Critical role for protein phosphatase 2A heterotrimers in mammalian cell survival.

The predominant forms of protein phosphatase 2A (PP2A), one of the major Ser/Thr phosphatases, are dimers of catalytic (C) and scaffolding (A) subunits and trimers with an additional variable regulatory subunit. In mammals, catalytic and scaffolding subunits are encoded by two genes each (alpha/beta), whereas three gene families (B, B', and B'') with a total of 12 genes contribute PP2A regulatory subunits. We generated stable PC12 cell lines in which the major scaffolding Aalpha subunit can be knocked down by inducible RNA interference (RNAi) to study its role in cell viability. Aalpha RNAi decreased total PP2A activity as well as protein levels of C, B, and B' but not B'' subunits. Inhibitor experiments indicate that monomeric C and B subunits are degraded by the proteosome. Knock-down of Aalpha triggered cell death by redundant apoptotic and non-apoptotic mechanisms because the inhibition of RNAi-associated caspase activation failed to stall cell death. PP2A holoenzymes positively regulate survival kinase signaling, because RNAi reduced basal and epidermal growth factor-stimulated Akt phosphorylation. RNAi-resistant Aalpha cDNAs rescued RNAi-induced loss of the C subunit, and Aalpha point mutants prevented regulatory subunit degradation as predicted from each mutant's binding specificity. In transient, stable, and stable-inducible rescue experiments, both wild-type Abeta and Aalpha mutants capable of binding to at least one family of regulatory subunits were able to delay Aalpha RNAi-induced death of PC12 cells. However, only the expression of wild-type Aalpha restored viability completely. Thus, heterotrimeric PP2A holoenzymes containing the Aalpha subunit and members of all three regulatory subunit families are necessary for mammalian cell viability.

The dynamic balance of protein kinase and phosphatase activities toward key substrates is crucial for cellular homeostasis. It is becoming increasingly apparent that phosphatases are just as tightly controlled as kinases, although mechanistic details largely remain to be explored.
Protein phosphatase 2A (PP2A) 1 is one of four major classes of serine/threonine-specific phosphatases (for a recent review, see Ref. 1) that, together with protein phosphatase 1, contributes most of the Ser/Thr phosphatase activity in cells. The PP2A catalytic subunit is a 36-kDa globular protein that associates predominantly with a 65-kDa scaffolding that associates predominantly with a 65-kD scaffolding A subunit (or PR65) to form the PP2A core dimer. In vertebrates, two genes referred to as C␣/␤ and A␣/␤ encode catalytic and scaffolding subunits, respectively. The significance of having two isoforms is unclear, although there are subtle differences in the tissue expression of the two A and C subunit genes, with ␣ isoforms generally being more abundant (2,3). The A subunits are hook-shaped proteins with 15 non-identical, tandem ␣-helical repeats (4, 5) that link catalytic subunits to a diverse group of regulatory subunits to form PP2A heterotrimers. Regulatory subunits have been divided into three gene families, called B (or PR55), BЈ (or B56 or PR61), and BЉ (or PR72/130, PR59, or PR48). Each gene family contains one to two genes in lower metazoans, such as flies and worms, and three to five members in vertebrates. The three gene families have no obvious sequence or predicted structural similarities, although loose consensus sequences for A subunit interaction have been reported (6). The B family of regulatory subunits consists of WD repeat-containing ␤-propeller proteins with variable N termini that dictate subcellular localization (7,8). Proposed functions include the regulation of cytoskeletal dynamics, mitogen-activated protein kinase signaling, and apoptosis (8 -12). BЈ family regulatory subunits are phosphoproteins that have been variously implicated in the control of Wnt/␤-catenin signaling, survival, and synaptic plasticity (11,(13)(14)(15)(16)(17). Lastly, BЉ regulatory subunits are nuclear, calciumbinding proteins that may regulate cell cycle progression (18 -20).
Although some regulatory subunits are expressed in a tissueand developmental stage-specific pattern, any given mammalian cell is believed to express several dozen distinct PP2A heterotrimers formed by the combination of the 16 PP2A subunit genes. This complexity presents a considerable challenge, particularly because functional redundancy can complicate the interpretation of genetic ablation experiments. Representing the only PP2A subunit knocked out to date, deletion of the major catalytic subunit (C␣) leads to early embryonic lethality, because the mesoderm does not form (21).
