Phosphorylation by Cyclin-dependent Protein Kinase 5 of the Regulatory Subunit of Retinal cGMP Phosphodiesterase

Retinal cGMP phosphodiesterase (PDE) is regulated by Pγ, the regulatory subunit of PDE, and GTP/Tα, the GTP-bound α subunit of transducin. In the accompanying paper (Matsuura, I., Bondarenko, V. A., Maeda, T., Kachi, S., Yamazaki, M., Usukura, J., Hayashi, F., and Yamazaki, A. (2000) J. Biol. Chem. 275, 32950–32957), we have shown that all known Pγs contain a specific phosphorylation motif for cyclin-dependent protein kinase 5 (Cdk5) and that the unknown kinase is Cdk5 complexed with its activator. Here, using frog rod photoreceptor outer segments (ROS) isolated by a new method, we show that Cdk5 is involved in light-dependent Pγ phosphorylation in vivo. Under dark conditions only negligible amounts of Pγ were phosphorylated. However, under illumination that bleached less than 0.3% of the rhodopsin, ∼4% of the total Pγ was phosphorylated in less than 10 s. Pγ dephosphorylation occurred in less than 1 s after the light was turned off. Analysis of the phosphorylated amino acid, inhibition of Pγ phosphorylation by Cdk inhibitors in vivo and in vitro, and two-dimensional peptide map analysis of Pγ phosphorylated in vivo and in vitro indicate that Cdk5 phosphorylates a Pγ threonine in the same manner in vivo and in vitro. These observations, together with immunological data showing the presence of Cdk5 in ROS, suggest that Cdk5 is involved in light-dependent Pγ phosphorylation in ROS and that the phosphorylation is significant and reversible. In an homogenate of frog ROS, PDE activated by light/guanosine 5′-O-(3-thiotriphosphate) (GTPγS) was inhibited by Pγ alone, but not by Pγ complexed with GDP/Tα or GTPγS/Tα. Under these conditions, Pγ phosphorylated by Cdk5 inhibited the light/GTPγS-activated PDE even in the presence of GTPγS/Tα. These observations suggest that phosphorylated Pγ interacts with and inhibits light/GTPγS-activated PDE, but does not interact with GTPγS/Tα in the homogenate. Together, our results strongly suggest that after activation of PDE by light/GTP, Pγ is phosphorylated by Cdk5 and the phosphorylated Pγ inhibits GTP/Tα-activated PDE, even in the presence of GTP/Tα in ROS.

with GTP/T␣, and the region is masked when P␥ is complexed with GTP/T␣. The same kind of inhibition was also observed in the ADP-ribosylation of P␥, because the P␥ ADP-ribosylation site (arginines 33 or 36) is masked when P␥ is complexed with GTP/T␣ (18,19). It is very likely that the phosphorylation of threonine 35 in P␥ occurs when P␥ is complexed with P␣␤. We have shown that arginine 33 or 36 in P␥ is ADP-ribosylated when P␥ is complexed with P␣␤ (18).
In contrast to these P␥ phosphorylations, P␥ phosphorylation by P␥ kinase appears to be light-dependent and thus can be brought into agreement with in the current model of phototransduction (16,17). In the phosphorylation, P␥ complexed with GTP/T␣ is the best substrate for P␥ kinase, and the P␥ phosphorylation is dependent upon GTP in ROS membranes. These results indicate that the P␥ phosphorylation occurs after PDE activation. Threonine 22 in P␥ is phosphorylated. The phosphorylated P␥ loses its affinity to GTP/T␣, but gains a 10ϳ15 times higher ability to inhibit PDE activity than that of nonphosphorylated P␥. Thus, the phosphorylated P␥ more effectively inhibits GTP/T␣-activated PDE than nonphosphorylated P␥, and the inhibition occurs even in the presence of GTP/T␣. These observations imply that 1) the P␥ phosphorylation is probably involved in the recovery phase of phototransduction to the dark state, 2) after activation of PDE, GTP/T␣ may interact with another effector and the interaction may be associated with mechanisms for the recovery of phototransduction, and 3) the lifetime of GTP/T␣-activated PDE can be regulated by the P␥ phosphorylation when the P␥ phosphorylation functions.
In this series of experiments, we showed that Cdk5 phosphorylates P␥ complexed with GTP/T␣ in vitro (in the accompanying paper (20)) and in vivo (in this paper). In the accompanying paper (20), we have shown that P␥ preserves an amino acid sequence required for the phosphorylation by Cdk5 and that the P␥ kinase is Cdk5 complexed with p35, a Cdk5 activator. We have also demonstrated that recombinant Cdk5/p35 phosphorylates P␥ in a GTP␥S-dependent manner in ROS membranes, suggesting that Cdk5 is involved in the phosphorylation of P␥ complexed with GTP/T␣. In the present study, we link these observations with light-dependent P␥ phosphorylation in vivo (21). Using frog photoreceptor outer segments isolated by a new method, we show that Cdk5 is involved in the light-dependent P␥ phosphorylation in vivo, that the P␥ phosphorylation is significant and reversible, and that the P␥ phosphorylation and dephosphorylation are rapid enough to be involved in the recovery phase of phototransduction. Moreover, in an homogenate of photoreceptor outer segments, the phosphorylated P␥ inhibits light/GTP␥S-activated PDE, even in the presence of GTP␥S/T␣. These observations suggest that the P␥ phosphorylation verified in the in vitro system (16,17,20) functions similarly in functional, isolated photoreceptor outer segments.
