|Home | About | Journals | Submit | Contact Us | Français|
Sophisticated models for the regulation of mitotic entry are lacking for human cells. Inactivating human cyclin A/Cdk2 complexes through diverse approaches delays mitotic entry and promotes inhibitory phosphorylation of Cdk1 on tyrosine 15, a modification performed by Wee1. We show here that cyclin A/Cdk2 complexes physically associate with Wee1 in U2OS cells. Mutation of four conserved RXL cyclin A/Cdk binding motifs (RXL1 to RXL4) in Wee1 diminished stable binding. RXL1 resides within a large regulatory region of Wee1 that is predicted to be intrinsically disordered (residues 1 to 292). Near RXL1 is T239, a site of inhibitory Cdk phosphorylation in Xenopus Wee1 proteins. We found that T239 is phosphorylated in human Wee1 and that this phosphorylation was reduced in an RXL1 mutant. RXL1 and T239 mutants each mediated greater Cdk phosphorylation and G2/M inhibition than the wild type, suggesting that cyclin A/Cdk complexes inhibit human Wee1 through these sites. The RXL1 mutant uniquely also displayed increased nuclear localization. RXL1 is embedded within sequences homologous to Crm1-dependent nuclear export signals (NESs). Coimmunoprecipitation showed that Crm1 associated with Wee1. Moreover, treatment with the Crm1 inhibitor leptomycin B or independent mutation of the potential NES (NESm) abolished Wee1 nuclear export. Export was also reduced by Cdk inhibition or cyclin A RNA interference, suggesting that cyclin A/Cdk complexes contribute to Wee1 export. Somewhat surprisingly, NESm did not display increased G2/M inhibition. Thus, nuclear export of Wee1 is not essential for mitotic entry though an important functional role remains likely. These studies identify a novel bifunctional regulatory element in Wee1 that mediates cyclin A/Cdk2 association and nuclear export.
Despite broad progress in studies of cell cycle control in eukaryotes, advanced models are lacking for the regulation of mitotic entry in human cells. This regulation is pivotal in cell cycle control, and a better understanding of it may be crucial to improving cytotoxic cancer chemotherapy, the mainstay of cancer treatment. Models of mitotic entry in higher eukaryotes revolve around activation of the cyclin B/Cdk1 (cyclin-dependent kinase 1 or Cdc2) complex, which drives the major events of mitosis. A rise in the cyclin B level triggers mitotic entry in Xenopus egg extracts but not in mammalian cells (15, 47). Inhibitory phosphorylation of Cdk1 on the ATP-binding site residue tyrosine 15 (Y15) has been recognized as a key constraint throughout eukaryotes (29, 42). Wee1 and Myt kinases perform this phosphorylation in vertebrate cells, where Wee1 appears to be dominant (34). Kim and Ferrell and others have recently developed an elegant model for ultrasensitive, switch-like inactivation of Wee1 by cyclin B/Cdk1 in a positive feedback loop that contributes to mitotic entry in Xenopus egg extracts (27).
Although cyclin A(A2)/Cdk2 is traditionally omitted from models of mitotic entry, accumulating evidence from several different approaches suggests that cyclin A/Cdk complexes play roles. Cyclin A levels rise during S phase and peak in G2 before falling abruptly in prometaphase of mitosis (60). Microinjection of cyclin A/Cdk2 complexes in human G2 phase cells was observed to drive mitotic entry (14). Conversely, microinjection of antibodies directed against cyclin A in S-phase cells inhibited mitotic entry without an apparent effect on bulk DNA synthesis (45). In complementary approaches that supported biochemical analyses, cyclin A RNA interference (RNAi) or induction of a dominant negative mutant of Cdk2 (Cdk2-dn), the major cyclin A binding partner, inhibited mitotic entry (13, 15, 21, 37). In these settings, cyclin B/Cdk1 complexes accumulated in inactive, Y15-phosphorylated forms (13, 21, 37). Cdc25 phosphatases, which can reverse this phosphorylation, show reduced activity in this context (37), but increased Cdc25 activity could not readily overcome the arrest (13). RNAi-mediated knockdown of Wee1 was found capable of overriding the arrest mediated by cyclin A RNAi, suggesting that Wee1 is a key rate-limiting factor (13). However, whether and by what mechanisms cyclin A complexes might regulate Wee1 and drive Cdk1 dephosphorylation and mitotic entry have remained unclear.
Recently, genetic studies in mice have reinforced these observations while providing evidence for some cell type differences (24). Although Cdk2 is not essential, in its absence Cdk1 binds more cyclin A and E and provides redundant functions (4, 25, 44). Deletion of the cyclin A gene is lethal for embryos and adults (24). Gene deletion in fibroblasts in vitro did not completely abrogate their proliferation but caused S and G2/M delays. In this setting cyclin E was upregulated, and combined deletion of cyclin E yielded arrest in G1, S, and G2/M phases. Cyclin A gene deletion was alone sufficient to block proliferation of hematopoietic stem cells, suggesting that cyclin A is essential for their proliferation.
Wee1 is regulated on multiple levels, including inhibitory phosphorylation in the amino-terminal regulatory domain (NRD), residues 1 to 292. This region is predicted to be intrinsically disordered (56), and few functional elements have been identified in it. The cyclin B/Cdk1 complex has been thought to be the principal or exclusive kinase responsible for NRD phosphorylation (18, 27, 28). Two sites in the Xenopus embryonic Wee1 NRD, Thr 104 and Thr 150 (referred to here by the homologous residue, T239, in human somatic Wee1), have been identified as Cdk phosphorylation sites that inhibit Wee1 activity (28). Recent studies of Xenopus somatic Wee1 suggest that T239 phosphorylation may antagonize the function of a surrounding motif, dubbed the Wee box (43). This small, conserved region appears to augment the activity of the carboxy-terminal kinase domain.
We show here that cyclin A/Cdk2 complexes directly bind Wee1 as a substrate in human cells. In particular, a conserved cyclin A/Cdk binding RXL motif in the Wee1 NRD is required for efficient T239 phosphorylation. Further analysis revealed that RXL1 is located within a Crm1 binding site that mediates Wee1 export during S and G2 phases. Cyclin A/Cdk2 activity appears to foster Wee1 export, but this export is not essential for mitotic entry. These findings further define roles of cyclin A/Cdk complexes in regulating Wee1 and mitotic entry in human cells and dissect the mechanisms and consequences of Wee1 redistribution during the run-up to mitosis.
