Interaction of p58PITSLRE, a G2/M-specific Protein Kinase, with Cyclin D3*

The p58PITSLRE is a p34cdc2-related protein kinase that plays an important role in normal cell cycle progression. Elevated expression of p58PITSLRE in eukaryotic cells prevents them from undergoing normal cytokinesis and appears to delay them in late telophase. To investigate the molecular mechanism of p58PITSLRE action, we used the yeast two-hybrid system, screened a human fetal liver cDNA library, and identified cyclin D3 as an interacting partner of p58PITSLRE. In vitro binding assay, in vivo coimmunoprecipitation, and immunofluorescence cell staining further confirmed the association of p58PITSLRE with cyclin D3. This binding was observed only in the G2/M phase but not in the G1/S phase of the cell cycle; meanwhile, no interaction between p110PITSLRE and cyclin D3 was observed in all the cell cycle. The overexpression of cyclin D3 in 7721 cells leads to an exclusively accumulation of p58PITSLRE in the nuclear region, affecting its cellular distribution. Histone H1 kinase activity of p58PITSLRE was greatly enhanced upon interaction with cyclin D3. Furthermore, kinase activity of p58PITSLRE was found to increase greatly in the presence of cyclin D3 using a specific substrate, β-1,4-galactosyltransferase 1. These data provide a new clue to our understanding of the cellular function of p58PITSLRE and cyclin D3.

The p58 PITSLRE is a p34 cdc2 -related protein kinase that plays an important role in normal cell cycle progression. Elevated expression of p58 PITSLRE in eukaryotic cells prevents them from undergoing normal cytokinesis and appears to delay them in late telophase. To investigate the molecular mechanism of p58 PITSLRE action, we used the yeast two-hybrid system, screened a human fetal liver cDNA library, and identified cyclin D3 as an interacting partner of p58 PITSLRE . In vitro binding assay, in vivo coimmunoprecipitation, and immunofluorescence cell staining further confirmed the association of p58 PITSLRE with cyclin D3. This binding was observed only in the G 2 /M phase but not in the G 1 /S phase of the cell cycle; meanwhile, no interaction between p110 PITSLRE and cyclin D3 was observed in all the cell cycle. The overexpression of cyclin D3 in 7721 cells leads to an exclusively accumulation of p58 PITSLRE in the nuclear region, affecting its cellular distribution. Histone H1 kinase activity of p58 PITSLRE was greatly enhanced upon interaction with cyclin D3. Furthermore, kinase activity of p58 PITSLRE was found to increase greatly in the presence of cyclin D3 using a specific substrate, ␤-1,4-galactosyltransferase 1. These data provide a new clue to our understanding of the cellular function of p58 PITSLRE and cyclin D3.
The eukaryotic cell division cycle is tightly regulated by the activation and deactivation of the cyclin-dependent kinases (CDKs). 1 Active CDK serves as a protein kinase subunit, the kinase activity of which is dependent on its association with a regulatory cyclin subunit (1)(2)(3). In mammalian cells both the CDKs and cyclins consist of numerous members, including cyclin A-H and at least nine different p34 cdc2 -related kinases (4,5). Among them, the CDKs 4 and 6 are first activated by binding to the D-type cyclins (cyclin D1, D2, and D3) and are believed to control progression through G 1 phase of the cell cycle, in response to cell cycle progression and mitogenic sig-nals (3, 6 -8). CDK2, subsequently, in combination with cyclin E and cyclin A, controls G 1 /S phase transition and S phase progression (9 -11). The p34 cdc2 (CDK1) in association with cyclin A is essential for the completion of S phase and entry into G 2 phase, whereas the transition through G 2 /M phase is regulated by p34 cdc2 -cyclin B complex (12). Therefore, the association of different CDK subunits with different cyclin subunits regulates progression through different stages of the cell cycle (1-3, 13, 14). Although cyclin binding is required for the activation of the CDK subunit of the complex, other means of modulating the activity of CDKs also exist, such as phosphorylation and dephosphorylation of the key residues on the CDK subunit and the binding of cyclin-dependent kinase inhibitors (2,3,14,15).
The PITSLRE protein kinases are parts of the large family of p34 cdc2 -related kinases whose functions appear to be linked with cell cycle progression, apoptotic signaling, and tumorigenesis (16 -25). The PITSLRE homologues exist in human, mouse, chicken, Drosophila, and Xenopus, suggesting that their functions may be well conserved (16,19,26,27). The small p58 PITSLRE isoform was originally isolated from a human liver cDNA library and has a 299-amino acid region with 68% homology to the p34 cdc2 protein kinase (16). During the study of p58 PITSLRE , 10 isoforms of the p58 PITSLRE subfamily of protein kinases including p110 PITSLRE have been isolated by molecular cloning (19). The discovery of multiple p58 PITSLRE isoforms has led to the renaming of these kinases according to an established nomenclature system, which is based on the single amino acid codon designation of the conserved PSTAIRE box region of p34 cdc2 (17). The p110 PITSLRE isoform can be detected in all phases of the cell cycle, whereas the p58 PITSLRE is mainly expressed in G 2 /M phase (28). Ectopic expression of p58 PITSLRE in Chinese hamster ovary fibroblasts leads to a late telophase delay, abnormal cytokinesis, and a reduced rate of cell growth (16). Conversely, the diminished expression of p58 PITSLRE mRNA is found to increase DNA replication and enhance cell growth (17). Further analysis of the Chinese hamster ovary cells ectopically expressed of p58 PITSLRE demonstrated that the reduced cell growth was due to apoptosis (20). In addition, it was shown that the p58 PITSLRE and p110 PITSLRE isoforms were cleaved by caspase proteases to generate smaller 46 -50-kDa proteins that could also phosphorylate histone H1 during tumor necrosis factor ␣and Fas-mediated apoptosis (21)(22)(23). Because of its ultimate function in cell growth control, the p58 PITSLRE and its family have been a target for alteration, translocation, and deletion during tumorigenesis (18,24,25).
