12-O-Tetradecanoylphorbol-13-acetate-induced dephosphorylation of protein kinase Calpha correlates with the presence of a membrane-associated protein phosphatase 2A heterotrimer.

Protein kinase C signaling is desensitized through a combination of dephosphorylation and proteolysis in intact cells. The process of dephosphorylation is analyzed here, as well as its relationship to degradation. It is established for protein kinase Cα that dephosphorylation occurs in a membrane compartment following activation and temporally preceding significant degradation. The phosphatase responsible for the dephosphorylation appears to be a heterotrimeric type 2A phosphatase, which is shown to be in part constitutively membrane associated. Consistent with a role for this activity, okadaic acid is shown to inhibit the phorbol ester-induced dephosphorylation of protein kinase C that occurs in intact cells. Furthermore, phorbol ester-induced down-regulation of protein kinase Cα is shown not to be dependent on the rate of dephosphorylation, indicating that these desensitizing pathways may operate in parallel.

Protein kinase C signaling is desensitized through a combination of dephosphorylation and proteolysis in intact cells. The process of dephosphorylation is analyzed here, as well as its relationship to degradation. It is established for protein kinase C␣ that dephosphorylation occurs in a membrane compartment following activation and temporally preceding significant degradation. The phosphatase responsible for the dephosphorylation appears to be a heterotrimeric type 2A phosphatase, which is shown to be in part constitutively membrane associated. Consistent with a role for this activity, okadaic acid is shown to inhibit the phorbol ester-induced dephosphorylation of protein kinase C that occurs in intact cells. Furthermore, phorbol esterinduced down-regulation of protein kinase C␣ is shown not to be dependent on the rate of dephosphorylation, indicating that these desensitizing pathways may operate in parallel.
Members of the protein kinase C (PKC) 1 family constitute a class of diacylglycerol-dependent protein kinases (reviewed in Refs. [1][2][3][4]. These ligand-dependent protein kinases display constitutive catalytic activity on partial proteolysis in vitro, indicating the presence of a fully functional kinase domain that does not require ligand-dependent modification for activity (see Ref. 5). Consistent with this view, point mutations in the autoinhibitory domain of PKC␣ render the protein constitutively active, i.e. ligand-independent (6). Thus PKC could be considered to behave like the cAMP-dependent protein kinase, albeit with tethered regulatory and catalytic domains. Indeed, much of our understanding of PKC function in vivo reflects its transient, allosteric activation by its effectors, a paradigm much exploited for cAMP-dependent protein kinase.
Although this view of a ligand-dependent protein kinase encompasses numerous observations of PKC behavior, it is well established that at least for PKC␣ and PKC␤, phosphorylation controls intrinsic catalytic potential (see Ref. 7 and references therein). This was first demonstrated for the PKC␣ isotype, both in vivo (8) and in vitro (9). In the former case, the dephos-phorylated, inactive form of PKC␣ accumulated following chronic activation by the phorbol ester TPA (8). It is clear that in other contexts, chronic activation by TPA also leads to the generation of dephosphorylated forms that migrate faster on SDS-polyacrylamide gel electrophoresis (PAGE) (for example, see Ref. 10); these forms usually accumulate in association with the cytoskeleton. In view of the loss of activity of PKC␣ on dephosphorylation, it is anticipated that in vivo this event serves an important role in desensitizing responses through the PKC pathway.
There is limited information available as to the specific components involved in the desensitization of PKC. In the case of down-regulation, it has been established that the loss of protein is a consequence of increased proteolysis (11). Furthermore, there is evidence that this is driven by some form of vesicle-dependent sorting process (12), but the molecular details remain elusive. Nothing is known about the dephosphorylation of PKC in intact cells or the relationship between this process and the coincident down-regulation that is observed. To understand the selective dephosphorylation of PKC␣ and its role, it is important to define the molecular mechanism(s) involved. This study describes the identification of a membrane-associated protein phosphatase 2A heterotrimer that acts on PKC␣ in vitro and in intact cells and further analyses the relationship between the proteolytic degradation and dephosphorylation of PKC␣.

EXPERIMENTAL PROCEDURES
Materials-Okadaic acid and other biochemical reagents were obtained from Sigma unless stated otherwise. Other chemicals were of the highest grade available. Antisera to PKC␣ and to subunits of protein phosphatases 1 and 2A were as described previously (13)(14)(15)(16). The horseradish peroxidase-conjugated donkey anti-rabbit IgG, ECL detection reagents, and [␥-32 P]ATP were from Amersham Corp.
Cell Culture and Transfection-COS cells were grown in culture as described previously (17). Where indicated, subconfluent COS cells were transfected by electroporation as detailed elsewhere (18). Where analyzed for down-regulation of wild-type and mutant PKC␣, cells were treated with TPA for the times indicated and harvested directly into sample buffer (19). Proteins were fractionated by SDS-PAGE on 10% acrylamide gels, and PKC␣ was detected by Western analysis.