Human Wee1A coding region cDNA (Origene Inc., Rockville, MD) was amplified by PCR. Primers used for PCR were the following: W5p, 5′-TTAGGATCCATGAGCTTCCTGAGCCGA-3′; w3p, 5′-GGCCTCGAGTCAGTATATAGTAAGGCT-3′. The PCR product was cloned into mammalian expression vector pCMV-3tag-2A (Stratagene, La Jolla, CA) at BamHI/XhoI sites. The constructs M1-4 (RXL1-4 mutation), M1A (RXL1 with RXL and the +5 leucine mutated to alanines) M2AGL (selective mutation of the R alanine in RXL2), qM (quadruple mutant with all four RXL sites mutated), KD (kinase defective), T239A, M1KD, and NESm (mutation of the proline and phenylalanine residues of the nuclear export signal [NES]) were generated by site-directed mutagenesis using a multimutagenesis kit (Stratagene). Primers for mutagenesis were the following: M1, 5′-CCACACAAGACCTTCGCGGCCGCGCGACTCTTCGACAC-3′; M1A, 5′-AAGACCTTCGCGGCAGCGCGAGCCTTCGACACCCCG-3′; M2, 5′-CTTTTGCAAGTTGGCGCGGGCGCCAGGTATATTCATTC-3′; M2AGL, 5′-CTCCTTTTGCAAGTTGGCGCCGGCTTGAGGTATATTC-3′; M3, 5′-GAAGAGGGCGATAGCGCTTTTGCTGCAAATGAAGTTTT-3′; M4, 5′-GCAGCTGAGGAAGCGGCCGCATTCACTGACCGGA-3′; KD, 5′-GTTATGTTTAAAATAGGTAACCTAGGGCATGTAACAAGGATCTCC-3′; T239A, 5′-CAAGTGAATATTAATCCTTTTGCGCCGGATTCTTTGTTGCTTC-3′; NESm, 5′-GGGCACCCCGGCACACAAGACCGCGCGCAAGCTGC-3′.
U2-OS cells and HeLa cells were cultured in Dulbecco's modified minimal essential medium supplemented with 10% fetal calf serum, penicillin, and streptomycin and maintained in 5% carbon dioxide and humidified air. Hydroxyurea (HU; Sigma-Aldrich, St. Louis, MO) block and release were described previously (21). For the mimosine (Sigma-Aldrich) block, culture medium and transfection complexes were removed and replaced by medium containing 0.5 mM mimosine for 24 h. Lipofectamine 2000 (Invitrogen, Carlsbad, CA) was used for transfection of small interfering RNA (siRNA) and plasmid according to the manufacturer's instructions. Wee1 siRNAs (sense, GGUAUUGCCUUGUGAAUUUUU; antisense, AAAUUCACAAGGCAAUACCUU) were designed by using the siDESIGN Center from Dharmacon Inc. (Chicago, IL). Cyclin A siRNAs (sense, CUAUGGACAUGUCAAUUGUTT; antisense, ACAAUUGACAUGUCCAUAGTA) are from Applied Biosystems Inc. (Foster City, CA). Cdk2 siRNAs (sense, GCUGAAGAGGGUUGGUAUAUU; antisense, UAUACCAACCCUCUUCAGCUU) are from Dharmacon. The U2-OS cell clones with inducible expression of Cdk2 wild type (Cdk2-wt) and Cdk2-dn have been described previously (21). Induction was achieved by removing cells from the dish with trypsin-EDTA, washing twice with phosphate-buffered saline (PBS), and replating in the absence of tetracycline. In synchronized experiments, induction was begun at the start of HU or mimosine treatment. Cells were released by incubation with standard culture medium for the designated times. Cells were treated with 200 μmol/liter olomoucine (LC Laboratories, Woburn, MA), 20 μmol/liter roscovitine (Sigma-Aldrich), 2.5 ng/ml leptomycin B (LMB; Sigma-Aldrich), or vehicle (dimethyl sulfoxide) alone 6 h after mimosine release. Total cell extracts were prepared in E1A lysis buffer as described previously (21) at the designated time points. Nuclear and cytoplasmic fractions were isolated using a CelLytic NuCLEAR extraction kit (Sigma-Aldrich). For gel electrophoresis, nuclear and cytoplasmic fractions were normalized to cell number.
Immunoblotting and immunoprecipitation were performed as described previously (21). Antibodies directed against cyclin A (H-432 or BF683), cyclin E (C-19 or HE111), cyclin B1 (C-19 or H-433), Cdk2 (M2 or D-12), Cdk1 (C-19 or 17), Crm1 (H-300 or C-1), and Wee1 (C-20 or B-11) were from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-CREB, anti-phospho-Cdk tyrosine 15, and anti-Myc-tag (9B11) antibodies were from Cell Signaling Technology (Danvers, MA). Anti-alpha-tubulin and antiactin antibodies were from Sigma-Aldrich. Anti-glutathione S-transferase (anti-GST) antibody was from GE Healthcare Life Sciences (Piscataway, NJ). The antihemagglutinin (anti-HA) antibody was mouse monoclonal clone 12CA5 (Roche, Indianapolis, IN). Fluorescein isothiocyanate-CD20 antibody was from BD Biosciences (San Jose, CA). Rabbit polyclonal anti-Wee1-phospho-T239 antibody was generated by Covance Immuno Technologies (Denver, PA). It was generated against and purified on a column with the phosphopeptide CKKVNINPFpTPDSL and counterselected against the unphosphorylated peptide. The anti-Wee1 phospho-S121/S123 (S121/123) antibody was kindly provided by Nobumoto Watanabe (Wako, Japan). Secondary antibodies were from GE Healthcare Life Sciences. ImageJ software (National Institutes of Health) was used for quantification of immunoblotting bands. Values for binding or phosphorylation of Wee1 mutants were expressed as ratios to those observed for the wild-type protein expressed in parallel and were calculated following normalization of signal to expression of each Wee1 protein, assayed by immunoblotting.
