Phosphorylation of Rhodopsin by Protein Kinase C in Vitro*

Calium/phospholipid-dependent protein kinase (pro- tein kinase C) was purified from bovine retinae rod outer segments (ROS). In the presence of 0.1-2 PM calcium protein kinase C binds tightly to ROS and phosphorylates rhodopsin in the absence or presence of illumination. This property of protein kinase C con-trasts with that of rhodopsin kinase, which in vitro phosphorylates only bleached rhodopsin. Peptide maps of rhodopsin phosphorylated by protein kinase C or rhodopsin kinase were compared using limited Staphylococcus aureus V8 protease digestion or complete tryptic digestion. Phosphorylation sites map to serine and threonine residues on the cytoplasmic carboxyl- terminal domain of rhodopsin for both kinases. The functional consequence of protein kinase C phosphorylation of rhodopsin was a reduced ability to stimulate the light-dependent rhodopsin activation of [36S]guan-osine 5’-O-(thiotriphosphate) binding to transducin, the GTP-binding regulatory protein present in ROS. Properties of the calcium-stimulated interaction of protein kinase C with membranes and in vitro phosphorylation of intrinsic proteins are discussed based upon the findings. in the absence of rhodopsin was less than 2% of that observed in the presence of rhodopsin. Rhodop- sin-stimulated [35S]GTPyS binding was dependent upon light and temperature. No binding was detectable if purified transducin was absent from the reaction mixture.

Photoreceptor excitation involves absorption of a photon by rhodopsin which triggers a set of events resulting in a membrane voltage change. The sensitivity of the photoreceptor becomes reduced after illumination, a phenomenon referred to as adaptation or attenuation (1,2). Phosphorylation of rhodopsin is one mechanism that has been proposed for the regulation of rhodopsin sensitivity (3-7). Light-dependent alterations in intracellular calcium concentration or calcium translocation from intradiscal sites to the cytoplasmic sirrface of photoreceptor membranes, as well as alterations of phosphatidylinositol metabolism, have suggested involvement of calcium in visual adaptation (2,(8)(9)(10)(11)(12)(13)(14)(15)(16)(17). At the molecular level, no link between possible calcium-dependent and phosphorylation-dependent regulation of rhodopsin function has been identified. We have now succeeded in demonstrating calciumdependent phosphorylation of rhodopsin in vitro and have resolved this activity from the previously described rhodopsin kinase that phosphorylates only bleached rhodopsin in a calcium-independent manner (3-7). The calcium-dependent rhodopsin-phosphorylating activity was protein kinase C, which we have purified from ROS' (18). Protein kinase C is * This work was supported by National Institutes of Health Grants GM30324 and NS18779. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

MATERIALS AND METHODS
Preparation of ROS Membranes-ROS were isolated from frozen dark-adapted bovine retinae. After thawing in 20 mM Tris-HC1, pH 7.4, 1 mM CaC12, 5 pg/ml leupeptin, 0.3 unit/ml aprotinin, 8.7 pg/ml phenylmethylsulfonyl-fluoride (buffer A), and 45% (w/w) sucrose the rod segments were disrupted by several passes through a 50-ml syringe. ROS were isolated by flotation in a SW 28 rotor centrifuged at 25,000 X g for 20 min. The ROS were collected and diluted in buffer A before being layered over a step gradient of 25 and 35% (w/ w) sucrose in buffer A and centrifuged at 100,000 X g for 20 min. ROS at the 25/35% sucrose interface were diluted in buffer A and pelleted by centrifugation at 25,000 X g for 20 min. The pellets were resuspended in 2 ml of 200 mM Na-Hepes, pH 8.0, 20 mM EDTA/BO retinae to extract extrinsic membrane proteins, pelleted, and repeated 2 times. The extracts were pooled and dialyzed against 20 mM Tris-HC1, pH 7.4,l mM EGTA, 1 mM EDTA, 1 mM DTT.
The pellets containing rhodopsin were recovered and represented 11.3 mg of rhodopsin and 50 mg of total protein. Stripped ROS membranes were prepared from this preparation by the method of Yamazaki et al. (25). The final preparation of stripped ROS yielded 8.8 mg of rhodopsin and 25 mg of total protein from 50 retinae. The stripped ROS were stored at a concentration of 50 PM rhodopsin in 20 mM Tris-HC1, pH 7.4,l mM EDTA, 2 mM MgC12,0.5 mM DTT at -80 "C. Purification and Assay of Protein Kinase C-Protein kinase C was purified by sequential chromatography on phenyl-Sepharose and DEAE-cellulose (18). Protein kinase C was assayed essentially as described by Vilgrain et al. (26).