Protein Analysis-Samples were subjected to SDS-PAGE (19). Separated proteins were transferred to Immobilon membranes (Millipore) and probed with PKC or protein phosphatase antisera as described previously (20). Detection of immunoreaction was by ECL. Protein concentration was measured by the method of Bradford (21).
Cell Extraction and Fractionation-For the partial purification of the membrane-associated protein phosphatase, COS cells were harvested and snap frozen in aliquots of ϳ10 8 cells. Extracts from frozen COS cells were prepared in 20 mM Tris-HCl, pH 7.5, containing 2 mM EDTA, 5 mM EGTA, 10 mM benzamidine, 100 g/ml phenylmethylsulfonyl fluoride, 100 g/ml leupeptin, 100 g/ml aprotinin, and 0.3% ␤-mercaptoethanol; 2 ml of extraction buffer was used per 10 8 cells. The extract was homogenized using a Dounce homogenizer, and the crude membrane fraction was prepared by centrifugation at 14,000 rpm in a bench-top microfuge.
The membrane fraction was either assayed directly for PKC␣-phosphatase activity, extracted in sample buffer for Western analysis, or extracted in the above extraction buffer additionally containing 1% Triton X-100. This neutral detergent extract was clarified by centrifugation at 14,000 rpm as above and then further fractionated on a HiTrap Q cartridge (Pharmacia Biotech Inc.) attached to a fast protein liquid chromatographic apparatus (Pharmacia) and equilibrated in the same Triton X-100-containing buffer. Proteins were eluted in the same buffer using a linear gradient from 0 to 1.0 M NaCl; 0.5-ml fractions were collected.
PKC␣ Protein Phosphatase Assays-PKC␣ was purified from PKC␣ baculovirus-infected Hi5 insect cells as previously documented for Sf9 cells (22). To assay cellular fractions and solubilized protein phosphatase, an aliquot of the purified PKC␣ (0.4 g) was mixed with 20 l of cell extract or column fraction (as indicated in the text) in the presence of 30 l of extraction buffer, 15 l of lipids (or buffer as indicated), and H 2 O to give a final volume of 73 l. Lipids were prepared by sonication of nitrogen-dried phosphatidylserine and TPA into 20 mM Tris-HCl, pH 7.5, yielding concentrations of 5 mg/ml and 5 g/ml, respectively. Incubations were carried out at 30°C for 10 min and then stopped by the addition of 27 l of 4 ϫ SDS-PAGE sample buffer. Dephosphorylation of the PKC␣ was monitored by its shift in migration on SDS-PAGE using a high acrylamide (7.5%)/bisacrylamide (0.06%) ratio. The different phosphoforms of the protein were visualized by Western analysis.

RESULTS
Preliminary data indicated that dephosphorylation of COS cell PKC␣ could be detected within 60 min of TPA treatment (see below). To assess the compartment in which PKC␣ is dephosphorylated, COS cells were stimulated with TPA and fractionated. Cytosol, the Triton X-100-soluble membrane fraction, and the Triton X-100-insoluble fraction were analyzed. As shown in Fig. 1, untreated cells showed an exclusively cytosolic localization of PKC␣. However, after 60 min of TPA treatment, the PKC␣ antigen was no longer present in the cytosolic fraction but was found in the Triton X-100-soluble membrane fraction. The antigen migrated as a doublet characteristic of phospho-and dephospho-PKC␣ species. No dephosphorylation was observed following much shorter treatments with TPA (15 min; not shown) even though PKC␣ becomes fully membrane asso-ciated (e.g. see Ref. 20). The phosphatase responsible for PKC␣ dephosphorylation thus appears to act in a membrane compartment.
To determine whether PKC-directed phosphatase activity is present in membranes from these cells, fractionation was carried out, and membranes were tested for their ability to dephosphorylate recombinant PKC␣. The presence of a PKC␣phosphatase in the membrane fraction is clearly illustrated by the data shown in Fig. 2. This activity was inhibited by microcystin (1 M), confirming the nonproteolytic nature of the altered PKC␣ migration and identifying the phosphatase activity as a type 1 or 2A phosphatase. Furthermore, the dephosphorylation was greatly enhanced by the presence of phosphatidylserine and TPA, suggesting that the active conformation of PKC␣ was the preferred substrate. To substantiate the substrate-directed nature of the effect of phosphatidylserine and TPA, an alternative substrate was assessed. Phosphatase activity toward 32 P-labeled histone could not be stimulated by phosphatidylserine and TPA; in fact, a modest inhibition was observed (data not shown).
