Raf-1-associated Protein Phosphatase 2A as a Positive Regulator of Kinase Activation*

The Raf-1 kinase plays a key role in relaying proliferation signals elicited by mitogens or oncogenes. Raf-1 is regulated by complex and incompletely understood mechanisms including phosphorylation. A number of studies have indicated that phosphorylation of serines 259 and 621 can inhibit the Raf-1 kinase. We show that both serines are hypophosphorylated during early mitogenic stimulation and that hypophosphorylation correlates with peak Raf-1 activation. Concentrations of okadaic acid that selectively inhibit protein phosphatase 2A (PP2A) induce phosphorylation of these residues and prevent maximal activation of the Raf-1 kinase. This effect is mediated via phosphorylation of serine 259. The PP2A core heterodimer forms complexes with Raf-1 in vivo and in vitro. These data identify PP2A as a positive regulator of Raf-1 activation and are the first indication that PP2A may support the activation of an associated kinase.

Mitogenic stimulation of cells typically induces hyperphosphorylation of Raf-1 and a retardation of its migration on SDS gels. This hyperphosphorylation correlates with the down-regulation of Raf-1 kinase activity (9,10) and may be implemented by a negative feedback mechanism depending on MEK activity (10,11). Serines 43,621, and 259 are phosphorylated in resting fibroblasts, albeit to different degrees (12). Phosphorylation of all three residues has been implicated in the negative regulation of Raf-1. Phosphorylation of serine 43 interferes with Ras binding and consequently with Ras-mediated activation (3). Phosphorylated serine 259 and serine 621 represent binding sites for 14-3-3 adaptor proteins (13,14), whose function in Raf-1 activation is controversial. While bivalent binding to Ser 259 and Ser 621 has been suggested to maintain Raf-1 in an inactive conformation (15,16), reversible association with 14-3-3 facilitates Ras-dependent activation in vivo and in vitro (17). In particular, binding to the Ser(P) 621 site appears to be necessary for kinase activity (16,18), a finding that contrasts with the studies indicating that phosphorylation of this site by PKA in vitro is inhibitory (19). Therefore, the significance of Ser 621 phosphorylation is still in question. Its investigation is hampered by the fact that Ser 621 is essential for the catalytic function of Raf-1 and cannot be replaced by other amino acids without loss of kinase activity (19,20). Serine 259 can be phosphorylated by protein kinase B, another Ras effector activated in parallel with Raf, and this phosphorylation correlates with the down-regulation of kinase activity (21). Consistent with an inhibitory role of Ser(P) 259 , mutation of this residue moderately activates the Raf-1 kinase in cultured cells (16,22); the corresponding point mutants display a gain of function phenotype in Drosophila (15,20). Taken together, these data raise the possibility that dephosphorylation of negative regulatory residues plays a role in Raf-1 activation.
Protein phosphatase 2A (PP2A) is a major form of serine/ threonine phosphatase involved in the regulation of signal transduction, growth, and development (23). This class of enzymes consists of a heterotrimer that exists in multiple forms. The core components of all trimeric forms are the 36-kDa catalytic subunit (PP2AC) and the 65-kDa regulatory subunit (A subunit, PR65). This core heterodimer is ubiquitous, and it forms complexes with "variable" subunits of cellular origin (some of which are expressed in a tissue-and/or developmentrestricted manner) as well as with transforming viral antigens (24 -26). Association with variable subunits of cellular and viral origin occurs via the N-terminal leucine-rich repeats of PR65 (27) and confers distinct properties to the enzyme (28). Recently, PP2A has been shown to form a complex with Ca 2ϩ / calmodulin-dependent protein kinase IV (29) as well as with PAK1, PAK3, and p70 S6 kinase (30). The isolated catalytic subunit can associate with casein kinase 2␣ (31). Where investigated (29), PP2A has been shown to contribute to the inactivation of the associated kinase. Here we show that the PP2A inhibitor okadaic acid inhibits full fledged Raf-1 activation. This effect is mediated by a change in the phosphorylation of Ser 259 of Raf-1. In addition, Raf-1 forms stable complexes with PP2A heterodimers. Our results are the first indication that PP2A may support the activation of an associated kinase and highlight the intimate relationship between kinases and phosphatases, which we are just beginning to understand.
