The Phosphorylation State of Phosducin Determines Its Ability to Block Transducin Subunit Interactions and Inhibit Transducin Binding to Activated Rhodopsin*

Heterotrimeric GTP-binding proteins (G-proteins) serve many different signal transduction pathways. Phosducin, a 28-kDa phosphoprotein, is expressed in a variety of mammalian cell types and blocks activation of several classes of G-proteins. Phosphorylation of phosducin by cyclic =-dependent protein kinase prevents phosducin-mediated inhibition of G-protein GTPase activity (Bauer, P. H., E. J. M., M. J. Nature 358, 73-76). In retinal rods, phosducin in- hibits transducin (G,) activation by binding its Py subunits. While rod phosducin is phosphorylated in the dark and dephosphorylated after illumination (Lee, R.-H., Brown, B. M., and Lolley, R. N. (1984) Biochemistry 23, 1972-1977), the significance of these reactions is still un-clear. The data presented here permit a more precise characterization of phosducin function and the consequences of its phosphorylation. Dephosphophosducin blocked binding of the G,a‘ subunit

Fax: 505-665-1464. tion pathways. Binding of G-proteins to their activated receptors results in exchange of GDP for GTP and dissociation of the GTP-bound Ga subunit from its GP-y subunits. Activated Gas (GwGTP's) regulate a variety of effector protein systems, including cyclic nucleotide processing enzymes, phospholipases, and cation-selective channel proteins (for review, see . In addition, GP-y subunits have recently been shown to regulate directly effectors such as phospholipase A, (4), P-adrenergic receptor kinase (5), type I1 adenylyl cyclase (6, 7) and phospholipase C-P (8,9). GTP hydrolysis by Ga allows the reassociation of Ga-GDP with GP-y, returning the system to its ground state in which G a p y G D P is again capable of binding to an activated receptor.
Phosducin is a highly conserved 28-kDa phosphoprotein that is expressed in a number of cell types including retinal rods and cones, pinealocytes, neurons, hepatocytes, and myocytes (10-13). Moreover, the retinal protein MEKA was found to be virtually identical to phosducin from retina and brain (11,12,14). Phosducin inhibits the GTPase activities of a number of Gproteins including Gi, G,, a n d Go (12); phosphorylation of phosducin with cyclic AMP-dependent protein kinase prevents this GTPase inhibition (12). Phosducin was first isolated in vertebrate retinal rod cells in a complex with GJ3-y subunits (151, and i t has been proposed that phosducin interferes with transducin (GJ function by binding to G$-y (16). While rod phosducin is known to be phosphorylated in the dark and dephosphorylated after illumination (171, the functional significance and biochemical details of these phosphorylation and dephosphorylation reactions have not been well defined. Here we report new observations that permit a more precise characterization of the functional biochemistry of rod phosducin and the consequences of its phosphorylation.

EXPERIMENTAL PROCEDURES
Preparation of Bovine Rod Outer Segments -Rod outer segments (ROS) were prepared by sucrose floatation (18). All manipulations of dark-adapted retinas were performed under infrared illumination using infrared image converters. Briefly, 250 fresh dark-adapted bovine retinas (J.A. & W.L. Lawson Co., Lincoln, NB) were suspended in the dark in 400 ml of HEPESRinger's buffer (10 m M HEPES pH 7.5, 120 m M NaC1, 3.5 m M KC1, 0.2 m CaCl,, 0.2 m M MgCl,, 0.1 m M EDTA, 10 m M glucose, 1 m M DTT, 0.2 ~l l~ PMSF) plus 45% (w/w) sucrose and disrupted by being drawn twice through a 60-ml syringe with a 0.7-cm diameter orifice. The disrupted retinas were passed through two layers of cheesecloth, and the retinal material that did not pass through the cheesecloth was supended in 50 ml of HEPESRinger's buffer, drawn through the syringe again, and reapplied to the cheesecloth. This process was repeated three times. The retinal material that did not pass through the cheesecloth was used for phosducin purification. The filtrate from the cheesecloth was centrifuged for 10 min at 4,000 x g. The resulting pellet was used for phosducin purification, and the supernatant, containing the ROS, was diluted to -15% sucrose with HEPESminger's buffer. This suspension was centrihged at 10,000 x g for 20 min. The resulting supernatant was used for phosducin purification, and the pellet was suspended in HEPESRinger's buffer, layered over 15 ml of 33% (w/w) sucrose in 6-8 tubes, and centrifuged at 40,000 x g for 15 min. The ROS were collected from the top of the 33% sucrose layer, diluted to -15% sucrose with HEPESRinger's, and centrifuged at 25,000 x g for 10 min. From these ROS, G, subunits and urea-stripped ROS membranes were prepared.
