Protein Phosphatase 2A Is Associated with Class C L-type Calcium Channels (Cav1.2) and Antagonizes Channel Phosphorylation by cAMP-dependent Protein Kinase*

Phosphorylation by cAMP-dependent protein kinase (PKA) regulates a vast number of cellular functions. An important target for PKA in brain and heart is the class CL-type Ca2+ channel (Cav1.2). PKA phosphorylates serine 1928 in the central, pore-forming α1C subunit of this channel. Regulation of channel activity by PKA requires a proper balance between phosphorylation and dephosphorylation. For fast and specific signaling, PKA is recruited to this channel by an protein kinase A anchor protein (Davare, M. A., Dong, F., Rubin, C. S., and Hell, J. W. (1999) J. Biol. Chem. 274, 30280–30287). A phosphatase may be associated with the channel to effectively balance serine 1928 phosphorylation by channel-bound PKA. Dephosphorylation of this site is mediated by a serine/threonine phosphatase that is inhibited by okadaic acid and microcystin. We show that immunoprecipitation of the channel complex from rat brain results in coprecipitation of PP2A. Stoichiometric analysis indicates that about 80% of the channel complexes contain PP2A. PP2A directly and stably binds to the C-terminal 557 amino acids of α1C. This interaction does not depend on serine 1928 phosphorylation and is not altered by PP2A catalytic site inhibitors. These results indicate that the PP2A-α1C interaction constitutively recruits PP2A to the channel complex rather than being a transient substrate-catalytic site interaction. Functional assays with the immunoisolated class C channel complex showed that channel-associated PP2A effectively reverses serine 1928 phosphorylation by endogenous PKA. Our findings demonstrate that both PKA and PP2A are integral components of the class C L-type Ca2+ channel that determine the phosphorylation level of serine 1928 and thereby channel activity.

ical and physiological criteria, high-threshold voltage-gated Ca 2ϩ channels are classified as L-type and non-L type channels. Class C (Ca v 1.2) and class D (Ca v 1.3) Ca 2ϩ channels, containing ␣ 1C and ␣ 1D , respectively, constitute most neuronal L-type channels in the brain (2). Ca 2ϩ influx through L-type channels is involved in regulation of membrane excitability, synaptic plasticity, and gene expression (3)(4)(5)(6)(7). In the heart, Ca 2ϩ influx through the class C channel is the critical first step that triggers myocardial contraction.
A crucial signaling pathway that regulates the heart beat is ␤-adrenergic stimulation which results in PKA 1 -mediated phosphorylation of this channel (8), thereby increasing the channel activity (9,10). Only ␣ 1C is required for stimulation of channel activity by PKA (11), although phosphorylation of the ␤ subunit also contributes to the up-regulation of channel activity by PKA (12). PKA phosphorylates ␣ 1C on serine 1928 in vitro and in vivo (8,(13)(14)(15). Mutation of serine 1928 to alanine eliminates PKA-mediated phosphorylation of ␣ 1C and inhibits up-regulation of the channel activity (16). This site is only present in the full-length, 220-kDa form of ␣ 1C . It is deleted by calpain-mediated proteolytic cleavage upon Ca 2ϩ influx through NMDA-type glutamate receptors in neurons, which results in the 180-kDa short form of ␣ 1C (17,18). C-terminal truncation increases the activity of this channel about 4-fold (19). This modification is permanent in contrast to that by PKA-mediated phosphorylation, which is readily reversed by protein phosphatases. The phosphatase inhibitor okadaic acid increased the PKA-stimulated activity of class C channels ectopically expressed in Chinese hamster ovary cells (11). In addition, forskolin did not elevate phosphorylation of ␣ 1C ectopically expressed in HEK293 cells unless cells were preincubated with phosphatase inhibitors (16).
Protein kinase A anchor proteins or AKAPs target PKA to various substrates (20 -22) including class C channels (Ca v 1.2) (15). Disruption of PKA binding to AKAPs prevents PKA-mediated regulation of AMPA-type glutamate receptors (23) and skeletal muscle L-type channels (Ca v 1.1) (24). Phosphorylation of ␣ 1C by PKA is observed when wt AKAP79 but not when a PKA binding-deficient mutant of AKAP79 is co-expressed with the class C channel in HEK293 cells (16). Furthermore, PKA is associated with ␣ 1C in the brain and this interaction may be mediated by the microtubule-associated protein MAP2B (15), the first AKAP to be recognized as such (25). Collectively, these data indicate that PKA anchoring at ␣ 1C is essential for efficient phosphorylation of the channel. Previous studies suggest that phosphorylation and dephosphorylation of ␣ 1C counteract each other in a dynamic way (11,26). We hypothesized that a phosphatase has to be localized at or near the channel for effective reversal of channel phosphorylation. In fact, run-down of class C channel activity in inside-out patches excised from ventricular myocytes was greatly slowed by okadaic acid (26). These results indicate that a phosphatase was attached to the patch, possibly through a direct interaction with the channel complex. We found that the serine/threonine phosphatase PP2A is associated with the class C L-type channel, reversing phosphorylation of ␣ 1C by PKA.

