Modulation of the G Protein Regulator Phosducin by Ca2+/Calmodulin-dependent Protein Kinase II Phosphorylation and 14-3-3 Protein Binding*

Phototransduction is a canonical G protein-mediated cascade of retinal photoreceptor cells that transforms photons into neural responses. Phosducin (Pd) is a Gβγ-binding protein that is highly expressed in photoreceptors. Pd is phosphorylated in dark-adapted retina and is dephosphorylated in response to light. Dephosphorylated Pd binds Gβγ with high affinity and inhibits the interaction of Gβγ with Gα or other effectors, whereas phosphorylated Pd does not. These results have led to the hypothesis that Pd down-regulates the light response. Consequently, it is important to understand the mechanisms of regulation of Pd phosphorylation. We have previously shown that phosphorylation of Pd by cAMP-dependent protein kinase moderately inhibits its association with Gβγ. In this study, we report that Pd was rapidly phosphorylated by Ca2+/calmodulin-dependent kinase II, resulting in 100-fold greater inhibition of Gβγ binding than cAMP-dependent protein kinase phosphorylation. Furthermore, Pd phosphorylation by Ca2+/calmodulin-dependent kinase II at Ser-54 and Ser-73 led to binding of the phosphoserine-binding protein 14-3-3. Importantly, in vivodecreases in Ca2+ concentration blocked the interaction of Pd with 14-3-3, indicating that Ca2+ controls the phosphorylation state of Ser-54 and Ser-73 in vivo. These results are consistent with a role for Pd in Ca2+-dependent light adaptation processes in photoreceptor cells and also suggest other possible physiological functions.

The phototransduction cascade of vertebrate photoreceptor cells is a well studied G protein-mediated signaling pathway that has been a model system for much of our understanding of G protein signaling. Sensitivity in this system is modulated to allow responsiveness over several orders of magnitude of light intensity. Signal response is maximized in the absence of light ("dark adaptation") and dampened as background illumination increases ("light adaptation"). Several independent molecular events, many involving Ca 2ϩ , are believed to underlie this regulation (see Refs. 1 and 2 for reviews). Cytosolic Ca 2ϩ levels in rod outer segments vary from their dark resting concentration of ϳ500 nM to below 50 nM when the Ca 2ϩ channels close as a result of the light signal (3). This decrease in Ca 2ϩ concentration triggers a number of feedback responses believed to be involved in light adaptation (2). These responses include an increase in phosphorylation of rhodopsin (4), which causes rhodopsin inactivation through arrestin binding, and an increase in guanylyl cyclase activity (5), which restores [cGMP] after its depletion by light-activated cGMP phosphodiesterase.
Phosducin (Pd), 1 an abundant protein in retinal photoreceptors and the developmentally related pineal gland (6,7), is believed to play a role in modulation of phototransduction and other G protein pathways (8 -10) by virtue of its ability to bind G protein ␤␥-heterodimers (G␤␥) with high affinity (11)(12)(13). When bound to Pd, G␤␥ is sterically blocked from interacting with G␣ subunits (10,14) or other G␤␥ effectors (15)(16)(17). Thus, Pd has been shown to down-regulate G protein signals in photoreceptors (9 -10) and other cell types in vitro (8) as well as in overexpression experiments (15,18). In other G protein systems that require G␤␥ for receptor phosphorylation to initiate receptor inactivation, similar in vitro and overexpression experiments have shown that Pd enhances G protein signaling by blocking the binding of G␤␥ to G protein receptor kinase-2 or -3 (18 -20).
Phosphorylation of Pd by cAMP-dependent protein kinase (PKA) significantly diminishes its ability to inhibit G protein signals (8,10,15). In photoreceptor cells, the phosphorylation state of Pd is light-dependent, with maximal phosphorylation occurring in the dark (21). It has therefore been proposed that Pd is a feedback regulator of the light signal in a phosphorylationdependent manner (22). In this hypothesis, dephosphorylation of Pd in response to light initiates its binding to transducin-␤␥ (G t ␤␥). Pd binding inhibits the association of G t ␤␥ with transducin-␣ (G t ␣) and dampens signaling. In the dark, phosphorylation of Pd results in its dissociation from G t ␤␥, allowing maximal signaling through the G protein.
In cells overexpressing Pd, stimulation of PKA activity results in a pronounced decrease in the ability of Pd to inhibit G protein signals. However, PKA phosphorylation decreases the binding affinity of Pd for G t ␤␥ by only 3-fold (13,23). In addition, PKA phosphorylation only partially diminishes the ability of recombinant Pd to inhibit G t ␤␥ interactions with G t ␣ (24). Structural studies of the phosphorylated Pd⅐G t ␤␥ complex explain the small difference in binding affinity. PKA phosphorylation occurs at Ser-73 (25) and results in disruption of a helixcapping motif in helix 2 of Pd, causing an order-to-disorder transition of a 20-residue stretch and producing a 15% loss of surface area contact between Pd and G t ␤␥ (24). The lost contacts overlap the G t ␣-binding site on G t ␤␥. Thus, Pd phosphorylation increases the access of G t ␣ to its binding site without opening it up entirely. However, these studies do not explain how these minor differences in G t ␤␥ binding that occur upon PKA phosphorylation translate into the striking inhibition of Pd activity that occurs in cells upon activation of PKA by 8-bromo-cAMP (8-Br-cAMP) (15). One possible explanation is that other cellular kinases also participate in the inactivation of Pd.
Analysis of the sequence of Pd revealed a consensus site for phosphorylation by Ca 2ϩ /calmodulin-dependent kinase II (CaMKII) at residues 51-54 (26). Pagh-Roehl et al. (27) have reported Ca 2ϩ /calmodulin (CaM)-dependent phosphorylation of fish Pd, but such phosphorylation was not observed in a mammalian rod outer segment (ROS) preparation (28). CaMKII is found in various tissues and is abundant in neural cells, where it plays a variety of roles (see Refs. 29 and 30 for reviews). In view of the role of Ca 2ϩ in light adaptation, the presence of a CaMKII consensus site in Pd, and the CaM-dependent phosphorylation seen in teleosts, inquiry into the possible phosphorylation of Pd by CaMKII is warranted.