Protein Purifications a n d Membrane Preparations-Gp, G,py, and urea-stripped ROS membranes were prepared as described (19) except that after separation of G, subunits on a Blue Sepharose column (Pharmacia Biotech Inc.), the subunits were further purified with a high performance liquid chromatography (HPLC) size exclusion column (Bio-Silect SEC-250, Bio-Rad) by isocratic elution in 10 rn HEPES pH 7.4, 500 mM NaCI, 5 mM MgCl,, 1 mM EDTA, 1 mM D m , a n d 0.1 m~ PMSF (buffer A). The resulting G,py was >98% pure and was free of G,a as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (PAGE). The G p was >95% pure and was likewise free of G$y subunits. G,a and G,py were labeled with lZ5I using Bolton-Hunter reagent (DuPont NEN) as described previously (19).
Phosducin was purified by a modification of previously described procedures (11,15). The retinal debris retained on the cheesecloth was released and homogenized in 5 mM TrisC1, pH 7.5, 62 mM NaC1, 5 mM MgCI,, 2 p~ leupeptin, and 2 p~ pepstatin A in a Potter-Elvehjem homogenizer. The homogenate was centrifuged at 48,000 x g for 10 min; the supernatant was saved and the pellet was homogenized once more. The pellet from the 45% sucrose centrifugation step above was also homogenized twice. The supernatants were combined with the supernatant from the 15% sucrose centrifugation step (-2 liters), and proteins were precipitated by the addition of 474 gAiter ammonium sulfate. The solution was centrifuged at 30,000 x g for 25 min, and the pellet was resuspended in 30 rn potassium phosphate, pH 7.0, 1 mM DTT, 1 m~ EDTA, and 0.1 mM PMSF (buffer B). The suspension was dialyzed overnight against 9 liters of buffer B with one change of buffer after 6 h. The suspension was adjusted to pH 5.7 with 1 M acetic acid and centrifuged at 48,000 x g for 20 min. The pH and the conductivity of the solution were adjusted to 7.0 and 6 millisiemendcm, respectively, and loaded at 4 mumin at 4 "C on TSK-Gel DEAE Toyopearl 650 M (Toso-Haas) column material packed in an HR16/50 column (Pharmacia), using a Pharmacia fast protein liquid chromatography system. After the unbound material was washed through with buffer B, phosducin was eluted with 600 ml of a linear gradient from 30-500 mM potassium phosphate in buffer B a t a flow rate of 4.5 ml/min. The fractions with conductivities between 13 and 31 millisiemenskm were pooled and dialyzed against 10 rn potassium phosphate, pH 7.0, 1 mM D m , and 0.1 mM PMSF (buffer C). The solution was loaded onto Macro-Prep ceramic hydroxyapatite (40 pm, Bio-Rad) column material packed in an XK26/40 column (Pharmacia) equilibrated with the buffer C using fast protein liquid chromatography a t a flow rate of 3 mumin at 4 "C. The phosducin was eluted with 300 ml of a linear gradient of buffer C to buffer B containing 150 mM potassium phosphate, pH 7.0, a t 3 mumin. Fractions with conductivities of G 1 4 millisiemendcm were pooled and concentrated to -1.5 ml with CentriPreplO (Amicon) concentrators. The concentrated phosducin fraction was chromatographed on a TSK G3000SW size exclusion column (TosoHaas, 21.5 mm (inner diameter) x 30 cm with a 21.5 mm inner diameter x 7.5 cm guard column) running isocratically with buffer A at 5 mumin using a Waters 625LC HPLC system at 23 "C. The phosducin-G,py complex eluted a t 18 min. Phosducin was separated from G,py and other contaminants by strong anion exchange chromatography using a Poros Q/M 4.6/100 column (Per-Septive Biosystems). The column was equilibrated in 50 mM Tris, pH 8.0, 0.1 mM PMSF, and 1 m~ DTT and was eluted by a linear gradient of 50-750 m~ Tris at 3 mUmin (30 ml). The phosducin eluted at -560 mM Tris. Phosducin was further purifed by sucrose density gradient centrifugation (5 ml of 5-20% sucrose, 100 mM potassium phosphate, pH 7.4,2 mM MgCl,, 1 mM EDTA, 1 mM D m , 0.1 mM PMSF, and -0.5 mg of proteidtube) in a Beckman SW 55Ti rotor running in a Beckman L8-70 M centrifuge for 15.3 h at 53,000 rpm (w2t = 1.7 x 10" rad%) at 4 "C. This phosducin was 95% pure and was free of G,py as determined by SDS-PAGE and Western blotting with a peptide-specific antibody for GJ3. Protein concentrations were determined using Coomassie Plus protein assay reagent (Pierce Chemical Co.).
