CDK2-Dependent Phosphorylation of FOXO1 as an Apoptotic Response to DNA Damage

See allHide authors and affiliations

Science  13 Oct 2006:
Vol. 314, Issue 5797, pp. 294-297
DOI: 10.1126/science.1130512


The function of cyclin-dependent kinase 2 (CDK2) is often abolished after DNA damage. The inhibition of CDK2 plays a central role in DNA damage–induced cell cycle arrest and DNA repair. However, whether CDK2 also influences the survival of cells under genotoxic stress is unknown. Forkhead box O (FOXO) transcription factors are emerging as key regulators of cell survival. CDK2 specifically phosphorylated FOXO1 at serine-249 (Ser249) in vitro and in vivo. Phosphorylation of Ser249 resulted in cytoplasmic localization and inhibition of FOXO1. This phosphorylation was abrogated upon DNA damage through the cell cycle checkpoint pathway that is dependent on the protein kinases Chk1 and Chk2. Moreover, silencing of FOXO1 by small interfering RNA diminished DNA damage–induced death in both p53-deficient and p53-proficient cells. This effect was reversed by restored expression of FOXO1 in a manner depending on phosphorylation of Ser249. Functional interaction between CDK2 and FOXO1 provides a mechanism that regulates apoptotic cell death after DNA strand breakage.

In response to DNA damage, mammalian cells activate checkpoint pathways that induce cell cycle delay or arrest. Delayed progression of the cell cycle allows time for either the repair of DNA damage or the elimination of genetically unstable cells by apoptosis. CDK2 is a key regulator of DNA damage–activated G1 phase and intra–S phase checkpoints (1). DNA damage leads to the activation of several protein kinases, such as ataxia telangiectasia mutated (ATM), ataxia telangiectasia and Rad3 related (ATR), Chk1, and Chk2 (2), which eventually causes the p53-dependent transcription of the CDK inhibitor p21WAF1 (3) or ubiquitin-dependent degradation of the protein phosphatase Cdc25A (4, 5), or both, thereby inhibiting CDK2 activity and DNA replication. However, it is not clear whether CDK2 plays a role in the regulation of DNA damage–induced cell death.

Activation of FOXO transcription factors induces apoptosis by up-regulating a number of cell death genes, including those encoding the ligand for the death receptor known as Fas or CD95, the Bcl-2–interacting mediator (Bim) of cell death, and the tumor necrosis factor–related apoptosis-inducing ligand (68). Increased expression of these pro-apoptotic proteins is required for cell death to be induced by the DNA damaging agent camptothecin or its derivatives (9, 10). Because CDK2 is a key mediator of most checkpoint functions, we hypothesized that the activities of FOXO proteins might be regulated by DNA damage signals through functional interactions with CDK2. We therefore sought to determine whether CDK2 could phosphorylate the FOXO1 protein. Endogenous CDK2 proteins were immunoprecipitated from NIH 3T3 cells, and in vitro kinase assays were conducted with glutathione S-transferase (GST)–FOXO1 fusion proteins (Fig. 1A) as substrates. Two fragments of FOXO1, FO1-2 and FO1-3, were phosphorylated by immunoprecipitated CDK2 but not by control immunoprecipitates (Fig. 1B). The amount of FOXO1 phosphorylation was comparable to that of the C-terminal segment of retinoblastoma protein (RB), a well-characterized CDK2 substrate (Fig. 1B). No phosphorylation of the control GST protein was detected. Reconstituted complexes of purified bacterially produced GST-CDK2 with either GST–cyclin E or GST–cyclin A phosphorylated the same fragments of the FOXO1 protein as did the endogenous CDK2 (Fig. 1C). These effects were abolished when the CDK inhibitor p27KIP1 was included (fig. S1A). These results indicate that CDK2 directly phosphorylates the FOXO1 protein in vitro.

Fig. 1.

