Mirk/dyrk1B kinase destabilizes cyclin D1 by phosphorylation at threonine 288.

The phosphorylation of cyclin D1 at threonine 286 by glycogen synthase kinase 3beta (GSK3beta) has been shown to be required for the ubiquitination and nuclear export of cyclin D1 and its subsequent degradation in the proteasome. The mutation of the nearby residue, threonine 288, to nonphosphorylatable alanine has also been shown to reduce the ubiquitination of cyclin D1, suggesting that phosphorylation at threonine 288 may also lead to degradation of cyclin D1. We now demonstrate that the G(0)/G(1)-active arginine-directed protein kinase Mirk/dyrk1B binds to cyclin D1 and phosphorylates cyclin D1 at threonine 288 in vivo and that the cyclin D1-T288A construct is more stable than wild-type cyclin D1. Transient overexpression of Mirk in nontransformed Mv1Lu lung epithelial cells blocked cells in G(0)/G(1). Depletion of endogenous Mirk by RNA interference increased cyclin D1 protein levels but not mRNA levels, indicating that Mirk destabilizes cyclin D1 protein. Destabilization was confirmed by induction of a stable Mirk transfectant of Mv1Lu cells, which blocked cell migration (Zou, Y., Lim, S., Lee, K., Deng, X., and Friedman, E. (2003) J. Biol. Chem. 278, 49573-49581), and caused a decrease in the half-life of endogenous cyclin D1, concomitant with an increase in Mirk expression. In vitro cyclin D1 was phosphorylated in an additive fashion by Mirk and GSK3beta. Mirk-phosphorylated cyclin D1 mutated at the GSK3beta phosphorylation site and was capable of phosphorylating cyclin D1 in the presence of the GSK3beta inhibitor LiCl. Mirk may function together with GSK3beta to assist cell arrest in G(0)/G(1) by destabilizing cyclin D1.

ing factors needed for the progression into S phase. Cyclin D1 is translocated into the cytoplasm during S phase where it is destroyed by the proteasome following phosphorylation at threonine 286 by GSK3␤ (2,3). Mutant cyclin D1-T286A, which cannot be phosphorylated by GSK3␤, is stabilized in the nucleus and is capable of transforming murine fibroblasts, whereas overexpression of wild-type cyclin D1 cannot act alone to transform such cells (4). A cyclin D1 isoform derived by alternative splicing was shown to lack threonine 286, enabling this cyclin D1 isoform to remain nuclear throughout the cell cycle, remain highly expressed, and function to facilitate transformation of NIH3T3 cells (5). This cyclin D1 splice variant was also found in tumor-derived cells and primary human esophageal tumors (5). Overexpression of cyclin D1 occurs in several cancers including breast, pancreatic, and esophageal (6), suggesting that either increased transcription, transcription of stable splice variants, or dysregulation of cyclin D1 turnover may frequently occur in cancer.
In this study, we have studied the interaction of the ubiquitously expressed protein kinase Mirk/dyrk1B with cyclin D1. Mirk/dyrk1B is an arginine-directed serine/threonine kinase (7), which functions as a transcriptional co-activator and is activated through the stress-activated mitogen-activated protein kinase kinase MKK3 (8). We have shown recently that Mirk stabilizes the CDK inhibitor p27kip1 in the G 0 phase of the cell cycle in NIH3T3 fibroblasts, whereas depletion of Mirk by RNA interference increases cell cycling as measured by increased PCNA expression (9). Mirk expression is decreased by mitogen activation of the MEK-ERK pathway during G 1 (10), restricting Mirk function primarily to G 0 and early G 1 . We now confirm that transient overexpression of Mirk in nontransformed Mv1Lu lung epithelial cells increases the length of G 0 /G 1 by FACS analysis and that Mirk targets the G 1 cell cycle regulator, cyclin D1, to maintain cells in growth arrest.

EXPERIMENTAL PROCEDURES
Materials-Affinity-purified rabbit polyclonal antibody to a unique sequence at the C terminus of Mirk and affinity-purified rabbit polyclonal antibody to a unique sequence at the N terminus of Mirk were raised as described previously (7). Antibodies to cyclin D1 and GSK3␤ were from Santa Cruz Biotechnology. Antibody to GSK3␤ phosphorylated at serine 9 was purchased from Cell Signaling Technology. Recombinant purified N-terminal histidine-tagged GSK3␤ was from Sigma. Polyvinylidene difluoride transfer membrane Immobolin-P was purchased from Millipore. PLUS reagent and LipofectAMINE were from Invitrogen. All of the radioactive materials were purchased from PerkinElmer Life Sciences, and ECL reagents were from Amersham Biosciences. All of the other reagents were from Sigma.
Plasmid Construction-The murine pFLEX-cyclin D1 construct and FLAG-tagged cyclin D1-T286A derivative were the kind gifts of Drs. Martine Roussel and Charles Sherr, and the wild-type GSK3␤ construct was the kind gift of Dr. Alan Diehl. Plasmids pcDNA3.1 (Mirk) and pcDNA3.1 (kinase-inactive YF Mirk) had been generated previously (7). Wild-type cyclin D1 was subcloned into pGEX-4T1 (Amersham Biosciences) to make GST-tagged mutants by site-directed mutagenesis by the GeneEditor system (Promega). Constructs were subcloned into pCMV-tag2B to generate FLAG-tagged cyclin D1 constructs. All of the mutant cyclin D1 constructs were sequenced to confirm the mutated sequence.
