The Effector Enzyme Regulates the Duration of G Protein Signaling in Vertebrate Photoreceptors by Increasing the Affinity between Transducin and RGS Protein*

The photoreceptor-specific G protein transducin acts as a molecular switch, stimulating the activity of its downstream effector in its GTP-bound form and inactivating the effector upon GTP hydrolysis. This activity makes the rate of transducin GTPase an essential factor in determining the duration of photoresponse in vertebrate rods and cones. In photoreceptors, the slow intrinsic rate of transducin GTPase is accelerated by the complex of the ninth member of the regulators of G proteinsignaling family with the long splice variant of type 5 G protein β subunit (RGS9·Gβ5L). However, physiologically rapid GTPase is observed only when transducin forms a complex with its effector, the γ subunit of cGMP phosphodiesterase (PDEγ). In this study, we addressed the mechanism by which PDEγ regulates the rate of transducin GTPase. We found that RGS9·Gβ5L alone has a significant ability to activate transducin GTPase, but its affinity for transducin is low. PDEγ acts by enhancing the affinity between activated transducin and RGS9·Gβ5L by more than 15-fold, which is evident both from kinetic measurements of transducin GTPase rate and from protein binding assays with immobilized transducin. Furthermore, our data indicate that a single RGS9·Gβ5L molecule is capable of accelerating the GTPase activity of ∼100 transducin molecules/s. This rate is faster than the rates reported previously for any RGS protein and is sufficient for timely photoreceptor recovery in both rod and cone photoreceptors.

The phototransduction cascade of vertebrate photoreceptor cells is a prototypic G protein-based signal transduction cascade. It is uniquely designed to ensure both a high degree of signal amplification and fast signal termination on the physiological subsecond time scale (reviewed in Refs. [1][2][3]. Vision begins upon light excitation of rhodopsin, which activates many molecules of the photoreceptor-specific G protein, transducin. Activated transducin binds to the inhibitory ␥ subunits of its effector, PDE, 1 increasing its catalytic activity; this leads to a decrease in the intracellular cGMP level, the closure of cGMP-gated ion channels of the photoreceptor plasma membrane, and progression of the photoresponse. Photoresponse termination requires the timely inactivation of PDE, which occurs when GTP bound to transducin is hydrolyzed to GDP and inorganic phosphate by the GTPase activity of transducin. However, the intrinsic rate of transducin GTPase is much slower than the rate of photoresponse recovery. In the photoreceptor, this problem is solved by the action of a powerful mechanism of transducin GTPase acceleration (reviewed in Ref. 4). Intensive studies over the past decade have indicated that this mechanism is based on a cooperative action between two protein components, the RGS9⅐G␤5L complex (5-7) and the transducin target, PDE␥ (8 -10). In vivo experiments with transgenic mice indicate that both components are essential to ensure photoresponse recovery on a physiological time scale (11,12). It is now established that the primary catalytic role in activation of GTP hydrolysis belongs to the RGS homology domain of RGS9⅐G␤5L. This domain stabilizes the conformation ("transition state") of the transducin ␣ subunit, which is most favorable for GTP hydrolysis. PDE␥ itself does not activate transducin GTPase but rather enhances the activity of RGS9⅐G␤5L.
The goal of this study was to elucidate the mechanism by which PDE␥ enhances the activation of transducin GTPase by the RGS9⅐G␤5L complex. In principle, two possibilities could be considered. First, PDE␥ could directly contribute to GTP hydrolysis, for example, by further stabilizing the G␣ t transition state beyond the action of the RGS domain. Second, it could act by increasing the affinity between activated G␣ t and RGS9⅐G␤5L. Previous efforts to answer this question using the expressed RGS homology domain of RGS9 yielded contradictory results. McEntaffer et al. (13) reported that PDE␥ acts catalytically by increasing the maximal GTPase rate ϳ2-fold and that it does not change the apparent affinity between the RGS9 homology domain and transducin. To the contrary, Skiba et al. (14) presented evidence that the entire PDE␥ effect consists of an ϳ3-fold increase in the affinity between the RGS9 homology domain and transducin. The reason for this incongruity is unknown. Furthermore, it is unclear whether the RGS9 homology domain could serve as a good model for studying the effects of PDE␥ because the degree of transducin GTPase activation by PDE␥ observed with native RGS9⅐G␤5L within the ROS membranes is at least one order of magnitude higher than the effect observed with the domain alone (10,9). This finding implies that the physiologically relevant mechanism of PDE␥ action should be addressed with the whole RGS9⅐G␤5L complex.