Although the p58 PITSLRE plays an important role in cell cycle progression, little is known about its interaction proteins. Meanwhile, study of the p110 PITSLRE isoform showed that it could interact with the RNA-binding protein RNPS1, RNA polymerase II, and multiple transcriptional elongation factors, regulating some aspects of RNA splicing or transcription in proliferating cells (29,30). Thus, the identification of the cellular proteins that interact with p58 PITSLRE is a useful approach for defining the cellular function and regulatory mechanism of p58 PITSLRE . To investigate this issue, a two-hybrid screening from human fetal liver cDNA library was carried out using the full length of p58 PITSLRE as bait. As a result, cyclin D3 was identified as a p58 PITSLRE -associated protein. This interaction between p58 PITSLRE and cyclin D3 is specific, as demonstrated by the inability of the other D-type cyclins to associate with p58 PITSLRE using in vitro binding assays and yeast two-hybrid assays and the inability of the p110 PITSLRE to associate with cyclin D3 using immunofluorescence cell staining and immunoprecipitation. More importantly, we showed that the p58 PITSLRE was associated with the cyclin D3 in vivo at G 2 /M phase by coimmunoprecipitation and immunofluorescence. Interestingly, the elevated expression of cyclin D3 affected p58 PITSLRE cellular distribution. Moreover, kinase activity of p58 PITSLRE was greatly enhanced upon cyclin D3 association. Taken together, the data suggest that cyclin D3 is important for some aspects of p58 PITSLRE regulation and function in G 2 /M phase.

EXPERIMENTAL PROCEDURES
Cell Lines and Reagents-7721 cells, a human hepatocarcinoma cell line, were obtained from the Institute of Cell Biology, Academic Sinica. The 7721 cells ectopically expressed of p58 PITSLRE (7721/p58 cells) were constructed and confirmed in our previous work (31). The rabbit polyclonal anti-PITSLRE antibody, the goat anti-rabbit-fluorescein isothiocyanate secondary antibody, and the goat anti-mouse-rhodamine secondary antibody were purchased from Santa Cruz Biotechnology, and the mouse monoclonal anti-cyclin D3 antibody was purchased from Signal Transduction Laboratories. Protein G-agarose, glutathione-Sepharose beads, the mouse monoclonal anti-HA (12CA5) antibody, and histone H1 were purchased from Roche Molecular Biochemicals. Bovine ␤-1,4-galactosyltransferase 1, leupeptin, aprotinin, and phenylmethylsulfonyl fluoride were purchased from Sigma. [␥-32 P]ATP (Ͼ3000 Ci/ mM), [ 35 S]methionine, Hybond polyvinylidene difluoride membrane, goat anti-mouse-horseradish peroxidase secondary antibody, goat antirabbit-horseradish peroxidase secondary antibody, and the enhanced chemiluminescence (ECL) assay kit were purchased from Amersham Biosciences.
Yeast Two-hybrid Assays-A genetic screen using the yeast interaction trap was performed as recommended by the manufacturers (according to CLONTECH Matchmaker LexA two-hybrid system user manual). The full-length of p58 PITSLRE was cloned in-frame into LexAcoding sequence to generate bait plasmid, pLexA-p58 PITSLRE . A human fetal liver cDNA library in the pB42AD plasmid (CLONTECH) was screened for proteins that interact with p58 PITSLRE using EGY48 yeast strain (Mat␣ trp1 ura3-52 leu2::pLeu2-lexAop6(⌬G-UAS leu2)). Yeast transformation was performed by the lithium acetate method. Plasmid DNA from LEU2 ϩ /LacZ ϩ colonies was isolated and recovered, and the true positives were sequenced with dideoxy sequencing according to the manufacturer's instructions (Amersham Biosciences). The fish plasmid, pB42AD harboring cyclin D3, was transformed back into yeast along with either the bait plasmid or other nonspecific bait plasmids to verify the specificity of the two-hybrid assay. For direct interaction tests, pLexA constructs with the full-length of p58 PITSLRE and the two mutants, were co-transformed with the D-type cyclin pB42AD constructs. The specific interaction was measured by the production of leucine and ␤-galactosidase.