Immunodetection-Following treatment as indicated and washing with cold phosphate-buffered saline, cells were lysed in radioimmune precipitation assay buffer (phosphate-buffered saline, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, and protease inhibitors 20 g/ml leupeptin, 20 g/ml aprotinin, 0.5 mM phenylmethylsulfonyl fluoride, 200 M sodium orthovanadate, and 20 mM sodium fluoride). Lysates were pelleted in a microcentrifuge for 15 min to remove insoluble material. Depending upon the experiment, 10 -50 g of cell lysates were blotted onto polyvinylidene difluoride membranes after separation by SDS-PAGE. The blots were blocked in 5% milk in TBST (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 0.1% Tween 20) for 1 h at room temperature and incubated for 1 h at room temperature with primary antibody in TBST plus 4% bovine serum albumin, and proteins were detected subsequently by enhanced chemiluminescence. All of the Mirk blots used affinity-purified polyclonal antibody directed to the Mirkunique C terminus.
Transient Transfections-Mv1Lu cells were transfected transiently by adding a complex of PLUS reagent (3 l/g DNA) and Lipo-fectAMINE (2-4 l/g DNA) in serum-free media for 3 h and then allowing the expression in growth medium for 24 -36 h, or they were transfected using LipofectAMINE 2000 (3 l/g DNA) in serum-containing medium for 24 -36 h as noted. The amount of total DNA used was kept constant by the addition of empty vector DNA.
Co-immunoprecipitation of Mirk and Cyclin D1-Mirk and cyclin D1 expression plasmids were co-transfected into MV1Lu cells and allowed to express for 48 h. After lysis in 50 mM Tris-HCl, pH 8.0, and 0.5% Nonidet P-40 with protease inhibitors, cyclin D1 was immunoprecipitated by adding antibodies to the FLAG epitope followed by rocking overnight at 4°C. 20 l of protein G-agarose conjugates (Santa Cruz Biotechnology) then was added and incubated for an additional 2 h at 4°C. The agarose beads in each tube were washed three times followed by SDS-PAGE and Western blotting.
Co-immunoprecipitation of GSK3␤ and Cyclin D1-Endogenous GSK3␤ and associated proteins were immunoprecipitated following lysis in 50 mM Tris-HCl, pH 8.0, and 0.5% Nonidet P-40 with protease inhibitors by adding antibody to GSK3␤ (2 g to 1 mg of total lysate) followed by rocking overnight at 4°C. 20 l of protein A-agarose conjugates then were added and incubated for an additional 2 h at 4°C. The agarose beads in each tube were washed three times followed by SDS-PAGE and Western blotting. The second immunoprecipitation of GSK3␤-associated cyclin D1 was performed by diluting the immunoprecipitates (in SDS sample buffer without dye) from 50 l to 1 ml by adding 2% bovine serum albumin. The diluted sample was then incubated with anti-cyclin D1-conjugated protein G-agarose. The agarose beads were washed three times followed by SDS-PAGE and autoradiography. This method was adapted from Borja, et al. (13).
In Vitro Kinase Reactions-GST-Mirk or its kinase-inactive YF mutant form was incubated together with GST-cyclin D1 as noted in kinase buffer (10 mM Tris, pH 7.4, 150 mM NaCl, 10 mM MgCl 2 , and 0.5 mM dithiothreitol with protease and phosphatase inhibitors containing 10 M cold ATP plus 2.5 Ci of [ 32 P]ATP) for 10 min at 30°C. In some experiments as noted, 40 mM LiCl was added to the reaction mixture to inhibit GSK3␤. Reaction mixtures then were analyzed by SDS-PAGE and autoradiography.
Peptide Mapping-The cyclin D1 mutants were subcloned into pGEX-6P1 to take advantage of the cold cleavage conditions using the PreScission enzyme (Amersham Biosciences). Following an in vitro kinase reaction, the phosphorylated cyclin D1 band was cut from the SDS-polyacrylamide gel. The gel slice was washed five times with water for 20 min each and then soaked in a solution of urea/water/acetic acid (1 g/1 ml/1 ml) for 20 min and then in this solution containing 0.015 M N-chlorosuccinimide for 1 h. After washing with water, the slice was loaded onto a precast 10 -20% gradient gel (Bio-Rad) and then subjected to electrophoresis and autoradiography. V8 protease mapping was performed exactly as detailed previously (14).
RNA Interference (RNAi)-The RNAi to Mirk (si1) was synthesized and used as described previously (10). Positively transfected cells were sorted by co-transfected GFP using a BD FACSVantage SE cell sorter.
Band Density in Autoradiograms and Western Blots-Band density was measured using a Lacie Silverscanner and Silverscanner III software and analyzed using the IP LabGel program.
Northern Analysis-3.5 g of total RNA from each cell line was electrophoresed in a 1.1% agarose-formaldehyde gel, transferred to nylon membranes by downward capillary transfer, and cross-linked by baking. The membranes were hybridized to a cDNA probe containing the cyclin D1 coding sequence in pCMV5,which had been labeled with 32 P by random priming. The blot was hybridized overnight at 68°C with at least 10 7 cpm of the labeled probe, washed at room temperature twice for 15 min with 1ϫ SSC, 0.1% SDS, and then washed for 15 min at 65°C in 0.2ϫ SSC, 0.1% SDS and autoradiographed.

