Regulation of the Multifunctional Ca 2 (cid:1) /Calmodulin-dependent Protein Kinase II by the PP2C Phosphatase PPM1F in Fibroblasts*

The regulation of the multifunctional calcium/calmod-ulin dependent protein kinase II (CaMKII) by serine/ threonine protein phosphatases has been extensively studied in neuronal cells; however, this regulation has not been investigated previously in fibroblasts. We cloned a cDNA from SV40-transformed human fibroblasts that shares 80% homology to a rat calcium/cal-modulin-dependent protein kinase phosphatase that en-codes a PPM1F protein. By using extracts from transfected cells, PPM1F, but not a mutant (R326A) in the conserved catalytic domain, was found to dephosphorylate in vitro a peptide corresponding to the auto-inhibitory region of CaMKII. Further analyses demonstrated that PPM1F specifically dephosphorylates the phospho-Thr-286 in autophosphorylated CaMKII substrate and thus deactivates the CaMKII in vitro . Coimmunoprecipitation of CaMKII with PPM1F indicates that the two proteins can interact intracellularly. Binding of PPM1F to CaMKII involves multiple regions and is not dependent on intact phosphatase activity. Fur-thermore, overexpression of PPM1F in fibroblasts caused a reduction in the CaMKII-specific phosphorylation of the known substrate vimentin(Ser-82) anti-Myc epitope Cruz (cid:4) and with monoclonal to en-sure that equal amounts of protein were loaded. The blots were then with horseradish peroxidase-conjugated rat anti-mouse (cid:5) light chain-specific IgG (Zymed Laboratories Inc.) as the secondary antibody. Enhanced chemiluminescence was performed by using Western Light-ning TM Chemiluminescence Reagent Plus (PerkinElmer Life Sciences). To determine the activity directly in lysates, the extract supernatant was diluted 1:4 in excess extraction buffer. Phosphatase reactions were initiated with the addition of an aliquot of diluted supernatant to a 30 °C preincubated reaction mixture consisting of the following: 7.5 (cid:4) M 32 P-AutoCaMtide3 solution (see “Preparation of Phosphatase Assay Substrate” for details), 10 m M MnCl 2 , and 10 (cid:4) M CaMKII-specific in- hibitory peptide (CaM-KIINtide) (39) (synthesized by Sigma-Genosys) with or without 10 (cid:4) M OA (Sigma, sodium salt). Various reaction times were used depending on the experiment and are provided in the figure for that experiment. Reactions were terminated by the addition of ice-cold 15% trichloroacetic acid. Recovery of 32 P-AutoCaMtide3 as well as analysis of samples by scintillation spectrometry was performed according to the same procedure used in the in vitro CaMKII assay described by Dhavan et al. A sample devoid of cell extract was included with each assay as a control to define the total amount of 32 P incorporated in the substrate. The difference of the sample (experimen- tal) values from control (starting) indicated the amount of PO 4 from the peptide as a result of phosphatase activity. To assess the activity of immunoprecipitated PPM1F-B (cid:1) HA, cell extracts were prepared

The dynamic interactions between kinases and phosphatases play a critical role in the regulation of cellular processes. Because of their extensive involvement in calcium signaling (1), the multifunctional Ca 2ϩ /calmodulin-dependent protein kinase II (CaMKII) 1 and the protein phosphatases that interact with it have been the focus of several studies. In response to Ca 2ϩ -mobilizing agents, CaMKII is autophosphorylated (at Thr-286 for CaMKII␣) generating a kinase with Ca 2ϩ -independent activity (autonomous activity); however, this activity rapidly declines on removal of the stimulus (2)(3)(4), suggesting that protein phosphatases may contribute to the regulation of CaM kinase. Dephosphorylation of Thr-286 by the serine/threonine protein phosphatases PP1 (5), PP2A (6), and PP2C (7) has been shown to occur in vitro, and the phosphatase(s) responsible for regulating the endogenous CaMKII activity have been identified in a variety of cell types (7)(8)(9)(10)(11)(12)(13)(14)(15).
Because CaMKII comprises up to 2% of the total protein in some regions of the brain, multiple investigations into the regulation of this kinase by phosphatases have been done with tissue from the central nervous system (7)(8)(9)(10)(11)(12)15). In the rat forebrain, the predominant phosphatase varies with the distinct cellular compartment; PP2A dephosphorylates soluble CaMKII, whereas PP1 acts on postsynaptic density-associated kinase (8). In the case of rat cerebellar granule cells, the predominant phosphatase activity in cytosolic extracts is PP2C (7). More limited studies have examined the regulation of CaMKII by phosphatases in non-neuronal cells, including canine vascular smooth muscle cells (16,17), mouse pancreatic ␤-cells (14), and rat pancreatic acinar cells (13).
However, CaM kinase regulation has not been evaluated in fibroblasts. CaMKII has been shown to be expressed in these cells based on the detection of kinase activity in the following rodent and human fibroblast cell lines: NIH 3T3 (mouse embryo) (18); 3Y1 (rat embryo) (19); WI38 (human embryonic lung) (20,21); and GM38 (human foreskin) (22). Hence, this kinase may play a critical role in the Ca 2ϩ -mediated signaling cascades that have been shown to be required for cell cycle progression in fibroblasts (23)(24)(25). In support of this possibility, treatment of proliferating NIH 3T3 cells with KN-93, a membrane-permeable synthetic inhibitor of CaMKII, induced G 1 arrest (18). In addition, CaMKII contributes to the S-phase delay in the cell cycle of ␥-irradiated human fibroblasts; ataxia telangiectasia fibroblasts fail to undergo this delay because of the inability of these cells to activate CaMKII following ␥-irradiation (22), and normal fibroblasts treated with KN-62, a pharmacological inhibitor of CaM kinases, exhibit an "ataxia telangiectasia-like" phenotype after ␥-ray treatment (26). As has been observed with neuronal cells, the increase in autonomous CaMKII activity was transient in the ␥-irradiated normal fibroblasts, suggesting that the rapid decline in kinase activity is the result of phosphatase regulation (22).