CDK2 phosphorylates FOXO1 at Ser249 in vitro. (A) Schematic diagram of the five overlapping GST-FOXO1 fusion proteins, designated as FO1-1 to FO1-5. Numbers in parentheses show the start and end amino acid number of the FOXO1 fragment in each fusion protein. The forkhead DNA binding domain (FKD), nuclear localization signal domain (NLS), nuclear export signal domain (NES), and transactivation domain (TAD) of the protein are indicated. (B) (Top) In vitro kinase assays with immunoprecipitated (IP) CDK2. Mr(K), relative molecular weight (kilodaltons); 32P, phosphorus-32. Immunoprecipitates of an antibody to CDK2 or nonspecific immunoglobulin G (IgG) from NIH 3T3 cells were used in kinase assays together with 2 μg of substrates: GST-FOXO1 fusion proteins, GST, or RB-C [a recombinant RB protein with its C-terminal segment (residues 701 to 928) fused to the maltose-binding protein]. (Middle) Protein substrates indicated by Coomassie blue staining. (Bottom) Immunoblotting (IB) of immunoprecipitated CDK2 in NIH 3T3 cells. (C) Reconstituted CDK2 kinase assays. Reconstituted cyclin E/CDK2 (0.12 μg, top) and reconstituted cyclin A/CDK2 (0.1 μg, bottom) were used in vitro in kinase assays with 0.2 μg of substrates. (D) Reconstituted cyclin E/CDK2 kinase assays with the use of the wild-type (WT) and S249A single–amino acid substitution mutant of FO1-2 as substrates.

CDK2 and other CDKs often recognize and phosphorylate the motif of serine or threonine followed by proline. Only one such site (Ser249-Pro250) exists in FO1-2. We demonstrated by in vitro kinase assays that CDK2 phosphorylates the FOXO1 protein at Ser249 and Ser298, with a preference at Ser249 (Fig. 1D and fig. S1, C to E). These findings suggest that CDK2 phosphorylates FOXO1 primarily at the Ser249 residue within the consensus CDK-phosphorylation motif (fig. S1B).

To test whether FOXO1 is phosphorylated at Ser249 in vivo, we generated a phosphorylation-specific antibody against a peptide containing the phosphorylated Ser249. The antibody specifically recognized the wild-type FOXO1 but not the Ser249→Ala249 (S249A) mutant (Fig. 2A). This activity was blocked by a peptide containing the phosphorylated Ser249 but not by the corresponding nonphosphorylated peptide (fig. S2A). The antibody-mediated reaction was sensitive to the treatment of proteins with protein phosphatase (fig. S2B). Silencing of endogenous CDK2 by a pool of small interfering RNAs (siRNAs) led to a decrease in Ser249 phosphorylation (Fig. 2B). Moreover, phosphorylation of Ser249 was increased in cells transfected with an active CDK2 mutant, CDK2-Thr14→Ala14-Tyr15→Phe15 (CDK2-AF), that is unable to undergo inhibitory phosphorylation (5) (fig. S2C). Furthermore, phosphorylation of Ser249 was low during the G1 phase and increased as cells progressed through the S phase (Fig. 2C). Thus, these data indicate that CDK2 phosphorylates FOXO1 at Ser249 in vivo. CDK2 also interacts with the FOXO1 protein in vitro and in vivo (fig. S3).

Fig. 2.

CDK2 phosphorylates FOXO1 at Ser249 in vivo. (A) LNCaP cells were transfected with wild-type FLAG-tagged FOXO1 or S249A mutant. Ectopically expressed FOXO1 proteins were immunoprecipitated with an antibody to FLAG and blotted with an antibody against the phosphorylated FOXO1 at Ser249 (S249-p) or an antibody to FOXO1. (B) LNCaP cells were transfected with CDK2-specific siRNA or a nonspecific control siRNA. At 48 hours after transfection, the levels of endogenous CDK2, phosphorylated Ser249, and total endogenous FOXO1 proteins were analyzed by Western blots. Immunoblotting of extracellular signal–regulated kinase 2 (Erk2) was included as a loading control. (C) Ser249 phosphorylation of FOXO1 during the cell cycle. LNCaP-FOC4 cells were arrested in the M phase by treatment with nocodazole (1 μg/ml) for 24 hours and then released from arrest for the indicated times. Cell lysates were immunoprecipitated with an antibody to FLAG (M2) and analyzed by immunoblotting as indicated. A cell cycle profile obtained from flow cytometry was also included.