Most PP2A catalytic subunits in the cell form heterotrimeric complexes followed by the AC core dimer and, finally, numerous lower abundance complexes (22)(23)(24)(25)(26)(27). Highlighting the role of PP2A in growth control, several viruses encode proteins that subvert PP2A activity by binding to either dimeric or trimeric forms of the phosphatase (1).
To begin to address the roles of the different forms of PP2A in mammalian cells, we report here the consequences of downregulating the major scaffolding (A␣) subunit by inducible RNAi. We find that A␣ RNAi destabilizes other PP2A subunits, impairs Akt signaling, and leads to apoptotic cell death. Rescue experiments involving A␣ mutants with different regulatory subunit binding specificities demonstrate that PP2A heterotrimers containing members of all three regulatory subunit families contribute to cell survival.
pcDNA3 plasmids expressing A␣ and A␤ wild-type and 〈␣ mutants harboring a C-terminal EE-epitope tag were described previously (3,29) and kindly provided by Gernot Walter (University of California at San Diego). The RNAi-resistant A␣ cDNAs were generated by doublestranded DNA mutagenesis according to the QuikChange protocol using Pfu Ultra polymerase (Stratagene) (sense mutagenic primer was 5Ј-GGGCCGCAGCCTCCCACAAAGTGAAGGAATTCTGTGAAAACC-TCTCAGC-3Ј; base changes are underlined). For Dox-inducible expression, A␣/A␤ cDNAs were excised with HindIII/XbaI and subcloned into pcDNA5/TO (Invitrogen). All PCR products were subjected to automated sequencing.
Stable Cell Line Generation-PC6-3 cells stably expressing the tetracycline repressor (12) were transfected with the linearized vector expressing A␣-directed hpRNA under the control of the H1-TO promoter and selected in 500 g/ml G418 and 2 g/ml blasticidin essentially as described (12). 36 clones were expanded and screened for Dox-dependent down-regulation of A␣ by immunoblotting. A-exchange cells with concomitant induction of A␣ mutants were established by transfecting an inducible A␣ RNAi clone with linearized, RNAi-resistant A␣ cDNAs in pcDNA5/TO, followed by selection in 500 g/ml hygromycin, 200 g/ml G418, and 2 g/ml blasticidin.
Protein Phosphatase Assays-A␣ RNAi PC6-3 cells growing in 6-well plates were lysed in buffer containing 1% Triton X-100 (v/v), 150 mM NaCl, 20 mM Tris, pH 7.5, 1 mM EDTA, 1 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 1 mM benzamidine, and insoluble material was removed by centrifugation at 22,000 ϫ g for 15 min. Cleared lysates were diluted 1:100 to 1:200 in 2 mg/ml bovine serum albumin, 50 mM Tris, pH 7.5, 2 mM EDTA, 2 mM EGTA, 2 mM dithiothreitol, 1 mM benzamidine, and 1 mg/ml leupeptin. Myelin basic protein was phosphorylated with [␥-33 P]ATP by protein kinase A as described (8) and diluted to ϳ10,000 cpm/l in the same buffer. Phosphatase reactions were started by the addition of 5 l of diluted lysate to 20 l of diluted substrate, incubated for 30 min at 30°C with intermittent agitation on an Eppendorf shaking incubator, and terminated by the addition of trichloroacetic acid to a final concentration of 20% (w/v). Following centrifugation at 22,000 ϫ g, acid-soluble 33 Pphosphate was quantified by liquid scintillation counting. PP2A activity was determined by subtracting the activity in the presence of 2.5 nM okadaic acid from the total activity (30). Less than 20% substrate dephosphorylation occurred under all assay conditions.
Survival and Apoptosis Assays-A␣ RNAi PC6-3 cells were cultured (37°C with 5% CO 2 ) in PC12 growth medium (10% horse serum and 5% fetal bovine serum in RPMI 1640) containing 200 g/ml G418 and 2 g/ml blasticidin. A-exchange PC6-3 cell medium additionally contained 200 g/ml hygromycin. For survival/apoptosis assays, cells were seeded in 24-or 96-well plates at 50,000 or 2,000 cells/well, respectively, followed by the addition of 1 g/ml Dox or ethanol vehicle (final 0.1%) at different time points.
For apoptosis assays based on nuclear morphology, fixed cells in 24-well plates were stained with 1 g/ml Hoechst 33342 and analyzed by epifluorescence microscopy as described (8). For trypan blue exclusion and annexin V assays, cells were trypsinized, stained with either a 0.2% trypan blue dye solution or Cy3-conjugated annexin V according to the manufacturer's protocol (United States Biological, San Antonio, TX) and counted in a hemocytometer under transmission or epifluorescence illumination.

