Identification and Functional Analysis of Two Ca -binding EF-hand Motifs in the B /PR72 Subunit of Protein Phosphatase 2A*

Protein phosphatase 2A (PP2A) is a multifunctional serine/threonine phosphatase that is critical to many cellular processes including cell cycle regulation and signal transduction. PP2A is a heterotrimer containing a structural (A) and catalytic (C) subunit, associated with one variable regulatory or targeting B-type subunit, of which three families have been described to date (B/PR55, B /PR61, and B /PR72). We identified two functional and highly conserved Ca -binding EF-hand motifs in human B /PR72 (denoted EF1 and EF2), demonstrating for the first time the ability of Ca to interact directly with and regulate PP2A. EF1 and EF2 apparently bind Ca with different affinities. Ca induces a significant conformational change, which is dependent on the integrity of the motifs. We have further evaluated the effects of Ca on subunit composition, subcellular targeting, catalytic activity, and function during the cell cycle of a PR72-containing PP2A trimer (PP2AT72) by site-directed mutagenesis of either or both motifs. The results suggest that integrity of EF2 is required for A/PR65 subunit interaction and proper nuclear targeting of PR72, whereas EF1 might mediate the effects of Ca on PP2AT72 activity in vitro and is at least partially required for the ability of PR72 to alter cell cycle progression upon forced expression.

Protein phosphatase 2A (PP2A), 1 an essential serine/threonine phosphatase present in all eukaryotic cells, is a multifunc-tional enzyme of fundamental importance in signal transduction that regulates a wide variety of cellular events, including cell cycle progression, development, transcription, translation, DNA replication, and viral transformation (reviewed in Refs. 1 and 2). This extraordinary functional diversity can be explained by the existence of several mechanisms by which PP2A activity, substrate specificity, and subcellular localization are regulated (reviewed in Ref. 1). One of the major ways to achieve this is by the interaction, through a scaffolding A/PR65 subunit, of the PP2A catalytic subunit (PP2A C ) with one of several regulatory B-type subunits. Although it has been suggested that a significant part of PP2A C can occur as a dimer with the A/PR65 subunit (PP2A D ) within cells (3), recent evidence argues that PP2A is an obligate trimer (PP2A T ) in vivo (4,5). Nevertheless, there are a number of well documented cases at variance with this central "dogma," such as the association of PP2A C with ␣4 without any other subunits (6), the complex formation of PP2A C , the BЈ subunit and cyclin G 2 , without the A subunit (7), and the association of PP2A D with some viral tumor antigens (8).
PP2A regulatory subunits are encoded by four multigene families, referred to as B, BЈ, BЉ, and Bٞ. All B-type subunits, with exception of Bٞ, share two motifs for A/PR65 subunit binding (9). Because Bٞ subunits lack these binding motifs, their status as real PP2A subunits needs further evaluation. Each B subunit is thought to confer a set of specific functional characteristics to the phosphatase. For example, the B subunits have been implicated in the regulation of cytoskeletal protein assembly (10,11), and in Drosophila S2 cells they mediate actions of PP2A on the mitogen-activated protein kinase signaling pathway (4,5). The BЈ subunits, on the other hand, mediate PP2A function in the Wnt signaling cascade (12)(13)(14) and are required for the protective function of PP2A against apoptotic cell death in S2 cells (4,5).
The functions of the BЉ subunit family are probably the least well understood. Five different BЉ isoforms have been identified in mammals: human PR72 and PR130 (the founding members of the family) (15), mouse PR59 (16), human PR48 (17), and the recently identified human G5PR (18). Although BЉ homologues have been described in plants (19), Xenopus laevis oocytes (20), Drosophila melanogaster (4,5) and Caenorhabditis elegans (protein CO6G1.5, GenBank accession number AAK32946), they are manifestly absent in yeast. PR72 and PR130 are splice variants generated from a single gene and share an identical C terminus. PR72 has a muscle-specific expression, whereas PR130 is ubiquitous (15). In an in vitro assay, PP2A T72 is the only PP2A trimer that specifically stimulates simian virus 40 large T-antigen-dependent origin unwinding, an essential step in the initiation of viral DNA replication (21). PR59 was iden-tified as interaction partner of the retinoblastoma-like p107 protein. Its overexpression results in G 1 /S arrest of cell cycle progression and coincides with increased amounts of hypophosphorylated and active p107 (16). PR48 was identified as a Cdc6-interacting protein (17). Cdc6 is an ATPase required for the initiation of DNA replication that primarily acts by recruiting the minichromosome maintenance complex to origins of replication (22). Recent results indicate, however, that PR48 is a partial clone that would better be renamed as "PR70," fitting with its real molecular mass (20). G5PR was identified as a germinal center-associated nuclear protein (GANP)-associated molecule by yeast two-hybrid screening (18). GANP carries DNA primase and minichromosome maintenance 3-binding activity and is thought to be involved in the regulation of DNA replication in activated germinal center B cells (23,24). Interestingly, G5PR also strongly interacts with the tetratricopeptide repeat domain of protein phosphatase 5, another eukaryotic serine/threonine phosphatase (18). In general, most of the functions of the BЉ subunits described so far argue for important roles of these proteins in the regulation of the G 1 /S transition of the cell cycle and the regulation of DNA replication during S phase.
