|Home | About | Journals | Submit | Contact Us | Français|
Proper control of entry into and progression through mitosis is essential for normal cell proliferation and the maintenance of genome stability1–4. The mammalian mitotic kinase Pololike kinase 1 (Plk1) is involved in multiple stages of mitosis5. Here we report that Forkhead Box M1 (FoxM1), a substrate of Plk1 (refs 6–8), controls a transcriptional programme that mediates Plk1-dependent regulation of cell-cycle progression. The carboxy-terminal domain of FoxM1 binds Plk1, and phosphorylation of two key residues in this domain by Cdk1 is essential for Plk1–FoxM1 interaction. Formation of the Plk1–FoxM1 complex allows for direct phosphorylation of FoxM1 by Plk1 at G2/M and the subsequent activation of FoxM1 activity, which is required for expression of key mitotic regulators, including Plk1 itself. Thus, Plk1-dependent regulation of FoxM1 activity provides a positive-feedback loop ensuring tight regulation of transcriptional networks essential for orderly mitotic progression.
Transitions through cell-cycle phases require the coordination of multiple events and are tightly regulated by protein kinases1,2,5. In addition, transcriptional control of key cell-cycle regulators are important for cellcycle progression9,10. Thus, defining the links between protein kinases and transcriptional networks is essential for an understanding of normal cell-cycle progression and how specific factors may contribute to the misregulation of this process in diseases.
In Saccharomyces cerevisiae, the polo kinase Cdc5p coordinates cellcycle-dependent transcription at G2/M by directly phosphorylating and regulating the activity of a co-activator of the forkhead transcription family member Fkh2p (ref. 11). The mammalian Polo-like kinase 1 (Plk1) is also essential for multiple events throughout M phase1,5 and several Plk1 targets involved in mitosis have been identified12–17. So far, no direct link between Plk1 and the transcription factors controlling gene expression has been identified, although the conservation of functions between polo kinases in yeast and higher eukaryotes makes it tempting to speculate that Plk1 may have a role in mammalian cells similar to that recently demonstrated for Cdc5p in yeast11.
The mammalian transcription factor Forkhead Box M1 (FoxM1) is important in regulating mitotic entry and subsequent execution of the mitotic programme by controlling the expression of a cluster of G2/M target genes6,8,18. However, the mechanisms regulating the activation of FoxM1 at specific cell-cycle phases, and how this contributes to G2/M progression are largely unknown.
Here, using yeast two-hybrid screening, we identified FoxM1 as a direct binding partner of Plk1. Six of the twenty-four positive clones encoded various lengths of the FoxM1 C terminus, with the smallest clone encoding residues 463–748 of FoxM1. These initial results suggest that there is a link between Plk1 and a transcription factor controlling a G2/M transcriptional programme.
In vitro translated FoxM1 associated with GST-fused Plk1 (GST–Plk1), but not with GST (Fig.1a), indicating that Plk1 binds directly to FoxM1. In contrast, bacterially expressed FoxM1 failed to interact with in vitro translated Plk1 (Supplementary Information, Fig. S1a), suggesting that post-translational modification of FoxM1 may be required for this interaction. In addition, ectopically expressed FoxM1 formed a complex with endogenous Plk1 (Fig. 1b), but not Plk2 or Plk3 (Supplementary Information, Fig. S1c). Plk1 and FoxM1 show periodic expression patterns, with both proteins increasing markedly in G2/M cells19 (Supplementary Information, Fig. S1b). Indeed, an endogenous Plk1–FoxM1 complex was readily observed in G2/M cells, whereas this complex was barely detectable in asynchronous cells (Fig. 1c).
A noticeable progressive decrease in FoxM1 mobility was observed when cells were released from G0/G1 arrest (Supplementary Information, Fig. S1b), indicating that FoxM1 undergoes an initial phosphorylation in S phase followed by hyperphosphorylation during G2/M, a time at which Plk1 activity is maximal. Spatial and temporal coordination of Plk1 activity is mediated by the C-terminal Polo-box domain (PBD) of Plk1 (refs 20–22). Furthermore, the PBD is required for recruitment of Plk1 to its substrates that have been ‘primed’ by phosphorylation at specific docking sites23. Therefore, FoxM1 may undergo an initial priming phosphorylation, allowing it to be recruited and directly phosphorylated by Plk1.