Knock-down of PP2A/Aa by Inducible
RNAi-Stable knockdown of gene expression is a prerequisite for carrying out long term growth and survival studies in a homogenous cell population, and stable expression of double-stranded RNA to downregulate endogenous mRNAs is increasingly used as a less time consuming alternative to traditional gene knock-outs (32). An inducible RNAi approach avoids a possible selection bias and allows for the analysis of essential gene products. We modified the H1 promoter, which is widely used to drive small hpRNA expression for RNAi (28), by replacing evolutionarily non-conserved sequences before and after the TATA box with the bacterial TO sequence. A similar inducible construct with a single TO site between the TATA box and the transcription start site was reported recently (33) and has since become commercially available (pSUPERIOR; OligoEngine, Seattle, WA). The induction principle is illustrated in Fig. 1A. In cells expressing the tetracycline repressor (TR) protein, TR silences the hybrid H1-TO promoter in the absence of tetracycline or its more stable analog, Dox. Upon the addition of an inducer, TR dissociates from the promoter, allowing hpRNA expression and RNAi to proceed. We found the H1-TO promoter to lack detectable leak expression and to effect gene knock-down with a potency comparable to the parental H1 promoter. Importantly, this inducible system appears to require the repressor function of the native TR protein in that TR cannot be replaced with the more widely used VP16 transactivation domain fusions of the TR DNA binding domain (34) (data not shown).
To study the role of PP2A holoenzymes in cell survival, we chose to target the scaffolding A␣ subunit because it is the best-characterized PP2A subunit in terms of structure/function (4, 29). We used TR-expressing PC6-3 cells as a host for the inducible RNAi cassette (12). PC6-3 cells are a subline of rat PC12 pheochromocytoma cells that differentiate to a sympathetic neuronal-like phenotype upon the addition of nerve growth factor (31). Cells were selected in G418, and multiple, independently isolated clones were scored for Dox-inducible down-regulation of the endogenous A subunit by immunoblotting. Two positive clones were characterized further and exhibited identical phenotypes in survival and PP2A subunit degradation assays. Experiments with one of these clones are reported here.
Proteosomal Degradation of PP2A Holoenzymes after A␣ RNAi-Dox addition resulted in a time-dependent (ϳ70% by day 4) decrease in A subunit expression as detected with an antibody that recognizes both ␣ and ␤ gene products (Fig. 1, B and C). An A␣-specific antibody showed the same time course of down-regulation, indicating that A␣ is the predominant scaffolding subunit isoform expressed in PC6-3 cells and that A␣ down-regulation does not cause an appreciable compensatory induction of A␤.
Previous studies in Drosophila S2 cells showed that RNAi of the A or C subunit leads to the loss of the other PP2A subunits (11,16). Consistent with this result, A␣ knock-down in mammalian cells was paralleled by a significant decrease in PP2A catalytic subunit immunoreactivity (Fig. 1, B and C), which was well matched by a decline in total PP2A phosphatase activity (Fig. 1D). In addition to the PP2A C subunit, levels of B family subunits (detected by a pan-B specific antibody) as well as BЈ family subunits (detected with BЈ␣-and BЈ␦-specific antibodies) fell dramatically following A␣ knock-down (Fig. 2). In contrast, neither the BЉ subunit PR59 nor the PP2A-associated protein striatin responded to Dox treatment for 3 days.
These data suggest that C, B, and BЈ family subunits are stable only when complexed to the A subunit, whereas BЉ and striatin proteins can exist independently of PP2A in the cell.
We have demonstrated previously that ectopically expressed, monomeric B␥ mutants are rapidly cleared by ubiquitination and proteosomal degradation (7). To examine the mechanism of PP2A subunit removal following 〈␣ knock-down, cells were treated with MG132, a proteosomal inhibitor, or leupeptin, which inhibits lysosomal degradation, in the absence or presence of Dox. Proteosome inhibition restored C and B family subunits almost to control levels, whereas leupeptin had no effect. Neither inhibitor prevented the loss of BЈ family subunits following A␣ RNAi (Fig. 2).

