Sequence of interactions in receptor-G protein coupling.

Guanine nucleotide exchange in heterotrimeric G proteins catalyzed by G protein-coupled receptors (GPCRs) is a key event in many physiological processes. The crystal structures of the GPCR rhodopsin and two G proteins as well as binding sites on both catalytically interacting proteins are known, but the temporal sequence of events leading to nucleotide exchange remains to be elucidated. We employed time-resolved near infrared light scattering to study the order in which the Galpha and Ggamma C-terminal binding sites on the holo-G protein interact with the active state of the GPCR rhodopsin (R*) in native membranes. We investigated these key binding sites within mass-tagged peptides and G proteins and found that their binding to R* is mutually exclusive. The interaction of the holo-G protein with R* requires at least one of the lipid modifications of the G protein (i.e. myristoylation of the Galpha N terminus and/or farnesylation of the Ggamma C terminus). A holo-G protein with a high affinity Galpha C terminus shows a specific change of the reaction rate in the GDP release and GTP uptake steps of catalysis. We interpret the data by a sequential fit model where (i) the initial encounter between R* and the G protein occurs with the Gbetagamma subunit, and (ii) the Galpha C-terminal tail then interacts with R* to release bound GDP, thereby decreasing the affinity of R* for the Gbetagamma subunit. The mechanism limits the time in which both C-terminal binding sites of the G protein interact simultaneously with R* to a short lived transitory state.

In eukaryotes, signal transduction across cell membranes is in many cases based on the interplay between G protein-coupled receptors (GPCRs) 1 and heterotrimeric guanine nucleotide-binding proteins (G proteins, G␣␤␥). Binding of extracellular signaling molecules like hormones, neurotransmitters, or odorants to GPCRs triggers structural rearrangements in the receptor, such that its intracellular domain becomes competent to catalyze nucleotide exchange in the heterotrimeric G protein (1).
Rhodopsin is the visual pigment in retinal rod photoreceptors, those cells responsible for seeing under dim light conditions, and is the prototypical GPCR of the large family of rhodopsin-like GPCRs. Rhodopsin's ligand, the chromophore 11-cis-retinal, is covalently bound and recognizes a photon as an extracellular signal. Within 200 femtoseconds, the energy of the photon causes cis 3 trans isomerization of the retinal, thereby triggering the conversion of inactive dark-adapted rhodopsin into the active receptor conformation (R*), which is reached after milliseconds and is capable of interacting with transducin, the G protein of the rod cell (2).
High resolution structures of transducin (G t (3) and the closely related heterotrimeric G protein G i ␣ 1 ␤ 1 ␥ 2 (4)) and rhodopsin (in the dark-adapted 11-cis-retinal bound state (5)) are available (Fig. 1). However, static crystal structures alone cannot elucidate the dynamics of the receptor-G protein interaction. Previous studies have focused on identifying structural domains involved in catalysis. The key binding sites on transducin are the C-terminal tails (CT) of the G␣ subunit and the farnesylated G␥ subunit of the G␤␥ dimer (CT␣ and CT␥-far, respectively), which specifically recognize and bind to R* (6 -8).
In current models of nucleotide exchange, it is assumed that both of these sites act simultaneously on the distant nucleotide binding domain by a pull or lever mechanism (8 -14).
In this work, we have investigated the unknown temporal sequence of interaction between the CT␣ and CT␥-far sites and R*. We used "mass-tagged" peptides (in which the key CTs are fused to functionally neutral maltose-binding protein (MBP)), modified G proteins (wild type and modified in their attached lipids or C-terminal G␣ amino acid sequence), and a kinetic near infrared light scattering assay (15)(16)(17) to monitor proteinprotein interactions in real time. This approach allowed us to investigate how individual receptor binding sites on the G protein are functionally linked with each other. Based on the experimental data, we propose that nucleotide exchange requires a sequential two-step interaction of the G protein with R*. An encounter of CT␥-far with R* initiates the interaction and thereby makes CT␣ available for binding to R*. In the second step, R* shifts interaction from CT␥-far toward CT␣. Now the nucleotide binding site has low affinity for GDP and is prepared for the uptake of GTP.

