Interactions of phosducin with defined G protein beta gamma-subunits.

Phosducin has recently been identified as a cytosolic protein that interacts with the β-subunits of G proteins and thereby may regulate transmembrane signaling. It is expressed predominantly in the retina but also in many other tissues, which raises the question of its potential specificity for retinal versus nonretinal β-subunits. We have therefore expressed and purified different combinations of β- and -subunits from Sf9 cells and have also purified transducin-β from bovine retina and a mixture of β complexes from bovine brain. Their interactions with phosducin were determined in a variety of assays for β function: support of ADP-ribosylation of α by pertussis toxin, enhancement of the GTPase activity of α, and enhancement of rhodopsin phosphorylation by the β-adrenergic receptor kinase 1 (βARK1). There were only moderate differences in the effects of the various β complexes alone on α, but there were marked differences in their ability to support βARK1 catalyzed rhodopsin phosphorylation. Phosducin inhibited all β-mediated effects and showed little specificity toward specific defined β complexes with the exception of transducin-β (β), which was inhibited more efficiently than the other β combinations. In a direct binding assay, there was no apparent selectivity of phosducin for any β combination tested. Thus, in contrast to βARK1, phosducin does not appear to discriminate strongly between different G protein β- and -subunits.

Phosducin has recently been identified as a cytosolic protein that interacts with the ␤␥-subunits of G proteins and thereby may regulate transmembrane signaling. It is expressed predominantly in the retina but also in many other tissues, which raises the question of its potential specificity for retinal versus nonretinal ␤␥-subunits. We have therefore expressed and purified different combinations of ␤and ␥-subunits from Sf9 cells and have also purified transducin-␤␥ from bovine retina and a mixture of ␤␥ complexes from bovine brain. Their interactions with phosducin were determined in a variety of assays for ␤␥ function: support of ADP-ribosylation of ␣ o by pertussis toxin, enhancement of the GTPase activity of ␣ o , and enhancement of rhodopsin phosphorylation by the ␤-adrenergic receptor kinase 1 (␤ARK1). There were only moderate differences in the effects of the various ␤␥ complexes alone on ␣ o , but there were marked differences in their ability to support ␤ARK1 catalyzed rhodopsin phosphorylation. Phosducin inhibited all ␤␥-mediated effects and showed little specificity toward specific defined ␤␥ complexes with the exception of transducin-␤␥ (␤ 1 ␥ 1 ), which was inhibited more efficiently than the other ␤␥ combinations. In a direct binding assay, there was no apparent selectivity of phosducin for any ␤␥ combination tested. Thus, in contrast to ␤ARK1, phosducin does not appear to discriminate strongly between different G protein ␤and ␥-subunits.
Guanine nucleotide-binding proteins (G proteins) are transducers between heptahelical receptors and various effectors. Traditionally, G proteins have been classified according to their ␣-subunits, but in the last years a plethora of effects have been assigned to the ␤␥-subunits. Thus, they have been shown to regulate adenylyl cyclases, phospholipases C-␤ and A 2 , PI3kinase, the ADP-ribosylation factor and several ion channels (for reviews, see Refs. 1 and 2). In addition, ␤␥-subunits provide a membrane attachment site for the ␤-adrenergic receptor kinase (␤ARK) 1 which allows translocation of the kinase from the cytosol to the cell membrane and thus in close proximity to its substrate receptors (3,4). Phosphorylation of G protein-coupled receptors by ␤ARK is thought to represent the first step of homologous desensitization; it is followed by binding of pro-teins of the arrestin family which results in uncoupling of receptors and their G proteins (5,6).
Another cytosolic protein which interacts with G protein ␤␥-dimers is phosducin, a phosphoprotein originally purified from bovine retina as a complex with the ␤␥-subunits of G t (7). By binding to G protein ␤␥-subunits phosducin can inhibit G protein-mediated signaling (8,9). Furthermore, phosducin can compete with ␤ARK for the ␤␥-subunits and can thereby impair ␤ARK-mediated phosphorylation of receptors (10). The interaction of phosducin with G proteins is markedly reduced following its phosphorylation by protein kinase A (8,11).
