Mitosis-specific phosphorylation and subcellular redistribution of the RIIalpha regulatory subunit of cAMP-dependent protein kinase.

Phosphorylation of the RII regulatory subunits of cyclic AMP-dependent protein kinases (PKAs) was examined during the HeLa cell cycle. Three RIIalpha isoforms of 51, 54, and 57 kDa were identified by RIIalpha immunodetection and labeling with 8-azido[32P]cAMP in different cell cycle phases. These isoforms were characterized as different phosphorylation states by the use of selective PKA and cyclin-directed kinase inhibitors. Whereas RIIalpha autophosphorylation by PKA caused RIIalpha to shift from 51 to 54 kDa, phosphorylation of RIIalpha by one other or a combination of several kinases activated during mitosis caused RIIalpha to shift from 51 to 57 kDa. In vivo incorporation of [32P]orthophosphate into mitotic cells and RIIalpha immunoprecipitation demonstrated that RIIalpha was hyperphosphorylated on a different site than the one phosphorylated by PKA. Deletion and mutation analysis demonstrated that the cyclin B-p34(cdc2) kinase (CDK1) phosphorylated human recombinant RIIalpha in vitro on Thr54. Whereas RIIalpha was associated with the Golgi-centrosomal region during interphase, it was dissociated from its centrosomal localization at metaphase-anaphase transition. Furthermore, particulate RIIalpha from HeLa cell extracts was solubilized following incubation with CDK1 in vitro. Our results suggest that at the onset of mitosis, CDK1 phosphorylates RIIalpha, and this may alter its subcellular localization.

Cyclic AMP-dependent protein kinases (PKAs) 1 are present in mammalian tissues as two major isozymes, type I and type II (for reviews, see Refs 1 and 2). The inactive holoenzyme is composed of a regulatory subunit dimer that binds two catalytic subunits. Binding of cAMP to the regulatory subunits results in dissociation of the catalytic subunits, which are then catalytically active and can also be translocated to the nucleus.
Four different regulatory subunits (RI␣, RI␤, RII␣, RII␤) and three different catalytic subunits (C␣, C␤, C␥) have been identified as separate gene products. The regulatory subunits RI␣ and RII␣ are present in almost all cell types, whereas the expression of RI␤ and RII␤ is tissue-specific. RII isoforms are associated with A kinase-anchoring proteins (AKAPs) (3) that target PKA type II to organelles, cytoskeleton, and membranes (3)(4)(5). Intracellular accumulation of RII has been observed in association with centrosome or the Golgi complex in several mammalian cell types (6 -8).
The activities and subcellular localization of protein kinases and phosphatases are known to be regulated by phosphorylation (for a review, see Ref. 9). A major difference between RI and RII is the ability of RII to be autophosphorylated by the catalytic subunit of PKA (10). Phosphorylation sites for casein kinase II, glycogen synthase kinases 3 and 5, and prolinedirected protein kinase have been identified in vitro on bovine RII␣ as well (11)(12)(13). More recently it was shown that bovine RII␤ can also be phosphorylated at Thr 69 in vitro by CDK1, the mitotic kinase p34 cdc2 associated with cyclinB (14). The physiological role of these phosphorylations is not fully understood. In vitro autophosphorylation of PKA type II decreases the affinity between RII and the catalytic subunit, and CDK1 phosphorylation of RII␤ decreases its ability to anchor to MAP-2 (a neuronal microtubule-associated protein (14)). However, RII␤ is expressed primarily in neuro-endocrine cells and is thus present in differentiated cells that divide slowly or not at all. For these reasons, we examined the cell cycle-dependent phosphorylation of the ubiquitously expressed RII␣. In the present study, we show that in dividing cells, the regulatory subunit RII␣ demonstrates different phosphorylation patterns during the cell cycle. In mitotic-arrested cells, RII␣-labeling with 8-azido[ 32 P]cAMP, immunostaining with a specific antibody, and in situ incorporation of [ 32 P]orthophosphate in HeLa cells revealed different states of RII␣ phosphorylation compared with that observed during interphase. Inhibition of PKA activity with a selective inhibitor, H89, or inhibition of mitotic kinase activity with olomoucine showed that RII␣ is mainly phosphorylated by the catalytic subunit during interphase and by a mitotic kinase on another phosphorylation site(s) during mitosis. Moreover in vivo incorporation of [ 32 P]orthophosphate into mitotic and interphase HeLa cells followed by RII␣ immunoprecipitation demonstrated that in mitosis RII␣ was hyperphosphorylated on at least two sites. Recombinant native human RII␣ was phosphorylated in vitro by purified CDK1 on a site located in the N-terminal domain of this protein. Sitespecific mutagenesis of the putative Thr 54 phosphorylation site demonstrated that this site was the target of CDK1. During mitosis, RII␣ was dissociated from centrosome at the met-aphase-anaphase transition concomitantly with its phosphorylation by PKA and CDK1.

MATERIALS AND METHODS
Cell Culture, Synchronization, and 32 P Incorporation-HeLa cells were grown as monolayer cultures at 37°C in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum. They were arrested or slowed down in their cell cycle by different techniques previously described (15). Briefly, cells enriched in late G 1 were obtained by a 2.5 mM thymidine block for 17 h, followed by 7 h of release from this block and a subsequent block with 400 mM L-mimosine for 16 h. For early S phase arrest, subconfluent cultures were arrested in G 0 by serum deprivation for 48 h followed by 5 g/ml of aphidicolin for 24 h. A double-thymidine block followed by 100 M olomoucine treatment for 6 h gave a mixed population of cells arrested in G 1 and in G 2 (16). HeLa cells were synchronized in mitosis by a double thymidine block followed by 1 M nocodazole treatment for 18 h. The cell cycle state was analyzed by flow cytometry. Human osteosarcoma (SaOS-2) and human skin fibroblast (HS27) cell lines were grown as monolayer cultures on glass slides for immunocytochemistry in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.