The same region of Pd that contains the potential CaMKII phosphorylation site also constitutes a consensus recognition site for the phosphoserine-binding protein 14-3-3. The family of 14-3-3 proteins was originally discovered in brain and is involved in numerous signal transduction pathways (see Refs. 31 and 32 for reviews). 14-3-3 proteins are widely expressed in neuronal tissues and many other cell types in a broad range of eukaryotes from yeast to humans. Searching the human genome data base revealed 20 putative 14-3-3 genes (33), of which two isoforms (⑀ and ) have been found at high levels in the retina (34).
Mass spectrometric analysis of proteins copurifying with Pd from the retina identified 14-3-3, suggesting a possible interaction between Pd and 14-3-3 (35). This observation led us to investigate the possibility that Pd could bind 14-3-3 proteins in an interaction regulated by CaMKII phosphorylation. We have found that Pd is indeed phosphorylated by CaMKII and that such phosphorylation dramatically inhibits the interaction of Pd with G t ␤␥ while enabling 14-3-3 binding. We have also found that in vivo modulation of the Ca 2ϩ concentration results in the phosphorylation of two serine residues that are required for interaction with 14-3-3 and that block G t ␤␥ binding.

Determination of CaM-dependent Kinase Phosphorylation of Pd-
Recombinant mouse CaMKII-␣ (a generous gift of Dr. Thomas R. Soderling, Vollum Institute, Oregon Health Sciences University) was produced in a baculovirus expression system as described (36). Phosphorylations were carried out at 1.5 M CaMKII in 50 mM HEPES (pH 7.5) with 10 mM magnesium acetate, 50 M calcium chloride, 0.1 M bovine CaM, 0.5 mM ATP, and 0.05 Ci/l [␥-32 P]ATP. The substrate (His 6 -tagged recombinant rat Pd (14,37) or Syntide 2 (Fluka Chemika-Biochemika)) was used at 10 M. Reactions were incubated at 30°C for the times indicated and quenched with 4ϫ Laemmli sample buffer (38). Samples were run on 12% SDS-polyacrylamide gels or 15% Tris/Tricine peptide gels (39), and the gels were stained with Coomassie reagent, dried, and exposed to a phosphor screen for analysis on a Storm 860-D PhosphorImager (Molecular Dynamics, Inc.). Data were fit to a firstorder rate equation: PPd ϭ PPd max (1 Ϫ e Ϫkt ), where PPd is phosphorylated Pd, k is the pseudo first-order rate constant, and t is the time in minutes. Values for k and PPd max were obtained from the curve fit.
Extracts were prepared from whole bovine retinas (obtained from Deseret Meat, Spanish Fork, UT) that were dark-adapted in their eye cups (Ͻ45 min postmortem) in HEPES/Ringer buffer (10 mM HEPES (pH 7.5), 120 mM sodium chloride, 3.5 mM potassium chloride, 0.2 mM calcium chloride, 0.2 mM magnesium chloride, 0.1 mM EDTA, 10 mM glucose, and 1 mM dithiothreitol) for 1 h. Single retinas were suspended in 1 ml of hypotonic buffer (20 mM HEPES (pH 7.5), 1 mM dithiothreitol, and 0.2 mM EGTA) and dissociated with a tissue homogenizer, followed by 10 passages through a 25-gauge hypodermic needle. Cellular debris was pelleted by centrifugation at 27,000 ϫ g for 20 min. Intact rod outer segments (IROS) were prepared as previously detailed (40), and IROS pellets were subsequently resuspended in hypotonic buffer, disrupted by 10 passages through a 25-gauge needle, and clarified of debris by spinning at 27,000 ϫ g for 20 min. Total protein was determined using Coomassie Plus reagent (Pierce). Phosphorylation of 5 M recombinant Pd (rPd) by kinases in these extracts (2 g/l and 66 ng/l total protein for retinal or IROS extract, respectively) was carried out in 20 mM HEPES (pH 7.5), 100 mM potassium chloride, 20 mM sodium chloride, 5 mM magnesium acetate, 1 mM dithiothreitol, 0.2 mM EGTA, 1 M microcystin LR, and 1 mM ATP (500 mCi/mmol [␥-32 P]ATP) at room temperature. Reactions were stopped between 1 and 20 min by addition of Laemmli sample buffer, electrophoresed, stained, dried, and analyzed with the PhosphorImager. Data were fit to a first-order rate equation as described above, and initial rates were determined from the derivative of the first-order rate equation at t ϭ 0, which is equal to PPd max ⅐k. Inhibitors were used as follows: PKA inhibitor (PKI)- Measurements of Pd/G␤␥ Interactions-Surface plasmon resonance binding experiments were performed as previously described (13) with unphosphorylated Pd, CaMKII-phosphorylated Pd, or PKA-phosphorylated Pd. CaMKII phosphorylation reactions were carried out for 10 min as described above. PKA phosphorylation reactions were as described previously (13). Pd was assayed for its ability to inhibit the binding of G t to light-activated rhodopsin (Rho*) and its inhibition of G t ␤␥ binding to non-illuminated, urea-stripped ROS disc membranes as previously detailed (10). Serine-to-alanine mutants were created using oligonucleotide-directed, polymerase chain reaction-based mutagenesis employing the QuikChange protocol (Stratagene).