Phosphorylation Assays-Specific conditions for phosphorylation assays are given in the figure legends. Incorporation of 32Pi from [y-32P]ATP into rhodopsin was quantitated by Cerenkov counting of excised rhodopsin bands from SDS-acrylamide gels or by densitometry of autoradiographs. Confirmation that rhodopsin was the phosphorylated protein was obtained by rhodopsin purification and by limited V8 protease digestion (27) and the characteristic rhodopsin phosphopeptide profile on 15% acrylamide-SDS gels. V8 Protease Digestion of Phosphorylated Rhodopsin-Striped ROS were phosphorylated with either rhodopsin kinase or protein kinase C. The membranes were then washed 2 times in ice-cold 20 mM Na-Hepes, pH 8.0, 2 mM EDTA, and then resuspended in 20 mM Tris-HC1, pH 7.4,2 mM MgC12,l mM EDTA, 1 mM DTT at a concentration of 42 Fg/ml. The membranes were then incubated with 2 pg/ml V8 protease for 120 min at 30 "C. Controls were incubated similarly in the absence of protease. The samples were then dissolved in 2% SDS sample buffer and electrophoresed on 15% acrylamide-SDS gels. The gels were then stained with Coomassie Blue and autoradiographed.
Tryptic Digestion of Phosphorylated Rhodopsin and HPLC Analysis 4749 4750

Rhodopsin Phosphorylation
of Phosphopeptides-Stripped ROS were phosphorylated with either rhodopsin kinase or protein kinase C. The membranes were dissolved in SDS and reduced with DTT overnight at 25 "C followed by alkylation with iodoacetamide (28). Samples were then electrophoresed on 10% acrylamide SDS gels and autoradiographed for 60 min to locate the rhodopsin band. The band was excised, washed for 5 h with distilled Hz0 to remove SDS, and then digested with three 50 pg/ml tosylphenylalanyl chloromethyl ketone-trypsin aliquots in 3 ml of 50 mM NHIHC03, pH 8.4, at 37 "C for 24 h. Eighty-five per cent of the label in the rhodopsin band was released from the gel, and the pooled fractions were lyophilized and then resuspended in 1% trifluoroacetic acid. Samples were chromatographed on a C18 reverse phase HPLC column (Waters pBondapak, 3.9 mm X 30 cm) equilibrated in 0.1% trifluoroacetic acid and 0.05% triethylamine. Half-ml fractions were collected during a 0-45% acetonitrile gradient followed by a 100% methanol wash of the column. Fractions were analyzed by Cerenkov counting.
Phosphoamino acid analysis of the 32Pi phosphopeptides was performed by partial acid hydrolysis for 1 h at 110 "C in 6 N HC1 followed by thin layer electrophoresis as described by Hunter and Sefton (29).
[35S]GTPyS (0.01-2.0 p~) binding to transducin as described in the legend to Fig. 6 was saturable and to a single high affinity site of K D = 0.05 p~. Binding of [35S]GTPyS in the absence of rhodopsin was less than 2% of that observed in the presence of rhodopsin. Rhodopsin-stimulated [35S]GTPyS binding was dependent upon light and temperature. No binding was detectable if purified transducin was absent from the reaction mixture.

RESULTS
Isolated ROS contain the necessary components for lightactivated phosphodiesterase stimulation as well as the phosphorylating activity that modifies bleached rhodopsin. All of the known components, except for rhodopsin, in these reactions are extrinsic membrane proteins which can be quantitatively stripped from the ROS membranes. Rhodopsin is an intrinsic membrane protein whose structure has been recently defined (31,33). The rhodopsin-regulated phosphodiesterase activation, as well as the rhodopsin-phosphorylating activities, can be reconstituted simply by mixing the stripped ROS membranes containing rhodopsin with the extrinsic membrane proteins. Fig. lA shows the results of such an experiment demonstrating that the extrinsic membrane protein fraction (referred to as extract) reconstitutes the light-dependent phosphorylation of rhodopsin. If ROS membranes are prepared in the presence of calcium and the extrinsic membrane proteins exhaustively stripped with chelator, dialyzed, and subsequently reconstituted with the stripped ROS in the presence of 200 pM CaC12, phosphorylation is observed in the absence of light and further enhanced by rhodopsin bleaching (Fig. 1B). Rhodopsin kinase is obviously present in the extract, but no rhodopsin phosphorylation is observed in the dark in the absence of added calcium (Fig. lA, lane 3). Rhodopsin kinase readily phosphorylates bleached rhodopsin and does not require calcium for its activation (3-7). These facts indicate that the calcium-dependent rhodopsin phosphorylation observed in Fig. 1 was not due to the presence of bleached rhodopsin in the dark-adapted ROS, because rhodopsin kinase would readily phosphorylate the bleached photopigment. The simplest interpretation of the calcium-dependent rhodopsin phosphorylation was the presence in the extract of two rhodopsin-phosphorylating activities. An alternative explanation for the finding was a calcium-dependent inhibition of either a phosphatase or even an ATPase. It seemed unlikely that phosphatase activity was being inhibited by calcium because the light-dependent rhodopsin phosphorylation in the absence of calcium was stable for long incubation periods at room temperature (not shown), indicating the rhodopsin phosphorylation was not readily reversed with these conditions. Similarly, increasing the ATP levels up to &fold did not enhance rhodopsin phosphorylation in darkadapted ROS suggesting ATP hydrolysis was not a problem in the protocol described in Fig. 1. For these reasons attention was turned toward identifying specific kinases in the extract in addition to the previously characterized rhodopsin kinase. Table I shows the soluble extract from ROS membranes contained significant calciumand phospholipid-dependent histone H1-phosphorylating activity. This activity was dependent on the simultaneous presence of calcium and a mixture of phosphatidylserine and diacylglycerol. The concentration of calcium required in order to obtain maximal histone phosphorylating activity was greatly reduced in the presence of the phorbol ester PMA (18). As described elsewhere (18) we have purified this histone kinase activity from the ROS extract and demonstrated it correlates with the presence of an 85-kDa protein on SDSacrylamide gels and has the properties of protein kinase C.