Phosducin Phosphorylation a n d Isoelectric Focusing-Phosducin was phosphorylated with the catalytic subunit of bovine heart CAMP&pendent protein kinase (Fluka) as follows. Phosducin (30 p~) was mixed with cAMP-dependent protein kinase (1 uniupl) and ATP (0.5 mM) in 50 mM HEPES, pH 7.5, 5 mM MgCI,, 1 nm EDTA, 1 mM D m , and 0.1 mM PMSF and incubated for 10 min at 23 "C. Phosphophosducin was im-mediately assayed for 1251-G,a binding (see below). For isoelectric focusing, phosducin was phosphorylated as described above in the presence of 0.5 mM (-1000 dpdpmol) [33PlATP, and then focused on a flat bed isoelectric focusing gel (0.4 mm) according to the manufacturer's (Bio-Rad) protocol using Pharmalyte 4.5-5.4 ampholines (Pharmacia) in the presence of 8 M urea. The gel was fixed in 20% trichloroacetic acid for 30 min. Ampholines were eluted from the gel by incubating in 0.25% SDS, 10% acetic acid, and 40% methanol for 30 min followed by 10% acetic acid and 40% methanol for 1 h. The gel was stained with 0.12% Coomassie Brilliant Blue R (Serva) in 10% acetic acid and 40% methanol for 30 min and then destained in the same solution without Coomassie Blue. Finally, the gel was autoradiographed to verify the incorporation of phosphate into phosducin.
Binding of lZ5I-G,a to Rho* in ROS Membranes-Light-induced binding of IZ5I-G,a and G,py to urea-stripped ROS membranes was carried out as described (19). Dephospho-or phosphophosducin, at the concentrations indicated, was mixed with 0.1 PM 1251-G,a and 0.1 p~ G,Py before the addition of urea-stripped membranes in the dark. Lightinduced binding was then initiated by bleaching 50% of the Rho. Samples were centrifuged, and the amount of '"I-Gp bound was quantified as described (19).
Phosducin Inhibition of *251-G,py Binding to ROS Membranes-The effect of phosducin on the binding of 1251-Gtpy to urea-stripped ROS membranes was measured by incubating 0.35 p~ 1251-G,py with increasing concentrations of phosducin in isotonic buffer (10 mM HEPES, pH 7.5, 100 mM KCl, 20 mM NaCl, 1.5 mM CaCl,, 1 mM EDTA, 1 m~ DTT, and 0.2 mM PMSF) for 10 min at 23 "C. Unilluminated urea-stripped membranes, a t 40 p~ Rho, were then added in the dark, and the mixture was incubated in the dark for 10 min. The final sample volume was 50 pl. For phosphophosducin, 0.5 unitdpl of CAMP-dependent protein kinase, and 0.5 mMATP were added to phosducin in isotonic buffer, and phosphorylation was allowed to continue for 5 min before the addition of the 1251-G,py and dark urea-stripped membranes. During the dark incubation, 20 pl of the sample were taken and counted for to obtain the total counts, The samples were then centrifuged a t 50,000 x g for 20 min to pellet the membranes, and 20 pl of the supernatant were removed and counted for lZ5I. The amount of 1251-G,py in the membrane pellet was calculated by subtracting the supernatant counts from the total counts in the sample. During these experiments, the phosphorylation state of phosducin was monitored in parallel samples that were incubated with 0.5 mM [32PlATP (500 cpndpmol). Incorporation of 32P into phosducin was quantified by precipitating 10 pl of the sample in 100 p1 of 20% trichloroacetic acid, filtering through type HA nitrocellulose filters in 96-well Multiscreen filtration plates (Millipore), washing three times with 100 pl of 20% trichloroacetic acid, and counting the 32P on the filter. Phosducin was completely phosphorylated by 2 min and remained phosphorylated throughout the experiment. Control samples without phosducin were run in order to correct for phosphorylation of the urea-stripped membranes under these conditions.
In order to obtain a binding constant for the interaction between 1251-Gtpy and phosducin, an equation for the binding of lZ5I-G,py to urea-stripped ROS membranes as a function of the total concentration of phosducin was derived from the following model (Equation I), which describes in simplest terms the binding of 1251-G,py to the membrane where M is the membrane binding site specific for 1251-G,py, and Pd is phosducin. From the expressions for the dissociation constants,

Effect of G,a on Phosducin Inhibition of 12'I-Gfpy Binding to ROS
Membranes-In the same experimental format used in the previous section, 2 PM phosducin was added to increasing concentrations of Gta followed by 0.35 p~ '*'I-G,py, and then unilluminated urea-stripped ROS (40 pv Rho) were added in the dark. The fraction of '*'I-G,py in the supernatant and the extent of phosducin phosphorylation were quantified as described above.