Protein Preparation-Recombinant bovine Cdk5/p35 was prepared by a Mono S column (22). Cdk5 and p35 cloned from bovine retina have the same cDNA sequences as those of bovine brain proteins (20). We note that recombinant p35 is expressed as a ϳ25-kDa protein, as described previously (22)(23)(24). The ϳ25-kDa protein (p25) is a truncated form of p35, but p25 has been shown to activate Cdk5. P␥ kinase was isolated from a soluble fraction of frog ROS using a Mono Q column (16). Frog P␥ (10) and its phosphorylated form (16) were isolated as described. Recombinant bovine P␥ was prepared as described previously (18). Frog GDP/T␣ was purified as described previously (9,10). GTP␥S/T␣ was prepared from GDP/T␣ by using GTP␥S, urea-treated ROS membranes, and T␤␥, as described previously (9). To prepare P␥ complexed with GTP␥S/T␣ or GDP/T␣, these T␣s were incubated with equimolar concentration of P␥ at 4°C overnight, and these complexes were isolated by gel-filtration columns.
Isolation of Frog Photoreceptor Outer Segments and Determination of the Level of P␥ Phosphorylation in the Outer Segments-Frogs were fully dark-adapted (Ͼ12 h) at room temperature. Before bleaching samples, all manipulations were done under infrared light. Photoreceptor outer segments were isolated as described previously (19). Briefly, the frontal hemisphere of an eyeball was removed by a razor blade, and the resulting eye-cup was vertically cut in half. The half-eye-cup was put on two layers of curved filter paper (Toyo Filter Paper 5B, 3 ϫ 6 cm). After its vitreous body was soaked by the filter papers, the retina layer adhered to the filter paper. The retina could be peeled off from the eye-cup when the curved filter paper was flattened. The pigment epithelium layer stayed on the eye-cup. The retina was covered with 100 l of Ringer's solution (105 mM NaCl, 2.5 mM KCl, 1.2 mM MgCl 2 , 10 mM HEPES (pH 7.5), 2 mM taurine, and 5 mM glucose) containing [ 33 P]phosphorus (ϳ1 mCi/ml) and incubated for 30 min under O 2 /CO 2 (95:5) atmosphere. [ 33 P]Phosphorus was used to avoid illumination of rhodopsin by the Cerenkov effect of [ 32 P]phosphorus. The pH value of the incubation medium was exactly adjusted to 7.50 by the addition of 0.1 N NaOH. After rinsing with Ringer's solution, the retina was placed on a flat end of a plastic plunger, and the excess Ringer's solution was removed by dry filter paper. The retina on the plastic plunger was exposed to white light (2% rhodopsin bleached/min) for the indicated periods and quickly frozen by attaching its photoreceptor outer segment layer to the surface of a copper block pretreated with liquid N 2 . The retina was moved into cold acetone (Ϫ20°C, 100 ml) containing 5 g of molecular sieve (4A 1/16) and incubated (Ϫ20°C, 3 days) with acetone. The acetone was replaced each day. Acetone was removed by decantation, and dehydrated retinas were dried under vacuum. A piece of an adhesive tape (Scotch 3M) was attached to the outer segment surface of the retina layer. After the backing filter paper was removed, another piece of tape was attached to the neural retina layer. By pulling these two tapes apart, the outer segment layer and the neural retina layer, including the photoreceptor inner segments, were separated. The outer segment layer attached to the tape is observed by light microscopy (ϫ 500) (Fig. 1A). The purity of the outer segments will be described later. The outer segment layer was solubilized with 100 l of buffer A (1% Triton X-100, 1% deoxycholic acid, 150 mM NaCl, 50 mM Tris/HCl (pH 7.5), 1 mM PMSF, 10 g/ml aprotinin, 10 g/ml leupeptin, 10 g/ml pepstatin A, and 50 nM okadaic acid) containing 1% SDS, heated at 95°C for 5 min, and diluted with 900 l of buffer A. Insoluble materials were spun down by ultracentrifugation. The protein concentration was determined by the bicinchoninic acid protein assay method (25) after proteins were precipitated with deoxycholic acid/trichloroacetic acid. P␥ and phosphorylated P␥ in each sample (1.0 mg of protein) were immunoprecipitated using a P␥-specific antibody and further purified by SDS-PAGE. Radioactive bands corresponding to phosphorylated P␥ were detected by using an image analyzer and were cut out. Phosphoamino acid analysis of the P␥ was carried out as described previously (16). 33 P-Labeled spots were detected on an image analyzer. Radioactive spots were recovered, and their radioactivities were measured.
The purity of photoreceptor outer segments isolated was investigated using antibodies against MAP kinase, which are believed not to be present in outer segments (20). Photoreceptor outer segments and the neural retinal layer were solubilized in 200 l of SDS-sample buffer containing 70 mM DTT and heated at 95°C for 5 min. Proteins (20 g) in these samples were isolated by SDS-PAGE, and MAP kinase was detected by Western blot. The protein concentration was measured as described previously (25).