EXPERIMENTAL PROCEDURES
Materials-The PKA inhibitory peptide PKI (5-24) was a gift from Dr. L. M. Graves (University of North Carolina, Chapel Hill, NC). The ECL detection kit and protein G-Sepharose were purchased from Amersham Pharmacia Biotech, microcystin LR and okadaic acid from Calbiochem (San Diego, CA), and purified catalytic subunit of PKA and protein A-Sepharose from Sigma. [ 3 H]PN200-100 (2948 GBq/mmol) was from PerkinElmer Life Sciences and digitonin from Gallard-Schlesinger (Carle Place, NY). Other reagents were obtained from commercial suppliers and were of standard biochemical quality.
Antibodies-The anti-␣ 1C antibody was produced by immunizing rabbits (2) with a fusion protein consisting of the ␣ 1C loop II sequence (see model in Fig. 3A) between residues 783 and 845 fused to the C terminus of GST (see below). The production, characterization, and purification of anti-CH1923-1932P, which specifically binds to phosphorylated serine 1928 in ␣ 1C , was detailed earlier (8,15). Antibodies against the AMPA receptor GluR1 subunit, AKAP150, and GST were as described earlier (15). Anti-PP1␥ was kindly provided by A. C. Nairn (Rockefeller University, New York, NY). Antibodies against PP2A/C were purchased from Transduction Laboratories (Lexington, KY), PP2A/A from Santa Cruz Biotechnology (Santa Cruz, CA), PP2B/calcineurin from Chemicon Int. (Temecula, CA), the HA tag from Babco (Richmond, CA), the T7 tag from Novagen (Madison, WI), and non-immune control antibodies from Zymed Laboratories Inc. (South San Francisco, CA).
Immunoprecipitation from Brain and HEK293 Cells and Immunoblotting-All extracts contained the protease inhibitors pepstatin A (1 g/ml), leupeptin (10 g/ml), aprotinin (20 g/ml), phenylmethanesulfonyl fluoride (200 nM), and calpain inhibitor I and II (8 g/ml each). Crude membrane fractions were prepared from rat brain as described (15), and the cytosolic fraction saved as a source of native PP2A for in vitro interaction assays with ␣ 1C -derived GST fusion proteins. Membranes and HEK293 cells were extracted with solubilization buffer, and insoluble material removed by ultracentrifugation. After preincubation with Sepharose CL-4B to remove proteins that nonspecifically bind to the resin, extracts (0.5 ml) were incubated on ice with either 10 g of affinity-purified anti-␣ 1C antibody, 10 g of control rabbit IgG, 2 l of anti-GluR1 or anti-AKAP150 antiserum, or 0.3 l of anti-HA tag ascites. After 1.5 h, 3-5 mg of protein A-Sepharose were added, samples tilted for 2.5 h, and the resins washed and extracted with 20 l of SDS sample buffer before immunoblotting as described (2,15). All experiments were performed at least three times with comparable results.
Determination of the Stoichiometry of the Class C Channel-PP2A Interaction-Class C channels were quantitatively labeled with the L-type channel-specific ligand [ 3 H]PN200-100 and solubilized with digitonin, which preserves the binding of the ligand to the L-type channels, as described (2,14). Briefly, crude membrane fractions were prepared in the absence of EGTA and EDTA and incubated with a saturating concentration of [ 3 H]PN200-100 (25 nM) for 1 h on ice in the dark. Membranes were collected by ultracentrifugation and channels solubilized for 20 min on ice with 1.5% digitonin, 150 mM NaCl, 10 mM Tris-Cl, pH 7.4, containing protease inhibitors plus 1,10-phenanthroline (200 g/ml). Non-soluble material was removed by ultracentrifugation. Immunoprecipitations were performed with 20 g of affinity-purified anti-␣ 1C antibody or control IgG from rabbit. Immunoresins were washed and split into equal portions. One portion was used for scintillation counting to determine the amount of [ 3 H]PN200-100 and the other portion for immunoblotting with antibodies against PP2A/C. In parallel, increasing amounts of affinity-purified GST-PP2A/C (see next paragraph) was loaded onto the same gel (the protein concentration of PP2A was determined with a BCA protein assay from Pierce). The amount of class C channel present in each immunoprecipitate was calculated from the values obtained by scintillation counting (counting efficiency was about 40% in our hands) after subtracting the amount of [ 3 H]PN200-100 nonspecifically precipitated with the control IgG (about 5% of the quantity obtained by immunoprecipitation with anti-␣ 1C ). We also took into account the off rate for channel-associated [ 3 H]PN200-100 (t1 ⁄2 is about 3 h under our conditions; data not shown). The amount of PP2A/C that coprecipitated with class C channels was determined by matching the PP2A/C immunosignal intensity present in the channel precipitates with signals from the GST-PP2A/C titration curve.