Cell fixation for flow cytometry with propidium iodide and CD20 staining was performed as described previously (21). Data were collected on a Becton Dickinson single laser three-fluorescence FACScan flow analyzer and analyzed with FlowJo and Dean/Jett/Fox (Free Star, Ashland, OR) or ModFit (www.vsh.com/products/mflt) software. Cells were fixed for immunofluorescence as described previously (21) and stained with a Myc-tagged mouse monoclonal antibody (Cell Signaling Technology, Danvers, MA), an Alexa Fluor 568 goat anti-mouse secondary antibody (Invitrogen), and 4′,6′-diamidino-2-phenylindole (DAPI; Sigma-Aldrich). Images were acquired on a Nikon Eclipse E800 microscope with a Nikon DXM1200C camera and using ACT-1C software (Nikon, Medville, NY). In each of two to three independent experiments, cells were scored by a blinded observer.
Full-length wild-type and mutant (M1, qM, KD, and M1KD) Wee1 coding regions were cloned as BamHI/XhoI fragments into the GST fusion protein expression vector pGEX-4T-1 (GE Healthcare Life Sciences) to generate GST-Wee1 fusion proteins. pGEX-4T-1 and pGEX-Wee1-wt/M1/qM/KD/M1KD were transformed into Escherichia coli strain Rosetta 2 (kindly provided by Yoshihiro Matsumoto, Fox Chase Cancer Center, Philadelphia, PA). Bacterial culture and GST fusion protein purification followed instructions in the GST SpinTrap purification module (GE Healthcare Life Sciences) manual except that an induction temperature of 20°C was used. For pulldown assays, 2 μg of GST or GST-Wee1-wt/M1/qM protein was immobilized with glutathione beads (BD Pharmingen, San Jose, CA) in binding buffer (50 mM HEPES, pH 7.0, 50 mM NaCl, 0.1% NP-40, 1 mM EDTA, 10% glycerol, 1% bovine serum albumin, and protease inhibitors). U2-OS cell extracts were precleared by incubation with GST-glutathione beads at 4°C for 30 min, and then cells were incubated with immobilized GST or GST-Wee1 fusion proteins at 4°C for 2 h. Beads were washed four times with binding buffer and eluted with 2× concentrated sample loading buffer for sodium dodecyl sulfate-gel electrophoresis.
DNA and protein sequences were analyzed by Lasergene (DNASTAR Inc., Madison, WI) and Vector NTI (Invitrogen). Predictions of structural order in Wee1 were made using the DIOSPRED server (http://bioinf.cs.ucl.uk.disopred/) (56). Multiple sequence alignments of NES residues from Wee1, cyclin D1, and Snurportin were performed with ClustalW (53). The sequence of Wee1 was aligned to the Snurportin NES templates using the program MolIDE (8), and side chain conformations of the NES peptide were predicted with the program SCWRL (8, 55). This program allowed all Wee1 side chains to move while avoiding clashes with the Crm1 NES binding cleft, whose residues were held rigid during modeling. Structure figures were produced with the program Chimera (University of California, San Francisco).
A paired t test was used to test each hypothesis comparing G2/M fractions obtained following expression of wild-type Wee1 or a specific mutant (or vector). This approach was taken to account for possible subtle differences in conditions among the several independent experiments analyzed for each wild-type-mutant pair.
We chose U2-OS human osteogenic sarcoma cells for these studies because they respond well to synchronization methods and are widely used for human cell cycle studies (12, 33, 36, 54). We first asked whether endogenous Cdk2 and Wee1 are physically associated. Cdk2 is the major Cdk partner of cyclin A in human cells and does not bind appreciably to cyclin B in the presence of Cdk1 (21, 35, 37, 38). Cdk2 was readily detected in somatic Wee1 immunoprecipitates (IPs) derived from asynchronous cells (Fig. (Fig.1A,1A, left). Conversely, Wee1 was readily detected in Cdk2 IPs (Fig. (Fig.1A,1A, right). Similar results were obtained using extracts from synchronized cultures enriched for cells in late S and G2 phase (Fig. (Fig.1B1B and data not shown; see below for comparable flow cytometry data). RNAi directed against Cdk2 confirmed that the Cdk2 antibody used is specific for Cdk2 and that the band detected in Wee1 IPs was indeed Cdk2 (Fig. (Fig.1C).1C). These results suggested that Wee1 and Cdk2 are physically associated in human cells, including cells in late interphase.
We then tested whether Wee1 associates with cyclin A or E complexes, using asynchronous cells. Wee1 was readily detected in cyclin A IPs but not cyclin E IPs, despite comparable precipitation of Cdk2 (Fig. (Fig.2A).2A). These results suggest that Wee1 associates preferentially with cyclin A complexes. Because Cdk2 binds most cyclin A present in U2-OS cells during late S and G2 phases (37), most Wee1 bound to cyclin A is likely in complex with Cdk2. Some complexes may also contain Cdk1, which associates with a low but increasing fraction of cyclin A as U2-OS cells approach mitosis (38). Indeed, low-level association of Wee1 with Cdk1 was detected (data not shown).
In principle, Wee1 might bind Cdk complexes as substrates, and/or Cdk complexes might bind Wee1 as substrate. The preferential association of Wee1 with cyclin A/Cdk complexes suggested that the cyclin may dictate association. Recent data have shown that catalytic activity of Cdc28, the major Cdk in Saccharomyces cerevisiae, is required for stable association with the budding yeast Wee1 homologue Swe1 (18). This observation, together with lack of binding of a phosphorylation-resistant Swe1 mutant, suggested the model that phosphorylation of Swe1 by Cdc28 stabilizes the association. We therefore asked whether Cdk2 kinase activity is required for stable association with Wee1. We used U2-OS clones with induction of Cdk2 wild type (Cdk2-wt) or a well-characterized catalytically inactive, dominant negative (Cdk2-dn) mutant (21). The inducible Cdk2 proteins are tagged with HA epitopes to distinguish them from endogenous Cdk2. Reciprocal immunoprecipitations confirmed association of endogenous Wee1 with the induced Cdk2-wt (Fig. 2B and C). In contrast, Wee1 association with Cdk2-dn was not readily detected (Fig. 2B, C), suggesting that Cdk2 kinase activity is required for robust association. These observations further confirm association of Wee1 with Cdk2 complexes and support the notion that Wee1 may be recognized as a substrate of cyclin A/Cdk2 complexes.