Identification of CaMKII Phosphorylation Sites-rPd that had been phosphorylated with CaMKII as described above was analyzed by LC-MS on a PE Sciex API III triple quadrupole mass spectrometer with an IonSpray source. On-line liquid chromatography employed a 250-m inner diameter fused silica capillary column packed with POROS R4 resin (PerSeptive Biosystems) with a 0 -80% acetonitrile gradient developed over 40 min in 0.1% formic acid using an Applied Biosystems Model 140A HPLC at 20 l/min. Scans were stepped at 0.1 m/z with a 0.5-ms dwell time. In addition to analyses of the intact protein, phosphorylated Pd was reduced with 4 mM dithiothreitol; alkylated with 40 mM 4-vinylpyridine under argon; acetone-precipitated; resuspended in 200 mM Tris (pH 8.0); and proteolyzed with modified trypsin (Promega), endoproteinase Lys-C (Wako Bioproducts), or endoproteinase Asp-N (Roche Molecular Biochemicals) at an enzyme/substrate ratio of 1:10 (w/w). Digests were then separated and analyzed by LC-MS using the conditions described. Parent ions were chosen following data analysis, and LC-MS/MS was subsequently performed using argon as a collision gas, with the scans stepped at 0.2 m/z with a 0.8-ms dwell time.
Determination of Phospho-Pd/14-3-3 Interactions-GST-Pd fusion proteins were created as previously described for the phosducin-like protein N-terminal domain (40). Fusion proteins were phosphorylated with CaMKII as detailed above. (CaMKII does not phosphorylate GST alone (data not shown).) Recombinant His 6 -tagged 14-3-3 (a gift from Dr. Haian Fu, Emory University) was expressed as described for other such constructs (14) and radioiodinated using the Bolton-Hunter method (41). A 200-l mixture of 1.5 M GST-Pd fusion protein and 1.5 M 125 I-labeled 14-3-3 was incubated in phosphate-buffered saline with 1 mg/ml bovine serum albumin at room temperature for 15 min, after which 100 l of a 50% slurry of glutathione beads (Amersham Pharmacia Biotech) was added. Samples were then incubated for 1 h at 4°C; beads were centrifuged briefly and washed three times with phosphate-buffered saline plus 1 mg/ml bovine serum albumin; and pellets were analyzed in a ␥-counter. Counts were converted to picomoles of 14-3-3 using the determined specific activity.
For the measurement of binding of Pd from cell extracts, 14-3-3 beads were prepared by adding 1 mg of His 6 -tagged 14-3-3 to a 200-l bed volume of Probond Ni 2ϩ resin (Invitrogen) in equilibration buffer (20 mM Tris (pH 7.5), 120 mM sodium chloride, 50 mM imidazole, Ϯ1 M microcystin LR) and incubated for 1 h at 4°C. The beads were then washed with 10 bed volumes of equilibration buffer. Fresh bovine eye cups (Ͻ45 min postmortem) were cut and incubated for 2 h in HEPES/ Ringer buffer at room temperature in the dark or light. The retinas were removed and homogenized in equilibration buffer. The samples were then centrifuged at 30,000 ϫ g for 20 min to remove debris. One milliliter of 10 mg/ml total supernatant protein was loaded onto the 14-3-3 beads and incubated with gentle mixing for 1 h at 4°C. After incubation, the beads were washed five times with 1 ml of equilibration buffer/wash, and the bound material was eluted in 1 ml of equilibration buffer containing 8 M urea. Samples were run on 12% SDS-polyacrylamide gels, and phosphorylated Pd was detected via Western analysis using a previously described antibody (40).
For Ca 2ϩ chelation experiments, eye cups from light-adapted eyes (Ͻ45 min postmortem) were incubated at room temperature in the dark for 2 h in HEPES/Ringer buffer without Ca 2ϩ , but with 50 M BAPTA/AM or 0.2% Me 2 SO vehicle. Retinas were extracted, and Pd binding to the 14-3-3 beads was determined as described above.

RESULTS
Pd Phosphorylation by CaMKII-To determine if Pd is a substrate for a CaM-dependent kinase, phosphorylation of purified rPd by recombinant CaMKII-␣ (36) was measured in vitro. Compared with phosphorylation of Syntide 2, a peptide substrate derived from synapsin I (42), CaMKII incorporated 5-fold more phosphate/mol of Pd, with a similar pseudo firstorder rate constant (1.4/min for Syntide 2 and 0.94/min for Pd) (Fig. 1A). Syntide 2 contains one phosphorylation site; thus, it appears that there are approximately five CaMKII phosphorylation sites on Pd. These data demonstrate that Pd is an excellent substrate for this kinase.
To begin to assess the physiological significance of CaMKII phosphorylation of Pd, extracts of whole bovine retinas and IROS were tested for Ca 2ϩ /CaM-dependent phosphorylation of Pd ( Fig. 1, B and C). In these experiments, the effect of Ca 2ϩ , CaM, or specific kinase inhibitors on the rate of phosphorylation of exogenously added rPd was measured. Addition of Ca 2ϩ or Ca 2ϩ /CaM caused an 8-fold increase in the initial rate of Pd phosphorylation in retinal extract (Fig. 1B). In IROS extract, addition of Ca 2ϩ alone caused a 4-fold initial rate increase, whereas addition of both Ca 2ϩ and CaM increased the phosphorylation rate 7-fold (Fig. 1C). This difference in the CaM . Data from the number of different extracts indicated were combined, and the initial rates of Pd phosphorylation were determined. Error bars represent the error associated with the non-linear least-squares fit of the data to obtain the initial rates. C, CaMKII in IROS extract phosphorylates Pd. Extracts of IROSs prepared from dark-adapted bovine retinas were incubated with rPd and other components as described for B. Data from three different extracts were combined, and the initial rates of Pd phosphorylation were determined as described for B. D, CaMKII and PKA in IROS extract phosphorylate Pd with similar rates. Extracts of IROSs prepared from dark-adapted bovine retinas were incubated with rPd in the presence of [ 32 P]ATP and (from left to right) buffer only, 0.5 mM Ca 2ϩ ϩ 10 M bovine calmodulin ϩ 10 M PKI, 50 M 8-bromo-cAMP (8-Br cAMP) ϩ 10 M AIP, 0.5 mM Ca 2ϩ ϩ 10 M bovine calmodulin ϩ 50 M 8-bromo-cAMP, or 10 M AIP ϩ 10 M PKI. Data from three different extracts were combined, and the initial rates of Pd phosphorylation were determined as described for B. requirement between whole retinal and IROS extracts suggests that a portion of the endogenous CaM may have been lost when the outer segments were dislodged from the retina. Inhibition of PKA with the specific peptide inhibitor PKI-(5-24) (43) had no significant effect on the Ca 2ϩ -dependent phosphorylation of Pd in either extract, indicating that Ca 2ϩ was not indirectly activating PKA through the Ca 2ϩ /calmodulin-dependent adenylyl cyclase (22) in these membrane-free extracts. In contrast, inhibition of CaM-dependent kinases using a peptide corresponding to the autoinhibitory domain of CaMKII ([Ala 286 ]CaMK inhibitor-(281-301)), which inhibits CaMKII and CaMKIV due to the relatedness of the respective autoinhibitory domains (30), caused a dramatic reduction in the Ca 2ϩdependent phosphorylation of Pd. Both retinal and IROS extracts exhibited a substantial decrease in initial rate in the presence of this inhibitor (Fig. 1, B and C). A second peptide inhibitor, AIP (44), which is highly selective for CaMKII, but not for CaMKIV, inhibited Pd phosphorylation very effectively in the IROS extracts, whereas it did not inhibit Pd phosphorylation in whole retinal extracts. At the 10 M concentration used in this experiment, AIP has been shown to completely inhibit CaMKII-␣, whereas it has no effect on CaMKIV-␣ and -␤ (44). These inhibitor results suggest that CaMKIV is responsible for Pd phosphorylation in whole retinal extracts, whereas CaMKII is responsible for Pd phosphorylation in IROS extracts. Together, the data demonstrate that Pd is a good substrate for CaMKII and that CaMKII is highly enriched in the same photoreceptor cells in which Pd is found.