For analysis of RII␣ phosphorylation, kinase inhibitors were incorporated into certain experiments. Cells in interphase (essentially in G 1 /S phases) were treated for 6 h with 20 M protein kinase A inhibitor H89 (17). Cells entering mitosis were accumulated by treatment with 1 M nocodazole for 18 h with additional 20 M H89 during the last 6 h of this period. Olomoucine (100 M), a cyclin-directed kinase inhibitor, was added 2 h before the end of mitotic synchronization to avoid inhibition of entry into mitosis.
For 32 P-labeling, HeLa cells arrested either in interphase at the G 1 /S border or in mitosis were incubated for 1 h in phosphate-free Dulbecco's modified Eagle's medium containing 10% dialyzed fetal calf serum, then 1 mCi/ml carrier free [ 32 P]orthophosphate (Amersham Pharmacia Biotech) was added for an additional 2 h at 37°C.
Cell Fractionation and Cdc2 Kinase Assay-Both interphase (G 1 , G 1 /S, and S) and mitotic cells were washed twice in 80 mM sodium ␤-glycerophosphate, 20 mM EGTA, 15 mM MgCl 2 , 100 mM sucrose, 1 mM dithiothreitol, pH 7.2 (EBS buffer). They were centrifuged at 800 ϫ g for 5 min, and the cell pellet was resuspended in EBS buffer containing 2 mM ATP, protease inhibitors (1 mM phenylmethylsulfonyl fluoride; 20 g/ml each leupeptin, pepstatin, and chymostatin; and 1 g/ml aprotinin) and phosphatase inhibitors (1 M okadaic acid and 0.1 mM sodium orthovanadate). Cells were homogenized with a Dounce homogenizer, and the homogenates were centrifuged at 800 ϫ g for 10 min to discard unbroken cells and nuclei. The post-nuclear fractions were then centrifuged at 400,000 ϫ g for 30 min in a TL100 Beckman centrifuge to obtain cytosol and particulate fractions.
Cell lysates were obtained as described above in the presence of 1% Nonidet-P40 in EBS buffer. After homogenization, lysates were centrifuged at 100,000 ϫ g for 30 min, and the supernatant was used for determination of cyclin-directed kinase activity as described below. Protein content of each fraction was quantified according to Bradford (18).
The activity of mitotic cyclin-directed kinase CDK1 was measured in cell lysates with [␥-32 P]ATP and the BIOTRAK Cdc2 kinase enzyme assay system (Amersham Pharmacia Biotech). This system is based on a highly specific substrate (SV40 large T antigen) for the Cdc2 kinase (19).
Photoaffinity-labeling of the Regulatory Subunits with 8-Azido-[ 32 P]cAMP-Photoaffinity incorporation of 8-azido[ 32 P]cAMP was performed as described (20). Subsequently proteins were resolved by SDS-PAGE on 6 -15% polyacrylamide gradient gels. For all cytosols prepared from synchronized cells, the incorporation of 8-azido[ 32 P]cAMP was linear with regard to the amount of protein used up to 150 g.
Antibodies-A monoclonal antibody (clone 4D7) raised against purified human testis RI␣ (21) was used as purified IgG 2a from mouse ascites at 1/1000 dilution for Western blotting. Polyclonal antibody against human RII␣ was raised against a specific peptide from the N-terminal end of the human RII␣ amino acid sequence (21), affinitypurified, and used at 1 g/ml dilution for Western blots and 10 g/ml for immunoprecipitation. Antisera raised against rat heart RII␣ (7) was used at 1/30 dilution for immunoprecipitation. A monoclonal antibody CTR 453 (IgG 2b ) obtained from a library of monoclonal antibodies raised against centrosomes isolated from human lymphoblasts was previously characterized as specific for the centrosome(22) and used for immunocytochemistry as purified immunoglobulins at 140 ng/ml dilution.
Electrophoresis and Immunoblotting-The cytosolic proteins solubilized in sample buffer were separated on 6 -15% polyacrylamide gradient gels and electrophoretically transferred to nitrocellulose filters as described previously (23). The blots were incubated with the appropriate primary antibodies and then with anti-rabbit immunoglobulins conjugated to alkaline phosphatase (for hRII␣ Ab) or with anti-mouse IgG horseradish peroxidase-linked antibody (for hRI␣ Ab).
Immunoprecipitation-Metabolically labeled HeLa cells in interphase and in mitosis were washed twice with PBS and lysed in 1 ml of modified radioimmune precipitation buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 1% Nonidet-P-40, 0.5% sodium deoxycholate, 0.1% SDS) containing protease inhibitors (as described above for cell homogenization) and phosphatase inhibitors (10 mM sodium orthovanadate, 1 M okadaic acid, and 50 mM sodium ␤-glycerophosphate). Immunoprecipitation of RII␣ from metabolically 32 P-labeled cells was performed as described previously (23). The radioactive polypeptide bands were cut out from the gel, and limited digestion of RII was carried out using trypsin-EDTA (Boehringer Mannheim) as described by Takio et al. (24). The reaction was stopped after 5 min at room temperature by the addition of phenylmethylsulfonyl fluoride, and the peptides were separated by SDS-PAGE on a 1-mm-thick slab gel consisting of a 4% staking gel and a 15% resolving gel. Labeled proteolytic products were visualized by autoradiography of the dried gel using Kodak BIOMAX films.