Determination of Cdk5 Localization in Frog Retinas-Fresh frog retina was immediately fixed with 4% paraformaldehyde in 0.13 M phosphate buffer (pH 7.4) for 2 h at 4°C. After washing with the same buffer (ϫ3), the retina was equilibrated with 30% sucrose solution. Cryosections (14 m thick) were prepared with a Cryostat (Leica CM3050 Bensheim, Germany), and mounted on glass slides. These specimens were blocked with the phosphate-buffered saline buffer containing 1% (w/v) bovine serum albumin and 0.5% Triton X-100 for 30 min at room temperature. Sections were incubated with a Cdk5 antibody at the IgG concentration of 1 ng/ml for 2 h. For controls, the same concentration of the antibody was mixed with the peptide antigen (20 ng/ml) prior to application. Specimens were incubated with a secondary antibody, alkaline phosphatase-conjugated goat anti-rabbit IgG (15 mg/ ml), for 1 h. These specimens were washed with the phosphate buffer (ϫ3). Finally, antibody binding sites were visualized with 5-bromo-4chloro-3Ј-indolyl phosphate p-toluidine salt and nitro blue tetrazolium chloride.
Immunological Detection of Phosphorylated P␥-Before illumination of samples, all manipulations were done under infrared light. Retinas were isolated from frog eye-cups as described above. These retinas were incubated in Ringer's solution (20 min). After exposure to white light (2% rhodopsin bleached/min) for indicated times, these retinas were quickly frozen as described above. As a control, all procedures were carried out without light. The photoreceptor outer segment layer was isolated from dried retina as described above and then solubilized with 50 l of 1% SDS containing 100 mM DTT. Solubilized outer segment sample was diluted by 10-fold with buffer B (2% Triton X-100, 1% Pharmalyte (pH 8 -10.5), 0.5% Bio-Lyte (pH 5-7), 0.5% Bio-Lyte (pH 6-8), 1 mM PMSF, 10 g/ml aprotinin, 10 g/ml leupeptin, 10 g/ml pepstatin A, 100 nM okadaic acid, and 9.2 M urea in final concentrations). P␥ and phosphorylated P␥ in the solubilized outer segments (50 g protein) were isolated by two-dimensional gel electrophoresis using Immobiline DryStrip gels, consisting of isoelectric focusing (pH 6 -11) in the first dimension and SDS-gradient gel (10 -20%) electrophoresis in the second dimension. P␥ and phosphorylated P␥ were blotted on nitrocellulose membranes, and membranes were blocked by 5% milk in Tris-buffered saline containing 0.1% Tween 20. A P␥-specific antibody diluted in the same blocking buffer was used to identify P␥, and the bound antibody was detected by a chemiluminescence detection kit. After development of x-ray films, P␥ spots were scanned by Paragon 1200A3 ProScanner and relative density (mm 2 ϫ OD) was calculated by Molecular Analyst software (Bio-Rad). The location of phosphorylated P␥ in gels was confirmed using 33 P-phosphorylated P␥.
Detection of Effects of Cdk Inhibitors on P␥ Phosphorylation in Vivo and in Vitro-In the in vivo system, Cdk inhibitor, olomoucine (100 M), or roscovitine (50 M) were applied to retinas incubated in Ringer's solution containing [ 33 P]phosphorus. Other experimental conditions were the same as those described above. An inactive analogue of olomoucine, iso-olomoucine (100 M), was used as a control. Isolation of P␥ and phosphoamino acid analysis of phosphorylated P␥ were carried out as described above. Effects of these inhibitors on P␥ kinase and recombinant Cdk5/p35 were also measured using P␥ phosphorylation in vitro (16,20).

Two-dimensional Peptide Map Analysis of P␥ Phosphorylated in Vitro and in Vivo-Phosphopeptide maps of P␥ phosphorylated in vitro
and in vivo were compared. As P␥ phosphorylated in vitro, frog P␥ (10 g) was phosphorylated by using ϳ10 Ci of [␥-33 P]ATP and P␥ kinase, as described previously (16,20). As P␥ phosphorylated in vivo, after incubation of retinas in Ringer's solution containing [ 33 P]phosphorus, P␥ was phosphorylated by 10-min light exposure (20% rhodopsin was bleached), as described above. These phosphorylated P␥s were isolated by immunoprecipitation using a P␥-specific antibody and SDS-PAGE. P␥ radioactive spots were identified by autoradiography and cut out. Extracted P␥s were digested with trypsin, and the resulting peptides were analyzed using two-dimensional peptide map analysis as described previously (26). Radioactive spots were detected by an image analyzer (BAS2000, Fuji Film).

Measurement of Effects of Nonphosphorylated and Phosphorylated P␥s on Light/GTP␥S-activated PDE-Dark-adapted intact frog ROS
was prepared from six frogs by Percoll density gradient centrifugation (13). The intact ROS was suspended in buffer C (65 mM KCl, 35 mM NaCl, 10 mM HEPES (pH 7.8), 2 mM MgCl 2 , 0.2 mM PMSF, 5 g/ml aprotinin), and its aliquot (10 l) was divided into siliconized plastic tubes. Each tube was tightly sealed with an aluminum foil and stored at Ϫ90°C. Each frozen ROS was thawed just before use and sheared by passing through a siliconized fine pipette tip. The ROS homogenate (final concentration of rhodopsin, 2.5 M) was added to 180 l of buffer C containing 0.45 mM GTP␥S and 1 mM BAPTA in a lucent glass chamber with magnetic stirrer. PDE activity was assayed by a pH electrode (Beckman S506A) in the presence or absence of 300 nM P␥, GDP/T␣/P␥, GTP␥S/T␣/P␥, and GTP␥S/T␣ and phosphorylated P␥. Reaction was started by adding 4 mM cGMP. To stimulate PDE activity, a light flash that bleached 0.003% rhodopsin was given at time 0. After each measurement, continuous white light was applied to see a maximum reaction rate.