PKA is an established regulator of Pd activity in photoreceptor cells. It has been shown to phosphorylate Pd in vivo (25) and in IROS preparations (45,46). Thus, it was important to compare the activity of CaMKII with that of PKA in IROS extracts to assess the relative abundance of each kinase. To assure that each kinase was measured independently of the other, CaMKII activity was measured in the presence of the PKA inhibitor PKI, whereas PKA activity was measured in the presence of the CaMKII inhibitor AIP. Ca 2ϩ /CaM (0.5 mM and 10 M, respectively) and 8-bromo-cAMP (50 M) concentrations were chosen that would maximally activate their respective kinase. Very similar initial rates of Pd phosphorylation were observed when CaMKII or PKA was activated (Fig. 1D), indicating that the extracts contain similar amounts of both kinases. Consistent with this observation, when both kinases were activated together, the initial rate of Pd phosphorylation doubled. Both CaMKII and PKA are soluble, cytosolic enzymes that would be expected to experience similar losses during the IROS preparation, suggesting that in the in vivo state, photoreceptor cells may express similar amounts of these two kinases.
CaMKII Phosphorylation of Pd Affects Its Interaction with G t ␤␥-We have previously reported that PKA-phosphorylated Pd is impaired in its ability to bind photoreceptor G t ␤␥ and to affect the interaction of transducin (G t ) with Rho* (10, 13). To assess the functional importance of CaM-dependent phosphorylation, we carried out similar investigations using CaMKII to phosphorylate Pd.
Direct binding of Pd to G t ␤␥ was monitored using BIAcore surface plasmon resonance measurements. Pd was phosphorylated to saturation with CaMKII or PKA, corresponding to ϳ5 mol of phosphate/mol of Pd with CaMKII (see Fig. 1A) or ϳ1 mol of phosphate/mol of Pd with PKA. CaMKII phosphorylation decreased the binding of Pd to G t ␤␥ dramatically, whereas PKA phosphorylation decreased binding only modestly ( Fig.  2A). Global fit analysis (13) of families of curves generated from four different G t ␤␥ concentrations with unphosphorylated Pd yielded rate constant values of 1.3 ϫ 10 5 M Ϫ1 s Ϫ1 for k on and 2.9 ϫ 10 Ϫ3 s Ϫ1 for k off , with a corresponding K d value of 2.3 ϫ 10 Ϫ8 M. With CaMKII-phosphorylated Pd, the k on was decreased ϳ50-fold to 2.5 ϫ 10 3 M Ϫ1 s Ϫ1 , whereas the k off was increased ϳ5-fold to 1.6 ϫ 10 Ϫ2 s Ϫ1 , resulting in a nearly 300-fold increase in K d to 6.5 ϫ 10 Ϫ6 M. In contrast, PKA phosphorylation of Pd caused only a 3-fold increase in K d . This means that CaMKII phosphorylation blocks the Pd/G t ␤␥ interaction nearly 100-fold better than does PKA phosphorylation.
The relevance of Pd phosphorylation by CaMKII in the G protein signaling pathway of rod photoreceptors was investigated using two functional assays of Pd that we had previously developed (10). The first of these assays measures the ability of Pd to inhibit the binding of G t to Rho*, which it does by preventing the association of G t ␤␥ with G t ␣ and the formation of the G t heterotrimer required for binding to Rho*. Unphosphorylated Pd effectively inhibited this interaction, with an IC 50 of 0.15 M (Fig. 2B). Phosphorylation with PKA resulted in a 17-fold decrease in Pd inhibition, whereas CaMKII phosphorylation reduced Pd inhibition by 57-fold. The striking inactivation of Pd by CaMKII phosphorylation is consistent with the large difference in binding affinity observed in the direct binding measurements.
The second functional assay measures the ability of Pd to inhibit G t ␤␥ binding to urea-stripped ROS disc membranes. The C-terminal domain of Pd interacts with the membranebinding face of G t ␤␥ (14) and thus blocks binding of G t ␤␥ to the membrane (10). Unphosphorylated Pd inhibited membrane binding of G t ␤␥, with an IC 50 of 0.5 M (Fig. 2C). PKA phosphorylation had little effect on this property of Pd, whereas CaMKII phosphorylation almost completely eliminated the ability of Pd to inhibit membrane binding. Thus, phosphorylation of Pd by CaMKII dramatically decreases its affinity for G t ␤␥ and its ability to block interactions of G t ␤␥ with G t ␣ and with membranes. Notably, the effect of CaMKII phosphorylation was much more pronounced than that of PKA phosphorylation.