Heterologous Expression of Recombinant Human RII␣ and Site-specific Mutagenesis-Heterologous expression, purification, and characterization of the human regulatory subunit RII␣ of PKA will be described elsewhere. 2 Briefly, a fragment of the human RII␣ cDNA sequence was amplified by polymerase chain reaction to generate a full-length open reading frame with convenient cloning sites, subcloned into the expression vector pGEX-KG, expressed as a fusion protein with GST in the Escherichia coli strain BL21/DE3, and purified from bacterial lysates by absorption to glutathione-agarose beads as described elsewhere (21,25). Subsequently, purified fusion protein was digested with thrombin, and GST was absorbed on glutathione-agarose beads to obtain soluble RII␣ protein that contained an extra two-amino acid N-terminal extension from the linker segment/thrombin cleavage site of GST.
Mutants RII␣-Thr 54 -Ala and RII␣-Thr 54 -Glu were made by polymerase chain reaction of two overlapping amplification products, both with the Thr 54 -Ala or Thr 54 -Glu mutation intoduced. This was accomplished using RII␣ wild type cDNA (26) as a template and primers covering nucleotides 189 -208 and 335-350 for the 5Ј amplification product and primers covering nucleotides 336 -351 and 1523-1542 for the 3Ј amplification product. Subsequently, the overlapping products were mixed, denatured to facilitate annealing of the overlapping part with mutations, filled with Klenow, and subjected to a second round of polymerase chain reaction as described above using only the outside primers. Polymerase chain reaction products of the full-length open reading frame were subcloned to pCRII vector (Invitrogen, Leek, The Netherlands) and sequenced to identify mutations. RII␣-mutated genes were then subcloned to pGEX-KG, and RII␣ mutants were expressed in E. coli and purified as described above.
Phosphorylation in Vitro of Human Recombinant Wild Type and Mutated RII␣ by Different Kinases-The catalytic subunit C of PKA type II (stock solution 0. 21 mg/ml) was purified from bovine heart as described by Lohmann et al. (27). The starfish cyclin B-p34 cdc2 kinase was purified from starfish oocytes by a protocol including affinity chromatography on p13 suc1 -Sepharose beads and consisted of two major polypeptides identified as p34 cdc2 catalytic subunit of Cdk1 and cyclin B (28). The specific activity of starfish CDK1 preparations varied from 150 to 1000 pmols of phosphate/min/l incorporated into histone H1 as a substrate. HeLa CDK1 was enriched by affinity chromatography on p13 suc1 -Sepharose beads from lysates of HeLa cells arrested in mitosis with nocodazole. The CDK1⅐p13 suc1 ⅐Sepharose beads complex (specific activity: 400 pmol of phosphate/min/mg incorporated into histone H1) was used to phosphorylate the particulate proteins isolated from HeLa cells arrested in G 1 phase of their cell cycle. Production and purification of yeast p13 suc1 protein were as described by Labbé et al. (29). Human recombinant extracellular signal-regulated kinase 1 (specific activity of 8.5 nmol of phosphate/min/mg of enzyme incorporated into myelic basic protein) or extracellular signal-regulated kinase 1 purified from sea star Piaster ochraceus (specific activity of 1 nmol of phosphate/min/mg of transferred to myelic basic protein) were from Upstate Biotechnology Inc. (Lake Placid, NY).
Recombinant human RII␣ subunits were phosphorylated in vitro at 20°C either in PKA phosphorylation buffer (50 mM Tris, pH 7.4, 1 mM MgCl 2 , 10 mM ␤-mercaptoethanol, 20 M cAMP, 100 M ATP, and 1 Ci of [␥-32 P]ATP (specific activity up to 5000 Ci/mmol (1 Ci ϭ 37 GBq; Amersham Pharmacia Biotech)) or for CDK1 and microtubule-associated protein kinase in EBS phosphorylation buffer without sucrose containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, 10 g/ml aprotinin), 100 M ATP, and 1 mCi [␥-32 P]ATP. The reaction was stopped by the addition of Laemmli buffer. For subsequent 8-azido[ 32 P]cAMP-labeling of the RII subunits, phosphorylation was performed in the absence of radioactive ATP, and the reaction was stopped at 4°C with the addition of 10 mM EDTA. After affinity-labeling, the samples were boiled in Laemmli buffer.
Phosphorylation of the Triton X-100-insoluble Fraction of HeLa Cells with CDK1 Isolated on p13 src1 -Sepharose Beads-Hela cells were arrested at the G 1 -S border with a double thymidine block. The cells were washed in 45 mM Pipes, 45 mM Hepes, 10 mM EGTA, 5 mM MgCl 2 , pH 6.9, (PHEM buffer (30)) containing the protease and phosphatase inhibitors described under "Cell Fractionation and Cdc2 Kinase Assay." Then they were extracted for 1 min with 0.5% Triton X-100 in PHEM buffer. The Triton X-100-soluble and -insoluble fractions were separated by centrifugation at 300 ϫ g for 10 min. The Triton X-100insoluble fraction (100 g of protein) was incubated for 30 min at 22°C in EBS phosphorylation buffer (as described above) containing 100 M ATP and human CDK1 (specific activity 400 pmol/min/g of protein for histone H1) that had been enriched 4 times on p13 suc1 -Sepharose beads from a cytosol of nocodazole-arrested cells (29). A control with the same amount of Triton X-100-insoluble protein was incubated with p13 suc1 -Sepharose beads without CDK1. The Sepharose beads and the particulate material were pelleted at 14,000 ϫ g for 20 min, and the proteins of the pellet and the supernatant were boiled in Laemmli buffer. Both fractions were examined for the amount of RII␣ by Western blotting as described above.