Since the dephosphorylation of PKC␣ in intact cells is observed following TPA treatment, it was possible that the activ- The membrane-associated PKC␣-phosphatase activity was extracted in Triton X-100 and fractionated on a Mono Q anion exchange column. A, the PKC␣-phosphatase activity detected by altered migration of dephosphorylated PKC␣ peaked in fractions 11 and 12. B, the fractions through the peak of phosphatase activity (9 -15) were analyzed for protein phosphatase 1 and 2A catalytic subunits. Whole cell extracts (extract) were reactive with both protein phosphatase 1c (PP1c) and protein phosphatase 2Ac (PP2Ac) antisera, demonstrating the presence of these 37-kDa subunits. Antiserum to protein phosphatase 1c detected no antigen in the fractionated membrane samples (not shown), whereas the 2Ac serum identified the 37-kDa subunit coincident with the PKC␣-phosphatase activity. C, the same Mono Q fractions as above were analyzed for the p55␣ and p65␣ subunits by Western blotting. Arrows, reactive antigens migrating at 55 and 65 kDa. ity of the membrane-associated phosphatase might increase following TPA treatment of cells. To investigate this, the recovery of the membrane-associated phosphatase activity was also determined after TPA stimulation (Fig. 3). Pretreatment of cells with TPA did not affect the PKC␣-phosphatase activity (see further below). It should be noted that the PKC␣ could be fully dephosphorylated in vitro under these conditions and that the phosphatase activity determinations are rate estimates and not end point assays. These data indicate that the TPAinduced dephosphorylation in intact cells is a PKC-directed control, consistent with the selective effect TPA has on PKC␣ dephosphorylation in vitro.
The membrane-associated phosphatase activity was found to be stable to extraction in Triton X-100 (not shown), permitting fractionation and analysis. The activity eluted as a single peak on anion exchange chromatography, eluting at ϳ450 mM NaCl (see Fig. 4A).
Western analysis with antisera to the catalytic subunits of type 1 and 2A phosphatases, although detecting antigens in the crude extract (Fig. 4B, left panel), demonstrated that only the phosphatase 2A catalytic subunit was detectable in these membranes, and this coeluted with the PKC␣-phosphatase (Fig.  4B). To assess the subunit composition of the phosphatase, antisera selective for the 130-and 72-kDa proteins and the 65-kDa ␣, 65-kDa ␤, 55-kDa ␣, and 55-kDa ␤ subunits were used. As shown in Fig. 3C the 65-and 55-kDa ␣ proteins were present; none of the other antigens was detected (data not shown). The 55-and 65-kDa subunits eluted precisely with the catalytic subunit, indicating the presence of the typical hetero-trimeric form of protein phosphatase 2A.
Having established the identity of the membrane-associated PKC␣-phosphatase, confirmation of the TPA-independent localization (see Fig. 3) was obtained. Control and TPA-stimulated cells were fractionated, and protein phosphatase 2A subunits were localized. It is evident (Fig. 5) that the heterotrimeric complex is in part constitutively membraneassociated; this distribution is entirely insensitive to TPA treatment. The lack of TPA response on the localization of the phosphatase is consistent with the effect of TPA being PKC␣ (i.e. substrate)-directed.
To assess the role of protein phosphatase 2A in intact cells, PKC␣-transfected COS cells were treated with TPA, and the effect of okadaic acid (OA) was determined. As illustrated in Fig. 6, okadaic acid inhibited TPA-induced dephosphorylation with an almost complete effect at 500 nM. The slow time-dependent increase in PKC␣ dephosphorylation even in the presence of OA (Fig. 5B) suggests either incomplete inhibition by OA or the action of an additional inhibitor-insensitive activity.
It would appear from Fig. 6A that OA may protect PKC␣ from TPA-induced degradation. However, the OA effect on antigen loss was variable, and furthermore, any response to OA would not distinguish between an effect on the proteolytic "machinery" and an effect via inhibition of PKC␣ dephosphorylation. To resolve the relationship between TPA-induced PKC␣ dephosphorylation and degradation, use was made of the T638A PKC␣ phosphorylation site mutant, which is rapidly dephosphorylated in vivo on TPA stimulation (23). It is evident from Fig. 7 that the wild-type PKC␣ and the T638A PKC␣ down-regulate at similar rates. In fact, wild-type PKC␣ is lost more rapidly than the T638A mutant, with 20% (wild-type) and 34% (T638A) of antigen remaining after 3 h of TPA treatment. DISCUSSION It is demonstrated here that the TPA-dependent dephosphorylation of PKC␣ at the membrane is okadaic acid-sensitive in vivo, correlating with the presence of a membrane-associated protein phosphatase 2A heterotrimeric complex. TPA does not appear to influence the membrane-associated heterotrimeric protein phosphatase 2A activity or its subcellular distribution; the effect on PKC␣ dephosphorylation correlates with TPAinduced PKC activation and the associated conformational change (7). It is further shown that although TPA-induced PKC␣ dephosphorylation is inhibited in vivo by okadaic acid, dephosphorylation is not rate-limiting with respect to downregulation; these two desensitizing pathways appear to operate independently. Thus, although it is possible that other okadaic acid-sensitive phosphatases may act on PKC␣ in vivo, the simplest interpretation is that the membrane-associated, heterotrimeric 2A complex is responsible for this desensitizing process.
