Signal Transfer from GPCRs to G Proteins

Catalysis of nucleotide exchange in heterotrimeric G proteins (Gαβγ) is a key step in cellular signal transduction mediated by G protein-coupled receptors. The Gα N terminus with its helical stretch is thought to be crucial for G protein/activated receptor (R*) interaction. The N-terminal fatty acylation of Gα is important for membrane targeting of G proteins. By applying biophysical techniques to the rhodopsin/transducin model system, we studied the effect of N-terminal truncations in Gα. In Gαβγ, lack of the fatty acid and Gα truncations up to 33 amino acids had little effect on R* binding and R*-catalyzed nucleotide exchange, implying that this region is not mandatory for R*/Gαβγ interaction. However, when the other hydrophobic modification of Gαβγ, the Gγ C-terminal farnesyl moiety, is lacking, R* interaction requires the fatty acylated Gα N terminus. This suggests that the two hydrophobic extensions can replace each other in the interaction of Gαβγ with R*. We propose that in native Gαβγ, these two terminal regions are functionally redundant and form a microdomain that serves both to anchor the G protein to the membrane and to establish an initial docking complex with R*. Accordingly, we find that the native fatty acylated Gα is competent to interact with R* even in the absence of Gβγ, whereas nonacylated Gα requires Gβγ for interaction. Experiments with N-terminally truncated Gα subunits suggest that in the second step of the catalytic process, the receptor binds to the αN/β1-loop region of Gα to reduce nucleotide affinity and to make the Gα C terminus available for subsequent interaction with R*.

ical GPCR and the eponym of the largest family of the GPCR superfamily is rhodopsin, the visual pigment of retinal rod photoreceptor cells. Photon capture by rhodopsin's covalently bound chromophore 11-cis-retinal causes retinal cis/trans isomerization and thus activates rhodopsin. Within milliseconds, subsequent conversions in the protein moiety lead to the active rhodopsin conformation (R*), which is capable of catalyzing GDP/GTP exchange in the retinal G protein transducin (G t ) (3,4). In the inactive state, G t consists of G␣ t ⅐GDP and the G␤ 1 ␥ 1 dimer (5) (see Fig. 1A).
A great deal of data regarding rhodopsin and G t are available: high resolution structures of both rhodopsin (6 -8) and G t (5,9,10); information on interaction sites on receptor and G protein (reviewed in Refs. 2, 4, and 11); experimental (e.g. see Refs. 2 and 12-16) and theoretical studies (17,18); and various concepts of R*/G t interaction (11, 15, 19 -28). Despite the abundance of this information, however, it has not been sufficient to comprehensively describe how R* catalyzes nucleotide exchange in the G protein. The side of G t facing R* (side oriented downward in Fig.  1A) (29) contains two key interaction sites, the C-terminal tail of G␣ t (CT␣) and the C-terminal tail of G␥ 1 with its farnesyl modification (CT␥). CT␣ and farnesylated CT␥ selectively recognize R* (30,31) and adopt helical conformations upon binding (32)(33)(34). Involvement of the fatty acylated N-terminal region of G␣ t (NT␣; see Fig. 1A) in catalytic R*/G protein interaction was suggested by a study on a form of G␣␤␥ lacking a farnesyl moiety (15). There, fatty acylation of the G␣ N terminus enabled interaction with R*. This and other results (35)(36)(37)(38) indicate the presence of NT␣ in the R*/G t interface, although its function there is not well understood.
In this work, we set out to elucidate the role of NT␣ in G protein/receptor coupling by investigating N-terminal truncated G␣ subunits. We confirm that lack of NT␣ has an effect on subunit interaction with G␤␥ (39 -41). We show that the G protein's two terminal regions carrying hydrophobic modifications (i.e. myristoylated NT␣ and farnesylated CT␥) are functionally redundant parts of an amphiphilic microdomain that serves to anchor the G protein to the membrane and affords collisional coupling with R*. The actual trigger of nucleotide exchange requires a subsequent step that makes the G␣ C terminus available for interaction with cytoplasmic binding sites of R* (15).