Immunocytochemistry-Cells grown on 35-mm plastic dishes or 22-mm glass coverslips were rinsed twice with PBS, fixed with methanol at Ϫ20°C for 6 min, then rinsed with PBS containing 0.1% Tween 20. Primary antibodies diluted in PBS containing 3% bovine serum albumin were added for 1 h at room temperature. The cells were then rinsed with PBS containing 0.1% Tween 20 three times to wash away unbound primary antibodies and incubated for 1 h at room temperature with mixed Texas red-conjugated goat anti-rabbit and fluorescein or/ and rhodamine-labeled goat anti-mouse secondary antibodies. For negative controls, the primary antibody was either omitted or replaced by a nonspecific IgG used at the same dilution as the primary antibody. Finally, the cells were mounted in Citifluor (Citifluor, UKC, Chemical Laboratories, Canterbury, CT27NH, United Kingdom), examined, and photographed on an Olympus BX60 epifluorescence microscope.

Photoaffinity-labeling with 8-Azido[ 32 P]cAMP and RII␣ Immunodetection Identify Different RII Forms in Mitotic and
Interphase HeLa Cytosols-8-Azido[ 32 P]cAMP affinity-labeling of R subunits was used to follow the different regulatory subunit isoforms present in cytosols from cells arrested during the cell cycle. The position of the cells in the cell cycle was monitored by flow cytometry analysis (Fig. 1A). The CDK1 mitotic kinase activity was measured in lysates of synchronized cells. A single thymidine block followed by L-mimosine arrest in late G 1 caused 73 Ϯ 12% of HeLa cells to accumulate in G 1 /S. Although it has been reported that mimosine-arrested cells have high levels of cyclin A-complexed CDK1 (31), we only detected low CDK1 activity in the lysate of these cells (200 Ϯ 58 pmol/mg/ min). Serum deprivation followed by an aphidicolin block yielded 73 Ϯ 5.8% of cells in S phase. The CDK1 activity was also low in these cells (349 Ϯ 5.7 pmol/mg/min). HeLa cells were synchronized in G 2 phase by a double-thymidine block followed by olomoucine arrest in G 2 . However olomoucine arrests the cells also in G 1 , and we obtained a mixed population of cells enriched in G 1 (29%) and in G 2 (70%). The CDK1 activity was very low (27.9 Ϯ 9.9 pmol/mg/min) in these cell extracts, even lower than that observed in G 1 -arrested cells. Cells were also arrested in a pseudometaphase state with nocodazole, which produced 94 Ϯ 9% mitotic cells. The CDK1 activity was quite high (2264 Ϯ 4.5 pmol/mg/min). The 8-azido[ 32 P]cAMP-labeling of the PKA regulatory subunits (Fig. 1B) was performed both in the absence and in the presence of an excess of unlabeled cAMP, the latter used as a check Use of unlabeled cAMP, which blocks 8-azido[ 32 P]cAMP-labeling in lanes 1 and 6 demonstrated the specificity of labeling of R subunits. C, immunodetection of RI␣ and RII␣ in HeLa cytosols. Proteins of HeLa cells arrested in G 1 , S, or M phases were separated by SDS-PAGE on 8% minigels and transferred to nitrocellulose. RI␣ was immunodetected with anti-human RI␣ monoclonal antibody diluted 1:1000, RII␣ with an affinity-purified anti-human RII␣ antibody diluted to 1:1000, both followed by a secondary antibody coupled to alkaline phosphatase. of the specificity of the labeling. In the cytosol of G 1 , of Senriched cells, or in the mixed population of G 1 and G 2 cells, three polypeptides of 49, 51, and 54 kDa specifically bound the 8-azido[ 32 P]cAMP (Fig. 1B, lanes 2-4). In the cytosol of mitotic cells, four polypeptides of 49, 51, 54, and 57 kDa were labeled (Fig. 1B, lane 5). No 8-azido[ 32 P]cAMP-labeling was observed in the presence of excess unlabeled cAMP (Fig. 1B, lanes 1 and  6).
In HeLa cells only two regulatory subunit isoforms have been identified, RI␣ and RII␣. This was observed with antibodies directed against nonhuman R isoforms by Weber et al. (32) and Nigg et al. (6) and more recently with specific antibodies directed against the known human isoforms of R subunits, RI␣, RI␤, RII␣, and RII␤ (data not shown). To identify the 8-azido[ 32 P]cAMP-labeled R isoforms in synchronized cell cytosols, RII␣ and RI␣ were immunodetected in different HeLa cytosols with an affinity-purified antibody directed specifically against the N-terminal domain of human RII␣ and a monoclonal antibody directed against human RI␣. RII␣ was always immunochemically detected in interphase HeLa cells as a polypeptide doublet of 51 and 54 kDa (Fig. 1C, right panel), corresponding to the well known unphosphorylated and autophosphorylated forms of this regulatory subunit (6,32). However in cells arrested in mitosis, an additional slower migrating polypeptide (57 kDa), which cross-reacted with anti-hRII␣ antibody, was always detected. Western blot analysis revealed no difference in the total amount of RII␣ detected during the cell cycle. In HeLa cytosols, the human RI␣ monoclonal antibody detected only one polypeptide at 49 kDa, which also did not vary in amount during the cell cycle (Fig. 1C, left panel).