Effect of Phosducin on 12sI-G,(r Binding to ROS Membranes-In a similar experimental format, increasing concentrations of phosducin were added to 0.35 p~ '2sI-G,a followed by 40 p~ Rho in heat-denatured urea-stripped ROS membranes. The urea-stripped membranes were heat denaturated at 70 "C for 5 min prior to the addition of "'I-Gp and phosducin to remove residual amounts of G,py from the membranes (see "Results").
Gel Analysis of G,a, G,py and Phosducin Binding to ROS Membranes-Phosducin (4 p~) was mixed with G,py (6 PM) in the presence or absence of G,a (16 p~) in isotonic buffer. The mixture was incubated with unilluminated urea-stripped ROS membranes (40 p~ Rho) for 10 min in the dark a t 23 "C. The samples were centrifuged a t 50,000 x g for 20 min. The membrane pellets were resuspended in the dark with 500 p1 of hypotonic buffer (10 mM HEPES, pH 7.5, 1 mM EDTA, 1 mM DTT, and 0.2 mM PMSF) and incubated for 30 min a t 23 "C. The suspension was again centrifuged a t 50,000 x g for 20 min, and the resulting supernatant was concentrated to 35 p1 using Microcon 10 (Amicon) concentrators. 15 pl of the concentrated supernatant were analyzed by SDS-PAGE on a 12% Laemmli gel (20). The gel was stained with 0.12% Coomassie Brilliant Blue R (Serva) in 10% acetic acid and 40% methanol for 30 min and then destained in the same solution without Coomassie Blue.
HPLC Size Exclusion Chromatography-A Bio-Silect SEC 250-5 (Bio-Rad) size exclusion column was used for analytical HPLC analysis. The column matrix consists of a silica-based gel. Samples containing phosducin, phosphophosducin, G,py and G,a! (as indicated in the figure legends) were injected into the column and eluted isocratically with buffer A a t a flow rate of 1.0 mumin. Protein peaks were detected by measuring the absorbance of the eluate at 254 nm. Fractions from the column were pooled as indicated and concentrated in Centricon 10 (Amicon) microconcentrators, and equivalent volumes were analyzed by 12'70 SDS-PAGE (20). For phosducin plus G,Py samples, the proteins were mixed and incubated for 5 min a t 23 "C before injection. Phosphophosducin plus G,py was prepared by phosphorylating phosducin as described above and then adding G& and incubating for 5 min a t 23 "C before injection into the column. The same incubation steps were followed in assessing G,a binding to phosducin.

Phosphorylation State-dependent Inhibition of '25Z-G,a Binding to Rho* by Phosducin
In an effort to understand the mechanism of phosducin inhibition of G-protein activation and to assess the consequences of phosducin phosphorylation, direct effects of phosducin on 1251-G,a binding to Rho" were measured. Phosducin or phos- phophosducin was added in increasing concentrations to Iz5I-G p , G,P-y, and urea-stripped ROS membranes in '251-G,a binding assays (19). The binding of lz5I-G,a to Rho* was prevented by phosducin (Fig. lA). At lzsI-Gta and GJ-y concentrations of 0.1 p~, half-maximal inhibition occurred at -0.3 V M phosducin, and binding was totally prevented a t 2 p~ phosducin. Inhibition was cooperative (nspp = 1.6 0.2), consistent with the cooperative nature of G, binding to Rho* (19) and indicating that phosducin does control the concentration of free GpPy. In light of these data, the observed inhibition of G-protein GTPase activity and effector enzyme activation by phosducin (12,161 is best explained by phosducin's blocking G-protein interaction with the activated receptor. When phosducin was phosphorylated with CAMP-dependent protein kinase and ATP before assaying for inhibiton of 1251-G,a binding to Rho*, no inhibition was found even at phoducin concentrations up to 6 PM (Fig. L4). This corresponds to a decrease of a t least 50-fold in phosphophosducin's ability to inhibit 1251-Gta binding compared with that of dephosphophosducin.
The extent of phosducin phosphorylation under these conditions was determined by isoelectric focusing (Fig. 1B). In isoelectric focusing gels, three major isoforms of phosducin were observed near PI 5.2. These isoforms may result from sequence variations in phosducin. In retinal rods, variants of phosducin at amino acid position 44 have been reported. One isoform contains a His residue (11, 12,21), whereas the second isoform contains a Pro residue (10, 12). The different isoforms could also result from modifications during purification such a s proteolytic cleavage. However, if proteolysis had been occurring, it could not have been extensive because no proteolytic products were observed by SDS-PAGE, which could detect the loss of fragments of 1 kDa or larger. Phosphorylation by CAMP-dependent protein kinase shifted all isoforms to more acidic PI 7;alues (compare lanes a and b ) and an autoradiograph of the gel (lanes c and d ) showed that all isoforms were phosphorylated (lane d ) . Moreover, no residual phosducin was detected in the bands corresponding to unphosphorylated phosducin. Thus, it appears that none of the isoforms are phosphorylated in the purified phosducin and that all isoforms are completely phosphorylated upon treatment with CAMP-dependent protein kinase and ATP.
