Residues within the Polycationic Region of cGMP Phosphodiesterase γ Subunit Crucial for the Interaction with Transducin α Subunit

Interaction between the γ subunit (Pγ) of cGMP phosphodiesterase and the α subunit (Tα) of transducin is a key step for the regulation of cGMP phosphodiesterase in retinal rod outer segments. Here we have utilized a combination of specific modification by an endogenous enzyme and site-directed mutagenesis of the Pγ polycationic region to identify residues required for the interaction with Tα. Pγ, free or complexed with the αβ subunit (Pαβ) of cGMP phosphodiesterase, was specifically radiolabeled by prewashed rod membranes in the presence of [adenylate-32P]NAD. Identification of ADP-ribose in the radiolabeled Pγ and radiolabeling of arginine-replaced mutant forms of Pγ indicate that both arginine 33 and arginine 36 are similarly ADP-ribosylated by endogenous ADP-ribosyltransferase, but only one arginine is modified at a time. Pγ complexed with Tα (both GTP- and GDP-bound forms) was not ADP-ribosylated; however, agmatine, which cannot interact with Tα, was ADP-ribosylated in the presence of Tα, suggesting that a Pγ domain containing these arginines is masked by Tα. A Pγ mutant (R33,36K), as well as wild type Pγ, inhibited both GTP hydrolysis of Tα and GTP binding to Tα. Moreover, GTP-bound Tα activated Pαβ that had been inhibited by R33,36K. However, another Pγ mutant (R33,36L) could not inhibit these Tα functions. In addition, GTP-bound Tα could not activate Pαβ inhibited by R33,36L. These results indicate that a Pγ domain containing these arginines is required for its interaction with Tα, but not with Pαβ, and that positive charges in these arginines are crucial for the interaction.

Interaction between the ␥ subunit (P␥) of cGMP phosphodiesterase and the ␣ subunit (T␣) of transducin is a key step for the regulation of cGMP phosphodiesterase in retinal rod outer segments. Here we have utilized a combination of specific modification by an endogenous enzyme and site-directed mutagenesis of the P␥ polycationic region to identify residues required for the interaction with T␣. P␥, free or complexed with the ␣␤ subunit (P␣␤) of cGMP phosphodiesterase, was specifically radiolabeled by prewashed rod membranes in the presence of [adenylate-32 P]NAD. Identification of ADP-ribose in the radiolabeled P␥ and radiolabeling of arginine-replaced mutant forms of P␥ indicate that both arginine 33 and arginine 36 are similarly ADP-ribosylated by endogenous ADP-ribosyltransferase, but only one arginine is modified at a time. P␥ complexed with T␣ (both GTP-and GDP-bound forms) was not ADP-ribosylated; however, agmatine, which cannot interact with T␣, was ADP-ribosylated in the presence of T␣, suggesting that a P␥ domain containing these arginines is masked by T␣. A P␥ mutant (R33,36K), as well as wild type P␥, inhibited both GTP hydrolysis of T␣ and GTP binding to T␣. Moreover, GTP-bound T␣ activated P␣␤ that had been inhibited by R33,36K. However, another P␥ mutant (R33,36L) could not inhibit these T␣ functions. In addition, GTP-bound T␣ could not activate P␣␤ inhibited by R33,36L. These results indicate that a P␥ domain containing these arginines is required for its interaction with T␣, but not with P␣␤, and that positive charges in these arginines are crucial for the interaction.
Cyclic GMP phosphodiesterase (PDE), 1 a key enzyme in phototransduction, is composed of P␣␤ and two P␥ subunits (1)(2)(3)(4)(5)(6). P␣␤ hydrolyzes cGMP (7,8) and binds cGMP to its high affinity, noncatalytic sites (9 -11). In amphibian ROS, P␥ regulates these P␣␤ functions as an inhibitor of cGMP hydrolysis (12) and as a stimulator of cGMP binding to noncatalytic sites (13,14). Different interactions between P␣␤ and P␥ have been suggested to be required to express these two functions (15,16). In bovine ROS, P␥ inhibits cGMP hydrolysis by P␣␤ (17); however, the effect of P␥ on the cGMP binding to noncatalytic sites has never been documented. In amphibian ROS, these P␥ functions are interrupted by P␥ release with GTP⅐T␣ from P␣␤ (12)(13)(14)18). We have recently suggested that these functionally different P␥s are released in the different steps of phototransduction (15,16). When [cGMP] is at the dark level, P␥ responsible for the inhibition of cGMP hydrolysis is released. Consequently, cGMP is hydrolyzed by the activated PDE for photoexcitation. When [cGMP] becomes low, P␥ responsible for the stimulation of cGMP binding is released, and the affinities of these noncatalytic sites to cGMP are drastically reduced. The resulting release of cGMP from these noncatalytic sites may facilitate the recovery of cytoplasmic [cGMP] to the dark level in ROS.