EXPERIMENTAL PROCEDURES
Native and Modified G␣ Subunits and G␤␥ Dimers-G proteins used in this study were either the G protein of the rod photoreceptor cell (transducin, purified from bovine eyes), purified recombinant G i ␣ 1 , or a G t ␣ construct. Because of the known very low expression of the soluble transducin G␣ subunit (G t ␣) in Escherichia coli and Sf9 cells, single amino acid substitutions were introduced into G t ␣ to yield a G t ␣/G i ␣ 1 chimera that contained 16 residues of G i ␣ 1 in a G t ␣ background (18). G i ␣ 1 is not present in photoreceptor cells but belongs to the same G␣ subfamily as G t ␣ and couples to light-activated rhodopsin. G␣ with an N-terminal myristoyl modification was obtainable for native G t ␣ and for * This work was supported by the Deutsche Forschungsgemeinschaft/ Sonderforschungsbereich 449. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Cloning, Expression, and Purification of Chimeric G t ␣/G i ␣ 1 and G i ␣ 1 Subunits-The expression vector pHis6-Gt␣*, which was generously provided by M. Natochin and N. Artemyev, contains a chimeric bovine G t ␣/rat G i ␣ 1 cDNA sequence, preceded by a nucleotide sequence encoding six histidines as an affinity tag (18). The chimeric G␣ protein contains only 16 residues from G i ␣ 1 and was shown to be similar to native G t ␣ in receptor interaction and basal nucleotide exchange. For construction of the mutants, the SpeI-HindIII fragment of pHis6-Gt␣* was introduced into the pLitmus 38 cloning vector (New England Biolabs), yielding the precursor pL-Gt␣*. An oligonucleotide duplex with the respective mutation was cloned into the Tth111I and SapI restriction sites of pL-Gt␣*, and the modified SpeI-HindIII fragment was subcloned into SpeI/HindIII-digested pHis6-Gt␣*. Recombinant G i ␣ 1 was expressed and purified (19) using the plasmid pQE-60 (Qiagen) harboring the rat G i ␣ 1 coding sequence (20). This protein was designed to contain an internal His 6 epitope (amino acid sequence GGHHHHH-HGGGMTA) after position 121, where the homologous GPA1 from yeast has a long insert compared with mammalian G␣ subunits. For expression of myristoylated G i ␣ 1 and mutants thereof, the respective G protein-encoding plasmids were cotransfected with pBB131 (coding for yeast N-myristoyltransferase; a generous gift of Jeffrey Gordon (21)) into E. coli JM109. Cultures were grown at 30°C, induced with 100 M isopropyl-1-thio-␤-D-galactopyranoside at an A 600 of 0.8 and harvested 12-14 h later. Expression and purification of the G t ␣/G i ␣ 1 chimera was performed as described (18,19,22). E. coli BL21(DE3) harboring plasmids encoding G␣ subunits were induced with 100 M isopropyl-1-thio-␤-D-galactopyranoside at an A 600 of 0.8 and harvested 5-7 h later (20°C). A TALON metal affinity resin (BD Biosciences Clontech) and an imidazole gradient were used for purification.
Expression and Purification of G␤␥ Dimers-Nonfarnesylated G␤␥ complexes were expressed in and purified from Sf9 cells infected with the respective baculoviruses. Baculoviruses encoding G␤ 1 and G␥ 1 C71S were a generous gift from A. G. Gilman and P. Gierschik, respectively. G␤ 1 ␥ 1 C71S was recovered from the cytosolic fraction of Sf9 lysates as described (23) and purified by Ni 2ϩ -nitrilotriacetic acid co-chromatography with His-tagged G i ␣ 1 (20).
Cloning, Expression, and Purification of MBP Fusion Proteins-The expression vector pMal-c2x was purchased from New England Biolabs. Synthetic oligonucleotide duplexes encoding a Gly-Gly linker followed by peptide sequences derived from the C-terminal regions of G␣ or G␥ ( Fig. 2A) were ligated into EcoRI/BamHI-digested pMal-c2x. Expression and purification of the fusion proteins was performed as described (26). MBP-CT␥-far was prepared by intein-mediated protein ligation (27) using the protein splicing element Mxe GyrA intein excised from vector pTXB3 (New England Biolabs) and the synthetic peptide CDKN-PFKELKGGC-farnesyl. After on-column ligation (chitin beads; New England Biolabs), the protein was purified by chromatography on an amylose resin (New England Biolabs). MBP-CT␥-far was identified by mass spectrometry and had a purity of Ͼ85% according to SDS-PAGE analysis.
Peptide Synthesis-Peptide synthesis, farnesylation, and purification were carried out as described before (28). The amino acid sequences of the peptides are given in Fig. 2A. The amino termini of the peptides were unmodified, and carboxyl termini of the G␥-derived peptides were amidated. The peptide CDKNPFKELKGGC-farnesyl used for preparation of MBP-CT␥-far was synthesized as C-terminal amide on a Rink resin (loading 0.25 mg/mol; Rapp Polymere, Tü bingen, Germany) with a Pioneer Synthesizer (Applied Biosystems) using an Fmoc (N-(9-fluorenyl)methoxycarbonyl) strategy. The C-terminal Cys-sulfur was protected by a trityl group, and the N-terminal Cys-sulfur, which was used for intein-mediated ligation, was protected by an S-(tert-butylsulfenyl) group. All other side chain functions were protected with trifluoroacetic acid labile groups. The peptide was cleaved from the resin with 95% trifluoroacetic acid using triisopropylsilane as a scavenger. The crude peptide was precipitated twice with ether and lyophilized from acetonitrile/water (1:3). For farnesylation, 100 mg of the crude peptide (0.065 mmol), dissolved in 3 ml of N,N-dimethylformamide, was treated with 1.4 eq of farnesyl bromide (0.09 mmol) and an equal amount of diisopropyl ethylamine (0.16 mmol). The formation of oxidation products was suppressed by degassing with N 2 . After 1 h, the crude farnesylation mixture was purified by preparative HPLC. For deprotection of the S-(tert-butylsulfenyl) group from cysteine, 20 eq of tris(2-carboxyethyl)phosphine hydrochloride were added to a solution of the peptide in 50% acetonitrile/water. The mixture was adjusted to pH 5 with NH 3 and stirred at room temperature for 5 h and then separated by HPLC. The farnesylated peptide obtained was characterized by mass spectrometry and analytical HPLC.