EXPERIMENTAL PROCEDURES
Materials-Rod outer segments and disk membranes were prepared from frozen dark-adapted retinas as described (42). G t heterotrimers were purified from rod outer segment preparations essentially as described previously (42,43). Subsequent separation into the G␣ t subunit and the G␤␥ dimer was performed by chromatography on a HiTrap Blue column (GE Healthcare Life Sciences) as described (42). The contamination of G␣ t preparations by G␤␥ was estimated to be less than 1%. This result was based on evaluation of the initial rates of R*-induced GTP␥S uptake by G␣ t with decreasing G␤␥ concentrations using the G t fluorescence activation assay.
Peptide Synthesis-Peptide synthesis, myristoylation, and purification were as described previously (44). Lyophilized peptides were dissolved in water, and the pH was adjusted with 1 M NaOH to pH 6. Peptides were either unmodified or myristoylated at the N-terminal Gly.
UV-visible Spectroscopy-The amount of "extra MII" was recorded by time-resolved UV-visible spectroscopy as described (45). Samples contained 10 M rhodopsin in disk membranes. Measurements were performed at pH 8.0 and 4°C. Cuvette path length was 2 mm. 11.5% of rhodopsin was flash-activated by 500 Ϯ 20-nm light.
Near Infrared Kinetic Light Scattering-Changes in intensities of scattered light were measured as described (43). All measurements were performed in 10-mm path cuvettes at pH 7.4 and 23 Ϯ 1°C. Reactions were triggered by flash photolysis of rhodopsin (3 M) in disk membranes using flashes of green light (500 Ϯ 20 nm; the flash activated 32% of rhodopsin).
Transducin Fluorescence Activation Assay-G t activation was monitored by relative changes in intrinsic fluorescence emission at 340 nm after excitation at 300 nm as described before (45). Samples containing rhodopsin in disk membranes (50 nM), G␣ (0.6 M), and G␤␥ (0.5 M) were illuminated with orange light (Ͼ495 nm, 10 s). Nucleotide uptake by the G protein was induced by adding GTP␥S (5 M final concentration) at 20°C and constant stirring. The intrinsic GDP/GTP exchange of G␣ subunits (0.6 M) was monitored in the absence of G␤␥ and rhodopsin (see Ref. 45 and references therein). To complete activation of the entire G␣ pool, NaF and AlCl 3 were added sequentially with ϳ20 s between additions (final concentrations: 3.7 mM and 100 M, respectively). Traces were normalized to the final fluorescence levels evoked by the G␣⅐GTP␥S/G␣⅐GDP⅐AlF 4 Ϫ mixture (15). Size Exclusion Chromatography-Size exclusion chromatography was used to characterize the G protein subunit association in solution using the molecular weight shift that occurs upon heterotrimer formation. The subunits (ϳ6 M, 50 l final volume) were loaded on a Superose 12 column (0.32 ϫ 30 cm; GE Healthcare Life Sciences) equilibrated with 20 mM, 1,3-bis-(tris(hydroxymethyl)-methyl-amino)propane (BTP), pH 7.5, 130 mM NaCl, 1 mM MgCl 2 , 2 mM dithiothreitol (10°C) at a flow rate of 40 l/min using a Smart System (GE Healthcare Life Sciences). The elution was monitored at 280 nm, and 40 l fractions were collected for subsequent SDS-PAGE analysis.