To understand physiological functions of protein-protein interaction, functional structures of the protein involved should be elucidated. P␥ functional structure is especially interesting because such a small protein (87 amino acids) plays important roles in phototransduction by interacting with various proteins. At least four different interactions are considered: (a) P␥-P␣␤ interaction for the inhibition of cGMP hydrolysis by P␣␤; (b) P␥-P␣␤ interaction for the stimulation of cGMP binding to P␣␤ noncatalytic sites; (c) P␥-T␣ interaction for the release of P␥ inhibitory strain from P␣␤; and (d) P␥-T␣ interaction for the release of P␥ to reduce the affinity of P␣␤ noncatalytic sites to cGMP. Previous studies have focused on interactions (a) and (c). Peptides have been used to identify the polycationic region of P␥ within residues 24 -45 and the car-boxyl-terminal region of P␥ corresponding to residue 46 -87 as the sites for the interaction (a) (19 -22). Mutational analysis of P␥ has also shown that the carboxyl-terminal residues and several positive charged residues in the polycationic region are also involved in the inhibition of cGMP hydrolysis (23,24). Without impairing interaction with P␣␤, a frameshift mutation of P␥ has also revealed that the carboxyl-terminal residues are involved in the cGMP hydrolysis inhibition (15). In addition, the polycationic region and a site near the carboxyl terminus in P␥ have been suggested as sites required for the interaction (c) (19,23,(25)(26)(27). However, little is known about P␥ domains involved in the interactions (b) and (d). The frameshift mutation of P␥ has suggested that the amino-terminal residues are involved in the stimulation of cGMP binding to noncatalytic sites on P␣␤ (15). However, a T␣ interaction site on P␥, which is required for the P␥ release to reduce the affinity of P␣␤ noncatalytic sites to cGMP, has not been identified.
In this study we have focused on identification of specific residues in the P␥ polycationic region for following reasons. (i) The polycationic region has been suggested to be involved in the interaction with both P␣␤ and GTP⅐T␣ for the regulation of cGMP hydrolysis. Identification of amino acid residues in the polycationic region seems to be crucial to reveal the mechanism for the P␥ release by GTP⅐T␣. However, residues required for these interactions have not been identified. (ii) We found that the specific arginines in the P␥ polycationic region were ADPribosylated by an endogenous enzyme. Thus, protein-protein interactions in which the polycationic region is involved may be monitored by tracing the P␥ ADP-ribosylation under physiological conditions. (iii) We also found that the P␥ ADP-ribosylation was regulated by the interaction between P␥ and T␣. Therefore, the ADP-ribosylation can be a useful tool to learn the interaction between P␥ and T␣. We describe that both arginine 33 and arginine 36 in the P␥, free or complexed with P␣␤, are ADP-ribosylated by endogenous arginine-ADP-ribosyltransferase. The P␥ ADP-ribosylation is inhibited when P␥ is complexed with T␣ (both GTP-and GDP-bound forms), suggesting that the domain including these arginines is not exposed to ADP-ribosyltransferase when P␥ is complexed with T␣. Then, using forms of P␥ mutated in these residues, we confirm that the domain is involved in the interaction with T␣. Moreover, we find that positive charges in these arginine are important for the interaction with T␣.

EXPERIMENTAL PROCEDURES
Materials-Mono Q (5 ϫ 50 mm), Pep RPC HR5/5 (5 ϫ 50 mm), TSK G2000SW (7.5 ϫ 300 mm), DEAE-Sephacel, SP-Sepharose Fast Flow, and Blue Sepharose CL-6B were purchased from Pharmacia Biotech Inc. AG 1-X2 resin was obtained from Bio-Rad. Other materials were purchased from the following sources: [adenylate- 32  Bleached ROS membranes from 20 frogs were suspended in 3 ml of Buffer B and passed through a no. 21 needle seven times. Membranes and soluble fractions were separated by centrifugation (200,000 ϫ g, 4°C, 15 min). ROS membranes were washed seven times in the same way, and these ROS membranes were termed prewashed ROS membranes. The prewashed ROS membranes were washed seven more times with 3 ml of Buffer C and seven times with 3 ml of Buffer C containing 400 M GTP. T␣ (more than 90%) and P␥ (about 50%) were released from these membranes. These membranes contain P␥-less (active) PDE and were termed as P␥-depleted ROS membranes. Residual P␥ in these membranes is not sensitive to GTP⅐T␣ (12), suggesting that the membrane preparation contains a distinct subset of P␥ which cannot be released by GTP⅐T␣. This subset of P␥ was termed GTP⅐T␣insensitive P␥. When P␥-depleted ROS membranes were used as a source for ADP-ribosyltransferase, these membranes were washed twice with 2 ml of Buffer C to remove residual GTP. When the P␥depleted membranes were washed an additional seven times with 3 ml of Buffer D, T␤␥ (more than 90%) was released. These membranes are termed P␥-and transducin-depleted ROS membranes. It should be emphasized that ADP-ribosyltransferase activity was detected in P␥depleted and P␥-and transducin-depleted membranes when frog or recombinant bovine P␥ was used as a substrate. However, ADP-ribosylation of the residual P␥ was not detected in these membranes. This suggests that GTP⅐T␣-insensitive P␥ cannot be ADP-ribosylated. Ureatreated ROS membranes were prepared as described (13).