Disk Membrane and Transducin Preparation-Preparations were performed as described previously (17).

Measurement of G Protein Conformational Changes by Fluorescence-
The basal GDP/GTP exchange of G␣ subunits (2 M) in the absence of G␤␥ and rhodopsin was monitored by detection of fluorescence emission at 340 nm (excitation at 300 nm, measured at 20°C with constant stirring) using a SPEX fluorolog II spectrofluorometer (see Ref. 16 and references therein). Activation of G␣ was started by adding GTP␥S (10 M final concentration). Activation of the whole G␣ pool was completed by adding NaF and then AlCl 3 (yielding 100 M AlF 4 Ϫ final concentration). Traces were normalized to show the same total increase of intensity of fluorescence emission induced by GTP␥S and AlF 4 Ϫ . Kinetic Light Scattering-Changes in intensities of scattered near infrared light were measured as described before (17). Light-induced binding of proteins from solution to activated rhodopsin in disk membranes leads to an increase of the size of the scattering particle and concomitant increase of the intensity of scattered light. The sensitivity of the light scattering assay is given by the measuring conditions and the experimental setup (15)(16)(17). As employed in the present study, the light scattering change is proportional to the gain of mass and monitors binding of "mass-tagged" peptides (MBP-peptide fusion proteins) to R* but not binding of short synthetic peptides to R* because of the small peptide mass. All measurements were performed in 10-mm path cuvettes at pH 7.4 (20 mM bis-tris propane, 130 mM NaCl, 1 mM MgCl 2 , pH 7.5) and 23 Ϯ 1°C. Reactions were triggered by flash photolysis of rhodopsin (3 M) using flashes of green light (500 Ϯ 20 nm). Binding signals were recorded with 32% flash-activated rhodopsin. For data evaluation, see "Appendix."

Monitoring the Interaction between R* and C-terminal Binding Sites of the G Protein by Kinetic Light
Scattering-To monitor in real time binding between R* in disk membranes isolated from the rod outer segment and interacting CTs of the heterotrimeric G protein (namely CT␣ and CT␥-far; Figs. 1 and 2A), the kinetic near infrared light scattering assay was employed (see "Experimental Procedures" and Refs. 15 and 16). "Mass-tagged" peptides, which can be obtained by expression and purification of MBP-peptide fusion proteins, allow convenient measurement of the binding to R* (Fig. 2B).
A typical binding signal triggered by flash activation of rhodopsin is shown in Fig. 2B for MBP-CT␣HA1 (top trace). This MBP fusion protein contains the well known G t ␣ high affinity analog CT sequence ( Fig. 2A, CT␣HA1), which was identified by Hamm and co-workers (26). Since both the small synthetic and MBP "mass-tagged" CT␣HA1 peptides bind to R*, binding of MBP-CT␣HA1 to R* can be inhibited by an excess of the competing small synthetic CT␣HA1 peptide (Fig. 2B, middle trace). A further control showing that the assay monitors the interaction between R* and MBP-CT␣HA1 is provided by the effect of hydroxylamine, which affects this interaction by hydrolyzing the retinal Schiff base in a competing reaction (Fig.  2B, second trace from top). Hence, the degradation of R* results in a transient interaction, which is reflected in a transient increase of the intensity of scattered light.
By testing peptides with modified CT␣t sequences, we were able to identify a single amino acid exchange (K341L in G t ␣) that increases the affinity for R* of the corresponding CT␣ peptide (termed CT␣HA2) to a level similar to CT␣HA1 (details to be published). "Mass-tagging" of CT peptides with MBP was extended to CT␣t and CT␣HA2 and the farnesylated CT␥-far. All of these MBP-CT fusion proteins showed binding to R*, were sensitive to competition by the corresponding synthetic CT peptide, and were sensitive to the effect of hydroxylamine ( Fig. 3C and data not shown). MBP lacking a fused G protein CT peptide tail showed no binding to R* (data not shown).