Monoclonal Antibody Production-N-terminal His 6 -tagged human PP2AC was expressed in the bacterial strain M15[pREP4], and the fusion polypeptide was purified by Ni 2ϩ -nitrilotriacetic acid chromatography (Qiagen). Balb/c ϫ CBA F1 mice (ϳ3 months old) were immunized by subcutaneous injection with 100 g of bacterially expressed PP2AC fusion protein emulsified with an equal volume of Titremax Gold (CytRx Corp.). The immunization was repeated four times at three monthly intervals, and then the mice rested for 6 months. A further 100 g of fusion protein in phosphate-buffered saline plus 0.1% SDS was injected intraperitoneally at 6 and 3 days prior to sacrifice. Hybridoma lines were then established by fusing splenocytes from the immunized animal with the myeloma line Sp2/0-Ag14 by polyethylene glycol treatment and conventional procedures. Cell lines were screened for antibody production by a modified dot blot procedure using the bacterially expressed fusion protein (34) and cloned three times before tissue culture fluid containing the monoclonal antibody was harvested. Clone F2.8F5 was typed as an IgG2K species using Isotype strips (Roche Molecular Biochemicals) and confirmed as specific for PP2AC by Western blot analysis using total cell lysates. Using deletion mutants of PP2Ac (35), the epitope has been mapped close to the C-terminal end of PP2AC.
Assay of Kinase and Phosphatase Activity-Raf-1 kinase activity was measured as the ability of immunoisolated Raf-1 to phosphorylate recombinant, catalytically inactive MEK-1 (MEK Ϫ ) or to activate recombinant MEK-1 in coupled assays using MBP (41) as the end point of the assay.
Protein Purification and Pull-down Assays-The GST-Raf-1 fusions used in this study have been previously described (42). The proteins were expressed by standard techniques in the baculovirus Sf9 cell system and purified as described previously (43) with the exception that 1% Triton X-100 was added before binding to GST-agarose. For pulldown assays, 100 -200 ng of GST-Raf-1 fusion proteins were incubated with either 1 mg of whole cell lysates or 200 ng of purified heterodimer, PR65␣, or PP2AC for 3-6 h at 4°C. The complexes were recovered by centrifugation and washed in solubilization buffer containing 0.03% SDS, 10 mM dithiothreitol, 0.5 M NaCl prior to Western blot analysis. PP2A heterodimer used in this study was provided by Prof. J. Goris (Leuven, Belgium). Catalytic subunit of PP2A was kindly provided by Dr. Lisa Ballou (I.M.P., Vienna). Both forms of the enzyme were purified from rabbit skeletal muscle (44). Recombinant PR65␣ was kindly provided by Dr. P. Turowski.
In Vivo Labeling of Cells and Phosphotryptic Peptide Mapping-32 P-Labeling of cells was performed as described previously (36). Cell lysis and immunoprecipitation of Raf-1 proteins were performed as described above. 32 P-Labeled proteins were resolved by 7.5% SDS-PAGE, extracted from the gels, and subjected to digestion with sequencing grade trypsin (Promega) according to the manufacturer's instructions prior to phosphopeptide mapping. Tryptic peptides were separated in the first dimension by electrophoresis using pH 8.9 buffer and in the second dimension (chromatography) using a buffer containing n-butanol/pyridine/acetic acid/water (12:10:3:15). Chromatography was allowed to proceed for 20 h. The amount of Raf-1 contained in the immunoprecipitates used for the mapping of the phosphotryptic peptides was determined by immunoblotting an aliquot of the immunoprecipitates, and it was equal in all samples. Phosphotryptic peptide mapping was repeated twice with comparable results.