Site-directed Mutagenesis of P␥-All DNA manipulations were carried out using standard procedures (28). Full-length bovine P␥ cDNA (29), which was ligated into the EcoRI-HindIII-digested plasmid pAL-TER-1 (Promega), was used in the mutagenesis steps. All mutagenic oligonucleotide primers used (Table I) were purchased from DNAgency Inc. (Malvern, PA). Mutagenesis was carried out using Altered Site System Mutagenesis Kit (Promega). Mutant clones were identified by in situ hybridization with 32 P-labeled mutagene oligonucleotides. The mutations were confirmed by double-stranded DNA sequencing using the fmol of DNA sequencing System (Promega). Two oligonucleotides: Up-5Ј-GCCAACCTGCATATGAACCTGGAGCC-3Ј and Down-5Ј GGGG-TCGGATCCTAGATGATGCCATACTG-3Ј were used in a polymerase chain reaction to introduce NdeI and BamHI sites at the ends of P␥ genes. The NdeI-BamHI fragment was cloned into the NdeI-BamHI- Oligo D ϩ oligo G ϩ oligo I a Underlined letters indicate mutation sites.
digested pET-IIA (Novogene). The vector was transferred to Escherichia coli BL21(DE3) (Novogene) for expression of P␥. Expression and Purification of Recombinant P␥-A fresh single colony was grown overnight at 37°C in the presence of 100 g/ml ampicillin. The overnight culture was diluted 1:100 into a medium (12 g of tryptone, 24 g of yeast extract, and 0.2 ml of 5 M NaOH/1,000 ml) containing 100 g/ml ampicillin, and the culture was grown at 37°C. At an A 600 nm Ϸ 0.6, protein expression was induced by the addition of 1 mM (final) isopropyl ␤-D-thiogalactopyranoside, and cells were incubated for another 4 h at room temperature. Then, cells were spun down (1,700 ϫ g, 20 min, 4°C) and resuspended in 0.10 volume of Buffer E. After sonication, insoluble material was spun down (200,000 ϫ g, 15 min, 4°C), and the supernatant was loaded into an SP-Sepharose Fast Flow column (6 ϫ 100 mm) that had been equilibrated with Buffer E. After washing with 10 bed volumes of Buffer E containing 100 mM NaCl, the protein was eluted using 10 ml of Buffer E with a gradient of 0.2-0.5 M NaCl. Following the measurement of PDE inhibitory activity, the active fractions were collected, heated at 80°C for 5 min, and centrifuged (345,000 ϫ g, 30 min, 4°C). P␥ and its mutants were further purified using Pep RPC HR5/5 column as described (12). The purity of P␥ and its mutants was greater than 90%.
Purification of Proteins-Frog P␥ was purified from the supernatant prepared by washing prewashed ROS membranes with Buffer C containing GTP (12). Siliconized tubes and pipette tips were used in all experiments using P␥ except SDS-gel electrophoresis. T␣ (GTP␥S-and GDP-bound forms) and T␤␥ were isolated as described (30). In some experiments partially purified ADP-ribosyltransferase was used. ADPribosyltransferase was solubilized in 10 ml of Buffer F containing phosphatidylinositol-specific phospholipase C (1 unit) from P␥-depleted membranes (120 mg of protein). Following incubation (37°C, 30 min), the sample was centrifuged (200,000 ϫ g, 20 min, 4°C). The supernatant was applied to a Blue Sepharose CL-6B column (3 ml) that had been equilibrated with Buffer F and washed with 12 ml of Buffer F. Flow-through and washing fractions were collected and applied to NAD-agarose (2 ml) that had been equilibrated with Buffer F. The column was washed with 15 ml of Buffer F, and ADP-ribosyltransferase was eluted with 6 ml of Buffer F containing 100 M NAD (its purity was about 80%). After dialysis against Buffer F, ADP-ribosyltransferase activity was measured.
ADP-ribosylation of P␥, Peptides, and Agmatine-ADP-ribosylation of P␥ was carried out with various membrane preparations or enzyme preparations solubilized by detergents or phosphatidylinositol-specific phospholipase C. Frog P␥ and recombinant bovine P␥ were used. Both P␥s were ADP-ribosylated in a similar manner. The amounts of each components were slightly different in each experiment (see each figure legend); however, P␥ ADP-ribosylation was performed in a similar way. The reaction mixture contained P␥ (0.2-0.6 g), NAD (10 -50 M; ϳ0.5 Ci) and proteins in the 50 l of Buffer G. P␥ ADP-ribosylation was initiated by the addition of [adenylate-32 P]NAD and terminated by heating with SDS-sample buffer (5 min, 80°C). These samples were analyzed by SDS-polyacrylamide gradient (8 -16%) gel electrophoresis and autoradiography. The band corresponding to P␥ (its apparent molecular weight is 13,000 in gels) was also excised from gels, and its radioactivity was measured. The radioactivity was proportional to the value obtained by densitometric scanning. As a control, the same procedure was performed without added P␥, and the radioactivity of the corresponding region was subtracted from the radioactivity of the P␥ band. Without added P␥, no 13,000 band was detected, and the radioactivity was negligible. If one amino acid in P␥ was radiolabeled, approximately 3-10% of P␥ was apparently radiolabeled in the regular experiments, and the apparent maximum incorporation of radioactivity was 23% (see Fig. 2). Possible reasons for such a low incorporation of the apparent radioactivity will be discussed later. When pertussis toxin was used, pertussis toxin was activated as described (31). A peptide corresponding to P␥ residues 30 -39 (FKQRQTROFK) and its mutant forms was also ADP-ribosylated by P␥-and transducin-depleted membranes (76 g of protein). After various amounts of these peptides (0.05-0.6 g) were ADP-ribosylated (1 h, 33°C) in 100 l of Buffer G containing [adenylate-32 P]NAD (10 M; ϳ0.25 Ci), 40 l of the reaction mixtures was spotted onto 2 ϫ 2-cm pieces of Whatman P-81 phosphocellulose (32). Then, all of these pieces were washed five times with 400 ml of 1% trichloroacetic acid, rinsed with ethanol, and the radioactivity of each paper was measured. ADP-ribosylation of agmatine (20 mM) was carried out in a manner similar to that described above. ADP-ribosylated agmatine was isolated as described previously (33). Samples (100 l) were applied to AG 1-X2 columns (1 ml) that had been equilibrated with 20 mM Tris⅐HCl (pH 7.5). [ 32 P]ADP-ribosylated agmatine was eluted with 5 ml of water.