A␣ Knock-down Kills Cells by Apoptotic and Non-apoptotic
Mechanisms-RNAi-mediated knock-down of the single A and C subunit genes in Drosophila S2 cells causes apoptotic cell death (11,16). We observed a drastic loss of PC6-3 cell viability starting 4 days after Dox treatment to induce A␣-directed hpRNA expression, with almost complete cell loss by 8 days after the Dox treatment (Fig. 3A). In a second, independently isolated clonal cell line with slower kinetics of A␣ down-regulation, cell death was first detectable 6 days after Dox addition (data not shown). These results indicate that PC6-3 cells can tolerate a ϳ70% drop in A␣ levels before viability is compromised. As a control, inducible RNAi of the B␣ regulatory subunit of PP2A was associated with a modest (ϳ20%) decrease in cell number over the same time course, which appeared to be due to a slowing of proliferation rather than mortality (Fig. 3A and data not shown). Starting 4 days after Dox treatment, A␣ RNAi was associated with cell rounding and the appearance of membrane blebs (Fig. 3B, inset). Additionally, we observed increases in membrane permeability, annexin V staining, and condensed/fragmented nuclear morphology (Fig. 3B), which occurred over the same time course as the loss of viability (Fig.  3A). As further evidence of apoptotic cell death, A␣ RNAi was found to result in a time-dependent increase in caspase-3/7 activity, which could be abolished by the general caspase inhibitor Z-VAD-fmk (50 M ; Fig. 3C). In the same experiments, however, caspase inhibition was unable to prevent the cell death caused by PP2A down-regulation (Fig. 3D). Importantly, the same dose of Z-VAD-fmk completely blocked apoptosis induced by 100 nM stauroporine (data not shown), as was reported previously for PC12 cells (35). The inability of Z-VADfmk to restore viability in PP2A-depleted cells stands in marked contrast to Drosophila studies, which showed that caspase knock-down overcomes PP2A RNAi-mediated cell death (16). It would appear that in mammalian cells the loss of PP2A activity initiates redundant caspase-dependent (apoptotic) and caspase-independent cell death programs.
A␣ Knock-down Impairs Akt Phosphorylation-The Ser/Thr kinase Akt/PKB transduces survival signals from a variety of growth factor receptors (36). We analyzed Akt signaling by immunoblotting with an antibody that recognizes Ser(P)-473, which is tightly correlated with activity. At the time point analyzed, 3 days with or without Dox, A␣ subunit level and cellular PP2A activity are diminished by ϳ60 and 20%, respectively (Fig. 1), whereas cells are still completely viable (Fig. 3). We found that Dox treatment decreased basal Akt phosphorylation by 50%. In cells stimulated for 5 and 15 min with epidermal growth factor (10 ng/ml), A␣ RNAi caused a similar reduction in phospho-Akt without changing total Akt protein levels (Fig. 4). These results indicate that PP2A holoenzymes are necessary for proper activation of the survival kinase Akt.
RNAi Resistant A␣ Mutants Prevent Holoenzyme Degradation-Gernot Walter's group has identified several mutations in the N-terminal half of A␣ that modulate its ability to interact with select regulatory subunit families without effecting C subunit association (29). For instance, Glu-100 and Glu-101 of A␣ form critical salt bridges with basic residues on regulatory subunits (7), and the A␣ EE100RR mutant can hence only form PP2A dimers in the cell. The A␣ DWF139HAA mutant was shown in in vitro assays to associate exclusively with B family regulatory subunits, whereas the DTP177AAA mutant A␣ binds to B and BЉ but not to BЈ family subunits (29). We reasoned that rescue experiments with these mutants should enable us to identify the PP2A holoenzyme(s) essential for cell survival. To render them resistant to knock-down, four noncoding, single nucleotide changes were incorporated into the RNAi target sequence of wild-type and binding-deficient A␣ cDNAs (Fig. 5A). Coimmunoprecipitation of epitope-tagged A␣ mutants and B, BЈ, and BЉ subunits ectopically expressed in COS cells confirmed the binding selectivity that was demonstrated previously in vitro (Fig. 5B). Wild-type and mutant A␣ subunits could be transiently expressed to similar levels in the A␣ RNAi cell line and associated with equivalent amounts of the catalytic subunit (Fig. 5C). Ectopic expression levels of A␤, on the other hand, reached at most 50% of A␣ levels, and only low levels of the C subunit could be detected in the A␤ immunoprecipitate. The latter finding is in agreement with previous studies that reported low affinity of A␤ for both catalytic and B family regulatory subunits (3,37). Next, we asked whether transient expression of RNAi-resistant, mutant A␣ subunits could restore PP2A subunit levels in cells subjected to A␣ RNAi. All A␣ cDNAs were expressed to levels approximating the endogenous scaffolding subunit, and all A␣ cDNAs stabilized levels of the C subunit (Fig. 5D). Regulatory subunits that degrade upon A␣ RNAi were stabilized in accordance with the binding-selectivity of each A␣ mutant; e.g. BЈ␣ levels were restored by wild-type but not by any of the mutant A␣ cDNAs, and B family regulatory subunit levels were rescued by all but the A␣ EE100RR mutant. Thus, ectopic expression of bindingdeficient A␣ mutants coupled with down-regulation of endogenous A␣ can be used to ablate select families of PP2A heterotrimers.
