Phosphorylation of the CCAAT Displacement Protein (CDP)/Cux Transcription Factor by Cyclin A-Cdk1 Modulates Its DNA Binding Activity in G 2 *

Stable DNA binding by the mammalian CCAAT displacement protein (CDP)/Cux transcription factor was previously found to be up-regulated at the G 1 /S transi- tion as the result of two events, dephosphorylation by the Cdc25A phosphatase and proteolytic processing, to generate an amino-truncated isoform of 110 kDa. In S phase, CDP/Cux was shown to interact with and repress the core promoter of the p21 WAF1 gene. Here we demon-strate that DNA binding by p110 CDP/Cux is down-mod-ulated as cells progress into G 2 . Accordingly, cyclin A- Cdk1 was found to bind to CDP/Cux and modulate its DNA binding activity in vitro and in vivo . Interaction with CDP/Cux required the presence of both cyclin A and a cyclin-dependent kinase (Cdk)-activating kinase-activated Cdk1 and involved the Cut homeodomain and a downstream Cy motif. Phosphorylation of serines 1237 and 1270 caused inhibition of DNA binding in vitro . In cotransfection studies, cyclin A-Cdk1 inhibited CDP/ Cux stable DNA binding and prevented repression of the p21 WAF1 reporter. Synchronization— cells G , cells maintained 3 days FBS/DMEM (serum-starved). Synchronization in S phase performed double thymidine block procedure (67). Cells were cultured overnight in DMEM plus 10% FBS supplemented with 2 m M thymidine, the next day, cultured for 10 h in DMEM plus 10% FBS, further incubated overnight in the presence of 2 m M thymidine (Thy (cid:2) 0 h). To allow cells to progress in the cell cycle, the medium was replaced with DMEM plus 10% FBS, and cells were harvested 2 and 4 h later (Thy (cid:2) 2 h, Thy (cid:2) 4 h). FACS analysis was performed as previously described (32). Plasmid Description— Plasmids for many bacterially expressed fu- sion proteins have been described: GST-CR3, GST-CR3HD, GST-HD( (cid:2) Cy) (18), GST-(1007-1505) (68). Plasmids for expression of CDP/ Cux proteins in both bacteria and mammalians cells were prepared by inserting various fragments from the murine Cux (GenBank TM acces- sion number NM 009986 ) or human CDP/Cux (GenBank TM accession number NM 001913) cDNA into the specified vectors. For expression of glutathione S -transferase (GST), the following fragments from the human CDP/Cux cDNA were inserted into the bacterial expression vector pGEX-3X (Amersham Pharmacia Biotech): SCCL, nt 613–1315, amino acids 191–424; CR1L, nt 1605–2273, amino acids 522–744; CR1HD, nt 1605–2019 (cid:2) 3737–3949 added in-frame, aa 522–744 (cid:2) 1232–1302; CR2L, nt 2861–3413 amino acids 938–1124; HD, nt 3626–3941 amino acids 1195–1299; CTD, Peptide Binding Assays— sequence each Control immunoprecipitation cyclin

The CDP 1 /Cux/Cut transcription factors are a family of evolutionarily conserved homeodomain proteins that contain four DNA binding domains, three regions called Cut repeats 1, 2, and 3 and the Cut homeodomain (reviewed in Ref. 1). The first member of the family to be discovered was the Drosophila melanogaster Cut protein. Several lethal and viable mutations within the cut locus have been reported to mediate phenotypic effects in tissues as diverse as the femur, wing margin, embryonic sensory organs, malpighian tubules, trachea, and some structures in the central nervous system (2)(3)(4). One mutation preventing the function of a distant wing-specific enhancer caused the formation of truncated or cut wings (5,6). This is the phenotype that gave its name to the locus cut.
In vertebrates, the CCAAT-displacement activity was first characterized in sea urchin, early B cells, and myeloid precursor cells (7,8). Purification of this activity from HeLa cells and subsequent cloning of the corresponding cDNA revealed that the CCAAT displacement protein, CDP, was the human ortholog of Cut (9,10). Other cDNAs have been isolated from other species. In particular, the mouse cDNA was called Cux (Cut homeobox) in accordance with the nomenclature rules (11). The term CDP/Cux will be used in the remainder of this manuscript to designate the mammalian protein.
Several reports have described CDP/Cux as playing a role in precursor cells as a transcriptional repressor that down-modulates genes which later become expressed in terminally differentiated cells (8,(12)(13)(14)(15). This function was ascribed to the ability of CDP/Cux to prevent the interaction of various transcriptional activators with their binding sites (16). DNA binding studies using recombinant fusion proteins suggested that this "CCAAT displacement activity" was carried out by the combined action of Cut repeats 1 and 2, which exhibit rapid but transient binding to DNA sequences containing two half-sites loosely conforming to the CGAT or CAAT consensus positioned in either orientation and at various distances from each other (17). Apart from this transient DNA binding activity, Cut repeat 3 and the Cut homeodomain can cooperate to allow stable interaction with DNA sequences resembling the ATCGAT consensus (18 -21). The carboxyl-terminal region of the protein was shown to contain two active repression domains that can recruit histone deacetylase activity (22,23). Thus, CDP/Cux can repress gene expression by two mechanisms, competition for binding site occupancy and active repression.
Other studies report that CDP/Cux represses gene expression in immature B and T cells via its binding to AT-rich sequences within matrix-attachment regions (24 -27). The exact mechanism of repression is still a matter of debate but could involve competition with activators and/or the recruitment of a chromosomal locus to a distinct nuclear compartment.
