Identification of a novel protein phosphatase 2A regulatory subunit highly expressed in muscle.

Differential association of regulatory B subunits with a core heterodimer, composed of a catalytic (C) and a structural (A) subunit, is an important mechanism that regulates protein phosphatase 2A (PP2A). We have isolated and characterized three novel cDNAs related to the B' subunit of bovine cardiac PP2A. Two human (B'alpha1 and B'alpha2) and a mouse (B'alpha3) cDNA encode for alternatively spliced variants of the B subunit. The deduced primary sequences of these clones contain 12 of 15 peptides derived from the purified bovine B' subunit. Differences between the deduced sequences of the B alpha splice variants and the cardiac peptide sequences suggest the existence of multiple isoforms of the B' subunit. Comparison of the protein and nucleotide sequences of the cloned cDNAs show that all three forms of B'alpha diverge at a common splice site near the 3'-end of the coding regions. Northern blot and reverse transcription-polymerase chain reaction analyses revealed that the B'alpha transcripts (4.3-4.4 kb) are widely expressed and very abundant in heart and skeletal muscle. The expressed human and mouse B'alpha proteins readily associated with the PP2A core enzyme in both in vitro and in vivo complex formation assays. Immunofluorescence microscopy revealed that epitope-tagged B'alpha was localized in both the cytosol and nuclei of transiently transfected cells. The efficiency of binding of all three expressed proteins to a glutathione S-transferase-A subunit fusion protein was greatly enhanced by the addition of the C subunit. Expression of the B'alpha subunits in insect Sf9 cells resulted in formation of AC.B'alpha heterotrimers with the endogenous insect A and C subunits. These results show that the B' subunit, which is the predominant regulatory subunit in cardiac PP2A, is a novel protein whose sequence is unrelated to other PP2A regulatory subunits. The nuclear localization of expressed B'alpha suggests that some variants of the B' subunit are involved in the nuclear functions of PP2A.

Protein phosphorylation is an essential mechanism regulating a wide variety of cellular processes. The coordinate activity of protein kinases and phosphatases is required for normal signal transduction. Protein phosphatase 2A (PP2A) 1 is a serine/threonine phosphatase that has been implicated in the control of the cell cycle (1), growth and proliferation (2), DNA replication (3), viral transformation (4), and morphogenetic events (5). The PP2A holoenzyme is a heterotrimer composed of a 38-kDa catalytic (C) subunit, a 63-kDa structural (A) subunit, and a third subunit termed B or phosphatase regulatory (PR) subunit (6,7). There are at least five distinct families of proteins that interact with and regulate the PP2A core enzyme. These include the B, BЈ, PR72 (BЉ), and ␦ families of regulatory subunits and the small and middle tumor antigens of DNA tumor viruses (8).
The diversity of PP2A regulatory subunits suggests specific physiological roles for individual holoenzymes. Genetic studies have shown that disruption of the yeast homolog of the B␣ subunit, CDC55, results in defects in budding and cytokinesis (9). Decreased expression of the Drosophila B␣ subunit causes defects in mitosis, duplication of wing imaginal discs and is lethal in the late larval/early pupal stage of development. (5,10). Biochemical studies have demonstrated that the B subunits and tumor antigens affect the substrate specificity and activity of PP2A (11)(12)(13). Sensitivity to polyamines, polycations, and ceramide is also dependent on the type of B subunit associated with the AC core complex (13)(14)(15). These data support an emerging hypothesis that differential association of B subunits regulates PP2A function in vivo.
Molecular cloning has revealed diversity within the PP2A regulatory subunit families. Multiple isoforms (␣, ␤, ␥) of the B subunit and a splice variant of B␣ have been isolated from mammalian sources (9,16,17). These proteins are 81-87% identical and diverge primarily at the amino termini. Two cDNAs encoding 72-and 130-kDa forms of the PR72 (BЉ) subunit were obtained from human heart and brain libraries (18). PR130 has an extension at the amino terminus that is generated either by alternative splicing or the use of an alternative promoter site. In this report, we describe the isolation and expression of novel cDNAs encoding human (BЈ␣1 and BЈ␣2) and mouse (BЈ␣3) members of the BЈ familiy of PP2A regulatory subunits. The recombinant subunits readily form heterotrimeric complexes with the AC core enzyme in vitro and in vivo. Transient expression of human BЈ␣ in mammalian cells revealed the presence of both cytoplasmic and nuclear populations of the regulatory subunits. The BЈ␣ subunits may be important for the localization/translocation of PP2A into the nucleus.

