Protein Phosphatase 2 A ( B 55 ) Prevents Premature Activation of Forkhead Transcription Factor FoxM 1 by Antagonizing Cyclin A / Cyclin-dependent Kinase-mediated Phosphorylation *

Mónica Alvarez-Fernández, Vincentius A. Halim, Melinda Aprelia, Shabaz Mohammed, and René H. Medema From the Department of Medical Oncology and Cancer Genomics Centre, UMC Utrecht, Universiteitsweg 100, Stratenum 2.118, Utrecht 3584 CG, The Netherlands and the Biomolecular Mass Spectrometry and Proteomics Group, Bijvoet Centre for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, Netherlands Proteomics Centre, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands

The forkhead transcription factor FoxM1 controls expression of a large number of genes that are specifically expressed during the G 2 phase of the cell cycle. Throughout most of the cell cycle, FoxM1 activity is restrained by an autoinhibitory mechanism, involving a repressor domain present in the N-terminal part of the protein. Activation of FoxM1 in G 2 is achieved by Cyclin A/Cyclin-dependent kinase (Cdk)-mediated phosphorylation, which alleviates autoinhibition by the N-terminal repressor domain. Here, we show that FoxM1 interacts with B55␣, a regulatory subunit of protein phosphatase 2A (PP2A). B55␣ binds the catalytic subunit of PP2A, and this promotes dephosphorylation and inactivation of FoxM1. Indeed, we find that overexpression of B55␣ results in decreased FoxM1 activity. Inversely, depletion of B55␣ results in premature activation of FoxM1. The activation of FoxM1 that is observed upon depletion of B55␣ is fully dependent on Cyclin A/Cdk-mediated phosphorylation of FoxM1. Taken together, these data demonstrate that B55␣ acts to antagonize Cyclin A/Cdk-dependent activation of FoxM1, to ensure that FoxM1 activity is restricted to the G 2 phase of the cell cycle.
Orderly progression through the cell cycle depends on sequential activation of distinct gene clusters (1). For example, during the late stages of the G 1 phase, phosphorylation of the retinoblastoma protein by Cyclin D/Cyclin-dependent kinase (Cdk) 2 4 -6 complexes relieves repression of E2F transcription factors that control expression of a cluster of genes required for DNA replication (2). Similarly, in the G 2 phase, a number of transcription factors have been shown to be activated by Cyclin A/Cdk2 to promote expression of a cluster of genes that controls progression through G 2 and mitosis (3)(4)(5). One of these transcription factors is the forkhead factor FoxM1 (6). Expression of FoxM1 itself is restricted to proliferating cells and is first activated in late G 1 (7). Despite its production in late G 1 and early S phase, the FoxM1 protein is not transcriptionally active before cells reach G 2 phase. During G 1 /S, FoxM1 activity is kept low through its N-terminal autoinhibitory domain, which interacts with the C-terminal transactivation domain. In G 2 , phosphorylation of its transactivation domain by Cyclin A/Cdk inhibits its interaction with the N-terminal autorepressor domain, allowing FoxM1-mediated gene transcription (4,8). However, CyclinA/Cdk2 activity is already present in cells as early as late G 1 phase, and it is unclear how phosphorylation and activation of FoxM1 are prevented at this stage and restricted to G 2 .
Here, we identify B55␣, a regulatory subunit of the protein phosphatase 2A (PP2A), as a novel interactor of FoxM1. PP2A is the major serine-threonine phosphatase in mammalian cells and plays critical roles in cell cycle, cell proliferation, development, and regulation of multiple signaling pathways (9). PP2A holoenzymes consist of a catalytic subunit C, a structural subunit A, and a variable regulatory subunit B that determines the substrate specificity and spatial and temporal functions of PP2A. There are at least four families of the regulatory subunit: B (also known as B55 or PR55), BЈ (B56 or PR61), BЉ (PR72), and Bٞ (PR93/PR110), with at least 16 members in these families (10). We show here that FoxM1 interacts with the B55␣ subunit, and this acts to prevent FoxM1 activation at the early stages of the cell cycle. We find that depletion of B55␣ from cells synchronized at the G 1 /S transition results in premature phosphorylation and activation of FoxM1. Importantly, this activation is fully dependent on the presence of Cyclin A, indicating that B55␣ acts to antagonize Cyclin A/Cdk-dependent activation of FoxM1 during G 1 and S phase, to limit the transcriptional activation of FoxM1 to G 2 phase.