Only One of the C-terminal Peptides, CT␣ or CT␥-far, Can Bind to R* at One Time-To investigate whether CT␣ and CT␥-far interact simultaneously with R* (see Fig. 1), competition of the two CTs for binding to R* was investigated with the kinetic light scattering assay. The interaction between the MBP-CT␣HA1 fusion protein and R* could be inhibited in a concentration-dependent manner by the presence of a synthetic peptide corresponding to CT␥-far (Fig. 3A). Analogously, binding of MBP-CT␣HA2 and MBP-CT␣t to R* could be suppressed by excess CT␥-far peptide (data not shown). The competition between the MBP-CT␣HA1 fusion protein and the CT␥-far peptide for R* could be supported by a centrifugation assay (Fig. 3B). In this biochemical binding assay, rhodopsin in disk membranes is pelleted in the presence of MBP-CT fusion pro- 1. Interface between the heterotrimeric G protein transducin and its receptor rhodopsin. Transducin (Protein Data Bank accession number 1GOT) consists of the nucleotide-binding G t ␣ subunit and the heterodimeric G t ␤␥ subunit (G t ␣ 1 and ␤ 1 ␥ 1 are the species found in the rod cell of the mammalian visual system). G␣ with the bound nucleotide GDP (space-filling model) is shown in green, G␤ in blue, and G␥ in magenta. Several putative interaction sites on the G protein have been reported (6,7,18,(42)(43)(44)(45) (reviewed in Ref. 13), but only two of them (shown in orange), the G␣ C-terminal tail (CT␣, residues 340 -350) and the farnesylated C-terminal tail of G␥ (CT␥-far, residues 60 -71), have been confirmed to specifically bind and stabilize the active rhodopsin conformation (6 -8). NMR studies suggest that these tails adopt helical structures upon binding to rhodopsin (33,46). For illustration, these regions were modeled with a helical conformation. The two lipid modifications of transducin, heterogeneous fatty acylation of the G␣ N-terminal Gly and farnesylation of the G␥ Cterminal Cys side chain, are not resolved in the crystal structure and were modeled (myristoylation and farnesylation are shown as a spacefilling model; amino acids 1-5 of G␣ and amino acids 67-71 of G␥ were modeled). These lipid modifications are thought to be in proximity (31). Binding sites for transducin on the GPCR rhodopsin (Protein Data Bank accession number 1HZX), which contains the bound chromophore 11-cis-retinal (gray, space-filling model), comprise the cytoplasmic region (gray) without the extreme C terminus (red) following two palmitoylated cysteines (not shown; reviewed in Ref. 47). The relative orientation of both proteins during catalytic receptor-G protein interaction is unknown. The inset gives a schematic representation of the two proteins with the orange triangle and the orange L shape depicting the binding conformations of CT␣ and CT␥-far, respectively. The gray trapezoid and rectangle depict the GDP and myristoyl moiety of G␣, respectively.

FIG. 2. Binding of C-terminal tails of G protein ␣ and ␥ subunits to light-activated rhodopsin can be monitored by kinetic light scattering.
A, the amino acid sequences of the C-terminal tails studied either as synthetic peptides, fused to MBP or contained in the different G protein subunits (see "Experimental Procedures") are as follows: CT␣t (residues 340 -350 of bovine G t ␣), CT␣i (residues 344 -354 of rat G i ␣ 1 ), and G␣ high affinity (HA) analog sequences, CT␣HA1 (26) and CT␣HA2. CT␥-far corresponds to amino acids 60 -71 of bovine G␥ 1 and was farnesylated at Cys-71. B, flash illumination (arrow) of disk membranes containing rhodopsin triggers complex formation between soluble MBP-CT fusion proteins and light-activated rhodopsin (R*) as seen in the increase of the relative intensity of scattered light (⌬I/I, binding signal (15)). As an example, binding of MBP-CT␣HA1 (0.5 M) to R* (1 M) is shown in the upper trace. Under the experimental conditions, the assay is not sensitive enough to monitor binding of R*-interacting small synthetic peptides (lower two traces; 1 M R*, 50 M peptide). The binding of MBP-CT fusion proteins to R* can be inhibited by corresponding short synthetic peptides (shown for the CT␣-HA1 peptide (50 M), middle trace). The interaction can also be terminated by inactivating R* with hydroxylamine (20 mM), which hydrolyzes the retinal Schiff base in a competing reaction (second trace from top). MBP-CT␣HA2 and MBP-CT␥-far show the same basic binding behavior (data not shown). teins in the dark or light, followed by subsequent analysis of the pellet and supernatant by SDS-PAGE and protein staining. The centrifugation assay also showed that the MBP-CT␣ fusion proteins investigated ( Fig. 2A) as well as a control MBP protein lacking a CT sequence did not bind to the disk membranes in the dark (data not shown).
Next, we measured binding of the MBP-CT␥-far fusion protein to R* by kinetic light scattering. As observed with the MBP-CT␣ fusion proteins, a binding signal was obtained with MBP-CT␥-far (Fig. 3C). The farnesyl moiety of the MBP-CT␥far fusion proteins results in considerable interaction of the fusion protein with the disk membrane already in the dark, which was detected by the centrifugation assay (data not shown). This is reminescent to physiologically completely lipidmodified G i ␣ 1 /G t ␤␥ that shows binding to the disk membrane in dark and light (Fig. 4, top). Since only fusion proteins, which bind from solution to the scattering disk vesicle, contribute to the increase of intensity of scattered light, the intensity change is smaller for MBP-CT␥-far.