MATERIALS AND METHODS
Molecular Cloning of the Human BЈ-Subunit-PP2A (AC⅐BЈ) was purified from bovine cardiac tissue, and partial amino acid sequence of the BЈ subunit determined as described previously (12). The peptides derived from the bovine cardiac BЈ subunit were used to search for similar sequences in the GenBank data base using the BLAST network service of the National Center for Biotechnology Information. This search revealed significant homologies between the peptides and a random cDNA cloned from human KG1 cells (accession number D26445). The sequence of this cDNA was used for the design of oligonucleotide primers. An 854-bp fragment was amplified by polymerase chain reaction (PCR) as described previously (13) from a human umbilical vein epithelial cell (HUVEC) cDNA library (kindy provided by Drs. J. Battey and Mark Akeson, NIDCD) using a sense primer 5Ј-CTGAGT-GTCTACCATCCCCAG-3Ј (nucleotides 775-795) and an antisense primer 5Ј-TGGAATTCGTTTGAACTTCTAATCT-3Ј (nucleotides 1605-1629) in which several bases were mutated (underlined) to generate an EcoRI restriction site. The thermal profile (94°C, 1 min; 50°C, 1 min; 72°C, 1 min) was carried out for 30 cycles. The PCR fragment was subcloned into pCRII (Invitrogen) and the sequence was verified. Random priming of this fragment was used to generate a radiolabeled probe (ϳ10 9 dpm/g) to screen the HUVEC ZAP cDNA library. Prehybridization and hybridization of filters containing 10 6 independent recombinants was performed at 65°C in 5 ϫ SSC, 5 ϫ Denhardt's solution, 0.1% SDS, and 100 g/ml herring sperm DNA. The filters were washed 4 times (5 min) in 2 ϫ SSC, 0.1% SDS at room temperature, followed by a 20-min wash in 0.4 ϫ SSC, 0.1% SDS at 65°C, and subjected to autoradiography at Ϫ20°C. Five positive clones (HBЈ-2, 3, 5, 6, 7) were obtained after two rounds of plaque purification. Recovery of plasmids from the HUVEC cDNA library was carried out according to the manufacturer's protocol (Stratagene). The cDNAs were sequenced on both strands by the dideoxynucleotide chain termination method (19).
Yeast Two-hybrid Screening-The A subunit of PP2A (20) was inserted into the SmaI/SalI site of pAS1-CYH2 to generate a fusion with the GAL4 DNA binding domain. Yeast transformed with this plasmid were used to screen a mouse T-lymphocyte cDNA library, constructed as fusions with the GAL4 activation domain (10 6 transformants), using the two-hybrid assay (21). Positive colonies were selected on Trp Ϫ , Leu Ϫ , and His Ϫ medium in the presence of 50 mM 3-aminotriazole and screened for ␤-galactosidase activity by the colony lift method (22). The activation domain pACT-fusion plasmids were extracted from ␤-galactosidase negative colonies (25) after growth on Leu Ϫ , cycloheximide (100 g/ml) medium to remove the pAS1-A subunit plasmid. Sequences of the fusion plasmids were determined and used in BLAST searches of the nucleic acid and protein data bases.
GAL4 DNA binding domain fusions of human BЈ␣, p53, and the A subunit were tested for binding specificity with GAL4 activation domain fusions of the A and C subunits of PP2A. The A and C subunits of PP2A were inserted into the EcoRI/XhoI and SmaI/SalI sites of pACTII, respectively. NcoI/BamHI fragments from the pRcCMV-BЈ␣ plasmids (see below) were subcloned into the appropriate sites in the GAL4 DNA binding domain vector, pAS1-CYH2. Expression of ␤-galactosidase activity in the two-hybrid assay was determined as described above (22).