Heterotrimeric PP2A Holoenzymes Are Necessary for Mammalian Cell Viability-Three assays were used to investigate rescue of viability in A␣ knock-down cells. For short-term sur-vival, cells were transfected with a combination of a "rescue plasmid" expressing RNAi-resistant PP2A/A subunits and a ␤-galactosidase marker plasmid. Cells were then incubated in the absence or presence of Dox for 7 to 11 days followed by a ␤-galactosidase assay to quantify cell survival. Results from a representative experiment are shown in Fig. 6A. Empty vectortransfected cells died with a time course similar to that of untransfected cells (compare with Fig. 3A), with 36% survival after 7 days and 4% survival after 10 days in Dox. A␣ wild-type transfected cells were significantly protected from A␣ downregulation (78% survival after 7 days and 66% survival after 10 days), especially considering the transient nature of the expression. This result is important, as it shows that the observed loss of viability is indeed a consequence of A␣ down-regulation as opposed to an off-target effect of the RNAi. The B family-only binding A␣ mutant (DWF139HAA) rescued almost as well as wild-type, whereas the binding-incompetent, "dimer-only" A␣ mutant (EE100RR) had strongly impaired survival-promoting activity. Fig. 6B summarizes the data from 2-7 independent short-term survival experiments, normalizing rescue activity to A␣ wild-type (100%) and vector control (0%). The B/BЉ binding A␣ mutant (DTP177AAA) rescued as well as wild-type did, A␤ and A␣ DWF139HAA rescued moderately well (70 -75%), and dimer-only A␣ EE100RR had poor activity (30%). ␣4, the mammalian homolog of yeast Tap42, is a component of the mammalian target of rapamycin signaling pathway and competes with the A subunit for binding to the PP2A catalytic subunit (22). ␣4 was unable to promote survival, nor was cyclin G2, which binds to PP2A/BЈ and C but not to the A subunit (23),

FIG. 5. Characterization of RNAi-resistant A subunit mutants with specific regulatory subunit binding deficits.
A, COS cells were cotransfected with wild-type (w.t.) or RNAi-resistant (res.) A␣ cDNAs and plasmids expressing A␣-directed hpRNA (ϩ) or empty vector (Ϫ), and extracts were immunoblotted for transfected A␣ (EE epitope-tagged) and endogenous extracellular signal-regulated kinase 1/2 (ERK1/2) as a loading control. The alignment of the A␣ RNAi target sequence shows the four non-coding base changes (underlined) that confer resistance to down-regulation. B, COS cells were cotransfected with the indicated EE epitope-tagged A␣ subunits and members of B subunit families (B␥, BЈ␤, and BЉ␣ (PR72)) carrying FLAG-epitope tags. A␣ immunoprecipitates were probed for transfected A and B subunits as well as the endogenous C subunit. C, PC6-3 cells with inducible A␣ RNAi were transfected with the indicated cDNAs, and immunoprecipitates of transfected A subunits were probed for A and C subunits. D, A␣ RNAi cells transfected with RNAi-resistant A␣ mutants or empty vector control were treated for 3 days in the absence (Ϫ) or presence (ϩ) of Dox followed by immunoblotting for the indicated proteins. Arrows point to transfected (top) and endogenous (bottom) A subunits, which are best separated in the A␣ DTP177AA transfected lanes. Blots are representative of at least three independent experiments. demonstrating that neither protein can functionally replace the A␣ subunit ( Fig. 6B and data not shown). Because A␣-depleted cells show decreased Akt activity (Fig. 4), we tested whether the expression of membrane-targeted (myristoylated) and, therefore, constitutively active Akt can rescue these cells. Although the expression of myristoylated Akt approached the levels of endogenous Akt (not shown), significant survival was not achieved (Fig. 6B).