Specific Phosphoserine Residues Involved in the Inhibition of G t ␤␥ Binding-Mass spectrometric techniques were employed to determine the CaMKII phosphorylation sites in rPd. LC-MS of intact CaMKII-phosphorylated rPd indicated that the most abundant population contained five phosphates (30,558 atomic mass units compared with an unphosphorylated mass of 30,163 atomic mass units) (data not shown), consistent with the comparison of phosphate incorporation into Syntide 2 mentioned above. The protein was digested with trypsin, endoproteinase Lys-C, or endoproteinase Asp-N protease, and the resultant digests were analyzed by LC-MS. Confirmation of phosphopeptides and identification of phosphorylated residues were accomplished by subsequent LC-MS/MS.
A tryptic fragment of 2854 atomic mass units was confirmed to be a peptide that spans the region linking the hexahistidine tag and the amino terminus of Pd, Ϫ3 GSHMEEAAS*Q-SLEEDFEGQATHTGPK 23 , by LC-MS/MS of the triply charged molecular ion at m/z 952.3 (Fig. 3A). Fragment ions of the b type (see Ref. 47 for nomenclature) were consistent with dehydroalanine (the product of loss of a labile phosphate) at Ser-6 in this peptide. Fragment ions alone leave some room for ambiguity between phosphorylation at Ser-6 or Ser-8 since there was no observable b ion at either m/z 882 or 900. A 1590-atomic mass unit proteolytic product that resulted from digestion with endoproteinase Asp-N was confirmed to be the peptide 28 (Fig. 3C) gave b-type ions showing a dehydroalanine at Ser-54. Notable is the peak at m/z 840.6 and the lack of a peak at m/z 858.5, indicating phosphorylation at Ser-54 and not at Ser-55. A 1714-atomic mass unit product of endoproteinase Lys-C digestion was confirmed as 72 MS*IQEYELIHQDK 84 . LC-MS/MS of the triply charged m/z 572.5 ion (Fig. 3D) gave b-type ions consistent with dehydroalanine at the only serine in this sequence. Finally, an endoproteinase Lys-C peptide of 3033 atomic mass units was shown to be 105 LS*FGPRYGFVYELETGEQFLETIEK 129 . LC-MS/MS of the triply charged ion at m/z 1012.0 (Fig. 3E) gave b-type ions consistent with dehydroalanine and phosphoserine at Ser-106 and not at the threonines in this peptide. To assure that these serines constituted a complete set of CaMKII phosphorylation sites, alanine substitutions were made at Ser-6, Ser-36, Ser-54, Ser-73, and Ser-106, and the ability of CaMKII to phosphorylate this variant was measured as described for Fig. 1A. No phosphorylation was observed (data not shown), indicating that serines 6, 36, 54, 73, and 106 are the complete set of sites. This result also suggests that Ser-8 (which was not ruled out in the mass spectrometric analysis) is not phosphorylated.
To determine the role of each of the five serine residues phosphorylated by CaMKII in regulating Pd function, combinations of serine-to-alanine mutations were prepared. They were tested for their ability to inhibit association of G t ␤␥ with G t ␣ and their subsequent binding to Rho* (Fig. 4A) and to inhibit G t ␤␥ binding to the ROS disc membranes (Fig. 4B). The elimination of any one of the phosphorylation sites caused little change in the phosphorylation-dependent decrease in the inhibition of G t ␤␥ binding to G t ␣ (Fig. 4A). Combinations of twophosphorylation site substitutions were somewhat less sensitive to phosphorylation than were single-site substitutions in this assay, whereas combinations of three or more substitutions began to markedly affect the ability of CaMKII phosphorylation to perturb G t ␤␥ binding. The contribution from all sites was not equivalent, however. Substitution of Ser-6 had little effect, whether in isolation or in combination with other mutants. This observation is consistent with a report from Lee 2 that Pd phosphorylated at Ser-6 purified from bovine retinas is not significantly different from unphosphorylated Pd in its affinity for G t ␤␥. Mutation of Ser-36 created a five-substitution variant that behaved significantly more like the unphosphorylated wild-type protein than did the four-substitution mutant, indicating that phosphorylation at this residue does have an effect on G t ␤␥ binding. A triple mutant (S6A/S36A/S106A) that left only Ser-54 and Ser-73 intact was as ineffective as phosphorylated wild-type Pd, indicating that phosphorylation at Ser-54 and Ser-73 plays a significant role. In control experiments, the unphosphorylated serine-to-alanine variants behaved identically to unphosphorylated wild-type Pd (data not shown). Combined, these data suggest that phosphorylation at four of the five sites contribute in a cumulative manner to the disruption of the ability of Pd to inhibit G t ␣/G t ␤␥ interactions, with phosphorylation at Ser-54 or Ser-73 appearing somewhat more effective than phosphorylation at Ser-36 or Ser-106. The contribution of each phosphorylation site to the disruption of the ability of Pd to block membrane binding of G t ␤␥ showed a similar cumulative effect, but phosphorylation at Ser-106 appeared to play a particularly important role (Fig.  4B). Single serine-to-alanine substitutions had no effect, except in the case of Ser-106, where a significant increase in inhibitory capacity over phosphorylated wild-type Pd was observed. This indicates that even if other phosphorylation sites are intact, CaMKII phosphorylation is significantly less effective in reducing Pd inhibition of G t ␤␥ association with membranes if Ser-106 is not phosphorylated. Little difference was observed between a triple substitution including Ser-54, Ser-73, and Ser- 106 and a quadruple substitution including Ser-6 as well, indicating that phosphorylation at Ser-6 plays a minor role. As was the case in the other assay, alanine substitution of all five phosphorylation sites was required to approach the percent inhibition achieved with unphosphorylated Pd, suggesting that phosphorylation at Ser-36, Ser-54, and Ser-73 also participates along with that at Ser-106 in Pd inactivation.