Following chronic activation of PKC by exposure to the poorly metabolized activator TPA, TPA-responsive PKC isotypes are frequently found to become down-regulated through increased proteolysis (e.g. Ref. 11). Associated with this downregulation is a shift in mobility on SDS-PAGE of the remaining antigen and an accumulation in the Triton X-100-insoluble cytoskeleton fraction (e.g. Refs. 8 and 10). The faster migrating forms that accumulate resemble the fully dephosphorylated species and, consistent with this, appear to be inactive (8). 2 The down-regulation of PKC␣ and cytoskeleton association is shown here to be preceded by a dephosphorylation of the membrane-associated, activated PKC. PKC␣ is phosphorylated on at least three sites in vivo (24), 3 and the extent of the shift in migration is consistent with the complete dephosphorylation of the protein. The time frame of these events is minutes or hours, not seconds; hence this would seem to be part of an adaptive process responding to chronic activation. It should be noted that prolonged activation and down-regulation is not unique to TPA stimulation but is observed also on exposure to physiological agonists (e.g. Refs. 25 and 26).
The inhibition of TPA-induced PKC␣ dephosphorylation by OA gave a variable effect on the subsequent loss of polypeptide, normally associated with chronic TPA exposure. To resolve any possible causal relationship between dephosphorylation and degradation of PKC␣, a mutant of PKC␣ (T638A), which is hypersensitive to dephosphorylation (23), was used. In this context, when the rate of dephosphorylation was very greatly increased, no difference was observed with respect to PKC␣ polypeptide loss. Thus the rate of dephosphorylation does not determine the rate of down-regulation.
As yet there is no evidence on the specific nature of the membrane compartment in which PKC␣ is dephosphorylated. Although there is no cause-and-effect relationship between dephosphorylation and down-regulation, it is evident that these events occur within the same time frame. We have shown previously that the process of down-regulation appears to be a vesicle traffic-directed event (12). It is thus quite likely that the membrane-associated heterotrimeric protein phosphatase 2A complex is also present in the same vesicle compartment and is not (solely) resident at the plasma membrane.
Although the heterotrimeric protein phosphatase 2A complex was first identified as a soluble activity, a membraneassociated form(s) of protein phosphatase 2A has been identified in T cell membranes (27). In COS cells, approximately 10% of the protein phosphatase 2A catalytic subunit is membraneassociated (see Fig. 5). Whether there is a specific membranetargeting protein or whether there are modifications to one or more subunits rendering the protein membrane-bound remains to be determined.
During the course of these studies, a similar particulate protein phosphatase 2A complex was identified as being responsible for the inactivation of c-Raf (28) and for the dephosphorylation of the ␤-adrenergic receptor (29). It would appear that this membrane-associated activity plays a key role in the desensitization and reversal of a number of distinct signaling events. In the case of PKC␣ the trigger for dephosphorylation is the TPA-induced membrane association and change in conformation. No change in the localization of the protein phosphatase 2A complex or in its extractable membrane-associated activity was observed following TPA treatment of cells.
The existence of a mechanism for the dephosphorylation (and inactivation) of PKC␣ suggests that either sustained activation has detrimental effects on the cell and hence must be desensitized or that the cytoskeleton-associated, dephosphorylated protein that accumulates has some unique function in vivo. Although this distinction remains to be resolved, the simple view would be that this is part of a desensitization process. Since ceramide has been shown to activate protein phosphatase 2A (30), it was possible that increased dephosphorylation of PKC may play a role in the antagonism between ceramide (or perhaps a metabolite) and PKC function. However, in COS cells treated acutely (15 min) with either short chain ceramides or sphingomyelinase, no difference in the rate of TPA-induced dephosphorylation of PKC␣ was observed (data not shown). This contrasts with recent data for T cell lines, in which on prolonged exposure in the absence of TPA, ceramide has been shown to induce PKC␣ dephosphorylation (31). Whether this distinction reflects cell specificity or subcellular localization remains to be determined.
In conclusion, PKC␣ is shown to be dephosphorylated by a membrane-associated heterotrimeric protein phosphatase 2A complex. This process is triggered by the membrane-associated activation of PKC␣ itself. Whether physiological modifiers of protein phosphatase 2A activity control the extent of activation of PKC␣ remains to be determined. However, irrespective of the acute control of the dephosphorylation itself, it would appear that this step temporally precedes the enhanced degradation observed in response to agonists and thus plays a primary role in desensitizing this pathway.