Calcium is an important second messenger within cells. Highly regulated changes in the intracellular Ca 2ϩ concentration control biological processes as diverse as muscle contraction, fertilization, cell proliferation and division, gene transcription, and apoptosis (reviewed in Ref. 25). Many of these effects are modulated by multiple classes of Ca 2ϩ -binding proteins, some of which can in turn regulate multiple downstream effectors. Among the major intracellular Ca 2ϩ receptors are calmodulin (CaM) and S100 proteins. Both are members of a large class of so-called "EF-hand" Ca 2ϩ -binding proteins, which share a common Ca 2ϩ -binding helix-loop-helix motif, the conformation of which essentially determines biological function. Specific binding of Ca 2ϩ to the loop alters the conformation of the motif, involving the rearrangement of both helices in threedimensional space (reviewed in Ref. 26). In many Ca 2ϩ sensor proteins this conformational change exposes a hydrophobic effector protein-binding surface, which in turn allows them to bind to and regulate an array of secondary effector proteins (27).
Although calcineurin (also known as protein phosphatase 2B) and protein phosphatase 7 are the only serine/threonine phosphatases known so far to be regulated directly by Ca 2ϩ (28,29), several reports have suggested a role for PP2A in Ca 2ϩdependent signaling. PP2A has been shown to associate with the CaM-dependent kinase IV (30) and with the CaM-binding proteins striatin and S/G 2 nuclear autoantigen (also known as the Bٞ or PR93/PR110 subunits) (31). Moreover, PP2A has been implicated in the regulation of Ca 2ϩ -activated potassium channels by glucocorticoids (32,33), in the regulation of L-type Ca 2ϩ channels (34,35), the cardiac ryanodine receptor RyR2 (36), and the inositol 1,4,5-trisphosphate receptor signaling complex (37). Inversely, calcium ions have also been shown to regulate PP2A activity in G-protein-coupled receptor signaling (38) or PP2A (re)localization in the establishment of cell-cell contacts between epithelial cells (39) and during mast cell secretion (40). However, the underlying mechanisms remain elusive.
In this report we identify two Ca 2ϩ -binding EF-hand motifs in human BЉ/PR72, demonstrating for the first time, the ability of calcium ions to directly interact with and regulate PP2A. By a mutational analysis of both domains, we evaluate the effects of Ca 2ϩ on PP2A T72 subunit composition, subcellular targeting, catalytic activity, and function during the cell cycle. The results obtained suggest that both motifs serve different functions. 45 CaCl 2 (2.5 mCi/ml, 110.8 g of Ca 2ϩ /ml), [ 35 S]methionine, and protein G-Sepharose beads were obtained from Amersham Biosciences. Restriction enzymes and DNA-modifying enzymes were purchased from Fermentas. Calcium ionophore A23187, propidium iodide, and Hoechst 33342 were from Sigma. BAPTA-AM was supplied by Alexis Biochemicals; nocodazole was from Applichem. The bacterial expression vector for the His-tagged domain I of calpastatin was a generous gift of Dr. J. Elce (Cancer Research Laboratories, Department of Biochemistry, Queen's University, Kingston, Ontario, Canada). The protein was expressed and purified on nickel-agarose beads (Affiland) following standard procedures. Anti-EGFP antibodies were a kind gift of Dr. M. Beullens (Division of Biochemistry, Faculty of Medicine, Katholieke Universiteit Leuven, Belgium). Anti-PP2A C and anti-PR65 monoclonal antibodies were generously supplied by Dr. S. Dilworth (Department of Metabolic Medicine, Imperial College Faculty of Medicine, Hammersmith Hospital, London, UK). Pwo proofreading polymerase (used in all PCRs) was purchased from Roche Molecular Biochemicals.

Materials-
Site-directed Mutagenesis of EF-hands 1 and 2-Two point mutations were introduced in each EF-hand using a PCR-based method. For the introduction of a single point mutation into EF1, two separate PCRs were performed with PR72 cDNA as template: the first with 5Ј-ATAT-CATATGATGATCAAGGAAACATCTC-3Ј (start primer) and 5Ј-GTAT-CGAGACAGAGCGGCCTGGC-3Ј (reverse mutated primer) and the second with 5Ј-GCCAGGCCGCTCTGTCTCGATAC-3Ј (forward mutated primer) and 5Ј-TATAGGATCCCTATTCTTCATCCACTGATTG-3Ј (stop primer). The combined reaction products of these two PCRs were then used as template for a second amplification round with the start and stop primers only. The resulting PCR product was cloned into the EcoRV-digested pBluescript vector (Stratagene) and sequenced to confirm the introduction of the mutation. The second point mutation within EF1 was introduced very similarly with the start primer, 5Ј-GGTCGT-GATCAGTAGCTAGTTCCCAG-3Ј (reverse mutated primer), 5Ј-CTGG-GAACTAGCTACTGATCACGACC-3Ј (forward mutated primer), and the stop primer using the single EF1 point mutant as template. Together, the PR72 EF1 sequence 290 DTDHDLYISQADL 302 was changed into 290 ATDHDLYISQAAL 302 . For mutation of EF2, the first mutation round was performed with the start primer, 5Ј-CATAGAAGTACTC-CAATGCATACATGGAGAG-3Ј (reverse mutated primer), 5Ј-CTCTC-CATGTATGCATTGGAGTACTTCTATG-3Ј (forward mutated primer) and the stop primer using PR72 as template. The second point mutation was generated very similarly with the start primer, 5Ј-GTCTCCATC-CACAGCCATGCAGCGG-3Ј (reverse mutated primer), 5Ј-CCGCTG-CATGGCTGTGGATGGAGAC-3Ј (forward mutated primer) and the stop primer using the single EF2 point mutant as template. In this way the wild-type EF2 sequence 364 DVDGDGVLSMYEL 376 was eventually changed into 364 AVDGDGVLSMYAL 376 . The double mutation of EF1 and EF2 was generated via four consecutive mutation rounds using the same primers and the same PCR-based method.