Phospho-Pd and 14-3-3 Protein-Phosphoserines in the proper context can be targets for the binding of proteins be-longing to the 14-3-3 family (48). The consensus 14-3-3 binding site is RSXS*XP or RX(Y/F)XS*XP, where S* denotes phosphoserine (49). Phosphoproteins that are bound by 14-3-3 usually contain two such sites, and 14-3-3 binds as a dimer (50,51). Dual binding of the two sites enables association even if one of the sites falls short of the consensus, as is seen in the case of the 120-kDa proto-oncogene product Cbl, where the sequences are NRHS*LP and RLGS*TF (52).
The sequence at the Ser-54 CaMKII phosphorylation site ( 51 RQMS*SP 56 ) fits the 14-3-3 binding consensus, and the region of Ser-73 ( 70 RKMS*IQ 75 ) fits somewhat more loosely. To determine if CaMKII-phosphorylated Pd binds 14-3-3, a GST-Pd fusion protein was prepared, and its binding to 125 Ilabeled recombinant 14-3-3 was assessed in a pull-down assay format using glutathione beads (Fig. 5A). 14-3-3 binding to GST-Pd was similar to that of the GST control in the absence of CaMKII phosphorylation. In contrast, phosphorylation caused a 5-fold increase in binding. To determine which CaMKII phosphorylation sites were required for binding, single serine-toalanine substitutions were prepared for the five sites and were tested for phosphorylation-dependent binding. Binding was abolished when either Ser-54 or Ser-73 was substituted, whereas little or no change was observed when any of the other sites was substituted. Thus, 14-3-3 binding to Pd requires phosphorylation at both Ser-54 and Ser-73. This result is analogous to that obtained with Cbl (52).
Binding of 14-3-3 was shown to further diminish the ability of Pd to block G t ␤␥ association with G t ␣ and Rho* (Fig. 5B). Although CaMKII-phosphorylated Pd was already significantly impaired, association with 14-3-3 resulted in a further reduction in G t ␤␥ binding. Thus, it appears that 14-3-3 competes with G t ␤␥ for the marginal binding to Pd that remains after CaMKII phosphorylation.
The in vitro binding of CaMKII-phosphorylated Pd to 14-3-3 merited further studies to assess the in vivo significance of this association. Immunoblot analysis of whole retina has demonstrated abundant expression of both 14-3-3⑀ and 14-3-3 (34). In addition, immunocytochemical localization experiments have shown large amounts of 14-3-3 in photoreceptor inner segments (35). To determine if phosphorylated Pd from photoreceptor cells would also bind 14-3-3, purified hexahistidinetagged 14-3-3 was immobilized on a nickel chelate resin and used as an affinity matrix, over which extract from darkadapted retinas was passed. Pd from this extract bound to 14-3-3 (Fig. 6A). The phosphatase inhibitor microcystin, which inhibits phosphatases type 1 and 2A, which have been shown to dephosphorylate Pd in vivo (27), was added to the extract to maintain the phosphorylation state of Pd at the time of extraction. When microcystin was omitted in these experiments, no binding of Pd was observed, indicating that binding was due to Pd phosphorylation. Importantly, 5-fold more Pd was bound to 14-3-3 from extracts of dark-adapted retinas compared with light-adapted retinas. When coupled with the fact that phosphorylation at Ser-54 and Ser-73 is required for Pd binding to 14-3-3, these results indicate that dark adaptation increases the phosphorylation state of Ser-54 and Ser-73 of Pd in vivo. The data are also consistent with the observation that 14-3-3 co-immunoprecipitates with Pd from retinal extracts of darkadapted (but not light-adapted) rats (35).
Given the increase in Ca 2ϩ concentration that occurs in rod photoreceptors upon dark adaptation and the Ca 2ϩ requirement for CaM kinase activity, it is possible that changes in photoreceptor Ca 2ϩ levels control Pd phosphorylation in vivo. Thus, we investigated the effect of photoreceptor Ca 2ϩ concentration on the binding of endogenous Pd to 14-3-3. When retinas were dark-adapted in the presence of the cell-permeant FIG. 4. CaMKII phosphorylation sites have a cumulative effect on Pd/G t ␤␥ interactions. A, inhibition of G t binding to light-activated rhodopsin. Assays were performed as described for Fig. 2B. Shown is the percent maximal G t ␣ bound to Rho* in the presence of 2 M wildtype (WT) Pd (last bar), CaMKII-phosphorylated Pd (P-Pd; first bar), or CaMKII-phosphorylated serine-to-alanine variants as indicated. Data shown are the means Ϯ S.D. of triplicates. B, inhibition of G t ␤␥ binding to membrane. Assays were as described for Fig. 2C. Shown is the percent maximal G t ␤␥ bound to urea-stripped ROS membranes in the presence of 7 M Pd (last bar), CaMKII-phosphorylated Pd (first bar), or CaMKII-phosphorylated serine-to-alanine variants as indicated. Data shown are the means Ϯ S.D. of triplicates.
Ca 2ϩ chelator BAPTA/AM, the amount of Pd from retinal extract that bound 14-3-3 decreased 5-fold compared with similarly dark-adapted retinas in the absence of BAPTA/AM (Fig.  6B). Thus, Ca 2ϩ appears to control the in vivo phosphorylation state of Pd at Ser-54 and Ser-73. The conclusion that can be derived from these data is that Pd is phosphorylated in vivo by a CaM-dependent kinase at serines 54 and 73 and that this phosphorylation enables binding of 14-3-3.