EXPERIMENTAL PROCEDURES
Cell Culture, Transfections, and Drugs-U2OS and 293-T cells were maintained in DMEM with 6% fetal calf serum and antibiotics. Thymidine and nocodazole were purchased from Sigma and used at 2.5 mM and 250 ng/ml, respectively. Okadaic acid (OA) was purchased from Calbiochem and used at 100 -200 nM. Cells were transfected with plasmid DNA using the standard calcium phosphate transfection protocol. Small inhibitory RNA (siRNA) oligonucleotides were transfected with HiPerFect (Qiagen) following the manufacturer's protocol.
Reporter Assays-Luciferase activity was determined using the dual luciferase kit (Promega) according to the manufacturer's instructions. Relative luciferase was expressed as the ratio of firefly luciferase activity to control Renilla luciferase activity.
RT-PCR-Total RNA was isolated with the Qiagen RNeasy kit, according to the manufacturer's instructions. cDNA was synthesized from 1 g of total RNA by using Superscript II reverse transcriptase (Invitrogen) and oligo(dT) primers. The resultant cDNA was used as a template for PCR amplification with specific primers.
Automated Image Analysis-Cells were grown in 96-well plates (Viewplate-96; PerkinElmer Life Sciences). At the indicated time points, cells were fixed with 4% formaldehyde and methanol, and were stained with DAPI and the indicated antibodies. Image acquisition was performed using a Cellomics ArrayScan VTI (Thermo Scientific) using a 20 ϫ 0.40 NA objective. Image analysis was performed using Cellomics ArrayScan HCS Reader (Thermo Scientific).
Immunoprecipitations-Cells were lysed in PLB buffer (PBS, 2 mM EDTA, 1 mM EGTA, 0.5 mM DTT, and 1% Triton X-100) supplemented with protease and phosphatase inhibitors. Cell lysates were incubated for 2 h with the appropriate antibody coupled to protein A/G-agarose beads. Immunoprecipitates were extensively washed, then denatured in sample buffer, loaded on SDS-PAGE and transferred to nitrocellulose blots. Blots were probed with the indicated antibodies.
Dephosphorylation Assays-For endogenous FoxM1 dephosphorylation, U2OS cells were lysed in phosphatase buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.25% Nonidet P-40, and protease inhibitors) and incubated at 30°C for the indicated times, in the presence or absence of OA. For in vitro PP2A/B55 dephosphorylation assays, FoxM1 was immunoprecipitated from 293-T cells overexpressing FLAG-FoxM1wt and used as a substrate. The dephosphorylation assay was performed at 30°C for 30 min in phosphatase buffer (20 mM Hepes, pH 7.4, 1 mM DTT, 1 mM MgCl 2 , 100 g/ml BSA, and protease inhibitors), containing 0.2 unit of purified PP2A A/C heterodimer (Millipore) and 0.25 g of recombinant GST-B55␣. Where indicated, OA was added at 200 nM.
Mass Spectrometry Analysis-To identify FoxM1-interacting proteins, FoxM1 was purified from U2OS cells stably expressing YFP-S peptide-FoxM1. Ten 15-cm dishes were harvested and cells were lysed in PLB buffer, after which FoxM1 was isolated by pull-down with protein S-agarose beads. FoxM1 complexes were subsequently subjected to SDS-PAGE, and protein bands were visualized by silver staining. Each lane was cut into 16 gel pieces and subjected to in-gel digestion with trypsin. Peptides were extracted with 10% FA. The extracted peptides were subjected to nanoscale liquid chromatography tandem mass spectrometry (nanoLC-MS/MS) analysis, performed on an Agilent 1100 HPLC system (Agilent Technologies) connected to an LTQ Linear Ion Trap Mass Spectrometer combined with an Orbitrap (ThermoFisher) (15). Raw MS data were converted to peak lists using DTASuperCharge version 1.27 (16). Spectra were searched against the International Protein Index (IPI) Human data base version 3.37 (69,164 sequences; 29,064,824 residues) using Mascot software version 2.2.0. Peptides with a mascot score of 35 or higher were considered. The results were combined and compared using Scaffold 2 (Proteome software).