Analytical Methods-P␥ phosphorylation by P␥ kinase (16) and recombinant Cdk5/p35 (20) was performed as described. Protein concentration was assayed with bovine serum albumin as a standard (27). Concentration of frog P␥ (10,20) and recombinant P␥ (20) was measured as described. 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.

Localization of Cdk5 in Frog
Retinas-In pervious studies (16,17,20), we showed that Cdk5 is in a soluble fraction of frog ROS, that P␥ complexed with GTP␥S/T␣ is the best substrate for Cdk5, and that P␥ phosphorylation by Cdk5 is dependent upon GTP␥S in frog ROS membranes. We also showed that all known P␥s have a special phosphorylation motif for prolinedirected kinase, including Cdk5 (20). These biochemical observations strongly suggest that Cdk5 is in photoreceptor outer segments. In this study, we carried out two experiments to confirm the localization of Cdk5 in photoreceptor outer segments: immunodetection of Cdk5 in outer segments isolated by a new method and an immunohistochemical search for Cdk5 in retina. In the first experiment, photoreceptor outer segments were separated from other neural retinal layers, including photoreceptor inner segments (Fig. 1). Light microscopy (ϫ500) showed that outer segment layer contains yellow rod-shape cells (Fig. 1A). The yellow color detected appears to indicate the presence of bleached rhodopsin. The neural retinal layer also contained similar rod-shape cells; however, the color of these cells was white (data not shown). We note that pigment epithelium cells were not contaminated in this preparation, because no black cells were detected in the preparation. To check the purity of the outer segment layer isolated, first we compared the protein profile in the outer segments with that in neural retinal layers. We found that protein profiles are different on SDS gels (Fig. 1B, lanes 1). In addition, the purity of the outer segment layer was also examined by comparing of the contents of MAP kinase in the outer segment layer with that in the other neural retinal layers, because, using an immunohistochemical method, we have already suggested that MAP kinase is present in the neural retinal layers, but not in the outer segment layer (20). We found that the immunological signal of MAP kinase was clearly observed in the neural retinal layers, but not in the outer segment layer (Fig. 1B, lanes 3), indicating that MAP kinase contents in the outer segment preparation are not enough to be detected by Western blotting. These observations suggest that the outer segment layer is reasonably separated from the other neural retinal layers. Under these conditions, Cdk5 was clearly observed in both outer segment and other neural retinal layers (Fig. 1B, lanes 2). This observation strongly suggests that Cdk5 is present in both the photoreceptor outer segment layer and the other neural retinal layers in frog retina. This observation also shows that Cdk5 contents in the outer segment layer appears to be less than that in the neural retinal layers. We note that the specificity and sensitivity of the Cdk5 antibody have already been shown in the accompanying paper (20).
In the immunohistochemical search for Cdk5 in frog retina, the strongest immunological signal was observed in the inner plexiform layer (Fig. 2B). Other strong signals were in the outer plexiform layer and at the interface between the inner segments and the outer nuclear layer. These observations are consistent with previous data showing the location of Cdk5 in rat retina (28). The immunological signal in the outer segment layer, on the contrary, was less intense than that of those layers. However, we believe that this signal is significant for the following reasons. 1) In the same photoreceptor cell, the signal in the outer segment layer is stronger than the signal in the outer nuclear layer; 2) in retinal cells, the signal in the outer segment layer is clearly stronger than that in the inner nuclear layer; and 3) the background signal seen between outer segments is similar to the signal in the control retina, and the signal in the outer segments is much stronger than the background signal. The similar immunological signals were also observed in bovine retinal cells (data not shown). Based on three different observations (biochemical observations in previous studies (16,17,20), detection of Cdk5 in an homogenate of highly purified outer segments (Fig. 1), and this immunohistochemical observation), it is reasonable to conclude that Cdk5 is present in photoreceptor outer segments.
Light-dependent P␥ Phosphorylation in Vivo-In previous studies (16,17,20), we have shown that P␥ complexed with GTP␥S/T␣ is the best substrate for P␥ kinase (Cdk5) and that GTP␥S is required for P␥ phosphorylation by Cdk5 in photoreceptor outer segment membranes. These observations imply that P␥ phosphorylation is light-dependent in photoreceptors, because binding of GTP␥S to T␣ is completely dependent upon illuminated rhodopsin (1, 7). Indeed, the light-dependent P␥ phosphorylation was detected in vivo (21). We combined and extended these observations. After incubation with [ 33 P]phosphorus, retinas were illuminated under various light conditions, quickly frozen, and dehydrated. The photoreceptor outer segment layer was then isolated from these retinas as described above. P␥ was isolated from these photoreceptor outer segments by a P␥-specific antibody and SDS-PAGE, and subsequently phosphorylated amino acids in the P␥ were identified and their radioactivities were measured. Under dark conditions the phosphorylated amino acid was barely detected (Fig.  3A), and only less than 0.5% of the total P␥ was found to be phosphorylated (Fig. 3C). However, a threonine residue in P␥ was clearly phosphorylated in illuminated photoreceptors (Fig.  3A). Two phases of phosphorylation were observed. After an initial rapid phosphorylation, the phosphorylation increased gradually during illumination (Fig. 3B). In the rapid phosphorylation, the P␥ phosphorylation was detected when 0.03% of rhodopsin was bleached (1 s). During the slow phosphorylation, 10 Ϯ 4% of the total P␥ (n ϭ 5) was phosphorylated after bleaching of 10% of rhodopsin (Fig. 3C), and 16 Ϯ 6% of the total P␥ (n ϭ 5) was phosphorylated after bleaching of 20% of rhodopsin (data not shown). Thus, as the rapid phosphorylation, ϳ4% of the total P␥ was estimated to be phosphorylated after less than 0.3% of the rhodopsin was bleached in less than 10 s (Fig. 3A). These observations indicate that P␥ phosphorylation in photoreceptor outer segments is light-dependent and significant and that the P␥ phosphorylation is fast enough to be involved in the recovery phase of phototransduction. When the light was turned off, the P␥ was dephosphorylated rapidly (Fig.  3, A and B). The P␥ dephosphorylation could be observed in less than 1 s. Together, these results indicate that the P␥ phosphorylation is reversible and that the dephosphorylation is also fast.