DISCUSSION
Physiological Implications of CaMKII-mediated Phosphorylation and 14-3-3 Binding-The results presented here provide new insights into the function of Pd in photoreceptor cells. Our data show that CaMKII phosphorylation is a principal means of regulating the Pd/G t ␤␥ interaction and that such phosphorylation initiates an interaction with 14-3-3. For these interactions to be physiologically significant, the components must co-localize in the photoreceptor cell. Immunolocalization studies have shown that Pd is found in both the outer and inner segments in dark-adapted photoreceptors and only in the inner segment in light-adapted cells (35,40,(53)(54)(55). G t ␤␥ is found in the outer segment in dark-adapted photoreceptors and both the outer and inner segments in light-adapted cells (54). Thus, light causes movement of both Pd and G t ␤␥ to the inner segment. 14-3-3 is located in the inner segment in light-adapted photoreceptors (35), whereas its location in dark-adapted cells has not been reported. Both Pd and 14-3-3 are also found in the synaptic terminal region in light-adapted cells. From these localization data, it appears that Pd, G t ␤␥, and 14-3-3 have ample opportunity to interact, although it will be important to localize 14-3-3 in dark-adapted cells since it is in the darkadapted state that Pd is phosphorylated and would be expected to interact with 14-3-3.
In the case of CaMKII, a retinal CaM-dependent kinase has been previously described (56), and its activity was greatly enhanced during dark adaptation (57). Immunocytochemical studies have generally shown CaMKII and CaMKIV to be expressed predominantly in the inner retina, with only modest FIG. 5. 14-3-3 protein interacts with Pd and affects its association with G t ␤␥. A, 14-3-3 binding to Pd is dependent on intact phosphorylation sites at Ser-54 and Ser-73 of Pd. The binding of 125 Ilabeled recombinant 14-3-3 to a GST control, GST-tagged Pd, phosphorylated Pd, or Pd with single serine-to-alanine substitutions was measured in a GST pull-down assay as indicated. Data from three experiments were combined by normalizing to the GST control, which varied between 0.8 and 1.2 pmol of 125 I-labeled 14-3-3 bound between experiments. Error bars represent the S.D. of the normalized data. B, association of Pd with 14-3-3 protein affects its inhibition of G t binding to light-activated rhodopsin. Assays as described for Fig. 2B  expression in the outer retina, where the photoreceptor cells are located (58,59). In addition, an examination of rat IROS showed a lack of Ca 2ϩ -stimulated phosphorylation of Pd (28), but the effect of CaM was not tested in this study. In contrast to these results, a CaMKIV-like protein termed reticalmin was shown by immunolocalization to be abundantly expressed in photoreceptor outer segments (60). Furthermore, Ca 2ϩ /CaMdependent phosphorylation of Pd in fish rod inner/outer segment preparations has been reported (27). Our observation that Pd is phosphorylated in rod outer segment preparations by a CaMK that is sensitive to the CaMKII-specific peptide inhibitor AIP is consistent with these latter findings. Moreover, our data indicate that both Ser-54 and Ser-73 of Pd are phosphorylated in vivo. CaMKII is the only kinase reported to phosphorylate Ser-54. PKA phosphorylates only Ser-73 (25), and G protein receptor kinase-2 phosphorylates between residues 204 and 245 (61). Consistent with these observations are the facts that the sequence surrounding Ser-54 fits the optimal substrate recognition motif for CaMKII (62) and that this sequence is located in a flexible loop of Pd (14). Taken together, most of the evidence indicates that CaMKII is found in photoreceptor cells and that it phosphorylates phosducin in vivo in a Ca 2ϩdependent manner. Further investigations are needed to verify this conclusion and to understand the specific properties of this kinase.
Previous conclusions that PKA was responsible for Pd phosphorylation in photoreceptor cells were based on comparison of two-and one-dimensional peptide maps of in vivo phosphorylated Pd and PKA-phosphorylated Pd (25). These maps were similar, leading to the conclusion that Pd is phosphorylated by PKA in vivo. However, the maps were not identical, and the resolution was not sufficient to rule out phosphorylation by CaMKII or other kinases. Our phosphorylation data showed comparable rates of phosphorylation of Pd by CaMKII and PKA in IROS preparations. Thus, it appears likely that both PKA and CaMKII phosphorylate Pd in vivo and contribute to the inhibition of G t ␤␥ binding and activation of 14-3-3 binding. PKA phosphorylates Ser-73, whereas CaMKII also phosphorylates Ser-73 as well as Ser-6, Ser-36, Ser-54, and Ser-106. This adds complexity to the regulation of Pd activity, which may require both Ca 2ϩ -dependent and cAMP-dependent signals for maximal phosphorylation.
Light-dependent Ca 2ϩ fluctuations orchestrate light adaptation processes in photoreceptors. The data in Fig. 6B show that Ca 2ϩ concentration controls the phosphorylation state of phosducin and its ability to bind 14-3-3. Previously, an indirect link between Ca 2ϩ and PKA phosphorylation of Pd was proposed through the effect of Ca 2ϩ on cAMP synthesis by a photoreceptor CaM-dependent adenylyl cyclase (22). In addition to those findings, CaMKII provides a more potent, direct link between Ca 2ϩ concentration, Pd phosphorylation, and G t ␤␥ availability that may contribute to light adaptation by decreasing the complement of G t heterotrimers and thereby inhibiting further light signals (22). A second possible role for Pd is in the photoreceptor inner segment and synaptic region, where it is found in abundance (35,40,(53)(54)(55). By cycling between its phosphorylated, 14-3-3-binding form in the dark and its dephosphorylated, G␤␥-binding form in the light, Pd could regulate G protein signaling events that occur in the inner segment and synaptic region. Little is known about these events, but there are many G protein-coupled receptors in these regions, including receptors for dopamine, serotonin, adenosine, and glutamate (63)(64)(65)(66). Interestingly, it has been recently reported that the G␤␥ released upon activation of dopamine receptors inhibits neurotransmission in giant motor synapses of lamprey by binding to components of the SNARE exocytotic machinery (67). If G␤␥ participates in a similar way in controlling glutamate release in photoreceptors, then Pd may regulate this G␤␥ function in a light-dependent manner.