Plk1 PBD recognizes a consensus sequence of S-pS/pT-P/X22,24. FoxM1 sequence contains two potential PBD-binding sites (Thr 596 and Ser 678). Our yeast two-hybrid results indicated that a C-terminal domain of FoxM1 (residues 463–748) is essential for Plk1 binding. Therefore, we examined whether Thr 596 and Ser 678 of FoxM1 would mediate the interaction between these two proteins. Thr 596 and Ser 678 were mutated to Ala individually (FoxM1T596A, FoxM1S678A) or in combination (FoxM1TSAA), and their ability to bind GST-tagged Plk1 PBD was analysed. FoxM1WT associated with GST–PBD, but not with GST or the GST–PBD mutant (Fig. 1d). Binding to the Plk1 PBD in the FoxM1T596A and FoxM1S678A single mutants was reduced; however, mutation of both residues abolished this interaction (Fig. 1d). Thus, both Thr 596 and Ser 678 of FoxM1 seem to be key sites involved in a phosphorylation-dependent interaction between FoxM1 and Plk1 PBD.
Many docking sites on Plk1-binding proteins are initially phosphorylated by Cdk1/cyclin B. FoxM1 preferentially binds to the Cdk1/cyclin B complex in G2, and Thr 596 of FoxM1 was previously identified as a Cdk1 phosphorylation site7. We determined whether phosphorylation of FoxM1 by Cdk1 is responsible for the formation of Plk1–FoxM1 complex. We performed pulldown assays using bacterially expressed GST–Plk1 PBD and MBP–FoxM1 fusion proteins, with or without Cdk1/ cyclin B. Phosphorylation of MBP–FoxM1 by Cdk1 promoted a strong interaction between FoxM1 and Plk1 PBD (Supplementary Information, Fig. S2a). Inhibition of Cdk1, but not Plk1 (ref. 25), markedly reduced the binding of endogenous FoxM1 to GST–Plk1 PBD (Supplementary Information, Fig. S2b, e). Thus, Cdk1-dependent phosphorylation of FoxM1 is required for its interaction with Plk1.
FoxM1 is hyperphosphorylated during G2/M phase, which correlates with its transcriptional activity being cell-cycle-restricted. Therefore, we hypothesized that Plk1 phosphorylates and activates FoxM1 during G2/M transition. We found that FoxM1 was phosphorylated by the constitutively active Plk1 (Plk1TD) in vitro, but not by the catalytically inactive Plk1KD mutant (Fig. 2a).
We next sought to map the major Plk1 phosphorylation sites on FoxM1. A series of FoxM1deletion-mutants were generated and tested for their capacity to be phosphorylated by Plk1 in vitro. Although the amino-terminal segments of FoxM1 were weakly phosphorylated by Plk1, C-terminal segments, especially those containing the transcriptional activation domain (TAD, residues 617–748; Supplementary Information, Fig. S3a, b), showed much higher levels of phosphorylation.
Plk1 substrates do not in general contain a strict consensus phosphorylation motif26,27. Therefore, we used mass spectrometry analysis for the identification of Plk1-dependpent FoxM1 phosphorylation sites. Using recombinant FoxM1 protein purified from bacteria and phosphorylated in vitro by Plk1, we identified four Ser residues (Ser 680, Ser 702, Ser 715 and Ser 724) located within the TAD of FoxM1 as potential Plk1 phosphorylation sites. Single- and double-site mutants were generated and their capacity to be phosphorylated by Plk1 was determined in vitro. Single mutation of Ser 715 or Ser 724 reduced Plk1-dependent phosphorylation of FoxM1 (Supplementary Information, Fig. S3c), whereas phosphorylation was almost abolished when the S715A;S724A double mutant was used (Fig. 2b). Thus, Ser 715 and Ser 724 seem to be the two major Plk1-dependent phosphorylation sites within the TAD region of FoxM1. Importantly, these two Ser residues, as well as the surrounding amino acids, are highly conserved among vertebrates (Fig. 2c), indicating that the phosphorylation of these two sites may have an evolutionarily conserved role in regulating FoxM1 activity.