Expression and Purification of Recombinant PR72 Polypeptides-Wild-type PR72, N-and/or C-terminal truncations of wild-type PR72, and the single or double EF hand mutants thereof were cloned into pET15b, pET3C (Novagen), pGEX-4T2 or pGEX-4T3 (Amersham Biosciences) following standard molecular biology procedures. The resulting plasmids were transformed into BL21-pLys(RP) Escherichia coli bacteria for protein expression. Inductions were performed with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 2 h 30 min at 30°C. Bacterial pellets were lysed by sonication in lysis buffer (50 mM Tris⅐HCl, pH 8.0, 2 mM EDTA, 1 mM dithiothreitol) supplemented with 2 mg/ml lysozyme, 10 g/ml leupeptin, 1 mM benzamidine, 10 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride. Inclusion bodies were purified (41) and solubilized in 7 M guanidinium hydrochloride. After dialysis against buffer B (200 mM Tris⅐HCl, pH 8.2, 500 mM NaCl), the majority of the proteins remained soluble. 45 Ca 2ϩ Overlay Assay-Ca 2ϩ binding to the recombinant (fusion) proteins was measured after transfer to Immobilon P membranes (42). Briefly, membranes were washed overnight at room temperature in nominally Ca 2ϩ -free incubation medium containing 10 mM imidazole, pH 7.0, 60 mM KCl, and 1 mM MgCl 2 . Binding was carried out in the same medium supplemented with 1.5 Ci/ml 45 CaCl 2 for 10 min at room temperature. Membranes were washed during 2 min in 50% ice-cold ethanol/water mixture and air-dried. Bands were visualized by autoradiography.
Fluorescence Spectroscopy-Protein concentrations of purified recombinant proteins, comprising amino acids (aa) 262-449 of PR72, were evaluated by absorbance measurement at 280 nm and by the BCA quantification method (Pierce). Equal amounts of PR72 aa262-449 or PR72 EF(1 ϩ 2)mut aa262-449 were diluted in buffer B and excited at 295 nm. The fluorescence spectra between 300 and 410 nm were recorded in a Photon Technology International spectrofluorometer in the absence or the presence of different amounts of CaCl 2 buffered by 1 mM EGTA. Free Ca 2ϩ concentration was calculated with the CaBuf program (available at ftp.cc.kuleuven.ac.be/pub/droogmans/cabuf.zip).
GST-Pull Downs-[ 35 S]Methionine-labeled proteins were obtained from pBluescript vectors containing the coding regions of PR72, PR72 EF1mut, PR72 EF2mut, or PR72 EF(1 ϩ 2)mut, using the TNT-coupled rabbit reticulocyte lysate system (Promega) with the appropriate RNA polymerase (T3 or T7). The GST-PR65␣ subunit of PP2A (PR65-GST) and the free GST protein were produced in E. coli BL21-pLys cells following standard procedures, and purified on glutathione-Sepharose beads (Amersham Biosciences) according to the manufacturer's instructions. The GST pull-down binding reactions contained 20 l of [ 35 S]methionine-labeled proteins, 1 g of GST or PR65-GST, 20 l of prewashed glutathione-Sepharose beads, and NENT 100 buffer (20 mM Tris⅐HCl, pH 7.4, 1 mM EDTA, 0.1% Nonidet P-40, 25% glycerol, 100 mM NaCl) to a final volume of 500 l. If appropriate, either 2 mM EGTA or 4 mM CaCl 2 was added to the reaction mix. Incubation was done for 4 h at 4°C on a rotating wheel. The beads were washed five times with 1 ml of NENT 300 (NENT with 300 mM NaCl) containing 2 mg/ml bovine serum albumin and either 2 mM EGTA or 4 mM CaCl 2 . Bound proteins were eluted by addition of 20 l of SDS sample buffer and boiling. The eluted proteins were analyzed by SDS-PAGE and imaged using an Amersham Biosciences PhosphorImager.
For yeast two-hybrid experiments, GAL4 binding and activation domain fusion proteins were generated by cloning PCR fragments of full-length PR72 (wild-type or mutated in either or both EF-hands) and of several PR72 deletions in the pGBT9 and pGAD424 vectors (Clontech), respectively. The PR72 deletion fragments used encompassed aa1-99, aa1-219, aa1-358, aa123-473, aa353-529, and aa219 -529. All transformations were performed in the PJ69 -4A Saccharomyces cerevisiae strain (46) harboring adenine, ␤-galactosidase, and histidine reporter plasmids under the control of GAL4 upstream-activating sequences. For the evaluation of positive interactions, only the expression of the adenine and ␤-galactosidase reporters were assayed.
EGFP Experiments-PCR fragments of PR72, PR72 EF1mut, PR72 EF2mut, and PR72 EF(1 ϩ 2)mut were cloned into the SmaI restriction site of pEGFP-C1 (Clontech). 1 g of each plasmid was transfected into COS7 cells grown on a glass coverslip in a 6-cm dish (Nunc) using FuGENE 6 transfection reagent. 24 -36 h after transfection cells were washed in phosphate-buffered saline and fixed in ice-cold methanol for 20 min. Following the addition of 3 M Hoechst 33342 for 10 min to stain the nuclei, cells were briefly rinsed in water, mounted in mounting medium (Sigma), and examined by a fluorescence microscope (Diaplan, Leitz, Germany) equipped with a digital camera (DC200, Leica Microsystems). For anti-EGFP immunoprecipitations, transfected cells were lysed in Tris-buffered saline supplemented with 0.1% Nonidet P-40. Before addition of the anti-EGFP antibody, lysates were precleared with 10 l of protein G-Sepharose. Immunoprecipitates were washed four times in lysis buffer and once with 0.1 mM LiCl before addition of SDS sample buffer, boiling, SDS-PAGE, and Western blotting. Blots were developed with a mixture of anti-PP2A C and anti-PR65 monoclonal antibodies using the enhanced chemiluminescence detection system (Amersham Biosciences).