To further address whether Wee1 is recognized as a substrate of cyclin A/Cdk2 complexes, we hypothesized that this interaction might be directed by RXL/Cy motifs. Such motifs mediate Cdk2 binding and phosphorylation of pRb, p27, and other substrates via binding to a hydrophobic pocket in cyclin A (1, 7, 52, 59). Most defined RXL interactions have been with cyclin A or, less commonly, cyclin E/Cdk2 complexes but not cyclin B/Cdk1 complexes. Direct evidence has recently been obtained that cyclin A/Cdk2 complexes interact more strongly in vitro with model RXL-containing peptides than do cyclin B/Cdk2 complexes (6). An apparent anomaly is Myt, which contains a C-terminal RXL sequence that mediates binding to cyclin B/Cdk1 complexes (5).
Wee1A contains four RXL sequences (Fig. (Fig.3A)3A) that are conserved throughout the known vertebrate somatic Wee1 family members (Fig. (Fig.3B).3B). These sites lie, respectively, within the long NRD of Wee1 (RXL1), the kinase domain (RXL2 and RXL3), and the short carboxy-terminal region (RXL4) (Fig. (Fig.3A).3A). In the two sites within the catalytic region (RXL2 and RXL3), the R and L residues are not conserved among protein kinases (17), suggesting that they may not be required for catalytic activity. We noted that RXL1 is the most conserved RXL sequence. It is also conserved in flies and, in addition, possesses a leucine in the +5 position relative to the arginine. A leucine residue is preferred in this position in cyclin A recognition motifs (6, 9). The NRD of Wee1 is overall poorly conserved and predicted to be largely disordered by the DISOPRED server (56). This method uses different criteria, judging in part whether the amino acid content is consistent with packing interactions needed for structured domains. In contrast, RXL1 was within a small region of potential order (Fig. (Fig.3C).3C). Of note, the Wee box containing T239 occupied a nearby region of potential order (Fig. (Fig.3C).3C). Such short regions predicted to be ordered within otherwise largely disordered regions have been identified as potential functional regions by Mohan and colleagues (39).
To test the potential role of the RXL sites in mediating Wee1 binding and/or phosphorylation by cyclin A/Cdk complexes, we mutated these sites. The RXL residues were mutated to alanines in each site, except the central (X) residues in sites 2 and 3, which are conserved in protein kinases. M1 refers to the RXL1 mutant, M2 to the RXL2 mutant, and other designations use the same form. Amino-terminal Myc epitope tags were added, and the primary structure of each coding region was confirmed by total sequencing. In transient expression studies, each mutant was expressed at levels comparable to the wild-type protein except M2, which displayed lower levels (data not shown). Examination of the crystal structure of the Wee1 kinase domain (46, 51) suggests that the RXL2 leucine is involved in packing interactions. Loss of these interactions might render the protein unstable. We therefore generated an RXL2 mutant with only the arginine replaced, designated M2AGL. This protein was expressed at levels comparable to the level of the wild-type protein (data not shown). We then generated a quadruple mutant (qM) with all four RXL sites altered as in the individual mutants, including M2AGL.
We expressed wild-type Wee1, M1, and qM by transient transfection in U2-OS cells and compared their stable association with cyclin A/Cdk2 complexes in cells enriched in late S and G2 phases (data not shown) (see below). Cdk2 immunoprecipitations recovered modestly and distinctly lower levels of M1 and qM, respectively (Fig. (Fig.4A).4A). Cyclin A immunoprecipitations yielded similar results (Fig. (Fig.4B).4B). Binding of cyclin A to the RXL mutants was reduced more than that of Cdk2, an observation consistent with the possibility of some reduction in binding of cyclin A/Cdk1 complexes and/or independent binding of cyclin E/Cdk2 complexes. These observations suggest that M1 and the other RXL sites contribute to stable Wee1 association with cyclin A/Cdk2 and possibly cyclin A/Cdk1 complexes. To exclude secondary effects (e.g., cell cycle effects) that might influence associations in vivo, we assayed binding in vitro. We generated amino-terminal GST fusion proteins with full-length wild-type Wee1, M1, and qM. We compared the ability of the respective GST-Wee1 fusion proteins and glutathione beads to pull down cyclin A and Cdk2 from extracts of cells enriched in S and G2 phase. GST fusions of M1 and qM showed modest and distinctly less binding, respectively, than GST-Wee1-wt, providing further evidence that the RXL sites in Wee1 bind cyclin A/Cdk2 complexes (Fig. (Fig.4C).4C). It was more difficult to detect binding of cyclin B/Cdk1 complexes to the GST-Wee1 fusion proteins, but no differences were observed between the Wee1 mutants (Fig. (Fig.4D),4D), consistent with the established binding preferences of RXL sites for cyclin A complexes.
RXL motifs have been implicated in directing phosphorylation of particular sites among local alternatives (9, 49). Multiple Cdk phosphorylation sites have been identified in budding yeast Swe1 and Xenopus somatic and embryonic Wee1 proteins. Phosphorylation of T239 stands out in the vertebrate proteins. This site was independently identified as a Cdk1 phosphorylation site in Xenopus somatic and embryonic Wee1 proteins (28, 43). T239 is within the Wee-box, and evidence has been obtained that phosphorylation at this site abrogates Wee-box-mediated stimulation of Wee1 activity, thereby inhibiting Wee1 (28).