Plasmid Construction-For the bait of two-hybrid system, the full-length of p58 PITSLRE (31) was cloned into the EcoRI/XhoI site of pLexA (CLONTECH) in-frame with the DNA binding domain of LexA. The glutathione S-transferase (GST) fusion expression vector pcDNA3-GST-p58 PITSLRE and pcDNA3-GST-CDK4 for in vitro translation and the HA epitope-tagged p58 PITSLRE eukaryotic expression vector pcDNA3-HA-p58 PITSLRE were obtained as described previously (31). To generate pEGFP-p58 PITSLRE , the full-length of p58 PITSLRE without the stop codon was cloned into pEGFP N3 in-frame with the EGFP. The deletion mutants of p58 PITSLRE were constructed by PCR with pLexA-p58 PITSLRE as the template using the primers NH 2 -p58 PITSLRE (sense, an EcoRI site for subsequent subcloning is underlined; 5Ј-gcgaattcgag-gaagaaatgagtgaaga-3Ј), NH 2 -p58 PITSLRE (antisense, an XhoI site for subsequent subcloning is underlined; 5Ј-gcctcgagcttttgctctgtagaccactc-3Ј), ⌬NH 2 -p58 PITSLRE (sense, an EcoRI site is underlined; 5Ј-gcgaattctgccggagcgtcgaggagtt-3Ј), and ⌬NH 2 -p58 PITSLRE (antisense, a SalI site is underlined; 5Ј-gggtcgacacaaagtaagacgaggagtt-3Ј). Full-length cyclin D3, obtained from yeast two-hybrid screening, was cloned into pcDNA3 for in vitro translation. By PCR amplification, we cloned cyclin D3 in-frame into pDsRed C1 at the site of EcoRI/BamHI using the primers cyclin D3 (sense, an EcoRI site is underlined; 5Ј-Gcgaattctatggagctgctgtgttgcga-3Ј) and cyclin D3 (antisense, a BamHI site is underlined; 5Ј-gcggatccagagggcctctccagggcta-3Ј). The cyclin D1 and cyclin D2 cDNAs were also generated by PCR with the human liver cDNA library cDNA (Invitrogen) as template using the primers cyclin D1 (sense, an EcoRI site is underlined; 5Ј-gcgaattcatggaacaccagctcctgtg-3Ј), cyclin D1 (antisense, an XhoI site is underlined; 5Ј-gcctcgagtcagatgtccacgtcccgca-3Ј), cyclin D2 (sense, an EcoRI site is underlined; 5Ј-gcgaattcatggagctgctgtgccacga-3Ј), and cyclin D2 (antisense, an XhoI site is underlined; 5Ј-gcctcgaggcccaactggcatcctcaca-3Ј. All the plasmids produced by PCR were confirmed by sequencing. In Vitro Protein Expression and Interaction-GST-p58 PITSLRE , GST-CDK4, GST, cyclin D1, cyclin D2 and cyclin D3 were [ 35 S]methioninelabeled in vitro with the TNT ® coupled reticulocyte lysate system (Promega) according to the user manual. Plasmid DNA purified with Wizard Plus Minipreps DNA purification system (Promega) was added to the TNT ® lysate reaction buffer with 0.4 Ci/l [ 35 S]methionine. After incubation at 30°C for 90 min, the labeled proteins were mixed together with 25 l of glutathione-Sepharose beads in the binding buffer (20 mM HEPES, pH 7.7, 150 mM NaCl, 0.5% Nonidet P-40, 2 mM EDTA, and 10% glycerol) for 4 h at 4°C. Then the beads were washed three times with the binding buffer and boiled in SDS sample buffer. The bound proteins were analyzed by autoradiography after they were resolved by SDS-PAGE.
Cell Culture and Synchronization-All the cells were cultured in RPMI 1640 medium supplemented with 10% (v/v) bovine calf serum, 100 units/ml penicillin, and 50 g/ml streptomycin at 37°C under 5% CO 2 in humidified air. G 1 /S phase-arrested 7721 cells were obtained by sequential thymidine treatment. First, the cells were treated with 2.5 mM thymidine for 24 h then changed to the fresh medium for another 24 h and replaced with the 2.5 mM thymidine medium for 24 h. To block cells in G 2 /M phase, cells were seeded in RPMI 1640 medium with 10% fetal bovine serum and 2.5 mM thymidine. After 24 h, the cells were washed twice with PBS and fed with medium containing camptothecin (0.5 M). One hour later, the cells were washed twice with PBS and fed with complete medium for additional 23.5 h.
Immunoprecipitation, Immunoblot Assays, and Cellular Fractionation-The 7721 cells grown in RPMI 1640 medium supplemented with 10% bovine calf serum were plated in 60-mm dishes (Nunc) at a concentration of 6 ϫ 10 5 cells/dish the day before transfection. Plasmid DNA (4 g) was transfected into 7721 cells with a calcium phosphate precipitation method. Two days after transfection, cells were washed three times with ice-cold PBS and solubilized with 1 ml of lysis buffer (50 mM Tris HCl, pH 7.5, 150 mM NaCl, 0.1% Nonidet P-40, 5 mM EDTA, 5 mM EGTA, 15 mM MgCl 2 , 60 mM ␤-glycerophosphate, 0.1 mM sodium orthovanadate, 0.1 mM NaF, 0.1 mM benzamide, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride). Detergent-insoluble materials were removed by centrifugation at 13,000 rpm for 15 min at 4°C. The whole cell lysates were incubated with mouse normal IgG or anti-HA monoclonal antibody at 4°C for 2 h. Pre-equilibrated protein G-agarose beads were then added, and after 4 h of incubation, they were collected by centrifugation and then gently washed three times with the lysis buffer. The bound proteins were eluted by boiling in SDS sample buffer and resolved on a 10% SDS-PAGE gel. The proteins were transferred onto a polyvinylidene difluoride membrane and probed with a 1:1000 dilution of a monoclonal anti-cyclin D3 antibody. Proteins were detected using the ECL kit.
The coimmunoprecipitation in 7721 cells under normal physiological situations was conducted with the normal 7721 cells and the 7721 cells synchronized at a different cell cycle phase. The method was the same as above except that the antibody used for immunoprecipitation was monoclonal anti-cyclin D3 antibody and for immunoblot was rabbit polyclonal anti-PITSLRE antibody. The coimmunoprecipitation for the HeLa cells was the same as that of the 7721 cells.
Cellular fractionation was performed as below. G 2 /M phase-arrested cells (8 ϫ 10 6 ) were suspended for 5 min on ice in 500 l of buffer (10 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 25 mM KCl, 250 mM sucrose, 1ϫ complete protease inhibitors, 0.3% Nonidet P-40). After gentle mixing, the lysate fraction was centrifuged at 1000 rpm for 2 min at 4°C. The resulting supernatants constituted the cytoplasmic fractions with the pellets representing the nuclear fractions. Coimmunoprecipitation was performed with anti-cyclin D3 monoclonal antibody, and immunoblot analysis was performed using rabbit polyclonal anti-PITSLRE antibody (29).