The second rescue assay involved selecting A subunit-transfected cells in the presence of Dox for 4 weeks and then deter-mining the number of viable cells. The results of four independent long-term survival assays are summarized in Fig. 6C. The dimer-only and B family-specific A␣ mutants (EE100RR and DWF139HAA) had essentially no long-term rescue activity; the B/BЉ-binding mutant A␣ DTP177HAA provided 16% rescue, whereas A␤ was most effective at 33% survival compared with wild-type A␣.
A potential problem inherent to the short-term survival assays is that A subunit mutants may differentially affect ␤-galactosidase expression, whereas the long-term viability results could be confounded by secondary effects on cell proliferation. To address these concerns and provide conclusive evidence for survival-promoting roles of different PP2A holoenzymes, we created neuronal PC6-3 cells in which Dox treatment leads to the simultaneous knock-down of endogenous A␣ and the induction of RNAi-resistant A␣ mutants. Fig. 7A shows that in these "A-exchange" cells the time course of mutant induction mirrors the down-regulation of endogenous A␣. Importantly, induced A␣ levels approximate endogenous protein levels prior to Dox treatment. Cell viability assays following the exchange of the A subunit confirmed the previous survival assays in that A␣ wild-type inducing cells showed no impairment, whereas cells selected for the integration of an empty vector died over a time course similar to that of parental A␣ RNAi cells (50% viability after 5 days in Dox; Fig. 7B). The induction of the two trimerforming A␣ mutants significantly attenuated death compared with vector control, with the B/BЉ-binding A␣ DTP177AAA offering the better protection (65% viability at 10 days with Dox). Inducible expression of the dimer-only-forming A␣ FIG. 6. Transient and stable viability rescue by A subunit mutants. A, A␣ RNAi cells were transiently transfected with the indicated RNAi-resistant A␣ cDNAs (see legend for panel C) or empty vector in combination with a ␤-galactosidase marker plasmid and incubated in the absence (ϪDox) or presence of Dox for 7 (7d ϩDox) or 10 days (10d ϩDox) to knock-down endogenous A␣. Cell survival was determined by ␤-galactosidase assays and is expressed as means Ϯ S.E. of quadruplicate wells relative to the uninduced control within each transfection group. B, summary of short-term (transient) rescue experiments. Survival (␤-galactosidase activity) after 8 to 11 days in Dox was normalized to A␣ wild-type (w.t.) (100%) and empty vector (0%). Shown are means Ϯ S.E. from the indicated number of experiments. Akt-ca, constitutively active (myristoylated) Akt. C, summary of long-term (stable) rescue experiments. A␣ RNAi cells transfected with the indicated RNAi-resistant A subunit cDNAs were cultured for 26 -30 days in the presence of Dox, after which cell survival was determined by colorimetric metabolism assays (tetrazolium salt reduction). Shown are means Ϯ S.E. from four independent experiments normalized to A␣ wild-type (w.t.) (100%) and empty vector (0%).

FIG. 7. Trimeric but not dimeric PP2A holoenzymes promote survival in A-exchange cells.
A, stable PC6-3 cell lines were generated in which A␣ knock-down is coupled to inducible expression of RNAi-resistant A␣ mutants (A-exchange). The immunoblots depict the time course (0 to 3 days in Dox) of A subunit replacement in the indicated polyclonal cell populations, with the EE-epitope antibody detecting the induced A subunit and the A subunit antibody detecting both endogenous (striped arrowhead in DTP177AAA samples) and induced (solid arrowhead) subunits. B, A-exchange cells were cultured for up to 10 days in the presence of Dox, and viability was determined by a colorimetric reduction to formazan assay. Data points are means Ϯ S.E. of triplicate determinations from one experiment representative of three. w.t., wild-type. EE100RR actually accelerated cell death compared with empty vector-selected cells. This dominant-negative effect is at odds with the transient transfection survival data, which show a small protective effect of the dimeric A␣ mutant (Fig. 6, A and  B). The reasons for this discrepancy are unclear.