In this study, we examined the mechanism by which PDE␥ potentiates the GAP activity of the native RGS9⅐G␤5L complex in bovine photoreceptor membranes. We used a combination of kinetic analysis and direct binding assays and found that the major effect of PDE␥ on the activation of transducin GTPase is an increase in the affinity between activated transducin and RGS9⅐G␤5L by Ͼ1 order of magnitude. This work provides the first explanation of how an RGS protein and a G protein effector cooperate in regulating the lifetime of G protein in its active state.

EXPERIMENTAL PROCEDURES
Purification of ROS and Various Photoreceptor Membrane Preparations-ROS were purified from frozen bovine retinas (TA & WL Lawson Co., Lincoln, NE) under infrared illumination as described (15). To obtain membranes lacking most peripheral proteins but retaining an active RGS9⅐G␤5L complex, ROS were washed under infrared illumination twice with isotonic buffer containing 100 mM KCl, 2 mM MgCl 2 , 1 mM dithiothreitol and 10 mM Tris-HCl (pH 7.5) and three times by a hypotonic buffer containing 0.5 mM EDTA and 5 mM Tris-HCl (pH 7.5). We will call this preparation "washed membranes" throughout the rest of the text. The amount of residual transducin in "washed membranes" did not exceed 1% of the original endogenous transducin content, as estimated from Coomassie-stained SDS-PAGE gels. Rhodopsin concentration in all membrane preparations was determined spectrophotometrically by measuring the difference in absorbance at 500 nm before and after rhodopsin bleaching (16).
Preparation of Transducin, PDE␥, and Peptide Encompassing PDE␥ Residues 63-87 (PDE␥-(63-87))-Transducin was purified from bovine ROS as described (17). Transducin concentration was first estimated by the Bradford method (18), using bovine serum albumin as the standard, and then the precise concentration of functionally active transducin was measured by the maximal amount of rhodopsin-catalyzed GTP␥S binding (19). Transducin concentration, as determined with GTP␥S, was used in all data analyses. Recombinant bovine wild type PDE␥ was expressed in Escherichia coli and purified as described in (20). Peptide corresponding to residues 63-87 of PDE␥ was synthesized and purified as described (10). The purity and chemical formula of PDE␥-(63-87) were confirmed by mass spectrometry and reversed-phase high pressure liquid chromatography. PDE␥ and PDE␥-(63-87) concentration was determined spectrophotometrically using a molar extinction coefficient ⑀ 280 of 7,100.
Cloning and Expression of Recombinant G␣ t -G␣ t cDNA, preceded by a nucleotide sequence encoding a hexa-histidine amino acid tag, was cut off pHis 6 G␣ t (21) with EcoRI and PstI and inserted into the baculovirus transfer vector pVL1393 digested with EcoRI and PstI. The recombinant baculovirus was generated using a BaculoGold transfection kit (PharMingen). The resulting recombinant G␣ t is N-terminally modified with the His 6 tag and therefore lacks N-terminal myristoyl moiety. Sf9 cells from a 2-liter shaking culture were collected 72 h post-infection. His 6 -G␣ t was purified from the cytoplasmic fraction using a combination of affinity chromatography on Ni-NTA-agarose (Qiagen) and anion exchange chromatography on MonoQ (Amersham Pharmacia Biotech) essentially as described for G␣ t /G␣ i1 chimeras (21). The concentration of active G␣ t able to interact with rhodopsin and G␤␥ was determined as the maximal amount of rhodopsin/G␤␥-dependent [ 35 S]GTP␥S binding (19).