RESULTS
Cyclin D1 Turnover Enhanced by the G 0 /G 1 Kinase Mirk-Mv1Lu cells were chosen to study the interaction between Mirk and cyclin D1, because Mv1Lu cells are nontransformed epithelial cells that exhibit some normal cell cycle regulation. Mv1Lu cells growth arrest in G 0 /G 1 in response to transforming growth factor-␤1 or to culture in serum-free conditions, whereas treatment with a cellular mitogen releases this arrest (15). In addition, we recently have made a stable Mirk-inducible subline from Mv1Lu cells and have shown that the induction of Mirk inhibited the migration of these cells in wounding experiments and inhibited their invasion through polycarbonate Transwell filters (12). Therefore, we know that Mirk exhibits biological activity in these epithelial cells. We transiently transfected Mirk into subconfluent Mv1Lu cells, allowed the expression of Mirk for 24 h in serum-free medium, and arrested cells in G 0 by continued culture in serum-free medium for an additional 48 h. In both Mirktransfected cultures and vector control transfectants, 78 -79% of cells were in G 0 /G 1 phase (Fig. 1A). An examination of cultures by fluorescence microscopy for co-transfected enhanced GFP showed that ϳ60 -70% of cells were transfected. The rate of entry into S phase was determined by examining cultures by flow cytometry 7 and 12 h after the culture medium was changed to serumcontaining growth medium. Both Mirk transfectants and control transfectants remained in G 1 at 7 h after release (data not shown). However, 66% of vector control transfectants entered S phase 12 h after the addition of serum, whereas only half of that many, 33% of Mirk transfectants, entered the S phase at this time (Fig. 1A). Thus, transient overexpression of Mirk in epithelial cells blocked cells in G 0 /G 1 . In earlier studies, we had made a stable Mirk-inducible subline from nontransformed Mv1Lu lung epithelial cells (12). The induction of Mirk caused a 30 -40% inhibition of Mv1Lu cell entry into S phase as measured by decreased uptake of [ 3 H]thymidine (Fig. 1B), confirming the results of our transient transfection experiments (Fig. 1A). Complete cessation of growth did not occur in the presence of serum mitogens, but increased levels of Mirk inhibited cell cycling.
Transient overexpression of Mirk in Mv1Lu cells decreased cyclin D1 protein levels as shown by Western blotting (Fig. 2A). Similarly, the induction of Mirk protein in the stable Mirkinducible Mv1Lu subline C9 decreased cyclin D1 levels by half, whereas treatment of cells with the proteasome inhibitor MG115 blocked the Mirk-induced reduction in cyclin D1 abundance (Fig. 2B). Translation arrest experiments with cycloheximide confirmed that the induction of Mirk decreased the halflife of cyclin D1 from ϳ40 to 20 min in Mv1Lu mink lung epithelial cells cultured in serum-free conditions (Fig. 2, C and D). The faster migrating cyclin D1 form (Fig. 2C, long arrow) is lost rapidly upon Mirk induction. Thus, Mirk enhanced the normal proteasomal turnover of cyclin D1 in Mv1Lu cells. Mirk is a kinase most active in G 0 /G 1 in NIH3T3 cells when it is predominately localized in the nucleus (9). The major kinase that controls cyclin D1 levels in mammalian cells is GSK3␤ (2, 3), which is localized in the cytoplasm and perinuclear region during G 1 in NIH3T3 cells and enters the nucleus during S phase. Possibly, Mirk maintains cells in G 0 by limiting the abundance of nuclear cyclin D1 in G 0 or, alternatively, functions in G 1 to enhance the later occurring phosphorylation of cyclin D1 by GSK3␤.
Mirk Binds to Cyclin D1-We further explored the interaction between Mirk and cyclin D1 by determining whether they interact in vivo. Nontransformed Mv1Lu mink lung epithelial cells were co-transfected for 24 h with Mirk, FLAG-cyclin D1, or an equal amount of vector control DNA, cultured for 48 h in serum-free medium, and then immunoprecipitated with affinity-purified polyclonal antibody directed to a unique sequence at the N terminus of Mirk (Fig. 3A), affinity-purified polyclonal antibody directed to a unique sequence at the C terminus of Mirk (Fig. 3B), or antibody to the FLAG epitope on the cyclin D1 expression plasmid. The immunoprecipitates were sepa-rated by SDS-PAGE, and the abundance of Mirk and of FLAGcyclin D were determined by Western blotting using the Cterminal directed antibody for Mirk and antibody to the FLAG epitope for cyclin D1. Five percent of the lysates were analyzed by Western blotting (Input panel in A). Antibody directed to the N terminus of Mirk was able to co-immunoprecipitate cyclin D1 with Mirk (Fig. 3A), whereas antibody directed to the C terminus of Mirk could only immunoprecipitate Mirk and not cyclin D1 (Fig. 3B). These results suggest that Mirk interacts with cyclin D1 through Mirk-unique C terminus. Antibody to the FLAG epitope could also co-immunoprecipitate cyclin D1 and Mirk (Fig. 3B, last lane). Thus, Mirk and cyclin D1 interact in vivo within Mv1Lu cells as shown by their co-immunoprecipitation with antibodies directed either to Mirk or to cyclin D1.
Depletion of Mirk by RNAi in Postmitotic C2C12 Myoblasts Increases Cyclin D1 Abundance-We wished to determine the effect of depleting endogenous Mirk on cyclin D1 stability. However, RNA interference experiments could not be performed reliably in Mv1Lu cells because Mirk had not been sequenced in the mink cell. We chose murine C2C12 myoblasts because Mirk mediates their differentiation into postmitotic fused myotubes when cells are placed into differentiation medium (10). The relative abundance of Mirk and cyclin D1 during the initial hours of differentiation was determined first.