The 54-and 57-kDa 8-Azido[ 32 P]cAMP-labeled RII Subunits Are Different Phosphorylation States of RII-In interphase cells, H89, a cell permeant inhibitor of cAMP-dependent protein kinase, led to the disappearance of the 54-kDa R-subunit (Fig. 2, lanes 1 versus 5). In mitotic cells treated with H89 (PKA activity in these cells was completely inhibited as verified using Leu-Arg-Arg-Ala-Ser-Leu-Gly (Kemptide) as a substrate), we detected no 57-kDa RII isoform. Although PKA activity was abolished, we still observed a faint 8-azido[ 32 P]cAMP-labeled R-subunit at 53-54 kDa (Fig. 2, lane 2). To distinguish between noncomplete H89 inhibition of RII autophosphorylation and another nonautophosphorylation event on RII in mitosis, we made use of olomoucine. Treatment of mitotic cells with 100 M olomoucine led to an inhibition of CDK1 (as previously measured in synchronized cell lysates) and to a disappearance of the 57-kDa R-subunit in mitotic cytosol (Fig. 2, lane 3). In contrast, olomoucine treatment of G 1 -arrested cells did not modify the 8-azido[ 32 P]cAMP binding pattern (not shown, see also Fig. 1B,  lane 4 for a mixed population of G 1 -and G 2 -arrested cells with olomoucine). Unfortunately the addition of both H89 and olomoucine on nocodazole-arrested cells led to cell death. Whereas phosphatase treatment of cytosols isolated from either the mitotic or the interphase cells abolished the 54-and 57-kDa RII-phosphorylated isoforms, only the 57-kDa RII isoform disappeared when the homogenization of mitotic cells was performed in the absence of okadaic acid (data not shown).
In Vivo 32 P Incorporation Reveals at Least Two Different Phosphorylation Sites on RII␣ in Mitotic Cells-To confirm the phosphorylation states of RII during mitosis, HeLa cells in interphase or arrested in mitosis were metabolically labeled with [ 32 P]orthophosphate. Cells were analyzed for the presence of labeled RII␣ by immunoprecipitation with anti-RII␣ antibodies raised either against rat RII␣ (Fig. 3A, lanes 1-4) or against a specific sequence of the N-terminal domain of human RII␣ (Fig. 3A, lanes 5-6). This last antibody was affinity-purified against the appropriate RII␣ peptide (33). The anti-rat RII␣ antibody has been shown to cross-react with human RII␣ (23). As described previously, R-subunits immunoprecipitated from unlabeled interphase HeLa cells were first monitored for their ability to serve as in vitro substrates of the catalytic subunit of PKA (23). RII␣ in the immunoprecipitate could be phosphorylated by the catalytic subunit of PKA and migrated as 54 kDa in the gel system used for our experiments (not shown; see also Ref. 23). After 32 P incorporation in mitotic and G 1 /S-arrested cells, RII␣ was immunoprecipitated with either the polyclonal antibody raised against rat heart RII␣ (Fig. 3A, lanes 1-2), the preimmune sera of the rabbit used for immunization with rat-RII␣ (Fig. 3A, lanes 3-4), and the affinity-purified polyclonal antibody raised against human RII␣ (Fig. 3A, lanes 5-6). Both antibodies gave similar results. Whereas only the polypeptide of 54 kDa was to some extent labeled in immunoprecipitates of interphase extracts, two polypeptides of 54 and 57 kDa were strongly labeled and reproducibly observed (5 separate experiments of [ 32 P] incorporation in vivo) in immunoprecipitates of mitotic extracts. RII was not immunoprecipitated with the preimmune serum of the rabbit immunized with rat RII␣. A proteolytic product of 46 kDa was often observed in immunoprecipitates of mitotic lysates. Several high molecular  1 and 2), preimmune sera from the rabbit later immunized with rat heart RII␣ (lanes 3 and 4), or with affinity-purified anti-human RII␣ (lanes 5 and 6). The immunoprecipitates were analyzed by SDS-PAGE and autoradiography. Both antibodies showed that RII␣ was hyperphosphorylated during mitosis. B, autoradiogram of tryptic digestions. The 57-kDa phosphorylated polypeptide band from the mitotic immunoprecipitate (M) and the 51-54-kDa area from the interphase immunoprecipitate (I) were excised from the gel and digested with trypsin-EDTA for 5 min at 4°C, and the tryptic digest was submitted to SDS-PAGE on a 20% polyacrylamide gel. Two proteolytic products of RII (36 and 17 kDa) were observed to be phosphorylated. weight phosphorylated polypeptides also coimmunoprecipitated with RII␣. Two of them were reproducibly observed at 350 and 150 kDa, even after immunoprecipitation with affinity-purified anti-hRII␣ Ab.
In 2 experiments, we next excised from wet gels the weakly labeled polypeptides at 51-54 kDa and the 57-kDa labeled polypeptide from interphase and mitotic cell immunoprecipitates, respectively. Excised polypeptides were then digested with trypsin-EDTA for 5 min. The trypsin digests were submitted to electrophoresis on a 20% polyacrylamide gel, and the gel was autoradiographed (Fig. 3B). Whereas no phosphorylation was observed in tryptic digest from interphase cells (lane 1), two major heavily 32 P-labeled proteolytic products of 36 and 17 kDa were observed in immunoprecipitates from mitotic cells (lane 2). When HeLa cells were arrested in G 1 and then stimulated with 10 M forskolin (to increase the cAMP concentration in these cells (34)), incorporation of 32 P was increased only in the 54-kDa RII isoform (not shown).