GTPase Measurements-Transducin GTPase activity was determined by using either a multiple-turnover ([GTP] Ͼ [transducin]) or single-turnover ([GTP] Ͻ [transducin]) technique as described previously (22). GTPase assays were conducted at room temperature (22-24°C) in a buffer containing 10 mM Tris-HCl (pH 7.8), 100 mM NaCl, 8 mM MgCl 2 , and 1 mM dithiothreitol. Washed membranes, used as the source of both rhodopsin (to activate transducin) and RGS9⅐G␤5L, were illuminated on ice immediately before the experiments. The reaction was started by adding 10 l of [␥-32 P]GTP at the desired concentration (approximately 10 5 dpm/sample) to 20 l of washed membranes (30 M rhodopsin) reconstituted with the proteins of choice. The reaction was stopped by the addition of 100 l of 6% perchloric acid. 32 P i formation was measured with activated charcoal as described (22). All data fitting and statistical analyses were performed with Sigmaplot software, version 6.
Binding of RGS9⅐G␤5L Complex to his 6 -G␣ t Immobilized on Ni-NTA-Agarose-Washed membranes containing 10 g of rhodopsin were solubilized in 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 1 mM dithiothreitol containing 1% lauryl sucrose (Calbiochem). Insoluble material was removed by a 10-min centrifugation at 120,000 ϫ g with a Beckman Airfuge, and the supernatant was mixed with 10 l of Ni-NTA-agarose beads pre-bound to ϳ5 g of his 6 -G␣ t in 50 l of 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 0.2% lauryl sucrose. 10 mM NaF, 30 M AlCl 3 , and/or 10 M PDE␥ was added if needed. Samples were incubated on ice for 30 min with occasional shaking. The agarose beads were spun down and washed twice for 5 min with 1 ml of 20 mM Tris-HCl (pH 8.0), 300 mM NaCl, 20 mM imidazole, 0.25% lauryl sucrose, or the same buffer containing 10 mM NaF and 30 M AlCl 3 (for the AlF 4 -activated samples). Bound RGS9 was eluted from the resin with the SDS-gel sample buffer, subjected to SDS-PAGE (23), and detected by Western blot analysis using specific antibodies and an ECL substrate kit (Amersham Pharmacia Biotech).

RESULTS
The goal of this study was to evaluate the relation between PDE␥ and the native RGS9⅐G␤5L complex in regulating transducin GTPase. We took advantage of the fact that native RGS9⅐G␤5L is tightly associated with ROS membranes (6), which enabled us to deplete ROS membranes of endogenous PDE, transducin, and other extractable proteins while preserving their entire RGS9⅐G␤5L content. Previous studies indicate that a significant cooperation between RGS9⅐G␤5L and PDE␥ is observed in experiments utilizing this type of washed ROS membrane preparation (10,9).
The unique feature of studying components of the phototransduction cascade is that the G protein transducin can be supplied by a practically unlimited amount of its activated receptor, bleached rhodopsin. This makes it possible to study the effects of RGS9⅐G␤5L on transducin GTPase under steadystate conditions in which hydrolysis of G␣ t -bound GTP is followed immediately by recycling of transducin to the GTPbound form (24). As illustrated in Reaction 1 below, this steadystate situation could be analyzed by the Michaelis analysis, where RGS9⅐G␤5L is considered the enzyme, G␣ t -GTP the substrate, and G␣ t -GDP and inorganic phosphate the reaction products (the rate of GTPase reaction under these conditions is not dependent on the concentrations of GTP and bleached rhodopsin to the extent that these concentrations are saturating).
We therefore could assess directly whether PDE␥ influences k cat or apparent K m in this reaction after measuring the rate of the GTPase reaction at a fixed concentration of RGS9⅐G␤5L and various transducin concentrations. However, before pursuing this analysis we needed to solve two experimental problems outlined below.