When we performed the competition experiment with the MBP-CT␥-far fusion protein and CT␣ peptides, an analogous inhibition by the CT␣ peptide in a concentration-dependent manner was seen in the light scattering assay (shown for CT␣HA1 in Fig. 3C). At higher concentrations of the competing CT␣ peptide (Ͼ1 M), a transient small binding signal becomes visible, which sits on the normal light scattering change with competitively reduced amplitude. It may reflect a delayed action of the CT␣ peptide, in agreement with the data presented below. The binding of MBP-CT␥-far to the disk membrane in the dark causes a high background in the centrifugation assay, thereby limiting the use of this assay in competition experiments involving MBP-CT␥-far (data not shown). Taken together, the competition experiments show that R* can interact with only one C-terminal binding domain of the G protein at one time, either CT␣ or CT␥-far.
The First Interaction Step Requires Lipid-modified G Protein-The competition experiments described above suggest that in the holo-G protein, the two CT binding domains interact sequentially and not simultaneously with R*. Therefore, the question arose which CT binding domain of the holo-G protein is involved in the initial encounter step with R*. For interaction of a synthetic CT␥-far peptide with R*, it is known that the farnesyl moiety is indispensable (7,8). However, for the transducin holoprotein, farnesylation is controversially discussed (29,30). Thus, we prepared the G␤␥ dimer with and without farnesylation. Furthermore, we prepared the G␣ subunit with and without myristoylation at the N terminus in order to investigate whether the hydrophobic modification on the N terminus of the G␣ subunit (Fig. 1) is involved in the interaction with R* as suggested by its close proximity to the farnesyl moiety (31). Hofmann, and C. Kleuss, manuscript in preparation). Samples contained lipid modifications as indicated (myristoylation (myr) of G i ␣ 1 or farnesylation (far) of G␥). Under the experimental conditions the G␣/ G␤␥ combinations shown formed holoproteins (data not shown), which were not bound to the disk membrane when at least one lipid modification was lacking. The amplitude of the binding signal of the ϩmyr/ ϩfar combination is reduced, because this holoprotein shows like native transducin considerable binding to the disk membrane in the dark, an effect, which is more pronounced for the holoprotein containing G i ␣ 1 . Consequently, the concentration of soluble G protein, which is available to contribute to changes of the intensity of scattered light (17), is reduced. Similar results were obtained for transducin when heterogeneously fatty acylated native G t ␣, the nonmyristoylated G t ␣/G i ␣ 1 chimera and G t ␤␥ with and without farnesyl moiety were used (data not shown). No difference in behavior was observed when enzymatically defarnesylated G t ␤␥ or the recombinant G␤␥(C71S) mutant was used.
Light scattering and centrifugation experiments were performed with native transducin subunits (G t ␣, G t ␤␥), G i ␣ 1 with its physiological myristoyl moiety, and G t ␤␥ lacking its farnesyl moiety. Recombinant nonmyristoylated G␣ subunits were obtained by expression in E. coli, either a G t ␣ construct or G i ␣ 1 , which can couple like transducin to R* (see "Experimental Procedures"). By light scattering, rapid and strong binding to R* was seen with G proteins that are missing one of the two lipid modifications, either myristoylation or farnesylation (Fig.  4). This was supported by the centrifugation assay, as seen by the intensity of the G␣ and G␤ protein bands, which reflect the amount of G protein bound to the disk membrane in dark and light, respectively (Fig. 4, protein bands are boxed). Native holo-G protein, which carries both lipid modifications, is like MBP-CT␥-far partially membrane bound and therefore yields only a smaller binding signal. No interaction with rhodopsin or R* is observed when both lipid modifications are lacking. This cannot be explained by a lack of holo-G protein formation, since the G␣/G␤␥ combinations tested formed heterotrimers in solution as determined by analytical gel filtration. 2 The data presented therefore imply that one lipid modification in the holo-G protein is required for interaction with R*. In the absence of G␥ farnesylation, myristoylation of G␣ can replace CT␥-far.
We have shown above that the MBP-CT␣ fusion protein interacts with R*. Therefore, CT␣ as part of the non-lipidmodified holo-G protein should be able to confer affinity to the G protein to bind to R*. However, this was not observed in the holo-G protein lacking both lipid modifications (Fig. 4), suggesting that this binding domain of G␣ is not available for interaction with R* in the holo-G protein. Taken together, this argues for an initial encounter of R* with the hydrophobic farnesyl/myristoyl pair, which starts nucleotide exchange catalysis and subsequently enables CT␣ to interact with R*.
Sequence of Interactions-Another line of evidence for the CT␣ binding site being the second in the interaction of the holo-G protein with R* is shown in Fig. 5. We investigated the influence of the CT␣ sequence on the reaction. By a spectroscopic binding assay ("Extra-MII" assay; see Ref. 16), a CT␣ peptide (CT␣HA2; Fig. 2A) was determined to develop an affinity for R* 2 orders of magnitude higher than the native CT␣ peptide (CT␣t or CT␣i), 3 similar to an already described high affinity CT␣ peptide (CT␣HA1) (26). Nonmyristoylated G␣ either with the native or the high affinity CT␣ (CT␣HA2) was tested for R* binding in combination with farnesylated G␤␥. In the absence of additional nucleotides (only the endogenous GDP was present in the nucleotide binding site), both combinations showed identical binding signals. The same rates of R* binding support the notion that CT␣ is not involved in the initial encounter step.