We note that incorporation of the radioactivity into a P␥ serine residue was sometimes detected in the in vivo system, especially when the level of the P␥ threonine phosphorylation was increased by high bleaching (Fig. 3A). The incorporation was weak because only less than 10% of the radioactivity incorporated into the threonine residue was observed in the serine residue. As described below, we will show that Cdk5 is involved in the light-dependent phosphorylation of the P␥ threonine. However, the serine phosphorylation seems not to be due to Cdk5, because only threonine 22 in P␥ (frog and bovine) is phosphorylated by Cdk5 in vitro (16,20). Since we do not know the full amino acid sequence of frog P␥, it is also possible that the P␥ contains another Cdk5 phosphorylation site, including a serine residue, and the phosphorylation of the serine was detected in vivo, but not in vitro, because of structural hindrance. However, this possibility is very slim, because P␥ appears not to have a rigid conformation, and such structural hindrance seems not to be present in P␥ in vitro (20). We also note that the FIG. 1. Isolation and purity of photoreceptor outer segment layer from frog retinas. A, isolation. Frog retinas were isolated and incubated in Ringer's solution as described. Then retinas were quickly frozen by attaching their photoreceptor outer segment layer to the surface of a copper block pretreated with liquid N 2 . Retinas were then dehydrated and dried as described. A piece of a Scotch 3M tape was attached to the outer segment surface of the retina layer, and another piece of tape was attached to the neural retina layer. By pulling these two tapes apart, the outer segment layer was isolated from the neural retina layers including the photoreceptor inner segments. The outer segment layer was observed by light microscopy (ϫ 500). The picture shown was taken on a dark background. The cleft shown was formed when dehydrated retina was dried in vacuo. B, purity of the outer segments prepared by the method. In addition to Cdk5, the presence of MAP kinase in the outer segment layer (a) and other neural retinal layers (b) was investigated. After solubilization of these preparations in 100 l of SDS-sample buffer containing 70 mM DTT, proteins (20 g) were separated by SDS-PAGE and detected by Coomassie Blue staining or Western blotting. Lanes 1, proteins stained by Coomassie Blue; lanes 2, Cdk5 detected by Cdk5 antibody; and lanes 3, MAP kinase detected by MAP kinase antibody.

FIG. 2. Localization of Cdk5 in frog retina.
Cdk5 in fresh frog retina was localized using a Cdk5 antibody and alkaline phosphataseconjugated anti-rabbit IgG. For a control, the antibody treated with the peptide antigen prior to application was used. The cell layers seen are: OS, outer segments; IS, inner segments; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; and IPL, inner plexiform layer. As shown in the accompanying paper (20), the specificity of the antibody was enough to distinguish Cdk5 in an homogenate of retina, and the antibody was sensitive to frog Cdk5. involvement of PKC and PKA in the serine phosphorylation is unlikely, because these protein kinases have been shown to phosphorylate only threonine 35 in P␥ in vitro (14,15). It is also possible that the frog P␥ contains a serine phosphorylation site for these protein kinases, and the site could not be phosphorylated by structural hindrance under the in vitro conditions. However, the possibility was small, because P␥ seems not to have a rigid conformation, as described above. In the case of PI-stimulated kinase (13), we suggested that threonine 35 or serine 40 in frog P␥ was phosphorylated in vitro. However, the possibility of serine 40 phosphorylation is also small, because under bleached conditions the site appears to be masked by GTP/T␣ in vivo, as described in the Introduction. At present, the mechanism and function of the serine phosphorylation are unknown.
Effects of Cdk Inhibitors on P␥ Phosphorylation in Vitro and in Vivo-We used Cdk inhibitors, olomoucine, and roscovitine to determine whether P␥ phosphorylation in vivo is due to a Cdk. These Cdk inhibitors have been used in vitro (0.2-10 M) and in vivo (10 -100 M) (29 -31). We found that under the in vitro conditions both olomoucine and roscovitine inhibited P␥ phosphorylation by frog P␥ kinase or by recombinant bovine Cdk5/p35 with 50% inhibition ϳ2 and ϳ7 M, respectively (Fig.  4A). Iso-olomoucine, an inactive analog of olomoucine, did not inhibit the P␥ phosphorylation. We also found that 100 M olomoucine completely inhibited the light-dependent phosphorylation of P␥ threonine in vivo, but iso-olomoucine did not (Fig.  4B). Moreover, 50 M roscovitine drastically reduced the P␥ threonine phosphorylation. These observations strongly suggest that a Cdk is involved in the light-dependent P␥ phosphorylation.