Production of GST and Poly-His Fusion Proteins and in Vitro Binding Assays-␣ 1C and PP2A/C-encoding cDNA templates (31,35) were amplified by PCR with oligonucleotides containing engineered endonuclease restriction sites for subcloning in frame with a 5Ј nucleotide sequence encoding glutathione S-transferase (GST) or a polyhistidine tag and T7 epitope by standard methods (36). Forward primers (containing a BamHI site) and reverse primers (carrying an EcoRI site), respectively, were for the following ␣ antibody production). PP2A/C forward primers were 5ЈGC GGA TCC GAT ATC ATG TAT CCA TAT GAT GTT CC3Ј (for HA tag) and 5ЈGC GGA TCC GGT ACC ATG GAC GAG AAG GTG TTC ACC3Ј (untagged), and the reverse primer was 5ЈGCG TCG ACA AGC TTA CAG GAA GTA GTC TGG3Ј. PCR fragments were cloned into the TA cloning vector PCR.1 (Invitrogen), and inserts were excised with BamHI/EcoRI and subcloned into pGEX 4T-1 (Amersham Pharmacia Biotech) and pTrcHisA (Invitrogen). DNA sequences were confirmed by PCR sequencing with the AmpliTaq system (PerkinElmer Life Sciences). Plasmids for the GST fusion proteins CT1507-1733 and CT1509 -1622 were created as described (37). GST and poly-His fusion proteins were expressed in Escherichia coli (NovaBlue; Novagen, Madison, WI), purified, and used for in vitro interaction assays as described earlier (38,39).
In Vitro Phosphorylation and Dephosphorylation of Serine 1928 -Immunoprecipitated class C channels were resuspended in 45 l of phosphorylation buffer (0.1% Triton X-100, 50 mM HEPES-NaOH, pH 7.4, 10 mM MgCl 2 , 0.5 mM EGTA, 0.5 mM dithiothreitol, pepstatin A (1 g/ml), leupeptin (10 g/ml), aprotinin (20 g/ml)). Phosphorylation was started by addition of 50 M ATP and 5 M cAMP to stimulate channel-associated PKA (15). After incubation for 15 min at 32°C in a thermomixer, endogenous PKA was blocked by adding 1 M PKI. One sample serving as the zero time point for dephosphorylation by the endogenous phosphatase was immediately washed with radioimmunoassay buffer (15) and extracted with SDS sample buffer. Mn 2ϩ (1 mM), which increases the activity of several phosphatases including PP2A in vitro, and, when indicated, protamine (1 mg/ml, final concentration) and okadaic acid (2 nM) were added. Samples were incubated for 2.5 or 40 min as before, washed with radioimmunoassay buffer, extracted with SDS sample buffer, and analyzed by immunoblotting with anti-CH1923-1932P and subsequently with anti-␣ 1C . For phosphorylation of GST-CT1584 -2140, this fusion protein was immobilized on glutathione-Sepharose and incubated in phosphorylation buffer containing 0.5-1 g of the purified catalytic subunit of PKA and 50 M ATP. After washing with 150 mM NaCl, 10 mM Tris-Cl, pH 7.4, it was used in parallel with unphosphorylated GST-CT1584 -2140 for pull-down experiments of PP2A out of brain cytosol.

RESULTS
Okadaic Acid and Microcystin Inhibit Dephosphorylation of ␣ 1C in HEK293 Cells and Brain Extract-An HEK293 cell line stably expressing the wt ␤ 2 -adrenergic receptor (34) was used for transient expression of class C channels (Ca v 1.2) consisting of ␣ 1C , ␣ 2 -␦, and ␤ 2A . Cells were co-transfected with AKAP75, the bovine homolog of human AKAP79 (20). Like class C channels (18), this AKAP may be present at postsynaptic sites (40) and had been used to reconstitute PKA-mediated phosphorylation of ␣ 1C in HEK293 cells (16). For the following experiments we used anti-␣ 1C , an antibody against an epitope in the central part of ␣ 1C (14,17), and anti-CH1923-1932P, an antibody that specifically recognizes ␣ 1C when phosphorylated at serine 1928 (15). Serine 1928 is the main if not only phosphorylation site for PKA on ␣ 1C . Phosphorylation of this site was determined after immunoprecipitation of class C channels with anti-␣ 1C by immunoblotting with anti-CH1923-1932P (15). Treatment of the cultures with 100 nM okadaic acid significantly increased the level of serine 1928 phosphorylation (Fig.  1A). The extent of this increase was comparable to that induced by stimulation of the ␤ 2 -adrenergic receptor with 10 M isoproterenol. These data suggest that phosphorylation by PKA and dephosphorylation by an okadaic acid-sensitive phosphatase of ␣ 1C are in a dynamic equilibrium that is influenced by inhibition of the phosphatase or stimulation of PKA.