Presently, very little is known about the phosphatases that regulate CaMKII in fibroblasts. There is indirect evidence to indicate that this regulation does occur. The finding that Swiss 3T3 fibroblasts, pretreated with KN-93, were protected from apoptosis induced by nodularin, a cyclic pentapeptide inhibitor of PP1 and PP2A, suggests that the inhibition of these phosphatases can lead to CaM kinase deregulation and subsequent apoptosis in fibroblasts (27). We previously identified a cDNA (termed J4-3) from a subtractive hybridization screen of SV40transformed human fibroblasts (28) that matched the 3Ј-untranslated region of a cDNA (KIAA0015) cloned from a human myeloid cell line (29) that contained a putative PP2C phosphatase motif. Since then, two groups of investigators have independently isolated the same cDNA from different sources and have named it hFEM-2 (30) and POPX2 (31), depending on its identified function. Both groups recognized what we had found earlier during the course of our investigation, 2 the human sequence shares 80% identity with the rat Ca 2ϩ /calmodulin-dependent protein kinase phosphatase (rCaMKPase) (32); however, neither group fully investigated this possible function of the gene, other than to demonstrate that a partially purified hFEM-2 could dephosphorylate CaMKII in vitro (30). The rCaMKPase has characteristics similar to members of the PP2C family, including a dependence on divalent cations for activity and insensitivity to okadaic acid (OA), and has been shown to dephosphorylate as well as deactivate autophosphorylated CaMKII in vitro (33). This phosphatase substantiates the findings in previous studies (7,8,10,14) that PP2C-like phosphatases play a role in regulating CaM kinase activity. To assess CaMKII regulation in fibroblasts, we determined whether J4-3, which we name PPM1F in accordance with the PP2C family designation as given by the Human Genome Nomenclature Commission, functions as CaM kinase phosphatase. Our results show that like its rat homolog PPM1F specifically dephosphorylates Thr-286 of CaMKII and deactivates the autophosphorylated kinase in vitro. Moreover, we demonstrate by coimmunoprecipitation (co-IP) that the two proteins interact intracellularly and, most important, that PPM1F can inhibit the intracellular phosphorylation of a CaMKII-specific substrate by the endogenous CaM kinase, indicating that PPM1F can contribute to the regulation of CaMKII in fibroblasts.
Cloning of PPM1F and Construction of Deletion Mutants-The PPM1F-A cDNA (full-length J4-3) was generated from an SV40-transformed nonimmortal human diploid fibroblast cell line by reverse transcription with an oligo(dT) primer by using the SuperScript TM Preamplification System (Invitrogen) followed by PCR with DNA primers flanking the largest open reading frame of the cDNA by using Taq polymerase (Roche Diagnostics). PPM1F-A was cloned into pIRES2EGFP (Clontech), and an HA epitope tag was introduced at the C-terminal end by ligating a double-stranded oligo to the BstAPI (SibEnzyme) site overhang, resulting in a 4-amino acid truncated protein that we termed PPM1F-B. The construct carrying the HA epitopetagged PPM1F-B (PPM1F-B⅐HA) under the control of a CMV promoter was named pPPM1F-B⅐HA/IG. The CMV-IE promoter/PPM1F-B⅐HA/ IRES/EGFP/SV40 poly(A) cassette was subcloned into a vector carrying a Pac gene under the control of an SV40 origin-defective early promoter resulting in the pPPM1F-B⅐HA/IGPac construct. The PPM1F cDNA was generated from the ovarian cancer cell line SKOV3 using the same protocol as described for PPM1F-A and was cloned into pVAX1 (Invitrogen). The nontagged version of the PPM1F ⌬13-94 mutant was generated by the ligation of MscI and filled in EcoRI overhang (New England Biolabs) within the PPM1F sequence. The PPM1F cDNA was subcloned into the SalI (Invitrogen) site of pCMV-Myc (Clontech). As a consequence of generating the construct pCMV-Myc⅐PPM1F, an additional 33 amino acids, including the Myc epitope tag, are present at the N terminus of the Myc⅐PPM1F protein. All subsequent deletion mutants of PPM1F were generated by using pCMV-Myc as the vector and consequently contain the additional amino acids unless otherwise specified. The PPM1F ⌬13-94 gene was subcloned into the SalI site to give the Myc⅐PPM1F ⌬13-94 mutant. The ligation of the SrfI (Stratagene) overhang from PPM1F with that of SalI from the vector generated the ⌬1-148 mutant that has only 23 of the 33 additional amino acids mentioned earlier. The large internal deletion mutant ⌬148 -397 was constructed by ligating together the two SmaI (Roche Diagnostics) sites in PPM1F that were the farthest apart from each other. To generate the ⌬300 -454 deletion mutant, the filled in overhangs of BssHII (New England Biolabs) from PPM1F and NotI (New England Biolabs) from the vector were ligated together. As a result of this ligation, the stop codon fell within the vector sequence 7 amino acids downstream of the PPM1F sequence. The remainder of the C-terminal deletion mutants of PPM1F, which include ⌬411-454, ⌬417-454, ⌬419 -454, ⌬422-454, ⌬433-454, and ⌬441-454, were constructed by Expand High Fidelity PCR (Roche Diagnostics) by using a common 5Ј DNA primer corresponding to PPM1F sequence directly proximal to the open reading frame together with an XhoI restriction site. Specific 3Ј primers encode the amino acids immediately upstream of the deletion together with a stop codon at the position of the truncation and a BglII (Roche Diagnostics) restriction site. The missense mutant R326A was generated by the same PCR method by using a 5Ј primer containing a conversion of the AGA codon (976 -978 bp of the PPM1F cDNA) to that of GCA as well as a HincII (New England Biolabs) restriction site, and a 3Ј primer corresponding to the distal open reading frame sequence, including the stop codon together with a BstAPI restriction site. The digested PCR product was incorporated into the PPM1F cDNA by replacing the original HincII/BstAPI fragment. All PPM1F constructs were sequenced in their entirety to verify that no other mutations were present. DNA primers and sequencing were provided by the Molecular Resource Facility at the UMDNJ-New Jersey Medical School.
Preparation of Phosphatase Assay Substrate-AutoCaMtide3 (Invitrogen) at a concentration of 100 M was phosphorylated in vitro with 500 ng of purified CaMKII (Upstate Biotechnology, Inc.), as described previously (37), with the addition of 1 mM CaCl 2 and 20 g/ml calmodulin to the reaction mixture ([␥-32 P]ATP (6,000 Ci/mmol, PerkinElmer Life Sciences)). The reaction was diluted with 3ϫ phosphatase reaction buffer containing 33.3 mM Tris-HCl (pH 7.5) and 0.33 mM EGTA (pH 7.5) to a final concentration of 25 M 32 P-AutoCaMtide3. Labeled peptide was separated from kinase protein by centrifugation through Microcon-10 microconcentrators (Millipore).