The data presented in Fig. 1 indicate that phosducin's ability to inhibit '*'I-G,a binding to Rho* is decreased more than 50fold when phosducin is in its phosphorylated state.
Dephospho-and Phosphophosducin Binding to G,Py-Phosducin is known to interact with G,Py and appears to dissociate solubilized GtaPy (16). Since GtPy is required for high affinity binding of G p to Rho* (22-241, phosducin could block G p binding to Rho" by sequestering G,Py in a phosducin-G,py complex. If this is the case, then phosphorylation of phosducin should inhibit phosducin-G,py interactions, since phosphophosducin did not block 1251-G,dG,py binding to Rho*. The effect of phosducin phosphorylation on the interaction with GtPy was directly examined in a silica gel-based HPLC size exclusion column. Unmodified phosducin eluted from an SEC 250-5 column with a retention time of 8.4 min, which was earlier than the predicted retention time for a 28-kDa protein (9.6 min, Fig. 2, A and E). This early elution could result from phosducin oligomerization or electrostatic repulsion of phosducin by the column matrix. The latter possibility appears likely because increasing the NaCl concentration from 0.5 to 1.0 M increased the retention of phosducin on the column to near normal for a 28-kDa monomer. Furthermore, sucrose density gradient centrifugation analysis of phosducin gave no apparent evidence for an oligomer. Using bovine serum albumin, ovalbumin, chymotrypsinogen, and cytochrome c as standards, phosducin exhibited an szo,w value of 2.2 S, which is somewhat lower than expected for a globular protein of 28 kDa. G,Py also exhibited an unexpected elution rate from the column. It was much delayed and eluted at 11.8 min, later than the total column volume (Fig. 2, B and E). This retention of G,Py on the column seems to be a result of hydrophobic interactions with the column matrix, since increasing the NaCl concentration further increased the retention time of G$y. These unusual interactions of phosducin and G,Py with the column matrix were exploited to obtain excellent separation of phosducin and G,Py on the column. When G,Py was incubated with phosducin prior to sample injection, G,Py eluted a t 8.4 min, overlapping the phosducin peak. This is evidenced by an increase in the area under the phosducin peak, by a shifting of G,Py into this peak (Fig. 2 F , lane c), and by the disappearance of G,Py at the 11.8 min elution time (Fig. 2, C and F , lane e).
Thus, binding to phosducin caused G,Py no longer to interact with the column matrix but to elute in a complex with phosducin that exhibited a retention time similar to that of free phosducin. This corresponds to the expected retention time for a Gt/3y.phosducin oligomer of 74 kDa. A very small fraction of the G,Py was found a t elution times between that of phosducin-G,Py and free G,Py (Fig. 2F, lane d ) , suggesting that the phosducin-G$y complex is quite stable on the column and that the column matrix is not slowly removing G,Py from the phosducin-G,Py complex as it passes through the column. HPLC size exclusion chromatography and gel analysis of peak fractlons were performed as described under "Experimental Procedures." Elution profiles from the size exclusion column are shown as follows:A, 3.3 nmol of phosducin; B, 1.1 nmol of G$y; C, 3.3 nmol of phosducin and 1.1 nmol G,Py; D , 3.3 nmol of phosphophosducin, 1.1 nmol of G,Py, 200 units of CAMP-dependent protein kinase, 12 nmol of ATP; E, molecular mass standards with the following retention times: 5.8 min, bovine thyroglobulin (670 kDa); 7.6 min, bovine y globulin (158 kDa); 8.9 min, chicken ovalbumin (44 kDa); 10.2 min, horse myoglobin (17 kDa); and 11.5 min, vitamin B-12 (1.35 kDa). F, SDS-PAGE of fractions (a-n) from the chromatograms as indicated. In panel C, phosducin and G,Py were incubated for 5 min a t 23 "C prior to sample injection. In panel D, phosphophosducin was prepared according to "Experimental Procedures," and then G,Py was added and incubated for 5 min a t 23 "C prior to sample injection. SDS-PAGE molecular mass standards are shown and correspond to the following molecular masses from top to bottom: 66,45,36,29, and 24 kDa.
Next, the interaction between phosphophosducin and GJy was examined in the presence of CAMP-dependent protein kinase and ATP, which were present to assure persistent phosducin phosphorylation. (Phosphophosducin was found to be ex- Procedures"). The total '?SI-Gtpy in the 2 0 4 sample was 7.0 pmol.