The possibility of a common motif for the interaction of all these proteins with ␤␥-subunits has attracted much recent interest. Binding to distinct types of adenylyl cyclases, phospholipase C-␤3, atrial K ϩ channels, and ␤ARK has been found to include a defined stretch of amino acids (12). However, this sequence cannot be found in phosducin. Binding to ␤ARK involves the kinase's C terminus, which partially overlaps with the pleckstrin-homology domain (PH domain) of ␤ARK (13,14). The binding site of phosducin responsible for interaction with ␤␥-subunits has been mapped to its N terminus (15,16) which contains neither a PH domain nor the sequence mentioned above. Thus, the interaction of phosducin with G protein ␤␥subunits must be different from those with other ␤␥-binding proteins.
Molecular cloning techniques have revealed multiple isoforms of ␤and ␥-subunits. The most recent data indicate the existence of five distinct ␤and 10 ␥-subunits (17,18). Compared to the highly conserved ␤-subunits, the ␥-subunits are more divergent, suggesting them to be the specificity-determining factor. The functional differences of defined ␤␥ combinations are limited; regulation of particular types of adenylyl cyclases or stimulation of phospholipase C-␤ isoforms and K ϩ channels as well as interactions with G protein ␣-subunits did not reveal striking differences except for ␤ 1 ␥ 1 , the ␤␥ complex found in retinal rods, which showed lower potencies in its actions (19 -21). Receptor phosphorylation by ␤ARK was stimulated more potently by ␥ 2 -containing dimers and a preparation of ␤␥-subunits from bovine brain than by complexes with ␥ 3 or by ␤ 1 ␥ 1 (22).
The proteins of the retinal signal transduction cascade are homologous but distinct from those of receptor-mediated signaling (23). They are restricted in their expression to the retina and the developmentally related pineal gland, and they show specificity toward each other when compared with the proteins of other receptor systems. This applies also to the G protein ␤␥-subunits, with ␤ 1 ␥ 1 being a retina-specific combination. Phosducin, on the other hand, is highly expressed in the retina and pineal gland, but also in other tissues (8). This suggests that it may show less specificity for the other components of the retinal signaling system than has been found for other members of this system. In fact, phosducin has been shown to inhibit the GTPase-activity not only of G t but also of G o , G i , and G s preparations (8,9). Since the ␤␥-subunits appear to be the primary interaction partners for phosducin, the aim of the present investigation was to examine the specificity of interactions between defined G protein ␤␥ combinations and phosducin.

EXPERIMENTAL PROCEDURES
Expression and Purification of G Protein ␤␥-Subunits-The generation of recombinant baculoviruses for the G protein subunits ␤ 1 , ␤ 2 , ␥ 2 , and ␥ 3 has been described earlier (22). For large scale preparations of recombinant ␤␥-subunits, 200 ml of suspension cultures of Sf9 cells (2 ϫ 10 6 cells/ml) were co-infected with the recombinant baculoviruses with a multiplicity of infection of 5 or 10 for the ␤and the ␥-subunits, respectively. The expression of the ␤and ␥-subunits was monitored with Western blots using antibodies recognizing conserved epitopes of the respective subunits. The blots were developed with peroxidasecoupled goat-anti-rabbit IgG and visualized using ECL reagents (Amersham Corp.). The cells were harvested 70 h after infection, and the ␤␥ complexes were purified from the membranes by solubilization with cholate and chromatography over C7 heptylamine-Sepharose and DEAE-Sephacel columns as described previously (22). The purified ␤␥ complexes were quantitated according to Bradford (24) and also from Coomassie-stained SDS-polyacrylamide gels. Purified ␤␥ complexes were concentrated to about 15 pmol/l and were stored at Ϫ80°C with 5% glycerol.
G protein ␤␥-subunits and ␣ o from bovine brain and transducin-␤␥subunits from bovine retina were purified according to established procedures (25,26). In order to reduce the contamination of ␣ o with ␤␥-subunits to a minimum, the usual purification process was extended by a further chromatography step on a C7 heptylamine-Sepharose column.
Phosducin was expressed in Escherichia coli and purified following the procedure of Bauer et al. (8).
ADP-ribosylation of ␣ o -The activity of the recombinant ␤␥-subunits was measured by their ability to support the ADP-ribosylation of ␣ o by pertussis toxin. The assay was performed as described by Hekman et al. GTPase Activity of ␣ o -The GTPase activity of ␣ o was determined as described elsewhere (8,28). Reaction mixtures (100 l) contained 0.1 pmol of ␣ o and 0.1-2.4 pmol of ␤␥ complexes (1-24 nM). The effects of phosducin were examined at a constant concentration of ␤␥-subunits (24 nM) and 24 -720 nM phosducin.