For phosphosites analysis, samples from an in vitro dephosphorylation assay were subjected to SDS-PAGE followed by Coomassie staining. Protein bands corresponding to FoxM1 were sliced out from the gels and subjected to in-gel digestion as described above. Using the dimethyl labeling system (17), extracted peptides from phosphatase-treated and untreated samples were labeled with intermediate and light label respectively, mixed with equal ratio, and subjected to MS analysis as described above. Identification was done against the IPI Human data base using Mascot. Peptides with a Mascot score of 25 or higher were considered. Both mascot result and raw MS data were used to quantify the relative abundance of each peptide. Quantification was performed by dimethyl version of MSQuant (16).

FoxM1
Interacts with B55␣-To identify proteins that interact with FoxM1, we generated cell lines that stably express a version of FoxM1 N-terminal tagged to YFP and an S peptide domain. The YFP-S-FoxM1 fusion protein was pulled-down with S-protein-agarose beads, and the complexes were analyzed by SDS-PAGE. A clear band corresponding to YFP-FoxM1 could be seen, in addition to several other additional bands (Fig. 1A). We subsequently subjected the immunocomplexes to LC-MS/MS, to try to identify the proteins that specifically co-purified with FoxM1. To this end, the control pulldown was compared with the pulldown from YFP-FoxM1expressing cells. Two independent experiments were performed, and several unique peptides were identified in the YFP-FoxM1 complexes in both purifications, corresponding to a total of 12 known proteins (Fig. 1A). These included several peptides from Plk1, a mitotic kinase known to interact with FoxM1 and enhance its transcriptional activity (18). Among the potential novel interactors of FoxM1, we identified PPP2R2A, also known as B55␣, a regulatory subunit of the PP2A phosphatase (Fig. 1A). Because FoxM1 activity is known to be restricted to specific stages in the cell cycle through Cyclin-and Plk1-dependent phosphorylation, we decided to investigate the possible contribution of B55␣/PP2A to the control of FoxM1 activity.
First, we validated the interaction between FoxM1 and B55␣ by performing co-immunoprecipitation experiments in 293-T cells using differentially tagged variants of FoxM1 and B55␣. When we expressed FLAG-tagged B55␣ with nontagged FoxM1, we could clearly detect FoxM1 in the B55␣ immuno-precipitates (Fig. 1B). Inversely, if FLAG-tagged FoxM1 was co-expressed with GFP-tagged B55␣ we could detect GFP-B55␣ in the FLAG-FoxM1 immunocomplexes (Fig. 1B). In addition to B55␣, we also found that the catalytic subunit of PP2A (PP2A/C) was present in complex with FoxM1 (Fig. 1B), suggesting that B55␣ forms a bridge between PP2A/C and FoxM1.
B55␣ Controls FoxM1 Activity-If the function of B55␣ is to bring FoxM1 in contact with the core enzyme of the PP2A phosphatase, this is likely to contribute to regulation of FoxM1 activity. Transcriptional activation by FoxM1 is highly cell cycle-dependent, and this is primarily controlled through phosphorylation by various cell cycle-dependent kinases, such as Cyclin A/Cdk2 and Plk1. Therefore, we reasoned that B55␣ could antagonize the cell cycle-dependent phosphorylation of FoxM1, causing a reduction in FoxM1 activity. To test this, we analyzed the effect of B55␣ overexpression on FoxM1 activity. For this purpose, we introduced a plasmid carrying FoxM1 consensus sites in front of a luciferase reporter gene into cells to assay FoxM1 transcriptional activity. We have previously shown that the basal activity of this promoter is due to its transactivation by endogenous FoxM1 (19). Interestingly, activity of this reporter was clearly reduced upon overexpression of B55␣, suggesting that B55␣ inhibits the transcriptional activity of endogenous FoxM1 (Fig. 2A). Similarly, B55␣ also suppressed FIGURE 1. Identification of B55␣ as a FoxM1-interacting protein. A, FoxM1 was purified from U2OS cells stably expressing YFP-S peptide-FoxM1, as described under "Experimental Procedures." FoxM1 complexes were subsequently subjected to SDS-PAGE, and protein bands were visualized by silver staining. Each lane was cut into 16 gel pieces and followed by tryptic digestion and MS analysis. The table shows the potential FoxM1-interacting proteins, for which a significant number of peptides were identified in FoxM1 samples but not in control samples from parental U2OS or U2OS cells expressing only YFP-S peptide. Two independent experiments were performed, and only the proteins identified in both of them are indicated. B, 293-T cells were co-transfected with plasmids expressing FoxM1 and FLAG-B55␣. B55␣ immunocomplexes were purified by immunoprecipitation with an anti-FLAG antibody, and FoxM1 was detected with an anti-FoxM1 antibody (left). Cells co-transfected with plasmids expressing FLAG-FoxM1 and GFP-B55␣ were subjected to immunoprecipitation (IP) with an anti-FLAG antibody, and B55␣ and PP2A/C were detected with anti-GFP and anti-PP2A antibodies, respectively (right).