Measurement of PP2A Activity-Protein phosphatase assays were performed with 32 P-labeled phosphorylase a as the substrate in the absence of protamine stimulation, essentially as described (15). PP2A T72 , purified from rabbit skeletal muscle (15), was preincubated with calpastatin (an inhibitor of the m-calpain protease) and different amounts of CaCl 2 for 10 min at 30°C. Subsequently, different amounts of EGTA were added together with the substrate. This mixture was further incubated for 10 min at 30°C. The reactions were stopped by trichloroacetic acid precipitation, and the amount of free 32 P-labeled phosphate was counted. Similar experiments were performed by including the Ca 2ϩ and/or EGTA concentrations directly in the phosphatase assay. Because this approach basically resulted in the same data, this indicated that Ca 2ϩ /EGTA effects are "instantaneous," and prior incubation of Ca 2ϩ and/or EGTA with the enzyme is not required.
FACS Analysis-Asynchronously growing U2OS cells were transfected with pEGFP-C1, PR72-EGFP, PR72 EF1mut-EGFP, PR72 EF2mut-EGFP, or PR72 EF(1 ϩ 2)mut-EGFP in four 6-cm dishes per plasmid. 24 h after transfection, cells were trypsinized, seeded into 10-cm dishes, and allowed to grow for another 24 h before FACS analysis. If nocodazole (1 g/ml) was used, it was added at this point for another 16 h before FACS analysis. If BAPTA-AM (10 M) was used, it was added 6 h before the addition of nocodazole. Cells were fixed for 5 min in 4% paraformaldehyde at room temperature to preserve the EGFP signal. After washing, the cell pellet was incubated in 0.5 ml of phosphate-buffered saline containing 100 g/ml propidium iodide and 0.1% RNase for at least 1 h at room temperature. The samples were analyzed with a Beckman Instruments Coulter Epics XL flow cytometer (Analis) on FL1 (for EGFP) and FL3 (for propidium iodide) using standard procedures and the System II TM software (Analis) for quantification.

RESULTS
Two Conserved EF-hands within PR72 Bind Ca 2ϩ with Different Affinities-Analysis of the primary amino acid structure of the BЉ family members revealed the presence of two well conserved EF-hand domains (termed EF1 and EF2) (Fig. 1A). These motifs are well known Ca 2ϩ -binding domains and are present, very often in tandem, in many Ca 2ϩ -binding proteins (with calmodulin being the most renown example) (reviewed in Ref. 26). Given their high degree of conservation, within the family as well as throughout evolution, and given their (almost) perfect match with the known consensus sequence (Fig. 1A), it was not unlikely to predict that the BЉ members are genuine Ca 2ϩ -binding proteins. To check this, we used the human BЉ/ PR72 protein as a "model" for the other family members, and we performed all further experiments with this particular BЉ isoform.
To test whether PR72 could bind Ca 2ϩ in vitro, a 45 Ca 2ϩ overlay assay was performed on several recombinant PR72derived proteins, expressed and purified from E. coli bacteria (see "Experimental Procedures") (Fig. 1B). The results show that full-length PR72, as well as GST fusion proteins of PR72 aa238 -358 (comprising EF1), of PR72 aa353-410 (comprising EF2), and of PR72 aa238 -420 (comprising both EF1 and EF2), but not GST alone bind Ca 2ϩ . Moreover, GST-PR72 aa353-410 (comprising only EF2) seemed to bind significantly more Ca 2ϩ than GST-PR72 aa238 -358 (comprising only EF1), although approximately equal amounts of both proteins were loaded (Fig. 1, B and C, compare lane 6 and lane 8). This could suggest that the affinity of Ca 2ϩ for EF2 is much higher than for EF1.
To ensure that Ca 2ϩ binding to full-length PR72 or the fragments thereof indeed occurred via the proposed EF-hands, several mutations known to destroy Ca 2ϩ binding to these motifs were introduced either in the full-length protein or in the fragments thereof (see "Experimental Procedures"). Briefly, in each motif the aspartate at position 1 plus the aspartate or glutamate at position 12 were changed into alanines. Similarly, these mutated proteins were expressed and purified from bacteria and subjected to 45 Ca 2ϩ overlay (Fig. 1B). The data show that mutation of EF1 abolishes Ca 2ϩ binding to GST-PR72 aa238 -358 (comprising only EF1) (lane 7), and similarly, mutation of EF2 abolishes Ca 2ϩ binding to GST-PR72 aa353-410 (comprising only EF2) (lane 9). However, single mutation of EF1 within GST-PR72 aa238 -420 (comprising both EF1 and EF2) has only minor effects on Ca 2ϩ binding to this protein, whereas mutation of EF2 almost completely destroys Ca 2ϩ binding (Fig. 1B, lane 3 and lane 4). This clearly demonstrates that both EF motifs bind calcium ions with different affinities.
Ca 2ϩ Binding Induces Conformational Changes in PR72-It is well known that Ca 2ϩ binding can result into drastic conformational changes within EF-hand proteins, changing the relative position of helix E and helix F from a closed to a more open configuration (reviewed in Refs. 26 and 47). This property of Ca 2ϩ forms the basis for its regulatory capacity, because conformational changes within proteins often affect their biological activities.