Immunofluorescence-The 7721 cells were plated onto coverslips the day before synchronization. After synchronization, they were fixed in ice-cold methanol for 1 h and blocked in PBS containing 10% normal blocking serum followed by an overnight reaction with the primary antibody at 4°C. The primary antibody consisted of monoclonal anticyclin D3 antibody and the rabbit polyclonal anti-PITSLRE antibody. After overnight incubation, the coverslips were rinsed 3 times in PBS and reacted for 1 h with goat anti-mouse IgG-fluorescein isothiocyanate and goat anti-rabbit IgG-R (from Santa Cruz) in the dark. The coverslips were washed as described above, inverted, mounted on slides, and sealed with nail polish. The coverslips were examined in a Leica confocal microscope. Digitized images of the fluorescent-antibody-stained cells were acquired with software provided by Leica.
Fluorescence Imaging of Living Cell-The 7721 cells were plated onto coverslips the day before transfection. The pEGFP-p58 PITSLRE and pDsRed-cyclin D3 or pEGFP-p58 PITSLRE with pDsRed C1 were transiently co-transfected into 7721 cells with LipofectAMINE-PLUS reagent (Invitrogen) according to the manufacturer's instructions. After 48 h, the transfected cells were fixed for 30 min with 3% paraformaldehyde in PBS and observed under the Leica confocal microscope as described above.

RESULTS
Identification of Cyclin D3 as p58 PITSLRE Protein Kinaseinteracting Protein-To identify proteins that interact with p58 PITSLRE , the yeast two-hybrid system was employed with p58 PITSLRE -fused LexA DNA binding domain as bait. The bait did not have any intrinsic activity of transcriptional activation for the two reporters. A human fetal liver cDNA library was screened as described under "Experimental Procedures." Ap-proximately 6 ϫ 10 6 independent transformants were pooled and spread on the selection media (UraϪ, HisϪ, TrpϪ, and LeuϪ) containing 2% galactose to induce the expression of library cDNA. In the selection media, 50 colonies showed LEU2 ϩ /LacZ ϩ . The plasmids were extracted by yeast miniprep for further study. False positive clones were eliminated with the following approach. The positive library plasmids were reintroduced into the yeast alone or with (a) pLexA, (b) pLexA-p58 PITSLRE , or (c) pLexA hybrid with an unrelated protein.
Only the transformants that co-transformed the library plasmid with pLexA-p58 PITSLRE were positive for ␤-galactosidase activity, indicating true positive interactions. Among the first 50 LEU2 ϩ /LacZ ϩ colonies, there were 19 true positive colonies. The cDNAs from the 19 true positive colonies were PCR-amplified with primers derived from the vector pB42AD followed by sequence determination. DNA sequencing and data base searching revealed that the nucleotide sequence of 5 clones encoded full-length of human cyclin D3. The other 14 clones are in progress in our lab.
To further confirm the interaction between p58 PITSLRE and cyclin D3, two cloning vectors were exchanged by moving cyclin D3 from the activation domain (pB42AD) to the DNA-BD vector (pLexA) and p58 PITSLRE from the pLexA to pB42AD. The repeated two-hybrid assay was also positive for the two reporters (data not shown).
Two-hybrid Interactions between p58 PITSLRE and D-type Cyclins-The fact that cyclin D3 was identified by two-hybrid screening using p58 PITSLRE as bait raised the question of whether p58 PITSLRE interacted preferentially with this D-type cyclin or it also interacted with the other two D-type cyclins. To answer this question, we used a direct two-hybrid experiment to compare cyclin D1, cyclin D2, and cyclin D3 for their ability to bind to p58 PITSLRE . As a positive control, cyclin D3 was included in this experiment. As the negative control, p58 PITSLRE alone did not permit growth of the yeast on nutrient-deficient medium. Subsequent transformation with either of the D-type cyclin constructs showed that neither of the other two D-type cyclins permitted activation of the reporter genes, whereas cyclin D3, in the presence of p58 PITSLRE , did activate the two report genes (data not shown). These data indicated that only cyclin D3 but not cyclin D1 or cyclin D2 interacted with p58 PITSLRE in the yeast two-hybrid system (Fig 1).
In Vitro Interactions between p58 PITSLRE and D-type Cyclins-The ability of D-type cyclins to interact with p58 PITSLRE was further tested using a GST pull-down experiment. The GST-p58 PITSLRE , GST-CDK4, GST, cyclin D1, cyclin D2, and cyclin D3 were synthesized and isotopically labeled in vitro. The labeled proteins were incubated together GST-p58 PITSLRE incubated with cyclin D1, cyclin D2, or cyclin D3 and GST-CDK4 incubated with cyclin D1, cyclin D2, or cyclin D3 as FIG. 1. Schematic diagram of the interaction between p58 PITSLRE and the D-type cyclins. Mapping of the p58 PITSLRE regions that interact with cyclin D3 and determination of the interaction between p58 PITSLRE and D-type cyclins. Deletion constructs of p58 PITSLRE (the domains and residue numbers are indicated) were tested for interaction with cyclin D3, and the fulllength of p58 PITSLRE was analyzed for its ability to interact with the other two Dtype cyclins using the two-hybrid system in yeast. Columns on the right summarize whether constructs did (ϩ) or did not (Ϫ) interact.
positive controls and GST incubated with cyclin D3 as a negative control. The protein mixtures were bound to glutathione-Sepharose beads, washed, and subjected to SDS-PAGE. The resulting gel was then exposed. Only the GST-p58 PITSLRE band was observed when GST-p58 PITSLRE was incubated with cyclin D1 or cyclin D2. A strong cyclin D3 signal was observed after incubation of GST-p58 PITSLRE with cyclin D3 (Fig. 2). For the positive control and negative control, cyclin D1, cyclin D2, and cyclin D3 were observed after incubation with GST-CDK4, and no cyclin D3 was observed after incubation with GST. These data showed that p58 PITSLRE interacted preferentially with the cyclin D3.