Preparation of ROS membranes in the absence of calcium resulted in the loss of protein kinase C activity that could be stripped by EDTA during the removal of extrinsic membrane proteins and resulted in the loss of the calcium-dependent light-independent rhodopsin phosphorylation observed in Fig.  1. These observations suggested that the protein kinase C might be responsible for the calcium-dependent phosphorylation of rhodopsin.
It should be noted that protein kinase C-dependent phosphorylation of intrinsic membrane proteins such as the recep-

TABLE I
Calcium-and phospholipid-dependent histone HI kinase activity extracted from ROS membranes Extrinsic proteins associated with the isolated ROS membranes were solubilized using 20 mM EDTA and after dialysis assayed for histone H1 kinase activity. Varying amounts of CaC12 were added to the incubation mixture and free calcium concentrations added were estimated using a computer-assisted program (54). Indicated samples also received 26 p g of phosphatidylserine and 0. tors for EGF (34) and transferrin' requires calcium but not the addition of phosphatidylserine or diacylglycerol. Presumably, the phospholipids are provided by the membrane with which the protein substrate is associated. Fig. 2 demonstrates the resolution of the calcium-dependent protein kinase C rhodopsin-phosphorylating activity from the light-dependent rhodopsin kinase activity. A calcium-and phospholipid-dependent phenyl-Sepharose chromatography procedure (18, 35) was utilized to give a highly enriched protein kinase C preparation. Protein kinase C binding to phenyl-Sepharose was nearly quantitative and dependent on the addition of phosphatidylserine and calcium (18). Fig. 2 shows that the phospholipid-dependent histone-phosphorylating activity was eluted from the column with 1 mM EGTA. This activity was also calcium dependent and stimulated by PMA, which was consistent with the properties of protein kinase C (18). The lower panel of Fig. 2 demonstrates that these 'EGTA-eluted fractions also reconstituted calcium-dependent rhodopsin phosphorylation. No light-dependent rhodopsin phosphorylation could be detected using the EGTAeluted fractions. The phosphorylation of rhodopsin occurred using either unbleached or bleached rhodopsin, and the ability of the EGTA-eluted fractions to phosphorylate rhodopsin correlated with the calcium-and phospholipid-dependent histone phosphorylation. Fig. 2 also shows that the breakthrough fractions contained light-dependent rhodopsin-phosphorylating activity. No calcium-dependent rhodopsin phosphorylation was detected in the breakthrough, and very little phospholipid-and calciumdependent histone kinase activity could be detected. Results from many experiments of this kind have suggested rhodopsin kinase is very unstable during purification. Nonetheless, Fig.  2 clearly demonstrates the resolution of the calcium-dependent rhodopsin phosphorylation from the light-dependent rhodopsin kinase activity.