To investigate T239 phosphorylation in human somatic Wee1, we generated and purified polyclonal antibodies directed against a T239-phosphorylated peptide (P-T239). This antibody recognized endogenous Wee1 in a manner sensitive to phosphatase treatment (Fig. (Fig.5A).5A). The antibody also efficiently recognized wild-type Wee1 expressed by transient transfection and showed decreased reactivity against M1 and qM (Fig. (Fig.5B).5B). To confirm the antibody's relative specificity for the P-T239 protein, we generated a T239A mutant. The P-T239 antibody showed only modest reactivity against this protein, likely due to some reactivity against surrounding sequences present in the peptide antigen (Fig. (Fig.5C).5C). Relative to T239A, P-T239 reactivity with M1 and qM was slightly greater, suggesting that T239 phosphorylation of M1 and qM, though reduced, was not eliminated. For comparison, we examined phosphorylation of another nearby site suggested to be Cdk dependent and to influence the stability of human somatic Wee1 (58). Reactivity with an antibody demonstrated to recognize this site (57) was unaffected in M1 (Fig. (Fig.5D).5D). Less T239 phosphorylation of M1 than the wild type was also seen when both exogenous proteins were expressed at lower levels, approaching those of endogenous Wee1 (data not shown).
We then tested whether cyclin A knockdown reduced T239 phosphorylation of endogenous Wee1 using synchronized cells harvested as the peaks entered mid-late S phase (data not shown). Cyclin A RNAi reduced T239 phosphorylation (Fig. (Fig.5E).5E). Cdk2 knockdown also reduced T239 phosphorylation of endogenous Wee1 (data not shown). We conclude that RXL1 and the other RXL sites contribute to Wee1 binding by cyclin A/Cdk2 complexes and that RXL1 is required for optimal phosphorylation of T239.
As an initial test of the function of M1 and T239A mutants, we assayed their impact on Cdk Y15 phosphorylation in transfected cells. For these studies, we generated a KD derivative. We substituted asparagine for aspartate 145 of the catalytic site, the residue that orients the terminal phosphates of ATP for the phospho-transfer and is mutated in our extensively validated Cdk2-dn (Cdk2 numbering) (21). In vitro auto-phosphorylation confirmed that the KD mutant was inactive (see below). Wild-type Wee1 increased Cdk Y15 phosphorylation compared to empty vector in U2-OS cells (Fig. (Fig.5F)5F) and 293 cells (data not shown). M1 was more potent than the wild type (Fig. (Fig.5F),5F), and T239A was as potent as M1 (Fig. (Fig.5F,5F, right). The KD mutant was less potent than the M1 or T239A mutant. Some Cdk Y15 phosphorylation seen with the KD mutant may be due to protection of the Cdk from phosphatase activity and/or trapping of cells in early S phase (see next section). These findings indicated that M1 and T239A might be more potent Cdk Y15 kinases in vivo.
These results prompted us to examine the ability of M1 and other Wee1 mutants to inhibit progression through G2/M. Asynchronous U2-OS cells were transiently transfected with the respective constructs and analyzed by flow cytometry. Wild-type Wee1 imposed a dose-dependent increase in G2/M fraction (Fig. (Fig.6A)6A) (see below). M1 was more potent than the wild type in increasing the G2/M fraction. In contrast, M3 was less potent than either the wild type or M1 (Fig. (Fig.6A).6A). M2AGL had a mild phenotype like M3, and M4 had no substantial effect (data not shown). An M1 mutant with the additional mutation of the +5 position leucine to alanine (M1A) showed an increase in the G2/M fraction similar to that of M1 (data not shown).
We explored the effects of Wee1 further using synchronized cells. Cells were transfected, synchronized in G1 (10), released, and harvested at intervals. Sample flow cytometry profiles are shown in Fig. Fig.6B6B from cells harvested at regular intervals. M3 again showed reduced accumulation of cells at G2/M, the opposite effect of M1, underscoring the specificity of the results. M3 also delayed S-phase progression, an effect phenocopied by the KD mutant (data not shown) (see below). Expression of M1 again resulted in a greater accumulation of cells in G2/M than expression of the wild type (Fig. (Fig.6B).6B). These results directly demonstrate that the increased G2/M fraction observed with expression of M1 is due to cell cycle delay, excluding the formal possibility of acceleration of passage through other phases of the cell cycle.
We sought to further characterize the kinase activity of the Wee1 mutants. Immunoprecipitation-kinase assays proved problematic, even with inactive Cdk complexes as substrates. We were, however, able to perform auto-phosphorylation, a well-established assay of intrinsic kinase activity in Wee1 proteins (28), on the GST fusion proteins. GST-Wee1-wt and GST-M1 were similarly active in this assay, whereas GST-KD, GST-M1KD, and GST-qM were inactive (data not shown). Thus, despite preservation of residues conserved in kinases, the kinase domain mutations in qM (most likely M3 and possibly also M2AGL) appear to compromise intrinsic kinase activity (data not shown). We infer that efficient S-phase progression appears to require Wee1 kinase activity, but this issue requires further investigation and is not the focus here.
We extended these studies to other cell types and clarified the stage of the delay. M1 also showed a stronger G2/M delay than the wild type in HeLa and 293 cells (data not shown). Coimmunofluorescent staining for Wee1 (Myc) and DNA (DAPI) in U2-OS and HeLa cells demonstrated that the vast majority of wild-type- and M1-expressing cells in each cell type contained uncondensed DNA and that M1 resulted in a greater fraction (data not shown) (see Fig. Fig.88 below), consistent with a delay in interphase (G2). These results suggest that M1 is a hyperfunctional mutant of Wee1 with enhanced ability to block mitotic entry and that RXL1 fosters inactivation of Wee1.
To test this notion further, we performed an extensive series of experiments in both asynchronous and synchronized cells in which pairwise transfections were performed in parallel, using wild-type Wee1 as an internal control. The wild-type protein consistently increased the G2/M fraction compared to the empty vector (Fig. (Fig.7A),7A), and M1 was more potent than the wild type (Fig. (Fig.7B7B).
Exogenous Myt1 has been shown to be capable of inhibiting mitotic entry in a kinase-independent fashion (31). We tested whether the augmented G2/M delay imposed by M1 was dependent on kinase activity. Transient transfection experiments showed that M1KD displayed lower G2/M inhibition than M1 (data not shown). These observations suggest that the M1 phenotype of enhanced G2/M arrest depends, at least in part, on its kinase activity.