A role for CDP/Cux in cell cycle progression has been inferred from a number of reports. CDP/Cux was found to be a component of the histone nuclear factor D (HiNF-D), whose presence on various histone promoters coincides with its up-regulation in the cell cycle (28 -31). Interaction of CDP/Cux with a consensus binding site or with histone H4 gene promoter sequences (as part of HiNF-D) was found to be up-regulated as cells progress from G 1 to S phase (32,33). CDP/Cux was shown to bind to the core promoter of the p21 WAF1/CIP1 gene, and in transient transfection assays, CDP/Cux repressed a p21 WAF1/CIP1 reporter, whereas an antisense CDP/Cux construct was able to restore p21 WAF1/CIP1 expression levels in S phase (32). An increase in DNA binding at the G 1 /S transition involved two regulatory events, dephosphorylation of the Cut homeodomain by the Cdc25A phosphatase and specific proteolytic cleavage of CDP/ Cux between CR1 and CR2 to generate an amino-terminaltruncated protein of 110 kDa (32,34).
Cell cycle progression is controlled by a series of cyclin-dependent kinases (Cdk), phosphatases, and Cdk inhibitors. Each phase of the cell cycle and the transition between any two phases is controlled by certain cyclin-Cdk complexes whose activation requires several steps (reviewed in Ref. 35). First, association of a Cdk with a specific cyclin confers a certain level of activation as well as specificity. Secondly, Cdk activity is modulated by the state of phosphorylation of specific residues. Kinase activity is further increased by the action of the Cdkactivating kinase (CAK), comprised of Cdk7 and cyclin H, which phosphorylates a threonine residue in the T loop. Finally, for some Cdks, full activity requires dephosphorylation of Tyr-14 and Thr-15 by one of the Cdc25 phosphatases. Checkpoint control, a process by which cell cycle progression is blocked in response to defects in DNA, is implemented by the phosphorylation of Cdc25 phosphatases by the Chk1 and Myt1 kinases (36,37). Finally, association with various Cdk inhibitors can either down-modulate or stimulate Cdk function depending on the Cdk inhibitor and the relative amount of it (38,39). For example, p21 WAF1/CIP1 can stimulate cyclin D-Cdk4 activity at low stoichiometric amounts but repress it at higher concentrations (40).
Although the mechanisms of Cdk regulation have been deciphered to some extent, less is known about the targets of cyclin-Cdks and their functions. Only a few targets have been identified, and for most cyclin-Cdks we still do not know what targets are necessary and sufficient for progression into the next phase of the cell cycle. For example, G 1 Cdks, the cyclin D-Cdk4 and -6 and cyclin E-Cdk2, have been shown to phosphorylate Rb, triggering its release from the E2F transcription factor (41,42). Yet, cyclin E was also found to induce the G 1 /S transition in an Rb-independent manner, suggesting that at least one other target is essential for progression into S phase (43). Very little is known about the transition between the S and G 2 phases. One Cdk complex that is not as well known is the cyclin A-Cdk1 complex. Association with cyclin A was shown to require prior phosphorylation by CAK (44). Monitoring the activity of this complex involves immunoprecipitation with either anti-cyclin A or anti-Cdk1 antibodies. In the former case, one measures the combined activities of cyclin A-Cdk2 and Cdk1 complexes; in the latter case, one measures cyclin A-Cdk1 and cyclin B-Cdk1. Despite these caveats, the combined results suggest that cyclin A-Cdk1 is first detected in S phase and increases moderately until prophase (45,46). So far, only a few substrates of cyclin A-Cdk1 have been characterized. Phosphorylation of the Cdc25B phosphatase by cyclin A-Cdk1 was shown to target Cdc25B for proteosome-dependent degradation (47). In addition, phosphorylation of p53 by cyclin A-Cdk1 and Cdk2 on serine 315 was shown not only to stimulate DNA binding activity but also to modify its specificity (48).
Previous studies aimed at identifying substrate recognition sequences for cyclin-Cdk complexes have revealed a number of interaction motifs. Cyclin binding motifs, called Cy, were iden-tified within p21, p27, pRB, p107, p130, E2F-1, Cdc25A, CDC6, Myt1, and p53 (49 -59). The Cy motifs contain the sequence ZZXL, where Z and X are usually basic. A ZRXL sequence is present in p21, p107, p130, and Cdc25A, whereas both pRb and p53 contain a KKL sequence. Peptides containing a Cy motif were shown to inhibit interaction between the substrate and the cyclinD-Cdk4, cyclinE-Cdk2, or cyclin A-Cdk2 complexes. Further studies revealed that the Cy sequence interacts with a hydrophobic patch containing the MRAIL sequence on the surface of cyclin A and reduces the K m (peptide) ϳ75-120-fold (60 -62). Apart from this docking site, the phospho-acceptor serine was also shown to contribute to the interaction (62). Altogether, results from these studies suggest that the Cy motif and the phospho-acceptor site constitute a bipartite substrate recognition sequence for certain cyclin-dependent kinases. However, targeting of other substrates or other phosphorylation sites within the same substrate appear to involve different interaction motifs. For example, whereas most Cdk phosphorylation sites within pRb displayed a strict dependence on the presence of the Cy motif for their phosphorylation, at least one phospho-acceptor site could be efficiently phosphorylated in the absence of the Cy motif, implying that other mechanisms can serve to target a cyclin-Cdk to a specific substrate (50). In agreement with this, a Cdk2 binding domain was mapped between residues 45 and 60 of p21 WAF1/CIP1 , a region that is fully conserved in the p27Kip1 inhibitor (63). Moreover, a distinct Cdk1 binding domain was identified within the carboxylterminal domain of p53 (48,64).