Northern Analysis and Reverse Transcription PCR-A 1.5-kb EcoRI fragment (nucleotides 2501-4064 of BЈ␣1 clone HBЈ-7) was used to probe a mRNA blot of human tissues (Clontech). Prehybridization was carried out in 50% formamide, 10 ϫ Denhardt's solution, 5 ϫ SSPE, 2% SDS, and 100 g/ml herring sperm DNA at 42°C. Hybridization was carried out in the same solution with the randomly primed, radiolabeled (1.5 ϫ 10 10 dpm/g) EcoRI fragment for 16 h at 42°C. The filter was washed twice (20 min) in 2 ϫ SSC, 0.05% SDS at room temperature followed by a 20-min wash in 0.2 ϫ SSC, 0.1% SDS at 50°C. The blot was subjected to autoradiography for 6 h at Ϫ80°C with an intensifying screen.
In Vitro Translation of BЈ␣ and GST-A Binding Assays-A PCR product was amplified from the mouse T-lymphocyte pACT-BЈ␣3 plasmid using a sense primer 5Ј-TAATACGACTCACTATAGGGAGACCA-CATGGATGATGTATATAACTATCATTTC-3Ј, containing a T7 promoter site and an in-frame initiator methionine and using an antisense primer against the pACT vector (5Ј-CTACCAGAATTCGGCATGCCGG-TAGAGGTGTGGTCA-3Ј). In vitro translation products were synthesized with the mouse BЈ␣3 PCR fragment, pRcCMV-BЈ␣1, and pRc-CMV-BЈ␣2 as templates using the TnT T7 coupled reticulocyte lysate system (Promega). GST and GST-A subunit fusion proteins were prepared from Escherichia coli and used in binding assays with the [ 35 S]methionine-labeled BЈ␣ proteins (2). In some cases, purified bovine cardiac C subunit (2.5 g) was preincubated for 30 min (4°C) with GST (16 g) and the GST-A subunit fusion protein (16 g) prior to addition of labeled BЈ␣ translation products. Bound proteins were eluted with 10 mM glutathione, resolved by SDS-PAGE, followed by fluorography.
Association of Expressed BЈ␣ with the A and C Subunits of PP2A in Sf9 Cells-Recombinant baculoviruses encoding the BЈ␣ subunits were generated using the BAC to BAC expression system according to the manufacturer's protocol (Life Technologies, Inc.). Insect cells (4 ϫ 10 7 ) were infected with recombinant BЈ␣ baculoviruses (20), harvested 48 h postinfection, and washed twice with phosphate-buffered saline (PBS). Subsequent operations were performed at 4°C. The cells were resuspended in 3 ml of HS buffer (10 mM imidazole, pH 6.5, 1 mM EDTA, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, 10% glycerol) and homogenized in a Duall homogenizer. The homogenate was centrifuged at 12,000 ϫ g for 15 min, and the supernatant was loaded onto a heparin-Sepharose column (5 ml) equilibrated in HS buffer at a flow rate of 1 ml/min. The column was washed with the same buffer until the A 280 returned to base line. The column was eluted with 20 ml each of 0.1, 0.3, and 0.5 M NaCl in HS buffer, and 1-ml fractions were collected. An aliquot of the fractions was analyzed by immunoblotting with BЈ␣ specific antiserum. An aliquot (0.25 ml) of fractions 40 (0.3 M NaCl elution) and 60 (0.5 M NaCl elution), which contained BЈ␣, was adjusted to 0.5 ml with HS buffer and applied to a Superdex 200 10/30 gel filtration column in HS buffer containing 0.5 M NaCl. A second aliquot of fraction 60 was incubated with the purified AC form of bovine cardiac PP2A (20 g) for 30 min prior to chromatography. Chromatography was carried out at a flow rate of 0.25 ml/min, and 0.5-ml fractions were collected. The fractions were analyzed by immunoblotting with antibodies against the C (24) and BЈ␣ subunits.