Cloning, Expression, and Purification of G␣ Subunits-Expression of non-fatty acylated G␣ subunits in Escherichia coli and purification were as described before (15). Experiments were performed with rat G␣ i1 (termed G␣ i ) and a G␣ t /G␣ i1 chimera (here termed rG␣ t ; originally termed G t ␣* in Ref. 46), which contains 16 residues from G␣ i1 and can be expressed in E. coli (46). The expression vector pHis6-G t ␣* for expression of rG␣ t was kindly provided by M. Natochin and N. Artemyev. Both G␣ subunits are known to couple in combination with G␤ 1 ␥ 1 , as transducin couples to activated rhodopsin (47,48). In order to produce ⌬25-rG␣ t , bases corresponding to residues 2-25 of the G␣ t sequence were deleted by a PCR using pHis6-G t ␣* as a template. The PCR product was digested with NcoI and BstEII and subcloned into pHis6-G t ␣*. For construction of G␣ i mutants with N-terminal deletions, oligonucleotide duplexes with respective deletions were cloned into the EcoRI and BssHII restriction sites of the plasmid pGiL, which harbors the sequence coding for rat G␣ i1 with an internal His 6 tag (15). The subunits termed G␣ i -t, ⌬29-G␣ i -t, ⌬31-G␣ i -t, ⌬33-G␣ i -t, and ⌬35-G␣ i -t are derivatives of G␣ i (vector pGiL) with a single Asn 3 Glu amino acid exchange to yield the G␣ t C-terminal sequence (Fig. 1B). For G␣ i -t and the respective deletion mutants, we obtained about 3 mg of pure protein/liter of cell culture.

Intrinsic Nucleotide Exchange and Heterotrimer Formation of G␣ Subunits with N-terminal
Deletions-In order to study the role of NT␣ in catalytic R*/G protein interaction and subunit interaction with G␤␥, we expressed three sets of G␣ subunits in E. coli. Each set was composed of full-length G␣ and mutants with N-terminal truncations of various length (see Fig. 1B). The most extensive truncation included the end of the N-terminal helix and first residues of the ␤1 strand, which were reported to be involved in receptor contact (36,37). Since light-activated rhodopsin can interact with both G␣ t /G␤ 1 ␥ 1 and G␣ i /G␤ 1 ␥ 1 (47,48), the first set was derived from recombinant G␣ t (rG␣ t and ⌬25-rG␣ t ). rG␣ t is a chimera containing 16 residues from G␣ i , which allows functional expression in E. coli (46). The other two sets were based on G␣ i (G␣ i and ⌬29-G␣ i ) and a G␣ i mutant containing the N346E substitution, which changes the C-terminal tail to the corresponding G␣ t sequence (G␣ i -t, ⌬29-G␣ i -t, ⌬31-G␣ i -t, ⌬33-G␣ i -t, and ⌬35-G␣ i -t). G␣ i -t was preferentially used in the study, because the protein's termini are not modified by an affinity tag, and interaction with R* is similar to G␣ t (11,47,48).
We measured intrinsic GTP␥S-induced fluorescence increase of isolated G␣ subunits to study the influence of the different N-terminal truncations on intrinsic nucleotide exchange. Almost the same rates of GTP␥S uptake were found for G␣ i and ⌬29-G␣ i , in agreement with published data (41) (data not shown). This was also observed for G␣ i -t, ⌬29-G␣ i -t, and ⌬31-G␣ i -t, whereas ⌬33-G␣ i -t showed a slightly faster GTP␥S uptake ( Fig. 2A). After the addition of AlF 4 Ϫ , all of these subunits formed G␣⅐GDP⅐AlF 4 Ϫ complexes, thus demonstrating functional folding of the G␣ subunits. However, ⌬35-G␣ i -t did not show any fluorescence increase, after the addition of either GTP␥S or AlF 4 Ϫ ( Fig. 2A, bottom trace), indicating that no GDP was bound in the nucleotide binding pocket.
We used size exclusion chromatography to study heterotrimer formation of truncated nonacylated G␣ subunits with G␤␥. Nonacylated G␣ i -t, ⌬29-G␣ i -t, ⌬31-G␣ i -t, and ⌬33-G␣ i -t as well as G␤␥ eluted as single peaks from the size exclusion column, demonstrating the homogeneity of the subunit prepa-rations (Fig. 2B). When combined with G␤␥, all four G␣ subunits showed single peak elution profiles and shorter retention times as compared with the isolated G␣ subunits. Almost identical results were obtained with nonacylated G␣ i in combination with G␤␥ (data not shown). Although truncation of the N-terminal region reduces the affinity of the subunits as previously reported for G␣ t and other G␣ subunits (39 -41, 49, 50), the truncated G␣ subunits can form heterotrimers under the experimental conditions used in this study (i.e. in the micromolar range). Only ⌬35-G␣ i -t showed a complex elution profile with multiple maxima, indicating the presence of larger aggregates. Also, G␤␥ had little effect on the retention time, suggesting that ⌬35-G␣ i -t was unstable or not correctly folded (data not shown).