Mapping of the p58 PITSLRE Region That Interacted with Cyclin D3-In addition to the conserved p34 cdc2 -related Ser/ Thr protein kinase catalytic domain, p58 PITSLRE also contains a unique 74-amino acid NH 2 -terminal region with a putative calmodulin binding site, nuclear localization sequence, and three tandem PEST sequences (16). During Fas-and tumor necrosis factor ␣-induced cell death, its NH 2 -terminal region is cleaved by multiple caspases (21)(22)(23). Furthermore, ectopic expression of its NH 2 -terminal deletion mutant, which resembles the final caspase-modified product, has also been shown to induce apoptosis (20). To investigate the region in p58 PITSLRE responsible for binding to cyclin D3, we constructed two p58 PITSLRE mutants (Fig 1), one containing NH 2 -terminal 100 amino acids (NH 2 -p58 PITSLRE ) and the other lacking NH 2 -terminal 74 amino acids (⌬NH 2 -p58 PITSLRE ) (20). These two mutant constructs were co-transformed either with the empty pB42AD plasmid or with pB42AD-cyclin D3 into yeast cells. Co-transformants were tested for growth in the absence of leucine and production of ␤-galactosidase. No growth occurred in all the co-transformants (data not shown), which indicated that neither p58 PITSLRE mutants interacted with cyclin D3. This result suggests that the full-length of p58 PITSLRE might be necessary for its binding to cyclin D3 (Fig. 1), which will be further described below.
Binding of p58 PITSLRE with Cyclin D3 at G 2 /M Phase in Mammalian Cells-To further investigate the interaction of p58 PITSLRE and cyclin D3, we tested whether they associated in mammalian cells. The p58 PITSLRE protein kinase was tagged at its amino terminus with an HA epitope and transiently expressed in 7721 cells, a human hepatocarcinoma cell line. The expression of p58 PITSLRE was confirmed by a monoclonal antibody against HA epitope (Fig. 3A). The whole cell lysates, with equal amounts of HA-p58 PITSLRE and cyclin D3 proteins, were immunoprecipitated with normal mouse IgG or anti-HA monoclonal antibody followed by immunoblot analysis using an anticyclin D3 monoclonal antibody. As shown in Fig. 3B, cyclin D3 was coimmunoprecipitated with HA-p58 PITSLRE , whereas no cyclin D3 was detected in the control mouse IgG immunoprecipitation.
The ectopic expression of p58 PITSLRE is not cell cycle-regulated, whereas in vivo, p58 PITSLRE is produced almost exclusively in G 2 /M. To investigate whether p58 PITSLRE and cyclin D3 can interact in a normal physiological situation, we synchronized the 7721 cells and did immunoprecipitation in different stages of the cell cycles. After sequential thymidine treatment, there were 91.25% cells in G 1 phase, 3% cells in S phase, and no cells in G 2 /M phase (Fig. 3C). To arrest cells in G 2 /M phase, we incubated cells first with thymidine (2.5 mM), then with camptothecin (0.5 M). Finally, there were 72.75% of the cells arrested in G 2 /M phase (Fig. 3C). After synchronization, much more p58 PITSLRE protein was found in the G 2 /M phasearrested cells than in the G 1 /S phase-arrested cells (Fig. 3D).
Cell lysates from different cell cycles were subjected to immunoprecipitation with anti-cyclin D3 antibody followed by immunoblot analysis using a rabbit anti-PITSLRE polyclonal antibody. As shown in Fig. 3E, p58 PITSLRE coimmunoprecipitated with cyclin D3 in G 2 /M phase but not in G 1 /S phase. For normal 7721 cells, there were about 15% cells in G 2 /M phase, so the interaction could still be observed. However, the amount of the p58 PITSLRE that coimmunoprecipitated with cyclin D3 in the normal 7721 cells was much less than that in the G 2 /M phase-arrested 7721 cells. In addition, we also detected this association between p58 PITSLRE and cyclin D3 in HeLa cells with coimmunoprecipitation (data not shown).
To further address the subcellular interaction of p58 PITSLRE with cyclin D3, we did coimmunoprecipitation after crude fractionation of the G 2 /M phase-arrested 7721 cell lysates into nuclear and cytoplasmic components (Fig. 3F). The results showed that p58 PITSLRE and cyclin D3 interacted mostly in the nuclear fraction but not in the cytoplasmic fraction.
The rabbit polyclonal anti-PITSLRE antibody used for immunoblotting was raised against a COOH-terminal peptide, PITSLRE, which is conserved in all the PITSLRE isoforms (19). Therefore, it can recognize all the PITSLRE isoforms in the 7721 cells, including p58 PITSLRE and p110 PITSLRE , with the expression of p110 PITSLRE much more than that of p58 PITSLRE . Cyclin D3 coimmunoprecipitated only with p58 PITSLRE in the G 2 /M phase but not with p110 PITSLRE in all the cell cycle (shown in Fig. 3E). Thereby it demonstrated that only p58 PITSLRE isoform could interact with cyclin D3.