Error bars represent the standard error of the data from three separate experiments. The data were curve fit as described under "Experimental Procedures," yielding a Kd2 value of 0.11 PM for phosducin binding to '"I-Gtpy and 0.18 PM for phosphophosducin binding to lZSI-G,py. The saturatlon points for the inhibition were at 0.48 and 0.58 pmol for dephospho-and phosphophosducin, respectively. The curve fitted value of Kdl for the binding of 1251-Gtpy to the urea-stripped membranes was 2.2 p~ for dephosphophosducin, and this value was fixed a t 2.2 p~ in fitting the phosphophosducin curve.
tremely sensitive to trace amounts of phosphatase activity.) After a phosducin sample was phosphorylated, G,Py was added, and the mixture was injected. In this case, the phosducin peak remained unchanged, and the G,py eluted with the ATP peak ( Fig. 2 0 ) . Gel analysis showed that G,py had been lost from the phosducin peak and had reappeared in the 11.8 min peak (Fig.   3F, lanes j and 1 ). Thus, in this column matrix, phosphorylation inhibits phosducin's ability to bind G,&.
The data from the size exclusion column gives a qualitative comparison of dephospho-and phosphophosducin binding to G$y. In order to further quantify the differences in binding affinity between the two forms of phosducin, a second experimental approach was employed. This method took advantage of the fact that, unlike free G,&, the phosducin-G,py complex does not bind to ROS membranes (15). G,Py was iodinated with ['2sIlBolton-Hunter reagent in order to accurately determine amounts of G,py on the membrane and in the supernatant. 12sI-G,py was incubated with increasing concentrations of phosducin in the presence of unilluminated urea-stripped ROS membranes a t a 10-fold excess of Rho over 12sI-G,py. After incubation, the ROS membranes were pelleted, and the amount of 'ZSI-G,py in the pellet was quantified. In the absence of phosducin, 3.2 pmol, or 46% of the total "'I-G,py was found in the pellet (Fig. 3). The addition of phosducin inhibited '2sI-G,py binding to the membrane almost completely. Fitting the data to a simple model for 1:l stoichiometric binding of 1251-G,py to phosducin indicated that, at saturation, 0.48 pmol of '2sI-G,py were in the membrane pellet and that the Kd for phosducin binding to 1251-G,py was 0.11 m. Inhibition of 12sI-Gtpy binding to the membrane was not caused by phosducin competing for the same membrane binding sites as '2sI-Gtpy because phosdu- cin did not bind to the membrane. SDS-PAGE analysis showed < 5% of the total phosducin was found in the membrane pellet (Fig. 4B, lane b). Thus, the most compelling explanation of the inhibition of 12sI-Gtpy binding to the membrane is that it reflects conformational changes resulting directly from phosdu-~in-"~I-G,py interaction.
When phosducin was phosphorylated by CAMP-dependent protein kinase, only small decreases in the binding of '2sI-G,py were observed (Kd = 0.18 VM), and the saturation point for inhibition of '2sI-Gtpy binding to the membrane was similar (0.58 pmol in the pellet) to that of dephosphophosducin. Thus, phosphophosducin bound "'I-G,py nearly a s well as did dephosphophosducin. Iodination did not appear to change the interaction of GJy with phosducin, because similar results were obtained with unlabeled G,Py (measured by SDS-PAGE analysis of membrane bound and supernatant fractions). Moreover, unlabeled Gtpy competed equally well with 1251-Gtpy for phosducin binding in membrane binding experiments (data not shown). Eff'ect of G p on Phosducin Binding to G,py-The small decrease in binding affinity of phosphophosducin for GtPy could blot explain the striking difference in migration of G$y with the two forms of phosducin on the size exclusion column, nor could it explain the large difference in the ability of dephospho-and phosphophosducin to inhibit 1251-GtoJGpy binding to Rho*. Thus, it appeared that phosphorylation of phosducin was regulating phosducin's inhibition of G, activation by Rho* at a site different from phosducin-G,py binding. Therefore, the effect of G p on the binding of dephospho-and phosphophosducin to Gtpy was measured. Increasing amounts of Gta were added to '251-Gtpy and phosducin in the membrane binding experimental format described above (Fig. 4A). In the absence of phosducin, Gta caused an increase of lZ5I-Gtpy binding to unilluminated urea-stripped ROS membranes, as evidenced by an increase in 1251-Gtpy in the membrane pellet from 2.8 to 4.0 pmol. This result suggests that GpPy binds ROS membranes with higher affinity than G& alone. In the presence of 2 I" phosphophosducin, the binding of 1251-Gtpy to the membrane was inhibited as expected. Only 0.7 pmol of the 1251-Gtpy remained in the membrane pellet. When Gta was added, a large increase in 1251-Gtpy binding to the membrane was observed. '251-Gtpy in the membrane pellet increased from 0.7 to 4.0 pmol, reaching the level found in the absence of phosducin. These data indicate that G p does compete with phosphophosducin for binding to G$y, and that at -5-fold excess G,a the phosphophosducin is completely displaced by G p . G p competition with dephosphophosducin for 1251-Gtpy binding was more complex. In the presence of 2 PM dephosphophosducin, G p caused a more modest increase in 1251-Gtpy binding to the membrane (from 0.5 to 2.0 pmol). The initial increase in bound 1251-Gtpy paralleled that of phosphophosducin, but then it leveled off at 2.0 pmol 1251-Gtpy in the pellet, compared with 4.0 pmol in the absence of phosducin or with phosphophosducin. Thus, this increase in bound 1251-Gtpy does not appear to result from direct competition of G p with dephosphophosducin for binding to 1251-Gtpy. The increase also does not appear to result from increased binding of free 1251-Gtpy in the presence of G p , as was observed in the absence of phosducin. That effect exhibited a Kv, of 0.4 w which is equal to the 1251-Gtpy concentration, whereas the increased 1251-Gtpy binding in the presence of dephosphophosducin had a Kv, of 1.8 w. Moreover, very little free 1251-Gtpy exists at 2 phosducin and could not account for the observed 1.5 pmol increase in membrane-bound 1251-Gtpy.