Preparation of ADP-ribosylated P␥-Although the amounts of components in each reaction mixture were slightly different (see each figure legend), ADP-ribosylated P␥ was prepared by incubation of purified P␥, P␥-and transducin-depleted ROS membranes, and [adenylate-32 P]NAD. ADP-ribosylated P␥ was isolated using a Pep RPC column (see Fig. 2) or SDS-gel electrophoresis (see Fig. 3). In the case of the Pep RPC column, the same volume of 0.2 M formic acid was added into the sample to terminate the reaction, and the mixture was heated for 5 min at 80°C. Then, the sample was centrifuged (345,000 ϫ g, 30 min, 4°C). The process of the P␥ extraction by formic acid was repeated two times. The pooled supernatant was applied to a Pep RPC HR5/5 column that had been equilibrated with 0.1% trifluoroacetic acid. Elution of P␥ was carried out with an acetonitrile gradient (0 -60%) containing 0.1% trifluoroacetic acid, as shown in Fig. 2. The flow rate was 1 ml/min, and the fraction volume was 0.5 ml. Following measurement of both radioactivity and P␥ activity in each fraction, the active fractions were dried, suspended in water, and kept at Ϫ80°C. In the case of SDS-gel electrophoresis, samples were heated with SDS-sample buffer to terminate the reaction. The gel was stained in 20% methanol containing 0.2% (w/v) Coomassie Blue and 0.5% (v/v) acetic acid (20 min) and destained with 30% (v/v) methanol (1 h). The visualized P␥ was excised from the gel.
Identification of ADP-ribose in the [Adenylate-32 P]NAD-radiolabeled P␥-To identify ADP-ribose in the radiolabeled P␥, the radiolabeled frog P␥ was incubated with glycine/NaOH buffer (pH 10.0) or treated with snake venom phosphodiesterase, as shown in Fig. 3. After centrifugation of these reaction mixtures (345,000 ϫ g, 30 min, 4°C), supernatants were applied with ADP-ribose (100 M) or AMP (100 M) to a Mono Q column that had been equilibrated with 10 mM sodium phosphate (pH 6.0). After washing the column, radioactive compounds were eluted. Chromatographic conditions were: A, 10 mM sodium phosphate (pH 6.0); and B, 10 mM sodium phosphate (pH 6.0) and 0.2 M NaCl; 0 -100% B in 16 ml on a linear gradient. The flow rate was 1 ml/min, and the fraction volume was 0.5 ml.
Measurement of Molecular Ion Mass of ADP-ribosylated P␥-Fifty l of ADP-ribosylated recombinant bovine P␥ (0.2 mg/ml) was injected into a high performance liquid chromatography system that consisted of Ultrafast Microprotein Analyzer model 600 (Microm BioResources, Auburn, CA) equipped with a Reliasil C 18 reverse phase column (5 m particle size, 300 Å pore size, 1.0 ϫ 150 mm) with a flow rate 40 l/min. Chromatographic conditions were: A, 0.1% trifluoroacetic acid, 2% acetonitrile, H 2 O; and B, 0.07% trifluoroacetic acid, 90% acetonitrile, H 2 O, 0 -56% B in 25 min on linear gradient. The flow was monitored at 215 nm, and the eluate was introduced directly into an API-III triple quadrupole mass spectrometer (Perkin-Elmer Sciex, Thornhill, Ontario, Canada) equipped with an electrospray atmospheric pressure ionization source. The tuning and calibration were done using polypropylene glycol. The mass spectrometer was set to scan in a positive ion mode at orifice potential of 90 V from m/z ϭ 500 -2,200 with a mass step of 0.4 Da. The mass data were analyzed by MacSpec 3.22 (Perkin-Elmer Sciex).
Analytical Methods-Activities of PDE and P␥ were assayed as described (12). GTPase activity of T␣ and GTP␥S binding to T␣ were measured as described (18). Immunological detection of P␥ was carried out as described (34). SDS-polyacrylamide gel electrophoresis was performed as described (30). When [adenylate-32 P]NAD was present in the electrophoresis, the gel was cut above the dye front to remove free radioactive NAD for the reduction of background in autoradiography. Therefore, we do not show the dye front in each picture of gel. Protein concentrations were assayed with bovine serum albumin as standard (35). The amount of P␥ was assayed by densitometric scanning (12). To calculate the P␥ concentration, 9,625 and 9,669 were used as molecular weights of frog (36) and recombinant bovine (see Fig. 4) P␥, respectively, although P␥ was detected as a 13,000 band in SDS-gels. It should be emphasized that all experiments were carried out more than two times, and the results were similar. Data shown are representative of these experiments.