Human Recombinant RII␣ Is Phosphorylated by CDK1 at Thr 54 in the N-terminal Domain-As observed on Coomassie
Blue-stained SDS-PAGE gels (Fig. 4, left), a preparation of recombinant RII␣ showed the presence of four polypeptides corresponding to the GST⅐RII␣ complex (75 kDa), the cleaved human RII␣ (51 kDa), a major (46 kDa) proteolytic product of RII␣, and some contamination of cleaved GST (26 kDa). The 46-kDa protein corresponded to the 51-kDa human RII␣ cleaved by thrombin at amino acid Gln 57 as demonstrated by peptide sequencing of this polypeptide. Both the full-length recombinant protein and the 46-kDa cleavage product were phosphorylated by the catalytic subunit of PKA (see below Fig.  5). The human RII␣ preparation was phosphorylated in vitro at 20°C by the starfish CDK1 kinase for 2.5-30 min. Fig. 4 shows that native RII␣ was phosphorylated in vitro by CDK1 with a modification in the electrophoretic mobility from 51 to 53 kDa. The GST⅐RII␣ complex was slightly phosphorylated. The 46-kDa cleaved RII␣ was not phosphorylated, suggesting that phosphorylation occurs in the N-terminal domain of RII␣. The phosphorylated polypeptide at 49 kDa observed in the presence of CDK1 alone corresponded to cyclin B. Cyclin B has been reported to be phosphorylated by the catalytic subunit of CDK1, the p34 cdc2 kinase, in Xenopus oocytes and eggs (35) and in cultured human cells (36). We also noted a shift in the electrophoretic mobility of cyclin B when phosphorylated for 30 min.
Because the 46-kDa RII␣ proteolytic product corresponding to the C-terminal part of RII␣ starting at amino acid Gln 57 was not phosphorylated by CDK1, we mutated a putative phosphorylation site at Thr 54 of the N-terminal sequence Ala 52 -Ala-Thr-Pro-Arg-Gln 57 to either a glutamic acid or an alanine residue.
The native RII␣ and mutated RII␣ Thr 54 -Glu recombinant proteins were incubated for 10 min in vitro with [␥-32 P]ATP and 3 different purified kinases, i.e. the catalytic subunit of bovine heart PKA type II, purified starfish CDK1, and human recombinant extracellular signal-regulated kinase 1. The phosphorylation of both native and mutated RII␣ Thr 54 -Glu by PKA induced an electrophoretic mobility shift from 51-54 kDa both on Coomassie Blue-stained gels (not shown) and on autoradiograms (Fig. 6A, lanes 1 and 3). Whereas a distinct mobility shift and 32 P incorporation were observed when native RII␣ was incubated with starfish CDK1 (Fig. 6A, lane 2), only the phosphorylation of cyclin B associated by starfish CDK1 was observed after incubation of mutated RII␣ with CDK1 (Fig. 6A, lane 4; see also the phosphorylation of starfish CDK1 alone Fig.  4). Human recombinant extracellular signal-regulated kinase 1, shown to be active with myelic basic protein, did not phosphorylate native or mutated RII␣ (not shown).
RII␣ Subcellular Redistribution in Mitosis-G 1 -arrested or mitotic cells were fractionated in situ with 0.5% Triton X-100 in cytoskeleton-stabilizing buffer containing phosphatase and protease inhibitors. RII␣ was immunochemically detected in Triton X-100-insoluble and -soluble fractions with anti-hRII␣ Ab. In G 1 cells, the 54-kDa RII␣ phospho isoform was predominant in the insoluble fraction (Fig. 7A, lane 1), whereas the soluble fraction (Fig. 7A, lane 2) contained a major immunoreactive band at 51 kDa corresponding to unphosphorylated RII ␣. In mitotic cells, a consistent increase in RII␣ in the soluble fraction was observed (Fig. 7A, lane 4). Moreover a slowly migrating polypeptide (57 kDa) cross-reacting with the anti-hRII␣ antibody was always detected (lane 4). To further analyze this, cells were homogenized, and the homogenate was fractionated into nuclear pellet and post-nuclear supernatant followed by post-nuclear supernatant fractionation to particulate and cytosolic fractions. By RII␣ immunoblotting of these fractions, we similarly found equivalent amounts of RII␣ in both particulate and cytosol fractions of G 1 cells but a greater amount of soluble RII␣ in mitotic cells (not shown). Furthermore, phosphorylation of the Triton X-100-insoluble fraction of G 1 cells (which contains the majority of 54-kDa RII␣) with mitotic HeLa cell CDK1, isolated on p13 suc1 -Sepharose beads, led to RII␣ solubilization (Fig. 7B). Whereas no or very little 54-kDa RII␣ was released to the supernatant by incubation with p13 suc1 -Sepharose beads alone (lane 1), a large amount of RII␣ was released when CDK1 bound to the same beads were used (lane 3). RII␣ was not present in the mitotic kinaseenriched preparation of p13 suc1 -Sepharose beads as verified by Western blot. The RII␣ solubilization was observed in three separate experiments with variable specific activity of the human mitotic kinase bound to Sepharose beads. Immunofluorescence studies showed that whereas RII␣ was associated with centrosomes during interphase, RII␣ was either no longer observed at mitotic poles (Fig. 7C). In mitotic HeLa cells, RII␣ was either not observed in the centrosome region or detected as FIG. 7. A, immunodetection of RII␣ in HeLa subcellular fractions. HeLa cells arrested at the G 1 /S border (G 1 ) or in mitosis (M) were extracted in situ with 0.5% Triton X-100 for 1 min and fractionated into Triton X-100-insoluble (I) and -soluble (S) fractions. Proteins were separated by SDS-PAGE on 8% minigels and transferred to nitrocellulose. RII␣ was immunodetected with an affinity-purified anti-human RII␣ antibody diluted 1:1000 and a secondary antibody coupled to alkaline phosphatase. Shown