Modulation of (cid:1) -Catenin Phosphorylation/Degradation by Cyclin-dependent Kinase 2*

(cid:1) -Catenin functions as a downstream component of the Wnt/Wingless signal transduction pathway, and in-appropriate control of cytosolic (cid:1) -catenin is a crucial step in the genesis of several human cancers. Here we demonstrate that cyclin-dependent kinase 2 (CDK2) in association with cyclin A or cyclin E directly binds to (cid:1) -catenin. In vivo and in vitro kinase assays with cyclin-CDK2 demonstrate (cid:1) -catenin phosphorylation on residues Ser 33 , Ser 37 , Thr 41 , and Ser 45 . This phosphorylation promotes rapid degradation of cytosolic (cid:1) -catenin via the (cid:1) -TrCP-mediated proteasome pathway. Moreover, cyclin E-CDK2 contributes to rapid degradation of cytosolic (cid:1) -catenin levels during G 1 phase by regulating (cid:1) -catenin phosphorylation and subsequent degradation. In this way, CDK2 may “fine tune” (cid:1) -catenin levels over the course of the cell cycle. (cid:1) is

The fact that abnormal expression of ␤-catenin or mutation of its regulatory region is a crucial step in the genesis of several human cancers (2,26,27) suggests that ␤-catenin may act as an oncogene in mammalian cells. Consistent with that idea, disruption of ␤-catenin/TCF activity in colorectal cancer cells (28) or overexpression of APC in normal cells induces rapid G 1 arrest (29). It thus appears that fine control of ␤-catenin levels probably contributes to the maintenance of a normal cell cycle. It is noteworthy in that regard that levels of cytosolic ␤-catenin fluctuate over the course of the cell cycle, increasing during S phase, peaking in late G 2 /M phase, and then abruptly declining in G 1 phase (30,31). The mechanism responsible for this oscillation in ␤-catenin levels remains unknown, however.
Cyclin-dependent kinases (CDKs) are key regulators of cell cycle progression in eukaryotic cells (e.g. cyclin E-CDK2 and cyclin A-CDK2 promote progression from G 1 phase into and through S phase) (32, 33), and periodic activation of cyclin-CDK complexes is largely responsible for the characteristic sequence of cell cycle events, including mitosis, DNA synthesis, chromatin assembly, and other biosynthetic processes.
It is known that a short sequence motif (RXL) present in cyclin-CDK2 substrates is necessary for the binding of the enzyme and phosphorylation of the substrates (34 -36). The presence of the RXL motif in the primary sequence of ␤-catenin suggests that it, too, is a substrate for the cyclin-CDK2 complex. The fact that most cyclin-CDK2 substrates are phosphorylated in a cell cycle-dependent manner (e.g. p27 and BRAC1 are phosphorylated by cyclin-CDK2 during the G 1 -to-S transition) (37)(38)(39) led us to ask whether phosphorylation of ␤-catenin is catalyzed by cyclin-CDK2 and whether that phosphorylation and subsequent degradation is cell cycle-dependent.
Cell Cultures and Transfections-HeLa, Rat-1, and HEK293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Invitrogen). Transient transfections were carried out using LipofectAMINE Plus (Invitrogen) according to the manufacturer's instructions. Total amounts of transfected DNA were equalized by the addition of empty pCMV vector.
Cell extracts were prepared by lysing the cells in binding buffer for 1 h at 4°C, followed by centrifugation at 12,000 ϫ g. GST binding experiments were then carried out as described above, except that 500 g of cell extract was used instead of the in vitro translated products. The samples were analyzed by SDS-PAGE followed by immunoblotting.
Cell Fractionation, Immunoprecipitation, and Immunoblotting-Cell fractionation was carried out as described (42) using subconfluent cells. The resultant supernatants were analyzed as the cytosolic lysates. For immunoprecipitation, subconfluent cells were extracted for 1 h at 4°C in buffer containing 1% Nonidet P-40, 50 mM Tris-HCl (pH 7.5), 0.25% sodium deoxycholate, 150 mM NaCl, and a mixture of protease and phosphatase inhibitors (Sigma), after which the mixture was centrifuged at 12,000 ϫ g for 15 min at 4°C.