Mirk is expressed at very low levels in proliferating myoblasts but is rapidly induced at least 10-fold when primary-cultured muscle satellite cells or myoblast cell lines are placed in differentiation medium (10). The induction of Mirk in C2C12 myoblasts was accompanied by a sharp decrease in cyclin D1 levels detected within 6 h (Fig. 4A).
Cyclin D1 mRNA and protein decline rapidly during myogen- Cycloheximide (CH) was added at 40 g/ml in serum-free medium to arrest translation, and the levels of Mirk and cyclin D1 were determined at the times indicated by Western blotting. D, the data from panel C and the data from two additional experiments were averaged and plotted. CT, control-untreated cells. Mean values are shown. The mean Ϯ S.E. of repeat measurements was 6% (not shown). Data were subjected to exponential curve fitting, and the y intercepts and r 2 values were calculated. esis (16). Although no cyclin D1 could be detected in the nuclei of fused myotubes or single myoblasts maintained in mitogen-free differentiation medium, the induction of the mutant form of cyclin D1-T286A, which is resistant to GSK3␤ phosphorylation, led to the accumulation of cyclin D1/CDK4 complexes within the nucleus (17). These studies indicated that the decline in cyclin D1 levels during myogenesis is caused by phosphorylation by GSK3␤ and transport to the cytoplasm for proteasomal degradation. We questioned whether Mirk played any role in the modulation of cyclin D1 levels in differentiating myoblasts. During the initial 6 h in differentiation medium, there was no change in the total amount of GSK3␤ ( Fig. 4A) but there was a small increase in the fraction of GSK3␤, which was phosphorylated at serine 9, a residue phosphorylated when GSK3␤ is inactivated by Akt (18). During myoblast differentiation, Akt is activated and mediates cell survival (19), suggesting that Akt may continue to inhibit GSK3␤ during myoblast differentiation and that the increasing levels of Mirk in differentiating myoblasts may play a role in regulating cyclin D1 levels.
We showed previously that depletion of Mirk in differentiating myoblasts by RNA interference prevented the induction of the muscle regulatory factor myogenin and the consequent differentiation program including fusion into myotubes (10). We tested whether down-regulating Mirk levels would stabilize cyclin D1 in C2C12 cells. In fact, depletion of Mirk protein levels to 3% of control levels by RNAi led to a substantial 5-fold increase in cyclin D1 protein levels (Fig. 4B). The effect of Mirk on cyclin D1 was predominately posttranscriptional as cyclin D1 mRNA levels increased only a modest 20% in Mirk-depleted cells (Fig. 4B, similar microarray data not shown). These data strongly suggest that Mirk mediates the rapid turnover of cyclin D1 in differentiating myoblasts. Depletion of Mirk in C2C12 cells also decreased the amount of [ 32 P]orthophosphate incorporated into immunoprecipitated cyclin D1 (Fig. 4C). These data suggest that in differentiating C2C12 myoblasts Mirk may destabilize cyclin D1 by phosphorylating it.
Mirk Phosphorylates Cyclin D1 in Vitro and in Vivo in Mv1Lu Epithelial Cells-We next considered the possibility that Mirk worked directly on cyclin D1. Mirk did not phosphorylate the cyclin D1 kinase GSK3␤ (data not shown). Purified GST-cyclin D1 was phosphorylated directly by Mirk in vitro (Fig. 5A) with kinase-inactive YF-Mirk serving as the control. Mirk is expressed at very low levels in proliferating myoblasts and is only induced when C2C12 cells begin to differentiate terminally (Fig. 4A) (10). Because of the lack of Mirk in cycling myoblasts, for the subsequent studies, we returned to Mv1Lu epithelial cells.
To determine whether Mirk phosphorylated cyclin D1 in vivo, Mv1Lu cells were co-transfected for 24 h with FLAGcyclin D1 and either wild-type Mirk or vector control and then metabolically labeled for 5 h with [ 32 P]orthophosphate. Cyclin D1 was immunoprecipitated from cell lysates with anti-FLAG antibody and then analyzed by autoradiography and Western blotting after SDS-PAGE. Co-expression of wild-type Mirk increased in vivo phosphorylation of cyclin D1 (Fig. 5B). Thus, Mirk directly phosphorylated cyclin D1 in vitro and co-expression of Mirk increased the phosphorylation of cyclin D1 in vivo.
The region within cyclin D1, which contained the Mirk phosphorylation sites, was then determined by peptide mapping after in vitro phosphorylation. The GST epitope tag was cleaved from GST-cyclin D1 by the PreScission protease before assay. After the in vitro phosphorylation by Mirk, cyclin D1 was cleaved by N-chlorosuccinimide. The resulting peptide fragments were separated by 10 -20% gradient SDS-PAGE and visualized by autoradiography. The largest phosphopeptide was 26 kDa, which encompassed the C terminus of cyclin D1, and was the most strongly labeled of the peptides (Fig. 5C). The 17-kDa peptide contained the N-terminal half of the molecule and was less strongly labeled. The smallest N-terminal fragment was 8 kDa and was not phosphorylated in vitro. These results indicated that Mirk must phosphorylate cyclin D1 at a site or sites within the C terminus or the central region.