Phosphatase Assays-Depending on the experimental conditions, phosphatase assays were performed to evaluate the activity of exogenous PPM1F in a variety of cell types. For the time course analysis as well as the immunoprecipitation experiments, CHO-K1 were seeded at 7.5 ϫ 10 5 cells per 60-mm plate and then transiently transfected using LipofectAMINE2000 TM (Invitrogen) with 10 g of pPPM1F-B⅐HA/IG or vector control plasmid for 5 h. These cells were then harvested 40 h post-transfection for assay. For the phosphatase assays involving deletion mutants of PPM1F, HEK293T cells were seeded at 3.5 ϫ 10 6 cells per 60-mm plate and then transiently transfected overnight using LipofectAMINE2000 TM with 10 g of pCMV-Myc⅐PPM1F or with specified deletion mutants. These cells were then harvested 18 h posttransfection for assay. Cell extracts were prepared in a manner similar to that described by Dhavan et al. (38) with exclusion of the phosphatase inhibitors from the extraction buffer (referred to as modified PIPES-buffered extraction solution). For each sample, an aliquot was used for protein estimation with the Bio-Rad protein assay and for Western blot analysis with the mouse monoclonal anti-HA epitope or anti-Myc epitope antibodies (Santa Cruz Biotechnology, 0.2 g/ml) and with mouse monoclonal anti-␣-actinin antibody (Sigma, 1:1000) to ensure that equal amounts of protein were loaded. The blots were then stained with horseradish peroxidase-conjugated rat anti-mouse light chain-specific IgG (Zymed Laboratories Inc.) as the secondary antibody. Enhanced chemiluminescence was performed by using Western Lightning TM Chemiluminescence Reagent Plus (PerkinElmer Life Sciences).
To determine the activity directly in lysates, the extract supernatant was diluted 1:4 in excess extraction buffer. Phosphatase reactions were initiated with the addition of an aliquot of diluted supernatant to a 30°C preincubated reaction mixture consisting of the following: 7.5 M 32 P-AutoCaMtide3 solution (see "Preparation of Phosphatase Assay Substrate" for details), 10 mM MnCl 2 , and 10 M CaMKII-specific inhibitory peptide (CaM-KIINtide) (39) (synthesized by Sigma-Genosys) with or without 10 M OA (Sigma, sodium salt). Various reaction times were used depending on the experiment and are provided in the figure for that experiment. Reactions were terminated by the addition of ice-cold 15% trichloroacetic acid. Recovery of 32 P-AutoCaMtide3 as well as analysis of samples by scintillation spectrometry was performed according to the same procedure used in the in vitro CaMKII assay described by Dhavan et al. (38). A sample devoid of cell extract was included with each assay as a control to define the total amount of 32 P incorporated in the substrate. The difference of the sample (experimental) values from control (starting) indicated the amount of PO 4 released from the peptide as a result of phosphatase activity. To assess the activity of immunoprecipitated PPM1F-B⅐HA, cell extracts were prepared as stated previously with the exception that the lysates were centrifuged at 14,000 ϫ g for 5 min at 4°C. An aliquot of each supernatant was retained at 4°C for the phosphatase reaction as well as for Western blot analysis by using anti-HA epitope antibody. To 500 g of total protein for each sample, 20 g of agarose conjugated anti-HA epitope (anti-HA-AC) antibody (Santa Cruz Biotechnology) was added, and the sample was incubated for 1 h at 4°C. After centrifugation, the agarose beads were washed twice with ice-cold modified PIPES-buffered extraction solution (5 min per wash) at 4°C. The immunoprecipitates were then resuspended in a minimal amount of extraction buffer and immediately added to the phosphatase reaction mixture. The reactions were completed following the protocol listed above.
In Vitro Dephosphorylation of CaMKII-HEK293T were transiently transfected overnight with 10 g of plasmid for the following genes: mCaMKII␣ (pME18S-CaM Kinase II, a gift from Dr. Thomas R. Soderling (Oregon Health Sciences University, Portland, OR)), Myc⅐PPM1F, Myc⅐PPM1F R326A, Myc⅐PPM1F⌬419 -454, Myc⅐PPM1F⌬422-454, or vector plasmids. Cells were then harvested 18 h post-transfection, and cell extracts were prepared in a similar manner as described under "Phosphatase Assays." To activate CaMKII exogenously, aliquots (ϳ40 g of total cellular protein per reaction) of mCaMKII␣-containing lysate were used in limiting autophosphorylation reactions as described previously by Brickey et al. (40) with the addition of 10 M OA. Further activation was inhibited by the addition of 50 M KN-93 (Calbiochem, water-soluble). To dephosphorylate CaMKII, aliquots of activated mCaMKII were added to the phosphatase reaction buffer containing 2 mM MnCl 2 , 10 M OA, and 50 M KN-93. These reactions were initiated by the addition of extract (ϳ50 g of total cellular protein per reaction) prepared from cells transfected with Myc⅐PPM1F (or related construct) and were incubated for 30 min at 23°C before being terminated with the addition of SDS loading buffer. Replicate samples were pooled and then analyzed by Western blot with rabbit polyclonal anti-ACTIVE TM CaMKII (pT286) (also known as anti-phospho-CaMKII␣/␤ (Thr-286/ Thr-287)) (Promega, 1:1000 dilution), rabbit polyclonal anti-CaMKII␣ (Sigma, 5 g/ml), and anti-Myc epitope-specific antibodies. Horseradish peroxidase-conjugated goat anti-rabbit IgG (Sigma) was used as the secondary antibody for the rabbit polyclonal antibodies.