We have previously shown that CDP/Cux DNA binding activity fluctuates in a cell cycle-dependent manner. Phosphorylation of serine residues in a region of CDP/Cux encompassing the Cut homeodomain correlated with reduced DNA binding in G 1 , whereas increased DNA binding in S phase coincided with dephosphorylation of the same region (32). In the present study, we have investigated CDP/Cux activity later in the cell cycle. We found that DNA binding by CDP/Cux decreases as cells progress from S to G 2 . We showed that activated cyclin A-Cdk1 can interact with CDP/Cux in vitro, phosphorylate it on two serines around the Cut homeodomain, and inhibit its DNA binding activity. Overexpression of cyclin A-Cdk1 inhibited of DNA binding by the wild type protein but not by a mutant CDP/Cux protein in which serines 1237 and 1270 were replaced with alanine. These results suggest that CDP/Cux is a physiological target of cyclin A-Cdk1 and that down-modulation of CDP/Cux activity is important for cell cycle progression in late S and in G 2 .

MATERIALS AND METHODS
Cell Culture and Transfection-293 cells were grown in DMEM medium supplemented with 10% horse serum, antibiotics and glutamine. NIH 3T3 cells were grown in DMEM medium supplemented with 10% fetal bovine serum (FBS), antibiotics, and glutamine. HS578T cells were grown in DMEM medium supplemented with 5% FBS, antibiotics, and glutamine. Transfections in 293 cells were performed by the calcium phosphate precipitation method (65). Transfections in HS578T were performed with ExGen500 (MBI Fermentas) according to the manufacturer's instructions.
Synchronization-To obtain cells in G 0 , cells were maintained for 3 days in 0.4% FBS/DMEM (serum-starved). Synchronization in S phase was performed using the double thymidine block procedure (67). Cells were cultured overnight in DMEM plus 10% FBS supplemented with 2 mM thymidine, washed the next day, cultured for 10 h in DMEM plus 10% FBS, and finally further incubated overnight in the presence of 2 mM thymidine (Thy ϩ 0 h). To allow cells to progress in the cell cycle, the medium was replaced with DMEM plus 10% FBS, and cells were harvested 2 and 4 h later (Thy ϩ 2 h, Thy ϩ 4 h). FACS analysis was performed as previously described (32).
Expression and Purification of CDP/Cux Fusion Proteins-pET-15bderived vectors encoding histidine-tagged CDP/Cux proteins were introduced into the BL21(DE3) strain of Escherichia coli and induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 1.5 h. Proteins were purified on nickel nitrilotriacetic acid-agarose (Qiagen) according to the manufacturer's instructions. Vectors encoding GST-CDP/Cux fusion proteins were introduced in the DH5 strain of E. coli, and protein expression was induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 1.5 h. Proteins were purified on glutathione-Sepharose (Amersham Pharmacia Biotech) according to the manufacturer's instructions.
Production of Cyclin-Cdk Complexes-Baculovirus vectors expressing His-cyclin A, Cdk7-HA, His-cyclin H, HA-Cdk1, and HA-Cdk2 were kindly provided by Dr. S. Larochelle and developed by Dr. D. O. Morgan and Dr. R. P. Fisher (44). Protocols were adapted from previously published protocols (44,71). Sf9 cells at a density of 2 ϫ 10 6 cells/ml were co-infected at 5 multiplicity of infection with baculovirus vectors encoding human cyclin A and human Cdk1 or human cyclin H and human Cdk7 (CAK). After 48 h, cells were harvested and lysed in buffer M. Cyclin A-Cdk complexes were activated by incubating with CAK and 50 M ATP at 24°C for 30 min in buffer M. For Figs. 3, B and C, Sf9 cells were lysed in 250 mM NaCl, 50 mM Hepes, pH 7.0, 0.1% Nonidet P-40.
GST Pull Down Assay-Lysates from baculovirus Sf9 cells were mixed together in the presence of ATP to activate the cyclin A-Cdk1 complex. This lysate was then incubated with glutathione beads bound to 1 g of GST-CDP/Cux fusion proteins at 4°C for 1 h in binding buffer (20 mM Tris, pH 7.5, 0.5% Nonidet P-40, 137 mM NaCl), washed 3 times in binding buffer, and loaded onto a 10% SDS-PAGE. Proteins were analyzed by Western blotting with a monoclonal antibody for cyclin A (E23, Neo Markers) followed by a monoclonal antibody for HA (clone 11, Covance).
Peptide Binding Assays-Peptides (purchased from Research Genetics) were incubated with 35 S-labeled cyclin A-Cdk2 from baculovirusinfected Sf9 cells. After washing, the membrane was exposed to film. The sequence of each peptide is indicated.