Expression of BЈ␣ in Mammalian Cells-Mouse NIH 3T3 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% calf serum. Cells were transfected with pFLAG-CMV2 constructs using the liposome method (2). The cells were harvested 48 h posttransfection and washed twice with PBS. Cells were Dounce homogenized, and nuclei were isolated by centrifugation (800 ϫ g, 10 min) through a 1 mM imidazole, pH 7.4, 0.25 M sucrose cushion. The supernatant was recentrifuged at 12,000 ϫ g (10 min), and the resulting supernatant was the cytosolic fraction. The nuclei were washed once, and nuclear extracts were prepared according to Krainer et al. (25). Cytosolic and nuclear proteins were resolved by SDS-PAGE and analyzed for the expression of BЈ␣1, BЈ␣2, and B␣ by immunoblotting.
Immunofluoresence Microscopy-NIH3T3 cells transfected with pFLAG-BЈ␣1, pFLAG-BЈ␣2, or pFLAG-B␣ were trypsinsized after 24 h and grown on glass coverslips. The cells were fixed 48 h post-transfection by incubation in absolute methanol for 5 min at Ϫ20°C and washed twice with PBS. Nonspecific sites were blocked by incubation in PBS containing 5% goat serum (1 h). Subsequent antibody incubations (1 h) and washes were carried out in PBS containing 1% goat serum. The cells were incubated with monclonal anti-FLAG 5 antibody (Kodak) at a concentration of 10 g/ml followed by a Cy3-conjugated affinitypurified goat anti-mouse antibody (Jackson ImmunoResearch Labs) at a dilution of 1:400. The coverslips were mounted with Fluoromount-G (Fisher) and examined with an Olympus AX70 microscope (600ϫ magnification).
Production of BЈ␣ Specific Antisera-A synthetic peptide (CPQAQK-DPKKDR) corresponding to amino acids 431-441 of BЈ␣2 and residues 471-480 of BЈ␣1 was coupled to keyhole limpet hemocyanin. The amino-terminal cysteine was added to facilitate the coupling. Peptide conjugation and polyclonal antibody production were carried out as described previously (13).

Molecular
Cloning of cDNAs Encoding BЈ␣-The major form of regulatory subunit in purified bovine cardiac PP2A is related to the BЈ subunit originally identified in a form of rabbit muscle PP2A termed PP2A 0 (17). Partial amino acid sequence of the bovine cardiac PP2A BЈ subunit was obtained from tryptic and cyanogen bromide peptides. 15 peptides were sequenced, and a total of 177 residues were determined. A search of the protein data bases with these peptides indicated a high degree of similarity with the translated sequence of a randomly cloned human cDNA from myeloid KG1 cells (accession number D26445). The KG1 cDNA was 3.7 kb and the longest open reading frame encoded for a protein of 475 amino acids. The deduced sequence of the KG1 cDNA contained 12 of the 15 peptides obtained from bovine cardiac BЈ. The similarity strongly suggested that the human cDNA was related to the bovine BЈ subunit. A Saccharomyces cerevisiae cDNA, RTS1 (accession number U06630) encoding a suppressor of a defect in the ROX3 transcription factor was also found to contain some of the bovine BЈ peptides. Although the similarity is much lower, it appears that this clone is a yeast homolog of the mammalian BЈ subunit (Fig. 1).
Oligonucleotide primers were designed from the human sequence in the data base, and a 854-bp PCR product was amplified from a HUVEC ZAP cDNA library. The PCR product was sequenced and found to be identical to nucleotides 775-1629 of the KG1 cDNA. Five positive clones were isolated from the HUVEC cDNA library with the 854-bp probe using stringent hybridization and wash conditions. The largest cDNA (HBЈ-7) was 4064 bp long and nearly identical in sequence to the KG1 cDNA. Two base changes were present in the 3Јuntranslated region of the HUVEC cDNA; nucleotide 2082 was G instead of A, and nucleotide 3847 was A instead of G. The HUVEC cDNA extends the 5Ј-and 3Ј-untranslated regions of the KG1 sequence by 62 and 182 nucleotides, respectively. The sequence around the putative initiator codon, AGCAGGATG-GTGG, conforms to the consensus motif for translation initiation in vertebrates (28). Four potential polyadenylation sites (AATAAA) are present at nucleotides 2480, 3677, 3851, and 4027. There is also an insertion of 117 nucleotides in the HBЈ-7 cDNA that is absent in the KG1 cDNA. The insertion occurs after nucleotide 1385 and encodes for an additional 39 amino acids (433-472). We have designated this cDNA BЈ␣1. The deduced amino acid sequence encodes for a protein of 514 residues with a predicted molecular weight of 59,995 and an isoelectric point of 6.1 (Fig. 1).
