BAD detects coincidence of G2/M phase and growth factor deprivation to regulate apoptosis.

BAD, a member of the Bcl-2 protein family, promotes mitochondria-dependent apoptosis. Here, we report that BAD dissociates from 14-3-3zeta at each G2/M phase of proliferating lymphoid cells. The cell cycle-dependent dissociation of BAD was associated with phosphorylation at Ser-128, whereas mutant S128A-BAD, in which Ser-128 was converted to alanine, remained associated with 14-3-3zeta throughout the cell cycle. Although the cell cycle-dependent dissociation of BAD per se did not induce apoptosis, growth factor deprivation induced prompt apoptosis at the G2/M phase but not at the G1 phase. In cells expressing S128A-BAD, growth factor deprivation-induced apoptosis was markedly delayed and was accompanied by a delayed dephosphorylation of growth factor-dependent regulatory serine residues. These results indicate that BAD induces apoptosis upon detecting the coincidence of G2/M phase and growth factor deprivation.

Growth factors are required for cell growth, proliferation, and differentiation. The deprivation of appropriate growth factors induces apoptosis, causing elimination of hematopoietic cells (1). A BH3 only protein, BAD, plays a gatekeeper role in growth factor-dependent cell survival by regulating the mitochondrial threshold for apoptosis (2,3). It has been shown that BAD is phosphorylated and associated with 14-3-3 proteins in the presence of growth factors. Upon deprivation of growth factors, BAD is dephosphorylated and dissociates from 14-3-3 to bind to Bcl-xL or Bcl-2 on the mitochondria, and the translocation of BAD will reduce the threshold for apoptosis and markedly enhance cell death (4 -6). These reactions are regulated by the phosphorylation states of BAD. The phosphorylation of Ser-136 of BAD is necessary for the association between BAD and 14-3-3 and cell survival (7)(8)(9)(10)(11). The phosphorylation of Ser-112 of BAD enhances the association between BAD and 14-3-3 (12), resulting in cell survival (8,(13)(14)(15)(16). Another phosphorylation site at Ser-155 of BAD blocks the binding of BAD to Bcl-xL or to Bcl-2 (17,18). The importance of these functional residues was demonstrated in BAD knock-in mice in which each of the regulatory serine residues (Ser-112, -136, and -155) was converted to alanine. Growth factor-dependent survival was attenuated in cells from the immune and nervous systems of the mutant knock-in mice (5).
Ser-128 of BAD is another functional site, the phosphorylation of which enhances neuronal apoptosis by preventing the association between BAD and 14-3-3. The phosphorylation of Ser-128 by Cdc2 was observed when granule cells of the developing rat cerebellum received death signals (19,20). Although the effect of Cdc2 was observed in nonproliferating neurons, Cdc2 is an essential enzyme for mitosis progression in proliferating cells. If Cdc2 also phosphorylates Ser-128 of BAD in proliferating cells, it is possible that the association between BAD and 14-3-3 is regulated not only by growth factors but also by the cell cycle. If this is the case, BAD-mediated apoptosis upon the withdrawal of a growth factor may be cell cycle-dependent. However, the cell cycle dependence of BAD-mediated apoptosis has not been clarified.
To study the association between BAD and 14-3-3 along with the cell cycle progression in proliferating cells, we developed a fluorescence resonance energy transfer (FRET) 1 -based assay system to monitor their association. We fused the variants of the green fluorescent protein to BAD and to 14-3-3, which binds to BAD and has an anti-apoptotic effect (21,22). A significant FRET signal was observed upon the association between BAD and 14-3-3. We show here that BAD and 14-3-3 are associated at the G 1 phase but are dissociated in the G 2 /M phase in interleukin-3 (IL-3)-dependent FL5.12 lymphoid cells. The dissociation of the two molecules was regulated by the phosphorylation of BAD at Ser-128. When the cells were deprived of IL-3, apoptosis occurred immediately at the G 2 /M phase but not at the G 1 phase. These results demonstrate a novel cell cycle-dependent regulation of apoptosis by BAD.