FOXO1 functions primarily as a transcription factor. Ectopic expression of CDK2 and cyclin E decreased the transcriptional activity of FOXO1 (Fig. 3A). The transcriptional activity of FOXO1 was largely enhanced by cotransfection of the tumor suppressor PTEN or by the conversion of its three Akt phosphorylation sites to alanine (FOXO1-AAA) in LNCaP prostate cancer cells, where phosphatase and tensin homolog (PTEN) is mutated and Akt is highly active. However, this effect was abolished by the expression of wild-type CDK2 but not by catalytically inactive kinase-dead CDK2-KD (Fig. 3, A and B). Notably, the effect of CDK2 was abrogated by treatment of cells with roscovitine, an inhibitor of CDK2 (Fig. 3B). Mutations in FOXO1 (S249A and S298A) that make it impervious to CDK2-mediated phosphorylation enhanced the activity of FOXO1-AAA, and the inhibitory effect of CDK2 was abrogated in this double mutant (Fig. 3B). The transcription activity of FOXO1 was largely diminished by a phosphomimicking mutation at Ser249 (Ser249→Asp249) (fig. S4D). Taken together, these results suggest that CDK2-induced inhibition of transcriptional activity of FOXO1 is mediated primarily by the phosphorylation of Ser249. Expression of endogenous FOXO-activated genes, including p27KIP1, Bim, and manganese superoxide dismutase (7, 11, 12), was increased by inhibition of CDK2 (figs. S5C and S6, A and B). Moreover, Ser249 was adjacent to the three-arginine motif (fig. S6D), which is critical for nuclear localization of FOXO proteins (13). Overexpression of CDK2 resulted in the cytoplasmic localization of wild-type FOXO1 but not the phosphorylation-resistant mutant S249A of FOXO1 (Fig. 3, C and D), suggesting that CDK2-induced trafficking of FOXO1 from the nucleus to the cytoplasm may be mediated by phosphorylation at Ser249.

Fig. 3.

CDK2 inhibits FOXO1 through Ser249 phosphorylation. (A) Effect of CDK2 on PTEN-induced activation of FOXO1. LNCaP cells were cotransfected with a FOXO1 luciferase reporter construct containing three copies of the insulin-responsive sequence (3xIRS), and the plasmids are indicated. Luciferase activities were measured at 36 hours after transfection. The relative luciferase units (RLU) were determined by normalizing the measured units of firefly luciferase with the measured renilla luciferase activity. *P = 0.024; **P = 0.027. (B) Effect of mutations at the CDK2 phosphorylation sites in FOXO1 on its transcriptional activity. LNCaP cells were transfected with plasmids as indicated. Cells were treated with the CDK2 inhibitor roscovitine (15 μM) or vehicle at 24 hours after transfection. Cell lysates were subjected to luciferase assays at 48 hours after transfection. Luciferase measurement was performed as in (A). *P = 0.017; **P = 0.005; ***P = 0.002. Error bars in (A) and (B) indicate SD among three individual experiments. (C) Cellular localization of ectopically expressed wild-type FOXO1 in DU145 cells, which is a PTEN-positive cell line, and cellular localization of FOXO proteins are minimally affected by Akt. Plasmids for FLAG-FOXO1 were cotransfected with cyclin E and CDK2, or with an empty vector into DU145 cells. At 36 hours after transfection, cells were serum-starved for 24 hours. Cells were treated with roscovitine (40 μM) or vehicle for 3 hours before cells were subjected to immunofluorescent chemistry. Quantification of a representative experiment is shown (bar graphs). C, cytoplasm; C+N, cytoplasm and nucleus; N, nucleus. Similar results were obtained from three independent experiments. Inset on merge picture in the middle panel shows the staining of a cell with fluorescein isothiocyanate (FITC)–conjugated antibody to FLAG in higher magnification. DAPI, 4′,6′-diamidino-2-phenylindole. (D) Effect of CDK2 on the cellular localization of ectopically expressed FOXO1-S249A mutant in DU145 cells. Cells were transfected with plasmids as indicated and analyzed as in (C).

Camptothecin is a DNA damaging agent that inhibits the religation function of topoisomerase I by inducing the covalent attachment of topoisomerase I to DNA. This effect results in the activation of the DNA double-strand break checkpoint pathways and inhibition of CDK2 activity. Camptothecin treatment inhibited the activity of CDK2 in DU145 prostate cancer cells (Fig. 4A). Phosphorylation of endogenous FOXO1 at Ser249 was also abolished (Fig. 4A). Camptothecin treatment also diminished the phosphorylation of FLAG-FOXO1 at Ser249 in both LNCaP and PC-3 prostate cancer cells (Fig. 4B and fig. S7, A and B). Phosphorylation of FOXO1 at Ser249 also decreased in LNCaP cells exposed to γ irradiation (fig. S7C). Because neither DU145 nor PC-3 cell lines express functional p53 (14), these data suggest that CDK2-mediated phosphorylation of FOXO1 can be inhibited by DNA damaging agents through a p53-independent mechanism.

Fig. 4.