To assess whether Ca 2ϩ binding to PR72 could result into conformational changes, measurements of the intrinsic tryptophan fluorescence of wild-type and mutated PR72 recombinant proteins were conducted in the presence of different amounts of calcium ions. Such fluorescence spectra are dependent on the micro-and/or macro-environment of the emitting Trp residues within the protein. Optimal results were obtained with a rather short PR72 fragment comprising both EF-hand motifs (aa262-449). This polypeptide contains only three Trp residues, two of which are in the very near vicinity of each EF-hand (one is found three aa N-terminally from EF1 and one 5 aa N-terminally from EF2). Using this fragment obviously diminishes the risk that (small) changes in fluorescence of only a few Trp residues in the near vicinity of the EF motifs may become undetectable because of the intrinsic fluorescence of a lot of other Trp residues of which the physicochemical environment is not changed by Ca 2ϩ binding. The data show that addition of Ca 2ϩ to the PR72 aa262-449 apoprotein, in any of the concentrations used, leads to a significant increase in fluorescence intensity ( Fig.  2A), whereas this is not the case for the PR72 aa262-449 protein in which both EF-hands are mutated (Fig. 2B). This is indicative for a conformational change induced by binding of calcium ions to the EF hands. Moreover, the spectra of PR72 aa262-449 EF(1 ϩ 2)mut and PR72 aa262-449 in the absence of Ca 2ϩ are very similar (overlay Fig. 2C), suggesting that the introduction of the four point-mutations per se does not affect the overall protein conformation of the wild-type apoprotein.
Effects of Ca 2ϩ on the Interaction of PR72 with the A/PR65 Subunit-The A/PR65 subunit binding domain of PR72 was determined by analysis of the interaction of several deletion mutants of PR72 with A/PR65␣ in yeast two-hybrid assays. This yielded a rough estimation of the interaction domain. The smallest PR72 fragment tested was still able to interact with the A subunit, composed of amino acids 219 -473 (results not shown). This domain contains about 80% of the proposed A Subunit Binding Domain 1 (ASBD1, aa197-302) and the complete ASBD2 (aa342-399) (9). Remarkably, also both Ca 2ϩbinding motifs are located within these ASBDs, suggesting that calcium ions might affect the interaction of PR72 with the A/PR65 subunit (and consequently, with the core enzyme).
To test this hypothesis, we performed mammalian two-hybrid assays in COS7 cells. This system has the advantage that interactions can be easily quantified and/or evaluated in the presence of extracellular stimuli, such as calcium ionophore or BAPTA-AM treatment. In the absence of any stimulus, the A/PR65 subunit interaction is observed with wild-type PR72 and PR72 EF1mut but not with PR72 EF2mut or PR72 EF(1 ϩ 2)mut (Fig. 3A). Identical results were obtained in the yeast two-hybrid system (results not shown), suggesting that the integrity of EF2 and therefore its Ca 2ϩ -binding capacity are vital for binding to A/PR65. Furthermore, the comparison of A subunit binding to BЉ/PR72 with that to B/PR55␣ and BЈ/ PR61␥1 in the mammalian two-hybrid system reveals that BЉ/PR72 relatively shows the strongest interaction with A/PR65, whereas the interaction with B/PR55␣ is the weakest (Fig. 3A). Treatment of the cells with 5-10 M BAPTA-AM, a cell-permeable Ca 2ϩ chelator, resulted in a slight (albeit nonspecific) decrease of the observed PR72-PR65 interaction, because a parallel decrease was also observed for the PR55-PR65 and the PR61-PR65 interactions upon BAPTA-AM addition (data not shown). Similarly, treatment of the cells with 2-10 M A23187, a calcium ionophore, led to inconclusive results, because in this case an overall inhibition of the transcriptional response of the reporter genes (luciferase as well as ␤-galactosidase) was observed (data not shown).
To confirm the former data in vitro, GST pull-down assays were performed with GST-PR65␣ and in vitro translated and radioactively labeled PR72 protein or the EF-hand mutants thereof, either as such (without any special treatment), in the presence of a Ca 2ϩ chelator (2 mM EGTA) or in the presence of 4 mM CaCl 2 . Irrespective of the binding conditions, a specific interaction of GST-PR65␣ was observed with PR72 and PR72 EF1mut but not with PR72 EF2mut or PR72 EF(1 ϩ 2)mut (Fig. 3B). Moreover, in the presence of extra calcium ions, the interaction of PR72 and PR72 EF1mut with GST-PR65␣ was slightly increased compared with the binding in the presence of a calcium chelator (Fig. 3B). Taking the limitations of the use of chelators into account, the data suggest that Ca 2ϩ binding to EF2, a high affinity Ca 2ϩ -binding site, is a prerequisite for interaction of PR72 with the A subunit.