MATERIALS AND METHODS
Protein Expression in Escherichia coli and Purification-DNAs encoding BAD (NM_007522) and 14-3-3 (U57312) were obtained from a mouse brain cDNA library (Clontech). The plasmids pECFP-N1 and pEYFP-N1, encoding enhanced cyan fluorescent protein (ECFP) and enhanced yellow fluorescent protein (EYFP), respectively, were obtained from Clontech. A point mutation in the BAD gene (Ser-128 3 Ala-128) was generated by a mutagenic PCR technique. All the PCRamplified DNAs were sequenced (LRT; Amersham Biosciences) to ensure no errors had been introduced during amplification. The DNAs were inserted into pET23d (Stratagene) and introduced into E. coli BL21(DE3) (Stratagene). The cells were grown to an A 600 of 0.5 in the 2xYT medium at either 16°C for BAD-CFP or 25°C for 14-3-3-YFP. Recombinant gene expression was induced by incubation in the presence of 1 mM isopropyl-␤-D-thiogalacto-pyranoside for 24 h. The cells were harvested by centrifugation and broken with a French press. The recombinant proteins were purified from the cell lysate using a resin with an affinity for His-tagged proteins, Talon (Clontech), and gel filtration chromatography (HiLoad 16/60 Superdex 200pg, Amersham * This work was supported by grants-in-aid for scientific research and by the Advanced and Innovational Research Program in Life Science from the Ministry of Education, Culture, Sports, Science, and Technology, Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Surface Plasmon Resonance-We used BIAcore2000 (BIAcore) for surface plasmon analysis. Purified BAD-CFP (20 g/ml) was coupled to the CM5 sensor chip surface using the standard amine coupling procedure to yield a signal of 80 resonance units. BAD-CFP immobilized on the chip was phosphorylated at 30°C by injecting PKA in HEPESbuffered solution containing 1 mM MgCl 2 for 60 min at a flow rate of 2 l/min. As an analyte, 300 nM 14-3-3-YFP in HEPES-buffered solution was injected at a flow rate of 5 l/min at 25°C.
Fluorescence Spectra-The fluorescence spectra of the proteins (concentration 10 M) in a solution containing 50 mM HEPES (pH 7.4), 150 mM NaCl, and 0.1% bovine serum albumin were obtained using a spectrofluorometer (FP-6500; JASCO). The excitation wavelengths were 425 nm for CFP and 470 nm for YFP.
Protein Expression and Fluorescence Measurement in FL5.12 Cells-DNAs encoding BAD-ECFP and 14-3-3-YFP were cloned into the retroviral vector pMX (a kind gift from Dr. T Kitamura, The University of Tokyo). To package the retrovirus, GP293 cells were cotransfected with 1 g of pVSV-G (Clontech) and 12 g of either pMX-BAD-CFP or pMX-14-3-3-YFP using Lipofectamine 2000 (Clontech). Forty-eight hours after transfection, the culture supernatant was collected and applied to 1 ϫ 10 5 FL5.12 cells. Clones coexpressing BAD-CFP and 14-3-3-YFP were isolated by limited dilution and subjected to FRET analysis. Fluorescent images of the cells in the RPMI 1640 medium on a coverslip were obtained under an inverted fluorescence microscope (Olympus IX-70) using a water immersion objective lens (OLYMPUS LUMPlanFl/IR, ϫ40, numerical aperture 0.8). The images were separated by a dichroic mirror into two channels and independently collected using the W-VIEW system and processed using Aquacosmos software (Hamamatsu Photonics).
Evaluation of FRET Signal-Because of the broad excitation and emission spectra of GFP variants, the 535-nm channel with excitation at 420 nm contains the direct excitation of both CFP and YFP in addition to FRET between CFP and YFP. To evaluate the FRET-excited YFP signal, we subtracted the following two values from the 420-nm excited 535-nm signal: 1) CFP signals at 535 nm estimated from the 420-nm excited 480 nm signal, and 2) intensities of the 420-nm excited YFP signal estimated from the 490-nm excited YFP signal. We evaluated the FRET signal within cells using the ratio of the FRET-excited YFP signal (F535) to the 420-nm excited CFP signal at 480 nm (F480) from the images taken using the W-VIEW system.