the activity of overexpressed FoxM1 toward this reporter ( Fig.  2A). Interestingly, whereas activity of the overexpressed FoxM1 was reduced, FoxM1 levels were increased by overexpression of B55␣ ( Fig. 2A), suggesting that B55␣ can positively affect FoxM1 stability while negatively influencing FoxM1 activity.
In previous studies, we have shown that cell cycle-dependent phosphorylation of FoxM1 alleviates intramolecular repression of FoxM1 activity by the N-terminal domain (4). In those studies we described an N-terminal deletion mutant of FoxM1, lacking both the N-terminal repressor domain, as well as the KEN motif that targets FoxM1 for degradation (12). This mutant was shown to be constitutively active and did not display any cell cycle-regulated activation (4,8). Importantly, the activity of the ⌬N/⌬KEN mutant of FoxM1 was unaffected by overexpression of B55␣ ( Fig. 2A). This was not due to a lack of interaction of this mutant with B55␣ or PP2A/C, as we could easily detect both proteins in co-immunoprecipitates with the ⌬N/⌬KEN mutant of FoxM1 (Fig. 2B). In agreement with the luciferase results, overexpression of B55␣ also reduced the level of the FoxM1 target gene Plk1. We detected a significant reduction in endogenous Plk1, both at protein and mRNA levels, upon overexpression of B55␣ (Fig. 2C). Importantly, the effects of B55␣ overexpression on FoxM1 activity were not due to a cell cycle arrest, as overexpression of B55␣ did not result in any dramatic changes in the cell cycle distribution (Fig. 2D). Taken together, these results indicate that B55␣ directly affects the cell cycle-dependent activation of FoxM1.
Phosphorylation of FoxM1 Is Regulated by B55␣-During the early stages of the cell cycle, FoxM1 is kept inactive by an autoinhibitory mechanism involving the N terminus of FoxM1. Upon phosphorylation of FoxM1 by Cyclin A/Cdk complexes, autoinhibition is relieved (4,8). Because B55␣ seems to inhibit FoxM1 activity, we reasoned that it acts to antagonize Cyclin A/Cdk-dependent phosphorylation. To test this, we examined the effect of overexpression and depletion of B55␣ on cell cycledependent phosphorylation of FoxM1. As cells progress through G 2 and enter mitosis, FoxM1 becomes hyperphosphorylated, leading to a shift in its electrophoretic mobility (4). Overexpression of B55␣ led to a reduction in that phosphoshift in cells synchronized in G 2 /M, which was reverted by treatment with the phosphatase inhibitor okadaic acid (Fig. 3A, white  arrowhead). This effect was specific to B55␣, because overexpression of another B subunit for PP2A, B56␥, did not affect phosphorylation of FoxM1 (supplemental Fig. S1). The effect of B55␣ overexpression on the phosphorylation of endogenous FoxM1 was also confirmed with a phosphoantibody that recognizes phosphorylated threonine residues only when followed by proline (Fig. 3A). Next, we analyzed the effect of B55␣ depletion on FoxM1 phosphorylation. We synchronized U2OS cells at the G 1 /S transition using a thymidine block and subsequently released these cells using fresh medium containing nocodazole to trap cells in mitosis. Under these conditions, a slower migrating form of FoxM1 became visible at 12 h after release, and its abundance increased at the 16-h time point (Fig. 3B). Importantly, the abundance of hyperphosphorylated FoxM1 was clearly enhanced at these time points when B55␣ was depleted by RNA interference (Fig. 3B). Coincident with the higher levels of hyperphosphorylated FoxM1, we observed a clear increase in the amount of mitotic cells in the cultures depleted of B55␣, which fits with the recently reported function of B55␣ in mitotic exit (20). This result indicates that depletion of B55␣ enhances phosphorylation of FoxM1, although from these data we cannot conclude whether this effect is direct or whether it is achieved indirectly via a more general effect of B55␣ inactivation on cell cycle progression. To address this issue, we analyzed the status of FoxM1 phosphorylation upon depletion of B55␣ in cells synchronized in G 1 /S with thymidine, when there are no differences in cell cycle profile between control and B55␣-depleted cells. In this case, we could also detect the appearance of a slower migrating band of endogenous FoxM1, which was not observed if we co-depleted Cyclin A (Fig. 3C). Thus, depletion of B55␣ does increase Cyclin A-dependent phosphorylation of FoxM1.