In Vitro Deactivation of CaMKII-The steps to the point of terminating the phosphatase reaction were done essentially the same way as described under "In Vitro Dephosphorylation of CaMKII." In place of SDS loading buffer, 400 mM ␤-glycerophosphate, a nonspecific inhibitor of protein phosphatases, was added to each phosphatase reaction to give a final concentration of 80 mM in the kinase reaction. Autonomous CaMKII activity was assessed (for 3 min) for aliquots of this mixture utilizing the in vitro CaMKII assay protocol previously described by Dhavan et al. (38) (kinase assay kit from Invitrogen) with the following additions: OA (10 M), KN-93 (50 M), protein kinase A (Sigma, protein kinase inhibitor), and protein kinase C (fragment 19 -36, Sigma) inhibitors were included in the reaction buffer at a final concentration of 0.4 M each. Bio-Rad protein assay was used for protein estimation.
Coimmunoprecipitations-HEK293T cells were transiently cotransfected with 5 g of pCMV-CaMKIIwt (CaMKII cDNA subcloned from pME18S-CaM kinase II (provided by Dr. Thomas R. Soderling) into pVAX1 (Invitrogen)) and 5 g of pCMV-Myc⅐PPM1F or specified deletion mutant plasmid. Cell extracts were prepared as described previously by Aplin and Juliano (41) using the modified radioimmunoprecipitation assay (RIPA) buffer with the exception that the protease inhibitors were substituted with the following: 5 g/ml aprotinin, 50 g/ml leupeptin, and 100 g/ml Pefabloc® SC (Roche Diagnostics). Protein estimation was performed on the supernatant using the DC Protein Assay kit (Bio-Rad). For each sample, 500 g of total cellular protein was pre-cleared with 20 g of anti-HA-AC antibody for 30 min at 4°C and then centrifuged. Twenty micrograms of agarose-conjugated anti-Myc epitope (anti-Myc-AC) antibody (Santa Cruz Biotechnology) was added to the supernatant, and the immunoprecipitation (IP) reaction was incubated for 1 h at 4°C. The immunoprecipitate was centrifuged and washed three times at 4°C as follows: once with ice-cold modified RIPA buffer containing 1 M NaCl for 10 min, and then twice more with ice-cold modified RIPA buffer containing 0.15 M NaCl. The immunoprecipitate was resuspended in modified RIPA buffer and SDS loading buffer and subjected to Western blot analysis with rabbit polyclonal anti-CaMKII␣ and anti-Myc epitope-specific antibodies. For each sample, 20 g of cell lysate was analyzed by Western blot with the same anti-CaMKII␣ antibody.
CaMKII-specific Phosphorylation of Vimentin-For the experiments with endogenous CaM kinase, NIH 3T3 cells were seeded at 1.5 ϫ 10 6 cells per 100-mm plate (3 plates per plasmid) and then were transiently transfected for 5 h using LipofectAMINE PLUS TM (Invitrogen) with 10 g per plate of pPPM1F-B⅐HA/IGPac or vector. For the experiments with the constitutively active CaMKII mutant, NIH 3T3 cells were transiently cotransfected with 3 g of pRSV-CaMKII-(1-290) (a gift from Dr. Thomas R. Soderling) and 10 g of pPPM1F-B⅐HA/IGPac or vector. Eighteen hours post-transfection, cells from each 100-mm plate were transferred into a single 150-mm plate with complete medium containing 4 g/ml puromycin (Sigma). Twelve hours prior to cell harvesting, surviving cells from all 3 plates of the same plasmid were consolidated into a single 60-mm plate containing selection medium. For the experiments with the endogenous CaM kinase, the cells, after 48 h of puromycin selection, were treated with 1 M ionomycin or Me 2 SO (for 5 min at 37°C). Cell extracts were prepared as described previously by Kitani et al. (42) with the additional step of passing each Nonidet P-40-insoluble pellet through a 25-gauge syringe needle (at least four times). An aliquot from each sample was removed prior to the addition of SDS loading buffer and was used for protein estimation with the DC Protein Assay kit. Ten micrograms of each extract was subjected to Western blot analysis with goat anti-vimentin antiserum (Sigma, 1:1000 dilution) and then with horseradish peroxidase-conjugated rabbit anti-goat IgG (Sigma) as the secondary antibody. The amount of sample extract loaded on subsequent gels was normalized according to the uninduced vector sample. Replicate Western blots were generated and stained with mouse monoclonal anti-phospho-vimentin(Ser-82) (MO82) antibody (1:1000 dilution) (43), gift from Dr. Masaki Inagaki (Aichi Cancer Center Research Institute, Nagoya, Aichi, Japan), or with anti-vimentin antiserum.

Clone J4-3 as a Potential CaM
Kinase Phosphatase-J4-3 (28) and its open reading frame were cloned (see "Experimental Procedures") from an SV40-transformed derivative of the human cell line HS74 (HF), a fetal bone marrow stromal fibroblast. The sequence from our complete cDNA, which will be referred to as PPM1F-A, was 98% identical to KIAA0015 (29) with the disparity between the two DNA sequences resulting in a change of two amino acid residues (S2C and S124N). We attributed this discrepancy to a possible allelic difference between the donors of the two cell lines. The PPM1F-A cDNA was subcloned into a glutathione S-transferase fusion protein expression plasmid to generate recombinant protein for in vitro analysis. In our preliminary experiments using 32 P-casein as a substrate, the fusion protein exhibited phosphatase activity consistent with a PP2C (PPM1) (data not shown), such as dependence on divalent cations for activity and resistance to 10 M OA (44). Because the PPM1F-A cDNA did encode for a functional phosphatase, we performed Western blot analysis on HF lysates with the antibody PPM1F-454C, which specifically recognizes the C-terminal end of the human phosphatase, and we found that the protein is being expressed in these cells (data not shown). Because of the lack of specific inhibitors and the failure of the PPM1F-454C antibody to immunoprecipitate sufficient levels of endogenous phosphatase, it was not possible to discriminate among PPM1 activities in these cells, and consequently, the endogenous level of PPM1F-A activity could not be assessed.
While these studies were ongoing, the putative PPM1F-A sequence was found to share 80% identity with rCaMKPase (32), suggesting that PPM1F could act as a CaM kinase phosphatase. The rCaMKPase has been shown to dephosphorylate specifically and to deactivate CaMKI, -II, and -IV (45). We focused our efforts on CaMKII as a potential substrate in HS74 based on the fact that it is a multifunctional kinase and is widely expressed. The expression of CaMKII in HF lysates was assessed by an in vitro CaMKII-specific assay, and we found that the kinase is present in these cells at a level of activity (data not shown) similar to that reported for the normal lung fibroblast cell line WI38 (20).