Cyclin-Cdk Kinase Assays-The method used was adapted from that of Desai et al. (71). Briefly, activated cyclin A-Cdk1 kinase complexes were isolated by incubating activated lysate with cyclin A monoclonal antibody (E72, Neo Markers) bound to protein G-agarose beads (Life Technologies, Inc.) for 1/2 h at 4°C followed by washing with both buffer M and kinase buffer. Control immunoprecipitation reactions did not include cyclin A antibody. Co-immunoprecipitated cyclin-Cdk complexes were incubated with 500 ng of various CDP/Cux fusion proteins, 2 g of GST (as nonspecific competitor), 5 Ci of ␥-[ 32 P]ATP (6000 Ci/mmol) (Amersham Pharmacia Biotech) in kinase buffer (50 mM Hepes, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol, 1 mM NaF). Positive control reactions included 500 ng of histone H1 as substrate. Reactions were allowed to proceed for 20 min at 24°C and were terminated by adding 6 l of SDS loading buffer and boiling for 5 min. Proteins were resolved on 10% or 12% SDS-PAGE, dried, and exposed to x-ray film to visualize phosphorylated species.
Phosphoamino Acid Analysis-CR3HDϩCy was phosphorylated using an activated cyclin A-Cdk1 complex as described above. Phosphorylated CR3HDϩCy proteins were migrated on a 10% SDS-PAGE and electrophoretically transferred to a polyvinylidene difluoride membrane. The membrane was washed in a large volume of water, dried, re-wetted with methanol and water, and dried again. The membrane was exposed to x-ray film to visualize the phosphorylated species, which were cut out. Phosphoamino acid analysis was performed as previously described (32).
Immunoprecipitation-Polyclonal antibodies for CDP/Cux (anti-N (34 and anti-C (72)) were incubated with protein A-Sepharose beads (Life Technologies, Inc.) for 2 h. Beads were washed with lysis buffer 3-5 times. Lysate was incubated with antibody bead preparation for 2 h followed by washing in lysis buffer 5 times. Immunoprecipitated products were analyzed by resolving products on a 10% SDS-PAGE. Kinase Assay followed by EMSA-Activated cyclin A-Cdk1 kinase complex or CAK was isolated by incubating the lysates with nickel nitrilotriacetic acid-agarose (Qiagen) for 1/2 h at 4°C followed by washing with both buffer M and kinase buffer. Kinase complexes were eluted from the nickel nitrilotriacetic acid-agarose by incubation for 1/2 h in 250 mM imidazole, 1ϫ kinase buffer. Eluted cyclin-Cdk complexes were diluted 1/4 and incubated with 500 ng of His-CR3HD (1125-1308) fusion protein in the presence of 400 M ATP for 30 min at 24°C. EMSAs using phosphorylated proteins were performed as above using 2, 5, and 10 ng of phosphorylated proteins and separated on a 5% polyacrylamide (29:1) 0.5ϫ TBE gel.
Calf Intestinal Phosphatase Treatment-1 l of 10 units/l of calf intestinal phosphatase enzyme (New England Biolabs) was added to each sample plus 1 l of 10ϫ calf intestinal phosphatase buffer (10 mM ZnCl 2 , 10 mM MgCl 2 , 100 mM Tris HCl, pH 8.3) for a total of 10 l. The reaction was incubated on ice for 30 min. EMSA was performed after calf intestinal phosphatase treatment.
Luciferase Assays-Luciferase assays were performed as previously described with minor modifications (32) in HS578T cells. As a control for transfection efficiency, the ␤-galactosidase protein (Sigma) was included in the transfection mix, and the luciferase activity was normalized based on ␤-galactosidase activity.

CDP/Cux DNA Binding Activity Is Post-translationally
Down-modulated in G 2 -We have previously shown that CDP/ Cux DNA binding activity increases as cells progress from the G 1 to the S phase of the cell cycle. In the present study, we investigated CDP/Cux activity in the later part of the cell cycle. NIH3T3 cells were synchronized in S phase using the thymidine block procedure (see "Material and Methods"). Cell cycle distribution was monitored by fluorescence-activated cell-sorting analysis (Fig. 1C), and CDP/Cux DNA binding was measured by EMSA using oligonucleotides containing a consensus CDP/Cux DNA binding site (Fig. 1A). Specific antibodies were added to the binding reaction to ascertain the presence of CDP/Cux proteins in two specific retarded protein-DNA com- plexes (Fig. 1A, lanes 6 and 7). An antibody raised against the amino-terminal portion of the protein, ␣-N-term, supershifted the slow mobility complex only (lane 6), whereas an antibody raised against an internal epitope, ␣-861, was able to supershift both the slow and fast migrating complex (lane 7). We have previously shown that the faster migrating species involved an amino-terminal-truncated protein of 110 kDa that is generated by proteolytic cleavage as cells progress into S phase (32). The results show that both complexes were more abundant in the population of cells synchronized with thymidine than in unsynchronized cells or cells made quiescent by serum starvation (Fig. 1B, compare lane 3 with 1 and 2). Upon removal of thymidine from the medium, the intensity of the lower complex decreased as cells progressed into the G 2 phase of the cell cycle, yet the amount of p110 was slightly increased during the same period (Fig. 1B). We conclude that CDP/Cux DNA binding activity is down-modulated at a post-translational level as cells progress into G 2 .