We then assessed the cell cycle inhibitory activity of the T239A mutant using synchronized cells. T239A also displayed a stronger G2/M arrest than the wild type (Fig. (Fig.7C).7C). The G2/M accumulation observed with T239A was on average modestly less than that of M1, but this difference was not statistically significant in pairwise comparisons (data not shown). These findings confirm that inhibitory phosphorylation of T239 is conserved in human somatic Wee1 and suggest that this modification is a functional target of cyclin A/Cdk2 complexes acting through RXL1.
Wee1 has been described as a nuclear protein that relocates to the cytoplasm around mitosis, a shift expected to facilitate mitotic entry (3, 19, 26). We therefore compared the subcellular localization of exogenously expressed wild-type Wee1 and RXL mutants using indirect immunofluorescence (IF). In asynchronous cultures, wild-type Wee1 displayed a range of localizations, from exclusively nuclear in 13% of stained cells to exclusively cytoplasmic in 18% (Fig. (Fig.8A,8A, left). The fraction of cells with partial or exclusive cytoplasmic localization (ca. 70%) was much greater than the mitotic fraction scored by flow cytometry, suggesting that this localization was not restricted to mitotic cells with disrupted nuclear envelopes and, therefore, was not a passive outcome of mitosis. In striking contrast, M1 displayed exclusively nuclear staining in more than 80% of cells (Fig. (Fig.8A).8A). This degree of nuclear localization was unique to M1. M2AGL, M3, and a KD derivative of the wild-type protein showed only modestly greater nuclear localization than the wild type (potentially due to their imposing early S phase delays) (Fig. (Fig.66 and and88 and data not shown), and M4 and T239A showed no consistent difference from the wild type (Fig. (Fig.8A).8A). These observations indicate that T239 phosphorylation is not necessary for export and that reduced T239 phosphorylation is unlikely to account for the reduced export of M1.
We then compared wild-type and M1 localization in cells enriched in mid-late S phase by mimosine block in G1 and release for 10 h. This time point was before a Wee1-mediated G2 arrest in most cells (Fig. (Fig.6).6). The wild type was largely (ca. 80%) cytoplasmic (Fig. 8B and C), out of proportion to the mitotic fraction (ca. 15%) (Fig. (Fig.66 and data not shown). Consistent with this notion, IF assays demonstrated that the vast majority of wild-type-expressing cells had an interphase DNA staining pattern (data not shown). Thus, the exogenous wild-type protein appears to be largely exported from the nucleus prior to mitosis. In contrast, M1 was again predominantly (>80%) nuclear (Fig. 8B and C).
We extended these experiments to another cell type, HeLa, with similar results (Fig. (Fig.8D8D).
To test whether the subcellular distribution of M1 might be secondary to increased kinase activity of this mutant, we compared the subcellular distribution of M1 and the KD version (M1KD). The subcellular distribution of the M1KD mutant was very similar to that of M1 itself (Fig. (Fig.8E)8E) and showed greater nuclear localization than the KD mutant (with an intact RXL1) (Fig. (Fig.8A).8A). These results suggest that the increased nuclear localization of M1 is not dependent on M1 kinase activity or increased inhibition of Cdk proteins.
Given the striking nuclear localization of M1, we considered the possibilities that RXL1 may itself regulate or be embedded within an NES or nuclear localization sequence (NLS). The surrounding sequences do not match a consensus for either signal (22, 30, 41), and the M1 mutation, which removes two basic residues, would be predicted to weaken rather than strengthen an NLS. Furthermore, a Wee1 mutant entirely lacking the NRD also localized largely to the nucleus (data not shown). This observation suggested that an NRD NLS, if present, was not essential for Wee1 nuclear localization and that the NRD might harbor an NES. A Crm1 consensus sequence can mediate Crm1 binding so tightly as to prevent dissociation (30). Many functional Crm1 binding sequences only loosely match such a motif. Examination of the Wee1 primary sequence surrounding RXL1 reveals such a loose match to the consensus sequence and the NES of the prototypic export cargo Snurportin (Fig. (Fig.9A).9A). These residues are conserved throughout metazoan Wee1 proteins and fill the potentially structured domain surrounding RXL1 (Fig. (Fig.3C).3C). Crm1-dependent export of cyclin D1 is directed by a similar loose match to the Crm1 consensus, including proline as the first (somewhat atypical) hydrophobic residue (Fig. (Fig.9A)9A) (2).
As an initial test of whether Wee1 might undergo Crm1-dependent nuclear export, we transfected synchronized cells with wild-type Wee1, allowed these cells to progress into S phase, treated half with the Crm1 inhibitor LMB, and assayed Wee1 localization by IF assay in cells enriched in late S/G2 phases (data not shown). LMB treatment abolished cytoplasmic localization of Wee1 in this assay (Fig. (Fig.9B).9B). This finding supports the notion of Crm1-dependent export of Wee1 but does not rule out indirect effects.
We then asked whether endogenous Wee1 and Crm1 are physically associated. Reciprocal coimmunoprecipitation from S-phase-synchronized cells demonstrated such association (Fig. (Fig.9C).9C). This observation supports direct Crm1-mediated export of Wee1.
Crystal structures of Crm1 in complex with Snurportin have recently been solved (11, 40). Using the Crm1-Snurportin structure as a template, we aligned the putative Wee1 NES to that of Snurportin. We used a side chain rotamer library (8) to predict the conformation of Wee1 residues that would best fit in the Crm1 NES binding cleft, which was held rigid during the modeling process. The resulting model is shown in Fig. Fig.9D9D (see Movie S1 in the supplemental material for the rotation of structures). Wee1 residues P175, F179, L182, and L184 readily fit the binding cleft as the hydrophobic residues of the NES (Fig. (Fig.9A),9A), while hydrophilic residues were more solvent exposed (Fig. (Fig.9E9E).
As an independent test of the role of the potential NES, we therefore mutated P175 and F179 (together) to alanine residues, leaving RXL1 intact (NESm) (Fig. (Fig.9A).9A). After transient expression in U2-OS cells, NESm was almost exclusively nuclear (Fig. (Fig.9F).9F). We conclude that RXL1 is embedded within a Crm1-dependent NES. Cyclin A/Cdk2 binding by RXL1 may also contribute to localization (see below).