RESULTS
An Arginine Residue (Arg-33 or Arg-36) in P␥ Is ADP-ribosylated by Endogenous Arginine-ADP-ribosyltransferase-In the presence of [adenylate-32 P]NAD a 39-kDa protein in prewashed ROS membranes was radiolabeled by pertussis toxin (Fig. 1). However, the radiolabeling almost disappeared if ROS membranes were washed with a buffer containing GTP (data not shown). These data indicate that T␣ in prewashed ROS membranes is ADP-ribosylated by pertussis toxin, as described previously (31,37,38). In the absence of pertussis toxin, T␣ ADP-ribosylation was not detected, indicating that endogenous T␣ ADP-ribosylation (39,40) is negligible under our conditions. Under the same conditions, a 13-kDa protein was also radiolabeled, and the radiolabeling was more clearly observed in the absence of pertussis toxin. The radiolabeling was increased by the addition of purified P␥. These data support the idea that the radiolabeled 13-kDa protein is P␥ and that both endogenous and exogenous P␥ are radiolabeled by an enzyme(s) in prewashed ROS membranes in the presence of [adenylate-32 P]NAD. Following densitometric scanning of the protein band and measurement of its radioactivity, we estimate that approximately 50% of endogenous P␥ was radiolabeled if one amino acid in P␥ was radiolabeled.
To confirm that the radiolabeled 13-kDa protein is P␥, the radiolabeling was conducted using purified P␥ and P␥-and transducin-depleted membranes in the presence of [adenylate-32 P]NAD. Then, the 13-kDa protein was isolated by using a reverse phase column ( Fig. 2A). In the column chromatography, both radioactivity and PDE inhibitory activity were detected in the same fractions (Fig. 2B). Analysis of the radioactive fractions by SDS-gel electrophoresis and autoradiography of the gel indicate that the 13-kDa protein was isolated with a purity of more than 95% in these fractions and that the radioactivity was incorporated into the 13-kDa protein (Fig. 2C). Without P␥, the 13-kDa protein was not observed in the column chromatography (data not shown). Without P␥-and transducin-depleted membranes, the radiolabeling of the 13-kDa protein was not detected (data not shown). These data indicate that P␥ is radiolabeled by a membrane-bound enzyme(s) in the presence of [adenylate-32 P]NAD. If one amino acid in P␥ was radiolabeled, approximately 23% of P␥ was radiolabeled in the reconstituted system. Under these conditions P␥ is roughly estimated as a mixture of free P␥ (95%) and P␥ complexed with P␣␤ (5%) if all of the P␣␤ in the membranes is occupied by exogenous P␥. Thus, these data indicate that free P␥ is radiolabeled. We also note that P␥ complexed with P␣␤ is radiolabeled. Under the conditions shown in Fig. 1, P␥ appears to be complexed with P␣␤ for because (i) PDE activity in the ROS membranes was low, and (ii) addition of GTP or GTP␥S stimulated PDE activity. We also radiolabeled bovine P␣␤␥2 using partially purified frog ADP-ribosyltransferase after separa-tion 2 of P␣␤␥2 from P␣␤␥ and P␣␤. We found that ϳ20% of P␥ in the complex was radiolabeled (data not shown). Thus, we conclude that P␥, free or complexed with P␣␤, is radiolabeled by a membrane-bound enzyme(s).
Radiolabeled P␥ was treated under the following conditions: (i) incubation in a glycine/NaOH buffer (pH 10), and (ii) incubation with snake venom phosphodiesterase. These treatments have been used for the identification of ADP-ribose in the ADP-ribosylated ␣ subunit of G-protein (41)(42)(43). The radioac-2 A. Yamazaki, unpublished method.

FIG. 2. Isolation of [adenylate-32 P]NAD-radiolabeled P␥.
Purified frog P␥ (10 g) was incubated with P␥-and transducin-depleted ROS membranes (276 g of protein) in 160 l of concentrated (ϫ1.25) Buffer G for 30 min at 0°C. Radiolabeling of P␥ was performed by the addition of 20 l of 500 M NAD (ϳ25 Ci) at 33°C. After a 1-h incubation, an additional 20 l of 500 M NAD (ϳ25 Ci) was applied to the reaction mixture, and the mixture was further incubated for 1 h. The reaction was terminated by the addition of 200 l of 0.2 M formic acid and heating at 80°C for 5 min. The supernatant was collected by centrifugation (345,000 ϫ g, 30 min, 4°C) and applied to a Pep RPC column. P␥ was eluted with an acetonitrile gradient as described. Panel A, profile of absorbance at 280 nm. The arrow indicates the position of P␥ eluted when purified frog P␥ was applied to the column. Panel B, radioactivity (q) and PDE inhibitory activity (E) of each fraction. Each fraction (20 l) was used to measure its 32 P radioactivity. After drying each fraction (20 l), its PDE inhibitory activity was measured using P␥-depleted ROS membranes. Panel C, purity of radiolabeled P␥. Lane A, M r standards: a, 94,000; b, 68,000; c, 43,000; d, 30,000; e, 20,000; f, 14,000. Lane B, radiolabeled P␥ (1 g) visualized by Coomassie Blue. Lane C, autoradiography of radiolabeled P␥. tive products were then fractionated using a Mono Q column. As shown in Fig. 3, the radioactivity was detected in ADPribose fractions when the radioactive P␥ was incubated in the glycine/NaOH buffer. In contrast, the radioactivity emerged in the AMP fractions when the P␥ was incubated with phosphodiesterase. No other radioactive peak emerged in any fractions. It should be emphasized that nonenzymatic binding of NAD to P␥ is not involved in the P␥ radiolabeling, since the radioactivity was not detected in NAD fractions (fractions 4 and 5). We also note that nonenzymatic binding of ADP-ribose to P␥ is excluded, since the radiolabeling of P␥ by [adenylate-32 P]NAD was not inhibited by preincubation of P␥ with ADP-ribose (data not shown). These observations indicate that P␥ is ADP-ribosylated.