CDP/Cux Co-immunoprecipitates with a CAK-activated Cyclin A-Cdk1 Complex-Inhibition of CDP/Cux DNA binding in G 1 was previously found to result at least in part from the phosphorylation of a region that encompasses the Cut homeodomain. We therefore envisaged that the down-modulation of CDP/Cux activity in G 2 might similarly be caused by the phosphorylation of one of its DNA binding domains. As a first step in verifying this possibility, we investigated whether CDP/Cux could interact with cyclin A-Cdk1, a kinase complex that is active during this period of the cell cycle. The connection between CDP/Cux and cyclin A-Cdk1 was also suggested from previous gel shift immunoassays indicating that CDP/Cux, cyclin A, and Cdk1 were part of the multi-protein complex HiNF-D that binds to the promoter of several histone genes (28,73). As a source of proteins for co-immunoprecipitation, Sf9 insect cells were infected with baculovirus vectors expressing cyclin A and HA-tagged Cdk1, cyclin H and HA-tagged Cdk7 (CAK), or full-length CDP/Cux. Previous studies have shown that the association of Cdk1 with cyclin A and the activation of its kinase function require that Cdk1 be phosphorylated by the CAK, cyclin H, and Cdk7 (44). Therefore, lysates were mixed together in the presence of cold ATP and immunoprecipitated or not using CDP/Cux polyclonal antibodies anti-N and anti-C. Western blot analysis with a monoclonal antibody for HA served to monitor the presence of Cdk1-HA and Cdk7-HA in total lysates and immunoprecipitates. Cdk1-HA, but not Cdk7-HA, was co-immunoprecipitated with anti-CDP/Cux antibodies (Fig. 2, lane 5). Interaction between Cdk1 and CDP/Cux required the presence of cyclin A (lane 2) and CAK (lane 4). That CAK was necessary but that Cdk7 was not co-immunoprecipitated suggests that the Cdk1 must be activated by CAK to be able to interact with CDP/Cux. Because cyclin A was also required for the formation of the cyclin A-Cdk1 with CDP/Cux, these results suggest that CDP/Cux binds only to an activated cyclin A-Cdk1 complex.

Efficient Interaction with Cyclin A-Cdk Complex Requires Both the Cut Homeodomain and a Cy Sequence in the CTD-To
determine the region of CDP/Cux that associates with cyclin A-Cdk1, pull-down assays were performed using total extracts from Sf9 cells containing CAK-activated cyclin A-Cdk1-HA complex and a panel of GST-CDP/Cux fusion proteins (Fig. 3A). Activation of the cyclin A-Cdk1-HA complexes was performed as described under "Material and Methods." After affinity chromatography on glutathione beads, proteins were separated by SDS-PAGE and analyzed by Western blot with monoclonal antibodies against HA and cyclin A. Cyclin A and Cdk1-HA were efficiently pulled down by three fusion proteins: CR3HDϩCy, HDϩCy, and 1007-1505 (Fig. 3A, lanes 5, 7, and   9). The common region shared by these proteins includes the Cut homeodomain plus a short extension of seven amino acids from the carboxyl-terminal domain. Interestingly, this region contains a sequence, RREL, that resembles the consensus cyclin binding motif (Cy), RXL. Both the Cut homeodomain and the region encompassing the Cy motif appear to be needed for efficient binding to cyclin A and Cdk1 since weak or no binding was observed with a protein, CR1HD, that contains the Cut homeodomain without the Cy motif or with a protein that contains the CTD without the Cut homeodomain (Fig. 3A, lanes  3 and 8). Cyclin A and Cdk2 were similarly able to interact with HDϩCy, but not with HD alone, confirming that the Cy sequence is important for binding (Fig. 3B, lanes 6 and 7). Interestingly, cyclin A was not able to interact on its own (Fig. 3B,  lanes 2 and 3). Both Cdk2 and cyclin A were required for the interaction with CDP/Cux, binding at similar levels, suggesting this interaction involves cooperative interactions by both the cyclin and Cdk. Alternatively, it is possible that cyclin A adopts a different conformation when part of a complex with a Cdk.
Examination of the amino acid sequence of CDP/Cux revealed the presence of five different Cy-related sequences. To investigate whether any or all of these sequences could mediate the interaction, a peptide binding assay was performed using 35 S-labeled cyclin A-Cdk2 from baculovirus-infected Sf9 cells. As controls, we tested two peptides containing previously characterized Cy motifs from p107 and E2F3 and a control peptide containing an unrelated sequence. As expected, the p107-and E2F3-derived peptides, but not the control peptide, were able to bind to cyclin A-Cdk2 (Fig. 3C, lanes 1, 7, and 8). Among the 5 CDP/Cux-derived peptides, only the one starting at amino acid 1298, downstream of the Cut homeodomain, was able to bind to cyclin A-Cdk2 (Fig. 3C, lanes 2-6). Altogether, results from pull-down and peptide binding assays indicate that the Cyrelated sequence downstream of the Cut homeodomain is a bona fide interaction motif for cyclin-Cdk complexes.

FIG. 2. Co-immunoprecipitation of CDP/Cux with a CAK-activated cyclin A-Cdk 1 complex.
Sf9 insect cells were infected in separate plates with baculovirus vectors expressing cyclin A and HAtagged Cdk1, cyclin H, and HA-tagged Cdk7 (CAK) or full-length CDP/ Cux. Lysates were mixed together in the presence of cold ATP and immunoprecipitated or not using CDP/Cux polyclonal antibodies N and C as indicated. Total lysates and immunoprecipitates (IP) were separated by SDS-PAGE and analyzed by Western blot with a monoclonal antibody for HA. Input represents 1% of total lysate.