For co-immunoprecipitation in vivo, 800 g of cell extract were incubated with 4 l of anti-CDK2, anti-cyclin E, or anti-cyclin A Ab or 2 l of anti-␤-catenin Ab at 4°C for 4 h. The resultant immune complexes were then immobilized on protein A-or G-Sepharose beads (Amersham Biosciences) and washed three times with extraction buffer. The immunoprecipitates were subjected to SDS-PAGE and immunoblot analysis. Proteins were separated by electrophoresis in an 8 or 10% acrylamide gel, after which Western blots were probed with the indicated Abs. Primary Abs were visualized using an ECL chemiluminescence kit (Amersham Biosciences).
Cell Synchronization and Cell Cycle Analysis-To synchronize Rat-1 cells at the G 1 or early S phase, subconfluent cells were cultured in the presence of 50 nM rapamycin (Calbiochem) for 18 h or 2 g/ml aphidicolin (Calbiochem) for 18 h. The DNA content of cells was determined by analyzing parallel cultures using flow cytometry. Cell cycle analysis was carried out as described (31).

Cyclin-CDK2
Interacts with ␤-Catenin-In order to identify cell cycle regulators interacting with ␤-catenin, we mixed bacterially produced GST-␤-catenin with cell extracts and then identified the bound proteins by immunoblot analysis. As shown in Fig. 1A, cyclin A and E were found to preferentially bind to ␤-catenin. Because both are known to associate with CDK2, the membrane was also probed with anti-CDK2 Ab, which indeed revealed the presence of CDK2 (Fig. 1A). To confirm that the detected CDK2 were interacting directly with ␤-catenin, we assessed the ability of GST-␤-catenin to bind in vitro translated, [ 35 S]Met-labeled CDK2. Fig. 1B shows that GST-␤-catenin bound CDK2, whereas GST alone did not, suggesting that ␤-catenin associates with cyclin A-CDK2 and/or cyclin E-CDK2 complexes in vitro. Finally, co-immunoprecipitation assays carried out using HEK293T and Rat-1 fibroblast cells ( Fig. 1C) confirmed that complexes containing ␤-catenin, cyclins, and CDK2 also exist in vivo.
The RXL motif is required for stable binding of cyclin-CDK2 complexes to substrates (34 -36). We showed this also to be true for ␤-catenin by testing the ability of a series of deletion mutants to interact with CDK2 ( Fig. 1D). Mutants BD2, BD3, and BD4, which all contained RXL motifs, were all able to bind GST-CDK2 (Fig. 1D). On the other hand, BD1, which was composed of the N terminus of ␤-catenin and lacked an RXL motif, did not bind GST-CDK2 (Fig. 1D).
Cyclin-CDK2 Phosphorylates ␤-catenin on Ser 33 , Ser 37 , Thr 41 , and Ser 45 -We next determined the extent to which ␤-catenin is phosphorylated by cyclin-CDK2 in vitro. Anti-CDK2 Ab was used to immunopurify cyclin-CDK2 complexes from Rat-1 cells, after which we examined the ability of the isolated complexes to phosphorylate purified GST-␤-catenin or histone H1, a known CDK2 substrate in the presence of either with 50 M roscovitine (a CDK2 inhibitor) (44) or with Me 2 SO (the vehicle). CDK2 immune complexes from Rat-1 cells in the presence of Me 2 SO efficiently phosphorylated both GST-␤-catenin and histone H1, whereas their phosphorylation was dramatically inhibited in the presence of roscovitine ( Fig. 2A). To identify the specific CDK2-associated cyclin responsible for phosphorylating ␤-catenin, we carried out in vitro kinase assays with purified recombinant cyclin A-CDK2 and cyclin E-CDK2 and found that ␤-catenin was phosphorylated efficiently by cyclin A-CDK2 and cyclin E-CDK2. Thus, ␤-catenin proteins can serve as substrates for cyclin A-CDK2 and cyclin E-CDK2 (Fig. 2B). We then carried out in vitro kinase assay to reveal whether the cyclin-CDK2 complex associated with ␤-catenin retain kinase activity. In vitro kinase assays results showed that the cyclin-CDK2 complex interacting with ␤-catenin has the kinase activity, and histone H1 phosphorylation was mediated by a ␤-catenin-associated cyclin-CDK2 complex, but not by GSK-3␤, since the effect was completely blocked by roscovitine, whereas LiCl had no effect (Fig. 2C).