To provide more evidence for a direct link between FoxM1 dephosphorylation and B55␣, we monitored the rate of in vitro dephosphorylation of endogenous FoxM1 by incubating mitotic cell extracts, in which FoxM1 is hyperphosphorylated, at 30°C for different periods of time. FoxM1 dephosphory- Lysates were subjected to Western blot analysis with the indicated antibodies. Phosphorylation of endogenous FoxM1 was detected by Western blotting with a phosphothreonine-proline antibody on FoxM1 immunoprecipitates. B, U2OS cells transfected with siRNA control (GAPDH) or a siRNA targeting B55␣ were synchronized with thymidine in G 1 /S and subsequently released from the thymidine block in fresh medium with nocodazole for the indicated times. Whole cell lysates were analyzed by Western blotting with antibodies against FoxM1 and B55␣, and actin was used as a loading control (upper). The mitotic index was determined by pH 3 staining and FACS analysis (lower). C, U2OS cells transfected with siRNA control (GAPDH) or a siRNA targeting B55␣, alone or in combination with a siRNA targeting Cyclin A, were synchronized with thymidine in G 1 /S. The expression pattern of endogenous FoxM1 was analyzed by Western blotting. Efficiency of B55␣ and Cyclin A depletion was also determined by Western blot analysis. D, cell extracts from control or B55␣-depleted cells were subjected to in vitro dephosphorylation for the indicated time points at 30°C, in the absence or presence of OA (100 nM). Phosphoshifts of endogenous FoxM1 and B55␣ depletion were determined by Western blotting with the indicated antibodies. Actin was used as a loading control.
lation was clearly delayed in B55␣-depleted extracts compared with control extracts, as demonstrated by longer persistence of slower migrating bands (Fig. 3D). This result indicates that B55␣ is indeed directly involved in FoxM1 dephosphorylation. Moreover, we found that FoxM1 can be efficiently dephosphorylated in vitro by incubation with purified PP2A, containing both A and C subunits and recombinant B55␣ (Fig. 4A). Importantly, PP2A/C can directly dephosphorylate FoxM1 in vitro (supplemental Fig. S2), indicating that B55␣ is not required for the catalytic activity of PP2A/C. However, the activity of PP2A/C toward FoxM1 is enhanced by the addition of B55␣ (supplemental Fig. S2), consistent with increased substrate recognition. Phosphorylation of FoxM1 detected with the phospho-threonine-proline antibody was dramatically reduced upon addition of recombinant PP2A/B55␣, and this was reverted by preincubation with OA (Fig. 4A). This result clearly demonstrates that PP2A can directly dephosphorylate FoxM1.