To understand the physiological significance of 14-3-3 binding to phosphorylated Pd, it is useful to examine the proposed role of 14-3-3 in other signaling systems. In the mitogen-activated protein kinase cascade, 14-3-3 associates with and maintains phosphorylated Raf in an inactive state in the absence of Ras⅐GTP. In contrast, 14-3-3 promotes Raf activation and stabilizes its active conformation when Ras⅐GTP is present (68 -70). In the pro-apoptotic pathway of the Bcl-2 family member BAD, phosphorylation of BAD by Akt kinase results in 14-3-3 binding and inhibition of the ability of BAD to heterodimerize with Bcl-2 family members and to neutralize Bcl-2 anti-apoptotic function (71). In the case of the cell cycle regulator Cdc25 phosphatase, kinase activation in response to DNA damage results in phosphorylation of Cdc25 and binding of 14-3-3. When bound to 14-3-3, phosphorylated Cdc25 is sequestered in the cytosol away from its substrate, phosphorylated Cdc2, which is localized to the nucleus. Cdc2 is thus maintained in its phosphorylated inactive form, and the cell cycle remains arrested at the G 2 checkpoint until DNA repair is complete (72). In the regulation of myocyte hypertrophy, CaMK phosphorylation of histone deacetylase results in 14-3-3 binding and export of the complex from the nucleus. As a result, the transcription factor myocyte enhancer factor-2 is relieved of inhibition by histone deacetylase, and its target genes are subsequently expressed (73). These examples indicate that a major function of 14-3-3 proteins is to bind specific phosphoserine-containing proteins and to sequester them in the cytosol away from their interacting partners. In addition to this function, 14-3-3 proteins may also serve as a molecular bridge between phosphoproteins given the fact that 14-3-3 is a dimer in its native state, with each monomer able to interact with a phosphoserinebinding site from a different protein (50,74).
In view of these possible roles, 14-3-3 binding may serve to sequester phosphorylated Pd from G t ␤␥. Pd binding to G t ␤␥ is already severely inhibited by CaMKII phosphorylation, but 14-3-3 binding completely blocks the interaction (Fig. 5B). In addition, 14-3-3 should decrease the rate of Pd dephosphorylation after a light stimulus by virtue of its interaction with phosphorylated Ser-54 and Ser-73, lengthening the time that Pd remains phosphorylated after a light exposure. A second role for 14-3-3 may be to protect phosphorylated Pd from degradation or aggregation. The structure of the N-terminal domain of Pd when bound to G t ␤␥ indicates that it is not a highly stable structure (24). Addition of several phosphates to this domain through CaMKII and PKA phosphorylation, coupled with dissociation from G t ␤␥, would further destabilize the structure of Pd, possibly making it a target for cellular proteases. Binding of 14-3-3 would stabilize the N-terminal domain and protect it from the degradation likely to occur as a result of phosphorylation. Notably, Pd has been shown to interact with the SUG1 subunit of the proteasome complex (75). It would be interesting to determine the effect of phosphorylation and 14-3-3 binding on this interaction. A third, more speculative possibility is that 14-3-3 binding may control the localization of Pd within the photoreceptor. Pd appears to move from the outer segment to the inner segment in response to light (54,55). Interestingly, 14-3-3 has been shown to associate with the kinesin-like motor protein KIF1C (76) and may possibly enable Pd translocation by bridging phosphorylated Pd and a motor protein. The precise physiological role of the association of 14-3-3 with CaMKII-phosphorylated Pd, among these and other possibilities, remains to be determined, but it is likely to be of significance in photoreceptor physiology.

Structural Implications of CaMKII Phosphorylation of Pd-
The observed contributions of the five phosphorylation sites to the inactivation of Pd can be understood by referring to the structure of the Pd⅐G t ␤␥ complex (14). The apparent lack of effect of phosphorylation at Ser-6 is consistent with that fact that this residue is found in a disordered region of the amino terminus not involved in G t ␤␥ interactions. The contribution from the phosphorylation at Ser-73 has been analyzed in detail (24) and involves disruption of the capping of helix 2, leading to a disordering of this region of Pd. Ser-36 is at the C terminus of helix 1, and its side chain points toward the small hydrophobic core of the N-terminal domain. Introduction of phosphate at this position may partially compromise this core. Serine 54 is part of an extended loop between helices 1 and 2, and there are no contacts with G t ␤␥ in the residues surrounding Ser-54. However, in the structure, this residue does point toward the core of the N-terminal domain, and perhaps phosphorylation at this residue would compromise the core, as with Ser-36. It is noteworthy that substitution of Ser-106 affects the inhibition of G t ␤␥ membrane binding more than any other single-site mutant, despite a lesser effect of this mutation in the assay for inhibition of G t ␤␥ binding to G t ␣ and Rho*. Proximal to Ser-106 are Leu-105, His-102, and Met-101 of helix 3, which all interact with Trp-332 of G t ␤ (14). Trp-332 is one of three residues of G t ␤ that undergo a large conformational change upon Pd binding. This conformational change has been proposed to create a pocket for the farnesyl group of G t ␥ (77), and the burying of the farnesyl group in this pocket has been proposed to contribute to the ability of Pd to disrupt membrane binding of G t ␤␥ (13). Phosphorylation of Pd at Ser-106 may block the interaction of Leu-105, His-102, and Met-101 with Trp-332 of G t ␤ and thus inhibit the conformational change. Without a pocket to bury the farnesyl group, Pd binding to G t ␤␥ and G t ␤␥ dissociation from the membrane may be blocked.
None of the phosphorylation site residues are themselves G t ␤␥ contact residues, and phosphorylation at any of them would appear to minimally disrupt the structure of Pd and the binding interface with G t ␤␥. This observation is consistent with our finding that the phosphorylation-induced decrease in G t ␤␥ binding is cumulative, dependent on the phosphorylation of multiple residues. The cumulative effect of phosphorylation at these sites explains the differences between the inhibition of G t ␤␥ binding induced by CaMKII and that induced by PKA. PKA phosphorylates only Ser-73, one of the five sites phosphorylated by CaMKII. Thus, the disruption in contacts between Pd and G t ␤␥ is much greater in the case of CaMKII phosphorylation.