PI-stimulated kinase, PKC or PKA, have been shown to phosphorylate P␥ in vitro (13)(14)(15). However, it should be emphasized that involvement of these protein kinases in the lightdependent phosphorylation of P␥ threonine (Figs. 3 and 4) is very unlikely for the following reasons. 1) As summarized in the Introduction, these protein kinases appear not to function in the light-dependent P␥ phosphorylation in vivo, because the amino acid(s) phosphorylated by these kinases seem to be masked by GTP/T␣ when P␥ interacts with GTP/T␣ (15,18,19). 2) It is possible that frog P␥ could have another threonine, other than threonine 35, as a phosphorylation site for these protein kinases, and the site would be available in vivo, but not in vitro, because of structural hindrance. However, this possibility is small, because P␥ appears not to have a ridged conformation, as suggested (20). 3) The Cdk inhibitors used are very specific to Cdks (31). Very high concentration of these inhibitors is required to inhibit PKC and PKA in vitro; IC 50 values for various isozymes of PKC are Ͼ100 M roscovitine and Ͼ800 M olomoucine, and IC 50 values for PKA (bovine heart) are Ͼ1,000 M roscovitine and Ͼ2,000 M olomoucine (31). In addition, similar high concentration of these inhibitors is also required for the inhibition of other protein kinases such as cGMP-dependent protein kinase (bovine tracheal smooth muscle, IC 50 Ͼ1,000) and casein kinase 2 (rat liver, IC 50 Ͼ2,000) (31). Al-

FIG. 3. Light-dependent P␥ phosphorylation in vivo.
A, light dependence. Frog retinas were incubated in a Ringer's solution containing [ 33 P]phosphorus under dark conditions. Under various light conditions, retinas were illuminated and rapidly frozen. Then photoreceptor outer segment layers were isolated from lyophilized retinas. For the illumination, light was adjusted to bleach 2% rhodopsin/min. Outer segment layers obtained from lyophilized retinas were solubilized with 100 l of buffer A containing 1% SDS (95°C) and diluted with 900 l of buffer A. P␥ in each sample (1.5 mg) was immunoprecipitated using a P␥-specific antibody. P␥ in the precipitate was further purified by SDS-PAGE, and phosphoamino acid analysis of the P␥ was carried out. P-Ser, phosphoserine; P-Thr, phosphothreonine; P-Tyr, phosphotyrosine; D, dark; L10", light 10 s; L60", light 60 s; L60"D10", light 60 s and then dark 10 s; L60"D60", light 60 s and then dark 60 s. B, radioactivity of the phosphorylated threonine residue isolated in A. The radioactivity was measured after scratching radioactive spots of the threonine residue. C, immunological detection of light-dependent P␥ phosphorylation. Retinas were illuminated for 5 min under the same conditions described in A. As a control, retinas were not illuminated. Then photoreceptor outer segments were isolated from these retinas. P␥s in these photoreceptor outer segments (50 mg) were isolated by two-dimensional gel electrophoresis (IEF, isoelectric focusing (pH 6 -11), in the first dimension, and SDS-PAGE, polyacrylamide gradient 10 -20%, in the second dimension) and detected using the P␥-specific antibody with a chemiluminescence detection kit. After development of x-ray films, P␥ spots were scanned, and relative densities were calculated. The location of phosphorylated P␥ in gels, indicated by a open arrowhead, was identified using P␥ phosphorylated by recombinant Cdk5/p35 in vitro. a, P␥ in dark outer segments; b, P␥ in bleached outer segments.

FIG. 4. Effect of Cdk inhibitors on P␥ phosphorylation in vitro and in vivo.
A, in vitro. P␥ (2 g) was phosphorylated with 200 ng P␥ kinase (E, Ⅺ, ‚) or 10 ng of recombinant Cdk5/p35 (q, f, OE) in 30 ml of the reaction mixture in the presence of various concentrations of olomoucine (E, q) and roscovitine (‚, OE). As a control for olomoucine, iso-olomoucine was used (E, q). P␥ kinase and recombinant Cdk5/p35 were prepared by using a Mono Q column and a Mono S column, respectively. One-hundred percent indicates the level of P␥ phosphorylation without an inhibitor. B, in vivo. After incubation of frog retinas in Ringer's solution containing [ 33 P]phosphorus in the presence or absence of olomoucine (100 M), iso-olomoucine (100 M), and roscovitine (50 M). P␥ phosphorylation was stimulated by illumination (1 min). Phosphoamino acid analysis of the phosphorylated P␥ was carried out as described. P-Ser, phosphoserine; P-Thr, phosphothreonine; P-Tyr, phosphotyrosine.
though the effect of these inhibitors on frog kinases is unknown, it is unlikely that frog PI-stimulated kinase, PKC, PKA, or unknown protein kinases, other than Cdks, could be inhibited by less than 100 M amounts of these Cdk inhibitors in vivo. Especially, it is very unlikely that any frog protein kinase was completely inhibited by 100 M olomoucine in vivo (Fig. 4B).
Evidence for the Involvement of Cdk5 in P␥ Phosphorylation in Vivo-As described above, a Cdk is involved in the P␥ phosphorylation in vivo. To identify the Cdk as Cdk5, two-dimensional phosphopeptide maps of both P␥ phosphorylated by P␥ kinase (Cdk5) in vitro and P␥ phosphorylated in illuminated retina (10-min illumination) were compared (Fig. 5). We found that only one kind of the radioactive peptide was produced from these P␥s and that the peptide derived from P␥ phosphorylated in vivo was found in the same location as that of P␥ phosphorylated in vitro. Together with data showing that only threonine 22 is phosphorylated by Cdk5 in vitro (16,20), these observations indicate that the same threonine residue in P␥ is phosphorylated in vivo and in vitro. These observations imply that the Cdk involved in the light-dependent P␥ phosphorylation in vivo is Cdk5. Since under our conditions, photoreceptors were illuminated for 10 min (ϳ20% rhodopsin illumination), these results also indicate that the same threonine residue is phosphorylated in these two phases of P␥ phosphorylation. These observations also strongly support our previous conclusion that protein kinases known to phosphorylate P␥ threonine 35 are not involved in the light-dependent phosphorylation of P␥.