Mirk Destabilizes Cyclin D1 by Phosphorylation at Threonine 288 -Potential Mirk phosphorylation sites at Ser 90 , Thr 105 , Thr 128 , Thr 184 , Ser 197 , Ser 219 , Ser 257 , and Thr 288 were mutated to alanine. These sites were within the 26-kDa cyclin D1 fragment phosphorylated by Mirk and were conserved between mouse and human. The closely related kinase to Mirk, Dyrk1A, has been reported to phosphorylate serines or threonines ϩ3 from an arginine residue (20). In addition, we have found that Mirk phosphorylates hepatocyte nuclear factor 1␣ at serine 247, which is ϩ3 from an arginine, R 243 GVS* 247 PS (asterisk indicates phosphorylation site) (8). However, Mirk has phosphorylated other substrates at serines or threonines ϩ5 to Ϫ5 from an arginine (data not shown), so we mutated a series of serines and threonines within ϩ7 to Ϫ9 of an arginine residue and, in one case, ϩ3 to a lysine. Initial in vitro experiments suggested that Mirk phosphorylated cyclin D1 at Thr 184 , Thr 288 , or both residues (data not shown). However, the activity of kinases can be very different in vitro and in vivo (21), so these experiments were taken only as guides for subsequent studies.
The in vitro phosphorylation studies were confirmed with in vivo expression studies. Co-expression of Mirk and FLAG  (panels A and B), affinity-purified polyclonal antibody directed to a unique sequence at the N terminus of Mirk (panel A), or affinity-purified polyclonal antibody directed to a unique sequence at the C terminus of Mirk (panel B). The immunoprecipitates were separated by SDS-PAGE, and the abundance of Mirk was determined by Western blotting (WB) using affinity-purified polyclonal antibody directed to a unique sequence at the C terminus of Mirk or antibody to the FLAG epitope to detect cyclin D1. Five percent of the lysates were analyzed by Western blotting as above (Input panel in A).
epitope-tagged cyclin D1-T288A, T184A, or wild-type cyclin D1 in Mv1Lu cells was followed by in vivo labeling with [ 32 P]orthophosphate, and an analysis of the cyclin D1 immunoprecipitates by autoradiography and Western blotting after SDS-PAGE. In cyclin D1, the most effective mutation in blocking in vivo phosphorylation in the presence of co-expressed Mirk was T288A (Fig.  6A), whereas the T184A mutation caused no diminution of phos-phorylation. Threonine 288 was Ϫ3 from an arginine residue within the sequence LACTPT 288 *DVR 291 DVDI (asterisk indicates the phosphorylation site).
Cyclin D1 migrates in two forms on SDS-PAGE, both of which are more highly phosphorylated in the presence of coexpressed Mirk (Fig. 6A, compare lanes 1 and 2). However, when cyclin D1 is mutated at T288A, only the upper band is phosphorylated (Fig. 6A). Thus, Mirk phosphorylates cyclin D1 in vivo at Thr 288 , yielding the faster migrating cyclin D1 species. It is of interest to recall that the faster-migrating cyclin D1 band in the half-life experiments was the form of cyclin D1 most rapidly decreased when Mirk was induced (Fig. 2C, long  arrow). These data are consistent with a model in which Mirk phosphorylates Thr 288 to produce the faster migrating cyclin D1 species.
The cyclin D1 double band was detected by Western blotting in several experiments in both Mv1Lu and C2C12 cells (Figs. 2, A and C, and 4A) and by immunoprecipitation after phosphorylation in vivo in Mv1Lu cells (Fig. 6A). We wanted to confirm that both bands were forms of cyclin D1, just posttranslationally modified in different ways. The phosphorylated double band in wild-type cyclin D1 and in cyclin D1 mutated at T184A was compared with the phosphorylated single band of cyclin D1 mutated at T288A by V8 protease peptide mapping (Fig. 6B). Digestion of all three cyclin D1 constructs yielded a similar phospho-peptide pattern following V8 cleavage, demonstrating that both the upper and lower bands were cyclin D1.
Translation arrest experiments with cycloheximide (Fig. 6C) demonstrated that a cyclin D1-T288A mutant construct was more stable than a wild-type construct. The turnover of Mirk was measured in both cultures as an internal control and showed similar stability when co-expressed with either wild-type cyclin D1 or the mutant T288A form. Thus, phosphorylation of cyclin D1 on Thr 288 by Mirk destabilizes cyclin D1 in vivo.
Mirk Phosphorylates Cyclin D1 Bound to GSK3␤ and Can Phosphorylate Cyclin D1 When GSK3␤ Is Inhibited by LiCl-The major kinase that controls cyclin D1 levels in mammalian cells is GSK3␤. Cyclin D1 is known to be rapidly degraded through the ubiquitin-proteasome pathway following phosphorylation by GSK3␤ at threonine 286, which facilitates the export of cyclin D1 to the cytoplasm where it is proteolyzed (2,3). Mirk is most abundant and most active as a kinase in G 0 /G 1 (9), so we hypothesized that Mirk functions in G 1 to enhance the later occurring phosphorylation of cyclin D1 by GSK3␤. Mirk and GSK3␤ phosphorylated cyclin D1 in an additive fashion when tested in in vitro kinase assays. The activity of recombinant Mirk and recombinant GSK3␤ on GST-cyclin D1 was measured with and without treatment with the GSK3␤ inhibitor LiCl. When added together, GSK3␤ and Mirk phosphorylated cyclin D1 twice as much as either Mirk or GSK3␤ alone (Fig. 7A). In addition to the expected inhibition of GSK3␤ (25-fold inhibition), LiCl also blocked Mirk phosphorylation of cyclin D1 (3-fold inhibition). Mirk and GSK3␤ exhibit some homology, so this inhibition is not surprising. In the presence of LiCl, only the phosphorylation of cyclin D1 by Mirk was seen (note similar activities in Fig. 7A, lanes 4 and 6). Therefore, Mirk can phosphorylate cyclin D1 when GSK3␤ is inhibited.