Additional evidence that protein kinase C phosphorylates rhodopsin is shown in Fig. 3. PMA at limiting calcium concentration stimulates rhodopsin phosphorylation. The phosphorylation of rhodopsin in the presence of submaximal concentrations of calcium or PMA alone was significantly less than in the presence of both. Since PMA is well known to bind and activate protein kinase C by increasing its affinity for calcium (36), this finding provides additional evidence that protein kinase C and not an effect of calcium on a phosphatase or ATPase was responsible for the enhanced rhodopsin phosphorylation. Fig. 4A shows that calcium induced a translocation of protein kinase C from a soluble to a membrane-associated state in the absence of rhodopsin bleaching. When protein kinase C was mixed with ROS membranes stripped of extrinsic proteins and all detectable kinase activity there was a calcium-induced increase in membrane-associated rhodopsin phosphorylation and a concomitant decrease in soluble histone kinase activity. The rhodopsin phosphorylation shown occurred in the absence of illumination, but essentially identical results were obtained using bleached rhodopsin. The binding of protein kinase C to the ROS membranes occurred when the free calcium added was raised above 0.1 PM and was complete at 2 PM. The calcium concentration required for rhodopsin phosphorylation (Fig. 4B) was similar to that for inducing translocation, consistent with the notion that interaction of protein kinase C with the membrane activated the enzyme in a calcium-dependent mechanism. The strong correlation of the supernatant loss in phosphatidylserine-and diacylglycerol-dependent histone phosphorylation with the increased rhodopsin phosphorylation of the particulate fraction also indicates that protein kinase C was responsible for both activities. Presumably, the protein kinase C binding involved an anionic phospholipid such as phosphatidylserine and possibly diacylglycerol in the ROS membranes. Fig. 5 demonstrates two independent peptide-mapping strategies to characterize the rhodopsin phosphorylation sites for protein kinase C relative to the well characterized sites phosphorylated by rhodopsin kinase (3-5, 27, 37, 38). Short phosphorylation periods were used with both kinases, and the concentration of each kinase was adjusted so that similar amounts of 32Pi were incorporated into rhodopsin. These conditions were chosen so that the results might provide information regarding the preferred serine and threonine residues on the rhodopsin molecule. Light-dependent phosphorylation was performed in the presence of added EGTA to ensure that calcium-dependent phosphorylation was completely absent. The calcium-dependent protein kinase C phosphorylation of rhodopsin was performed in the dark. The autoradiographs in Fig. 5A show that the light-dependent kinase activity was specific for rhodopsin. Calcium-dependent protein kinase C phosphorylation of ROS membrane proteins was somewhat less specific. The preferred substrate, however, was rhodopsin which is labeled much more intensely than any other band. Previous experiments indicated the labeled band at about 68 kDa was actually a rhodopsin dimer (not shown). After phosphorylation the membranes were washed and treated with Staphylococcus aureus V8 protease. Labeled peptides were then analyzed on acrylamide-SDS gels. V8 protease has been demonstrated to cleave membrane-bound rhodopsin specifically at positions GluZ3@ and Gld41 (27,37). Residue GluZ3' resides in one of the three predicted cytoplasmic loops of rhodopsin (27, 31, 32), and G~u~~~ is 7 residues from the carboxyl terminus (27, 31). As judged by Coomassie Blue staining the V8 cleavage of rhodopsin was similar after phosphorylation by the two kinases (not shown). Labeled phosphate could not be detected in the largest V8-generated peptide (met1-GluZ3') after rhodopsin phosphorylation by either kinase, even after autoradiography for up to 2 weeks. For light-dependent phosphorylation by rhodopsin kinase this finding is in agreement with previous work (27, 37, 38) that FIG. 2. Purification of protein kinase C by calcium-and phosphatidylserine-dependent hydrophobic chromatography. One ml of dialyzed extract prepared from ROS membranes was brought to 0.035 mM phosphatidylserine, 3 mM MgC12, and 3 mM CaCI2 and then applied to a 1-ml phenyl-Sepharose C1-4B column equilibrated in 20 mM Tris-HCI, pH 7.4, 0.1 mM CaC12, 1 mM DTT. The column was then washed with 10 ml of equilibration buffer (arrow A ) . Protein kinase C activity was eluted using 20 mM Tris-HC1, pH 7.4, 1 mM EGTA, 1 mM DTT (arrow B ) . One-ml fractions were collected, and 10-pl aliquots were assayed for histone H1 kinase activity with an excess of calcium over the chelator and in the presence or absence of phosphatidylserine and diacylglycerol (PSIDAG). Thirty-pl aliquots of each fraction were then assayed for their ability to phosphorylate rhodopsin. Stripped ROS membranes devoid of measurable kinase activity were used as substrate. Reactions were performed for 10 min at 30 "C either with constant illumination (+hu) or in the dark (-hu). CaCI2. The samples represented by lanes 2 and 3 also contained 0.2 p~ PMA. The reaction was stopped with SDS and the samples electrophoresed as described for Fig. 1. indicated rhodopsin phosphorylation mapped to the carboxylterminal region of the molecule. Fig. 5A shows that phosphorylation by either kinase resulted in the appearance of two phosphorylated peptides near 14 kDa. The doublet corresponds to Ser240-Ala348 for the upper band, and the lower band has the 7-amino acid peptide removed so that it corresponds to residues Ser240-G1~341 (27,37). These results demonstrate that protein kinase C in the presence of calcium and the absence of light phosphorylates rhodopsin in the carboxylterminal portion of the molecule, similar to the light-dependent phosphorylation by rhodopsin kinase. The 7-amino acid peptide was not detectable by this technique since it runs off the bottom of the gels. However, the apparent lower recovery of label in the 14-kDa doublet generated after light-dependent rhodopsin phosphorylation probably reflects significant labeling of the 7-amino acid peptide at Ser343 (27,37). Longer exposure of the autoradiographs indicated there was label in the lower band of the doublet, although less than that in the upper band for the light-dependent rhodopsin phosphorylation. This is in agreement with previous reports that rhodopsin kinase phosphorylates sites on rhodopsin within the Ser334-Thr340 domain as well as Ser343 (37,38). Both bands of the doublet are equally labeled after digestion of rhodopsin phosphorylated by protein kinase C. This finding suggested that the preferred site on the carboxyl-terminal tail of rhodopsin for protein kinase C was not on the Thr342-Ala348 7amino acid peptide but within the 14-kDa peptide (SerZ4O-Glu3") containing the serine-and threonine-rich domain.