PP2A Is Associated with the Class C Channel-Immunoprecipitation of this channel from Triton X-100 extracts of rat cerebral cortex resulted in specific coprecipitation of PP2A ( Fig.  2A) but not of PP1 or PP2B (Fig. 2, B and C). The specificity of this coprecipitation is indicated by the absence of any PP2A immunoreactivity in precipitations with non-immune antibody and in AMPA-type glutamate receptor complexes isolated with antibodies against the GluR1 subunit of this receptor. PP2A is a heterotrimeric holoenzyme that exists in multiple forms composed of a core structure, which interacts with regulatory B subunits. Usually the core enzyme complex consists of the 36-kDa catalytic C subunit (PP2A/C) and the 65-kDa scaffolding A subunit (PP2A/A) and interacts with a B subunit, although the C subunit has recently been found in complexes lacking an A or B subunit (46). 2 PP2A/C and PP2A/A are ubiquitously expressed; each subunit is encoded by two highly related genes (97% and 87% identity, respectively) (47). Both the A and C subunit were detected in the channel complex ( Fig.  2A, middle and bottom panels).
To determine the stoichiometry of the class C channel-PP2A interaction, channels were labeled with saturating amounts of the tritiated form of the L-type-specific ligand isradipine, [ 3 H]PN200-110. After solubilization with digitonin, which affords purification of L-type channels with the ligand bound, and immunoprecipitation of class C channels, the amount of channel in each immunoprecipitate was determined by scintillation counting (see "Experimental Procedures" for more details). An equal portion of each sample was analyzed by immunoblotting with the antibody against PP2A/C. Defined amounts of affinity-purified GST-PP2A/C were loaded onto the same gel to provide a standard curve for quantification of the PP2A/C immunosignal. Our quantitative analysis indicates that, on average, 423 fmol of class C channel were specifically immunoprecipitated with the anti-␣ 1C antibody and that 337 fmol of PP2A/C coprecipitated. Accordingly, 79.7 Ϯ 11.1% (average Ϯ S.E. of three experiments) of the class C channel complexes contain PP2A/C. Because GST-PP2A/C was produced using human cDNA as a template and immunoprecipitations were performed from rat brain, it is important to emphasize that the PP2A/C antibody was made against a sequence (residues 153-2 M. C. Horne, unpublished results.  (34). Cells were treated with control medium or medium containing 100 nM okadaic acid (OA) or 10 M isoproterenol (ISO) for 5 min before lysis, immunoprecipitation (IP) with anti-␣ 1C , and immunoblotting with anti-CH1923-1932P and subsequently anti-␣ 1C . Immunosignals for anti-CH1923-1932P and ␣ 1C long form were quantified by densitometry and the former signals corrected for differences in relative amounts of ␣ 1C long form and normalized to control treatment which corresponds to 100% basal phosphorylation level of serine 1928 (given is the average Ϯ S.D. obtained in three or four experiments or the average Ϯ range from the two experiments with isoproterenol). B, rat forebrain membranes were prepared and solubilized in the presence and absence of 2 M microcystin (Mc) before immunoprecipitation with anti-␣ 1C or nonspecific control antibody and immunoblotting with anti-CH1923-1932P. Blots were stripped with SDS (2) and reprobed with anti-␣ 1C , which detects ␣ 1C long and short form, the latter of which is missing the serine 1928 phosphorylation site. Cont., control. 309) that is 100% conserved in the human ␣ and ␤ and rat ␣ and ␤ isoforms and, therefore, permits a direct comparison of all the PP2A/C immunosignals on a quantitative level.
PP2A/C Binds Directly to the C-terminal Region of ␣ 1C -Voltage-gated Ca 2ϩ channels consist of multiple subunits, and various interaction partner proteins have been identified (1,48). For example, ␣ 1C binds MAP2B (15). MAP2B is an AKAP (25) and may recruit PKA to the channel complex (15). Therefore, we investigated whether PP2A can directly associate with ␣ 1C or whether this association might require an adapter protein. ␣ 1C consists of four domains, which are homologous to each other. The loops between these domains as well as the N and the C terminus are intracellular (Fig. 3A) and available for interaction with intracellular proteins (1,48). GST fusion proteins covering these domains were immobilized on glutathione-Sepharose and incubated with cytosolic rat brain extracts as a source of native PP2A. PP2A/A and PP2A/C associated exclusively with the fusion protein that carries the C-terminal 557 residues of ␣ 1C (Fig. 3A). Loading of the glutathione-Sepharose was comparable for all fusion proteins, as indicated by immunoblotting with antibodies against GST (Fig. 3B).