Cyclin A-Cdk1 Phosphorylates Serines 1237 and 1270 of CDP/
Cux-Sequence analysis of CDP/Cux reveals the presence of 23 potential Cdk phosphorylation sites, SerPro, or ThrPro (Fig. 4D). Only two sites, Ser-1237 and Ser-1270, are situated within or close to a DNA binding domain, in this case the Cut homeodomain. These two sites are also in close proximity to the putative Cy site at position 1301-1303, and it is also noteworthy that phosphorylation within a region encompassing the Cut homeodomain was previously shown to correlate with inhibition of DNA binding (32). We therefore investigated whether cyclin A-Cdk1 could phosphorylate a histidine-tagged fusion protein containing Cut repeat 3 and the Cut homeodomain, CR3HD. Activated cyclin A-Cdk1 was immunoprecipitated from insect cells using an antibody against cyclin A and was incubated with CR3HD and an excess GST, included as a competitive substrate. Although GST was not phosphorylated at all, CR3HD was phosphorylated to a level comparable with that of histone H1, a well characterized substrate of Cdks (Fig. 4A, lanes 1, 5, 9, and 11). The level of phosphorylation of cyclin A was weak because activation of the cyclin A-Cdk1 complex by CAK required prior incubation with cold ATP. Phosphoamino acid analysis revealed that phosphorylation of CR3HD occurred on serine residues (Fig. 4C). To confirm the identity of the phosphorylation sites, we prepared mutant His-CR3HD fusion proteins in which serines 1237 and 1270 were replaced with alanines: CR3HD S1237A , CR3HD S1270A , and CR3HD S1237/1270A . Mutation of serine 1270 only slightly reduced

FIG. 3. Interaction with cyclin A-Cdk1 requires a Cy motif downstream of the Cut homeodomain.
A, GST pull-down assay using GST-CDP/Cux fusion proteins and total extracts from Sf9 cells containing CAK-activated cyclin A-Cdk1 complex. Activation of the cyclin A-Cdk1 complex was performed as described under "Material and Methods." After affinity chromatography on glutathione beads, proteins were separated by SDS-PAGE and analyzed by Western blot with monoclonal antibodies against HA and cyclin A. Input represents 10% of total lysate. A schematic representation of GST-CDP/Cux fusion proteins is shown on the left. B, pull-down assay using anti-cyclin A antibodies as a control (lanes 1 and  5), GST-CR3HD with or without the Cy sequence immediately downstream, and total extracts from 35 S-labeled Sf9 extracts containing cyclin A with or without Cdk2. After affinity chromatography, proteins were separated by SDS-PAGE and revealed by autoradiography. A diagram of the CR3HD ϩ/Ϫ Cy is shown on the right. C, peptide binding assays. Peptides were synthesized on a propylene membrane to which they were covalently bound via a carboxyl-terminal linker. The peptides were incubated with 35 S-labeled cyclin A-Cdk2 from baculovirus-infected Sf9 cells. After washing, the membrane was exposed to film. The sequence of each peptide is indicated. the level of phosphorylation, whereas mutation of serine 1237 had a greater effect (Fig. 4A, compare lane 1 with lanes 2 and 4). Mutation of the two serines at the same time further reduced the level of phosphorylation (Fig. 4A, compare lanes 2-4). We conclude that serines 1237 and 1270 represent major and minor sites, respectively, of phosphorylation by cyclin A-Cdk1.
The Cy Motif in the CTD Increases the Efficiency of Phosphorylation by Cyclin A-Cdk1-To assess the contribution of the putative Cy motif, the phosphorylation assay was performed using His-tagged CR3HD fusion proteins that differed by the presence or absence of a seven-amino acid carboxyl-terminal extension as substrates (schematic, Fig. 4D). The amount of FIG. 4. Cyclin A-Cdk1 phosphorylates CR3HD at positions S1237 and S1270. Cyclin A-Cdk1 complex from Sf9 cells was activated by CAK as described under "Materials and Methods" and immunoprecipitated using an anti-cyclin A antibody. 500 ng of the indicated proteins were incubated with activated cyclin A-Cdk1 complexes or beads only (lanes 10, 12, 17, and 19) for 20 min in the presence of 5 Ci of [␥-32 P]ATP and 2 g of GST as a nonspecific competitor. To control for phosphorylation efficiency, 500 ng of histone H1 was used as a substrate in parallel experiments (lanes 9 -12). Proteins in each sample were separated by electrophoresis on a 12% SDS-polyacrylamide gel, stained with Coomassie Blue, and revealed by autoradiography. A, wild type or mutated his CR3HD. Mutated CR3HD are proteins in which a serine residue was replaced for alanine at position 1237, 1270, or both as indicated. B, CR3HD with or without the Cy motif. C, phosphoamino acid analysis. p-, phosphorylated. D, schematic representation of His-CDP/Cux fusion proteins used in A and B. phosphorylation was measured both in a scintillation counter and by densitometric analysis of the autoradiogram and was corrected for the amount of substrate protein in the sample, as judged from Coomassie staining (Fig. 4B, compare lanes 15 and  16). The presence of the seven-amino acid extension was found to increase the level of phosphorylation by ϳ3-fold (Fig. 4B,  lanes 13 and 14). These results together with the pull-down assays described above suggest that the RREL sequence acts as a bona fide Cy motif.