Phosducin Binding Assay-Wells of microtiter plates were coated with 300 ng of ␤␥-subunits for at least 4 h at 4°C in 100 l of 20 mM Hepes, 20 mM NaCl, 0.1 mM EDTA, pH 7.6, and 0.05% cholate (ϭincubation buffer). The wells were then washed several times with the same ice-cold buffer supplemented with 0.05% Tween 20 (ϭwash buffer). After blocking with 3% bovine serum albumin in wash buffer, 0.1-20 g of phosducin (30 nM to 6 M) were incubated in the wells at 4°C for 2 h in 100 l of incubation buffer plus 5 mM MgCl 2 . The wells were then washed and blocked as above. Bound phosducin was determined by addition of affinity-purified rabbit anti-phosducin antibodies for 1 h at room temperature. After incubation with peroxidase-coupled goat-antirabbit IgG a color reaction was evoked with o-phenylenediamine dihydrochloride (Sigma) and stopped with 50 l of 3 M sulfuric acid, and absorption was measured at 490 nm.

RESULTS
In order to investigate effects of ␤␥-subunits of defined composition, the proteins were produced in Sf9 cells by co-infection with recombinant baculoviruses directing the expression of defined ␤and ␥-subunits. The resultant ␤␥ complexes were then purified to Ͼ95% purity by chromatography on C7 hep-tylamine-Sepharose and DEAE-Sephacel as described previously (22). The ␤␥ complexes used in this study were ␤ 1 ␥ 2 , ␤ 1 ␥ 3 , ␤ 2 ␥ 2 , and ␤ 2 ␥ 3 . Mock preparations of cells infected with wildtype baculoviruses were prepared in the same manner for control purposes. Transducin-␤␥ (␤ 1 ␥ 1 ) and a bovine brain ␤␥ preparation containing multiple ␤as well as ␥-subunits were prepared by established procedures to serve as controls.
All ␤␥ preparations were able to support pertussis toxincatalyzed ADP-ribosylation of ␣ o (Fig. 1A). There were no major differences between the different recombinant ␤␥ complexes. The two native preparations, transducin-␤␥ and the mixed preparation of ␤␥-subunits from bovine brain, were more effective at lower concentrations than the recombinant preparations; at higher concentrations this difference disappeared for transducin-␤␥ and became small for the bovine brain ␤␥ preparation.
Phosducin inhibited these enhancing effects of all ␤␥ complexes on pertussis toxin-catalyzed ADP-ribosylation of ␣ o .  tion in the presence of a constant, submaximally effective concentration (16 nM) of the various ␤␥ complexes. In this assay, only transducin-␤␥ showed a distinct inhibition curve; its effects were antagonized to a greater degree than those of other ␤␥ complexes. This occurred even though the enhancing effects of 16 nM transducin-␤␥ on ␣ o -ADP-ribosylation were not larger than those of the recombinant ␤␥ complexes and even smaller than those of the bovine brain ␤␥ preparation (Fig. 1, A and B). Among the recombinant ␤␥ complexes there were only very small differences in their inhibition by phosducin, with a slight tendency toward better inhibition by ␥ 3 -containing complexes versus ␥ 2 -containing complexes. These minor differences were not statistically significant.
Interactions between the ␤␥ complexes and ␣ o were also investigated by measuring the enhancement of the intrinsic GTPase activity of ␣ o which is exerted by ␤␥-subunits in the presence of high concentrations of Mg 2ϩ (31). All ␤␥ preparations were able to enhance the GTPase activity of ␣ o (Fig. 2A). Again, there were small differences between the different recombinant ␤␥ complexes; those containing the ␤ 1 -subunit appeared Ϸ40% more effective than those containing the ␤ 2subunit. Of the native preparations, the brain preparation of ␤␥-subunit was very effective, whereas transducin-␤␥ was the least effective of all combinations. None of the combinations reached saturation within the concentration range employed in these experiments; this was particularly the case for the most potent combinations, ␤ 1 ␥ 2 and the bovine brain ␤␥ preparation.