To find out which sites in FoxM1 are targets of PP2A/B55␣, we analyzed PP2A dephosphorylation of FoxM1 by MS. FoxM1 bands from an in vitro phosphatase assay were isolated from the gel and subjected to tryptic digestion. Peptides from phosphatase-treated and untreated control were dimethyl-labeled with intermediate and light isotopes, respectively, mixed in equal ratio, and subjected to LC-MS/MS. Quantification was performed by comparing the relative abundance of intermediate versus light peptides. We identified 79 peptides of FoxM1 pres-ent in both samples, of which 15 were phosphopeptides (supplemental Table 1 and Fig. 4B). Those phosphopeptides contained 10 unique phosphosites, most of which we previously found to be phosphorylated in vivo (4). Six of those phosphopeptides showed a ratio lower than 0.4, of which 5 contained phosphorylated Ser 35 and one contained a phosphorylated residue in position 609, 610, or 611 (Fig. 4B). Importantly, the ratio of their unmodified counterparts was equal to or higher than 1 (supplemental Table 1), indicating that the low abundance of the phosphopeptide in the phosphatase-treated sample is indeed due to a reduction in phosphorylation. We previously showed that mutation of Ser 35 does not have any significant effect on the transcriptional activity of FoxM1 in G 2 (4). In the same study we found that Thr 611 was phosphorylated in vitro and in vivo by Cyclin A/Cdk complexes, and this phosphorylation was required to relieve the autoinhibition of FoxM1 by the N-terminal repressor domain (4). Therefore, we generated a phosphomimetic mutant of FoxM1 in which the T611 residue was substituted by a glutamic acid (FoxM1 T611E) to test the effect of B55␣ overexpression on FoxM1 transcriptional activity in G 2 . Interestingly, we found that the activity of FoxM1 T611E was not inhibited upon B55␣ overexpression, indicating that the effect of PP2A/B55 on FoxM1 transactivation is mediated via dephosphorylation of Thr 611 (Fig. 4C).
B55␣ Suppresses Premature Activation of FoxM1-Because depletion of B55␣ increases the phosphorylation of FoxM1, we decided to analyze whether it also affects its transcriptional activity. In line with the inhibition of FoxM1 activity we observed upon overexpression of B55␣ ( Fig. 2A), we found that depletion of B55␣ leads to enhanced FoxM1 activity in cells arrested at the G 1 /S transition (Fig. 5A). Again, the activity of the ⌬N/⌬KEN mutant of FoxM1 was unaffected by perturbation of B55␣ expression (Fig. 5A). In addition, the activating effect of B55␣ depletion was no longer seen in cells released from the thymidine block and synchronized in G 2 /M (Fig. 5A), indicating that inhibition of FoxM1 by B55␣ occurs at the early stages of the cell cycle.
In agreement with the luciferase results, depletion of B55␣ at the G 1 /S transition resulted in higher mRNA levels of Plk1 and Cyclin A, two well known transcriptional targets of FoxM1 (Fig. 5B). This effect correlated with an increase in the protein level of Cyclin A in B55␣-depleted cells, as detected by immunofluorescence at the single cell level (Fig.  5B).
Cyclin A/Cdk activity was shown to peak in G 2 , coincident with the timing of activation of FoxM1-dependent target genes (4). However, the Cyclin A/Cdk2 complex is first activated at the G 1 /S transition, but is somehow prevented from activating FoxM1. Our findings that the effect of B55␣ on FoxM1 activity depends on a Cyclin A/Cdk phosphorylation site and that depletion of B55␣ leads to activation of FoxM1 specifically at the G 1 /S transition suggest that it acts to prevent premature activation of FoxM1 by low levels of Cyclin A/Cdk2. Indeed, when Cyclin A was co-depleted with B55␣, activation of FoxM1 activity was no longer observed in cells synchronized at the G 1 /S transition (Fig. 5C), demonstrating that the effect of B55␣ FIGURE 5. Effect of B55␣ depletion on FoxM1 activity. A, U2OS cells expressing the 6XDBE reporter together with FoxM1 wt or FoxM1 ⌬N/⌬KEN were co-transfected with siRNA control (GAPDH) or siRNA targeting B55␣. Luciferase activity was determined in cells synchronized at the G 1 /S transition by thymidine treatment and in G 2 /M by releasing them from the thymidine block in fresh medium for 14 h. FoxM1 transcriptional activity in control (siGAPDH) was set as 1. FoxM1 protein levels and efficiency of B55␣ depletion were assessed by Western blotting. CDK4 was used as a loading control. B, cells were transfected with siRNA control (GAPDH) or siRNA targeting B55␣ and synchronized in G 1 /S with thymidine for 24 h. mRNA levels of FoxM1 targets were analyzed by RT-PCR (left). Cyclin A protein levels were determined by immunofluorescence and automated image analysis. At least 500 cells were counted per well. Measurements were performed in triplicate, and error bars represent S.D. (right). C, cells transfected with the 6XDBE reporter and FoxM1wt were co-transfected with siRNA control (GAPDH) and siRNAs targeting either B55␣ or Cyclin A. FoxM1 transcriptional activity was determined in G 1 /S, 24 h after treatment with thymidine. Protein levels were determined by Western blot analysis with the indicated antibodies. Actin was used as a loading control. Graphs represent the average of three independent experiments, and error bars indicate S.D. depletion on FoxM1 activity is mediated through Cyclin A/Cdk-dependent phosphorylation.