Cyclin-CDK2 Down-regulates Cytosolic ␤-Catenin-To determine the extent to which phosphorylation by cyclin-CDK2 facilitates degradation of cytosolic ␤-catenin, we transfected HEK293T cells with WT ␤-catenin along with cyclin A-CDK2 or cyclin E-CDK2, after which cytosolic lysates from the transfectants were analyzed by immunoblotting. Overexpression of WT ␤-catenin led to increased levels of cytosolic ␤-catenin; however, cytosolic WT ␤-catenin was then efficiently down-regulated by cyclin A-CDK2 and cyclin E-CDK2 (Fig. 4A), whereas ␤-catenin S37A was largely resistant to cyclin E-CDK2 and cyclin A-CDK2 (Fig. 4B), suggesting that degradation of cytosolic ␤-catenin is dependent on its phosphorylation by cyclin-CDK2.
Cyclin-CDK2 Down-regulates ␤-Catenin via the ␤-TrCPmediated Proteasome Pathway-Phosphorylated ␤-catenin is known to be subject to ubiquitination, which is reflected in the accumulation of high molecular weight species (7,14). To determine whether cyclins-CDK2-mediated down-regulation of ␤-catenin proceeds via the ubiquitin/proteasome system, Rat-1 cells were treated with MG-132, a 26 S proteasome inhibitor, and/or roscovitine, after which the cell lysates were immunoblotted with anti-␤-catenin or anti-phospho-33/37/ 41-␤-catenin Ab. Treatment with MG-132 led to the accumulation of high molecular weight ␤-catenin species detectable by both Abs, indicating ubiquitination of the phosphorylated protein (Fig. 5A). The necessity for phosphorylation was confirmed by the finding that accumulation of high molecular weight ␤-catenin species was dramatically diminished by treatment with roscovitine (Fig. 5A). That ␤-TrCP-mediated ubiquitination was the pathway via which ␤-catenin is degraded was confirmed by the observation that expression of a dominant-negative ␤-TrCP F-box deletion mutant able to bind to phosphorylated ␤-catenin, but unable to form a SCF ␤-TrCP ubiquitin ligase complex (17), inhibited cyclin A-or cyclin E-CDK2-induced degradation of ␤-catenin (Fig. 5B). Consistent with the above results (Fig. 5B), treatment with MG-132 suppressed cyclin E-CDK2-mediated degradation of ␤-catenin in HEK293T cells (Fig. 5C).
Cyclin E-CDK2 Might Be Implicated in Rapid Degradation of Cytosolic ␤-Catenin during G 1 Phase-Previous reports revealed that cytosolic ␤-catenin fluctuates over the course of the cell cycle, increasing during S phase, peaking in late G 2 /M phase, and then abruptly declining in G 1 phase (30,31).
To determine whether the oscillation of cytosolic ␤-catenin levels reflects changes in the rate of its degradation over the course of the cell cycle, we treated cells synchronized at G 1 or Cytosolic lysates were immunoblotted with anti-␤-catenin Ab or anti-GFP Ab. C, HEK293T cells were co-transfected with cyclin E-CDK2 or empty vector (Ϫ) and WT ␤-catenin and then exposed to 20 M MG-132 for 12 h. pEGFP vector (250 ng) was used as an internal transfection efficiency control. Cytosolic lysates were immunoblotted with anti-␤catenin or anti-GFP Ab.

␤-Catenin Phosphorylation by Cyclin-dependent Kinase 2
early S phase in the presence of either Me 2 SO or proteasome inhibitor MG-132 for 2 h, after which levels of cytosolic ␤-catenin were assayed by immunoblotting. We found that inhibition of 26 S proteasome substantially increased levels of cytosolic ␤-catenin in G 1 -arrested cells but had no effect in cells arrested at early S phase (Fig. 6A). Thus, degradation of cytosolic ␤-catenin occurs mainly during G 1 phase.