We next determined whether FoxM1 is phosphorylated by Plk1 in vivo. We generated phospho-specific anti-FoxM1 antibodies. As shown in Fig. 2d, Both the anti-p715 and anti-p724 antibodies immunoreacted with wild-type FoxM1; however, only the anti-p715 antibody recognized the FoxM1S724A mutant and vice versa. Furthermore, neither antibody detected λ-phosphotase-treated wild-type FoxM1 (Supplementary Information, Fig. S3d). Thus, the anti-p715 and anti-p724 antibodies are highly specific and only recognize their corresponding targets. Using these phospho-specific antibodies, we showed that phosphorylation of endogenous FoxM1 could be readily detected in G2/M but not in asynchronous HeLa cells (Fig. 2e), indicating that FoxM1 is phosphorylated at Ser 715 and Ser 724 sites at G2/M. To determine whether Plk1 is required for these phosphorylation events, we depleted endogenous Plk1 by siRNA and observed a marked reduction of FoxM1 phosphorylation at Ser 715 and Ser 724 sites (Fig. 2f), supporting the suggestion that Plk1 is the kinase that phosphorylates FoxM1 in vivo.
We directly examined the possibility that Cdk1 could function as a priming kinase of FoxM1. To test this, we performed in vitro kinase assays using Cdk1, Plk1, or both as kinase sources. Consistent with earlier studies7, Cdk1 could phosphorylate FoxM1 to some extent (Supplementary Information, Fig. S2c). Interestingly, in the presence of Cdk1 the phosphorylation of FoxM1 by Plk1 was markedly elevated when compared with Plk1 alone. Furthermore, Cdk1-mediated phosphorylation was initiated as early as S/G2 phase in vivo, whereas Plk1-mediated phosphorylation occurred later, mainly at G2/M phases (Supplementary Information, Fig. S2d), suggesting that FoxM1 undergoes an initial priming phosphorylation by Cdk1, which allows its subsequent phosphorylation by Plk1.
To better understand the functional significance of Plk1-dependent phosphorylation of FoxM1, we tested whether Plk1 regulates FoxM1 transcriptional activity, assayed using the 6× FoxM1–TATA–luciferase reporter plasmid7. Wild-type Plk1, but not the kinase-inactive mutant of Plk1, significantly upregulated FoxM1 transcriptional activity in a concentration-dependent manner (Fig. 3a), suggesting a direct regulation of FoxM1 by Plk1-dependent phosphorylation.
We next investigated whether the association between Plk1 and FoxM1 is necessary for Plk1-dependent activation of FoxM1. Mutation of one of the two residues involved in the docking of Plk1 to FoxM1 resulted in only a partial decrease in Plk1-enhanced FoxM1 transcriptional activity (T596A, 55%; S678A, 30%). In contrast, mutation of both residueled to a marked reduction (>90% decrease) in Plk1-dependent FoxM1 transcriptional activity (Fig. 3b). These results are consistent with those shown in Fig. 1d and indicate that phosphorylation of either Thr 596 or Ser 678 of FoxM1 can independently mediate the binding to Plk1 and this binding is essential for Plk1-dependent activation of FoxM1.
To test the hypothesis that Plk1-mediated phosphorylation of FoxM1 enhances its transcriptional activity, we mutated the two major Plk1 phosphorylation sites of FoxM1 (Ser 715 and Ser 724) to Ala (FoxM1AA) or to glutamic acid (FoxM1EE) and examined the effects on FoxM1 transcriptional activity. We included a FoxM1 mutant lacking the TAD (FoxM11–616) as a negative control in these experiments. Indeed, the phospho-mutant of FoxM1 was not activated by Plk1 (Fig. 3c), whereas FoxM1EE showed robust transcriptional activity with or without Plk1 (Fig. 3c). Taken together, these results indicate that after an initial priming phosphorylation of FoxM1 by Cdk1, Plk1 is recruited to and directly phosphorylates FoxM1, thereby enhancing FoxM1 transcriptional activity.