DISCUSSION
In this study we aimed to identify new FoxM1-interacting proteins that contribute to FoxM1 function. We have found that B55␣, a regulatory subunit of the PP2A phosphatase, interacts with FoxM1 and modulates its transcriptional activity. We have previously shown that phosphorylation by Cyclin A/Cdk complexes is required to activate FoxM1 in the G 2 phase of the cell cycle (4). Here, we provide evidence that the phosphatase B55␣/PP2A counteracts that phosphorylation and negatively regulates FoxM1 activity. Whereas overexpression of B55␣ inhibits FoxM-dependent transactivation in G 2 , depletion of B55␣ leads to premature activation of FoxM1 in G 1 /S, in a Cyclin A-dependent manner. We also provide evidence that Thr 611 in FoxM1 is the main phosphoresidue through which PP2A/B55␣ influences FoxM1 transcriptional activity in G 2 . We and others have previously shown that phosphorylation of this residue by Cyclin/Cdk complexes is important for FoxM1 transactivation, allowing recruitment of the transcriptional coactivator CREB-binding protein and mediating the release of the repression of the N-terminal domain of FoxM1 (4,21). Moreover, Thr 611 has also been shown to be one of the two key sites of FoxM1 involved in its phosphorylation-dependent interaction with Plk1 (18). It is therefore likely that PP2A/B55␣ might counteract indirectly the activation of FoxM1 by Plk1.
It is unclear how the PP2A complexes are regulated and whether their activity changes throughout the cell cycle. Our results suggest that B55␣/PP2A complexes are active during G 1 /S, but they could be somehow inactivated in G 2 , because depletion of B55␣ in this phase does not affect FoxM1 transactivation. However, we find that overexpression of B55␣ reduces FoxM1 activity in G 2 and that FoxM1 also interacts with B55␣ in G 2 (data not shown). This indicates that, rather than the phosphatase being inactivated in G 2 , it is more likely that the increase in Cyclin A/Cdk2 kinase activity in this phase results in a shift in the balance between phosphorylation/dephosphorylation toward a hyperphosphorylated active form of FoxM1, in the presence of active B55␣/PP2A. All together, our results suggest that B55␣/PP2A-dependent regulation of FoxM1 is more important during the early stages of the cell cycle, when Cyclin A/Cdk2 activity is low.
In addition to FoxM1, other transcription factors, such as B-Myb or NF-Y, also play an important role in the expression of genes required for G 2 /M progression (22). Similar to FoxM1, their transcriptional activity is regulated by Cyclin A/Cdk phosphorylation, and so far no phosphatase has been found to regulate their activity. An interesting possibility would be that B55/ PP2A phosphatase exerts a more general role in regulating gene expression during the cell cycle by limiting also the activity of these other transcription factors to G 2 .
In summary, our data show that the combined action of Cyclin A/Cdk and B55␣/PP2A fine tunes FoxM1 transcriptional activity to ensure that FoxM1 is only active in G 2 . At the G 1 /S transition, the phosphatase PP2A, through the subunit B55␣, prevents premature activation of FoxM1 when there is low level of Cyclin A/Cdk activity. Later in G 2 , when Cyclin A/Cdk reaches its maximum activity, phosphorylation of FoxM1 releases its autorepressor domain, allowing FoxM1 activation and expression of the genes required for G 2 /M progression.