The remaining cDNAs were found to be partial clones, three of which corresponded to BЈ␣1. The fourth cDNA (HBЈ-5) lacked 617 nucleotides of the 5Ј sequence, terminated at nucleotide 3138 of BЈ␣1, and did not contain the 117-nucleotide insertion. This cDNA corresponded to the KG1 sequence in the data base, and we have designated it BЈ␣2. The deduced amino acid sequence encodes for a protein of 475 residues with a predicted molecular weight of 55,958 and an isoelectric point of 6.5 (Fig. 1).
An additional BЈ␣ cDNA was isolated in an independent screen of a mouse T-lymphocyte library for proteins that interact with the A subunit of PP2A using the yeast two-hybrid assay. One of the cDNAs that interacted with the GAL4 DNA binding domain-A subunit fusion protein was a mouse homolog of human BЈ␣. This cDNA (BЈ␣3) was 1.35 kb, lacked a portion of the 5Ј sequence, and contained 37 nucleotides of the 3Јuntranslated sequence. The partial open reading frame encodes for a protein of 435 residues, with a predicted molecular weight of 51,158. An alignment of the human, mouse, and a portion of the yeast RTS1 primary sequences is shown in Fig. 1. There are three conserved substitutions in the mouse sequence compared with human BЈ␣1 (88% identity) and BЈ␣2 (92% identity). The mouse sequence terminates 10 residues after the BЈ␣1 insertion begins.
Based on the nucleotide sequence identity, the two human forms are likely to arise by alternative splicing of a single gene. A putative splice acceptor site boundary, CCCAGG (nucleotides 764 -769 in BЈ␣2, 1319 -1324 in the KG1 cDNA, and 1379 -1384 in BЈ␣1), is present at the point of divergence between the BЈ␣1 and BЈ␣2 sequences. Consistent with these observations is the conservation of the same splice site junction in the mouse BЈ␣3 cDNA (nucleotides 1231-1236). The presence of the 39amino acid insertion in BЈ␣1 produces a bipartite nuclear targeting signal (29) that is absent in both BЈ␣2 and BЈ␣3. The yeast homolog of BЈ␣ diverges from its mammalian counterparts at the amino and carboxyl termini, which are not shown in Fig. 1. The overall identity between the yeast protein and BЈ␣1, BЈ␣2, and mouse BЈ␣3 is 39, 40, and 43%, respectively. However, if the divergent termini are omitted and the comparison done is on residues 127-561 of the yeast protein, the identity increases to 68%.
Expression and Distribution of BЈ␣ mRNA-The levels of transcripts encoding human BЈ␣ were analyzed in poly(A) ϩ RNA isolated from human tissues ( Fig. 2A). A single major transcript of 4.4 kb was detected in all tissues examined. The BЈ␣ mRNA was very abundant in heart, skeletal muscle, and brain. Lower levels were present in the pancreas, kidney, lung, and placenta. Liver had a very low level of the BЈ␣ mRNA that was detectable with longer exposure times (data not shown).
The probe used in the Northern analysis does not distinguish between the alternatively spliced forms of human BЈ␣. Therefore, expression of the alternatively spliced forms of BЈ␣ mRNA was examined in mouse tissues by reverse transcription PCR (Fig. 2B). Two major bands were amplified from each tissue that had mobilities identical to the BЈ␣1 (717 bp) and BЈ␣2 (600 bp) controls. Similar amounts of cDNA were present in each PCR reaction, since the ubiquitously expressed mRNA for cyclophilin was amplified to similar levels with all of the cDNAs (data not shown). Approximately the same ratios of BЈ␣1 and BЈ␣2 were observed in mouse tissues except liver and especially brain, where BЈ␣2 was more prevalent. In contrast to the human Northern blot, mouse skeletal muscle contained significantly lower levels of the BЈ␣ transcripts. Interestingly, bands of ϳ650 and ϳ750 bp were also amplified with the heart, brain, and skeletal muscle cDNAs. Although we have not identified these products, they may be due to the presence of additional splice variants of BЈ␣ or other isoforms of the BЈ regulatory subunit.