To test whether phosphorylation of ␤-catenin is also cell FIG. 6. Cyclin E-CDK2-dependent cytosolic ␤-catenin degradation during G 1 phase. A, cells were synchronized at G 1 phase by treatment of 50 nM rapamycin for 18 h or at early S phase by treatment with 2 g/ml aphidicolin for 18 h. Both G 1 -arrested and early S-arrested cells were treated with either Me 2 SO (DMSO) or 20 M MG-132 for 2 h, after which cytosolic lysates were immunoblotted with anti-␤-catenin or anti-tubulin Abs. The level of cytosolic ␤-catenin was quantified by densitometry and normalized to the amount of tubulin. For cell cycle analysis, the DNA content of the cells was determined at each time point using flow cytometry. B, cells were synchronized at G 1 phase by treatment of 50 nM rapamycin for 18 h or at early S phase by treatment with 2 g/ml aphidicolin for 18 h. The arrested cells were lysed and immunoprecipitated (IP) with ␤-catenin Ab and then immunoblotted with anti-phospho-33/37/41-or anti-phospho-41/45-␤-catenin Abs. C, the arrested cells were immunoprecipitated with ␤-catenin Ab, after which the immune complexes were subjected to in vitro kinase assays to monitor ␤-catenin-associated CDK2 and GSK3␤ kinase activities. D, The arrested cells were immunoprecipitated with anti-CDK2 Ab, anti-cyclin A Ab, or cyclin E Ab and then immunoblotted with anti-␤-catenin Ab. The cell lysates were immunoblotted with anti-CDK2 Ab, anti-cyclin A Ab, or cyclin E Ab. cycle-dependent, ␤-catenin was immunoprecipitated from G 1or early S-arrested cells, after which the immunoprecipitates were analyzed using anti-phospho-33/37/41-or anti-phospho-41/45-␤-catenin Abs. As expected, phosphorylation of ␤-catenin was up-regulated to a greater extent in G 1 -arrested cells than in early S-arrested cells (Fig. 6B).
In vitro kinase assays revealed that the phosphorylation of ␤-catenin during G 1 phase was assisted by a ␤-catenin-associated histone H1 kinase (Fig. 6C). To identify the specific cyclin-CDK2 responsible for phosphorylating ␤-catenin during G 1 phase, we did perform a co-immoprecipitation assay. CDK2, cyclin E, or cyclin A was immunoprecipitated from G 1 -or early S-arrested cells, after which the immunoprecipitates were analyzed using anti-␤-catenin Abs (Fig. 6D). As shown in Fig. 6D, the association of ␤-catenin with CDK2 and cyclin E was upregulated to a greater extent in G 1 -arrested cells than early S-arrested cells (Fig. 6D). However, no remarkable change of ␤-catenin-associated cyclin A level was detected throughout the cell cycle. Therefore, cyclin E-CDK2 appears to be responsible for ␤-catenin phosphorylation/degradation during G 1 phase and in this way may regulate the cell cycle-dependent turnover of ␤-catenin.

DISCUSSION
SW480 cells expressing an inactive, truncated form of APC contained high levels of cytosolic and nuclear ␤-catenin (30,52). Association of APC into ␤-catenin is dependent on ␤-catenin phosphorylation by GSK3␤ (53), and their association is prerequisite for degradation of ␤-catenin (12,20,54). In addition, strong interaction of ␤-catenin with APC was detected in the cells entering a new G 1 phase (30), supporting the involvement of APC or GSK3␤ in the modulation of ␤-catenin levels during the cell cycle. However, no remarkable change of ␤-catenin-associated GSK3␤ activity and APC level was detected throughout the cell cycle. Moreover, down-regulation of ␤-catenin could be induced by overexpression of full-length APC in SW480 cells, but inhibition of GSK3␤ activity did not suppress that recovery (55); another mechanism is involved in ␤-catenin phosphorylation/degradation during cell cycle.