are RII␣ forms of 51, 54, and 57 kDa. B, solubilization of RII␣ from Triton X-100-insoluble fraction of HeLa cells after incubation in vitro with human CDK1 bound to p13 suc1 -Sepharose beads. HeLa cells arrested at the G 1 -S border of the cell cycle were extracted in situ with 0.5% Triton X-100 for 1 min in cytoskeletonstabilizing buffer (PHEM, see "Materials and Methods"). Then the Triton X-100-insoluble and -soluble cell fractions were separated by centrifugation. Proteins (100 g) from the Triton X-100-insoluble fraction were incubated for 30 min in phosphorylation buffer with either CDK1 bound to p13 suc1 -Sepharose beads (histone H1-phosphorylating activity: 2000 pmols/min) or with p13 suc1 -Sepharose beads alone. At the end of the incubation, the mixture was centrifuged at 14,000 ϫ g for 20 min, and the pellet and supernatant were analyzed by SDS-PAGE and Western blot for the presence of RII␣ with the polyclonal antibody hRII␣ Ab. Whereas the total amount of supernatant was loaded on the gel, only 1/5 of the amount of pellet was loaded. Lanes 1 and 3, proteins from the supernatant (S) obtained after incubation with p13 suc1 -Sepharose beads alone (lane 1) or with CDK1 bound to beads (lane 3). Lanes 2 and 4, pellet proteins (P) after incubation with p13 suc1 -Sepharose beads alone (lane 2) or with CDK1 bound to beads (lane 4). The upper band observed in the pellet did not correspond to RII as this staining was not competited by the hRII␣ peptide used for immunization. The major 54-kDa RII␣ form was observed both in supernatant and pellet. C, exponentially growing HeLa cells were fixed with methanol and then double-labeled with human RII␣ antibody (hRII Ab (A)) and a specific centrosomal marker (monoclonal Mab453 (C)) and were counterstained with 4Ј,6-diamino-2-phenylindole dihydrochloride (Dapi (B)). The white arrowhead shows the staining of centrosome at mitotic poles in metaphase cell with monoclonal Ab 453 but no staining with hRII Ab was observed. The thin white arrow indicates that a weak staining with hRII␣ Ab was sometimes observed at centrosome as a tiny dot. Large white arrows show the staining of centrosome in interphase cells both with hRII Ab (A) and monoclonal Ab 453 (C). Bar, 20 m. very low levels at mitotic poles compared with the staining of RII␣ at centrosome in interphase (Fig. 7C, thin white arrow). In HeLa cells arrested in mitosis with nocodazole, no RII␣ was detected at unsplit centrosomes (not shown). This was observed for several cells such as SaOS2 osteosarcoma cells, neuroblastma cells, and human skin fibroblasts in primary cultures. DISCUSSION cAMP-dependent protein kinase has been directly implicated in cell cycle regulation, and its down-regulation may possibly play a role in the induction of mitosis and nuclear envelope breakdown in mammalian cells (37). Subsequently, during the metaphase-anaphase transition, concomitant with a drop in mitotic CDK1 activity, PKA is up-regulated to stimulate exit from mitosis (38,39). Although most of the regulation of cAMPdependent protein kinases derives from modifications in cAMP concentration, the phosphorylation states of the catalytic and regulatory subunits may also modulate the properties of these proteins. Because an increasing number of kinases are the targets of the mitotic kinase CDK1 (40 -43), we questioned whether PKA might be also a target for CDK1 or a kinase activated only during mitosis. We previously reported that the regulatory subunit RII␤ of PKA type II␤ is phosphorylated in vitro by CDK1 (14). However PKA type II␤ is poorly expressed in dividing cells, whereas PKA type II␣ is ubiquitously expressed. Moreover the amino acid sequences of RII␤ and RII␣ differ in the N-terminal domain (2).
Both 8-azido[ 32 P]cAMP-labeling and RI␣ or RII␣ immunodetection with specific antibodies were performed to follow the regulatory subunits of PKA in HeLa cytosols during the cell cycle. First, care was taken to prevent changes in the phosphorylation state of RII during cell homogenization (44). Okadaic acid and orthovanadate were added to prevent the action of serine/threonine type-1 protein phosphatase, type-2 protein phosphatase class A (45), and tyrosine phosphatases (46), respectively. Phosphorylated RII has been shown to be a substrate for type-1 protein phosphatase and type-2 protein phosphatase class A (47). Calcineurin (calcium-dependent type-2 protein phosphatase class B) has been shown to associate with RII-anchoring protein, AKAP 79 (48). The catalytic subunit of type-2 protein phosphatase class B has been shown to dephosphorylate RII on the site phosphorylated by the catalytic subunit of PKA (49). EGTA was present in the homogenization buffer to prevent the activation of calcineurin. Homogenization carried out in the presence of phosphatase inhibitors, ATP, and dithiothreitol generates mitotic cytosols and extracts with a more stable CDK1 activity (50). Both 8-azido[ 32 P]cAMP-labeling and RII␣ immunodetection showed that the number of RII␣ forms was dependent on the cell cycle phase in which cells were arrested. In cells arrested in late G 1 or S phases or in a mixed population of G 1 ϩ G 2 -arrested cells, a major RII isoform at 51 kDa was observed in HeLa cytosols. This was not the case in the cytosol of mitotic-arrested cells in which the amount of the 54-kDa RII form was more abundant than the 51-kDa isoform, and moreover, a third RII isoform appeared at 57 kDa. Immunodetection with specific antibodies identified only one RI␣ isoform at 49 kDa.