Activation of FOXO1 is essential for cell death in response to DNA damage. (A) DU145 cells were treated with (+) or without (–) camptothecin (CPT) (1.25 μM). At 16 hours after treatment, endogenous FOXO1 proteins were subjected to immunoprecipitation (IP) with antibody to FOXO1 and immunoblotted (IB) with antibodies for Ser249-p and FOXO1. CDK2 activities were measured by in vitro kinase assays with histone H1 as substrate. (B) LNCaP cells were cotransfected with an expression vector of FLAG-FOXO1, a pool of Chk1-specific siRNAs, or a control siRNA. At 48 hours after transfection, cells were treated with CPT (1.25 μM) or mock-treated for 16 hours. Immunoprecipitation and immunoblotting were performed as in (A). Immunoblotting of Erk2 was used as a loading control. (C) DU145 cells were transfected with expression plasmids for enhanced green fluorescent protein as well as plasmids and siRNAs as indicated. At 24 hours after transfection, cells were sorted and replated in medium containing 2.5% of fetal bovine serum. At 24 hours after replating, cells were collected and immunoblotted for FOXO1 expression (top lane) or treated with CPT (1.25 μM). At 48 hours after CPT treatment, three sets of cells were analyzed for apoptosis (bar graphs). *P = 0.002; **P = 0.0018. Error bars indicate SD among three individual experiments. (D) Model of the role for CDK2-mediated phosphorylation and regulation of FOXO1 in the apoptotic response to DNA damage.

Genotoxic agents inhibit the activity of CDK2 through a checkpoint signaling pathway that includes the protein kinases ATM and ATR, Chk1 and Chk2, and the protein phosphatase Cdc25A independently of the p53-p21WAF1 pathway (4, 5). To determine whether the Chk1 and Chk2 proteins function in regulating FOXO1 phosphorylation at Ser249, we silenced endogenous Chk1 in LNCaP cells or treated them with the Chk1 inhibitor UCN-01. Not only did the basal amount of FOXO1 phosphorylation at Ser249 increase, but the camptothecin-induced decrease in FOXO1 phosphorylation was also partially blocked (Fig. 4B and fig. S7D). Moreover, silencing of Chk2 by two independent siRNAs diminished the camptothecin-induced decrease in Ser249 phosphorylation of FOXO1 (fig. S7E). These findings indicate that DNA damaging agents regulate FOXO1, at least in part, by controlling the Chk1- or Chk2-dependent signaling pathways.

Silencing of FOXO1, but not of FOXO3a or FOXO4, markedly diminished camptothecininduced cell death (Fig. 4C and figs. S9F and S10A). Camptothecin-induced apoptosis was restored by forced expression of silencing-resistant FOXO1 proteins (FOXO1-WTsr or FOXO1-S249Asr). The effect of FOXO1-WTsr, but not FOXO1-S249Asr, was inhibited by the CDK2-AF mutant (Fig. 4C). Forced expression of FOXO3a or FOXO4 also restored DNA damage–induced apoptosis in cells with silencing of FOXO1, and this effect was not blocked by the expression of CDK2-AF (Fig. 4C and fig. S10B). This observation was not unexpected, given that FOXO factors share the same binding sequences of target genes (15) and that the functions of FOXO3a and FOXO4 were not regulated by CDK2 (fig. S9, A to F). In addition, the effect of FOXO1 on DNA damage–induced apoptosis was consistent with its transcriptional activities (fig. S10C). FOXO1-mediated cell death was also observed in p53-proficient cells, including the MCF7 breast cancer line, the “normal” (nontumorigenic) MCF10A breast epithelial cell line, and wild-type mouse embryonic fibroblasts (fig. S11, A to D). Thus, FOXO1 contributes to DNA damage–induced apoptosis in both p53-deficient and p53-proficient cells, especially in tumor cells. Activation of FOXO proteins (for example, FOXO4) but not p53 also mediates tumor-specific cell death that results from the silencing of sirtuin homolog 1, a regulator of p53 and FOXO pathways in response to stress (16). Thus, it is important to dissect the roles of individual FOXO proteins in the selective killing of tumor cells under genotoxic and metabolic stress conditions.

The DNA damage response functions as a biological barrier that inhibits cancer progression (17, 18). Defects in this pathway often allow human cancers to progress through the CDK2-dependent pathways (5). Cyclin E, another regulator of CDK2, is often deregulated in cancer (19). Deregulation of cyclin E and defects in the DNA damage signaling networks may synergistically favor the survival of nascent cancer cells and therefore promote tumorigenesis through the inhibition of FOXO1. Given that the FOXO1 mechanism can be restored by cancer therapeutic agents, such as γ irradiation and chemotherapeutic drugs, CDK2-mediated regulation of FOXO1 represents a previously unrecognized pathway (Fig. 4D) that links DNA damage to cell death (4, 5, 20).

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S11


References and Notes

View Abstract

Navigate This Article