Immunofluorescence Analysis of the p58 PITSLRE 35 S-labeled cyclin D1, 35 S-labeled cyclin D2 as positive control, and [ 35 S]GST was incubated with 35 S-labeled cyclin D3 as negative control. After incubation, the beads were washed three times with the binding buffer and analyzed by autoradiography after SDS-PAGE. Lanes from left to right are GST ϩ cyclin D3, GST-CDK4 ϩ cyclin D1, GST-CDK4 ϩ cyclin D2, GST-CDK4 ϩ cyclin D3, 20% cyclin D1 input, 20% cyclin D2 input, 20% cyclin D3 input, GST-p58 PITSLRE ϩ cyclin D1, GST-p58 PITSLRE ϩ cyclin D2, GST-p58 PITSLRE ϩ cyclin D3. chronized in different cell cycles were fixed and reacted with anti-PITSLRE and anti-cyclin D3 antibodies as described under "Experimental Procedures." The secondary antibodies tagged with fluorescein isothiocyanate and rhodamine, respectively, were used to stain and detect the localization of PITSLRE protein kinases and cyclin D3. When the staining images of the PITSLRE (Fig. 4A, II) and cyclin D3 (Fig. 4A, I) were merged in G 1 /S phase-arrested cells, the PITSLRE isoforms, most of which was p110 PITSLRE isoform, were found not to colocalize with cyclin D3, for no yellow color was visualized in the merged image (Fig. 4A, III). In G 2 /M phase, the staining image of PITSLRE isoforms (Fig. 4B, II), including p58 PITSLRE isoform, was shown to colocalize with that of cyclin D3 (Fig. 4B, I). The yellow color visualized in the merged image represents colocalization of p58 PITSLRE and cyclin D3 (Fig. 4B, III). All these data verified that cyclin D3 did associate with p58 PITSLRE in G 2 /M phase but not associate with p110 PITSLRE . Because the p110 PITSLRE isoforms contain the entire p58 PITSLRE sequence, all the anti-p58 PITSLRE antibodies can recognize p110 PITSLRE isoforms at the same time, which may interfere the colocalization between p58 PITSLRE and cyclin D3 observed by the anti-p58 PITSLRE antibody. To further confirm this colocalization, we co-transfected the 7721 cells with pEGFP-p58 PITSLRE and pD-sRed-cyclin D3. The cells double-transfected with EGFP-p58 PITSLRE and pDsRed or with pEGFP and DsRed-cyclin D3 were used as control (Fig. 5, A and B). Forty-eight hours after transfection, the cells were harvested, washed, fixed, sealed, and analyzed under confocal microscopy. Merging the separate fluorescent images obtained from EGFP-p58 PITSLRE and DsRed-cyclin D3 emission detection, we observed that the double-transfected cells contained yellow, indicating colocalization of p58 PITSLRE and cyclin D3 (Fig. 5C). Moreover, compared with the mock-transfected cells (Fig. 5A), the elevated expression of cyclin D3 affected p58 PITSLRE cellular distribution (Fig. 5C). In the cells double-transfected with pEGFP-p58 PITSLRE and pD-sRed, the fluorescent signals of p58 PITSLRE were detected both in nucleus and in cytoplasm, with the signal in nucleus much higher than that in cytoplasm (Fig. 5A, I). Although upon co-transfection with cyclin D3, p58 PITSLRE localized exclusively in the nuclear region, with no signal detected in the cytoplasm (Fig. 5C).
Enhanced p58 PITSLRE Kinase Activity upon Cyclin D3 Interaction-Cyclin D3 is well known as a regulatory cyclin of CDK 4 and CDK 6, regulating their kinase activities (6,7). To investigate whether the association with cyclin D3 would also influence the kinase activity of p58 PITSLRE , we used an immunodepletion kinase assay with histone H1 as the substrate to analyze this effect. The 7721/p58 cells in which HA-p58 PITSLRE was stably expressed were used for the following assay. The whole cell lysates from 7721/p58 cells containing equal amounts of HA-p58 PITSLRE were immunoprecipitated with an anti-HA monoclonal antibody in the presence of cyclin D3 or in the absence of cyclin D3 (immunodepleted by the monoclonal anti-cyclin D3 antibody). In vitro kinase assays of the anti-HA-  The secondary antibodies were antimouse IgG-conjugated to fluorescein isothiocyanate and anti-rabbit IgG-conjugated to rhodamine red. The images were captured with a Leica confocal microscope and software provided by Leica. A, the 7721 cells synchronized in G 1 /S phase were observed. I, the cyclin D3 image captured. II, the PITSLRE image of the same frame as in I. III, the merge of I and II. B, the 7721 cells synchronized in G 2 /M phase were observed. I, the cyclin D3 image captured. II, the PITSLRE image of the same frame as in I. III, the merge of I and II.
p58 PITSLRE immunoprecipitates revealed that p58 PITSLRE kinase activity was significantly decreased in the absence of cyclin D3 (Fig. 6, A and B). The rabbit polyclonal anti-PITSLRE antibody was also used for immunoprecipitation. The in vitro kinase assays of the anti-PITSLRE immunoprecipitates confirmed this decrease, whereas the latter decrease was smaller than the former one (Fig. 6, A and C).
In previous work, it was reported that p58 PITSLRE could copurify with ␤-1,4-galactosyltransferase 1, phosphorylate it, and modulate its activity (16,32). ␤-1,4-Galactosyltransferase 1, the key enzyme transferring galactose to the terminal Nacetylglucosamine-forming Gal␤134GlcNAc structure in the Golgi apparatus (33,34), might be more specific as the substrate for p58 PITSLRE kinase assay than histone H1. The kinase activity of p58 PITSLRE was also greatly decreased in the absence of cyclin D3 using ␤-1,4-galactosyltransferase 1 as substrate (Fig. 7, A and B). Together, these observations suggest that cyclin D3 plays an important role in the regulation of p58 PITSLRE kinase activity. DISCUSSION For a long time, ␤-1,4-galactosyltransferase 1 was considered the only protein that could interact with p58 PITSLRE (16). Through this binding, p58 PITSLRE phosphorylates ␤-1,4-galactosyltransferase 1 and enhances its activity (16,31,32). Actually, ␤-1,4-galactosyltransferase 1 serves as a substrate for p58 PITSLRE . As a p34 cdc2 -related protein kinase, p58 PITSLRE plays an important role in cell cycle control by leading to a late mitotic delay in response to minimal overexpression of this protein kinase (16,20). In addition, expression of p58 PITSLRE is G 2 /M phase-specific, resulting from translation controlled by an internal ribosome entry site (29). Based on its sequence homology and function, p58 PITSLRE might be considered a CDK in G 2 /M phase, but its partner cyclin and substrates other than ␤-1,4-galactosyltransferase 1 remain unknown. In this study, we demonstrate that cyclin D3 interacts with p58 PITSLRE in vitro and in vivo, and this interaction is found only in G 2 /M phase but not in the G 1 /S phase of the cell cycle. The elevated expression of cyclin D3 leads to an exclusively accumulation of p58 PITSLRE in the nuclear region. Moreover, kinase activity of p58 PITSLRE is greatly decreased without cyclin D3 binding. All of these suggest that cyclin D3 may function as a regulatory partner of p58 PITSLRE .