Effects of Phosphorylated P␥ on Light/GTP␥S-activated PDE-In previous studies (16,17), using systems reconstituted with isolated proteins, we showed that P␥ phosphorylated by P␥ kinase (Cdk5) more effectively inhibits GTP␥S/T␣-activated PDE than nonphosphorylated P␥. This is because the phosphorylated P␥ loses its affinity to GTP␥S/T␣, but gains a 10ϳ15 times higher ability to inhibit GTP␥S/T␣-activated PDE than that of nonphosphorylated P␥. In this study, we investigated whether these phenomena were also observed in an homogenate of frog photoreceptor outer segments, a system containing all proteins of outer segments. We checked the effects of nonphosphorylated and phosphorylated P␥s, alone or complexed with T␣, on the time course of light/GTP␥S-dependent activation of PDE (Fig. 6). We found that P␥ complexed with GDP/T␣ could not inhibit the PDE activity; however, the same concentration of P␥ alone inhibited the PDE activity (Fig. 6A). This implies that GTP hydrolysis is not sufficient for the deactivation of light/GTP-activated PDE in a frog ROS homogenate and that P␥ must be released from GDP/T␣ for the PDE deactiva-tion. This is consistent with previous results obtained in systems reconstituted by isolated proteins (10,32). We also found that P␥ complexed with GTP␥S/T␣ did not inhibit the light/ GTP␥S-activated PDE; however, P␥ phosphorylated by P␥ kinase (Cdk5) inhibited the light/GTP␥S-activated PDE, even in the presence of GTP␥S/T␣ in the homogenate (Fig. 6B). Moreover, we found that in the homogenate, lesser amounts of phosphorylated P␥ were required to inhibit light/GTP␥S-activated PDE than that of nonphosphorylated P␥ (data not shown). These results indicate that the nonphosphorylated P␥ retains its complex with GTP␥S/T␣ in the homogenate; however, after the P␥ is phosphorylated by Cdk5, the phosphorylated P␥ cannot keep its complex with GTP␥S/T␣ and inhibits effectively GTP␥S/T␣-activated PDE. We note that the effect of P␥ phosphorylation on the time course of light/GTP␥S-dependent PDE activation could not be measured by adding ATP to the homogenate, because rhodopsin in the homogenate is also phosphorylated, and the phosphorylated rhodopsin may affect the light/GTP␥S-activated PDE activity (33,34). We also note that 300 nM amounts of these P␥s, free or complexed with GDP/T␣ or GTP␥S/T␣, were used in these experiments, because by using the high concentration of P␥, we tried to clearly show that even a small portion of nonphosphorylated P␥ was not released from its complexes with GDP/T␣ or GTP␥S/T␣. Under similar conditions, previous studies used 35-200 nM P␥ to inhibit light/GTP-activated PDE activity measured with a pH electrode (35,36). DISCUSSION In previous studies (16,17), we showed that P␥ is phosphorylated by P␥ kinase in a GTP-dependent manner and that the phosphorylated P␥ loses its affinity to GTP/T␣, but gains high affinity to P␣␤ to inhibit cGMP hydrolysis by P␣␤. In the study described in the accompanying paper (20), we have demonstrated that P␥ has a special phosphorylation motif for Cdk5 and that P␥ kinase is Cdk5/p35. In the present study, using an in vivo system, we have shown the following results: 1) P␥ is phosphorylated in a light-dependent manner, 2) Cdk5 is involved in the P␥ phosphorylation, 3) the P␥ phosphorylation is FIG. 5. Two-dimensional peptide map analysis of P␥ phosphorylated in vitro and in vivo. As P␥ phosphorylated in vitro, frog P␥ was phosphorylated by P␥ kinase with [␥-33 P]ATP. As P␥ phosphorylated in vivo, frog retinas were incubated in Ringer's solution containing [ 33 P]phosphorus, and P␥ in these retinas was phosphorylated by illumination (10 min). These phosphorylated P␥s were isolated by a P␥specific antibody and SDS-PAGE. These phosphorylated P␥s were digested with trypsin, and resulting peptides were analyzed using twodimensional peptide map analysis. a, peptides from P␥ phosphorylated in vivo; b, peptides from P␥ phosphorylated in vitro; c, a mixture of peptides from P␥ phosphorylated in vivo and P␥ phosphorylated in vitro. * indicates the starting point of the two-dimensional peptide map. significant and rapid enough to be involved in the recovery phase of phototransduction, 4) dephosphorylation of P␥ is also rapid, indicating that the P␥ phosphorylation is reversible. In addition, using immunodetection in highly purified photoreceptor outer segments and an immunohistochemical search of the retina, we have detected significant signals of Cdk5 in photoreceptor outer segments. Moreover, using an homogenate of photoreceptor outer segments, we have demonstrated that the phosphorylated P␥ inhibits light/GTP␥S-activated PDE, even in the presence of GTP␥S/T␣. Although factors that regulate the P␥ phosphorylation are not identified yet, these observations clearly indicate that the P␥ phosphorylation by Cdk5 can play important roles in PDE regulation.