Because both proteins are present in HF, we proceeded to determine whether CaMKII could be a substrate for PPM1F. Our first approach was to use 32 P-AutoCaMtide3 as the substrate for in vitro phosphatase assays. An analogous peptide substrate, which also contained the Thr-286 autophosphorylation site of CaMKII␣, had been used to identify (37) and characterize (33) rCaMKPase as a CaM kinase phosphatase in vitro. A C-terminal hemagglutinin (HA) epitope-tagged PPM1F-A was generated in the mammalian expression plasmid pIRES2EGFP (see "Experimental Procedures") and was termed PPM1F-B⅐HA. The resulting construct (pPPM1F-B⅐HA/ IG) and the parent vector were transfected separately into CHO-K1 cells, and extracts were prepared 40 h post-transfection. Because cell lysates contained endogenous CaMKII as well as several phosphatases, we included CaM-KIINtide to block CaMKII phosphorylation of AutoCaMtide3 and OA to inhibit the majority of protein serine/threonine phosphatases with the notable exception of those belonging to the PPM1 class. The substrate for the phosphatase assay was generated by the kinase reaction using commercially available CaMKII with AutoCaMtide3 in the presence of [␥-32 P]ATP. The final product of the phosphatase reaction, released 32 PO 4 , was determined by the difference between filter-bound substrate with total incorporation (starting control) and the residual 32 P following exposure to lysate (experimental sample). The data in Fig. 1 were calculated after normalization of the activity of each sample to its total protein concentration. An increase in total phosphatase activity was observed as a result of PPM1F-B⅐HA transfection. Furthermore, in the presence of 10 M OA and 10 M CaM-KIINtide, the lysate from cells overexpressing PPM1F-B⅐HA showed 45% released 32 PO 4 after 2.5 min as compared with 0% from vector-transfected cells, indicating that PPM1F is directly or indirectly responsible for the dephosphorylation of 32 P-AutoCaMtide3. This type of analysis was used to test phosphatase activity of various full-length and deletion mutant clones of PPM1F generated during the course of this investigation (data summarized in Table I), and expression of each protein was verified by Western blot analysis of cell extracts.
Other studies have identified an OA-resistant/Mg 2ϩ -dependent phosphatase in the extracts of different cell types able to dephosphorylate CaMKII (7,8,14). To address whether PPM1F was dephosphorylating the peptide directly or acting through another phosphatase in the lysate, PPM1F-B⅐HA was immunoprecipitated by using anti-HA antibody, and the immunoprecipitate was tested for phosphatase activity in vitro in the presence of 10 M OA. The PPM1F-B⅐HA immunoprecipi-tate released 70% of the incorporated radioactivity, whereas there was no radioactivity released with the vector sample (Fig.  2), indicating that PPM1F is directly responsible for the removal of 32 PO 4 from AutoCaMtide3. These data therefore show that PPM1F can function as a phosphatase when overexpressed in mammalian cells and support the possibility that the phospho-Thr-286 of CaMKII␣ could be a target for dephosphorylation by this phosphatase.
In Vitro Dephosphorylation and Deactivation of CaMKII by PPM1F-The autophosphorylation of the Thr-286 residue of CaMKII␣ has been shown to play an important role in the activation process of this kinase (46,47), and the dephosphorylation of this residue by a variety of phosphatases has been demonstrated to be a critical step in regulating CaMKII activity (5-7). We therefore evaluated the ability of PPM1F to specifically dephosphorylate the Thr-286 residue of the CaMKII polypeptide by utilizing Western analysis with antibody that specifically recognizes the Thr(P)-286 modified form of the protein. To eliminate the possibility that the two amino acid difference in PPM1F-A may result in aberrant phosphatase function toward CaMKII, we used a PPM1F cDNA that we had isolated earlier from the human ovarian cancer cell line SKOV3 and had found to be a 100% match to KIAA0015. When the phosphatase activity of the newly cloned PPM1F was tested along with PPM1F-A in the in vitro phosphatase assay, the activities of PPM1F and PPM1F-A were essentially the same, 9.42 and 11.7 pmol 32 P/min/mg total cellular protein, respectively. To facilitate detection and IP of PPM1F, an N-terminal Myc epitope-tagged PPM1F (Myc⅐PPM1F) was generated. Deletions in PPM1F were generated to determine the regions of the phosphatase that may be required for proper CaMKII regulation. Comparison of sequence and structural domains of PPM1F with the crystallized PP2C␣ (PPM1A) indicated that amino acids within 140 -454 of PPM1F were conserved. By taking this into consideration, an N-terminal deletion mutant Myc⅐PPM1F⌬1-148 was made by using a convenient EcoRI restriction site. This mutant corresponds to a similar deletion made in hFEM-2 that was shown to enhance the apoptotic effect associated with its overexpression in HeLa cells (30). C-terminal deletion mutants Myc⅐PPM1F⌬419 -454 and Myc⅐PPM1F⌬422-454 were generated by PCR cloning and were included in this analysis based on the absence or presence of the phosphatase activity of these two proteins, respectively, toward 32 P-AutoCaMtide3 (see Table I). According to sequence homology analysis, the ⌬419 -454 of PPM1F is quite similar, although not identical, to the ⌬413-450 mutant (referred to as CaMKP-(1-412)) generated in rCaMKPase (48). In addition to deletion mutants, a catalytically inactive full-length version of Myc⅐PPM1F (Myc⅐PPM1F R326A), which was modeled on the identical missense mutation made in hFEM-2 (30), was gener-ated by converting the highly conserved Arg-326 to an alanine residue, and this mutant served as control for nonenzymatic effects resulting from overexpression of this protein.