Phosphorylation of CR3HDϩCy by Cyclin A-Cdk1 Inhibits Its DNA Binding Activity-To assess the effect of phosphorylation by cyclin A-Cdk1 on the DNA binding activity of CDP/ Cux, an EMSA was performed using CR3HDϩCy fusion proteins. Increasing amounts of CR3HDϩCy were incubated first with an activated cyclin A-Cdk1 complex in the presence of either cold ATP or H 2 O and then with oligonucleotides containing an ATCGAT binding site. The intensity of the retarded complex was reduced in the samples that had been incubated with cyclin A-Cdk1 in the presence of ATP (Fig. 5A, compare  lanes 1-3 with lanes 4 -6). In contrast, a similar treatment with cyclin H-Cdk7 did not affect DNA binding by CR3HD. Because cyclin H-Cdk7 does not interact with CDP/Cux (Fig. 2) and does not phosphorylate CR3HD (data not shown), we conclude that the specific phosphorylation of CR3HD by cyclin A-Cdk1 inhibits its DNA binding activity.
Co-expression of Cyclin A-Cdk1 with CDP/Cux Inhibits Its DNA Binding Activity in Vivo-To investigate whether cyclin A-Cdk1 may modulate CDP/Cux activity in vivo, 293 cells were co-transfected with vectors expressing wild type or mutated CDP/Cux proteins, cyclin A, and Cdk1. Whole cell extracts were prepared and analyzed in EMSA using oligonucleotides containing an ATCGAT binding site and in Western blots using anti-cyclin A and HA antibodies. Co-expression of cyclin A-Cdk1 with a wild type CDP/Cux protein caused a reduction in DNA binding activity (Fig. 6, A, lanes 5 and 8, B, lanes 2 and 3). Treatment of cell extracts with calf intestinal phosphatase restored DNA binding, suggesting that the inhibition of DNA binding in the sample co-transfected with cyclin A-Cdk1 was due to phosphorylation of the recombinant CDP/Cux protein (Fig. 6A, lanes 5 and 6). If this assumption was correct, then mutated CDP/Cux proteins in which serines 1237 and 1270 are replaced with alanine should not be affected by cyclin A-Cdk1. This is indeed what we observed (Fig. 6, A, lanes 1 and 4, B,  lanes 4 and 5). These results in conjunction with the in vitro studies suggest that cyclin A-Cdk1 down-modulates CDP/Cux DNA binding activity by phosphorylating serine 1237 and 1270.

Removal of the Cy Motif Reduces but Does Not Prevent the Effect of Cyclin A-Cdk1 on CDP/Cux-Pull-down and in vitro
kinase assays indicated that the Cy motif present in the CTD plays a role in the interaction between CDP/Cux and cyclin A-Cdk1. To assess the importance of the Cy motif in vivo, we compared the effect of cyclin A-Cdk1 on recombinant CDP/Cux proteins that differed by the presence or absence of the Cy motif. To this end, we engineered an eight-amino acid deletion encompassing the Cy motif. Because the Cy motif overlaps the predicted carboxyl-terminal end of the Cut homeodomain, this deletion may reduce its DNA binding affinity (Fig. 7A, compare  lanes 3 and 4). Co-expression of CDP/Cux with cyclin A-Cdk1 caused a reduction in DNA binding, albeit to a slightly lesser extent in the case of the ⌬Cy CDP/Cux protein (Fig. 7, compare lanes 2 and 3 with 4 and 5). This result may mean that the Cy motif does not play an essential role in the interaction in vivo. Alternatively, it is possible that overexpression of cyclin A-Cdk1 and CDP/Cux reduces the dependence on multiple cooperative interactions.
Co-expression of cyclin a-Cdk1 with CDP/Cux Prevents Repression of the p21 WAF1/CKI Reporter-CDP/Cux has previously been shown to bind to the core promoter of the p21 WAF1/CKI gene and to repress expression of reporter constructs driven from this promoter. Because phosphorylation by cyclin A-Cdk1 inhibits CDP/Cux DNA binding, we would predict that its transcriptional activity should be affected as well. To verify this hypothesis, a p21 WAF1/CKI reporter construct was cotransfected with CDP/Cux in the presence or absence of cyclin A-Cdk1. Expression from the p21 WAF1/CKI reporter was repressed by CDP/Cux as previously described. However, in the presence of cyclin A-Cdk1, expression of the p21 WAF1/CKI reporter was almost returned to its original level. In contrast, the mutated CDP/Cux (S1237A/S1270A) still repressed the p21 WAF1/CKI reporter in the presence of cyclin A-Cdk1 (Fig. 8). We conclude that co-expression of cyclin A-Cdk1 with CDP/Cux inhibits its activity as a transcriptional repressor. DISCUSSION We have shown that cyclin A-Cdk1 can associate with CDP/ Cux in vitro, phosphorylate it on serines 1237 and 1270 and inhibit its DNA binding activity (Fig. 1, 2 , 3, 4, 5). In cells, overexpression of cyclin A-Cdk1 inhibited CDP/Cux DNA binding and repression activities, whereas a mutant CDP/Cux protein in which serines 1237 and 1270 were replaced with alanine was not affected by cyclin A-Cdk1. These results are in accordance with the findings that CDP/Cux DNA binding activity decreases in G 2 , the phase of the cell cycle when the cyclin A-Cdk1 complex becomes prominent. Altogether these results strongly suggest that CDP/Cux is a target of cyclin A-Cdk1 and that progression through S and G 2 in the cell cycle is facilitated by the down-modulation of CDP/Cux activity.