Because a protein in the cytosolic brain extract could mediate such an interaction, we expressed the last 557 residues of ␣ 1C as poly-His-tagged fusion protein and full-length PP2A/C as a GST fusion protein in two different versions (one contained as additional N-terminal tag an HA epitope). The GST-PP2A/C fusion proteins were adsorbed onto glutathione-Sepharose and the resins incubated with lysates of bacterial cultures expressing the poly-His-tagged C terminus of ␣ 1C . The C terminus effectively bound to both PP2A/C fusion proteins but not GST (Fig. 3C, top panel), indicating that PP2A/C can directly attach to the C terminus of ␣ 1C . Equal amounts of the two PP2A/C fusion proteins and of GST itself, the negative control, were present on the glutathione resins (Fig. 3C, bottom panel). To further corroborate that the PP2A/C-␣ 1C interaction is direct, GST-PP2A/C and, as a negative control, GST was affinitypurified on glutathione-Sepharose and eluted with glutathione. In parallel, poly-His-tagged C terminus, N terminus, and loop II of ␣ 1C were purified on a nickel resin. GST-PP2A/C and GST were added to the poly-His fusion proteins, which were left on the affinity resins to allow subsequent washing steps. The resulting interaction of poly-His-C terminus with GST-PP2A/C was detected by immunoblotting with anti-GST (Fig. 3D, left  panel). It was specific with respect to the C terminus because GST-PP2A/C did not bind to the other two poly-His fusion proteins and with respect to GST-PP2A/C because GST by itself did not interact with any of the three poly-His fusion proteins including the C terminus of ␣ 1C . The presence of similar quantities of GST and GST-PP2A/C in these experiments was indicated by direct immunoblotting of an aliquot of each of these two proteins (Fig. 3D, right panel). The amount of poly-His-C terminus in these experiments was somewhat smaller than that of the other two poly-His fusion proteins, as determined in parallel by immunoblotting with an antibody against the T7 epitope that is part of the N-terminal poly-His tag expressed from pTrcHis vectors (Fig. 3E). The use of quantitatively more control poly-His-N terminus and loop II than poly-His-C terminus in these experiments further bolsters the argument for a specific interaction of PP2A/C with the C terminus.
PP2A/C Binding to ␣ 1C Is Not Driven by a Catalytic Site Interaction-A typical enzyme-substrate interaction is shortlived and ends with the usually fast release of the product. In contrast, the PP2A-␣ 1C association is stable arguing that this association recruits PP2A more permanently to the channel complex and has, therefore, the character of an adapter function for PP2A rather than that of a transient catalytic interaction. Furthermore, PP2A/C directly binds to the C-terminal GST fusion protein of ␣ 1C , which is not phosphorylated at serine 1928 unless pretreated with PKA and ATP (Fig. 4A,  upper panel). Because the dephosphorylated form constitutes a phosphatase product, which would be expected to be released from the catalytic site of PP2A, we would not at all expect that PP2A binding to the non-phosphorylated C terminus of ␣ 1C would be strong and long-lasting if the interaction were purely catalytic. In contrast, the pull-down assays of PP2A with the unphosphorylated C-terminal fusion proteins suggest a constitutive interaction (Figs. 3A and 4A). Phosphorylation of the fusion protein with PKA on serine 1928 did not at all increase PP2A binding (Fig. 4A, middle and bottom panels; compare  lanes 2 and 4). Such an increase in PP2A binding would have been expected if PP2A would only interact with this region in a catalytic manner.
We also evaluated the effect of microcystin on the PP2A-␣ 1C interaction. The crystal structure of the microcystin-PP1 complex shows that microcystin binding overlaps with the catalytic site of PP1, and evidence exists that indicates that the same is true for PP2A (45,49). Inclusion of microcystin did not at all alter binding of cytosolic PP2A to the C-terminal ␣ 1C fusion protein (Fig. 4A, middle and bottom panels; compare lanes 2  and 3). In addition, coprecipitation of PP2A with the class C channel from Triton X-100 brain extracts was not reduced by the presence of microcystin (Fig. 4B), although microcystin has full access to the channel-associated PP2A, as indicated by its ability to block dephosphorylation of serine 1928 (Fig. 1B). Collectively, these data indicate that the PP2A-␣ 1C association is not merely a catalytic site interaction. It is rather stable and FIG. 2. Co-immunoprecipitation of PP2A/A and PP2A/C with class C channels. Triton X-100 extracts of brain membrane fractions were employed for immunoprecipitations of class C channel complexes (anti-␣ 1C ), AMPA-type glutamate receptor complexes (anti-GluR1), and nonspecific control IgG from rabbit. Immunoblotting with antibodies against PP2A/A and C (A) but not with those against the ␥ isoform of PP1 (B) or PP2B (C) yielded positive results showing the specific coprecipitation of these two PP2A subunits with ␣ 1C . Molecular mass markers are indicated on the right (in kDa). IP, immunoprecipitation; Cont., control.
allows the constitutive recruitment of PP2A to the channel complex.