HEK293T cells were transiently transfected separately with these plasmids as well as with mCaMKII␣. The CaM kinase within the lysate from CaMKII-transfected cells was activated under limiting autophosphorylation conditions to allow for the specific phosphorylation of Thr-286 (49,50). KN-93, a specific inhibitor of CaMKII, was included to prevent further activation of the kinase. Aliquots of activated mCaMKII were then added to the phosphatase reaction containing 10 M OA and 2 mM MnCl 2 followed by the addition of extracts prepared from cells transfected with Myc⅐PPM1F. The reaction mix was incubated for 30 min at 23°C to allow for the dephosphorylation of CaM kinase by PPM1F. Duplicate samples, terminated with the addition of SDS loading buffer, were pooled for Western blot analysis with anti-phospho-CaMKII␣/␤ (Thr-286/Thr-287) and anti-CaMKII␣-specific antibodies. In the absence of PPM1F, Thr(P)-286 polypeptide was detected (Fig. 3, lane 2), but only if exogenous CaMKII␣ was present in the extract because the endogenous CaM kinase was not detectable (Fig. 3, lane 1). Following exposure to PPM1F (Fig. 3, lanes 3 and 6) or to the active ⌬422-454 mutant (Fig. 3, lane 8), the levels of Thr(P)-286 were undetectable even though equivalent amounts of CaMKII were present. In the case of exposure to catalytically inactive mutants of PPM1F such as R326A (Fig. 3, lane 4) and ⌬419 -454 (Fig. 3, lane 7), the levels of autophosphorylated CaM kinase were unaffected. These results indicate that PPM1F is able to specifically remove the phosphate from the Thr-286 of CaMKII and that this process requires a catalytically active PPM1F.
Because PPM1F was found to dephosphorylate CaM kinase, we assessed whether this had an effect on kinase activity. We exogenously activated mCaMKII␣ and subjected it to phosphatase treatment with PPM1F in vitro in a manner similar to that described above. Following the phosphatase reaction, ␤-glycerophosphate, a nonspecific phosphatase inhibitor shown to inhibit the rCaMKPase (33), was added to inhibit further phosphatase activity during the remainder of the assay. (The inclusion of this inhibitor was found to be necessary based on preliminary results indicating that PPM1F was dephosphorylating the product of the CaM kinase reaction.) We determined that 80 mM ␤-glycerophosphate (as the final concentration in the CaMKII reaction) could inhibit nearly 100% of PPM1F activity (data not shown) and had no detrimental effect on CaM kinase activity (data not shown) as reported previously (22,51). Aliquots of the phosphatase-treated mCaMKII were tested for autonomous kinase activity by the in vitro CaMKII assay in the presence of KN-93 and OA. In the absence of PPM1F (Fig. 4, CaMKII/Vec2 extract), the level of CaM kinase activity was 159 ϫ 10 2 pmol 32 P/min/mg greater than background; however, this level was reduced to 25 ϫ 10 2 pmol 32 P/min/mg, nearly that of background, upon treatment with PPM1F (Fig. 4, CaMKII/PPM1F extract), indicating that PPM1F was able to deactivate effectively the autophosphorylated CaMKII. The failure of R326A to affect CaM kinase activity indicates that the effect of PPM1F is not because of the phosphatase blocking CaMKII access to its substrate but rather because of its enzymatic activity. Intracellular Binding of CaMKII by PPM1F-A variety of proteins has been shown to interact with CaMKII such as ␣-actinin-1 (38), syntaxin 1A (52), and the NR2A subunit of N-methyl-D-aspartate receptor (53). In particular, the catalytic subunit of protein phosphatase 2A has been found to coimmunoprecipitate with CaMKII from rat ventricular myocyte lysates (54), demonstrating a physical association of CaM kinase with one of its regulating phosphatases. The ability of PPM1F to dephosphorylate and deactivate CaMKII in vitro suggests that the two proteins may also interact. To determine whether this interaction could occur intracellularly, we performed co-IP experiments with mCaMKII␣ and Myc⅐PPM1F (see Fig. 5). In addition, several deletion mutants of PPM1F were assessed to identify the region responsible for binding. Because there are no amino acid sequences in PPM1F that match those identified in other proteins to be responsible for CaMKII binding (38,52,53), we initially utilized the mutants described earlier. Fulllength as well as mutants of Myc⅐PPM1F were transiently cotransfected with mCaMKII␣ into HEK293T, and immunoprecipitates were generated using anti-Myc antibody under stringent wash conditions. The presence of CaMKII was determined by Western blot analysis of immunoprecipitates with anti-CaMKII␣ antibody. Exogenous CaM kinase was found to co-IP with full-length PPM1F (Fig. 5, lanes 3 and 6), whereas we did not detect binding of an endogenous protein to the phosphatase (Fig. 5, lane 1), indicating that PPM1F was interacting with CaMKII␣. As for identifying the binding domain of PPM1F, we began our analysis by investigating whether the N terminus of PPM1F played a role in binding to CaMKII based on the uniqueness of this region as well as on the finding that POPX2 (identical to PPM1F) bound to ␤1-PIX through this domain (31). To test this hypothesis, we examined the ⌬1-148 mutant in co-IP experiments and found that this mutant was able to bind CaM kinase (Fig. 5, lane 4). Initial attempts to construct a truncated protein with only the N terminus of PPM1F failed to generate stable protein (data not shown). We therefore assessed the ⌬148 -397 mutant, which was devoid of most of the conserved phosphatase domain but contained the entire N-terminal region, to determine whether a protein consisting mainly of this domain could bind to CaMKII. Even though the ⌬148 -397 mutant protein was relatively unstable, it reproducibly bound CaM kinase to a markedly enhanced extent (Fig. 5, lane 5), i.e. at a level similar to that of the highly stable ⌬1-148 mutant. Taken together, these results indicate that neither the N terminus nor an intact phosphatase domain is solely required for the binding of PPM1F to CaMKII.