To address the functional relevance of Plk1-dependent FoxM1 phosphorylation in vivo, we developed a rescue assay using U2OS cells stably expressing various wild-type or mutant forms of FoxM1. Endogenous FoxM1 levels were depleted using siRNA and rescue of the FoxM1 knock down phenotype was analysed in cells stably expressing Myc-tagged wild-type FoxM1 or siRNA-resistant Myc-tagged wild-type (WT-r) or S715A;S724A mutant (AA-r) FoxM1. As controls, U2OS cells stably transfected with empty vector were treated with siRNAs targeting FoxM1 or the firefly luciferase gene. Endogenous FoxM1 and Myc-tagged wild-type FoxM1 were significantly depleted but WT-r and AA-r remained abundant (Fig. 4b, Supplementary Information, Fig. S4a), confirming that the siRNA-resistant FoxM1 constructs were effective.
Previous studies have shown that cells lacking FoxM1 have multiple cellcycle defects, including a delay in G2 and reduced expression of G2/M FoxM1 target genes. Although some FoxM1-depleted cells eventually enter mitosis, additional mitotic defects may occur, resulting in the formation of polyploid and aneuploid cells6,8. Indeed, FACS, morphological analyses and time-lapse microscopy showed an increase in 4N cells and an accumulation of polyploid (>4N) cells (Fig. 4a; Supplementary Information, Fig. S4a, c) after FoxM1 depletion. Numerous mitotic defects, including misaligned and/or mis-segregated chromosomes, aberrant furrow formation, failed cytokinesis and prolonged mitotic progression, were also observed, confirming that FoxM1 is required for G2/M transition and proper mitotic progression. A delay of cells in G2/M transition and an increase in the number of binucleated cells (10.4%, Supplementary Information, Fig. S4b) seems to account for the accumulation of cells with 4N DNA content in the absence of FoxM1. Expression of wild-type and siRNA-resistant wild-type FoxM1 partially or completely restored normal cell-cycle progression (Fig. 4a). However, the siRNA-resistant FoxM1 S715A;S724A mutant was not able to rescue these phenotypes, suggesting that Plk1-dependent phosphorylation and activation of FoxM1 is crucial for FoxM1 function at G2/M.
To provide an in vivo correlation, we next examined the expression of FoxM1 target genes, in the context of the rescue assay described above. Indeed, in FoxM1-depleted cells, we observed downregulation of FoxM1 target genes, including Plk1, Cyclin B1, and Aurora B, all of which are required for mitotic progression; expression of siRNA-resistant wild-type FoxM1 rescued this downregulation (Fig. 4b, c). In contrast, Plk1, Cyclin B1 and Aurora B levels remained low in cells expressing the siRNA-resistant FoxM1 S715A;S724A mutant cells (Fig. 4b, c). Next, we blocked Plk1 activity using a chemical inhibitor25. Although inhibition of Plk1 activity reduced the expression of downstream targets of FoxM1, no effect was observed in cells expressing the constitutively active FoxM1EE mutant (Fig. 4d). Moreover, time-lapse imaging analysis revealed that inhibition of Plk1 activity leads to mitotic defects, such as prolonged early mitotic phases and prometaphase arrest in wild-type cells. These defects were partially rescued in cells expressing FoxM1EE (Fig. 4d; Supplementary Information, Movies 1, 2). Together, these data indicate that Plk1 regulates FoxM1 transcriptional activity by direct phosphorylation and thereby controls execution of a transcriptional programme required for mitotic progression.