This observation might suggest the formation of phosducin-GpPy complexes that bind G,a with lower affinity than does Gtapy, and which associate with ROS membranes with higher affinity than does phosducin-G,py. While phosducin-Gppy complexes have been proposed (12), no compelling evidence for such a complex has been presented. SDS-PAGE analysis of phosducin binding to the membrane pellet in the presence of excess G$y showed no bound phosducin in the absence or presence of G p , indicating that Gta does not increase phosducin binding to the membrane (Fig. 4B). In contrast, significant amounts of both G p and Gtpy were found in the membrane pellet, and G p caused an increase in Gtpy binding to the membrane as expected. Thus, the observed increase in 1251-Gtpy binding in the presence of dephosphophosducin and Gta cannot be explained by the formation of a phosducin-Gppy complex.
An alternative explanation for the data is that the different with the following retention times; 5.6 min, bovine thyroglobulin; 7.8 min, bovine y globulin; 9.2 min, chicken ovalbumin; 10.5 min, horse myoglobin; 12.5 min, vitamin B-12. In panel C, phosducin and G p were incubated for 5 min at 23 "C prior to sample injection. In panel D, phosphophosducin was prepared (see "Experimental Procedures"), and then G p was added and incubated for 5 min at 23 "C prior to sample injection.
isoforms of phosducin compete with G p for Gtpy binding with different affinities. One isoform might be more readily displaced from G,Py by Gta in a manner analogous to the displacement of phosphophosducin, whereas other isoforms might not be displaced by G p . This could account for the partial increase in 1251-Gtpy in the membrane pellet when G p is added in the presence of dephosphophosducin. Regardless of the reason behind partial competition between dephosphophosducin and Gta, the principal conclusion drawn from these experiments is that the Gta.Gtpy interaction is the locus at which phosphorylation controls phosducin inhibition of G, activation. Phosphophosducin is displaced from G,Py by G p and, therefore, does not inhibit G, activation by Rho*. In contrast, dephosphophosducin is not readily displaced from Gtpy by G p and, therefore, blocks G, activation by Rho* through sequestration of G,Py Assessment of G p Binding to Phosducin-Phosducin has been shown to inhibit the GTPase activity of Goa in the absence of GPy (121, suggesting that phosducin can bind Goa alone. Size exclusion chromatography (Bio-Silect SEC250) was used to assess phosducin binding to Gta in the absence of Gtpy (Fig. 5). G p eluted a t 10.0 min in the absence of phosducin (Fig. 5 A ) and phosducin eluted at 8.8 min in the absence of G p (Fig. 5B). When the two were incubated together prior to injection into the column, no change in elution times or peak areas for either phosducin or G p was observed (Fig. 5C). This is in sharp con- trast to the elution profile of phosducin plus G&, in which all the G,Py co-eluted with phosducin (Fig. 2C). In addition, phosphorylation of phosducin with CAMP-dependent protein kinase and ATP did not change the phosducin plus Gta elution profile (Fig. 5D). The possibility of a phosducin-G,a interaction was further probed by examining the effect of phosducin on the intrinsic affinity of Gta for ROS membranes in the absence of G,Py. Fig.  6 shows data for '251-G,a binding to ROS membranes in the presence of increasing phosducin concentrations. Since ureastripped ROS membranes contain residual amounts of endogenous G$y, the membranes were heat denatured prior to measuring 1251-Gta binding.2 Phosducin had no effect on 1251-Gta binding to the heat denatured membranes. Despite the addition of up to -20-fold excess of phosducin, no change in the amount of 1251-Gta in the membrane pellet was observed. Moreover, phosphorylation of phosducin did not modify the binding. This result should be contrasted with the large decrease in membrane bound 1261-GtPy (from 3.2 to 0.5 pmol) upon the addition of phosducin. Thus, the data from Figs. 5 and 6 give no evidence for a direct interaction between phosducin and Gp. The reason for the apparent discrepancy between these data and the Goa GTPase data mentioned above (12) is not obvious.