Effects of Ca 2ϩ on Phosphatase Activity of PP2A T72 in Vitro-In order to test the effect of CaCl 2 on phosphatase activity, different (buffered) concentrations of Ca 2ϩ were tested on purified PP2A T72 in an in vitro assay with phosphorylase a as the substrate. Calpastatin was added in order to block any residual m-calpain activity, known to be present in some PP2A T72 preparations (15). It should be noted that the standard purification of PP2A T72 from rabbit skeletal muscle is performed in buffers containing 1 mM EGTA all through the procedure (15). We therefore presume that EF1 (the low affinity binding site) may be (partially) in a Ca 2ϩ -free state, whereas EF2 (the high affinity binding site) is likely still loaded with Ca 2ϩ , because, according to our data, this is necessary for the interaction with PP2A D . Only high concentrations of Ca 2ϩ seem to have an inhibitory effect on the phosphorylase phosphatase activity of PP2A T72 in vitro (Fig. 4A). These inhibitory effects are therefore probably mediated by the low affinity binding site EF1. In the absence of any added Ca 2ϩ , EGTA is without effect or only slightly stimulatory (Fig. 4B). This slight stimulation could correlate with some dissociation of PR72 from the trimer because it is known that PP2A T72 has a lower specific activity than PP2A D with phosphorylase a as the substrate (48). However, because the effect is so small, these data seem to confirm that EGTA, even at very high concentrations, is hardly able to affect the interaction of PR72 with the core enzyme. In the presence of both Ca 2ϩ and EGTA, complex titration curves are observed (Fig. 4C); lower Ca 2ϩ concentrations stimulate the activity maximally 2-fold, depending on the EGTA concentration, whereas inhibition by the higher Ca 2ϩ concentrations seems to be more pronounced at higher EGTA concentrations. To investigate the individual contribution of EF1 and EF2 on the inhibitory as well as on the stimulatory effects of Ca 2ϩ in the in vitro phosphatase assay, we tried to reconstitute PP2A T72 from purified PP2A D and the bacterially expressed and purified PR72, PR72 EF1mut, PR72 EF2mut, and PR72 EF(1 ϩ 2)mut proteins, but these reconstitution experiments failed.
Effects of Ca 2ϩ on the Subcellular Localization of PR72-The localization of PR72 and the putative Ca 2ϩ effects on this phenomenon were initially evaluated by expression of EGFP fusion proteins of wild-type PR72 and of the PR72 EF-hand mutants in COS7 cells. The data show that wild-type PR72 and PR72 EF1mut are predominantly nuclear, whereas PR72 EF2mut and PR72 EF(1 ϩ 2)mut are clearly excluded from the nucleus (Fig. 5A). Moreover, immunoprecipitations with anti-EGFP antibodies revealed co-immunoprecipitation of PP2A C and A/PR65 with PR72 and PR72 EF1mut but not with PR72 EF2mut or PR72 EF(1 ϩ 2)mut (Fig. 5B), nicely confirming our former observations. Treatment of the transfected cells with 5-10 M BAPTA-AM or 2-10 M A23187, however, failed to affect the distribution of EGFP-PR72 or any of the EGFP-PR72 mutants (data not shown). The presence of PR72 in the nucleus was further confirmed by cell fractionation of L6 cells (rat myoblasts), where the majority of endogenous PR72 is present in the nuclear fraction (Fig. 5C). FIG. 4. Effects of calcium ions on phosphorylase a phosphatase activity of PP2A T72 in vitro. PP2A T72 purified from rabbit skeletal muscle was appropriately diluted and assayed in the presence of calpastatin and different amounts of Ca 2ϩ (A) or EGTA (B) for 10 min at 30°C with 32 P-labeled phosphorylase a as the substrate. C, the enzyme was preincubated in the presence of calpastatin and different amounts of calcium for 10 min at 30°C. Different amounts of EGTA were added together with 32 P-labeled phosphorylase a, and the incubation was continued for another 10 min at 30°C before trichloroacetic acid precipitation and measurement of the free 32 P-labeled phosphate. The final EGTA and Ca 2ϩ concentrations in the assay mixture are indicated. FIG. 3. A, mammalian two-hybrid results in COS7 cells. PR72, the PR72 EF mutants, PR55␣, and PR61␥1 were fused to the DNA binding domain of GAL4 (pAB-GAL4-derived plasmids). PR65␣ was fused to the transactivating region of VP16 (pSNATCH-II-derived plasmids). The indicated combinations of both expression plasmids were transfected into COS7 cells together with the luciferase reporter plasmid pUAS-TATA-luc and the ␤-galactosidase vector pEF1-␤Gal as internal control. The indicated luciferase values represent the mean of at least three independent experiments and are normalized against the measured ␤-galactosidase values, and eventually against the mean of the luciferase values measured for the PR72-PR65 interaction, which was set to 100. B, GST pull-down assay with A/PR65␣-GST and 35 S-labeled PR72 or the PR72 EF-mutants. 35 S-Labeled proteins were produced by in vitro coupled transcription-translation in reticulocyte lysates. Bacterial recombinant GST or GST-PR65␣ immobilized on glutathione-Sepharose beads were incubated with the labeled proteins, in the presence of 2 mM EGTA or 4 mM CaCl 2 . After stringent washings, the resin-bound proteins were eluted with SDS-loading buffer, analyzed by SDS-PAGE, and visualized by autoradiography.
These remarkable results suggest that the integrity of EF2 is not only required for binding to the core enzyme but also for proper subcellular (nuclear) targeting of PR72. Apparently, PR72 cannot enter the nucleus unless incorporated within a trimeric PP2A complex. Moreover, these data suggest that withdrawal of BЉ/PR72 from the trimer does not necessarily lead to its degradation within living cells, which is in contrast with some observations done for the B/PR55 subunit (49). To ensure that the EGFP tag did not influence our observations, we also transiently overexpressed wild-type PR72, PR72 EF1mut, PR72 EF2mut, and PR72 EF(1 ϩ 2)mut in human U2OS cells from the pCEP4 vector, and in all cases the wildtype or mutant PR72 proteins were expressed (results not shown). This confirms that PR72 mutants unable to bind to the A/PR65 subunit are not rapidly degraded within cells.