RESULTS AND DISCUSSION
The PP2A Inhibitor Okadaic Acid Decreases Mitogen-induced Raf-1 Activation and Dephosphorylation: Role of Ser 259 -We have studied Raf-1 phosphorylation and activation in BAC-1.2F5 macrophages stimulated by CSF-1 (45,46). The addition of CSF-1 to quiescent BAC-1.2F5 cells induced robust and transient activation of Raf-1 ( Fig. 1A; Ref. 45). Peak (16fold) Raf-1 activation correlated with the transient dephosphorylation of serine 621, the main residue phosphorylated in quiescent macrophages (Fig. 1B, b). Conversely, the decay of Raf-1 activity to a 4-fold stimulation 15 min post-CSF-1 treatment was accompanied by the hyperphosphorylation of serines 43 and 621. A number of minor yet unidentified residues (spots 2-4) as well as Ser 259 (Fig. 1B, c) were also phosphorylated at this time. Pretreatment with okadaic acid (100 nM; at this concentration a specific inhibitor of PP2A but not PP1␣ (47)), caused a slight increase in basal kinase activity (2-fold) and in Raf-1 phosphorylation in general. Notably, both the peptide pattern and the amount of activity associated with the Raf-1 immunoprecipitates are identical in cells treated with okadaic acid alone and in cells treated with CSF-1 for 15 min (Fig. 1, A  and B, compare c and d). This suggests that both positive and negative regulatory phosphorylation sites are targets of an okadaic acid-sensitive phosphatase and that the balance between phosphorylation of the positive and negative regulatory sites determines the extent of Raf-1 activation. Remarkably, in the presence of okadaic acid, maximal Raf-1 stimulation by CSF-1 was prevented, and only a moderate (5-6-fold) activation could be obtained. Inhibition of early serine 621 dephosphorylation and hyperphosphorylation of several other sites (most prominently serine 259) could be observed concomitantly. Phosphorylation increased further after 15 min of CSF-1 stimulation; by this time, Raf-1 activation had decayed despite the continuous presence of the phosphatase inhibitor (Fig. 1, A and B). Although these changes are complex, it can be appreciated that Ser 259 and Ser 621 are less phosphorylated (hypophosphorylated) in cells treated with CSF-1 alone than in cells treated with mitogen plus okadaic acid and that this "hypophosphorylation" correlates with maximal kinase activation. Thus, an okadaic acid-sensitive phosphatase is involved in Raf-1 dephosphorylation and activation in macrophages.
To determine whether okadaic acid influenced Raf-1 stimulation by receptor tyrosine kinases other than the CSF-1 recep-tor, we analyzed the effect of the inhibitor on COS-1 cells stimulated with EGF. In agreement with the data shown in Fig. 1A, okadaic acid significantly decreased EGF-mediated activation of wild type Raf-1 ( Fig. 2A). To assess whether this effect depended on Ser 259 , COS-1 cells transfected with a Ser 259 3 Ala Raf-1 mutant (RafS259A) were treated with EGF in the presence or absence of okadaic acid. As described previously (12,15,16,20,22), the basal activity of RafS259A was modestly increased with respect to wild type Raf-1. EGF efficiently stimulated RafS259A, but in contrast with the wild type, activation of RafS259A was not decreased by pretreatment with okadaic acid (Fig. 2B). These data confirm the importance of the okadaic acid-sensitive phosphatase in Raf-1 activation by receptor tyrosine kinases and identify Ser 259 as the Raf-1 site relevant for activation. The effect of okadaic acid on the activation of wild type Raf-1 in EGF-treated COS-1 cells was less dramatic than the one observed in BAC-1.2F5 macrophages stimulated by CSF-1. It is possible that the importance of dephosphorylation in Raf-1 activation may vary depending on the cell type; alternatively, Raf-1 overexpression might adversely affect the outcome of the experiment, as is often the case when multienzyme complexes are involved (see below).