In addition, the data suggest that the C-terminal domain, which is present in both of these mutants, may be involved. To examine this possibility specifically, we generated two mutants, ⌬411-454 and ⌬300 -454, with relatively large deletions FIG. 4. In vitro deactivation of CaMKII by PPM1F. HEK293T cells were transiently transfected with 10 g of plasmid for the following sequences: mCaMKII␣, Myc⅐PPM1F, Myc⅐PPM1F R326A, or vector (Vec) plasmids. In vitro activation of mCaMKII␣ and its dephosphorylation by Myc⅐PPM1F were performed as described in Fig. 3 with the exception that the phosphatase reactions were terminated with the addition of ␤-glycerophosphate (final concentration of 80 mM in the kinase reaction). Assay for autonomous CaMKII activity was performed with aliquots of the phosphatase-treated CaM kinase for 3 min at 30°C in a kinase reaction mixture containing 10 M OA and 50 M KN-93. Bars for each condition represent the average of three independent experiments assayed in duplicate; the error bars represent the range of values. within the C terminus of PPM1F. Both mutants failed to bind mCaMKII efficiently under these conditions (Fig. 5, lanes 8 and  9); however, the result may be due, at least in part, to the low level of mutant proteins in the lysates as compared with fulllength protein. Several attempts were made to use equivalent amounts of Myc⅐PPM1F protein by adding increasing quantities of lysate containing only the mutant protein to the co-IP reaction; however, these experiments were unsuccessful. Therefore, we could not properly assess the involvement of the C-terminal domain of PPM1F in binding to CaMKII. During these experiments, we did detect a faint band corresponding to CaMKII upon prolonged exposure of the Western blot, suggesting that a potentially weaker binding site may exist in the N terminus of PPM1F because this region was present in both of the C-terminal truncation mutants. The catalytically inactive R326A mutant (Fig. 5, lane 7) confirmed the finding made with the ⌬1-148 mutant that a functional phosphatase was not required for binding. Several other mutants of PPM1F were evaluated by this method to identify the CaMKII binding domain of PPM1F, and the results of those experiments along with the results from in vitro phosphatase assays for these mutants are summarized in Table I. Overall, the results of these experiments demonstrate that full-length PPM1F is able to interact physically with CaMKII␣ under stringent conditions when both proteins are overexpressed.
PPM1F Inhibits Intracellular Phosphorylation of a CaMKIIspecific Substrate-Because PPM1F was found to interact with CaMKII within the cell, we hypothesized that PPM1F could regulate the activity of the endogenous kinase. Our initial attempts to evaluate the effect of transiently overexpressed PPM1F on autonomous CaMKII activity were unsuccessful; therefore, we took an indirect approach by evaluating the effect of PPM1F on CaMKII-specific phosphorylation of the known endogenous substrate vimentin. It has been reported that vimentin is phosphorylated by CaMKII in the rat fibroblast cell line 3Y1 (55) and that the Ser-82 site of this intermediate filament was found to be a unique phosphorylation site for CaMKII (56). Furthermore, the phosphorylation state of vi-mentin has been shown to be affected by OA indicating that phosphatases are involved in its regulation (57).
To determine whether our phosphatase could inhibit CaM kinase activity within the cell, we evaluated the effect of PPM1F on the phosphorylation state of the CaMKII-specific site Ser-82 of vimentin as detected by an antibody (MO82) that solely recognizes the phosphorylated form of this residue (43). For these experiments, we used PPM1F-B⅐HA, based on previous experiments showing the functionality of this construct in NIH 3T3 (data not shown), subcloned into a vector expressing GFP and containing the Pac gene. This construct (pPPM1F-B⅐HA/IGPac) as well as parent vector were separately transfected into NIH 3T3, and 18 h post-transfection, the cells were subcloned into selection medium containing 4 g/ml puromycin. The surviving population, nearly 80% of which was positive for GFP, was treated with 1 M ionomycin, a Ca 2ϩ ionophore known to increase intracellular calcium levels, to activate the endogenous CaMKII, thereby inducing Ser-82 phosphorylation. Because ionomycin is a nonspecific activator of CaM kinases, KN-93 pretreatment was used in preliminary experiments to identify CaMKII as the kinase responsible for this phosphorylation (data not shown). Western blot analysis of extracts (see "Experimental Procedures") was used to identify the levels of Ser(P)-82 and total vimentin protein in each sample by staining with the MO82 antibody or anti-vimentin antiserum, respectively. As seen in Fig. 6A, ionomycin treatment resulted in a greater than 2-fold increase in the phosphorylation of vimentin at Ser-82; however, this level was reduced to the uninduced level as a result of PPM1F overexpression. This experiment was repeated three times, and the band intensities were quantified using a densitometer. The chart values (Fig.  6B) represent the average ratio of the optical density values for the MO82 band to the optical density values for the corresponding total vimentin band. Because of differences in binding affinities of the two antibodies as well as in the exposure intensities, these results are intended only to reflect visual differences observed between samples, and the comparison to total vimentin is to serve as a means of normalization and not as a quantitative assessment of the amount of phosphorylated vimentin. For this reason, the optical density ratio value for the untreated vector sample can be greater than or equal to 1. As seen from these results, PPM1F was consistently able to reduce the level of Ser-82 phosphorylation induced by ionomycin treatment, suggesting that PPM1F was inhibiting CaMKII. However, an alternative explanation is that PPM1F is directly dephosphorylating vimentin. To rule out this possibility, we assessed whether PPM1F would block the phosphorylation of vimentin by a cotransfected constitutively active CaMKII mutant (mCaMKII␣1-290). As seen from Fig. 7, PPM1F did not significantly affect Ser-82 phosphorylation by the constitutively active CaM kinase mutant, indicating that vimentin is not being dephosphorylated by PPM1F. Taken together, these results demonstrate that exogenous PPM1F is able to regulate the intracellular activity of the endogenous CaMKII. DISCUSSION CaMKII is a multifunctional kinase that is ubiquitously expressed (58) and is able to phosphorylate a variety of substrates in vitro (1), suggesting that this kinase may contribute to the transduction of Ca 2ϩ -activated signals in multiple cell types. CaM kinase activity has been detected in human fibroblasts and has been correlated with cell cycle regulation in these cells (18,22). However, the regulation of this kinase has not been examined previously in fibroblasts even though extensive studies have been done on neuronal tissue (7)(8)(9)(10)(11)(12)15). We cloned a cDNA (PPM1F) from HF that appeared to be the human homolog of the rat CaMKPase, suggesting that this gene may be involved in the regulation of CaM kinase in human cells. Because we were able to detect the co-expression of the two proteins in HF, we believed that CaMKII was a potential substrate for PPM1F in these cells. The dephosphorylation of 32 P-AutoCaMtide3 in vitro by overexpressed PPM1F provided the first indication that PPM1F was a CaM kinase phosphatase. As reported for rCaMKPase (33) as well as for hFEM-2 (30), PPM1F was found to dephosphorylate CaMKII in vitro; however, unlike the experiments with hFEM-2, the phosphorylated CaM kinase substrate used in the phosphatase reactions with PPM1F was generated under limiting autophosphorylation conditions (5 min at 5°C as compared with 1 min at 30°C), thereby restricting the majority of phosphorylating events to Thr-286. Furthermore, the dephosphorylation of this residue by PPM1F was monitored using the anti-Thr(P)-286 antibody.