In this study, we used frog outer segments isolated by a new method. We have shown that the purity of the outer segments isolated by the method is high enough to show the localization of Cdk5 in the outer segments. Nishizawa et al. (37) also recently reported a novel method to isolate bovine photoreceptor cells containing outer segments and the majority of the inner segments. The concept of their method is similar to that of ours, because they used nitrocellulose membranes to attach the photoreceptor cell monolayer. Then, the photoreceptor cell layer was separated from other retinal layers attached to a filter paper. It is unknown why photoreceptor cells isolated by their method contain both outer and inner segments. We, however, speculate that there are several reasons for the difference as follows. 1) The species used to obtain photoreceptors may be critical because the form and size of photoreceptors, especially the structural strength of the ciliary connection, may be different in different species. We used frog retina and they used bovine retina. We have never tried to isolate bovine outer segments by our method. Thus, we do not know whether our method is fitted to isolate inner segment-free outer segments from bovine retina. 2) Dehydration of retina may be important to weaken the structure of ciliary connection. We dehydrate retinas before separation of outer segments from inner segments, but they did not. In any case, both methods will be useful to isolate photoreceptors and their segments and can be improved by attention to the differences in methods and results.
It should be emphasized that this series of studies does not exclude the turnoff mechanism of GTP/T␣-activated PDE by hydrolysis of GTP. We anticipate that both P␥ phosphorylation and GTP hydrolysis are involved in the deactivation of GTP/ T␣-activated PDE in phototransduction and that the P␥ phosphorylation functions under some special conditions, although the relationship between PDE turnoff by P␥ phosphorylation and by GTP hydrolysis is unknown now. However, it is clear that GTP hydrolysis is not enough to turnoff GTP/T␣-activated PDE in frog ROS membranes. After hydrolysis of GTP by T␣, P␥ is still complexed with GDP/T␣, and the P␥ complex cannot inhibit GTP/T␣-activated PDE (Ref. 10; Fig. 6A). The GDP/ T␣/P␥ complex is easily extracted and isolated from frog ROS membranes (10), indicating that the interaction between GDP/T␣ and P␥ is very tight and specific. Indeed, the GDP/ T␣/P␥ complex is easily prepared by mixing GDP/T␣ with P␥ (32). We have shown that T␤␥ is required to release P␥ from the GDP/T␣ complex in a frog system (10). However, large amounts of T␤␥ may be required for the P␥ release because of the tight interaction of P␥ with GDP/T␣. Thus, in addition of the stimulation of GTP hydrolysis by RGS9 (38), the PDE turnoff by GTP hydrolysis may require an unknown mechanism to stimulate the P␥ release from its complex with GDP/T␣. We do not exclude the possibility that phosphorylation of P␥ complexed with GDP/T␣ is involved in the unknown mechanism, because P␥ complexed with GDP/T␣ is also phosphorylated by P␥ kinase (Cdk5), and the phosphorylated P␥ appears to be released from GDP/T␣ (16).
Previous studies have suggested that GTP hydrolysis is required for the deactivation of GTP/T␣-activated PDE in in vivo conditions (39,40). However, it should be emphasized that these studies do not exclude the PDE deactivation by P␥ phosphorylation, because ATP was omitted from their experimental conditions (39), or under their conditions the P␥ phosphorylation might have been suppressed due to the use of a P␥ mutant that can not interact with GTP/T␣ (40). We also note that using hydrolysis-resistant GTP analogs, electrophysiological studies have already suggested a GTP hydrolysis-independent PDE turnoff (41,42). Under their conditions, P␥ phosphorylation by Cdk5 could be a mechanism for the GTP hydrolysis-independent PDE turnoff. We also note that the P␥ phosphorylation mechanism might be involved in previous studies reporting that PDE activity is suppressed by rhodopsin phosphorylation in an homogenate of photoreceptor outer segments.
There are many unknown questions in the P␥ phosphorylation by Cdk5. We have shown that P␥ phosphorylation by Cdk5 has two phases, rapid and slow. At present the meaning of these two phases of P␥ phosphorylation is unknown. We also need to explore the regulatory mechanism of the P␥ phosphorylation by Cdk5. Moreover, electrophysiological study is required to show that Cdk5 is involved in the GTP hydrolysisindependent PDE turnoff described previously (41,42). Cdk inhibitors we used in this study will be useful for the electrophysiological study. In the P␥ phosphorylation mechanism, a phosphatase(s) is also crucial for the rapid dephosphorylation of P␥, as shown here. In a previous study (17), we suggested that phosphatase 2A might be involved in the dephosphorylation. However, solid data were not obtained for the identification of phosphatase. Moreover, regulation of the phosphatase should also be investigated to understand the total mechanism of the P␥ phosphorylation. Thus, many experiments are needed to explore the function of the P␥ phosphorylation in phototransduction. We have shown clues for answers to these questions in this series of studies. Now it is clear that P␥ is the second protein found to be phosphorylated in a light-dependent manner in phototransduction (rhodopsin is the first protein). The light dependence appears to be due to the requirement of GTP/T␣ for the phosphorylation. Thus, the mechanism of the light dependence is different from that of rhodopsin. Studies about rhodopsin phosphorylation clearly show that light-dependent protein phosphorylation is crucial in the regulation of phototransduction such as shut off and adaptation. We anticipate the P␥ phosphorylation is also important for shut off and adaptation in phototransduction.