Influence of the N-terminal Region of G␣ on the Interaction between the Heterotrimeric G Protein and R*-We applied kinetic light scattering, which is a monitor of R*/G protein interaction (42,43) to study R* binding of nonacylated G␣ subunits in the presence of G␤␥ (Fig. 3). All full-length G␣ subunits investigated, namely rG␣ t and G␣ i -t (Fig. 3A), as well as G␣ i (not shown) bound to R* in the presence of G␤␥. Similar binding to R* in the presence of G␤␥ was observed for the N-terminal truncated subunits ⌬25-rG␣ t , ⌬29-G␣ i -t, ⌬31-G␣ i -t, ⌬33-G␣ i -t (Fig. 3A), and ⌬29-G␣ i (not shown). As expected, G␤␥ and the apparently misfolded ⌬35-G␣ i -t did not bind to R* (Fig.  3A, bottom).
In the presence of G␤␥ and excess GTP, light activation of rhodopsin did not give rise to binding signals of G␣ i -t, ⌬29-G␣ i -t, ⌬31-G␣ i -t, or ⌬33-G␣ i -t (Fig. 3B). This result indicates that for all four G␣ subunits, binding of G␣␤␥ heterotrimers to R* is followed by rapid GTP uptake and G␣␤␥ dissociation from both the receptor and the membrane. Furthermore, the lack of any light scattering change suggests that G␣␤␥ heterotrimers were not bound to the disk membrane in the dark (see Refs. 15 and 42). The same results were obtained with rG␣ t and ⌬25-rG␣ t as well as with G␣ i and ⌬29-G␣ i (data not shown).
As another approach to address receptor-catalyzed G protein activation, GTP␥S uptake by G␣ subunits in the presence of G␤␥ and catalytic amounts of R*-containing disk membranes was monitored by fluorescence spectroscopy (Fig. 4). Consistent with the results from the light scattering assay, the lack of the N-terminal helix in rG␣ t and G␣ i -t had no severe effect on R*-catalyzed nucleotide exchange (Fig. 4). Similar observations were made for G␣ i and ⌬29-G␣ i (data not shown). A gradual decrease in the kinetics was observed as the length of the deletion increased (compare G␣ i -t, ⌬29-G␣ i -t, ⌬31-G␣ i -t, and ⌬33-G␣ i -t), which may arise from reduced G␣␤␥ formation. No GTP␥S uptake was observed for ⌬35-G␣ i -t/G␤␥, again demonstrating the complete functional loss of this G␣ mutant.
Since the effect of N-terminal truncation of G␣ on R* interaction was small, we tested two peptides derived from the N terminus of G␣ t and G␣ i1 , G␣ t -(2-28) and G␣ i1 -(2-32), both prepared with and without a myristoyl moiety at the N-terminal Gly (see Fig. 1B). G t holoprotein and peptides corresponding to the key binding sites of G t , CT␣ and farnesylated CT␥, are known to stabilize active metarhodopsin II (MII; "extra MII" assay) (30,31). In contrast to these two peptides, the G␣ N-terminal peptides did not stabilize MII (peptide concentrations up to 3 mM; data not shown). Nevertheless, the myristoylated G␣ N-terminal peptides inhibited G t -induced MII stabilization. Similar inhibition was observed for the interaction between R* and CT␣ (measured with MBP-CT␣HA1 fusion protein as in Ref. 15) (data not shown). Importantly, the inhibition was only observed with myristoylated peptides. On the other hand, changes in the sequence of the G␣ N-terminal peptides had only minor effects on the inhibition, suggesting that interaction FIGURE 1. Model of heterotrimeric G t and G␣ subunits used. A, G t is composed of G␣, G␤, and G␥ subunits, shown in green, blue, and magenta, respectively (5) (Protein Data Bank code 1GOT). CT␣ and the farnesylated CT␥ (both shown in orange) are known as the key binding sites for the active receptor and were modeled according to NMR data (32,34). The crystal structure reveals a helical conformation of the N-terminal region of G␣ (NT␣, residues 2-28; red). In