RNA Interference-We used small hairpin RNA (shRNA) to knock down the BAD expression (23). The sequence of the shRNA (5Ј-AGTC-CTGGTGGGATCGAAACT-3Ј) corresponds to mouse BAD nucleotide positions 563-583. The shRNA was retrovirally expressed in FL5.12 cells, and the expression level of BAD was examined by Western blotting.
Immunoblotting and Co-immunoprecipitation-Whole-cell samples for phosphorylation analysis were prepared as cells suspended in SDScontaining buffer (10 4 cells/l) and boiled before immunoblot analysis. The samples for immunoprecipitation were prepared as described by Zha et al. (4). Briefly, cells were lysed in Tris-buffered solution containing 0.3% Nonidet P-40. The lysates were incubated with 1 g/ml anti-BAD antibody (Cell Signaling) for 1 h followed with protein G beads for 1 h at 4°C. The immunoprecipitate was washed and resuspended in loading buffer for further immunoblot analysis.
The prepared samples were separated by polyacrylamide gel electrophoresis; separated products were transferred to a polyvinylidene difluoride membrane and probed with appropriate antibodies. Antibodies against BAD, BAD phospho-Ser-112, -Ser-136, and -Ser-155 (Cell Signaling), and 14-3-3 (Santa Cruz Biotechnology) were purchased from the indicated suppliers. The antibody against BAD phospho-Ser-128 was a kind gift from Dr. A. Bonni and Dr. Konishi (Harvard Medical School). Immunoreactive bands were detected using the ECL system (Amersham Biosciences).
Flow Cytometry-To analyze DNA content, cells were fixed with cold 70% ethanol and treated with RNase A. DNA was stained with propidium iodide (25 g/ml), and cytometric analysis was performed on a FACScan (BD Biosciences). For the measurement of mitochondrial membrane potential (⌬⌿ m ), cells were incubated at 37°C with 10 nM tetramethylrhodamine methylester for 10 min and analyzed using FACScan. Apoptosis was quantified by FACScan using annexin V-FITC (VioBision Inc.).

RESULTS
In Vitro Evaluation of FRET-based Assay of Association between BAD and 14-3-3-To monitor the association and dissociation between BAD and 14-3-3, we used a FRET-based assay system to facilitate time lapse measurements. We fused the ECFP as a FRET donor to BAD and the enhanced yellow fluorescent protein as a FRET acceptor to 14-3-3 (Fig. 1A). We first evaluated the functional properties of recombinant BAD-CFP and 14-3-3-YFP in vitro. The PKA phosphorylated serine residues at positions 112, 136, and 155 but not 128 of recombinant BAD-CFP (Fig. 1B) in accordance with the previous results (5,20). The interaction between 14-3-3-YFP and BAD-CFP was analyzed using surface plasmon resonance. BAD-CFP associated with 14-3-3-YFP only when BAD-CFP was prephosphorylated by PKA (Fig. 1C). These results indicate that the fusion proteins retain the properties of their parent molecules in terms of phosphorylation sites and molecular interaction.
We then examined whether we could observe FRET between BAD-CFP and 14-3-3-YFP upon their association. The fluorescence spectrum of nonphosphorylated BAD-CFP showed only a minimal change regarding the presence or absence of 14-3-3-YFP (Fig. 1D, left panel). In contrast, when PKA-treated BAD-CFP was mixed with 14-3-3-YFP, there was a clear change in the fluorescence spectrum: the fluorescence intensity of BAD-CFP at 480 nm (F480) decreased and that of 14-3-3-YFP at 535 nm (F535) increased (Fig. 1D, right panel). These results indicate that the intensity of the FRET signal between BAD-CFP and 14-3-3-YFP increases when the two molecules associate after the phosphorylation of BAD-CFP by PKA. Thus, the FRET signal (F535/F480) can be used to monitor the extent of association between BAD-CFP and 14-3-3-YFP.