First, PDE␥ has been documented to inhibit rhodopsin-dependent transducin recycling from the GDP-to the GTP-bound form (25,26). This is most likely because the sites of transducin interaction with rhodopsin and PDE␥ overlap, allowing competition between PDE␥ and rhodopsin for binding to transducin (27). As a result, transducin activation is partially blocked (27), making it essentially impossible to know the concentration of G␣ t -GTP in the steady-state Michaelis analysis when PDE␥ is present. We solved this problem by substituting PDE␥ by its C-terminal peptide, PDE␥-(63-87), which has been shown to represent the GAP activity of PDE␥ (10). This peptide essentially lacks the ability to bind to the GDP-bound form of transducin because a different site on PDE␥, PDE␥-(24 -45), is re-sponsible for this interaction (21,27). To confirm that PDE␥-(63-87) potentiates the GAP activity of RGS9⅐G␤5L to the same extent as the full-length PDE␥, we conducted a singleturnover GTPase assay in which only one synchronized turnover of transducin GTPase is allowed and transducin recycling does not take place (Fig. 1). Our data indicate that PDE␥ and PDE␥-(63-87) cause the same degree of transducin GTPase activation at their saturating concentrations of 1 M.
Second, we needed to distinguish between the basal transducin GTPase activity and the activity stimulated by RGS9⅐G␤5L. Because it is not possible to remove RGS9⅐G␤5L from the washed membranes under nondenaturing conditions, we blocked RGS9⅐G␤5L activity with specific affinity-purified sheep antibodies derived against a fragment of RGS9 between residues 226 and 484 (5,7). The data presented in Fig. 2 indicate that treatment of washed membranes with these antibodies blocked the ability of RGS9⅐G␤5L to activate transducin GTPase both in the presence and absence of PDE␥-(63-87). The effect was dose-dependent with complete inhibition observed at 500 nM antibodies. The specificity of anti-RGS antibodies to inhibit GTPase activation was confirmed in control experiments in which a total fraction of nonimmune sheep IgG did not affect transducin GTPase activity.
A typical experiment addressing the dependence of transducin GTPase rate on transducin concentration under the steadystate conditions is shown in Fig. 3A. The lower data set represents the basal GTPase activity of transducin obtained after blocking the GAP activity of RGS9⅐G␤5L with anti-RGS9 antibodies. This basal activity is a linear function of transducin concentration. The other two data sets represent GTPase activity in the presence of active RGS9⅐G␤5L measured with or without PDE␥-(63-87). The "accelerated" GTPase activities are re-plotted in Fig. 3B after a point by point subtraction of the basal activity from these data sets (closed symbols). Open symbols on the same graph show the data from an experiment carried out under identical conditions but with another set of protein/membrane preparations. The striking difference between the data obtained with and without PDE␥-(63-87) is that the data with peptide could be fitted by a hyperbola yielding the average value of apparent K m of ϳ3.2 M transducin, whereas the data without peptide do not approach saturation over the range of transducin concentrations used. Note that the K m value in the intact photoreceptor could not be derived from the experiment shown in Fig. 3 because in this experiment most of the substrate (G␣ t -GTP-PDE␥-(63-87)) is soluble (not shown), whereas in the intact rod, the substrate (G␣ t -GTP-PDE) is membrane-attached because of the isoprenoid modification of the PDE ␣ and ␤ subunits (28).
To emphasize that the rate of GTP hydrolysis did not saturate without PDE␥-(63-87), we re-plotted the data without peptide from both experiments in Fig. 3C and compared the best hyperbolic fits of these data (same as in Fig. 3B) with the fits forced to the K m value of 3.2 observed with the peptide (dotted lines). Whereas the coefficients of determination (R 2 ) of the best fits were equal to 0.99 and 0.97, the corresponding parameters for the forced fits were only 0.75 and 0.54. Although we were not able to concentrate the components of this reconstituted system sufficiently to obtain a highly reliable determination of both K m and V max values in the absence of PDE␥, the difference in the apparent K m values with and without PDE␥ 63Ϫ87 in these experiments was at least 15-fold. Hyperbolic fits of the data obtained without PDE␥ yielded V max values within an ϳ25% range of the values observed with PDE␥-(63Ϫ87); this result should be treated cautiously, however, because only ϳ30% of the calculated V max was achieved at the highest transducin concentrations used. The finding that PDE␥ primarily affects the value of the apparent K m rather than the V max is further illustrated in Fig. 3D where the data from Fig. 3B are re-plotted in double-reciprocal coordinates. Thus, we conclude that the major effect of PDE␥ in potentiating the GAP activity of RGS9⅐G␤5L is to increase the affinity between G␣ t -GTP and RGS9⅐G␤5L.