Association of Recombinant BЈ␣ Subunits with PP2A-Sev-eral approaches were taken to show that the human and mouse BЈ␣ cDNAs encode proteins that interact with the PP2A core enzyme. Yeast strains containing GAL4 activation domain fusions (GAD) of the A and C subunits of PP2A were transformed with plasmids encoding fusions between the GAL4 DNA binding domain (GDB) and human BЈ␣1, BЈ␣2, p53, and the A subunit of PP2A (Table I). Both of the GDB-fusions of human BЈ␣ interacted with the GAD-A subunit fusion and activated transcription of the HIS3 and lacZ genes in the two-hybrid assay. No hybrids were formed between either GDB-BЈ␣ fusions and activation domain fusions of the PP2A C subunit or p53. As a positive control, ␤-galactosidase expression was activated by the GDB-A subunit in the strain containing the GAD-C fusion.
In vitro transcription/translation of BЈ␣ cDNAs in reticulocyte lysates directed the synthesis of [ 35 S]methionine-labeled human and mouse BЈ␣ proteins. The expressed proteins had apparent molecular masses of 60 kDa (Fig. 3A, lane 1, BЈ␣1), 56-kDa (Fig. 3A, lane 2, BЈ␣2), and 52-kDa (Fig. 3A, lane 3,  BЈ␣3), which were nearly identical to the predicted molecular weights of each splice variant. The interaction of the expressed proteins with the A subunit was assayed using a GST-A subunit fusion protein. A low level of binding to GST-A was observed with BЈ␣1 (Fig. 3B, lane 1), BЈ␣2 (Fig. 3C, lane 1), and BЈ␣3 (Fig. 3D, lane 1). However, preincubation of the GST-A fusion protein with the C subunit of PP2A significantly enhanced the binding of all three forms of BЈ␣ (lane 2). The interaction of the BЈ␣ proteins with the A subunit was specific, as no binding was observed with GST alone (lane 3) or when GST was preincubated with the C subunit (lane 4). The minor bands present in the GST-A lanes may be due to proteolysis of BЈ␣1 and BЈ␣2 or initiation from an internal methionine.
Recombinant BЈ␣ proteins were also expressed in the baculovirus-insect cell system. Homogenates from infected cells were partially purified by heparin-Sepharose and size exclusion chromatography to determine if the expressed BЈ␣ proteins interacted with the endogenous AC core enzyme. The BЈ␣ proteins in infected Sf9 cell extracts were eluted from heparin-Sepharose with both 0.3 and 0.5 M NaCl (data not shown). When BЈ␣1 in the 0.3 M NaCl fraction was applied to a gel filtration column, the peak of immunoreactivity co-eluted with the C subunit of PP2A (Fig. 4A) in fractions 24 -26, which correspond to the elution position of purified cardiac AC⅐BЈ (M r ϭ 156,000). The 0.5 M NaCl heparin-Sepharose fraction, which did not contain any detectable C subunit (data not shown), eluted in fractions 27-29, corresponding to the predicted molecular weight of monomeric BЈ␣1 (Fig. 4C; left half of panel). Incubation of the 0.5 M NaCl heparin-Sepharose fraction with purified bovine cardiac AC dimer shifted the peak of BЈ␣1 to fractions 24 -26, corresponding to heterotrimeric PP2A ( Fig.  4C; right half of panel). Similar results were obtained with recombinant BЈ␣2 (Fig. 4, B and D). Formation of the AC⅐BЈ␣ complexes was not disrupted by chromatography in 0.5 M NaCl. These properties of the expressed BЈ␣1 and BЈ␣2 proteins are identical to those described previously for the BЈ subunit purified from cardiac tissue (20). These data demonstrate that the cloned BЈ␣ proteins bind directly to the A subunit of PP2A and readily form heterotrimeric complexes with the AC core enzyme in vitro and in vivo.