Heat denaturation decreased the amount of lZ5I-Gta in the membrane pellet from 4.2 -c 0.3 pmol ( n = 3) to 2.0 -c 0.2 pmol ( n = 4) under these conditions. Adding back exogenous G$y restored lZ5I-Gta binding to the membrane, with a Kv, of 0.1 VM, and the effect saturated at 5.6 pmol of the lZ5I-Gta in the pellet. Heat denaturation of the membranes had little effect on G,py binding. 3.1 -c 0.05 pmol ( n = 4) and 3.6 -c 0.06 pmol ( n = 2) of the 1251-G,py were in the membrane pellet in unheated and heat denatured membranes, respectively. Thus, it appears that heating urea-stripped ROS membranes in this manner denatures the -4% of the endogenous G,Py that remains after urea stripping. These results also suggest that GtPy does not bind to proteins in the ROS membrane, which would be denatured by heating. Moreover, endogenous Gtpy does not significantly affect the binding of exogenous Iz5I-Gtpr to the membranes under the condition of excess G,py binding sites used in these experiments. However, it is possible that G p and Goa differ in their ability to bind phosducin so that phosducin interacts differently with different G-proteins. DISCUSSION Phosducin inhibits G, activation by blocking its binding to Rho*, apparently as a result of G,Py sequestration in a phosducin-G$y complex. A single G,Py can act catalytically to activate many Gps, because once the Rho*.G,c@y complex dissociates upon GTP binding, the G,Py is freed to interact with other Gps (23). However, the rate of G p activation by Rho* depends on the concentration of free G&, with maximal activation occurring at a 1:l ratio of G,a to G,Py (24). Therefore, sequestering free G,Py in a complex with phosducin would decrease the rate of G p activation by Rho* without totally abolishing it. This type of down-regulation is precisely what is observed in light-adapted retinal rods (reviewed in Ref. 25).
The phosducin concentration in retinal rods is estimated to be equal to that of G, (151, so there is sufficient phosducin present to substantially decrease G, activation.
Although it has been suggested that phosphorylation of phosducin may regulate its function in retinal rods (161, the consequences of phosphorylation had been unclear. This may reflect the fact that the binding affinities of phospho-and dephosphophosducin for purified G,Py are similar. The data presented here show that phosphorylation of phosducin regulates its ability to block G p interaction with G$y. Phosphophosducin binds G,Py with similar affinity to dephosphophosducin but in a conformation that allows Gta to interact with G,Py and displace phosphophosducin. Once formed, the GpPy can be effectively activated by Rho*. In the unphosphorylated state, phosducin interacts with GtPy in a conformation that excludes Gta from accessing its binding site on G,Py, and, therefore, G,a cannot displace dephosphophosducin.
Evidence for two different conformations of phosducin-G$y, depending on the phosphorylation state of phosducin stems from their contrasting behavior on the size exclusion column. G,Py in the absence of phosducin interacted with the column matrix; but in the presence of phosducin, G,Py was bound in a conformation that did not allow it to interact with the column matrix, and the two proteins migrated in a phosducin-G$y complex. When phosducin was phosphorylated, despite having a similar affinity for G&, phosphophosducin and GtPy were separated on the column. The best explanation for this result is that phosphophosducin bound G,Py in a different conformation, which allowed G,Py to interact with the column matrix. Such an interaction would effectively pull GtPy from phosphophosducin during the column run. It is possible that G p acts in a manner similar to the column matrix in distinguishing between dephospho-and phosphophosducin-G,Py complexes.
From these findings, a model depicting phosducin as a phosphorylation-dependent regulator of light-activated G-protein mediated-signaling in retinal rods can now be proposed (Fig. 7). In unilluminated rods, cyclic nucleotide levels are elevated (26-28), cAMP-dependent protein kinase is active (29, 301, and phosducin remains phosphorylated (17). Under these conditions, phosphophosducin cannot compete with G p for G$y binding. Therefore, G p and G,Py are maximally associated in their active heterotrimeric state. After illumination, cyclic nucleotide levels fall (28, 31, 321, CAMP-dependent protein kinase is inactivated, and phosducin is dephosphorylated by unopposed rod phosphatases (17,331. Dephosphophosducin binds GtPy and prevents Gta from associating with G$y, thereby constraining Rho* activation of G p and slowing the rate of cGMP-phosphodiesterase activation by Gta.GTP. Thus, phosducin could act in concert with increased guanylyl cyclase (34) and rhodopsin kinase (35) activities (following light-induced clic nucleotides, the CAMP-dependent protein kinase activity, and, thus, the phosphorylation state of phosducin.