In addition to dephosphorylating CaMKII, we found that PPM1F could deactivate this kinase in vitro. Unlike the find-ings made with rPP2C␣ (7) and rCaMKPase (33), the reduced level of CaMKII activity could not be attributed, in part, to a concomitant reaction product dephosphorylation by PPM1F in the assay because an effective phosphatase inhibitor was included. All the in vitro analyses suggested that CaMKII was a potential substrate for PPM1F intracellularly; therefore, we proceeded to assess the interplay of these two enzymes within the cells. This is the first report of the intracellular interaction of PPM1F with CaMKII, and we were able to show that the overexpression of PPM1F in fibroblasts resulted in a reduction of the intracellular phosphorylation of a CaMKII-specific substrate (i.e. vimentin), indicating that the endogenous CaMKII can be regulated by PPM1F in fibroblasts.
In the course of identifying PPM1F as a CaM kinase phosphatase, we generated a variety of mutants within the N-and C-terminal regions of PPM1F to determine the sequences required for CaMKII regulation. The N-terminal region of PPM1F (amino acids 1-145) has been suggested to be an inhibitory domain of hFEM-2 (30). Contrary to the results reported for hFEM-2, a deletion made to the N terminus of PPM1F, such as that generated in the ⌬1-148 mutant, resulted in a catalytically inactive protein even though stable protein was made. By generating a predicted protein structure of PPM1F based on a comparison of amino acid sequence and composition of PPM1F against the resolved crystal structure of human PP2C␣ (also termed PPM1A), we believe that the failure of ⌬1-148 mutant to function as a phosphatase may be explained by the fact that this truncation removes a highly conserved ␤-strand known to be essential in the architecture of the catalytic domain of PP2C␣ (59). Because this mutant was able to bind CaMKII, the deletion of this amino acid sequence did not appear to fully disrupt the protein structure. Our result with the N-terminal deletion mutant implies that the 157-454 mutant of hFEM-2 is most likely an inactive phosphatase mutant; therefore, the apoptotic effect associated with its overexpression is not a result of its catalytic function as proposed by Tan et al. (30).
Similar to findings made with the N-terminal mutants, large deletions in the C-terminal region of PPM1F also resulted in catalytically inactive proteins. A smaller deletion in this region, such as that generated in the ⌬419 -454 mutant, also disrupted phosphatase activity but did not affect binding because this mutant was found to interact with the kinase, indicating that the catalytic function of the phosphatase is not required for complex formation. Based on the ability of the ⌬148 -397 and ⌬419 -454 mutants to bind CaMKII, sequential deletions between residues 397 and 419 may provide the proper mutants to study the involvement of the C-terminal end in associating with CaM kinase. However, the interpretation would be complicated by the fact that these mutations would disrupt the catalytic function of the protein in addition to the potential loss of CaMKII binding. We believe that the overall data from both types of deletion mutants suggest that CaMKII may interact intracellularly with PPM1F near its C-terminal end and to a limited extent at its N terminus. This interaction may occur directly or through a common co-factor that binds to PPM1F, because our experimental conditions do not eliminate this possibility.
The intracellular binding and regulation of CaMKII by PPM1F imply that the interaction of these two proteins plays a role in cellular processes. The importance of this interaction has been suggested by the finding that the overexpression of hFEM-2, a gene identical to PPM1F, induces apoptosis in HeLa as well as NIH 3T3 cells (30). A similar cellular response has been observed with the exogenous expression of other PPM1 phosphatases such as PP2C␣ (60), PP2C␦ (61), and FIN13 (62) FIG. 7. Inability of PPM1F to reduce the phosphorylation of vimentin by constitutively active CaMKII mutant. NIH 3T3 cells were transiently cotransfected with 3 g of pCaMKII-(1-290) and 10 g of pPPM1F-B⅐HA/IGPac or vector; at 18 h post-transfection, cells were refed with medium containing 4 g/ml puromycin. Forty eight hours post-selection, cells extracts were prepared as described in Fig. 6 without ionomycin treatment. Replicate Western blots were generated and stained with anti-phospho-vimentin(Ser-82) antibody (MO82) (upper panel) and anti-vimentin antiserum (lower panel). Vec, vector. in other cell types. However, over the course of our analyses, we did not observe any detrimental effects associated with the overexpression of PPM1F in CHO-K1, HEK293T, NIH 3T3, or HeLa cells. 3 The pro-apoptotic effect reported for hFEM-2 might therefore be specific to subtle culture conditions. Although we have not been able to identify a specific cellular phenotype, our finding that the exogenous expression of PPM1F can inhibit CaMKII-specific phosphorylation of vimentin indicates that this phosphatase can regulate components of the cellular cytoskeleton through its interaction with CaMKII. Because vimentin has been shown to play an important role in cell contractility, migration, and proliferation (63), the modulation of its assembly by the interplay of CaMKII and PPM1F may contribute to the alterations in the cytoskeletal architecture during these cellular events. It has been shown that the phosphorylation of vimentin by CaM kinase in vitro results in the disassembly of reconstituted filament structures (64), and this effect has been observed in cultured astrocytes in response to exogenous expression of constitutively active CaMKII (43). These findings demonstrate the importance of CaM kinase in the restructuring of vimentin filaments, and because vimentin is phosphorylated by CaMKII in fibroblasts (55), the regulation of this kinase by PPM1F may allow for the proper maintenance of the filament structures in these cells.
In addition to CaM kinase, the p21 Cdc42/Rac-activated kinase PAK has been shown by Koh et al. (31) to be a substrate in vitro for POPX2, a gene identical to PPM1F. Furthermore, the overexpression of this phosphatase was found to reverse the loss of stress fibers induced by the exogenous expression of an active PAK mutant, indicating that POPX2 (PPM1F) can regulate the activity of this kinase within the cell. Interesting, PAK also phosphorylates vimentin, and this phosphorylation results in the disruption of filament formation both in vitro as well as within cells following the overexpression of the constitutively active kinase (65). Cell migration in fibroblasts requires activation of PAK1 (66), suggesting that this kinase may be affecting the organization of vimentin filaments and, consequently, may also be a potential substrate for PPM1F in these cells. The regulation of CaMKII and PAK1 by PPM1F may therefore serve as an effective signaling mechanism to alter the cytoskeletal architecture of fibroblasts in response to a variety of stimuli.