Evaluation of the FRET-based Assay within Cells-To examine the FRET signal within living cells, we chose FL5.12 cells. FL5.12 cells were coinfected with BAD-CFP and 14-3-3-YFP (Fig. 1E). In cells treated with staurosporine, a general inhibitor of protein kinases including PKA (24), F480 increased at the expense of F535, resulting in a decrease in FRET signal intensity (Fig. 1F). The change in FRET signal intensity indicates the dissociation of BAD-CFP from 14-3-3-YFP and is accompanied by a decrease in the level of BAD-CFP phosphorylation (Fig. 1G). We also examined the effect of okadaic acid, a phosphatase inhibitor (10,25). In cells treated with 100 nM okadaic acid, the FRET signal intensity increased, indicating an enhanced association between BAD-CFP and 14-3-3-YFP (Fig. 1H) with a concomitant increase in the level of BAD-CFP phosphorylation at Ser-112 and Ser-136, but not at Ser-128 (Fig. 1I). However, when the okadaic acid concentration was increased to 1000 nM, the FRET signal intensity decreased, indicating the dissociation of BAD-CFP from 14-3-3-YFP (Fig.  1H). The decrease in the FRET signal intensity was not accompanied by any decrease in phosphorylation level at Ser-112 and Ser-136 but was accompanied by an increase in phosphorylation level at Ser-128 (Fig. 1I), which was shown to inhibit the association between BAD and 14-3-3 (20). These results show that the association between BAD and 14-3-3 is regulated by the balance among multiple phosphorylation sites on BAD and that the FRET signal reports their association state, which may be difficult to estimate by measuring only the phosphoryl-ation state of BAD. We also examined whether BAD has a growth factor deprivation-dependent proapoptotic effect in FL5.12 cells. The endogenous BAD expression was significantly reduced by the expression of small hairpin RNA (supplemental Fig. 1A). FL5.12 cells underwent cell death after withdrawal of IL-3 from the culture medium, and the extent of apoptosis was significantly reduced after the knock down of BAD (supplemental Fig. 1B). The results indicate that FL5.12 cells are appropriate for the study of BAD-mediated apoptosis.
Cell Cycle Arrest and Association between BAD and 14-3-3-Having established the method of monitoring the association between BAD and 14-3-3 in live cells, we examined whether the level of association between the two molecules changes with the cell cycle. FL5.12 cells were treated with a reagent that arrests cell cycle progression for 6 h before both the DNA content of the cells and the FRET signal were examined ( Fig. 2A). Using mimosine (26) or aphidicolin (27), most of the cells were synchronized at the G 1 phase, and simultaneously the FRET signal intensity increased. In contrast, treatment with nocodazole (28) or paclitaxel (29) synchronized most of the cells at the G 2 /M phase and decreased the FRET signal intensity. These results indicate that BAD-CFP associates with 14-3-3-YFP at the G 1 phase, whereas the two molecules dissociate at the G 2 /M phase. We then studied the level of phosphorylation at Ser-128 on BAD. It decreased at the G 1 phase and increased at the G 2 /M phase reciprocally with the change in the FRET signal intensity (Fig. 2B). The phosphorylation level at other functional sites, Ser-112, -136, and -155, showed a minimal change in terms of the cell cycle phase in the presence of IL-3 (Fig. 2B). The enhanced phosphorylation at Ser-128 was also observed in endogenous BAD when cells were synchronized at the G 2 /M phase by nocodazole not only in FL5.12 cells but also BW (murine thymoma) and NIH3T3 cells (fibroblasts) (Fig.  2C). To study the role of Cdc2 in the phosphorylation at Ser-128, we examined the effects of a Cdc2 inhibitor, roscovitine, in FL5.12 cells. The nocodazole-induced phosphorylation at Ser-128 was reversed in the presence of roscovitine (Fig. 2D). The nocodazole-induced dissociation of BAD from 14-3-3 was also reversed in the presence of roscovitine (supplemental Fig. 2A). These results indicate that Cdc2 is involved in the phosphorylation of BAD at Ser-128 in FL5.12 cells.