To compare the relative effects of Mirk and GSK3␤ in vivo, Mv1Lu cells were co-transfected for 24 h with wild-type Mirk, kinase-inactive YF-Mirk, or vector control together with wildtype FLAG-cyclin D1, mutant FLAG-cyclin D1-T184A used as a control, or mutant FLAG-cyclin D1-T288A. GSK3␤ activity was inhibited in some cultures by LiCl, and cells were then metabolically labeled with [ 32 P]orthophosphate. Cyclin D1 was immunoprecipitated from cell lysates with anti-FLAG antibody and then analyzed by autoradiography and Western blotting FIG. 4. Increased Mirk expression in differentiating C2C12 myoblasts is associated with reduced cyclin D1 levels, whereas depletion of Mirk by RNA interference increases cyclin D1 protein levels and decreases phosphorylation of cyclin D1. A, C2C12 cells were cultured in growth medium containing 20% fetal bovine serum overnight and then switched to serum-free Dulbecco's modified Eagle's medium for 1-6 h to begin differentiation to postmitotic myotubes. Cell lysates were examined by Western blotting (WB) for Mirk, cyclin D1, GSK3␤, and phosphorylated GSK3␤. B, RNAi to Mirk increases cyclin D1 abundance by a posttranscriptional mechanism. C2C12 cells were co-transfected with an expression plasmid for Mirk RNAi and enhanced GFP, or vector DNA and enhanced GFP selected by cell sorting for enhanced GFP were placed in growth medium for 1 day and then switched to differentiation medium for 2 days. Vc, vector. Left panel, cell lysates were examined by Northern blotting for cyclin D1. 28 and 18 S rRNA staining by ethidium bromide is shown. The ratio of the abundance of cyclin D mRNA to the abundance of the 28 S rRNA is given below the appropriate lanes. Right panel, cell lysates were examined by Western blotting for Mirk and cyclin D1. The abundance of a cross-reactive protein was similar in both lanes and served as an internal control (not shown). The ratio of Mirk protein and cyclin D1 protein in control and Mirk-directed RNAi-treated cultures is given to the right of the appropriate lanes. C, C2C12 cells were transfected for 24 h with either pSilencer plasmid encoding si1 to Mirk or control sequences and then switched to differentiation medium for 5 h containing [ 32 P]orthophosphate before cell lysis. CT, control-untreated cells. Cyclin D1 immunoprecipitates were analyzed by autoradiography and then Western blotting for cyclin D1. The ratio of labeled cyclin D1 to total immunoprecipitated (IP) cyclin D1 is listed under the appropriate lanes.
after SDS-PAGE (Fig. 7B). Co-expression of wild-type Mirk increased the phosphorylation of both wild-type cyclin D1 and mutant cyclin D1-T184A by 40%, whereas phosphorylation was decreased by 80% when cyclin D1 was mutated at the Mirk phosphorylation site of Thr 288 . A loss of the lower cyclin D1 phosphorylated band was seen in the T288A mutant (long arrow in Fig. 7B, similar to the results in Fig. 6A). LiCl treatment decreased the total amount of [ 32 P]orthophosphate incorporated into cyclin D1, but both bands remained phosphorylated, indicating that Mirk could phosphorylate cyclin D1 when GSK3␤ was inhibited (Fig. 7B, lanes 6 and 7). Moreover, only the mutation of cyclin D1 to Ala 288 , not LiCl treatment, caused a loss of phosphorylation of the lower band. The phosphorylation of cyclin D1 was decreased to a background level 20 -30% of control values by mutation to T288A both in the absence and in the presence of LiCl (Fig. 7B, compare lanes 5 and 8). We conclude that, similar to their in vitro interaction, Mirk was capable of phosphorylating cyclin D1 at Thr 288 in vivo, even when GSK3␤ was inhibited. These data suggested that Mirk might phosphorylate cyclin D1 before GSK3␤ and that blocking Mirk action on cyclin D1 by mutation at Thr 288 could dramatically inhibit in vivo phosphorylation of cyclin D1 at other positions, possibly including the GSK3␤ site of Thr 286 (compare lane 5 with 3).
We further explored the relationship between GSK3␤ and Mirk by measuring the kinase capacity of Mirk on immunoprecipitated GSK3␤ and its associated proteins (Fig. 7C). Mirk had little or no detectable kinase activity on immunoprecipitated GSK3␤ (Fig. 7C). However, a highly phosphorylated lower molecular weight band was observed in the GSK3␤ immunoprecipitates (Fig. 7C, upper blot), which co-migrated with cyclin D1 immunoprecipitated from the same lysates and run in parallel. To confirm the identity of this band, a portion of the immunoprecipitate, which had not been analyzed by SDS-PAGE, was re-precipitated this time with anti-cyclin D1 conjugated to agarose beads. Phosphorylated cyclin D1 was detected after SDS-PAGE and autoradiography (Fig. 7C, lower blot). Therefore, Mirk can phosphorylate cyclin D1 bound to GSK3␤.