Several lines of evidence in this study indicate that cyclins-CDK2 mediate phosphorylation/degradation of ␤-catenin, and cyclin E-CDK2, in particular, facilitates ␤-catenin turnover during G 1 phase. First, ␤-catenin associates with cyclins-CDK2 in vivo and in vitro. Previous studies have shown that a short sequence motif (RXL) is present in a number of cyclin-CDK2 substrates, which is necessary for the binding of cyclin-CDK2 (34 -36). RXL motifs present in ␤-catenin seem to mediate its binding to cyclin-CDK2 complexes. Second, cyclin-CDK2 specifically phosphorylates ␤-catenin on the same residues acted upon by GSK3␤ (Ser 33 , Ser 37 , Thr 41 , and Ser 45 ). It is interesting that cyclin-CDK2 phosphorylates ␤-catenin on residues Ser 33 , Ser 37 , Thr 41 , and Ser 45 , residues that are not in the CDK2 consensus phosphorylation sequence, serine-proline (SP), and threonine-proline (TP) sequence. The consensus phosphorylation sequence for CDKs is a serine or threonine residue followed by a proline residue (56). In fact, ␤-catenin contains three CDK2 consensus sequences (Ser 191 , Ser 246 , and Ser 605 ). Among these CDK2 consensus sequences, Ser 191 and Ser 246 residues are also phosphorylated by cyclin-CDK2 in vitro (data not shown). In this study, we only focused on Ser 33 , Ser 37 , Thr 41 , and Ser 45 phosphorylation. Phylogenetically, GSK3␤ is closely related to CDK2 (57), and therefore the active form of GSK3␤ and its complex are structurally very similar to those of CDK2 and ERK2 (58). Furthermore, GSK3␤ also is one of the proline-directed kinases, which phosphorylate serine or threonine residue flanked by a C-terminal proline residue. For example, GSK3␤ phosphorylates the Ser 501 -Pro 502 sequence of neurofilament-H subunit (59), the Ser-Pro sequence of tau (60), and the Thr 286 -Pro 287 residues of cyclin D1 (61,62). Thus, GSK3␤ can phosphorylate residues such as the SP and TP sequence and also other sequence (e.g. ␤-catenin), suggesting that ␤-catenin would be expected to be a potent substrate for both GSK3␤ and CDK2. On the other hand, phosphorylation of ␤-catenin by CDK2 is quite different from that by GSK3␤, which requires ␤-catenin to be first "primed" by phosphorylation on residue Ser 45 by CKI␣ or protein kinase A (45,46,63,64). In more general terms, GSK3␤ can only phosphorylate substrates that have a priming phosphate at position n ϩ 4 (where n is the target site for phosphorylation) (65). This reflects the fact that GSK3␤'s unique N terminus acts as a pseudosubstrate that is only displaced from the catalytic site by suitably primed substrates (57). However, cyclin-CDK2 has no such requirement. Third, most cyclin-CDK2 substrates are phosphorylated in a cell cycle-dependent manner, particularly during the G 1 /S transition. It is then, for example, that p27 and BRAC1 are phosphorylated by CDK2 (37)(38)(39). Several different mechanisms (cyclin binding, phosphorylation, and association of cyclin-dependent kinase inhibitor) are employed to regulate CDK activity in order to ensure that the cell's normal cycle is tightly controlled (66). In the present study, the association of ␤-catenin with cyclin E-CDK2 and its associated H1 kinase activity was high during G 1 phase, which coincides with maximal levels of ␤-catenin phosphorylation/degradation. By contrast, ␤-catenin-associated GSK3␤ activity remained constant throughout the cell cycle. Fourth, cyclin-CDK2 mediates downregulation of cytosolic ␤-catenin through the mechanism dependent upon the ␤-TrCP-mediated 26 S proteasome pathway. GSK3␤ also phosphorylates ␤-catenin, which is recognized by ␤-TrCP, a specificity component of an ubiquitination apparatus (15,16), and is in turn degraded by 26 S proteasome. Cotransfection of cyclin A-CDK2 or cyclin E-CDK2 with ␤-catenin WT markedly reduced cytosolic ␤-catenin level. In addition, treatment with MG-132 and the introduction of dominantnegative ␤-TrCP (F box deletion mutant) appeared to suppress cyclin-CDK2-mediated degradation of ␤-catenin, indicating that cyclin-CDK2 promotes degradation of ␤-catenin through a mechanism dependent upon cyclin-CDK2-mediated phosphorylation and the ␤-TrCP-mediated proteasome pathway, which seems to be in the same fashion as GSK3␤-mediated ␤-catenin degradation. Fifth, it is suggested that cyclin-CDK2 may regulate cell cycle events in both nucleus and the cytoplasm (67)(68)(69)(70)(71), suggesting that cytosolic cyclin-CDK2 may regulate cytosolic ␤-catenin phosphorylation. A previous report that both cyclin A-and cyclin E-CDK complexes are able to shuttle between the nucleus and the cytoplasm (72) suggests that they could directly phosphorylate both nuclear and cytoplasmic substrates.
Taken together, our findings imply that cyclin-CDK2 probably modulates cytosolic ␤-catenin phosphorylation and its subsequent TrCP-mediated proteasomal degradation, and ␤-catenin phosphorylation/degradation by cyclin E-CDK2 may mediate the fine tuning of ␤-catenin level in G 1 phase.