transducin, NT␣ is heterogeneously fatty acylated at the N-terminal Gly residue (myristoylation as the predominant modification of G␣ t was modeled, shown in gray). B, structure alignment shows NT␣ of bovine G␣ t (red) and rat G␣ i (subtype G␣ i1 ; blue) with the ␣N helix linked by a short loop (␣N/␤1-loop) to the ␤1 strand (green). The dashed lines represent residues not resolved in the structures of the heterotrimeric G proteins ( 2 GAGA 5 for G␣ t and 2 GCTLS 6 for G␣ i ; Protein Data Bank codes 1GOT and 1GP2, respectively). The G␣ subunits investigated were native rod cell-specific bovine G␣ t , a G␣ t /G␣ i chimera termed rG␣ t (containing an N-terminal His 6 tag and 16 residues from G␣ i1 in the C-terminal half), and rat G␣ i (subtype G␣ i1 with an internal His 6 tag). The alignment shows N-terminal amino acid sequences of G␣ t (residues 1-35), G␣ i (residues 1-39), and mutants with N-terminal deletions (⌬25-, ⌬29-, ⌬31-, ⌬33-, and ⌬35-, respectively). Subunits termed G␣ i -t were derived from G␣ i and contained the G␣ t C-terminal sequence due to an Asn 3 Glu mutation (IKENLKDCGLF). Residues comprising the ␣N helix and ␤1 strand are underlined and shaded.
of the N-terminal peptide with the receptor is mainly driven by the myristoyl moiety. However, we cannot entirely exclude the possibility that inhibition is due to a more indirect effect induced by membrane partitioning of the peptides.
G␤␥-independent Interaction of G␣ Subunits with R*-We performed kinetic light scattering to investigate the interaction of the nonacylated (Fig. 5A) and fatty acylated (Fig. 5B) G␣ subunits with R* in the absence of G␤␥. Even at high concentrations (10 M), the nonacylated subunits rG␣ t , ⌬25-rG␣ t , G␣ i -t, ⌬29-G␣ i -t (Fig. 5A, upper four traces), and G␣ i , as well as ⌬29-G␣ i (not shown), failed to bind R*. Interestingly, and in contrast to all other nonacylated G␣ subunits investigated, the further truncated subunits ⌬31-G␣ i -t and ⌬33-G␣ i -t bound to R* even in the absence of G␤␥. Binding occurred with low affinity, as indicated by the small amplitude of the binding signal, despite the high concentration of the G␣ subunit (10 M ; Fig.  5A). The low but significant light-induced binding was abolished in the presence of GTP (Fig. 5A), indicating a G␤␥-independent activation of these G␣ subunits.
We then tested native G␣ t , which is heterogeneously acylated at its N-terminal Gly (51,52). Whereas no binding to R* was observed for nonacylated rG␣ t (Fig. 5A, upper trace), native G␣ t bound to R* in a dose-dependent manner even in the absence of G␤␥ (Fig. 5B, three upper traces). From the amplitude of the binding signal (⌬I/I ϳ 0.01), we can calculate that with 0.5 M G␣ t in the sample, about 40% of the subunit was bound to R* (⌬I/I ϭ 0.01 corresponds to ϳ0.1 M G␣␤␥ or ϳ0.2 M G␣ bound to the membrane and/or receptor (42)). Hence, the binding could not be attributed to G␣␤␥ formed from trace amounts of G␤␥ contamination (Ͻ1%) present in the G␣ t and rod outer segment membrane preparation (see "Experimental Procedures"). Furthermore, G␤␥-independent   OCTOBER 6, 2006 • VOLUME 281 • NUMBER 40 binding to R* was also seen with recombinant myristoylated G␣ i , which is essentially G␤␥-free (data not shown). Light activation of rhodopsin did not evoke binding signals of acylated G␣ t in the presence of GTP (Fig. 5B, bottom). These results indicate that fatty acylation at the N-terminal Gly residue of G␣ t or G␣ i enables G␤␥-independent binding to R*. Furthermore, binding is followed by GDP/GTP exchange, leading to G␣ activation and subsequent fast dissociation of the R*⅐G␣ complex.

