin 8 h before cells were released from the second thymidine-block. Cells transfected with shRNA expression plasmids were selected in the presence of zeocin (250 mg/ml) [45]. siRNA duplexes (40-100 nM) were reverse transfected with lipofectamin 2000 (Invitrogen). siRNA and shRNA sequences are given in Table  S1 and S2. Cdc14 transcripts were measured by reverse transcription using random hexamer primers and quantitative real-time PCR (LightCycler 480, Roche) essentially as described [46]. Data were normalized to the level of actin-mRNA. Primer sequences are given in Table S3.

Immunocytochemistry
Cells grown on poly-lysine coated coverslips were fixed with 2% paraformaldehyde, permeabilized with ice-cold methanol, and incubated with the indicated first antibodies and fluorescencelabelled secondary antibodies (Invitrogen, Dianova). DNA was stained using Hoechst 33342. Apoptotic cells were detected by TUNEL assay using the In Situ Cell Death Detection Kit (Roche).

Purification of Recombinant Phosphatases
GST and recombinant GST-tagged proteins were expressed in E. coli Codon Plus-RP cells (Stratagene), affinity-purified over glutathione-Sepharose and eluted according to standard procedures.
Cdc25A activity was measured essentially as described [49]. Cdk1/cyclin B was immunopurified from S-phase cells using anticyclin B antibodies. U2OS-fl-hCdc14B cells were cultured in the absence or presence of doxycyclin (2 mg/ml, 8 h). Cdc25A was immunoprecipitated from equal amounts of cells and incubated with bead-bound Cdk1/cyclin B in 100 ml of phosphatase assay buffer (15 min, 30uC). Control reactions contained the same amount of cell lysate and rabbit IgGs (Dianova). After washing, beads were incubated for 15 min at 30uC in 50 ml histone H1 kinase assay buffer (50 mM Tris-HCl pH 7.5, 10 mM MgCl 2 , 1 mM DTT, 5 mg histone H1 (Roche), 50 mM ATP, 5 mCi 32 P-ATP), subjected to SDS-PAGE, and radioactive proteins were detected by PhosphorImaging.

Analysis of Phosphopeptides by Mass Spectrometry
GST-Cdc25A was phosphorylated with immunopurified Cdk1/ cyclin B for 1 h at 30uC in kinase buffer supplemented with 50 mM ATP in the absence or presence of radioactive 32 P-ATP. GST-Cdc25A beads were washed three times in high salt buffer (20 mM Tris-HCl, pH 7.9, 1 M KCl, 5 mM MgCl 2 , 0.2 mM EDTA, 10% glycerol, 0.5 mM DTT), equilibrated twice in phosphatase assay buffer, and incubated for 1 h at 30uC in phosphatase assay buffer containing equal phosphatase activities (1.2 U) of hCdc14A or hCdc14B, or no phosphatase for control. GST-Cdc25A beads were again washed in high salt buffer, equilibrated in 50 mM ammonium bicarbonate, and proteins were digested with trypsin overnight at 37uC. The resulting tryptic peptides were analyzed by nanoLC ESI-MS/MS. ESI-MS/MS analysis and database search were performed essentially as described [50].

Supporting Information
References S1 A. GST-tagged Cdc25A, B, and C were immobilized on Glutathione-Sepharose and phosphorylated in vitro using extracts from mitotic HeLa cells and 32P-ATP. Phosphorylated Cdc25 was incubated with increasing amounts of GST-hCdc14B or the corresponding activity of calf intestine alkaline phosphatase (CIAP, 0.025, 0.05, and 0.125 U), or was left untreated (-). After SDS-PAGE, GST-Cdc25 and GST-hCdc14B proteins were stained with Coomassie Blue (C panels), and the level of Cdc25 phosphorylation was monitored by PhosphorImaging (32P panels). B. GST-tagged Cdc25A and B were immobilized on Glutathione-Sepharose and phosphorylated in vitro using Cdk1/cyclin B complexes purified from Sf9 cells and 32P-ATP. After washing, the beads were incubated for 45 min at 30uC with 1 mg of GST-hCdc14B, 1 mg of the phosphatase-dead mutant GST-hCdc14BSL, or left untreated. Reactions were stopped, samples separated by 8% SDS-PAGE, and stained with Coomassie Blue (bottom panels). 32P-labelled Cdc25 (top panels) was detected by autoradiography. Found at: doi:10.1371/journal.pone.0014711.s006 (2.03 MB EPS) Figure S6 Mass spectrometry-based identification of GST-Cdc25A peptides phosphorylated by Cdk1/cyclin B and dephosphorylated by hCdc14A/B. Recombinant GST-Cdc25A was phosphorylated in vitro by Cdk1/cyclin B immunopurified from mitotic HeLa cell extracts in the absence or presence of 32P-ATP, followed by incubation with recombinant hCdc14A or B. A. 32Pphosphorylated GST-Cdc25A was incubated with 1.2 or 0.6 U of recombinant His-hCdc14A or B or was left untreated (-).
Following SDS-PAGE, GST-Cdc25A phosphorylation was monitored by PhosphorImaging (left panel). Phosphatase activity of recombinant His-hCdc14B (white bars) and His-hCdc14A (grey bars) was determined in vitro using pNPP as substrate (right panel). Values are derived from three independent assays (mean 6SD). B. Summary of tryptic GST-Cdc25A phospho-peptides identified with nanoLC ESI-MS/MS and Mascot search engine (top panel). Schematic presentation of the relative abundance of GST-Cdc25A phosphopeptides in the absence of hCdc14A or B (black bar) and after incubation with hCdc14B (white bar) or hCdc14A (grey bar). Found at: doi:10.1371/journal.pone.0014711.s007 (0.53 MB EPS) Figure S7 hCdc14B interacts with all three isoforms of Cdc25. GST or GST-Cdc25A, B, and C were incubated for 4 h at 4uC with 35S-labeled fl-hCdc14B. Bead-bound proteins were separated by SDS-PAGE and visualized by PhosphorImaging. The first lane shows 10% of the input proteins. Found at: doi:10.1371/journal.pone.0014711.s008 (0.37 MB EPS) Table S1 Sequences of the siRNAs used in this study. All sequences are written from 5 to 3 orientation. Chemically synthesized siRNA duplexes were obtained from Eurogentec (siCdc14B), MWG (siGFP), and Dharmacon (siCdc14A and siCtrl corresponding to the non-targeting siRNA pool 1). Sequences of the non-targeting siRNA pool 1 are available from Dharmacon.