Our data are consistent with the hypothesis that mitogen treatment induces a kinase that inactivates Raf-1 by phosphorylating Ser 259 and is counteracted by an okadaic acid-sensitive phosphatase. The strength of Raf-1 activation (or the proportion of Raf-1 molecules that can be activated) would depend on the balance between the activity of these two enzymes. Such a regulation would be particularly opportune in the case of Raf-1, whose moderate activation elicits proliferation, while a strong, prolonged Raf-1 signal leads to cell cycle arrest (48,49). In line with this hypothesis, while this manuscript was in preparation, activated protein kinase B was shown to phosphorylate Ser 259 of Raf-1, to suppress the activation of the mitogen-activated protein kinase cascade, and to shift a cell line from cell cycle arrest to proliferation (21).
Raf-1 Forms Complexes with PP2A in Vivo and in Vitro-Okadaic acid specifically blocked Raf-1 activation at a concentration that selectively inhibits PP2A (47). We therefore examined whether PP2A associated physically with Raf-1. Both the catalytic (PP2AC) and the regulatory (PR65) subunit of the PP2A core enzyme were detected in Raf-1 immunoprecipitates prepared from quiescent as well as CSF-1-treated cells. Neither PP2AC nor PR65 could be detected in precipitates prepared with nonimmune sera (lanes 1 and 2) or with protein A (Fig.  3A). In addition, significant amounts of Raf-1 were present in immunoprecipitates prepared with a monoclonal antibody against PP2AC but not with nonimmune mouse IgG (Fig. 3B), demonstrating the specificity of the interaction. The ubiquitously expressed PP2A core heterodimer binds to different variable subunits of cellular or viral origins, which are involved in regulating its substrate specificity and/or localization. None of the cellular variable subunits tested (p55␣ and -␤, and p72 plus p130) was present in Raf-1 immunoprecipitates (not shown). Consistent findings were obtained in fibroblasts stimulated with EGF (not shown).
We next verified the interaction between Raf-1 and the PP2A heterodimer in COS-7 cells transfected with vectors directing the expression of HA-tagged PP2A subunits (Fig. 4A). Anti-HA immunoprecipitates from cells transfected with the HA-tagged PP2AC contained low amounts of endogenous PR65 and Raf-1 (Fig. 4A, lane 1). Anti-HA immunoprecipitates from cells transfected with the HA-PR65, on the other hand, contained significant amounts of endogenous PP2AC subunit and more Raf-1 (Fig. 4A, compare lanes 1 and 2). Therefore, the amount of Raf-1 detected correlated with the amount of heterodimer pres- The Raf-1-PP2A complex formation was further analyzed by in vitro reconstitution experiments with purified proteins (Fig. 4C). GST-tagged Raf was expressed in Sf-9 cells alone or in combination with v-Ras plus Lck in order to activate it (Raf*). Raf proteins immobilized on glutathione-Sepharose beads were incubated with PR65, PP2AC, or the heterodimer PR65-PP2AC. Consistent with the lack of effect of mitogens on in vivo complex formation, we did not observe significant differences between PP2A binding to Raf or Raf*. PP2AC displayed only weak binding to Raf-1. In contrast, both PR65 and the core heterodimer associated strongly. Thus, as it is the case for cellular and viral subunits (28), PR65 probably plays the key role in the association between Raf-1 and the PP2A het-erodimer. This association, however, is not likely to be direct. Raf-1 does not interact with PR65, PP2AC, or p55 in the yeast two-hybrid system, 2 and size fractionation experiments indicate that in vivo Raf-1 and PP2A are part of a large protein complex (data not shown). Additional proteins, and possibly a variable subunit not detected in our experiments, might also be present in small amounts in the purified enzyme preparations used in the GST pull-down experiments and might be facilitating or even mediating the interaction observed in vitro. In this context, a variable subunit of PP2A has been recently shown to positively regulate Ras signaling upstream of raf during vulval development in Caenorhabditis elegans (50).