The human cyclin D3 gene was cloned from a placental cDNA library by cross-hybridization with cyclin D1 probe (35). Compared with cyclin D1 and cyclin D2, little is known about the function of cyclin D3 (36). Cyclin D1 knockout mice are slightly smaller and exhibit a lack of normal mammary gland development in adult female mice as well as retinopathy (37,38), whereas mice lacking cyclin D2 are infertile due to lack of development of ovarian granulosa cells (39). Successful disruption of the cyclin D3 gene in mice has not been reported. The overexpression of cyclin D3 in fibroblast cells leads to accelerated passage through G 1 phase with no effect on the overall doubling time (36). Moreover, cyclin D3 is found to not only play a crucial role in progression through the G 1 phase but also to regulate apoptosis induced by T cell receptor activation in leukemic T cell lines (40).
As cells enter cell cycle from quiescence, one or more D-type cyclins (cyclins D1, D2, D3) are induced and subsequently expressed throughout the cell cycle in response to mitogen stimulation, whereas cyclin A, B, and E (mitotic cyclins) are expressed periodically (3,6,7). Considerable attention has been paid to the role of D-type cyclins in controlling the G 1 phase progression by regulating CDKs 4 and 6 activation and Rb function (3,7,41). There is currently little evidence of a role for them in the later cell cycle. Here, we show that cyclin D3 may function in G 2 /M phase, serving as an interaction partner of p58 PITSLRE and regulating some parts of its function. This interaction linked a G 1 cyclin (cyclin D3) with a G 2 /M CDK (p58 PITSLRE ). No interaction between the p58 PITSLRE protein kinase and the other two D-type cyclins was observed in direct two-hybrid assay and GST pull down experiments. This indi- FIG. 5. Overexpression of cyclin D3 in 7721 cells lead to an exclusively accumulation of p58 PITSLRE in the nuclear region. The full-length of p58 PITSLRE was inserted into the pEGFP N3 to be expressed as a fusion protein with EGFP in 7721 cells, and full-length of cyclin D3 was inserted into the pDsRed C1 as a fusion protein with DsRed. After transfection of the indicated plasmids, the 7721 cells were cultured for 48 h and observed by confocal microscopy. A, the cells co-transfected with pEGFP-p58 PITSLRE and pDsRed C1 were observed by confocal microscopy. I, the pEGFP-p58 PITSLRE image of the cells co-transfected with pEGFP-p58 PITSLRE and pDsRed C1. II, pDsRed image of the same frame as in I. B, the cells co-transfected with pEGFP and pDsRed-cyclin D3 were observed by confocal microscopy. I, the pEGFP image of the cells co-transfected with pDsRedcyclin D3. II, pDsRed-cyclin D3 image of the same frame as in I. C, the cells cotransfected with pEGFP-p58 PITSLRE and pDsRed-cyclin D3 were observed by confocal microscopy. I, the pEGFP-p58 PITSLRE image of the cells co-transfected with pEGFP-p58 PITSLRE and pDsRed-cyclin D3. II, the pDsRed-cyclin D3 image of the same frame as in I. III, the merge of I and II.
cates that the binding between p58 PITSLRE and cyclin D3 might be specific. The high homology between the three D-type cyclins has suggested redundancy in their functions. However, there is more and more evidence that the three D-type cyclins are not equivalent in many ways, such as the tissue-specific expression patterns (7), different affinities to CDKs (42), different inductions by various signals in a cell lineage-specific manner (3,7), and different phenotypes of the knock-out mice (37-39) (homozygous disruption of cyclin D3 is not obtained by now). Given our results, it is likely that the interaction with p58 PITSLRE plays a distinct role of cyclin D3 in cell cycle control.
The p58 PITSLRE belongs to a large family that contains many isoforms. Among them the p58 PITSLRE and p110 PITSLRE are mostly studied and described. The p110 PITSLRE protein kinase was shown to participate in a signaling pathway that potentially regulates transcription and RNA-processing events, whereas the p58 PITSLRE plays an important role in the cell cycle progression control. Although the p110 PITSLRE isoform contains the entire p58 PITSLRE sequence, it did not associate with cyclin D3 by immunoprecipitation (Fig.. 3) and immunofluorescence cell staining (Fig. 4). This suggests that the NH 2 terminus of p110 PITSLRE may interfere or block the conformation of the COOH terminus so that the p58 PITSLRE sequence in the p110 PITSLR could not reach and interact with cyclin D3. These data are in agreement with the different functions of the two PITSLRE isoforms.