In summary, transient, stable, and inducible rescue experiments show that A subunits capable of forming PP2A heterotrimers support cell viability better than an A␣ mutant that can only bind to the catalytic subunit. However, not even the A␣ mutant that binds two of the three regulatory subunit families can sustain long-term growth. These results indicate that all three families of PP2A regulatory subunits are essential for mammalian cell viability. DISCUSSION The two scaffolding, two catalytic, and 12 regulatory subunit-encoding genes in mammals can potentially give rise to 48 distinct heterotrimeric PP2A holoenzymes, with alternative splicing adding further complexity. Here, we have employed an RNAi-based gene replacement strategy to investigate which PP2A holoenzymes are necessary for mammalian cell viability. RNAi of the major PP2A scaffolding subunit in neuronal PC6-3 cells leads to proteosomal degradation of catalytic and B family regulatory subunits, with a concomitant loss of total PP2A activity. When A␣ levels are depleted to ϳ30%, cells begin to die by both apoptotic and non-apoptotic mechanisms, possibly involving defective Akt signaling. Transient or inducible expression of mutant A␣ subunits that form PP2A heterotrimers rescued survival more effectively than a dimer-only-forming mutant. Only RNAi-resistant wild-type A␣ and, to a lesser extent, A␤ were able to support long-term survival of A␣ knockdown cells, strongly suggesting that none of the PP2A regulatory subunit families is dispensable for mammalian cell viability.
PP2A Subunit Stability-We show that monomeric catalytic and B and BЈ family regulatory subunits are lost, whereas PR59 (BЉ) and striatin remain stable following knock-down of the PP2A/A␣ subunit ( Figs. 1 and 2). These findings are in complete agreement with two studies showing that RNAi of the single A and C subunit gene in Drosophila S2 cells promotes loss of B and BЈ subunits (11,16). Drosophila BЉ subunit (dPR72) levels were not investigated in these studies, and striatin has no known Drosophila ortholog. Interestingly, B and BЈ subunits are rapidly removed in response to A␣ RNAi, whereas loss of the catalytic subunit and phosphatase activity is less advanced by the time cell death commences ( Figs. 1 and  2). This finding could reflect differences in the half-life of the monomeric subunits; alternatively or in addition, catalytic subunits in complexes not involving the A subunit (22, 23) may be protected from degradation.
Ectopically expressed, monomeric B␥ mutants are rapidly degraded (7), whereas monomeric BЉ/PR72 mutants are reportedly stable (20). The latter finding is consistent with the observed persistence of endogenous BЉ/PR59 levels after A␣ subunit RNAi (Fig. 2). The classification of striatin and the related SG2NA as members of a fourth class of PP2A regulatory subunits (Bٞ) is questionable, because these proteins lack the signature A subunit interaction motifs common to B, BЈ, and BЉ family subunits (6) and because only a fraction of cellular striatin is associated with PP2A. 2 Striatin's stability in the face of PP2A/A␣ down-regulation (Fig. 2) is thus better reconciled with the behavior of an AC dimer-binding protein as opposed to that of a bona fide PP2A regulatory subunit.
The ability of the proteosome inhibitor MG132 to stall the A␣ RNAi-induced loss of catalytic and B family regulatory sub-units strongly implicates regulation by ubiquitination and proteosomal degradation (Fig. 2) and is consistent with the literature (7,38). The lack of effect of MG132 on BЈ subunit loss could be interpreted to mean that this class of PP2A regulatory subunits is down-regulated by a mechanism different from proteosomal degradation. Alternatively, the rate of new protein synthesis of BЈ␣ and BЈ␦ may not be high enough to replenish protein levels during the ϳ16 h incubation with the proteosome inhibitor (longer incubations were toxic to cells). We favor the latter explanation, because we have observed that MG132 increases levels of ectopically expressed BЈ␤ mutants incapable of binding to the AC dimer. 2 Whether dimeric PP2A is a physiological form of the enzyme rather than an in vitro breakdown product of trimeric PP2A is still not entirely resolved. On the one hand, Gernot Walter's laboratory used a monoclonal antibody specific for the core PP2A dimer to demonstrate that dimeric PP2A constitutes about a third of the total PP2A pool in mammalian cells (39). On the other hand, RNAi of all Drosophila PP2A regulatory subunits results in loss of A and C subunits (11,16). In this report, we show that an A␣ mutant (EE100RR) with strongly compromised binding to all regulatory subunit families can be forcibly expressed to levels similar to those of wild-type A␣ (Fig. 5C). In addition, upon down-regulation of endogenous A␣, expression of this mutant stabilizes catalytic but not regulatory subunits, indicating selective reconstitution of the PP2A dimer (Fig. 5D). Taken together, it would appear that Drosophila PP2A is an obligate heterotrimer, whereas mammalian PP2A can exist in both dimeric and heterotrimeric forms.
PP2A in Cell Survival-Although PP2A also has various death promoting activities (8, 40 -43), this report and previous studies (11,16,21) establish an essential role for this class of enzymes. Furthermore, our rescue experiments underscore the importance of specific PP2A heterotrimers, as opposed to a critical level of total PP2A activity.