Another important parameter calculated from the data presented in Fig. 3B is the k cat value for RGS9⅐G␤5L observed in the presence of PDE␥-(63-87). Based on the 1:1640 Ϯ 620 RGS9⅐G␤5L:rhodopsin ratio measured by He et al. (5) in bovine ROS, we calculated that the RGS9⅐G␤5L concentration in this experiment was 0.012 Ϯ 0.005 M, and therefore the average k cat value was equal to 98 Ϯ 40 s Ϫ1 , which makes RGS9⅐G␤5L the most efficient RGS protein studied so far.
The next set of experiments illustrated in Fig. 4 provides further proof for our conclusion that PDE␥ serves as an "affinity enhancer" between G␣ t ϪGTP and RGS9⅐G␤5L. We monitored the binding of RGS9⅐G␤5L to the His 6 -tagged G␣ t -GDP immobilized on the Ni-NTA-agarose in the absence or presence of PDE␥. Native RGS9⅐G␤5L was solubilized from washed ROS membranes by the mild, non-ionic detergent lauryl sucrose and FIG. 1. PDE␥ and PDE␥-(63-87)  incubated with the Ni-NTA-agarose beads in the presence of the same detergent. The beads were rinsed, and bound proteins were extracted with the SDS-PAGE sample buffer (see "Exper-imental Procedures"). The amount of RGS9⅐G␤5L in the extract was then detected by immunoblotting with anti-RGS9 antibodies. We chose to use full-length PDE␥ in this experiment because its higher affinity for transducin than the PDE␥-(63-87) peptide allowed better protein retention upon agarose washes. The leftmost lane shows the amount of RGS9 loaded on the beads. The next two lanes provide a control showing that RGS9⅐G␤5L did not bind to inactive GDP-bound G␣ t both with and without PDE␥. As evident from the last two lanes, transducin activation by AlF 4 Ϫ , which mimics the transition state for GTP hydrolysis most favorable for interacting with RGS proteins (29 -31), results in RGS9⅐G␤5L binding to transducin, and significant binding was observed only in the presence of PDE␥.

DISCUSSION
The dual requirement for an RGS protein (RGS9⅐G␤5L) and an effector (PDE␥) for the timely inactivation of G protein is a unique feature of the vertebrate phototransduction cascade. It is now established that the major catalytic role in activating transducin GTPase belongs to the RGS homology domain of RGS9 and that PDE␥ acts as a facilitator of RGS activity (5,14). Although PDE␥ was historically the first identified protein capable of activating G protein GTPase activity (8), its exact role in this process remained unknown. The data reported in this study indicate that PDE␥ acts as an enhancer of the affinity between G␣ t -GTP and RGS9⅐G␤5L. What remains to be elucidated in the future is whether PDE␥ acts allosterically by optimizing the conformation of G␣ t -GTP switch regions, suggested to bind to the RGS9 homology domain or whether it interacts directly with RGS9⅐G␤5L.
Our finding that PDE␥ enhances the affinity between G␣ t -GTP and RGS9⅐G␤5L provides a solid mechanistic basis for our hypothesis of why both RGS9⅐G␤5L and PDE␥ are needed for timely transducin inactivation in photoreceptor cells (4,11). We noted that when a rod photoreceptor is hit by a photon of light, it needs to accomplish two nearly opposite tasks. First, it must transduce the signal from excited rhodopsin to PDE with high efficiency. Second, it has to inactivate the whole cascade within a fraction of a second. If transducin were allowed to be discharged by RGS9⅐G␤5L before it formed a complex with PDE, then some transducin molecules would never activate PDE and signal amplification would be diminished. Therefore, the dependence of GTPase activation on transducin association with PDE␥ ensures both high efficiency of signal transmission between transducin and PDE and timely photoresponse recovery. It is tempting to speculate that similar mechanisms may be utilized in other G protein-based signal transduction cascades to solve the same general problem of preserving both response sensitivity and time resolution.