DISCUSSION
In the crystal structures of G␣␤␥, NT␣ forms an ␣-helical stretch that protrudes from G␣ and links it to the G␤␥ dimer (Fig. 1). This prominent structural feature was found in heterotrimers of G␣ t and G␣ i and is probably shared by all G proteins (5,53). NT␣ is also part of the R*/G protein interface (35,36), and its characteristic fatty acylation at the N-terminal Gly is mandatory for membrane association of the G t heterotrimer (15,54). N-terminal truncations of G␣ were therefore found to have effects on subunit association and R* interaction (35,39,40,49). Here we resolve the function of NT␣ into membrane binding, receptor binding, and nucleotide exchange catalysis.
Membrane Attachment and R* Docking-The available G␣␤␥ crystal structures suggest that the N terminus of G␣ and the C terminus of G␥ are close to one another in space. Accordingly, NT␣ appears to serve as a spacer that enables the two hydrophobic modifications of G␣ and G␥ (both not resolved in the crystal structures) to form a common site for membrane anchoring (Fig. 1). This notion is supported by binding studies of G t to model membranes (55) and electron crystallography of G t in the membrane-bound state (29). The latter investigation further revealed that only about 2% of the G protein surface is in contact with the membrane. This conclusion coincides with the weak lipid/protein interactions found for G t , which engages a small amount of phosphatidylserine (56). Taking these facts into account, it is easy to understand that the lack of one of the two hydrophobic modifications (15,54) or N-terminal truncation of G␣ renders the G protein heterotrimer soluble.
The G␣ N terminus and the G␥ C terminus with their hydrophobic modifications are also involved in receptor binding (15,30,35,36) and may be viewed as an amphiphilic microdomain that is engaged in R* interaction. However, removal of the N-terminal myristoyl moiety or even truncation of whole NT␣ did not abolish R*/G protein coupling (Figs. 4 and 5). The same was observed when the farnesyl moiety or the C-terminal Gly-Gly-Cys-farnesyl fragment was removed from G␥ (15). This indicates that fatty acylated NT␣ and farnesylated CT␥ are functionally redundant and that one site is sufficient for interaction with R*. The initial docking to R* appears to be of predominantly hydrophobic nature with rather low specificity, since the sequences of CT␥ and NT␣ are different and vary between subtypes (e.g. see Fig. 1B). The NT␣-derived peptides fail to selectively stabilize MII. Furthermore, farnesylated CT␥derived peptides exhibit a lower MII-stabilizing efficiency compared with CT␣-derived peptides (30). 5 These observations suggest that the G␣ N terminus and G␥ C terminus, particularly   their hydrophobic modifications, may recognize binding sites that are already present in precursors of MII or even in the inactive ground state. The apparently weak contribution of the NT␣ sequence in R* interaction suggests that the main role of the N-terminal helix is to position the myristoyl moiety in the vicinity of the farnesyl moiety.