The predominant B family regulatory subunits expressed in dividing PC6-3 cells are B␣ and B␦ (12). Because the expression of a selective B family-binding A␣ mutant delays cell death associated with knock-down of the endogenous scaffolding subunit (Figs. 6 and 7), B␣/␦-containing PP2A heterotrimers must have pro-survival functions. Furthermore, because B␣ knockdown cells are viable (Fig. 3A), B␣ and the 89% identical B␦ appear to be redundant, at least for the viability of cultured cells.
Drosophila BЈ subunits (B56-1 and B56-2/widerborst) are important for cell survival because combined RNAi of both subunits triggers rapid caspase-dependent apoptosis in S2 cells (11,16). Similarly, because we find that uncoupling the BЈ subunits from PP2A impairs survival (Figs. 6 and 7), at least one of the five mammalian BЈ subunits must be essential. The rather protracted loss of viability upon the replacement of wild-type A␣ with the BЈ binding-defective mutant is consistent with the recently demonstrated anti-proliferative role of the BЈ␥ subunit (44). Mammals thus appear to have expanded their BЈ subunit repertoire to include tumor suppressors as well as positive growth regulators.
Although they may not be obligatory PP2A subunits in the sense of remaining stable after A␣ knock-down, BЉ subunits clearly contribute to cell viability, because the addition of BЉ binding ability significantly improves the rescue activity of ectopically expressed A␣ mutants (Figs. 6 and 7). Although nuclear forms of PP2A, likely including those containing BЉ subunits, are required for cell cycle progression (18 -20, 45, 46), this requirement cannot be the sole reason for the death of A␣-deficient PC6-3 cells, because nerve growth factor-differentiated A␣ RNAi cells lose viability at the same rate as dividing cells (data not shown).
We show that A␣ down-regulation impairs Akt phosphorylation (Fig. 4), implicating PP2A as a positive regulator of the PI3-kinase/Akt survival signaling cascade. Mechanistically, these results could be explained by PP2A dephosphorylating inhibitory Ser/Thr phosphorylation sites on the epidermal growth factor receptor (47)(48)(49) or its associated proteins (50,51). Intriguingly, Akt has also been shown to be a PP2A substrate (52,53), suggesting that the phosphatase can inhibit Akt signaling under certain conditions. Although stunted Akt activation probably contributes to the cell death associated with PP2A depletion, it is likely only one of many critical dysfunctions in PP2A-depleted cells, because constitutively active Akt failed to rescue cell viability (Fig. 6B).
The second PP2A scaffolding subunit, A␤, is 86% identical to A␣ but is much less abundant than A␣ and appears to bind less tightly to catalytic and B family regulatory subunits (3, 37) (Fig. 5C). On the other hand, the A␤ gene is frequently mutated in human cancers, suggesting a tumor suppressor role (54). We did not detect any compensatory up-regulation of A␤ upon A␣ knock-down in PC6-3 cells. However, transfected A␤ partially rescued in long-term survival assays, which is significant in view of our inability to express A␤ to the same level as A␣. It is thus conceivable that A␤ can substitute for A␣ in certain cell types in which it is highly expressed.
Although A␣ knock-down cells show multiple signs of apoptosis, including the activation of caspases, a broad-spectrum caspase inhibitor commonly used to stall apoptotic cell death fails to do so when A␣ levels drop. This result contrasts with studies of Drosophila S2 cells in which RNAi of several proteins involved in apoptosis prevented death due to RNAi of the PP2A subunit (16). Thus, whereas insect cells can apparently tolerate some loss of PP2A as long as apoptosis is blocked, mammalian cells engage multiple parallel cell death pathways under these conditions.
In apparent contradiction to our study, Collela et al. (55) analyzed 58 human brain tumor samples and found that nearly half expressed little or no detectable A␣/␤ subunit protein. Intriguingly, all samples had near normal levels of catalytic and B family regulatory subunits. Thus, some primary tumors have bypassed the requirement for the A subunit, perhaps by expressing an as yet unidentified alternate scaffold that couples PP2A catalytic and regulatory subunits in a way that facilitates tumorigenesis. The inducible A␣ RNAi cell line utilized in the present study may prove instrumental in identifying such a protein.
Although functional redundancy may exist within PP2A regulatory subunit families, this report provides evidence that each regulatory subunit family has unique and vital functions in the cell. Elucidating these functions will be the subject of future studies involving cell lines with inducible knock-out of select classes of PP2A holoenzymes.