Another important finding of this study is that the PDE␥enhanced activity of RGS9⅐G␤5L reaches the rate of ϳ100 s Ϫ1 , which is at least one order of magnitude higher than most reported rates for various RGS proteins (cf. Refs. 24 and 30). However, this rate is on the same order of magnitude as the ϳ25 s Ϫ1 rate constant reported by Mukhopadhyay and Ross (32) for the RGS4-stimulated GTP hydrolysis by G q . It is interesting to consider the RGS9⅐G␤5L k cat value of 100 s Ϫ1 in the context of photoresponse recovery kinetics. This number sets the upper limit for the rate of PDE inactivation. It appears to be not only sufficient but even excessive when compared with the ϳ7 s Ϫ1 rate of the recovery from single-photon responses of dark-adapted mammalian rods (33,34). Although the slowest rate-limiting step in the photoresponse recovery may not be transducin GTPase (35), the rate of 100 s Ϫ1 is too high when compared with the quenching rates of any phototransduction step derived from physiological experiments (36). This implies FIG. 3. Catalytic properties of native RGS9⅐G␤5L complex. The GAP activity of membrane-bound RGS9⅐G␤5L was determined in multiple-turnover GTPase assay (A). Washed ROS membranes were reconstituted with various amounts of transducin in the presence (circles) or absence (triangles) of PDE␥-(63-87). In the control experiment, 500 nM anti-RGS9 antibody was added to the reaction mixture to completely block the GAP activity of RGS9⅐G␤5 (squares). GTPase reactions were initiated by the addition of 150 M GTP and stopped after a 20-s incubation. The lower curve (squares) was fit with a straight line (k ϭ 0.029 s Ϫ1 ). Two upper curves (triangles and circles) were fit with the sum of a hyperbola (Y ϭ V max ⅐X/(K m ϩ X)) and straight line (with the slope of 0.029 s Ϫ1 determined from the control with anti-RGS9 antibodies), where V max is the rate of GTP hydrolysis at saturation, K m is the apparent Michaelis constant, and X is the transducin concentration.   4. Binding of RGS9⅐G␤5L to his 6 -G␣ t -GDP immobilized on Ni-NTA-agarose. RGS9⅐G␤5L was solubilized from washed ROS membranes with 1% lauryl sucrose. Approximately 10 ng of RGS9⅐G␤5L was mixed with 5 g of his 6 -G␣ t -GDP bound to 10 l of agarose beads in 50 l of 20 mM Tris-HCl (pH 8.0), 300 mM NaCl, 0.25% lauryl sucrose in the presence or absence of AlF 4 Ϫ and PDE␥, as indicated below the gel picture. Samples were incubated for 30 min on ice. Beads were then washed twice for 5 min with the same buffer containing 20 mM imidazole. RGS9⅐G␤5L bound to G␣ t was detected by Western blotting using anti-RGS9 antibody. The leftmost lane shows the total amount (ϳ10 ng) of RGS9 added to his 6 -G␣ t beads. The data represent one of two similar experiments. that transducin GTPase in rods does not work at its maximal rate, either because each RGS9⅐G␤5L has to sequentially stimulate the GTPase activity of several transducin molecules and/or because the concentration of activated transducin remains below saturation for RGS9⅐G␤5L during the course of the photoresponse.
Remarkably, the GTPase rate of 100 s Ϫ1 is in the same range as the photoresponse recovery observed in mammalian cones. For example, the flicker fusion frequency of human cone vision, which reflects the speed of the photoresponse, exceeds 60 s Ϫ1 and is ϳ5-fold higher than in rod vision (37). In this context, Cowan and colleagues (6) reported that bovine cones contain a significantly larger amount of RGS9 than rods and suggested that this difference underlies one of the biochemical mechanisms contributing to faster recovery kinetics of cones. Our finding that RGS9⅐G␤5L has the potential to stimulate transducin GTPase at the rate of ϳ100 s Ϫ1 provides a strong support for their idea. Higher RGS9⅐G␤5L concentration in cones should result in a higher rate of transducin GTPase and photoresponse recovery by providing a higher frequency of RGS9⅐G␤5L encounters with G␣ t -GTP-PDE.