We also analyzed the radiolabeled P␥ by electrospray ionization mass spectrometry. Two different molecular ion masses were detected in the purified sample (Fig. 4). The estimated molecular ion mass of the major peak is 9,668.48 (standard deviation 0.97). The major peak is believed to be nonmodified P␥, since (i) without radiolabeling, P␥ was detected as a single peak with the exact same molecular ion mass (data not shown); and (ii) the calculated molecular ion mass of the nonmodified P␥ is 9,670.28. The estimated molecular ion mass of the second peak (approximately 20% of the major peak) was 10,209.90 (standard deviation 1.08). The difference in the observed masses of these two peaks is 541.52. This value is in fair agreement with gain in molecular ion mass of G-protein ␣ subunit by ADP-ribosylation (541.3). Taking into account the purity of radiolabeled P␥ and the level of radiolabeling (Fig. 2), we conclude that the second peak is ADP-ribosylated P␥. These data indicate that a single ADP-ribosyl moiety is incorporated into P␥. These observations also confirm that nonenzymatic binding of NAD to P␥ (663.4 increase in molecular ion mass) is excluded.
To identify an ADP-ribosylated amino acid, we treated the radiolabeled P␥ under different conditions. Neither low pH (HCl, 0.1 M) nor HgCl 2 (10 mM) reduced the radioactivity from the P␥ (data not shown), suggesting that a cysteine in P␥ is not ADP-ribosylated. This conclusion is also supported by the observation that L-cysteine methyl ester did not inhibit the radiolabeling of P␥ (up to 20 mM) (data not shown). In contrast, as shown in Fig. 3, the radiolabeled P␥ is sensitive to high pH (pH 10.0). The radioactivity was also decreased when the radioactive P␥ was incubated with hydroxylamine (Fig. 5A). Moreover, the P␥ radiolabeling was inhibited by novobiocin (Fig. 5Ba), an inhibitor of arginine-ADP-ribosyltransferase (42), and by Larginine methyl ester (Fig. 5Bb). These observations indicate that an arginine in P␥ is ADP-ribosylated.
Bovine P␥ contains five arginine residues: Arg-11, Arg-15, Arg-24, Arg-33, and Arg-36 (29). We attempted to isolate a peptide(s) containing an ADP-ribosyl arginine after proteolytic digestion of the radiolabeled P␥, but we failed. This is probably because the ADP-ribosyl moiety was released from the radiolabeled P␥ during proteolytic digestion. The radioactive P␥ may be sensitive to longer incubation (room temperature, 18ϳ24 h)

FIG. 3. Identification of ADP-ribose in P␥ radiolabeled with [adenylate-32 P]NAD.
Purified frog P␥ (4.0 g) was incubated with [adenylate-32 P]NAD (15 M; ϳ4 Ci) and P␥-and transducin-depleted ROS membranes (85 g of protein) in 50 l of Buffer G for 60 min at 33°C. The reaction was terminated by the addition of SDS-sample buffer (30 l) and heating for 5 min at 80°C. Following SDS-gel electrophoresis of the sample, the gel was rapidly stained and destained as described. After visualization, P␥ was excised from the gel and cut into two pieces, and these two gel pieces were immersed overnight in 2 ml of water at 0°C and then in 600 l of acetonitrile for 20 min at room temperature (twice). These two gel pieces were dried under reduced pressure. One gel piece was incubated overnight in 50 mM glycine/ NaOH (pH 10.0) at room temperature and spun (345,000 ϫ g, 30 min, 4°C). After the supernatant (275 l) was adjusted to pH 7.0 with HCl, the volume of the sample was also adjusted to 1 ml with water and spun (345,000 ϫ g, 30 min, 4°C). The supernatant was analyzed by a Mono Q column as described. The other gel piece was swollen with 5 l of Buffer H, and then snake venom phosphodiesterase (5 g) suspended in 5 l of Buffer H was added for the further swelling of the gel (10 min, room temperature). After the addition of Buffer H (590 l) the gel piece was incubated at 37°C for 3 h. Then, 400 l of Buffer H was added to the mixture. The sample was spun (345,000 ϫ g, 30  in these buffers (pH 8.0 -8.5). In fact, after incubation of the radiolabeled P␥ under the same conditions (except without proteinase), a large portion of radioactivity was detected in the flow-through fractions in the reverse phase column chromatography. Therefore, to identify an arginine for ADP-ribosylation, we created mutant forms of P␥ in which an arginine was replaced by a lysine. PDE inhibitory activities of these P␥ mutants, summarized in Table II, show that all mutants have inhibitory activities similar to that of wild type P␥. These observations suggest that the mutation does not cause drastic change in the P␥ conformation required for the inhibition of cGMP hydrolysis. As shown in Fig. 6, these P␥ mutants were radiolabeled if each arginine was singly replaced by lysine (R11K, R15K, R24K, R33K, and R36K). However, the P␥ radiolabeling was abolished if both Arg-33 and Arg-36 are re-placed by lysines (R11,15,24,33,36K; R11,15,33,36K, and R33,36K). In contrast, the P␥ radiolabeling was detected even if two other arginines are replaced by lysines (R11,15K and R15,24K). We also confirmed these data using a peptide corresponding to residues 30 -39 of P␥ (FKQRQTROFK) and its mutant forms. The wild type peptide and mutant forms of the peptide (R33K and R36K) were similarly ADP-ribosylated by P␥-and transducin-depleted membranes. However, a mutant form of the peptide (R33,36K) was not modified under the same conditions (data not shown). Together with data that one ADPribose is incorporated in the radiolabeled P␥ (Fig. 4), these results indicate that Arg-33 and Arg-36 in P␥ can be ADPribosylated; however, only one ADP-ribose is incorporated at a time into one of these two arginines. The time course of radiolabeling on both R33K and R36K mutants appeared similar (Fig. 6), suggesting that the possibility for the radiolabeling of Arg-33 and Arg-36 is similar under these conditions.