Localization of BЈ␣ in Mammalian Cells-Indirect immunofluorescence of NIH3T3 cells expressing FLAG-BЈ␣ revealed both cytoplasmic and nuclear populations of the recombinant proteins. Transient expression of both BЈ␣1 (Fig. 5A) and BЈ␣2 (Fig. 5C) led to diffuse cytoplasmic and pronounced nuclear staining. In contrast, transient expression of FLAG-B␣ (Fig.  5E) led only to cytoplasmic staining. The predominantly cytoplasmic localization of FLAG-B␣ is consistent with the distribution of endogenous B␣ in CV1 cells (30). Similar amounts of FLAG-BЈ␣ and FLAG-B␣ were present in whole cell extracts as determined by immunoblotting with the anti-FLAG antibody (data not shown). The differential localization of FLAG-BЈ␣1 and FLAG-BЈ␣2 in the nucleus and cytoplasm and FLAG-B␣ in the cytoplasm was confirmed by immunoblots of cytosolic and nuclear extracts from the transfected cells (data not shown). Control experiments in which the primary or secondary antibody was omitted led to a loss of immunostaining. Preadsorbtion of the anti-FLAG antibody with the antigenic peptide also abolished the immunoreactivity (data not shown). Phase contrast images (Fig. 5, B, D, and F) showed that nontransfected cells lacked any significant immunofluorescence. DISCUSSION We have isolated three novel cDNAs related to the bovine cardiac BЈ regulatory subunit of PP2A. All three forms appear to be generated from a single gene by alternative splicing. Transcripts of BЈ␣ are widely expressed in human and mouse tissues and are especially abundant in muscle. The high level of BЈ␣ transcripts detected in heart and skeletal muscle is consistent with the biochemical composition of PP2A purified from these tissues (17,24). These BЈ subunits have no apparent homology to the B or PR72/130 (BЉ) regulatory subunits or other proteins that interact with and regulate PP2A, including viral tumor antigens and the phosphotyrosyl phosphatase activator protein (31). Interaction of these proteins with PP2A is apparently not due to regions of conserved sequence but may involve domains of similar higher order structure.
Heterotrimeric complexes of PP2A were formed when recombinant human and mouse BЈ␣ proteins were reconstituted with the AC core enzyme in vivo and in vitro. All three forms of BЈ␣ interacted with a GST-A subunit fusion protein and with a FIG. 2. Expression and tissue distribution of B␣ mRNA. Panel A, Poly(A ϩ ) RNA (2 g) from the indicated human tissues was hybridized with a BЈ␣ specific probe as described under "Materials and Methods." Panel B, first strand cDNA was prepared from poly(A ϩ ) RNA isolated from mouse tissues and used in a PCR with BЈ␣ specific primers as described under under "Materials and Methods." An aliquot of each reaction was resolved on a 1% agarose gel and stained with ethidium bromide (0.5 g/ml). The migration of molecular size standards (kb) are indicated to the left of each panel.

TABLE I
Interaction of BЈ␣ with PP2A in yeast Yeast strains were assayed for ␤-galactosidase activity in the twohybrid system as described under "Materials and Methods." GAL4 activation domain-A subunit fusion protein in the yeast two-hybrid assay. A role for the C subunit in stabilizing oligomeric complexes was originally suggested by reconstitution experiments with porcine cardiac PP2A (32). This idea is supported by chemical cross-linking studies (13,20), and analysis of A subunit mutants (33,34). The increased binding of recombinant BЈ␣ proteins to GST-A in the presence of C is also consistent with a role in stabilizing the AC⅐BЈ heterotrimer. We do not know if the endogenous yeast A subunit participates in the two-hybrid interaction. However, no hybrids were formed between GAD-C and GDB-BЈ␣, suggesting that BЈ␣ has very weak affinity for C and that the yeast A subunit homolog does not participate in hybrid formation. Another possibility is that the fusion with the DNA binding domain to the amino terminus of BЈ␣ hinders interaction with GAD-C.
The interaction assays suggested that the BЈ␣ subunits have different apparent affinities for PP2A. Of the three proteins, BЈ␣1 displayed the strongest interaction in both the yeast two-hybrid and GST-A binding assays. It is not known whether the absence of 8 amino acids from the amino terminus of mouse BЈ␣3 affects binding to PP2A. However, there is evidence that a portion of the amino terminus of the B␣ and B␤ subunits is involved in binding to the AC core enzyme (13). The fact that BЈ␣1 and BЈ␣2 have different apparent affinities for the AC complex suggests that regions near the carboxyl terminus are also important for subunit interactions.