To examine whether the phosphorylation of BAD at Ser-128 is indeed directly involved in the cell cycle-dependent dissociation of BAD from 14-3-3, we generated a mutant BAD in which Ser-128 was replaced by alanine (S128A). The FRET signal intensity between S128A mutant BAD and 14-3-3 did not change even when cells were synchronized at the G 2 /M phase by nocodazole treatment for 6 h (Fig. 2A). The result indicates that phosphorylation at Ser-128 is essential for the cell cycle-dependent dissociation of BAD from 14-3-3.
To further confirm the dissociation of BAD from 14-3-3 at the G 2 /M phase, immunoprecipitation experiments were carried out (supplemental Fig. 3). In FL5.12 cells coexpressing both BAD-CFP and 14-3-3-YFP, an anti-BAD antibody coimmunoprecipitated the two proteins, and the level of co-immunoprecipitated 14-3-3-YFP decreased following the cell cycle arrest at the G 2 /M phase by paclitaxel. On the other hand, in the cells expressing S128A mutant BAD-ECFP, there was no difference in the level of co-immunoprecipitated 14-3-3-YFP between control and G 2 /M-arrested cells. Thus, the observed Ser-128-dependent dissociation of BAD from 14-3-3 at the G 2 /M phase was supported by an independent method.
Temporal Correlation of Cell Cycle with Interaction between BAD and 14-3-3-We then examined whether the association/ dissociation of BAD and 14-3-3 would change along with the cell cycle progression. The cell cycle was first synchronized at either the G 2 /M phase in the presence of nocodazole or at the G 1 phase in the presence of mimosine for 5 h, and then the reagent was withdrawn to allow the cell cycle to proceed. The progression of the cell cycle was examined on the basis of the DNA contents of the cells. The population of cells at the G 2 /M (or G 1 ) phase increased (or decreased) during nocodazole treatment, and there was a rebound change after the withdrawal of nocodazole (Fig. 3A, left). Essentially, the same but opposite changes were observed in mimosine-treated cells (Fig.  3A, right). At the same time, the association between BAD and 14-3-3 was monitored using the FRET signal (Fig. 3B). In both nocodazole-and mimosine-treated cells, the levels of association between BAD and 14-3-3 followed a time course similar to that of the population of cells in the G 1 phase (Fig. 3, A versus  B). A correlation analysis between the G 1 cell population and FRET signal supported this observation (supplemental Fig. 4). We also examined the level of BAD phosphorylation at Ser-128 during the cell cycle progression. Indeed, the level of phosphorylation at Ser-128 increased during the G 2 /M phase and decreased during the G 1 phase (Fig. 3C). These results indicate that both the extent of association between BAD and 14-3-3 and the level of phosphorylation of BAD at Ser-128 change with cell cycle progression.
Link between Cell Cycle and Apoptosis after IL-3 Deprivation-Next we examined the relationship between the cell cycle phase and apoptosis after IL-3 deprivation. Apoptosis was mon- itored on the basis of mitochondrial membrane potential (⌬⌿ m ), which was measured using the fluorescence intensity of tetramethylrhodamine methylester as an indicator (30). When the cell cycle was arrested at the G 1 phase in the presence of mimosine, the extent of association between BAD and 14-3-3 increased, and the association state continued after the withdrawal of IL-3 in the continued presence of mimosine (Fig. 4A,  filled circles). Interestingly, apoptosis was not observed even after IL-3 deprivation for 2-3 h in the presence of mimosine (Fig. 4B, filled circles). Only after washing out of mimosine did BAD dissociate from 14-3-3 and apoptosis was enhanced (Fig.  4, A and B, filled circles). Therefore, when IL-3 was withdrawn at the G 1 phase, the apoptotic process was not initiated immediately.