To confirm that Crm1-dependent export of Wee1 does not depend upon overexpression, we examined the subcellular distribution of M1 following siRNA-mediated knockdown of endogenous Wee1 and low-level expression of exogenous proteins. Immunoblotting confirmed that Wee1 knockdown was effective and reduced Y15 phosphorylation of Cdk proteins (Fig. 10A). We titrated expression of exogenous Myc-tagged proteins down to levels within twofold of the endogenous protein (Fig. 10B). At this level of expression, the distribution of exogenous wild-type Wee1 was more nuclear than after expression at higher levels, consistent with the literature on the endogenous protein (3). This observation suggests that a factor needed for Wee1 nuclear localization may be limiting. However, M1 was still markedly more nuclear than the wild type (Fig. 10C), indicating that this difference is maintained over a wide expression range.
The subcellular localization of endogenous Wee1 could not be assessed well by IF assay even though a variety of fixation methods and antibodies were tried (data not shown). We therefore assayed its distribution by biochemical fractionation. LMB treatment of cells enriched in late S phase resulted in increased retention in the nucleus (Fig. 10D). Adjusted for the loading controls, LMB increased the nuclear-to-cytoplasmic ratio from 0.84 to 3.57. Residual Wee1 in the cytoplasm fraction after LMB treatment may reflect contamination of this fraction with nuclear proteins.
The NESm mutant allowed us to test whether nuclear localization per se affected T239 phosphorylation or Wee1-mediated cell cycle inhibition. We first confirmed that this mutation compromised Crm1 binding. When expressed to similar levels, NESm associated much less strongly with Crm1 IPs than wild-type Wee1 (Fig. 11A). Association of M1 with Crm1 was also reduced (Fig. 11A). However, T239 phosphorylation in NESm was unaffected (Fig. 11B). Somewhat surprisingly, we also observed no apparent effect of the NESm mutation on the ability of Wee1 to impose a G2/M delay (Fig. 11C). The simplest interpretation of these results is that the reduced T239 phosphorylation in M1 is due to loss of RXL1-mediated Cdk binding and not reduced nuclear export. Furthermore, the enhanced G2/M inhibition displayed by M1 appears to be derived largely from reduced T239 phosphorylation rather than increased nuclear localization (see Discussion).
We then tested further whether cyclin A/Cdk2 complexes foster Wee1 cytoplasmic redistribution. We expressed wild-type Wee1 by transient transfection in synchronized cells enriched in mid-S phase, with or without treatment with two well-characterized Cdk1/2 inhibitors, olomoucine and roscovitine (50). The Cdk inhibitors were added 6 h after mimosine release so that progression through S phase was only moderately delayed. In addition, the Cdk inhibitor-treated cells were fixed 4 h later than untreated cells to match cell cycle positions as closely as possible (15% and 18% more S phase cells for olomoucine and roscovitine) (data not shown). Wee1 localization was examined by IF assay. Both Cdk inhibitor treatments substantially increased the nuclear localization of wild-type protein (about 40% more cells with exclusively or largely nuclear staining) (Fig. 12A). These findings provide additional evidence that Cdk1/Cdk2 activity contributes to net nuclear export of Wee1. There is little cyclin B/Cdk1 activity in the cell cycle phases examined, modest cyclin A/Cdk1 activity, and abundant cyclin A/Cdk2 activity (21, 37, 38). The simplest interpretation is that cyclin A/Cdk2 and possibly cyclin A/Cdk1 complexes contribute to nuclear export of Wee1. In complementary assays, partial knockdown of cyclin A by RNAi in synchronized cells also substantially shifted the distribution of cotransfected wild-type Wee1 toward the nucleus (Fig. 12B).
We then examined localization of endogenous Wee1. Although the IF signal was weak, increased nuclear localization was seen in synchronized, roscovitine-treated cells (data not shown). As another assay, we performed subcellular fractionation with and without inhibition of cyclin A/Cdk activity by induction of Cdk2-dn. We again used synchronized cells enriched for late S and G2 phases (data not shown). Cdk2-dn induction during S phase was achieved without a dramatic shift in cell cycle profile (data not shown). Wee1 distribution was shifted in favor of the nucleus (Fig. 12C). Adjusted for the loading controls, the nuclear-to-cytoplasmic ratio changed with induction from 0.65 to 1.64. Note that the mobility of cytoplasmic Wee1 is also increased by Cdk2-dn induction, consistent with reduced Wee1 phosphorylation. With roscovitine treatment, the nuclear-to-cytoplasmic ratio changed from 1.0 to 2.13 (data not shown). Although Wee1 levels were reduced by Cdk2-dn induction, they were not reduced by Cdk inhibitor treatment (data not shown), suggesting that the localization shift is not secondary to reduced expression. These results suggest that efficient export of endogenous Wee1 is also dependent on Cdk activity.
Wee1 was identified in yeast from mutants that displayed accelerated entry into mitosis, resulting in small cell size (16, 48). These and related studies established that Wee1-mediated Y15 phosphorylation of cyclin B/Cdk1 complexes is a fundamental constraint on mitotic entry in eukaryotes. Evidence that Wee1 can, in turn, be inactivated by Cdk-dependent phosphorylation led to a model in which cyclin B/Cdk1 complexes in higher eukaryotes inactivate Wee1 in a positive feedback loop (47). However, the events that drive the feedback loop past the threshold needed for commitment to mitotic entry have remained obscure in human somatic cells, where a rise in cyclin B levels near mitosis is not routinely observed. Mounting evidence has implicated the dominant Cdk complexes in the preceding cell cycle phases, cyclin A/Cdk2 complexes, and, to a lesser extent, cyclin A/Cdk1 complexes as regulators of mitotic entry and Y15 phosphorylation of Cdk1 (24). These observations have suggested a modified model in which cyclin A/Cdk complexes drive cyclin B/Cdk1 activation past the threshold (13, 14, 15, 21). However, the mechanisms have not been fully defined. Recent data suggest that Wee1 is rate limiting when cyclin A/Cdk activity is compromised (13).