We have developed a model to describe the role of Plk1-mediated regulation of FoxM1 at G2/M transition (Fig. 4e). This suggests that at late S and G2 phase, FoxM1 is initially phosphorylated by Cdk1 kinases, creating docking sites for the PBD of Plk1. Subsequently, Plk1 binds and directly phosphorylates FoxM1. This activates FoxM1 transcriptional activity, resulting in the enhanced expression of key mitotic regulators. As Plk1 is a target gene of FoxM1, this mode of regulation can generate a positive-feedback loop, leading to a further increase in Plk1 levels and FoxM1 activity. Furthermore, as the major transcription factor during G2/M transition, full activation of FoxM1 ensures the coordinated expression of transcriptional networks essential for timely entry into M phase. This mode of regulation is reminiscent of the Rb-E2F pathway during G1/S transition, where Rb is hyperphosphorylated and releases E2F, allowing for transactivation of target genes that promote S phase entry28.
The polo kinase Cdc5p also regulates a G2/M transcriptional programme in S.cerevisiae11. Here we have extended these observations and revealed a link between Plk1 and regulation of a G2/M transcriptional network in mammalian cells. In contrast to yeast Cdc5p-mediated phosphorylation of a co-activator Ndd1p, Plk1 directly phosphorylates and thereby activates FoxM1 transcriptional activity. Our data do not exclude the possibility that other transcription factor(s) may collaborate with FoxM1 to coordinate the appropriate expression of transcriptional networks. As various cell-cycle kinases are upregulated in many cancers, the results presented here should help to provide further insights into the link between aberrant expression of these kinases and alterations in cell proliferation as well as genomic integrity, potentially through misregulation of transcriptional programmes.
Details of plasmid constructs, siRNA and primer sequences used in this study can be found in Supplementary Information Methods. Rabbit polyclonal antibodies recognizing phosphorylated Ser 678, phosphorylated Ser 715 or phos-phorylated Ser 724 of FoxM1 were raised against peptides CPFGNSpSPSDID, DTMNDpSLSKILLDC and KILLDIpSFPGLDC, respectively, and affinity-purified using their corresponding peptide columns. Other antibodies used in this study were polyclonal anti-FoxM1 (1:200) and anti-Cyclin B1 (1:1000) (Santa Cruz Biotechnology), monoclonal anti-Plk1 (1:1000; Zymed), monoclonal antiMyc (1:1000; Invitrogen), monoclonal anti-Aurora B (1:250; BD Biosciences), and anti-β-actin (1:5000; Sigma) antibodies.
293T and HeLa cells were cultured in RPMI 1640 medium (Invitrogen) containing 10% fetal bovine serum (FBS; Biosource), 100 units ml−1 penicillin, 100 units ml−1 streptomycin and 0.25 µg ml−1 amphotericin B (Fungizone; Invitrogen). U2OS cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen) containing 10% FBS, 100 units ml−1 penicillin, 100 units ml−1 streptomycin and 0.25 µg ml−1 amphotericin B. For cell synchronization, T24 cells were arrested in G0 by contact inhibition. After at least 3 days of confluence, the cells were split 1:10 by seeding in multiple 100-mm dishes at a concentration of 106 cells per dish. Cells were collected at various time points after replating30. For synchronization of U2OS cells, cells were treated with 2 mM thymidine for 24 h and then released in fresh medium for 10 h. To arrest exponentially growing HeLa or U2OS cells at prometaphase, nocodazole was added to a final concentration of 100 ng ml−1 for 16–24 h. Mitotic cells were collected by mechanical shake-off.
Cells were lysed in NETN lysis buffer (150 mM NaCl, 1 mM EDTA, 20 mM Tris pH 8, 0.5% Nonidet P-40) containing protease and phosphatase inhibitors for 20 min at 4 °C. Cell lysates were centrifuged for 10 min at 4 °C. For immunoprecipitation or co-immunoprecipitation experiments, cleared lysates (1–2 mg protein) were immunoprecipitated with the indicated polyclonal antibodies and protein A-Sepharose for 1 h. In phosphatase treatment experiments, FoxM1 immunoprecipitates were treated with 400 units of λ phosphatase in the accompanying buffer (New England Biolabs) for 1 h at 30 °C. Proteins were separated on 7.5% SDS–PAGE, transferred to Immobilon P membranes and immunoblotted with the indicated antibodies.