Role of NT␣ in Nucleotide Exchange Catalysis-We concluded previously that upon formation of the docking complex between R* and G␣␤␥, via the amphiphilic microdomain defined above, the second key binding site of the G protein, namely CT␣, becomes available for R* interaction (15). The experiments described in this study on truncated versions of isolated G␣ subunits (i.e. without G␤␥) suggest a role of NT␣ in this process (Fig. 5). We found a weak but significant interaction between R* and native fatty acylated G␣ (i.e. in the absence of G␤␥). This interaction did not occur when fatty acylation was missing. However, truncation of NT␣, including the loop linking the ␣N helix and the ␤1 strand (␣N/␤1-loop; see Fig. 1), rendered isolated G␣ competent to couple to R*, even without the initial hydrophobic interaction described above. This result suggests that in such truncated forms of G␣, the CT␣ binding site is already unleashed for interaction with R*. What then makes the CT␣ binding site available in the native holoprotein? Based on our results, we propose that structural alterations in the ␣N/␤1loop region contribute to the unleashing of CT␣ in G␣␤␥. It is conceivable that cytoplasmic receptor loops directly target the ␣N/␤1-loop region upon formation of the initial docking complex between R* and G␣␤␥. Experimental evidence was provided by cross-linking experiments on rhodopsin/G t interaction. An interaction between the third cytoplasmic receptor loop and the G␣ t fragment Leu 19 -Arg 28 covering the ␣N/␤1-loop was mapped (36). Furthermore, it was reported that the ␣N/␤1-loop allows GPCRs to distinguish between G␣ 15 and G␣ 16 subtypes (37). The ␣N/␤1-loop region also appears to be an element in controlling the affinity of GDP, because truncation of this region in G␣ increased intrinsic GDP/GTP exchange (Fig. 2). Interestingly, a similar effect was found previously for a point mutation close to this loop in the ␤1 strand (57).
EPR studies on G␣ i1 showed that NT␣ adopts an ordered structure in the myristoylated G␣ subunit (58,59), suggesting that even in isolated G␣ a structure corresponding to the ␣N helix in G␣␤␥ can be formed. NT␣ may act as spacer between the lipid anchor site and the ␣N/␤1-loop region, thus establishing a geometric constraint between these two regions. Such a constraint, enhanced by G␣/G␤␥ association in the heterotrimer, would be employed to bring the ␣N/␤1loop region and respective sites on R* in proximity for subsequent interaction to unleash CT␣. NMR studies and fluorescent labeling of Cys 347 in G␣ t showed that CT␣, which is not resolved in the crystal structures, adopts different conformations in solution (16, 60 -62). The conformational changes were dependent on the type of nucleotide bound by G␣, association of G␣ and G␤␥, and interaction of G␣␤␥ with R*. Further studies will be required to identify the specific structural changes needed to make CT␣ available for R*. Interestingly, a point mutation close to the nucleotide-bind-ing site in Switch 3 of a G␣ t /G␣ i chimera yielded a nucleotide-free G␣ mutant that blocked rhodopsin-catalyzed G protein activation (63). This suggests that in this mutant CT␣ is available for R*, indicating a link between the occupancy of the G␣ nucleotide-binding site and the availability of CT␣ for R*. Similarly, a mutation in CT␣ of nonmyristoylated G␣ i led to G␤␥-independent binding to R*, even in the presence of GTP␥S (15).
Freissmuth and co-workers (41,64) showed G␤␥-independent activation of G␣ i1 and G␣ o subunits by a receptor-mimetic peptide derived from the N-terminal portion of the third cytoplasmic loop of the D 2 -dopamine receptor. Interestingly, G␣ activation by the peptide was not affected by the lack of myristoylation or a 30-amino acid truncation of the G␣ i1 N terminus (41). Furthermore, the presence of G␤␥ reduced G␣ activation, which could be overcome by increasing the receptor-mimetic peptide concentration. This shows that (i) the peptide can directly target its binding site on G␣, which was mapped to be CT␣ in close proximity to the ␣N/␤1-loop, and (ii) G␤␥ contributes to shielding of CT␣. In the present study, G␣ and heterotrimeric G protein lacking hydrophobic modifications failed to interact with light-activated rhodopsin. We thus conclude that, in contrast to the receptor-mimetic peptide, the corresponding endogenous cytoplasmic third loop fragment of rhodopsin cannot directly target CT␣. Accordingly, this interaction would require the above described initial docking via the G protein's amphiphilic microdomain and subsequent conformational changes within the receptor.