Again, a submaximally effective concentration of ␤␥-subunits was chosen to examine the inhibitory effects of phosducin (Fig. 2B). As was the case with the ADP-ribosylation assays, phosducin was most efficacious in inhibiting the effects of transducin-␤␥. There was Ͼ40% inhibition at the lowest concentration of phosducin, at which the concentrations of phosducin and ␤␥-subunits were identical (24 nM), and the inhibition was almost complete at higher concentrations of phosducin. Transducin-␤␥ was followed by the various recombinant ␤␥ complexes, which were all inhibited in a very similar manner by phosducin; maximal inhibition amounted to Ϸ60% in all cases, and the IC 50 values were not significantly different. The effects of the bovine brain ␤␥ preparation, in contrast, were not well antagonized by phosducin; there was almost no effect at the lowest concentration of phosducin, and even at the highest concentration of phosducin (720 nM) the inhibition was only about 30%. These data suggest the possibility that the bovine brain ␤␥ preparation might contain some ␤␥ complexes which are not well recognized by phosducin.
In order to study this hypothesis further, we investigated another effect of G protein ␤␥ complexes, the enhancement of the phosphorylation of G protein-coupled receptors by the ␤-adrenergic receptor kinase 1 (␤ARK1). We had reported earlier that there are marked differences between the ability of defined ␤␥ complexes in enhancing receptor phosphorylation by ␤ARK1 (22). Using rhodopsin (Ϸ830 nM) as the substrate and a submaximally effective concentration of ␤␥-subunits (50 nM), the combination ␤ 2 ␥ 2 caused the largest enhancement, followed by the modestly effective combination ␤ 1 ␥ 2 and the bovine brain ␤␥ preparation, while ␤ 1 ␥ 3 and ␤ 2 ␥ 3 as well as transducin-␤␥ had only very small effects (Fig. 3A).
The enhancing effects of ␤␥-subunits on ␤ARK1-mediated rhodopsin phosphorylation were inhibited by phosducin (Fig.  3B). The presence of high concentrations of phosducin (1000 nM, i.e. 20-fold above ␤␥) essentially abolished the effects of all ␤␥ combinations. The enhancing effects of several ␤␥ combinations (␤ 1 ␥ 3 , ␤ 2 ␥ 3 , and transducin-␤␥) were too small to allow a reliable determination of the inhibitory effects of phosducin. For the remaining combinations, it can be seen that phosducin inhibited the effects of the two recombinant ␤␥ combinations, i.e. ␤ 2 ␥ 2 and ␤ 1 ␥ 2 , more potently than those of the bovine brain ␤␥ preparation (Fig. 3C). The EC 50 values were 105 and 123 nM for ␤ 2 ␥ 2 and ␤ 1 ␥ 2 , respectively, and 780 nM for the bovine brain ␤␥ preparation. In contrast, the maximal extent of inhibition appeared to be similar for the three ␤␥ preparations.
Finally, we developed an enzyme-linked immunosorbent assay procedure to directly determine the binding of phosducin to the various ␤␥ complexes. The various ␤␥-dimers were immobilized in wells of microtiter plates and phosducin was then incubated in the wells to allow binding to the immobilized ␤␥-subunits. Phosducin bound to all ␤␥ complexes in a saturable manner with similar characteristics (Fig. 4). Contrary to the data shown above, there were no differences in the affinity of phosducin for the different ␤␥ combinations. There were minor differences in the maximal binding, but most notably transducin-␤␥ and the bovine brain ␤␥ preparation had equal binding capacity for phosducin. Thus, phosducin appeared to show no selectivity in the direct interaction with ␤␥-subunits, whereas in functional assays the effects of transducin-␤␥ were antagonized more efficiently. (panel B) G protein-mediated signaling systems comprise a series of proteins, including the receptors, the G protein subunits, the effectors and various partially cytosolic proteins. The steadily increasing number of isoforms for all of these proteins that are being discovered raises the question of protein-protein specificity. A fairly consistent pattern is the specificity of the proteins that are involved in phototransduction: rhodopsin, G t , cGMPphosphodiesterase, rhodopsin kinase, and arrestin are all expressed only in the retina (and in the developmentally related pineal gland). Furthermore, in in vitro assays they show marked specificity toward each other. For example, rhodopsin kinase is much better in phosphorylating rhodopsin than are ␤-adrenergic receptors, whereas the reverse is true for ␤-adrenergic receptor kinase (32). Similarly, arrestin binds much better to rhodopsin than to ␤-adrenergic receptors, while ␤-arrestin prefers ␤-adrenergic receptors (28). At the level of the signaling chain itself (receptor/G protein/effector) the issue of specificity is less clear-cut. While a series of experiments with antisense oligonucleotides against individual G protein ␣-, ␤and ␥-subunits have suggested a remarkable specificity for certain receptor/ion channel coupling (33)(34)(35)(36), reconstitution experiments have shown only moderate specificity of receptor/G protein coupling and various degrees of specificity for regulation of effectors by G protein subunits (19,20,22,37,38).