Mirk Phosphorylates Cyclin D1, Which Is Mutant at the GSK3␤ Site-We next tested whether Mirk could phosphorylate cyclin D1 independently of GSK3␤ by co-expressing Mirk with FLAG-cyclin D1-T286A (mutated at the GSK3␤ site), FLAG-cyclin D1-T288A (mutated at the Mirk site), or wild-type FLAG-cyclin D1 in both Mv1Lu lung epithelial cells and in NIH3T3 fibroblasts (Fig. 8). Following metabolic labeling with [ 32 P]orthophosphate, the cyclin D1 constructs were immunoprecipitated by antibody directed to their FLAG epitope tag and their abundance was detected by Western blotting. There   FIG. 5. Mirk phosphorylates cyclin D1 in vitro and in vivo. A, GST-cyclin D1 was phosphorylated by Mirk in vitro. Kinase-inactive YF-Mirk served as the control. The in vitro kinase reaction mixtures were analyzed by SDS-PAGE and autoradiography, and the phosphorylated cyclin D1 band is shown (autorad). The total amount of purified cyclin D1 on the gel is shown by Ponceau S staining (Cyc D1). B, Mv1Lu cells were co-transfected for 24 h with FLAG-cyclin D1 and either wild-type (WT) Mirk or vector control and then labeled for 5 h with [ 32 P]orthophosphate in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. After lysis in buffer containing 0.5% Nonidet P-40, cyclin D1 was immunoprecipitated with anti-FLAG antibody and then analyzed by autoradiography and Western blotting (WB) for the FLAG epitope after SDS-PAGE. C, the phosphorylation sites in cyclin D1 were peptide-mapped after in vitro phosphorylation. The GST epitope tag was cleaved from GST-cyclin D1 by the PreScission protease before assay. After the in vitro kinase assay, the phosphorylated cyclin D1 was cleaved by Nchlorosuccinimide at Trp 63 and Trp 150 and the resulting peptide fragments were separated by 16 -25% gradient SDS-PAGE and visualized by autoradiography. The sizes of cyclin D1 (36 kDa) and its derivative phosphopeptides of 26, 17, 16, and 10 kDa are indicated. IP, immunoprecipitation.
FIG. 6. Mirk phosphorylates cyclin D1 at Thr 288 in vivo. A, Mv1Lu cells were co-transfected for 24 h with Mirk and FLAG-cyclin D1 wild type (WT), FLAG-cyclin D1-T184A, FLAG-cyclin D1-T288A, or FLAG vector (Vc) and then labeled for 5 h with [ 32 P]orthophosphate. After lysis in buffer containing 0.5% Nonidet P-40, cyclin D1 was immunoprecipitated (IP) with anti-FLAG antibody and then analyzed by autoradiography and Western blotting (WB) after SDS-PAGE. One of duplicate experiments with similar results is shown. B, gel slices containing the FLAG-cyclin D1 (WT), FLAG-cyclin D1-T184A, and FLAG-cyclin D1-T288A phosphate-labeled immunoprecipitates in panel A were digested with V8 protease as described under "Experimental Procedures," and then the peptides were separated by SDS-PAGE and visualized by autoradiography. Arrows indicate the major V8 protease-generated peptides. One of three studies with similar results is shown. C, Mv1Lu cells were co-transfected for 24 h with Mirk and either FLAG-cyclin D1 wild type or FLAG-cyclin D1-T288A. Cycloheximide (CH) was added at 40 g/ml in serum-free medium to arrest translation, and the levels of Mirk and cyclin D1 were determined at the times indicated by Western blotting. Cross-reacting (CR) protein is a stable cellular protein cross-reacting with the FLAG epitope antibody. One of duplicate experiments with similar results is shown. was a 30% increase in the abundance of cyclin D1-T286A compared with wild-type cyclin D1, probably because of the greater stability of the mutant form. The enrichment was predominately in the slower migrating cyclin D1 form and most clearly seen in NIH3T3 cells (Fig. 8B, small arrow). Mirk phosphorylation of cyclin D1 was not blocked by the T286A mutation. In fact, cyclin D1-T286A was phosphorylated twice as much as wild-type cyclin D1 in both cell types. Phosphorylation of cyclin D1 by co-expressed Mirk was inhibited by the T288A mutation as seen before with the loss of phosphorylation of the faster migrating band (autoradiography panels in Fig. 8, long arrows). These data show that the phosphorylation of cyclin D1 by GSK3␤ is not a precondition for phosphorylation by Mirk. We speculate that after Mirk phosphorylates cyclin D1 at Thr 288 , there is another modification of this phosphorylated form, which confirms its faster mobility. Because the faster migrating form of cyclin D1 is rapidly lost in translation arrest experiments (see Fig. 2C) and mutation to T286A stabilizes the slower form (Fig. 8), we speculate that both Mirk and GSK3␤ phosphorylate the slower migrating form. In this model, cyclin D1 phosphorylated at both 286 and 288 might then be rapidly modified to the faster migrating form and then proteolyzed. DISCUSSION Mirk/dyrk1B is a member of the Minibrain/dyrk family of arginine-directed protein kinases. Mirk, similar to Dyrk1A, functions as a transcriptional activator (8,22,23). We have shown in the current study that Mirk also functions by mediating protein degradation. Mirk phosphorylates cyclin D1 at Thr 288 , which is close to the GSK3␤ phosphorylation site of Thr 286 that is known to mediate cyclin D1 ubiquitination and degradation. Furthermore, we have shown that depletion of endogenous Mirk by RNA interference increases the abundance of cyclin D1 by a posttranscriptional mechanism and that up-regulation of Mirk levels leads to a faster turnover of cyclin D1 and decreases the rate of entry into the S phase. These effects were seen in two nontransformed cell types, Mv1Lu lung epithelial cells and C2C12 myoblasts. A related Dyrk family kinase, MBK-2, has been shown to coordinate the degradation of several maternal proteins, which is essential for Caenorhabditis elegans zygotes to complete cytokinesis (24). Depletion of MBK-2 by RNA interference arrested the development at the one-cell stage with multiple nuclei. The mechanism of MBK-2 action is not known, but MBK-2 was not a general activator of protein degradation by the proteasome. MBK-2 was postulated to target some unknown factor but not an E3 ubiquitin ligase common to the specific group of maternal proteins, which were degraded (24).