Association of a BЈ Subunit with the PP2A-␣ 1C Complex in a Heterologous Expression System-The PP2A A-C core complex usually associates with one subunit of a heterogeneous class of regulatory B subunits. More than twenty B subunits are now known, many of which are expressed in a tissue and cell specific manner. B subunits have been divided into the B, BЈ, and BЉ classes, although some of the more recently identified B subunits do not belong to any of these three classes (32, 50 -52). These subunits are thought to be involved in determining the substrate specificity of the PP2A holoenzyme (50,53,54). As a first attempt to identify whether a B subunit might be present in the ␣ 1C -associated PP2A complex, class C channel immunoprecipitates were analyzed by immunoblotting with antibodies against various B type B subunits including B␣, B␤, and B␥, with negative results (data not shown).
Because expression of BЈ␥ (B56␦ in Ref. 51) determined by Northern analysis is most prominent in the brain (32), we investigated whether this B subunit may associate with the channel complex. An HA-tagged form of BЈ␥ (32) was ectopically co-expressed with class C channels in HEK293 cells. Immunoprecipitation of ␣ 1C resulted in specific coprecipitation of BЈ␥ (Fig. 5A). The anti-HA immunosignal associated with the channel was comparable to that obtained with 5% of the input. Accordingly, about 5% of the total amount of HA-tagged BЈ␥ subunit is associated with ␣ 1C . A similar estimate is obtained by comparing the anti-HA signal in the channel complex with that resulting from immunoprecipitation of HA-tagged BЈ␥ with the anti-HA antibody. The HA-BЈ␥ immunosignal on the blot obtained after immunoprecipitation with anti-HA is more than 10-fold higher than that in the 5% input lane, showing that the anti-HA antibody precipitate most of the HA-tagged BЈ␥ (Fig. 5A).
To get an impression of how much PP2A/C subunit relative to BЈ␥ might be present in the channel complex, HA-tagged PP2A/C and untagged PP2A/A was coexpressed with the channel in HEK293 cells. Although overexpression of the PP2A A and C subunits cannot usually be up-regulated by more than 1.5-2-fold because PP2A is toxic at higher levels (31,55,56), the presence of the HA tag in PP2A/C affords immunoprecipitation of HA-PP2A with high efficacy as shown above for HA-BЈ␥ (see Fig. 5A). Immunoprecipitation of ␣ 1C resulted in a weak immunosignal for the PP2A/C HA tag, which was 1% or less of the HA tag signal obtained by immunoprecipitation of HA-PP2A/C with the anti-HA antibody (Fig. 5B). Despite the low level of the immunoreactivity in the channel precipitation, the signal was clearly specific because immunoprecipitation with non-immune control antibody from the same cell extracts did not result in any HA immunosignal. Furthermore, the HA signal was also absent when the ectopically expressed channel was immunoisolated from HEK293 cells, which had been cotransfected with green fluorescent protein instead of HA-PP2A/C. Similar results were obtained when PP2A/A levels in  A and B, resin samples were incubated with rat forebrain cytosol as a source of native PP2A complexes, washed, and analyzed by immunoblotting with antibodies against PP2A/A and C (A) and against GST to evaluate whether comparable amounts of fusion proteins were present on each resin sample (B). C, GST and two different PP2A/C GST fusion proteins (which were identical except one carried an HA tag) were immobilized on glutathione-Sepharose, incubated with 200 l of E. coli lysates containing ␣ 1C C terminus as poly-His fusion protein, washed, and analyzed by immunoblotting with an antibody against the T7 epitope present in the poly-His C-terminal fusion protein (upper panel). Probing the lower portion of the blot with antibodies against GST indicated the presence of equal amounts of the two PP2A/C fusion proteins and GST. As expected, only the T7 epitope of the poly-His C-terminal fusion protein (upper panel) but no GST fusion protein (lower panel) is detectable when 10 l of the bacterial lysate containing the poly-His fusion protein of the ␣ 1C C terminus is loaded (left lane; this amount corresponds to 5% of what was used for the pull-down experiments). D and E, GST-PP2A/C and plain GST were purified on glutathione-Sepharose and eluted with glutathione. Aliquots with equal amounts of GST-PP2A/C and GST were either directly applied to SDS-polyacrylamide gel electrophoresis and subsequent immunoblotting with anti-GST (D, right side of blot; of note, similar quantities of GST and GST-PP2A/C were present) or after a pull-down assay with poly-His fusion proteins of the N and C terminus and loop II of ␣ 1C purified and immobilized on nickel-nitrilotriacetic acid resin (Qiagen, Chatsworth, CA; D, left side of blot). In parallel, aliquots of the affinity-purified poly-His fusion proteins were directly compared by immunoblotting with an antibody against the T7 epitope (E). Arrowheads indicate the position of the full-length fusion proteins. Molecular mass are indicated on the right side of each blot (in kDa).