Effect of T␣ on P␥ ADP-ribosylation-ADP-ribosylation of P␥ by partially purified ADP-ribosyltransferase was inhibited by both GTP␥S-and GDP-bound forms of T␣ (Fig. 7). ADP-ribosylation of P␥ by the enzyme solubilized from ROS membranes by n-dodecyl-␤-D-maltoside was also inhibited by T␣ (data not shown). We note that P␥ forms a complex with both GTP␥S⅐T␣ and GDP⅐T␣ (12,34). These observations suggest the following two possibilities: (i) after complex formation with T␣ (both GTP-and GDP-bound forms), P␥ is not a substrate for ADPribosyltransferase; and/or (ii) ADP-ribosyltransferase is inhibited directly by T␣.
Agmatine is a simple arginine derivative often used as an artificial substrate for arginine-ADP-ribosyltransferase (33). ADP-ribosylation of agmatine by partially purified ADP-ribosyltransferase was carried out in the presence or absence of T␣ (GTP␥S-or GDP-bound forms). As shown in Fig. 8, agmatine was ADP-ribosylated; however, the ADP-ribosylation was not affected by T␣. Using a TSK-250 column (34), we confirmed that agmatine does not form a complex with T␣ (data not shown). These observations indicate that ADP-ribosyltransferase is not inhibited by T␣. Therefore, we conclude that P␥ is no longer a substrate for ADP-ribosyltransferase after complex formation with T␣. The simplest explanation for these phenomena is that both Arg-33 and Arg-36 in P␥ are masked by T␣. Thus, a domain including these arginines is involved directly in the P␥ interaction with T␣. In contrast, P␥ complexed with P␣␤ is a substrate for ADP-ribosyltransferase, as described above. Thus, these arginines seem to be exposed to the enzyme when P␥ is complexed with P␣␤, suggesting that these arginines are not directly involved in the P␥ interaction with P␣␤.
Effect of Site-directed mutagenesis of Arg-33 and Arg-36 in P␥ on the interaction between P␥ and T␣-To confirm the role of Arg-33 and Arg-36 in P␥ in the interaction with T␣, both Arg-33 and Arg-36 were replaced by lysine or leucine. These mutants inhibited PDE activity similarly to wild type P␥ (Table  II). These data support our conclusion that these arginines are not crucial for the interaction with P␣␤ to inhibit PDE activity. Then, GTPase activity of T␣ and GTP␥S binding to T␣ was measured in the presence of various amounts of these P␥ mutants. We have already shown that wild type P␥ inhibits both GTPase activity of T␣ and GTP␥S binding to T␣ under our conditions and that these phenomena are used as evidence for the interaction between T␣ and P␥ (34,44). As shown in Fig.  9A, the P␥ mutant R33,36K inhibited GTPase activity; however, the P␥ mutant R33,36L did not inhibit GTPase activity. Moreover, the R33,36K mutant inhibited GTP␥S binding to T␣, but the R33,36L mutant did not inhibit GTP␥S binding (Fig.  9B). Furthermore, GTP␥S⅐T␣ activated PDE that had been inhibited by the R33,36K mutant, but not PDE that had been inhibited by the R33,36L mutant (Fig. 10). These data indicate that arginines 33 and 36 are involved in the P␥ interaction with T␣ and that positive charges of these arginines are important for the interaction between T␣ and P␥. We note that another arginine in the polycationic region, Arg-24, is not involved in the interaction with T␣. A mutant form of P␥, R24E, inhibited GTPase activity in the same manner as wild type P␥ (data not shown). This indicates that arginines 33 and 36 in the P␥ polycationic region have a special function for the interaction with T␣. DISCUSSION PDE, a key protein to regulate the level of cGMP in retinal photoreceptors, is composed of P␣␤ and two P␥ subunits. P␥ has two roles in P␣␤ regulation: inhibiting cGMP hydrolysis by P␣␤ (12,17) and stimulating cGMP binding to high affinity, noncatalytic sites on P␣␤ (13,14). We have recently indicated that an identical P␥ expresses these different functions by FIG. 6. ADP-ribosylation of arginine-mutant P␥s. In the presence of P␥and transducin-depleted ROS membranes (320 g), mutant P␥s (0.8 g) were incubated at 33°C in 250 l of Buffer G. P␥ radiolabeling was initiated by the addition of [adenylate-32 P]NAD (50 M; ϳ2 Ci). Following incubation for various periods, an aliquot (50 l) was mixed with 20 l of SDS-sample buffer and heated at 80°C for 5 min. After SDS-gel electrophoresis and autoradiography, the P␥ band was excised from gel, and its radioactivity was measured. Based on the radioactivity of wild type P␥ after a 60-min incubation, the radioactivity of each mutant was calculated.   ϪIRSATRVMGGPVTPRKGPPKFKQRQTRQ- a Concentration of P␥ and its mutants for 50% inhibition of P␥-depleted PDE activity. Inhibitory effects of bovine recombinant P␥ and its mutants were investigated using P␥-depleted ROS membranes and various amounts of P␥. The specific inhibitory activity values (inhibition/mg of protein) and maximal inhibitory effect on P␣␤ were similar for wild type and mutant P␥s (Ϯ5%).