Further treatment of synchronized cells with kinase inhibitors showed that the slowly migrating forms of regulatory subunits were phosphorylated forms. In interphase cells treated with the PKA inhibitor H89, the autophosphorylated RII isoform (54 kDa) was not observed. In mitotic cells treated with H89 (PKA activity in these cells was completely inhibited as verified using Kemptide as a substrate) we still detected a minor amount of the 53-54-kDa RII isoform but no 57-kDa RII isoform. This can be explained either by a noncomplete H89 inhibition of RII autophosphorylation by PKA or another mod-ification occurring on RII during mitosis. Olomoucine, which has been shown to inhibit mainly the cyclin-directed kinases in several cell types (16), suppressed the 57-kDa RII isoform. In this case, only autophosphorylated RII was observed at 54 kDa. Both 57-and 54-kDa 8-azido[ 32 P]cAMP-labeled RII subunits were absent after alkaline phosphatase treatment of the mitotic cytosols, also suggesting that these R subunits are phospho isoforms of RII␣. From these data we conclude that during mitosis RII␣ is phosphorylated, not only on the autophosphorylation site (migration as 54 kDa) but also on another site that also induces a similar change in the electrophoretic mobility of this protein (to 53-54 kDa). The 57-kDa RII phospho isoform might correspond to RII␣ which is both autophosphorylated and phosphorylated on a second site and displays even slower mobility. When autophosphorylation is blocked, the 57-kDa RII form is not present. In vivo 32 P incorporation into synchronized HeLa cells also indicated hyperphosphorylation of RII␣ during mitosis. An RII phospho isoform with a slower electrophoretic mobility (57 kDa) than RII phosphorylated only on its autophosphorylation site (54 kDa) was observed. Tryptic digestion of the phosphorylated RII polypeptide gave two major phosphorylated proteolytic products of 34 and 17 kDa. HeLa cell RII␣ has been shown to be degraded to a 35-36-kDa product (32), and trypsin digestion of bovine RII␣ phosphorylated by PKA (56 kDa) gave two 39-and 17-kDa products (24). From these observations, we can conclude that RII␣ was phosphorylated on at least two different sites during mitosis, one most likely corresponding to the autophosphorylation site and the other localized in the N-terminal domain of the protein. Analysis of the human RII␣ sequence showed a putative phosphorylation site (P) for cyclin-directed kinases in the sequence (Pro 51 -Ala-Ala-Thr 54 -P-Arg-Gln 57 ) of the N-terminal domain (2). Human wild type RII␣ was an in vitro substrate for CDK1, and upon phosphorylation, changed its electrophoretic mobility from 51 to 53 kDa. Human RII␣ was not phosphorylated by CDK1 when Thr 54 was mutated to glutamic acid or to alanine by sitedirected mutagenesis, suggesting that CDK1 phosphorylated RII␣ in vitro on Thr 54 . A putative phosphorylation site for cyclin-directed kinases is also present in RII of several species including Saccharomyces cerevisiae (bcy 1 gene) (51), which, like mammalian RII, also displays an autophosphorylation site.
Phosphorylation of RII during mitosis could potentially affect the subcellular localization of PKA type II. PKA type II has been shown to associate with centrosome and with the Golgi apparatus (6 -8) via different anchoring proteins (23,52). Autophosphorylated RII␣ (54 kDa) is always more abundant in the particulate fraction than in the soluble fraction of cells. It is also interesting to note that, when associated with centrosome or with Golgi membranes isolated from different exponentially growing cells, RII␣ displays various phosphorylation states (23,52). The different phosphorylations of RII␣ might correspond to modifications in either PKA or other kinase activities during the cell cycle. These phosphorylation modifications appear to lead to subcellular redistribution of the PKA type II within the cell, perhaps to prevent PKA from reaching protein targets that are also substrates for other kinases, particularly mitotic kinases. During G 1 and S phases, the pool of RII␣ associated with membranes and cytoskeleton consists mainly of the 54-kDa autophosphorylated form. During mitosis, we consistently observed an increase amount of RII␣ in cytosol, suggesting a redistribution of this regulatory subunit. Moreover incubation of cytoskeleton fractions with CDK1 bound to p13 suc1 -Sepharose beads led to 54-kDa RII␣ solubilization. This is also the major RII␣ form found in the cytosol of mitotic cells. Immunostaining of different cell lines with anti-human RII␣ antibody suggested that the majority of RII␣ that is associated with centrosomes during interphase is removed at metaphase just before anaphase. Bailly et al. (22) show that CDK1 associates with centrosome at the onset of mitosis. The removal of centrosomal RII␣ cannot be explained by only a redistribution of the pericentriolar material at metaphase-anaphase transition, as several pericentriolar antigens are still present at the end of metaphase. Furthermore, the RII␤ subunit, which is expressed in differentiated normal cells and in neoplastic dividing cells, was not subject to cell redistribution from the centrosome at mitosis (data not shown). This serves as a control for redistribution of centrosomal RII␣ and indicates that RII redistribution is not linked to a modification in the amount of anchoring protein or A kinase-anchoring protein in the pericentriolar material during the cell cycle. Future imunofluorescence studies on in situ expressed mutated RII␣ in cells will address the question of whether the phosphorylation of Thr 54 in human RII␣ alters the subcellular localization of PKA type II at the onset of mitosis.