FIG. 2. Stimulation of the GTPase activity of ␣ o by ␤␥-subunits (panel A) and its inhibition by phosducin
Phosducin differs from the other proteins of such signaling chains in its expression pattern: while it is very abundant in the retina and in the pineal gland, it is also expressed in many other tissues (8). Since phosducin interacts primarily with the G protein ␤␥-subunits, and since the retinal ␤␥-dimer ␤ 1 ␥ 1 is not found in other tissues, the presence of phosducin in nonretinal tissues would make sense only if it exhibited little or no selectivity toward specific ␤␥-subunits.
This question was addressed in the present study by investigating the interaction of phosducin with a number of defined G protein ␤␥ complexes in several functional assays as well as in a direct binding assay. For this purpose, defined ␤␥ complexes were produced in Sf9 cells with the help of recombinant baculoviruses. The baculovirus expression system has been used by several groups for producing functional G protein subunits in high quantities (19,22,39,40). For the synthesis of functional ␤␥-subunits it is necessary to express both subunits simultaneously. Their functionality can be ascertained by their ability to support the ADP-ribosylation of G protein ␣-subunits by pertussis toxin. In our hands, all the combinations tested so far were able to enhance the ADP-ribosylation of ␣ o equally well. We interpret this finding as evidence for equal functional activity of the different ␤␥ preparations. Similar observations have been made by others regarding the ADP-ribosylation of native ␣ o and ␣ t , whereas for ␣ i1 produced in E. coli (which was not myristoylated) transducin-␤␥ was less efficient and potent (19, 20, 37). Another functional test for ␤␥ complexes which likewise involves interaction with the ␣-subunits is their ability to alter the intrinsic GTPase activity of the ␣-subunit. In the case of ␣ o , ␤␥-subunits activate the GTPase in the presence of high concentrations of Mg 2ϩ whereas in the presence of low concentrations of Mg 2ϩ (below 1 mM) they exert an inhibitory effect (31). In our studies, we measured modest differences in the activity of the different ␤␥-dimers on ␣ o , which were about 2-fold for the best (␤ 1 ␥ 2 ) versus the worst (transducin-␤␥) combination. The pattern of selectivity was somewhat different from the one found in the ADP-ribosylation assays, suggesting that there are different requirements for the ␤␥-subunits in the two different types of experiments. Investigations by Ueda et al. (20) on the inhibition of ␣-subunit GTPase activity by ␤␥-subunits have also failed to reveal significant selectivity with the exception of ␤ 1 ␥ 1 , which was less potent in all assays.
A third, quite different, test for the functionality of the ␤␥ complexes is their ability to serve as membrane anchors for ␤ARK to enhance agonist-dependent receptor phosphorylation (3,4). In an earlier study we had found differences between defined ␤␥ complexes in their ability to enhance ␤ARK-mediated rhodopsin phosphorylation (22). Similar differences were found here, with ␤ 2 ␥ 2 being most efficient and transducin-␤␥ least.
Phosducin had inhibitory effects on all of these ␤␥-mediated effects. Among the different ␤␥ combinations, the stimulatory effects of transducin-␤␥ both on ADP-ribosylation and GTPase activity were more efficiently antagonized, while we did not detect any real differences between the recombinant dimers consisting of ␤ 1 , ␤ 2 , ␥ 2 , and ␥ 3 . Phosducin also inhibited the ␤␥-mediated enhancement of rhodopsin phosphorylation. However, the divergent enhancing potential of the ␤␥ complexes allowed a characterization of this inhibition only in the case of ␤ 1 ␥ 2 , ␤ 2 ␥ 2 , and ␤␥ B .