Model of R*/G Protein Coupling-Our data are consistent with a coupling mechanism in which an initial contact is established between R* and a microdomain on the G protein composed of the hydrophobic modifications of G␣ and G␥ and adjacent N-terminal residues (Fig. 6). This microdomain therefore not only constitutes the membrane anchoring site of the G protein but also serves to establish a docking complex with R*. This complex in turn allows the G protein to sample different orientations until receptor loops can engage with the ␣N/␤1-loop region to lower nucleotide affinity and to unleash the CT␣ key binding site. Eventually, the interaction of CT␣ with R* triggers GDP release and weakens the contact between the initial R*/G protein docking sites (15). A gain in fidelity of signal transduction might be the reason for the multiple steps involved in G protein activation. Future work will aim at testing this working model with G proteins containing all native hydrophobic modifications.
An outcome of the present study as well as a study by others (41) is that G␤␥ is not unconditionally required for nucleotide exchange catalysis. A G␤␥-independent G␣ activation could allow the cell to maintain intracellular signaling even in the case where G␤␥ is involved in effector interaction. Notably, the mechanism proposed above can also be used to describe G␤␥-independent activation of G␣: initial contact between R* and the hydrophobically modified N-terminal region of G␣ enables unleashing of CT␣ and subsequent CT␣ interaction with R*. However, lack of G␤␥ in the amphiphilic microdomain employed in the docking step decreases efficiency of R*/G protein coupling. The question remains whether in the course of nucleotide exchange catalysis G␤␥ also directly affects nucleotide affinity. Models were proposed in which, upon receptor/G protein interaction, (i) G␤␥ undergoes a conformational switch that creates an escape route for GDP (19) and (ii) G␤␥ induces conformational changes in switch regions of G␣ to destabilize nucleotide binding (12,22,25). Such contributions by G␤␥ could improve nucleotide exchange catalysis, although CT␣ and the preceding ␣5 helix are considered to be the key elements needed to prompt nucleotide release from the "back side" of the nucleotide binding pocket (13,14,28,41,65). Our sequential fit mechanism of G protein activation includes a transitory state with multisite interaction between R* and the G protein (Fig. 6) (15, 30). This state enables GDP release and accounts for contributions by G␤␥ to enhance GDP release.
It should be further noted that the functional unit of active rhodopsin is not yet clear (e.g. see Refs 17 and 26). However, our docking concept is compatible with various scenarios (Fig. 6B). Among them are docking between (i) G t and an activated rhodopsin monomer, (ii) G t and a preformed rhodopsin dimer or higher oligomer with at least one activated rhodopsin molecule, and (iii) a preformed G t ⅐rhodopsin complex with an activated rhodopsin monomer. It is further conceivable that docking between G t and an activated rhodopsin monomer induces receptor dimerization or oligomerization.
As to the docking site on rhodopsin, previous work suggests a potential hydrophobic site comprised of the NPXXY(X) 5,6 F motif in transmembrane helix VII and cytoplasmic helix VIII of rhodopsin that includes the palmitoylated Cys 322 and Cys 323 (44, 66 -68). The crystal structure of the rhodopsin ground state revealed two lipid molecules bound close to this region (8). Furthermore, it was shown that rhodopsin/lipid interactions change upon rhodopsin illumination (69,70). It is conceivable that upon this change, a docking site for the G protein's amphiphilic microdomain is formed. A high resolution structure of the R*⅐G protein complex should provide information regarding specific interactions between GPCR and G protein as well as structural alterations enabling nucleotide exchange. However, as this work shows, interaction between the two proteins requires several steps to be productive. Inherent problems with crystallizing such a complex may be attributable to the dynamic nature of signal transfer. Step 2, interaction between the receptor and the ␣N/␤1-loop region of G␣ (2) makes the CT␣ binding site (3) available for binding to R*.
Step 3, R* interaction shifts from the G protein docking site (1) to CT␣ (3) concomitant with GDP release. The dotted lines represent nonresolved parts in the crystal structure. B, potential candidates for the docking step: R* monomer and G␣␤␥, preformed R*/R dimer and G␣␤␥, and R* and preformed R/G␣␤␥. The relative orientation between G␣␤␥ and membrane/GPCR is suggested by the electrostatic potentials of the corresponding G protein surface (29) and residual dipolar couplings of solution NMR measurements with the CT␣ fragment in the R*-bound state (33).