We have previously shown that CDP/Cux DNA binding activity is up-regulated at the G 1 /S transition as a result of dephosphorylation by the Cdc25A phosphatase (32). More recently, we discovered that the up-regulation of stable CDP/Cux DNA binding activity involves specific proteolytic cleavage of the full-length protein to generate an amino-truncated isoform of 110 kDa (34). Importantly, the proteolytic processing of CDP/Cux was found to take place preferentially in populations of cells that were synchronized in S phase. In the present study, we demonstrated that CDP/Cux DNA binding activity is down- modulated as cells progress in G 2 . Altogether, these results point to a specific role of CDP/Cux in S phase or at the G 1 /S transition. Because CDP/Cux functions as a transcription factor, we can speculate that it regulates the expression of genes whose products play a role in the control of DNA replication. A systematic analysis of CDP/Cux target genes will be necessary to understand its role in cell cycle progression. However, a few potential targets have already been identified and will be discussed here.
CDP/Cux was shown to bind to the core promoter of the p21 WAF1/CIP1 gene and repress its expression (32). Interestingly, a reporter construct containing the p21 promoter was shown to be down-regulated in S phase, and co-expression of a CDP/Cux antisense transcript restored expression of the reporter to the level seen in G 1 . The relevance of these observations is hard to assess in light of the fact that p21 protein expression, unlike its transcription, was not found to decrease in S phase. It can be argued that although the steady-state level of p21 may remain stable in G 1 and S phases, there is an important difference between a dynamic system where there is active synthesis and degradation of the protein and a static system where there is no synthesis and no degradation. In the dynamic system, there is an influx of protein that can associate with new partners, whereas in the static system we can envisage that the protein remains sequestered within the same complexes, thereby allowing newly synthesized cyclin-Cdk complexes to assemble without p21. In this line of thinking, one role of CDP/Cux may be to prevent de novo synthesis of p21 in S phase. Other potential targets of CDP/Cux are some of the histone genes. Several studies have described the S phasespecific association of CDP/Cux as part of the HiNF-D complex with the promoter of histone H4 genes (28,29). The coincidence between the up-regulation of histone H4 genes and the presence of the HiNF-D complex on their promoters raised the  , lanes 2, 3, 6, and 7), cell extracts were preincubated in the presence of calf intestinal phosphatase (CIP) before EMSA. C, schematic representation of recombinant CDP/Cux proteins. Short vertical lines on top of each diagram represent potential CDK phosphorylation sites, Ser/Thr Pro. possibility that CDP/Cux can play a role as an activator at least in the context of the HiNF-D complex. Interestingly the association of cyclin A and Cdk1 with HiNF-D was revealed from gel shift immunoassays at a time when it was not yet known that CDP/Cux was the DNA binding partner within this complex (73). In this framework, we speculate that the association with cyclin A-Cdk1 would serve to down-modulate the activity of HiNF-D once DNA and histone synthesis has taken place.
Our binding assays indicated that the interaction with CDP/ Cux required both cyclin A and a Cdk (Fig. 3). On one hand, Cdk1 bound to CDP/Cux only in the presence of cyclin A and after activation by CAK (Fig. 3). The role of CAK could be to allow the formation of the cyclin A-Cdk1 complex, since it was not itself co-immunoprecipitated with CDP/Cux. Indeed, cyclin A and Cdk1 were shown to form a high affinity complex only after CAK-mediated phosphorylation of Cdk1 (44). On the other hand, CDP/Cux bound to cyclin A and Cdk2 but not to cyclin A alone (Fig. 3B). The requirement for both cyclin A and a Cdk could merely reflect the fact that these proteins change conformation upon binding to each other. However, the fact that two regions of CDP/Cux were also needed for the interaction supports the notion that cooperative binding was at play. In pull-down assays the CTD was not sufficient for efficient interaction (Fig. 3C). It could only weakly bring down cyclin A. In contrast, a recombinant protein containing both the Cut homeodomain and the Cy sequence brought down both cyclin A and Cdk1 or Cdk2 efficiently. Removal of the Cy sequence prevented the interaction. Moreover, in the in vitro kinase assay, removal of the Cy sequence reduced the efficiency of phosphorylation by approximately 3-fold (Fig. 4B). Yet cyclin A-Cdk1 was still able to phosphorylate a CR3HD fusion protein in which the Cy sequence had been removed. In accordance with this result, in cotransfection assays the removal of the Cy sequence from the recombinant CDP/Cux protein diminished but did not abolish the effect of cyclin A-Cdk1 on the activity of CDP/Cux. Regarding the latter two experiments, it should be stressed that the proteins involved were present at high concentrations. It is likely that at physiological levels the modulation of CDP/Cux by cyclin A-Cdk1 requires cooperative interactions between cyclin A and the Cy motif and between Cdk1 and the Cut homeodomain.
Phosphorylation has been shown to have a profound impact on protein functions. Only two other proteins have been shown to be the target of modulation by cyclin A-Cdk1: Cdc25B and p53. Phosphorylation of Cdc25B induces its degradation by the proteosome, whereas phosphorylation of p53 stimulates its DNA binding activity and modulates its specificity. In the case of CDP/Cux, phosphorylation by cyclin A-Cdk1 inhibits its DNA binding activity. As for the substrates of other kinases, the effect of phosphorylation by cyclin A-Cdk1 depends essen- tially on the position of the phospho-acceptor site within the substrate. We would predict that a myriad of molecular consequences will be discovered as more targets of cyclin A-Cdk1 are identified. At the cellular level, the results of the present study and those of previous reports describing the activation of CDP/ Cux at the G 1 /S transition suggest that the phosphorylation of CDP/Cux by cyclin A-Cdk1 serves to down-modulate an activity that is required for passage through S phase but should be prevented to allow cells to progress through G 2 and into mitosis.