Conclusion-Concentrations of okadaic acid that specifically affect PP2A reduced Raf-1 activation, and PP2A was found in Raf-1 immunoprecipitates from quiescent and mitogen-treated cells. Therefore, while an effect of the drug on other phosphatases cannot be formally excluded, PP2A presumably represents the okadaic acid-sensitive phosphatase involved in Raf-1 regulation. Our current working model is that Raf-1-associated PP2A facilitates kinase activation by maintaining Ser 259 in a dephosphorylated state and thereby preventing the formation of inactive 14-3-3-Raf-1 complexes (14 -18). This may permit the activation of a larger number of Raf-1 molecules and prolong it by counteracting the mitogen-induced kinase (probably protein kinase B) that phosphorylates Ser 259 . Ultimately, the Ser 259 kinase must outpace PP2A to terminate Raf-1 activation. Inhibition of PP2A by okadaic acid has been reported to selectively impair Raf-dependent transformation (51). Furthermore, in genetically dissectable organisms, hypomorphic alleles of PP2A suppress the effects of activated Raf (52) or enhance the loss of function phenotype of Raf mutations (50). On the basis of these results, PP2A might have been considered either a Raf-1 effector or a positive regulator of Raf-1 activation. Our findings provide a mechanistic explanation for these observations. By identifying PP2A as a positive regulator of Raf-1, our data define a new function for this phosphatase and add a new facet to the complexity of Raf-1 regulation.

FIG. 2. EGF-stimulated activation of wild type Raf-1, but not of a S259A
Raf-1 mutant, is decreased by okadaic acid in COS-1 cells. COS-1 cells were transfected with a vector encoding human wild type Raf-1 (pCMV5c-raf) or S259A Raf-1 mutant (pCMV5-S259A). 48 h after transfection, cells were treated with okadaic acid (100 nM, 45 min) prior to stimulation with EGF (33 nM, 10 min), lysis, and immunoprecipitation. Raf-1 immunoprecipitates (duplicates) were assayed for kinase activity in a coupled assay. The amount of Raf-1 in the immune complexes was determined by immunoblotting. One representative kinase assay out of three is shown. Differences between samples were below 5% in all cases.

FIG. 3. Endogenous Raf-1 interacts with PP2A heterodimers in vivo.
A, the PP2A core heterodimer is present in Raf-1 immunoprecipitates from quiescent and mitogen-stimulated BAC-1.2F5 cells. Quiescent BAC-1.2F5 cells were stimulated with 6.3 nM mouse recombinant CSF-1 at 37°C for different times prior to solubilization. Raf-1 immunoprecipitates from 1 mg of whole cell lysates were analyzed by Western blotting with antisera directed against the 36-kDa catalytic subunit of PP2A, PR65␣, and Raf-1. 25 g of whole cell lysates (WCL) were loaded as a control. Neither PP2AC nor PR65 were detected in immunoprecipitates prepared using nonimmune rabbit sera (NI) or protein A beads (A) instead of the Raf-1-specific antiserum. The background bands observed in the PP2AC Westerns could also be detected by anti-rabbit antibody alone and represent IgG heavy chains. B, Raf-1 is present in PP2AC immunoprecipitates. BAC-1.2F5 were solubilized, and 1 mg of whole cell lysates were subjected to immunoprecipitation using a monoclonal antibody against PP2AC (PP2AC), nonimmune mouse IgG (NI), or an anti-Raf-1 serum. The immunoprecipitates were analyzed with monoclonal antibodies directed against PP2AC, PR65, or Raf-1. activated protein kinase; Prof. Jozef Goris, Dr. Patric Turowski, and Dr. Lisa Ballou for providing purified preparations of PP2A heterodimer, PR65␣, and PP2AC, respectively; and Dr. Jolanda Schreurs (Chiron Corp. Emeryville, CA) for supplying human recombinant CSF-1. We are indebted to Drs. Egon Ogris and Thomas Decker (Vienna Biocenter) for many helpful discussions and for critically reading the manuscript.