We have reported here that cyclin A/Cdk2 complexes physically associate with Wee1 in S and G2 phase human somatic cells. The preferential association of Wee1 with cyclin A complexes over cyclin E complexes and active over inactive Cdk2 complexes (or inactive Cdk1 complexes [data not shown]) suggests that Wee1 may be recognized as a cyclin A/Cdk2 substrate. The lack of association in vivo with inactive CDKs could reflect either a need for Cdk phosphorylation of Wee1 to stabilize the complex, as observed for Swe1 (18), and/or an inability of inactive Cdk1/Cdk2 complexes to free themselves from competitor substrates (see model below). In Swe1, elimination of many potential phosphorylation sites was required to prevent stable association (18). In any event, these observations are consistent with the notion that Wee1 is recognized as a cyclin A/Cdk substrate.
In further definition of this interaction, we identified an RXL motif, found in many cyclin A substrates, that is conserved in the relatively unstructured amino-terminal regulatory domains of metazoan Wee1 proteins and located in a small region independently predicted to have potential structure. A mutant of this site, M1, demonstrated reduced binding to cyclin A/Cdk2 complexes and reduced phosphorylation of the nearby residue T239. T239 phosphorylation has been implicated in negative regulation of Xenopus Wee1 proteins by Cdk1 complexes and appears to inactivate the Wee box, a domain that augments Wee1 kinase activity (43). RXL1 was the most conserved RXL sequence. Moreover, an RXL1 mutant uniquely displayed enhanced ability to arrest cells in G2. A T239A mutant imposed similarly enhanced G2/M delays, confirming that this phosphorylation is inhibitory in human Wee1 and a functionally significant target of RXL1. Thus, we have identified a molecular interaction through which cyclin A/Cdk complexes can regulate Wee1 and drive mitotic entry.
Evidence has been obtained by others that degradation of human somatic Wee1 is fostered by Cdks (57). Although we did not observe obvious differences in stability between wild-type Wee1 and RXL mutants, this property was not a major focus of our studies, and differences could have been obscured in settings of overexpression.
The reduced cyclin A/Cdk2 binding of the qM mutant compared to M1 suggests that one or more of the other RXL sites may be functional binding sites. T239 phosphorylation does not seem to be further reduced in the qM mutant beyond that of M1. Although expression of M3 (and to a lesser degree M2AGL) delays S phase transit, so does the KD mutant, and initial studies suggest that M3 is catalytically inactive. Therefore, the early S-phase delay observed for M3 may be secondary to misfolding of the kinase domain rather than loss of an RXL3 function that is opposite to that of RXL1. The packing interactions of the RXL2 leucine evident in the crystal structure of the kinase domain (51) argue against its serving as a binding site. M4 has little phenotype.
M1 was also unique in showing dramatically increased nuclear localization in S- and G2-phase cells. This led us to discover that RXL1 is embedded within a Crm1-dependent NES. The function of the NES was supported by the physical association of Wee1 with Crm1, disruption of Crm1 binding in M1 and NESm, and independent blockade of Wee1 nuclear export in M1, NESm, or LMB treatment. Molecular modeling of Crm1-Wee1 NES binding based on the recently defined Crm1-Snurportin crystal structures revealed a good fit within the Crm1 hydrophobic cleft, extending over the predicted structured region around RXL1.
How does cyclin A/Cdk activity contribute to cytoplasmic redistribution of Wee1? Possibilities include activation of nuclear export or cytoplasmic retention signals or inactivation of an NLS. RXL1 may direct additional Wee1 phosphorylation that mediates such an event. Canonical (S/T-P-X-R/K) and noncanonical (S/T-P) Cdk phosphorylation sites have been identified in Swe1 and Xenopus Wee1 NRDs, and a number of such potential sites are present in the human somatic Wee1 NRD. Other kinases may also contribute to Wee1 redistribution. A carboxy-terminal site in Wee1 has been reported to be phosphorylated by Akt, fostering binding to 14-3-3θ and localization in the cytoplasm (26). Cyclin A/Cdk activity might facilitate this mechanism, potentially through the RXL4 site.
It is somewhat surprising that the NESm did not show enhanced G2/M inhibition in our assays. Export of Wee1 is evidently not essential for mitotic entry and may not foster it. One explanation for the lack of an observed delay in mitotic entry of the NESm is that cytoplasmic Wee1 may be needed to prevent early activation of cyclin B/Cdk1 complexes on centrosomes (23). Thus, the role of Wee1 export is currently obscure. However, it is quite possible that such export fosters mitotic entry under more limiting conditions, such as in cell types or conditions (e.g., DNA damage) with less robust Cdk activity. Further work will be needed to define such potential conditions. Alternatively, Wee1 export may render subsequent mitosis more faithful.
A major caveat to models in which cyclin A/Cdk2 complexes drive mitotic entry has been the fact that their activity rises gradually through S and G2 phases. Mitotic entry must be prevented until DNA synthesis and centrosome duplication are complete and then must be fully and promptly executed to avoid genomic instability (20). This conundrum can be resolved if mitotic entry driven by cyclin A/Cdk complexes exhibits an ultrasensitive response, with bistable states separated by a threshold level of activity (27, 61). Kim and Ferrell have provided evidence for ultrasensitivity in inactivation of Xenopus embryonic Wee1 by cyclin B/Cdk1 complexes (27). Competition between Wee1 and alternative cyclin B/Cdk1 substrates contributed to the switch-like inactivation of Wee1. We suggest a revised model for human somatic cells in which cyclin A/Cdk complexes contribute to Wee1 inactivation (Fig. (Fig.13).13). In this model, an ultrasensitive response is based in part on competition between alternative S and G2 RXL-containing substrates and RXL1-mediated cyclin A/Cdk2 phosphorylation of Wee1. Inactivation of Wee1 could be completed by cyclin A/Cdk1 and cyclin B/Cdk1 complexes (data not shown). This model is by no means exclusive. Activation of Cdc25 phosphatases by Cdks (32, 37) and other events are also likely to contribute to mitotic entry.
We thank Tina Chen for contributing to the analysis of Wee1 localization and Sam Litwin for statistical analysis.
This work was supported by NIH grant R01 GM065514 (to G.H.E.) and benefited from NCI institutional support for the Flow Cytometry and Molecular Modeling Facilities at Fox Chase (P30 CA006927).
None of the authors has any conflict of interests to disclose.
Published ahead of print on 26 October 2009.
†Supplemental material for this article may be found at http://mcb.asm.org/.