In an earlier study, we had observed that stable overexpression of Mirk in colon carcinoma cells enhanced the turnover of cyclin D1 and the CDK inhibitor p27kip1 (25). These results appeared contradictory and may have resulted from the changes in cell physiology that occur in the presence of elevated expression of an active kinase with many potential cellular targets. However, the present study has confirmed the role of Mirk in inducing rapid turnover of cyclin D1 by three methods: RNA interference; transient overexpression of Mirk; and induction of a stable inducible Mirk construct. Moreover, we have identified threonine 288 in cyclin D1 as a specific substrate for Mirk kinase and have shown that the nonphosphorylatable mutant construct T288A is more stable in vivo than wild-type cyclin D1.
Cyclin D1 was phosphorylated almost exclusively on threonine in vivo in NIH3T3 cells or in Sf9 insect cells co-expressing FLAG-cyclin D1 and CDK4, whereas the mutation of cyclin D1 to T286A blocked the phosphorylation of a FLAG-cyclin D1 construct in both cell types (2). These data would appear to rule out any role for endogenous Mirk in the phosphorylation of cyclin D1 at Thr 288 . However, we have found that depletion of Mirk by RNA interference in C2C12 myoblasts stabilized cyclin D1 protein without affecting cyclin D1 mRNA (Fig. 4). Furthermore, induced overexpression of Mirk in Mv1Lu lung epithelial cells led to the phosphorylation of cyclin D1 at Thr 288 (Fig. 6) and a more rapid turnover of cyclin D1 (Fig. 2), all of which argue that Mirk can modulate cyclin D1 stability, at least in lung epithelial cells and in myoblasts. In this study, CDK4 was not co-expressed with exogenous cyclin D1, leaving most of the overexpressed cyclin D1 as a free monomer, possibly presenting a different substrate to cellular kinases. GSK3␤ is a more efficient kinase for cyclin D1 when cyclin D1 is bound to CDK4 (3). Thus, in our in vivo studies, the ectopic cyclin D1 would largely be in free form not complexed to CDK4, so we may have amplified the kinase effect of Mirk while decreasing the relative effect of GSK3␤.
Another possibility is that Mirk and GSK3␤ may function together, so elimination of the Mirk site may block the function of GSK3␤. In support of this hypothesis, the Mirk related kinase Dyrk1A has been hypothesized to act as a "primer" kinase for GSK3. Dyrk1A phosphorylates eIF2B⑀ at Ser 539 , greatly increasing the ability of GSK3 to phosphorylate the nearby residue Ser 535 (26). Purified recombinant Mirk and GSK3␤ phosphorylated cyclin D1 in an additive fashion in in vitro kinase assays in this study, and blocking GSK3␤ activity in vivo and in vitro with LiCl did not prevent Mirk from phosphorylating cyclin D1. Supporting this interpretation is the finding that Mirk was capable of phosphorylating cyclin D1 mutated at the GSK3␤ site of Thr 286 (Fig. 8), which Mirk phosphorylated more avidly than wild-type cyclin D1 in vivo. Both Mirk/dyrk1B and the Mirk-related kinase Dyrk1A phosphorylate the site 3a (Ser 640 ) in glycogen synthase, a site also phosphorylated by GSK-3, showing some overlap in function between the two kinase families (27).
The Mirk phosphorylation site of Thr 288 is a regulatory site for cyclin D1 ubiquitination. GSK3␤ initiates ubiquitination and nuclear export of cyclin D1 by phosphorylation at the nearby site, threonine 286 (2, 3). Other investigators have shown that both cyclin D1 residues, threonine 286 and threo-nine 288, are important for its ubiquitination (28). Much less ubiquitination of the cyclin D1-T286A/T288A construct was seen compared with wild-type cyclin D1, and the half-life of cyclin D1-T286A/T288A was three times longer than the halflife of wild-type cyclin D1 (28), consistent with the conclusions of the current study. Depletion of endogenous Mirk by RNA interference led to a 5-fold increase in cyclin D1 protein levels by a posttranscriptional mechanism (Fig. 4B), so we conclude, at least in the cell types we have tested, that phosphorylation by Mirk can increase the turnover of cyclin D1. In recent studies, we have shown that Mirk is a G 0 kinase, which aids cells in maintaining a G 0 arrest by stabilizing the CDK inhibitor p27 by phosphorylation at Ser 10 (9). This study demonstrates that Mirk phosphorylates the cell cycle regulator, cyclin D1, at Thr 288 , which enhances its rapid turnover. Thus, Mirk has the novel function of both stabilizing a CDK inhibitor and destabilizing a G 1 cyclin to assist cells in remaining arrested in G 0 .