immunoprecipitates with antibodies against ␣ 1C and PP2A/A were compared; only a small portion of PP2A/A was associated with the channel compared with the amount precipitated with the anti-PP2A/A antibody (Fig. 5C).
Channel-associated PP2A Antagonizes Phosphorylation of ␣ 1C by PKA-To evaluate the physiological relevance of the constitutive interaction of PP2A with the class C channel, we tested whether channel-associated PP2A can effectively dephosphorylate serine 1928. Class C channels were solubilized and immunoprecipitated in the absence of phosphatase inhibitors, which may block PP2A activity in the subsequent steps. The channel immunocomplex containing PKA (15) was incubated for 15 min with ATP and cAMP. Under these conditions the channel-associated PKA phosphorylates serine 1928 to near completion (15). Further phosphorylation was then stopped by adding the PKI (5-24) peptide, which effectively blocks the channel-associated PKA (15). In parallel Mn 2ϩ was added to increase the phosphatase activity (57,58). Dephosphorylation was not very effective under these conditions (Fig.  6A). However, when protamine was added, substantial dephosphorylation occurred within 2-3 min (Fig. 6A). Protamine is a polyamine that selectively activates PP2A (58 -61). Although protamine is selectively expressed in sperm cells, it has been speculated that it may be able to mimic endogenous as yet unidentified PP2A activating factors. Our results indicate that PP2A stably bound to the class C channel effectively dephosphorylates serine 1928 after its activation. Protamine-stimulated dephosphorylation of serine 1928 was highly sensitive to okadaic acid. This toxin blocked the dephosphorylation at a concentration of 1 nM over an extended incubation time of 40 min (Fig. 6B). Because PP2A/A and C are coprecipitating with ␣ 1C and because only PP2A (and its close relatives PP4 and PP5) are blocked by such low concentrations of okadaic acid (41,42,45,54), these functional studies argue that the phosphatase that dephosphorylates serine 1928 is PP2A. DISCUSSION Our results indicate that about 80% of the class C channel (Ca v 1.2) in rat forebrain are associated with PP2A. The rest of the channel complexes that lacked PP2A may represent an intracellular channel pool that had not yet bound PP2A. It is also possible that some PP2A dissociated from the channel during the whole immunoprecipitation procedure. The channel interacts with a PP2A holoenzyme that consists of the A and C subunit and likely a BЈ type B subunit. A comparison of the PP2A/A and PP2A/C immunosignals present in ␣ 1C immuno- precipitates from brain with those from samples of the load suggest that 1% or less of total PP2A might be associated with ␣ 1C in brain ( Fig. 2A). This result is not unexpected because PP2A is an abundant ubiquitous enzyme with multiple substrates, ␣ 1C being only one of many targets. We were not able to detect a specific B subunit in the channel complex isolated from rat brain due, in part, to a lack of availability of antibodies with high affinity for most B subunit types. However, in HEK293 cells, the coexpressed HA-tagged BЈ␥ subunit interacted with the channel. Of note, a significantly larger portion of the total BЈ␥ than of the PP2A/A and C pool coprecipitated with the channel, probably as much as 5% of the total amount of BЈ␥ present in the cell lysates. The fact that a larger fraction of the pool of ectopically expressed BЈ␥ than of the ectopically expressed PP2A/C and PP2A/A pool is associated with the channel suggests that the PP2A holoenzyme containing BЈ␥ targets only a subset of all PP2A substrates in these cells. It has been hypothesized that B subunits regulate targeting and substrate specificity of the trimeric PP2A holoenzyme (54).