To further analyze the phosphorylation sites on rhodopsin the 39-kDa rhodopsin band from gels corresponding with lanes 1 and 2 in Fig. 5A was excised and exhaustively digested with trypsin and analyzed by reverse phase HPLC (Fig. 5B). The labeled peptides derived from rhodopsin phosphorylated by either rhodopsin kinase or protein kinase C eluted with similar mobilities from the reverse phase column. Similar mobilities for the three major peaks observed for each kinase , 100 pl of protein kinase C eluate, and varying amounts of CaC12. The samples were then centrifuged for 2 min and the pellets and supernatants separated in the dark. Phospholipid-dependent histone kinase activity (M) was measured in the supernatants and expressed as the per cent of control. Control activity (1.6 X lo4 dpm/50 pllmin) was determined using supernatant from samples originally incubated without added CaC12. The pellets were resuspended to 150 pl in 40 mM Tris-HC1, pH 7.4,lO mM MgC12, 733 p~ CaC12, 330 ~L M EDTA, 0.5 mM DTT, and 10 p~ [y3'P]ATP (7000 dpm/pmol). After incubation for 5 min at 30 "C in the dark the reaction was stopped with SDS and the samples electrophoresed on 10% acrylamide-SDS gels. Note that all phosphorylation reactions were in the presence of a saturating concentration of added calcium. The 32P incorporation into rhodopsin was determined by Cerenkov counting of the excised 39-kDa rhodopsin band and is expressed as per cent of control (A---A). The control activity (9.5 X lo3 dpm incorporated into rhodopsin in 5 min/ROS pellet) was determined using pellets from samples originally incubated with a saturating concentration of free calcium (112 pM). B, calcium concentration dependence of rhodopsin phosphorylation. Stripped ROS and the EGTA-eluted protein kinase C fraction from the phenyl-Sepharose column were incubated as described above, and after 30 min at 4 "C in the dark 10 pl [y3'P]ATP (2000 dpm/pmol) was added and the samples were incubated for 5 min at 30 "C. The reaction was stopped with SDS and the samples electrophoresed on 10% acrylamide-SDS gels and autoradiographed. reaction were confirmed by mixing the two 32P-labeled rhodopsin digests and demonstrating co-elution of the peptides in peaks I, 11, and I11 (not shown). Peak I was predominantly labeled in the rhodopsin kinase reaction. In contrast, peak I1

F R A C T I O N N U M B E R
FIG. 5. Peptide mapping of phosphorylated rhodopsin. A, rhodopsin was phosphorylated in a light-dependent manner using rhodopsin kinase with EGTA in excess of CaC12. Alternatively protein kinase C was used to phosphorylate rhodopsin in darkness and in the presence of CaC12. The kinase preparations were those resolved by phenyl-Sepharose chromatography as shown in Fig. 2. Phosphate incorporation/mol of rhodopsin was 0.046 mol for rhodopsin kinase and 0.051 mol for protein kinase C. After digestion of rhodopsin with V8 protease as described under "Materials and Methods" the samples were analyzed on 15% acrylamide gels. The gels were stained with Coomassie Blue to visualize the characteristic V8 digestion pattern of membrane-associated rhodopsin (27) and then autoradiographed to visualize the 32P phosphopeptides. B, rhodopsin which had been phosphorylated as described above was reduced, alkylated, and electrophoresed on a 10% acrylamide-SDS gel. The rhodopsin band was visualized by autoradiography, excised from the gel, washed in H20 followed by 50 mM ammonium bicarbonate, and digested for 24 h with trypsin. Phosphopeptides were analyzed by reverse phase HPLC as described under "Materials and Methods." Three 32P-phosphopeptide peaks were identified at 18, 22, and 26% acetonitrile. Recovery of loaded radioactivity from the C18 column was approximately 50%. Phosphoamino acid analysis was performed on the peak fractions. was predominant in the protein kinase C reaction, but significant phosphorylation was also observed in peaks I and 111.
Phosphoamino acid analysis of each peak indicated P-Ser in peak I, primarily P-Thr with small amounts of P-Ser in peak 11, and a mixture of P-Ser and P-Thr in peak I11 for the phosphopeptides derived from both the light-dependent and calcium-dependent phosphorylation protocols.
The combined results in Fig. 5 of exhaustive tryptic digestion of rhodopsin and V8 protease digestion of membraneassociated rhodopsin indicate the phosphorylation domains are similar for the two kinase reactions. The preferred phosphorylation sites on rhodopsin, however, appear to be different in the two kinase reactions. Our findings are consistent with the extensive work of others (37-40) that the phosphorylation sites for rhodopsin kinase map to the carboxylterminal region of rhodopsin.