In the presence of additional GDP (ϳ800-fold excess), the amount of R*⅐G complex formed was reduced as seen in the lower final level of the binding signals at completion (Fig. 5B). The effect was less pronounced in the case of the modified holo-G protein, containing the high affinity CT␣. Kinetic anal-

FIG. 5. C-terminal tails of G␥ and G␣ subunits bind sequentially in a two-step reaction to light-activated rhodopsin (R*).
A, reaction scheme used to fit the kinetic data and to describe the steps of R*-catalyzed nucleotide exchange. After the initial encounter of R* with farnesylated G␥ (Equil. 1), the G␣ C-terminal domain enters its R* binding site accompanied by GDP release and dissociation of the binding domain on G␥ (Equil. 2). Uptake of GTP by the nucleotide-free R*⅐G complex leads to dissociation into R*, G␣⅐GTP, and G␤␥. GTP is hydrolyzed to GDP and inorganic phosphate (P i ) due to the intrinsic GTPase activity of G␣, allowing the recombined holo-G protein to re-enter the reaction cycle. B, binding of 0.6 M nonmyristoylated G i ␣ 1 /G t ␤␥ with low affinity or high affinity G␣ C terminus (CT␣i or CT␣HA2, respectively) to R* (1 M) monitored by kinetic light scattering. The arrow indicates time of flash illumination to generate R*. The presence of GDP (500 M) reduces the amount of R*⅐G complex formed to a different degree, depending on the affinity of the G␣ C-terminal binding domain (CT␣i or high affinity CT␣HA2). C, measurements as in B but in the presence of GTP (0.6 M) instead of GDP. The data were fitted according to the reaction scheme given in A and plotted as dotted lines (Table I). Similar results were obtained with an attenuated flash (0.09 M R*). ysis according to the reaction model (Fig. 5A) revealed that the high affinity CT␣ has no influence on the initial encounter step but specifically affects the second step (GDP release). It favors the formation of the nucleotide-free R*⅐G complex, which is seen in a shift of equilibrium 2 by a factor of 7 (shown for G i ␣ 1 in Fig. 5, B and C, Table I, k 2,on /k 2,off ).
When the measurements were performed in the presence of low amounts of GTP, the holoprotein with high affinity CT␣ showed a pronounced effect on GDP release and GTP uptake as observed in the reduced amplitude of the traces but not on the initial encounter step as seen in the initial rise of the traces after photolysis of rhodopsin. Due to GTP-induced dissociation, the R*⅐G complex forms only transiently (Fig. 5C, Table I). The presence of the high affinity CT␣ on the G protein accelerates significantly GTP uptake and R*⅐G complex dissociation by a factor of 3 (Table I). This suggests that upon GDP release, CT␣ is the site that remains engaged with R*, thereby determining the G protein conformation in which GTP can be taken up.
G␤␥-dependent Binding of CT␣-The single Lys 3 Leu substitution in CT␣t (K341L in G t ␣) not only elevated the affinity of CT␣ for R* but also revealed that CT␣ is apparently not available for interaction with R* in the isolated G␣ subunit. Nonmyristoylated G␣ with wild type CT␣ is incapable of interacting with R* in the presence of farnesyl-lacking G␤␥ (Fig. 4, lowest trace) or absence of farnesylated G␤␥ (Fig. 6A, lowest  trace), whereas the corresponding MBP-CT␣ fusion protein binds readily to R* (data not shown). However, the mutant G␣ subunit containing the high affinity Lys 3 Leu substitution (G␣-CT␣HA2) showed slow G␤␥-independent binding to R* (Fig. 6A). Interestingly, this mutant G␣ subunit does not interact with rhodopsin in the dark but was detected in rhodopsincontaining membranes after illumination (Fig. 6B). Light-induced binding of the mutant G␣-CT␣HA2 subunit to R* was also observed in the presence of GTP and when this mutant G␣ was preactivated with AlF 4 Ϫ (Fig. 6A). The mutation had very little effect on the basal nucleotide exchange rate of G␣ and the conformational change induced by AlF 4 Ϫ uptake as measured with a fluorescence activation assay (Fig. 6C). The mutant G␣-CT␣HA2 subunit showed binding to R* similar to the MBP-CT␣ fusion proteins, indicating that in this mutant G␣, the high affinity CT␣ is now available for interaction with R* in any conformational state of G␣. However, the capability of this mutant G␣ to bind to R* is lost when a holo-G protein is formed with G␤␥ lacking its farnesyl moiety (Fig. 6A). In contrast, farnesylated G␤␥ in the mutant holoprotein containing G␣-CT␣HA2 accelerated binding to R* and enabled R*-catalyzed GDP/GTP exchange, corroborating that the initial interaction of R* with G␤␥ makes CT␣ available for interaction with R* (Fig. 6A). DISCUSSION The Interaction Model-We performed kinetic experiments to elucidate how in the rhodopsin/transducin model system the G protein heterotrimer is presented to its receptor at the membrane surface, to illuminate the sequence of interactions of key binding sites of the G protein with the activated