Overexpression in mammalian cell lines can promote the interaction of proteins, which are usually not associated with each other in vivo. However, ectopic expression of PP2A/A and C does not increase their level of interaction with ␣ 1C as compared with brain; it appears that a significantly smaller portion of the total PP2A/A and C pool coprecipitated with ␣ 1C from HEK293 cells than from brain (compare the ratios of PP2A immunosignals of load or PP2A immunoprecipitates versus ␣ 1C immunoprecipitates of Figs. 2A and 5 (B and C)). These results validate our experimental approach of co-expression of BЈ␥ as a first step in demonstrating that this or a structurally related BЈ type subunit may be the B subunit present in the class C channel complex in brain. This notion is further supported by the observation that a significantly larger fraction of the BЈ␥ pool is associated with ␣ 1C than of overexpressed PP2A/A or PP2A/C. Because BЈ type subunits are very similar to each other (amino acid identity is usually higher than 50 -60% (Refs. 32 and 51)), it is quite possible that another BЈ type subunit is associated with class C channels in vivo but that the overexpression promoted the interaction with BЈ␥. It is important to note that of the BЈ type B subunits described to data, BЈ␥ is most abundantly expressed in brain. However, confirmation of the presence of BЈ␥ or another BЈ type subunit has to await the production of antibodies with higher affinity that can specifically identify BЈ␥ or its isoform in channel immunoprecipitates.
PKA has to be anchored by AKAPs in close proximity to substrates such as AMPA-type glutamate receptors (23) and L-type channels (16,24) for their efficient phosphorylation. Some AKAPs can directly bind to these PKA targets. Examples include the interaction of MAP2B with the class C channel (15), of AKAP79/150 and AKAP250 with the ␤ 2 -adrenergic receptor (62,63), and of AKAP350/yotiao with the NMDA receptor (64). Phosphorylation is only an efficient way for regulating the activity of ion channels or enzymes if it is counteracted by dephosphorylation. Because PKA is stably bound to the class C channel (15), we hypothesized that a phosphatase may also have to be constitutively associated with the channel to effectively balance PKA-mediated phosphorylation. We show that the phosphatase PP2A not only forms a stable complex with ␣ 1C but also reverses its phosphorylation by PKA.
It has been known for a while that AKAP79/150 can act as a scaffolding molecule for PKA, protein kinase C, and PP2B/ calcineurin, suggesting that kinase-phosphatase complexes are anchored next to their target proteins to ensure fast phosphorylation and dephosphorylation (22). However, only a few examples have been described, which demonstrate that anchor-ing of phosphatases in close proximity to their targets is as important for dephosphorylation as kinase anchoring is for phosphorylation. These include docking of PP1 by spinophilin near AMPA-type glutamate receptors (65) and by AKAP350/ yotiao at NMDA-type glutamate receptors (64) and binding of PP2A to the ␤ 2 -adrenergic receptor (60), and Ca 2ϩ -and calmodulin-dependent kinase IV (66).
The functional relevance of an anchored okadaic acid-sensitive phosphatase regulating the class C channel activity had already been suggested by earlier experiments in heart tissue, where the class C channel induces and regulates myocardial contraction. Okadaic acid inhibited run-down of channel activity in inside-out patches of ventricular myocytes (26). These results suggest that a phosphatase was attached to the patch at or near the channel complex. Class C channels exist in three different modes (67). They are not available for activation in mode 0, show brief openings in mode 1, and long-lasting openings and brief closings in mode 2 (67). Transition from mode 0 to mode 1 or 2 is induced by the ␤-adrenergic receptor-PKA signaling pathway (10). Okadaic acid inhibited while application of PP2A/C to excised inside-out membrane patches promoted the reversal of mode 2 and 1 in cardiac and smooth muscle cells (68 -70). These electrophysiological findings together with our biochemical evidence demonstrating that PP2A reverses phosphorylation of serine 1928 in ␣ 1C raise the possibility that phosphorylation of this site by PKA and its dephosphorylation by PP2A is involved in controlling the transition between mode 1 and 2. Switching from mode 1 to mode 0 is less sensitive to okadaic acid but more susceptible to inhibition by calyculin A than the transition from mode 2 to mode 1 (68,69). Calyculin A blocks PP1 with higher potency than PP2A. These results indicate that PP1 may play a larger role in reversing mode 1 to mode 0 than PP2A. PKA phosphorylates not only serine 1928 in ␣ 1C but also serine 478 and serine 479 in the ␤ 2a isoform of the Ca 2ϩ channel ␤ subunit (12). Phosphorylation of serine 478 and serine 479 in ␤ 2a contributes to up-regulation of the channel activity by PKA (12). We do not know whether phosphorylation of serine 1928 in ␣ 1C or serine 478 or 479 in the ␤ 2a subunit are at all regulating the transitions between modes 0, 1, and 2 and whether PP1 dephosphorylates the latter two sites. However, it is tempting to speculate that phosphorylation of serine 478 or 479 of ␤ 2a is involved in the transition from mode 0 to mode 1 and that these sites are dephosphorylated by PP1, whereas phosphorylation of serine 1928 in ␣ 1C , which is dephosphorylated by PP2A, may play a role in the transition from mode 1 to mode 2. In any case, our findings indicate that precise localization PP2A and regulation of its activity in brain and heart is likely to be crucial for the correct physiological function of class C L-type Ca 2ϩ channels.