Effects of Ca 2ϩ on the Ability of PR72 to Induce Cell Cycle Arrest-Similar to BЉ/PR48 (17) and BЉ/PR59 (16), we noticed that forced expression of BЉ/PR72 in U2OS cells leads to a G 1 /S phase arrest. This arrest became apparent by comparing the propidium iodide-stained DNA profiles of PR72-EGFP-transfected and pEGFP-C1-transfected cells by flow cytometry (results not shown). This effect was more pronounced after prior blockage of the dividing cells in G 2 /M by the addition of nocodazole, a spindle de-polymerizing agent (Fig. 6A). Interestingly, if the same experiment was repeated in PR72 EF1mut-EGFP, PR72 EF2mut-EGFP, and PR72 EF(1 ϩ 2)mut-EGFP expressing cells, the EF1 and EF(1 ϩ 2) mutants partially lost the ability to induce the G 1 /S arrest, whereas the EF2 mutant even generated a more pronounced G 1 /S arrest (Fig. 6A). These results suggest that the ability of PR72 to induce a G 1 /S cell cycle arrest is at least partially dependent on the integrity of EF1. The administration of 10 M BAPTA-AM induced a slight increase in the amount of cells in the G 1 phase in EGFP-, PR72 EF1mut-EGFP-, and PR72 EF(1 ϩ 2)mut-EGFP-transfected cells, whereas the opposite is true for the PR72-EGFP-and PR72 EF2mut-EGFP-transfected cells. These results confirm that Ca 2ϩ binding to the low affinity EF1-binding site is (partially) necessary to generate the growth arrest, probably via a Ca 2ϩ -dependent interaction with a substrate or another binding partner. The more pronounced G 1 /S arrest in cells where PR72 EF2mut is overexpressed suggests that the mechanism by which PR72 induces the G 1 /S arrest likely occurs via competition of monomeric PR72 with a BЉ-containing PP2A holoenzyme for binding to this substrate or binding partner. The EF2 mutant indeed lacks the interaction with the core enzyme and therefore would have a more pronounced dominant negative effect than the wild-type protein. DISCUSSION This is the first report of a direct regulatory effect of calcium ions on protein phosphatase 2A. In the present study, we demonstrate that the BЉ/PR72 regulatory subunit of PP2A is a "classical" Ca 2ϩ -binding protein of the EF-hand type. PR72 contains two well conserved EF-hand motifs, which apparently exhibit different affinities for Ca 2ϩ in an overlay assay. Because it is known that this type of assay merely detects high affinity binding sites (42), EF1 (aa290 -302) could be catalogued as a low affinity binding site because of its poor Ca 2ϩ binding, whereas EF2 (aa364 -376) is clearly a high affinity binding site. Although both amino acid sequences of EF1 and EF2 conform to the overall EF-hand consensus, the difference in affinity could be explained by the presence of a glycine residue at position 6 within EF2, whereas EF1 has a leucine at this position. According to some reports (47), it is important that a relatively small amino acid is present at this particular position in between both helices E and F in order for calcium ions to bind effectively. Moreover, we have shown that Ca 2ϩ binding to PR72 results in a significant conformational change, which is dependent on the integrity of the EF-hands. This structural change could be visualized by an increase in the intrinsic tryptophan fluorescence of the PR72 aa262-449 apoprotein upon the addition of Ca 2ϩ .
We further investigated the functionality of both EF-hands by a site-directed mutagenesis approach. We mutated in each EF-hand the glutamate or aspartate residues at positions 1 and 12, which are both involved in the Ca 2ϩ coordination (47), into FIG. 5. A, localization of wild-type and mutant EGFP-PR72 fusion proteins in COS7 cells. COS7 cells were transfected with pEGFP-C1derived plasmids encoding EGFP fusions of PR72 or the PR72 EF mutants. After fixation of the cells and nuclear staining with Hoechst 33342, EGFP expression was examined with the fluorescence microscope. B, co-immunoprecipitation of PP2A C and A/PR65 with PR72derived EGFP fusion proteins. Lysates of COS7 cells transfected with the upper EGFP fusion constructs were immunoprecipitated (IP) with a polyclonal anti-EGFP antibody. Whole cell lysates and the immunoprecipitations were then subjected to Western blotting and blots were developed with a mixture of anti-PP2A C and anti-PR65 monoclonal antibodies. C, Western blot of PR72 in nuclear and cytoplasmic extracts of L6 cells. Nuclear and cytoplasmic extracts of rat myoblast L6 cells, expressing PR72 were subjected to Western blotting. Blots were developed with a polyclonal anti-PR72 antibody raised against a C-terminal peptide of PR72 (15).
alanines. These mutations did not only abolish Ca 2ϩ binding to these motifs, but also changed the overall protein structure relatively little. This is an important observation, making it highly unlikely that specific defects of the EF mutants (as observed in further experiments) can be explained by conformational changes, which are sometimes inherent to the introduction of the mutations themselves.
By having established by a yeast two-hybrid approach that the minimal A subunit interacting domain of PR72 comprises both EF-hands, we were prompted to investigate whether these motifs are directly involved in A subunit binding. By a combination of both yeast and mammalian double-hybrid experiments, in vitro GST-PR65 pull-down assays, and anti-EGFP co-immunoprecipitations, we have shown that mutation of Asp-290 and Asp-301 within EF2 destroys binding of PR72 to the A subunit. Given that these mutations also destroy Ca 2ϩ binding to EF2, this could suggest that Ca 2ϩ binding to EF2 and the resulting conformational change are required for A subunit binding. Alternatively, Asp-290 and Asp-301 may be structurally important amino acids that are directly involved in protein-protein contacts with the A subunit. However, the identical intrinsic fluorescence spectra of PR72 aa262-449 and PR72 aa262-449 EF(1 ϩ 2)mut argue against this explanation. Because EF2 is a high affinity Ca 2ϩ -binding site in vitro, it would not be unlikely that it constitutively binds calcium ions in vivo, where the normal intracellular Ca 2ϩ concentration is a few hundred nanomolars. An accurate determination of the affinity constant of EF2 for Ca 2ϩ binding would make this more clear. The fact, however, that calcium chelators, even at high concentrations, are hardly capable of affecting the PR72-PR65 interaction in vitro as well as in vivo suggests that either the affinity of EF2 for Ca 2ϩ is very high or that Ca 2ϩ is inaccessible for these chelators because it is embedded within the PP2A T72 or PR72 protein structure.