GPCR (R*) and how this sequence is associated with nucleotide exchange. We performed the experiments with rhodopsin embedded in disk membranes of rod cells and two G proteins, consisting of G t ␣ or G i ␣ 1 and G t ␤␥, obtaining comparable results. We have provided four lines of experimental results, namely (i) the competition of isolated C-terminal sequences of G␣ (CT␣) and G␥ (CT␥-far) for binding to R*, (ii) the necessity of a hydrophobic lipid modification on either the G␥ C terminus or G␣ N terminus ( Fig. 1) to initiate the interaction with R*, (iii) the influence of the CT␣ sequence on GTP-dependent dissociation of the R*⅐G complex, and (iv) shielding of a constitutively available C terminus of a mutant G␣ subunit by holoprotein formation with G␤␥. Our results are consistent with a model in which the interaction of the G protein with R* is initiated by the encounter of R* with the G␤␥ subunit. Either the membrane interaction of the G protein or, more likely, the contact of the G protein with R* is required to make CT␣ available for interaction with R*. GDP may be released when both CT␣ and CT␥-far are transiently engaged in the interaction with R*. Eventually, however, the affinity of the receptor for the G␤␥ subunit decreases, relocating the main interaction of R* to CT␣ and resulting in a nucleotide-free R*⅐G complex into which GTP can be taken up. It is important to note that such a mechanism does not exclude simultaneous interaction of both sites with R* but embeds it into the sequence of events as a short lived transitory state (Fig. 5A). This may provide a general scheme according to which GPCRs activate their G proteins.
The initial encounter of the G protein with R* is likely to be associated with a conformational change in G␤␥. A conformational switch in G␤␥ was previously proposed on the basis of a lack of accessibility of CT␥ to carboxypeptidase Y in the solubilized G␤␥ or holotransducin, suggesting that CT␥ is masked at least in the non-membrane-bound state of the G protein (32). In a recent NMR study of a CT␥-far peptide, it was reported that CT␥ is unstructured in the presence of an inactive receptor but forms an amphipathic helix upon rhodopsin activation (33). In the initial encounter complex, the contact of R* with the lipid modifications of the holo-G protein appears to trigger the process of making the second CT binding site, CT␣, available for interaction. Interestingly, myristoylation at the N terminus of G␣ can functionally replace farnesylation of G␤␥ (Fig. 4). The precise role of the myristoylated N terminus of G␣ in the physiological process with farnesylated G␤␥ remains to be studied. Also unresolved is how dissociation of the R*⅐G complex occurs after GTP uptake. We described a mutant G␣ subunit, containing a high affinity CT␣, which binds to R* independent of its GDP-or GTP-bound conformational state but dissociates like native transducin from R* when G␤␥ is present in the complex with R* (Fig. 6; see below). This suggests that the GTP-induced R*⅐G complex dissociation passes through a similar but reversed sequence of interaction steps as established for R*⅐G complex formation. The subsequent disso- ciation of the active G protein into G␣⅐GTP and G␤␥ makes the reaction quasi-irreversible.
The Physiological Necessity of a Sequential Interaction-We have described that nucleotide exchange catalysis in G␣ can be improved by increasing the affinity of CT␣ for R* (Fig. 5, Table  I). An intriguing question that arises from this finding is why the native system does not employ a G␣ subunit with a high affinity CT␣. An important consequence of the high affinity mutation is that nonmyristoylated G␣ subunits can bind independent of G␤␥ to R*. Binding to R* is possible and strong in both the GDP-bound and the activated GDP⅐AlF 4 Ϫ conformation, suggesting that this mutation in the G␣ monomer uncouples the nucleotide binding site from CT␣ and exposes CT␣ for interaction with R*. It takes the complex with G␤␥ to make this high affinity CT␣ unavailable for direct binding to R* (shown by the farnesylation-dependent binding to R*) (Fig. 6A). Any high affinity G␣ subunit that is not bound to G␤␥ would, through its more accessible CT␣, represent a severe threat to signal transduction in the cell, since it would block the active GPCR and thereby poison the catalyst. Both efficient shielding of CT␣ in G␣ and holo-G proteins and a secure mechanism for making CT␣ accessible upon catalytic interaction with R* are therefore essential for maximal effectiveness of signal transduction. One would hope to learn the structural basis for the inaccessibility of CT␣ in monomeric inactive G␣⅐GDP and active G␣⅐GTP subunits from the crystal structure of G i ␣ 1 (34). Unfortunately, for the holo-G proteins the crystal structures are not informative, because most of CT␣ is not resolved (3,4), leaving molecular details open for further studies.