We then examined the effect of IL-3 deprivation on apoptosis in cells synchronized at the G 2 /M phase by nocodazole treatment. Upon the application of nocodazole the extent of association between BAD and 14-3-3 decreased (Fig. 4A, open circles). In sharp contrast with cells synchronized at the G 1 phase, an enhancement of apoptosis started immediately after IL-3 deprivation (Fig. 4B, open circles).
The above results suggest that dissociation of BAD from 14-3-3 is required for the initiation of IL-3 deprivation-induced apoptosis. If indeed the acceleration of apoptosis depends on the association between BAD and 14-3-3 cells expressing the S128A mutant, BAD would show a delayed initiation of apoptosis because S128A-BAD did not dissociate from 14-3-3 even when the cells were synchronized at the G 2 /M phase ( Fig.  2A). We therefore examined this possibility. In cells expressing S128A-BAD, IL-3 deprivation at the G 2 /M phase induced a much delayed enhancement of apoptosis in comparison with that in cells expressing wild-type BAD-CFP (Fig. 4B, triangles versus open circles). In accordance with this result, the inhibition of phosphorylation at Ser-128 by roscovitine, which also inhibited the dissociation of BAD from 14-3-3, delayed the initiation of apoptosis after IL-3 deprivation (supplemental Fig. 2B).
Effect of 14-3-3 Association on Dephosphorylation of BAD-As mentioned in the Introduction, the binding of BAD to 14-3-3 is also dependent on the phosphorylation of BAD at the growth factor-dependent regulatory serine residues (Ser-112 and -136). Using the R18 peptide, which interacts with the amphipathic groove of 14-3-3 to displace 14-3-3 targets, Chiang et al. (12) showed that dissociation of 14-3-3 from BAD accelerates dephosphorylation of the regulatory serine residues in vitro. Thus, we examined whether the Ser-128-dependent dissociation of BAD from 14-3-3 also has a potentiating effect on the rate of dephosphorylation of the regulatory serine residues in FL5.12 cells. The phosphorylation of BAD at Ser-112 and Ser-136 was analyzed in cells expressing BAD or S128A-BAD. Four hours after the cells were deprived of IL-3, the level of phosphorylation of BAD at Ser-112 and Ser-136 decreased, whereas that of S128A-BAD remained nearly constant (Fig. 5). These results suggest that the Ser-128-dependent dissociation of BAD from 14-3-3 facilitates the dephosphorylation of BAD at the growth factor-dependent sites after IL-3 deprivation. This would contribute to the progression of apoptosis by facilitating the dissociation between BAD and 14-3-3.
Relation between Apoptosis and Cell Cycle-dependent Dissociation of BAD from 14-3-3-The above results showed that the cell cycle-dependent dissociation of BAD from 14-3-3 promotes the proapoptotic function of BAD in the absence of IL-3. However, the cell cycle-dependent dissociation of BAD from 14-3-3 per se did not result in apoptosis in the presence of IL-3. Therefore, an additional IL-3-dependent step seems to be re-FIG. 6. Scheme for cell cycle-and growth factor-dependent BAD-mediated apoptosis. At the G 2 /M phase, when BAD is dissociated from 14-3-3, the dephosphorylation of growth factordependent sites and apoptosis proceeded promptly in the absence of the growth factor. At the G 1 /S-phase, when BAD associates with 14-3-3, the progression of apoptotic process was delayed even in the absence of the growth factor. quired for the induction of apoptosis after the dissociation of BAD from 14-3-3 at the G 2 /M phase. It has been shown that the phosphorylation of Ser-155 of BAD suppresses the proapoptotic function of BAD by eliminating its affinity for Bcl-xL (17,18). Indeed, there was no appreciable change in the phosphorylation level of BAD-CFP at Ser-155 at any cell cycle phase (Fig. 2B, bottom row). We further analyzed the phosphorylation state of endogenous BAD (supplemental Fig. 5). In wild-type cells synchronized at the G 2 /M phase by nocodazole treatment, Ser-128 was phosphorylated but Ser-155 remained phosphorylated in the presence of IL-3. Ser-155 was