An important observation during the course of these experiments was that protein kinase C phosphorylates rhodopsin in the absence of light (Fig. 5 ) . This suggests that the car-

Rhodopsin Phosphorylation
boxyl-terminal phosphorylation sites on rhodopsin are exposed on the cytoplasmic surface of the rhodopsin molecule in the absence of bleaching. The conformational change occurring in the rhodopsin molecule upon photon absorption stimulates both the binding and activation of rhodopsin kinase. This mechanism might serve to target rhodopsin kinase to bleached rhodopsin, rather than to expose the phosphorylation sites. A similar targeting mechanism involving conformational changes in rhodopsin is not used for protein kinase C since its activation is independent of light. Two mechanisms described to occur in vivo could function to activate protein kinase C and amplify the covalent modification of rhodopsin beyond that observed with the bleached molecules and rhodopsin kinase. First, it has been proposed that light induces a translocation of calcium from intradiscal sites to the surface of discs (8, 9). Second, light has been demonstrated to stimulate the breakdown of phosphatidylinositols which will result in the generation of diacylglycerol (12). Both mechanisms could increase the binding of protein kinase C to membranes and stimulate its activity. Because these mechanisms are activated by light in vivo they could provide the necessary targeting specificity for protein kinase C binding to membrane domains near the sites of rhodopsin bleaching for aniplification of the adaptation signal.
In support of this hypothesis, previous work by other laboratories has demonstrated that ROS preparations containing rhodopsin kinase activity phosphorylate rhodopsin upon bleaching and that the phosphorylated rhodopsin has a diminished ability to activate the retinal GTP-binding protein transducin (39,40). The mechanisms and enzymes mediating these responses i n vivo are not clearly defined; however, calcium and changes in phosphatidylinositol metabolism appear important in the i n vivo adaptive responses. The peptide mapping shown in Fig. 5 indicated that both rhodopsin kinase and protein kinase C will phosphorylate sites within the carboxyl-terminal domain of rhodopsin. This result predicts that similar functional changes might be seen in rhodopsin's ability to activate transducin. To test this prediction, control and protein kinase C-phosphorylated rhodopsin was used in measurements of light-dependent rhodopsin stimulation of [35S]GTPyS binding to the 01 subunit of purified transducin. Conditions were chosen so that rhodopsin was limiting in the activation of [35S]GTPyS binding to transducin (Fig. 6). The absence of calcium prevented the binding and activation of protein kinase C phosphorylation of rhodopsin and was used as a control to show that the purified protein kinase C preparation did not contain an inhibitor of transducin activation. Similarly, calcium in the absence of protein kinase C was without effect. When protein kinase C was incubated with the stripped ROS in the presence of calcium and Mg-ATP, there was a 40-45% decrease in the ability of rhodopsin to activate [35S]GTPyS binding to transducin. However, part of this inhibition was observed with protein kinase C and calcium alone, in the absence of added Mg-ATP. Using the conditions described in Fig. 6 about 0.15 mol of 32Pi was incorporated per mol of rhodopsin, and this value correlates well with the decrease in transducin activation attributable to phosphorylation which was approximately 20 f 5% (average f S.D., N = 4 experiments). The reduction in rhodopsin activation of transducin in the absence of ATP appeared related to protein kinase C binding to the membranes. If the protein kinase C bound to sites near rhodopsin, it could sterically inhibit transducin binding. This is the simplest explanation for this observation (see below), and a similar observation has been made for rhodopsin kinase-inhibiting transducin activation in the absence of phosphorylation.
Nonetheless, there is a positive correlation with rhodopsin phosphorylation and a decreased ability of light absorption to activate [35S]GTPyS binding to transducin. High concentrations of protein kinase C (0.8-1.2 units/ml, where 1 unit is defined as 1 nmol of phosphate transferred to histone Hl/min) were required in the experiments shown in Fig. 6 when 25-50 nM rhodopsin was included in the reaction mixture. The high protein kinase C to rhodopsin ratios were required to observe sufficient phosphorylation to measure significant decreases in transducin activation, similar to the problems reported for rhodopsin kinase using similar protocols. An initial unexpected observation was that phosphorylation incubations for longer times did not increase the incorporation of phosphate into rhodopsin. As shown in the inset of Fig. 7, rhodopsin phosphorylation was rapid and plateaued after approximately 5 min. Addition of EGTA to release the membrane-bound protein kinase C indicated it was still active as measured by its ability to phosphorylate soluble histone H1 in the presence of phosphatidylserine, diacylglycerol, and additional calcium. In fact, the histone phosphorylation was linear for greater than 20 min (not shown). This finding indicated that protein kinase C was still functional, and the "turn-off" of rhodopsin phosphorylation was not due to kinase denaturation. Furthermore, the phosphorylation was stable and no evidence for a contaminating phosphatase was observed. Fig. 7 also characterizes the stoichiometry of the protein kinase C phosphorylation of rhodopsin intrinsically associated with disc membranes. For all concentrations of protein kinase C and membrane-associated rhodopsin used, the phosphorylation reaction reached a plateau within 5-8 min, and further addition of ATP was without effect. With increasing protein kinase C, the stoichiometry of rhodopsin phosphorylation increased with higher stoichiometries obtained as the ratio of protein kinase C/rhodopsin increased. Similar characteristics of rhodopsin phosphorylation by rhodopsin kinase have been recently described by Sitaramayya and Liebman (41). They used "rhodopsin kinase-enriched membranes'' and found that with 22% bleaching the stoichiometry of phosphorylation was 0.07 mol of phosphate/mol of bleached rhodopsin. As the percentage of bleached rhodopsin was decreased the stoichiometry of phosphorylation of bleached rhodopsin increased. The functional consequence of decreasing the percentage of bleaching was to effectively raise the ratio of rhodopsin kinase to bleached rhodopsin, similar to the change in ratio of protein kinase C to total rhodopsin observed in Fig. 7. The properties of rhodopsin phosphorylation by the two kinases, therefore, appears both qualitatively and quantitatively similar. The majority of studies characterizing protein kinase C have utilized mixtures of phosphatidylserine, diacylglycerol, and calcium to phosphorylate substrates that have generally been soluble, such as histone H1. Under these conditions, a unit of protein kinase C transfers 1 nmol of phosphate/min.