Receptor Dimerization-Another open question is whether R* is represented by a receptor monomer or dimer. The mechanism of receptor-G protein interaction given above is compatible with an interaction between the G protein and a receptor monomer as well as a receptor dimer and does not require receptor dimerization for activation of the G protein. Recently, atomic force microscopy showed that rhodopsin can form dimers in disk membrane preparations (35). Future studies will have to address this point and determine the molecular identity of R*, which may consist of one or two activated receptor molecules or a dimer between one activated and one inactive receptor molecule. It even appears possible that the G protein induces receptor dimerization during the interaction process. It can be anticipated that a dimer between two lightactivated rhodopsin molecules is not required for G protein activation. Such dimers do not exist in the single photon working range (36) of the rod photoreceptor cell. Furthermore, it remains to be studied how the sequence of catalytic steps in R*-G protein coupling identified here relate to different conformations of the activated receptor, which are separated by protonation changes in rhodopsin (2,(37)(38)(39)(40)(41). The G protein may use different receptor conformations, which can bind either CT␣ or CT␥-far, to allosterically control the affinity of the two binding sites on the receptor for efficient nucleotide exchange catalysis.
Acknowledgments-We thank Paul Hargrave for critically reading the manuscript and Helena Seibel for technical assistance.

Kinetic Analysis of Receptor-G Protein
Coupling-Data analysis of the binding signals of G␣␤␥ (Fig. 5, B and C) was performed on the basis of the reaction scheme depicted in Fig.  5A. The time dependence of the concentrations of each reactant (the first derivative over time) can be described by the following differential equations.  6. G␤␥-independent binding of a G␣ subunit with high affinity C terminus to R* and basal nucleotide exchange time course of the G␣ subunit. A, binding of nonmyristoylated G i ␣ 1 with high or low affinity G␣ C terminus (CT␣HA2 or native CT␣i, respectively) to R* (1 M) was started by flash illumination (indicated by an arrow) of rhodopsin. Experiments were performed with G i ␣ 1 -CT␣HA2 alone (10 M) or in the presence of GTP (50 M) or after preactivation with AlF 4 Ϫ (100 M). AlF 4 Ϫ binds to inactive G␣⅐GDP near the site occupied by the ␥-phosphate in G␣⅐GTP (48) and induces a conformation very similar to the activated G␣⅐GTP (49,50). The interaction with R* is accelerated in the presence of G␤␥ (0.6 M G t ␤␥ plus 0.6 M G i ␣ 1 -CT␣HA2). When an excess of GTP (50 M) is present in addition, the R*⅐G i ␣ 1 -CT␣HA2⅐G t ␤␥ complex goes through nucleotide exchange and subsequent complex dissociation and therefore does not accumulate. R*⅐G i ␣ 1 -CT␣HA2⅐G t ␤␥ complex formation does not occur when defarnesylated G␤␥ is used (2 M G t ␤␥ (C71S mutant) plus 2 M G i ␣ 1 -CT␣HA2). In contrast to nonmyristoylated G i ␣ 1 -CT␣HA2, the nonmyristoylated G i ␣ 1 -CT␣i alone (10 M) does not bind to R* (lowest trace). B, rhodopsin in disk membranes (10 M) was incubated with nonmyristoylated G i ␣ 1 -CT␣HA2 (0.6 M) in complex with G t ␤␥ (0.6 M, upper panel) or as isolated subunit (10 M, lower panel) in the presence or absence of GTP (50 M) and pelleted. Supernatants (s) and pellets (p) were analyzed by SDS-PAGE and proteins were visualized by Coomassie Blue staining. Rhodopsin (R) was not completely removed during the SDS sample preparation and a small fraction is observed in the pellet lanes (17). Similar results in both assays (light scattering and centrifugation) were obtained when recombinant G t ␣ with CT␣HA2 was used. C, basal nucleotide exchange of nonmyristoylated G i ␣ 1 with high or low affinity G␣ C terminus (CT␣HA2 or native CT␣i, respectively) was monitored by measuring the increase in fluorescence emission. Traces of G i ␣ 1 -CT␣HA2 and G i ␣ 1 -CT␣i are superimposed. At the time indicated by the arrow, the reaction was started by the addition of GTP␥S. Activation of remaining G␣⅐GDP was started by the addition of NaF and AlCl 3 from separate stock solutions to form AlF 4 Ϫ at the times indicated by the arrows. The light scattering change ⌬I/I is proportional to the sum of the concentrations of the formed R*-G protein complexes, [R*⅐G⅐GDP] and [R*⅐G]. Therefore, the experimental values for ⌬I/I could be fitted using a scaling factor f (for further details, see Ref. 17). The set of three binding signals of the G protein (with either native or high affinity CT␣) in the presence of different nucleotides (no nucleotide, GDP added, or GTP added; Fig. 5, B and C) was fitted simultaneously using the differential equations above and a numerical multiple least square fit procedure (Scientist software; MicroMath Scientific Software, Salt Lake City, UT). The initial concentrations of G protein, R*, GDP, and GTP were fixed for each individual experiment, and the other parameters (k 1,on , k 1,off , k 2,on , k 2,off, k 3 , k 4 , f) were allowed to vary. The initial conditions were as follows: A first fit yielded kinetic parameters for G i ␣ 1 /G t ␤␥. Data fitting of G i ␣ 1 -CT␣HA2/G t ␤␥ was performed with fixed rate constants for equilibrium 1 (Fig. 5A, k 1,on and k 1,off ) using the values of G i ␣ 1 /G t ␤␥. Parameters obtained are given in Table I.