dephosphorylated when IL-3 was withdrawn. These results suggest that, after dissociation from 14-3-3 in the G 2 /M phase, BAD may not bind to Bcl-xL in the presence of IL-3 and that an additional step (dephosphorylation at Ser-155 due to IL-3 deprivation) is required for the induction of apoptosis. DISCUSSION We showed that BAD dissociates from 14-3-3 at the G 2 /M phase of each cell cycle. The cell cycle-dependent dissociation of BAD from 14-3-3 is primarily regulated by the phosphorylation of BAD at Ser-128 by Cdc2. The dissociation of BAD from 14-3-3 at the G 2 /M phase does not immediately promote apoptosis but facilitates subsequent apoptosis only when the cells are deprived of growth factors. Indeed, the dissociation of BAD from 14-3-3 enhanced the dephosphorylation of BAD at the growth factor-dependent sites, probably allowing phosphatases to gain access to BAD at regulatory serine residues because it was shown that the dephosphorylation of BAD at the functional serine residues (Ser-112, -136, and -155) is inhibited when BAD binds to 14-3-3 (12). These results indicate that cell cycle-dependent BAD phosphorylation facilitates apoptosis through the facilitation of growth factor deprivation-induced dephosphorylation of BAD (Fig. 6). Thus, BAD detects the coincidence of the G 2 /M phase and growth factor deprivation to regulate apoptosis.
This framework provides an explanation for why the dissociation of BAD from 14-3-3 at the G 2 /M phase does not induce immediate apoptosis. It has been shown that the phosphorylation of one of the functional serine residues (Ser-155) of BAD suppresses the pro-apoptotic function of BAD by preventing association of BAD and Bcl-xL (17,18). There was no change in the phosphorylation level of BAD at Ser-155 in any cell phase in the presence of IL-3, but the phosphorylation was markedly reduced after IL-3 deprivation ( Fig. 2B and supplemental Fig.  5). Thus, after the dissociation of BAD from 14-3-3 at the G 2 /M phase, BAD has to undergo an additional step (dephosphorylation at Ser-155) before the induction of apoptosis.
Paclitaxel is one of the most important anti-cancer drugs and is used to treat various solid tumors. The drug arrests the cell cycle at the G 2 /M phase by stabilizing microtubules (29). It was shown that the inhibition of Cdc2 activity reduces paclitaxelinduced apoptosis (31). Interestingly, BAD has been implicated in paclitaxel-induced cytotoxicity: the expression of BAD in ovarian cancer cells enhances the cytotoxic effects of paclitaxel (32). These results suggest that both Cdc2 and BAD are involved in paclitaxel-induced cytotoxicity. Therefore, our findings may underlie the molecular mechanism of the anti-cancer effect of the drug: the prolongation of the G 2 /M phase by paclitaxel enhances the Cdc2-mediated phosphorylation of BAD at Ser-128, which in turn decreases the mitochondrial threshold for apoptosis.
The activation of Cdc2 promotes cell proliferation (33,34). On the other hand, our results indicated, rather paradoxically, that activated Cdc2 increases the probability of cell death. However, many oncoproteins induce both cell cycle progression and apoptosis (34). The expression of transcription factors c-Myc or E2F1, which promotes the progression of cell cycle from the G 1 to the S-phase, was shown to enhance apoptosis in the absence of growth factors (35)(36)(37)(38). In our present study, we showed that Cdc2 activated at the cell cycle progression from the S-phase to the G 2 /M phase enhances apoptosis. These observations suggest that these mechanisms may function as a safety device for suppressing inappropriate cell proliferation. Indeed, the roles of BAD have been suggested in the regulation of cell proliferation. BAD-deficient mice suffer from large B cell lymphoma (6). On the other hand, BADoverexpressing mice have a severely decreased number of T cells in the thymus (39). Therefore, the cell cycle-dependent regulation of apoptosis by BAD may have an important role in regulating cell proliferation under physiological and pathological states.