Another remarkable result was the abolishment of the proper nuclear localization of PR72 by mutation of EF2, sug- gesting that its incorporation within the trimer is required for its subcellular targeting. This could indicate that the trimer context is required for the interaction with a specific import protein or, alternatively, with a modifying enzyme that takes care of a specific modification necessary for nuclear import. It should be noticed that PR72 contains two putative nuclear localization signals within its primary structure (15), which may be functional nuclear targeting motifs as well. This issue awaits further clarification.
A role for EF1, the low affinity calcium-binding site, is suggested by the in vitro measurement of PP2A T72 phosphatase activity toward phosphorylase a, where only the addition of high Ca 2ϩ concentrations resulted in inhibitory effects. Whether calcium ions also affect PP2A activity toward the as yet unknown physiological substrate(s) of PP2A T72 awaits further investigation. Unfortunately, the effects of Ca 2ϩ on the phosphatase activity of a PP2A T72 enzyme with a mutated EF1 or EF2 motif could not be assessed, because we were unable to reconstitute such an enzyme in vitro with PP2A D and bacterially expressed and purified PR72 or its mutants. Nevertheless, EF1 is (due to its relatively low affinity for Ca 2ϩ ) likely the site with the highest regulatory potential in vivo, because it might act as a "calcium sensor" that transiently binds Ca 2ϩ upon local or temporal rises in the intracellular Ca 2ϩ concentration. But whether Ca 2ϩ binding to EF1 would result in activation or inhibition of PP2A T72 toward a particular in vivo substrate cannot be predicted from our in vitro experiments.
However, that EF1 may operate as a calcium sensor is supported by the cell cycle experiments, where it became clear that the ability of PR72 to induce a G 1 /S arrest in U2OS cells is at least partially dependent on the integrity and the Ca 2ϩ recruiting ability of EF1. Note that in this case the addition of BAPTA-AM did have an effect on the properties of the wildtype protein. We propose that forced expression of monomeric PR72 may act as a dominant negative, influencing phosphatase activity indirectly. Because PR72 has a higher affinity for the A/PR65 subunit than PR55 or PR61 (Fig. 3A), it could be integrated in a trimer by expelling B/PR55 and BЈ/PR61 from their holoenzymes or by interacting with a pre-existing dimer. Dephosphorylation of specific substrates for PP2A T55 , PP2A T61 , or PP2A D would therefore be inhibited, whereas specific PP2A T72 substrates might be stimulated. On the other hand, "free" PR72 may compete with PP2A T72 for an interacting protein that would normally mediate a specific PP2A T72 effect. This would lead to inhibition of dephosphorylation of specific PP2A T72 substrates. Therefore, here again, it cannot be predicted whether overexpression of PR72 would lead to inhibition or stimulation of dephosphorylation of some PP2A substrates. However, PP2A activity is required for the firing of replication origins (50,51), and this may be mediated by PR72 (21) or another BЉ family member. Binding of PR72 with a relevant substrate in this case likely occurs via the Ca 2ϩ -bound EF1 motif, and Ca 2ϩ binding to EF1 would likely stimulate PP2A activity. Cdc6 is one of the candidates to be this binding partner, because it interacts with PR48 and with aa354 -1150 of PR130 (a region that encompasses the common part of PR130 and PR72) in yeast two-hybrid assays (17). The more pronounced dominant negative effect of the PR72 EF2 mutant, which lacks the interaction with the core enzyme and the proper subcellular localization, additionally suggests that this binding partner is (initially) not necessarily present in the nucleus.
Compared with BЈ/PR61␥1 and especially with B/PR55␣, BЉ/PR72 strongly interacts with the A/PR65 subunit in the mammalian two-hybrid system. This opens up the possibility that PR72 effectively competes with both PR61␥1 and PR55␣ for binding to PP2A D . Moreover, and in contrast with the reported data on the PR55 subunit (49), our data suggest that PR72 mutants failing to interact with the core enzyme are not highly unstable within cells. This suggests that in contrast with PP2A C , A/PR65 (4,5), and B/PR55 (49), the cell lacks a control mechanism to avoid the presence of free BЉ/PR72, provided of course that such a population would exist. These findings therefore contribute to a better understanding of the dynamics of the various PP2A complexes in vivo.
Together, we have demonstrated a role for Ca 2ϩ in PP2A T72 subunit assembly, nuclear targeting, catalytic activity, and PR72-mediated cell cycle regulation. Whether the other BЉ/ PR72 family members also bind Ca 2ϩ will have to be determined but is highly likely, given the conservation of the EF motifs within these proteins. This is obviously true for the PR130 subunit, a splice variant generated from the same gene as PR72 that shares both EF-hands. In the light of a recent report (52) demonstrating an interaction of PR130 with the ryanodine type 2 receptor, this is certainly an interesting observation. Moreover, given the muscle-specific expression of PR72 and the important role of calcium in muscle-specific processes, such as muscle contraction, our data open up a new avenue for further research at the interface between calcium signaling and protein phosphorylation/dephosphorylation.