However, the consensus from several studies suggests that protein kinase C targets are usually membrane-associated proteins and that the generation of diacylglycerol in the membrane activates protein kinase C associated with the membrane. This is obviously different from the phosphorylation of a soluble protein in the presence of high concentrations of phospholipids. Table I1 indicates that 1 unit of protein kinase C measured using the standard histone H1 assay will transfer 1-4 pmol of phosphate/min to rhodopsin on average during a 5-min phosphorylation assay. To date, we are aware of only two other intrinsic membrane proteins that have been shown to be substrates for protein kinase C with functional Nonspecific binding to the filters was determined by filtering samples at zero time in the absence of illumination and subtracted from the total binding. In order to confirm phosphorylation of rhodopsin during the incubations [y-32P]ATP was included in parallel samples and analyzed on SDS-acrylamide gels followed by autoradiography (not shown). Rhodopsin phosphorylation was similar to that observed in Fig. 4. None of the transducin suburiits appeared to be appreciably phosphorylated under these conditions. The ATP-dependent protein kinase C-mediated decline in the initial rate of rhodopsin-stimulated binding of [35S]GTPrS binding to transducin appeared to be directly related to the degree of rhodopsin phosphorylation. B, rhodopsin concentration dependence of transducin activation. Light-dependent rhodopsin activation of [%3]GTPrS binding to transducin was measured as described for A except that protein kinase C was omitted from the incubations. Varying amounts of stripped ROS were added to the incubations to vary the amount of functional rhodopsin to demonstrate the linear relationship between rhodopsin concentration and transducin activation. The values shown to the right of each time course represent the pmol of rhodopsin present per 105-pl assay mixture.

Rhodopsin Phosphorylation
consequences resulting from the phosphorylation. These are the EGF receptor (34) and transferrin receptor.' Table I1 also summarizes the protein kinase C phosphorylation properties of these receptor proteins. For both receptors, protein kinase C was added to membranes and calcium, and neither phosphatidylserine nor diacylglycerol was required for the binding of protein kinase C to the membranes and receptor phosphorylation. The work of Cochet et al. (34) used very high protein kinase C levels relative to EGF receptor and transferred approximately 0.3-0.6 pmol of phosphate/min/unit of protein kinase C to the EGF receptor on average during the initial phase of the phosphorylation reaction. Values for the transferrin receptor were intermediate from those for the EGF receptor and rhodopsin, but in all cases the reactions are rapid, reaching a plateau in minutes, and demonstrate a much lower apparent efficiency of phosphate transfer/unit of enzyme compared to soluble substrates (Table 11).

DISCUSSION
In the presence of calcium protein kinase C binds tightly to ROS membranes and phosphorylates rhodopsin. Protein kinase C is relatively abundant in ROS preparations, and our calculations suggest there is about 1 mol of protein kinase C/ 2000 mol of rhodopsin (18). Protein kinase C is, therefore, expressed at about the same level as rhodopsin kinase (42) in the ROS. The phosphorylation sites on rhodopsin are near the carboxyl terminus for both kinases. Thus, the functional consequences of rhodopsin phosphorylation by either kinase are predicted to be similar. This is, in fact, what is observed in that the catalytic activation of transducin is diminished when rhodopsin is phosphorylated. These findings make protein kinase C a strong candidate for an amplifying mechanism to regulate rhodopsin activity, since it is now thought that adaptation in the photoreceptor involves calcium (2, 8-11) and changes in phosphatidylinositol metabolism (12-17).