The purified bovine cardiac BЈ subunit has a very high affinity (IC 50 ϭ 0.58 nM) for the AC form of PP2A (20). Although cardiac BЈ and BЈ␣2 have the same mobility in SDS-PAGE, functional data and other lines of evidence suggest that human BЈ␣2 and bovine BЈ are not the same protein. Three of the peptides derived from the bovine cardiac protein were not found in the human or mouse BЈ␣ sequences. Although these peptides may have been generated from a contaminant(s), they were not similar to any other sequences in the data bases. Anti-peptide antibodies raised against the human BЈ␣ proteins cross-reacted with the bovine cardiac subunit; however, no cross-reactivity with the human proteins was observed with an antiserum (E005) raised against purified bovine cardiac BЈ (data not shown). Taken together, these data indicate that the human BЈ␣ and bovine cardiac BЈ subunits are closely related isoforms of a larger BЈ family. 2 PP2A has generally been regarded as a soluble cytoplasmic enzyme, but significant amounts of PP2A are also present in the nucleus (35,36). Partial purification of rat liver nuclear PP2A suggests that the nuclear enzyme is a heterotrimer; however, the type of B subunit in nuclear PP2A has not been identified (37). A significant feature of all three BЈ␣ variants is the presence of a cluster of basic residues near the carboxyl 2 While this manuscript was under review, McCright and Virshup (39) reported the isolation of a new family of PP2A regulatory subunit (accession numbers L42373-L42375). Comparison of the sequences indicates that their cDNAs are members of the BЈ family of regulatory subunits. The B56␥ isoform that they report corresponds to our BЈ␣ cDNAs. terminus and a consensus bipartite nuclear localization signal in BЈ␣1 (residues 462-478). The bipartite motif has been found in 60% of all known nuclear proteins and less than 4% of nonnuclear proteins (29). Although the bipartite nuclear localization signal is absent in BЈ␣2, localization of FLAG-BЈ␣2 was similar to that of BЈ␣1 in our transient assays. While some nuclear proteins contain large T antigen-type nuclear targeting signals or the bipartite motifs, there are many nuclear proteins that do not contain a concensus sequence for nuclear localization (29). The only significant feature that has been recognized in these proteins is the presence of clusters of basic residues within the signaling domain. The basic residues present in the carboxyl termini of the BЈ␣ proteins may serve as a signal for nuclear import. Mutational analysis of this region will be required to determine the critical residues for nuclear localization.
Exclusion of transiently expressed FLAG-B␣ from the nucleus is consistent with previous immunofluorescence data showing that AC⅐B␣ is largely cytoplasmic and that a subpopulation is associated with the microtubule cytoskeletal network (30). We have not quantitated the amount of FLAG-B␣ that is associated with endogenous AC in the transient assays. However, other studies have shown that expressed small t antigen interacts with endogneous AC in mammalian cells (2,38). In addition, we have shown here that expressed BЈ␣ subunits form heterotrimeric complexes with endogenous AC in Sf9 cells. Presumably, some FLAG-B␣ associates with AC and is bound to microtubules, preventing nuclear import. We have no evidence that AC⅐BЈ␣ binds to microtubules, and failure of the AC⅐BЈ␣ complex to interact with the cytoskeleton may allow nuclear uptake. Regardless of the mechanism, it is clear that expressed BЈ␣ subunits are highly localized within the nucleus, while the B␣ subunit is not. This result suggests that some members of the BЈ family may be present in nuclear PP2A and that BЈ may be important in nuclear functions of PP2A. PP2A has been implicated in the dephosphorylation of nuclear cAMPresponse element binding protein (37,38). The AC⅐BЈ␣ heterotrimer may be involved in the control of cAMP-mediated changes in gene transcription. Yeast RTS1 is a suppressor of a mutation in the ROX3 transcription factor. The similarity of RTS1 and BЈ␣ also suggests a role for the mammalian homolog in regulating the activity of transcription factor(s).