While stimulation of rhodopsin phosphorylation by ␤ 1 ␥ 2 or ␤ 2 ␥ 2 were equally well inhibited by phosducin, the effect on ␤␥ B was less pronounced and appeared to require higher concentrations of phosducin. Similarly, we also observed only a weak inhibition on the ␤␥ B -mediated stimulation of ␣ o -GTPase. We speculate that the mixture of ␤␥-subunits from bovine brain might contain ␤␥ complexes which are only poorly recognized by phosducin.
Whereas the functional assays suggested a preference of phosducin for transducin-␤␥, we found no real differences in affinity or maximal binding when we assayed phosducin binding to immobilized ␤␥-subunits directly. Thus, the apparent selectivity of phosducin for transducin-␤␥ in the functional assays may be caused by events which occur subsequent to the initial binding step. An alternative explanation is that transducin-␤␥ has a better solubility than the other ␤␥ complexes and that the binding assay was done with immobilized ␤␥subunits while the ADP-ribosylation and GTPase assays were carried out in solution. In this context it is important to note that the affinities measured in the direct binding assay with immobilized ␤␥-subunits were Ϸ10-fold lower than those determined in the functional assays in solution. Likewise, the affinities of phosducin for ␤ 1 ␥ 2 and ␤ 2 ␥ 2 found in the ADPribosylation and GTPase assays were Ϸ3-fold higher than those determined in the ␤ARK assay, where the ␤␥-subunits were integrated in the rod outer segment membranes. Similar discrepancies have been noted earlier. For example, Heithier et al. (41) reported severalfold higher affinities of ␤␥ complexes for ␣ o in detergent-containing solution than in lipid vesicles.
Since transducin-␤␥ differs from ␤ 1 ␥ 2 or ␤ 1 ␥ 3 only in its ␥-subunit, the small functional specificity of phosducin for transducin-␤␥ must be due to the presence of ␥ 1 . ␥ 1 contains a different isoprenylation, farnesyl instead of geranylgeranyl in all other G protein ␥-subunits. The functional importance of this farnesyl, compared to other prenyl modifications for the interaction with receptors, has very recently been elucidated by Kisselev et al. (42). Alternatively, the slightly different effects of ␥ 1 may be due to its rather different amino acid sequence: the amino acid identity of ␥ 1 with ␥ 2 or ␥ 3 is less than 40%, while the sequence identity between ␥ 2 and ␥ 3 is about 80% (18).
Overall, however, the interactions of phosducin with the various ␤␥ complexes appear to be quite similar. This relative lack of specificity goes in line with the nonretinal expression of phosducin, since it enables an interaction of phosducin with multiple G proteins and thus an interference with multiple signaling pathways. It contrasts with the more pronounced selectivity of ␤ARK for certain ␤␥ complexes where ␤ 2 ␥ 2 was the preferred partner and transducin-␤␥ ineffective (22) (see also Fig. 3A). This agrees with our hypothesis mentioned in the introduction that the ␤␥-binding domains of ␤ARK and phosducin are dissimilar. While some homologies between the N terminus of phosducin and the C terminus of ␤ARK have been noted (14,15), phosducin contains neither a PH domain nor the ␤␥-binding consensus sequence Gln/Asn-X-X-Glu/Asp-Arg/Lys proposed by Chen et al. (12). Furthermore, phosducin-like protein (43) with its much shorter N terminus which shows no similarity at all to the C terminus of ␤ARK, also interacts with G protein ␤␥-subunits (44), supporting the differences in the ␤␥-binding mechanisms between phosducin and ␤ARK.
While these data suggest that the binding sites on G protein ␤␥-subunits for ␤ARK and phosducin might be different, they must clearly be overlapping, since phosducin and ␤ARK have been shown to compete for ␤␥ binding (10). Furthermore, the binding sites for phosducin and for ␤ARK appear to overlap also with the binding site for the ␣-subunits, since both proteins interfere with coupling between the ␤␥and the ␣-subunits. Thus the ␤␥-subunits appear to possess various distinct but overlapping binding sites for different proteins.
Reconstitution experiments employing defined G protein ␤␥ complexes have largely failed to reveal much specificity. This contrasts with antisense experiments which indicate that G proteins of very specific ␣-, ␤-, and ␥-composition are required for efficient coupling of certain receptors to ion channels (33)(34)(35)(36). This raises the possibility that selectivity may be encoded in the cellular organization rather than in protein-protein interactions. Further studies are required to determine whether similar rules might also govern the interactions of phosducin with G protein ␤␥-subunits.