Recently, a 48-kDa protein, identified as the 48K protein which probably corresponds to the retinal S antigen, has been shown to bind to ROS in a light-dependent manner similar to rhodopsin kinase (43)(44)(45). The 48K protein appeared to have enhanced binding to bleached phosphorylated rhodopsin. It is possible that the 48K protein could also bind to unbleached phosphorylated rhodopsin and further uncouple its ability to activate transducin. Alternatively, the 48K protein may require the bleached conformatin of rhodopsin and could be involved in the regeneration of the photopigment. The ability to isolate phosphorylated unbleached rhodopsin now allows this question to be addressed specifically.
The finding that protein kinase C phosphorylated residues Tris-HC1, pH 7.4. Reactions were terminated by the addition of SDS, and samples were electrophoresed on 10% acrylamide-SDS gels. The rhodopsin band was located by short autoradiograph exposures and the band excised and counted. The results are expressed as mol of phosphate incorporated per mol of rhodopsin. The inset shows the time course of phosphorylation with samples containing 18.8 pmol of rhodopsin and 0.16 unit/ml protein kinase C. Aliquots at appropriate times were removed and analyzed on SDS-acrylamide gels followed by autoradiography. Rhodopsin bands were quantitated by densitometry and expressed in arbitrary units. The time course is indicative of every concentration of protein kinase C used although the extent of rhodopsin phosphorylation was directly related to protein kinase C concentration. The time course was not altered when reaction mixtures were preincubated for 30 min at 4 "C prior to addition of [y-32P]ATP. Protein kinase C activity, as measured by histone H1 phosphorylation, was stable during the course of these reactions if the enzyme was stripped from the ROS with EDTA (not shown), indicating that enzyme denturation was not occurring to a significant extent. near the carboxyl terminus of rhodopsin and probably within the serine-and threonine-rich phosphorylation domain on the amino-terminal side of Lys339 is consistent with the reported site specificity of protein kinase C which appears to commonly recognize such residues (46). The EGF receptor phosphorylation site at Thr654 for protein kinase C is within a very basic sequence on the cytoplasmic domain near the membrane-spanning region for the protein (47). Few other membrane proteins have been determined to be substrates for protein kinase C so no consensus sequences are apparent.
Like other protein kinases, however, secondary and tertiary structural determinants are probably very important for protein kinase C recognition (48). In this regard, the protein kinase C phosphorylation site is obviously close to the membrane. Interestingly, of the three predicted cytolasmic loops and carboxyl-terminal tail for rhodopsin, the greatest conservation is within the first loop structure (31-33,49,50). Six of the 12 amino acids are conserved between Drosophila and bovine rhodopsins, and two of the six conserved residues are basic. No acidic residues are present on this loop in any of the rhodopsins whose sequence has been determined. If basic residues are indeed important for protein kinase C recognition it is easily seen how the carboxyl-terminal phosphorylation sequences could be in close proximity to the basic residues in the first cytoplasmic loop. Furthermore, the conservation in sequence of this first loop indicates it must be important for rhodopsin function and regulation. We have succeeded in developing antisera to synthetic peptides that bind to the cytoplasmic domains of rhodopsin which will allow us to address this problem directly using site-directed probes and purified kinases. The final issue our results address is the mechanism of protein kinase C regulation of membrane proteins. Since it is assumed that diacylglycerol is important in activation and regulation of protein kinase C (51,52), then protein kinase C probably is activated in vivo when associated with a membrane. The breakdown of phosphatidylinositols is one mechanism to generate diacylglycerol and is thought to be involved in the regulation of protein kinase C (51, 52). Nishizuki and co-workers (53) and studies in our laboratory3 have demonstrated that protein kinase C bound to plasma membranes does not readily phosphorylate soluble substrates such as histone H1. The calcium-dependent binding and activation of protein kinase C in the presence of membranes is very different from the phosphatidylserine and diacylglycerol mixtures used to stimulate protein kinase C phosphorylation of soluble proteins. The apparent tight binding of protein kinase C to the membrane appears to actually sequester the enzyme in a localized region. The consequence of such a sequestration appears to be to limit the access of available substrates for phosphorylation. This could serve to target the enzyme and allow specificity of regulation for an enzyme that appears, in vitro, to have broad substrate recognition (51, 53). Inactivation of protein kinase C would require additional metabolism of membrane components so that the stability of the enzymecalcium-membrane complex would be reduced allowing the enzyme to dissociate. In the cell, this process would be predictably fast, and as long as a stimulus exists that generates diacylglycerol and/or calcium redistribution the enzyme could cycle by binding to the membrane followed by its release. In the ROS, for example, light might serve as the stimulus. In isolated membranes the ability to metabolize the membrane components required for the targeting and cycling of protein kinase C is apparently lost. In the presence of calcium protein kinase C is tightly bound to the membrane and remains sequestered until the calcium is chelated. The consequence of this sequestration in the presence of calcium is an apparently low turnover number for protein kinase C bound to mem-D. 3. Kelleher and G. L. Johnson, unpublished observation.