binding to different sites on P␣␤ (16) and that different regions in P␥ are involved in these functions (15). Since these P␥ functions are expressed by interaction with P␣␤ and interrupted by P␥ release with GTP⅐T␣, the functional structure of P␥ required for these interactions should be clarified to understand the regulation of these P␥ functions. In this study we have shown that two arginines, 33 and 36, in the P␥ polyca-tionic region are equally ADP-ribosylated by endogenous ADPribosyltransferase, but only one arginine is ADP-ribosylated at a time. We speculate that steric hindrance may contribute to ADP-ribosylation of one arginine. The ADP-ribosylation was detected when P␥ is complexed with P␣␤. However, the ADPribosylation was inhibited when P␥ is complexed with T␣ (GTPand GDP-bound forms). These data imply that these arginines are masked when P␥ is complexed with T␣. Then, site-directed mutagenesis was applied to replace these arginines with lysines or leucines, and the effects of these P␥ mutants on T␣ functions were measured. These experiments confirm that these arginines are crucial for the interaction with T␣. We have also shown that arginine 24, another arginine in the P␥ polycationic region, is not involved in the P␥ interaction with T␣. Thus, it is concluded that the polycationic region in P␥ may be divided into at least two subdomains, and a subdomain containing arginines 33 and 36 appears to be involved in the interaction with T␣, but not in the interaction with P␣␤.
As summarized in the Introduction, various methods have been applied to identify specific domains in P␥. Proteolytic digestion of P␥ has also been applied to identify a specific domain in P␥ (25). Although these methods are potentially useful in identifying specific domains in proteins, they have also serious problems. For example, one cannot be sure that conformation of the peptide corresponding to the specific region is the same or similar to that of the region in the protein.
Site-directed mutagenesis of the specific residues does not necessarily change only the conformation of the region in which these residues are involved. Moreover, deletion of the specific residues does not necessarily delete functions of the specific residues. We have already shown that deletion of the carboxylterminal residues of P␥ reduces not only its inhibitory activity of cGMP hydrolysis but also its ability to interact with P␣␤ (15). Therefore, we seek a method to identify specific residues in the P␥ polycationic region under more physiological conditions. In this study we have used ADP-ribosylation to identify two arginines in the polycationic region. The ADP-ribosylation was carried out by endogenous ADP-ribosyltransferase under physiological conditions. Then, we used peptides and site-directed mutagenesis to confirm data obtained by the ADP-ribosylation.
In a system reconstituted by exogenous P␥ and P␥-and transducin-depleted ROS membranes, the maximum level of P␥ ADP-ribosylation is estimated about 20% by the measurement of P␥ ADP-ribosylation after SDS-gel electrophoresis (Fig. 2). Although we do not think that the conclusions in this study are affected by the low level of ADP-ribosylation, we try to specify the reasons for such a low level of ADP-ribosylation. One possible interpretation is that P␥ complexed with P␣␤ may be a better substrate for ADP-ribosyltransferase, especially in membranes, because high P␥ ADP-ribosylation (about 50%) was detected in native membranes (Fig. 1). We note that all P␥ appeared to be complexed with P␣␤ under conditions described in Fig. 1; however, ϳ95% of added P␥ was estimated to be free under the conditions described in Fig. 2. Moreover, we anticipate that an activator of ADP-ribosyltransferase may be present in these membranes because ADP-ribosyltransferase might be regulated by several proteins. 3 Another possible interpretation is that the apparent incorporation of ADP-ribose to P␥ measured after SDS-gel electrophoresis may be underestimated. In this study, the pH values of the separating gel buffer and the running buffer are 8.8 and 8.4, respectively. SDS-gel electrophoresis was carried out without a cooling system. It is possible that these conditions accelerate the release of ADPribose from P␥. This speculation is supported by the observation that ADP-ribose-arginine linkage was sensitive to in buffers (pH 8.0 -8.5) as described above. Zolkiewska and Moss (43) also described possible breakdown of ADP-ribose-arginine linkage during electrophoresis.
We have shown previously that P␥ complexed with GTP⅐T␣ is phosphorylated by a kinase; however, P␥ complexed with P␣␤ is not a substrate for the kinase (36,44). The phosphorylation of P␥ inverts the relative affinities of P␥ to GTP⅐T␣ and to P␣␤, and the change in the relative affinities may function in the turnoff mechanism of GTP⅐T␣-activated PDE without GTP hydrolysis. In this study we have shown that P␥ complexed with P␣␤, but not with T␣, is ADP-ribosylated by arginine-ADP-ribosyltransferase in ROS membranes. We have utilized the P␥ ADP-ribosylation as a tool to identify arginines in the polycationic region which are involved in the interaction with T␣. However, the physiological significance of the P␥ ADPribosylation in phototransduction remains unsolved. We anticipate that P␥ ADP-ribosylation may control phototransduction through regulation of P␥ interaction with specific proteins involved in phototransduction. The information obtained in this study will also be useful to reveal the physiological significance of the P␥ ADP-ribosylation.