Human FoxM1 was transcribed and translated in vitro using TNT quick-coupled transcription/translation system (Promega). GST–Plk1 or GST control bound to glutathione-sepharose resin was incubated with in vitro translated and labelled with FoxM1 protein for 1 h at 4 °C, and the washed complexes were boiled in SDS–PAGE sample. Alternatively, glutathione-sepharose resin-coupled GST–Plk1 PBD wild-type, mutant or the GST control were incubated with cell lysates containing Myc-tagged wild-type or mutant FoxM1 for 1 h at 4 °C The resins were washed four times using the NETN buffer. Resin-bound complexes were eluted by boiling, separated by SDS–PAGE, and analysed by autoradiography or western blotting.
Cells grown on coverslips were fixed for 15 min with 3% paraformaldehyde solution and then permeabilized with 0.5% Triton X-100 for 5 min. Slides were incubated with primary antibodies for 20 min at 37 °C. Fluorescein isothiocyanate-conjugated goat anti-rabbit and/or rhodamine-conjugated goat anti-mouse serum (Jackson ImmunoResearch) were used as secondary antibodies. All antibodies were diluted in 5% goat serum. Cells were counterstained with 4–6-diamidino-2-phenylindole (DAPI) dye for 30 s.
Constitutively active human (Plk1TD), wild-type or kinase-defective (PlklKD) Plk1 were expressed in the baculovirus/insect cell system. Cdk1/cyclin B proteins were purchased from Invitrogen. GST–FoxM1 and its deletion mutant proteins were expressed in Escherichia coli BL21 strain. Kinase and substrates (FoxM1 proteins) were incubated in kinase buffer (20 mM Hepes, pH 7.4, 150 mM KCl, 10 nM MgCl2, 1 mM EGTA, 0.5 mM dithiothreitol, 5 mM NaF, 10 µM ATP, 5 µCi γ-32P) for 30 min at 30 °C. Reactions were stopped by the addition of SDS sample buffer. Then samples were heated for 5 min to 95 °C before analysis by SDS–PAGE and autoradiography.
U2OS cells were transfected with the indicated plasmids using Lipofectamine 2000 according to manufacturer’s protocol. Luciferase activity was determined 24 h after transfection using the dual luciferase kit (Promega). Luciferase levels were normalized to Renilla luciferase activity. Relative luciferase activity was expressed as percentage induction of promoter activity by the FoxM1 expression vector and Plk1, where promoter activity resulting from transfection with FoxM1 wild-type was set at 100%. Experiments were performed in triplicate and statistical analysis was performed using Microsoft Excel tools.
Cells were collected, washed with PBS and fixed with icecold 70% ethanol for at least 1 h. Cells were washed twice in PBS and treated for 30 min at 37 °C with RNase A at 5 µg ml−1 and propidium iodide at 50 µg ml−1, and analysed on a FACScan flow cytometer (Becton Dickinson). The percentage of cells in different cell-cycle phases was calculated using ModFit LT for Mac (BD Biosciences).
Experiments were performed using a Zeiss Axio Observer system with: CO2 Module S, TempModule S, Heating Unit XL S, Pln Apo ×10/1.4 DICIII objective, AxioCam MRm camera and AxioVision 4.6 software. The imaging medium was DMEM containing 10% FBS. The temperature of the imaging medium was kept at 37 °C. The exposure times was 400 ms at 2 × 2 binning. Total time of mitosis was defined as the interval between NEBD (nuclear envelope breakdown) and cytokinesis. Mitotic defects were scored.
Note: Supplementary Information is available on the Nature Cell Biology website
We are grateful for extensive discussions with R.H. Medema and J. Laoukili and their input. We thank H.M. Thompson for editing the manuscript. Mass spectrometry analysis was performed by Taplin Biological Mass Spectrometry Facility at Harvard University. This work was supported in part by grants from the National Institutes of Health (NIH RO1 CA113381 to JC), Mayo SPORE P50 (CA116201 project no. 1 to J.C.) and the T.J. Martell Foundation (to D.J.T.). J.C. is a recipient of an Era of Hope Scholars award from DOD. Z.F. is a recipient of a Ruth L. Kirschstein NRSA individual Fellowship from NIH.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/