Our studies have demonstrated that cyclin D3 interacted and colocalized with p58 PITSLRE at G 2 /M phase, and the elevated expression of cyclin D3 affected p58 PITSLRE cellular distribution. In addition, we speculate that this interaction and colocalization mainly existed in the nucleus for the biochemical fractionation study, which showed that p58 PITSLRE and cyclin D3 interacted mostly in the nuclear fraction but not in cytoplasmic fraction (Fig. 3F), and the yellow color visualized in the merged image was mainly localized in the nucleus (Figs. 4 and 5). When co-transfected with a control plasmid, p58 PITSLRE was shown to localize predominantly in the nucleus, with a little cytoplasm distribution (Fig. 5). This is consistent with the protein structure and function of p58 PITSLRE , which contains a nuclear localization sequence in its NH 2 -terminal region (16). For the p110 PITSLRE , it primarily localized in the nucleus (19,29,30). Upon co-transfection with cyclin D3, p58 PITSLRE appeared completely nucleus-localized without any signal de-FIG. 6. Activation of the p58 PITSLRE kinase activity on histone H1 by cyclin D3 association. A, a, immunoblot analysis of cyclin D3 immunodepletion efficiency. After immunodepletion, the precipitates were immunoblotted with anti-cyclin D3 antibody. More than 90% depletion was achieved by cyclin D3 immunodepletion. b, anti-HA monoclonal antibody (lanes 3 and 4) or anti-PITSLRE polyclonal antibody (lanes 1 and 2) was used to precipitate p58 PITSLRE from 200 g cell lysates of 7721/p58. After immunodepletion of cyclin D3 (Ϫcyclin D3) or directly (ϩcyclin D3), kinase activity of the precipitates was measured with histone H1 as the substrate. The figure is representative of three independent experiments performed. B and C, for anti-HA precipitates (B) or anti-PITSLRE precipitates (C), relative kinase activity of p58 PITSLRE was determined by quantitation of the labeled histone H1 bands with the ImageQuant software. Phosphorylation activity is presented as percent where kinase activity of HA precipitates (B) or anti-PITSLRE precipitates (C) in the presence of cyclin D3 is arbitrarily set at 100%.

FIG. 7.
Phosphorylation of ␤-1,4-galactosyltransferase 1 by p58 PITSLRE was greatly suppressed in the absence of cyclin D3. A and B, anti-HA monoclonal antibody (A) or anti-PITSLRE polyclonal antibody (B) was used to precipitate p58 PITSLRE in the presence of (ϩcyclin D3) or absence of cyclin D3 (Ϫcyclin D3). Kinase activity of the precipitates was measured with ␤-1,4-galactosyltransferase 1 as the substrate as described under "Experimental Procedures." The measurements are representative of three independent experiments performed. The relative phosphorylation activity is presented as percent where kinase activity of HA precipitates (A) or anti-PITSLRE precipitates (B) in the presence of cyclin D3 is arbitrarily set at 100%. tected in the cytoplasm (Fig. 5). However, it is preliminary to say that cyclin D3 can enhance p58 PITSLRE nuclear translocation, because many factors can make increased nuclear accumulation. This issue is currently under investigation in our lab. From Fig. 4 and Fig. 5, we found that there were still plenty of cyclin D3 that did not interact with p58 PITSLRE , because cyclin D3 acts as a regulatory subunit of CDKs 4 and 6 as well as an interaction partner of two distinct types of transcription factors, estrogen receptor and DMP1 (43,44). Through direct binding, cyclin D3 can enhance the growth-promoting activity of the estrogen receptor and inhibit the growth-restraining capacity of the DMP1 (43,44). The other issue raised from Fig.  4A is that cyclin D3 does not show any tendency toward nuclear localization in G 1 /S phase-arrested cells, which might be due to the different abundance or affinities of the D-type cyclins to the CDKs in 7721 cells (3,7,42). The other two D-type cyclins may occupy most of the CDKs so that cyclin D3 distributes all over the cells instead of a tendency toward the nucleus. From Fig.  5B, we can also observe the cytoplasmic distribution of the cyclin D3 in control cells, but upon co-expression with p58 PITSLRE , cyclin D3 localizes exclusively in the nucleus.
The in vitro immune complex kinase assay showed that kinase activity of p58 PITSLRE was significantly decreased when the binding between p58 PITSLRE and cyclin D3 was abrogated by immunodepletion with a monoclonal anti-cyclin D3 antibody. We used two different antibodies for immunoprecipitations in this assay; one is the mouse anti-HA monoclonal antibody, and the other is the rabbit polyclonal anti-PITSLRE antibody. The observed slight decrease in the kinase activity of the anti-PITSLRE immunoprecipitates in the absence of cyclin D3 could be due to its low specificity for p58 PITSLRE . All together, it is speculated that cyclin D3 may function as a regulatory partner of p58 PITSLRE in G 2 /M phase, which is a good explanation for the results of Herzinger and Reed (36). In their study, they found that the overexpression of cyclin D3 in fibroblast cells led to accelerated passage through G 1 phase with no effect on the overall growth rate, which suggested that the accelerated passage through G 1 phase might be compensated for by expanding subsequent cell cycle phases. Here we partly confirmed their postulation and demonstrated that p58 PITSLRE might be the target molecule for the subsequent expanding G 2 /M cell cycle phase.
In summary, this study demonstrates that cyclin D3, a G 1 cyclin, specifically interacted with p58 PITSLRE , a G 2 /M CDK. This binding happened in G 2 /M phase instead of G 1 /S phase and resulted in enhanced kinase activity of p58 PITSLRE . Therefore, cyclin D3 functioned not only in G 1 phase as a regulatory subunit of CDKs 4 and 6 but also in G 2 /M phase as a partner of p58 PITSLRE during cell cycle progression. Further analysis of this interaction along with past studies might result in a much more generalized understanding of the regulation and function of cyclin D3 and p58 PITSLRE , thereby providing new insights into the control of G 2 /M phase cell cycle progression.