Structural and kinetic modeling of an activating helix switch in the rhodopsin-transducin interface

Abstract

Extracellular signals prompt G protein-coupled receptors (GPCRs) to adopt an active conformation (R*) and catalyze GDP/GTP exchange in the α-subunit of intracellular G proteins (Gαβγ). Kinetic analysis of transducin (G_tαβγ) activation shows that an intermediary R*·G_tαβγ·GDP complex is formed that precedes GDP release and formation of the nucleotide-free R*·G protein complex. Based on this reaction sequence, we explore the dynamic interface between the proteins during formation of these complexes. We start from the R* conformation stabilized by a G_tα C-terminal peptide (GαCT) obtained from crystal structures of the GPCR opsin. Molecular modeling allows reconstruction of the fully elongated C-terminal α-helix of G_tα (α5) and shows how α5 can be docked to the open binding site of R*. Two modes of interaction are found. One of them – termed stable or S-interaction – matches the position of the GαCT peptide in the crystal structure and reproduces the hydrogen-bonding networks between the C-terminal reverse turn of GαCT and conserved E(D)RY and NPxxY(x)_5,6F regions of the GPCR. The alternative fit – termed intermediary or I-interaction – is distinguished by a tilt (42°) and rotation (90°) of α5 relative to the S-interaction and shows different α5 contacts with the NPxxY(x)_5,6F region and the second cytoplasmic loop of R*. From the 2 α5 interactions, we derive a “helix switch” mechanism for the transition of R*·G_tαβγ·GDP to the nucleotide-free R*·G protein complex that illustrates how α5 might act as a transmission rod to propagate the conformational change from the receptor-G protein interface to the nucleotide binding site.

G protein coupled receptors (GPCRs) use the free energy of agonist binding to transmit physical or chemical signals into the cell. Bound agonists stabilize the 7 transmembrane (7TM) helix bundle of the receptor in an active conformation (R*). R* in turn interacts with intracellular heterotrimeric G proteins (Gαβγ, G) to catalyze the exchange of GDP for GTP in the Gα subunit and thus activate downstream effectors. According to classical receptor theory, R* is also stabilized by G protein binding (1). We used a synthetic peptide derived from the C terminus of the α-subunit of transducin (GαCT), the key binding site for the receptor (2, 3), to stabilize and crystallize the R* conformation of opsin (4, 5). Opsin is the ligand-free form of the photoreceptor rhodopsin, and transducin (G_tαβγ, G_t) is its cognate G protein. X-ray structure analysis of the R*·GαCT complex revealed that the cytoplasmic side of the TM5/TM6 helix pair (corresponding to the cytoplasmic loop C3 of the 7TM bundle) forms a mitt-like structure in which GαCT is held. The contacts with R* induce in GαCT an α-helical conformation with a C-terminal reverse turn (C-cap) (4, 6, 7), which is recognized by the receptor on the basis of its geometry (4).

In GPCR mediated signal transduction, the signal is eventually established in the GTP-bound form of the G protein, which is the form that activates downstream effectors. The key step in which the signal transits the membrane is represented by formation of the receptor-G protein complex, with the nucleotide binding site empty and ready for uptake of GTP (R*·G_t[empty]) (8, 9). In the absence of GTP or GRP, the R*·G_t[empty] complex is stable (10–12). The R*·GαCT structure likely shows part of the R*·G_t[empty] complex (4), consistent with earlier EPR spectroscopic work (13). Comparison with the G_t crystal structure (14) indicates a rotational and translational movement of the G_tα C-terminal α5 helix in R*·G_t[empty], which is a structural perturbation in the G protein necessary for GDP release (4, 13, 15–17). Kinetic analysis has suggested, and it will be further confirmed in this study, that GDP release is triggered by conversion of an intermediate complex (R*·G_t·GDP) into the nucleotide-free R*·G_t[empty] complex (18). It is the objective of this study to explore the structure of the R*/G_t interface in the R*·G_t·GDP and R*·G_t[empty] complexes. Because the intermediate complex does not accumulate under realistic biochemical conditions, crystallization and x-ray analysis cannot be employed. Therefore, we used a computational modeling approach to gain insight into the dynamic changes of the R*/G_t interface linked to the conversion of R*·G_t·GDP to R*·G_t[empty].

The first step in this investigation was the elongation of the GαCT peptide and the reconstruction of the G_tα C-terminal α5 helix, to provide the necessary substrate for docking α5 to R*. Two well defined docking modes were found, one reproducing the mode of interaction seen in the R*·GαCT crystal structure and assigned to the R*·G_t[empty] complex and a second one assigned to the R*·G_t·GDP complex. Together with information on cytoplasmic rhodopsin loops interacting with G_t (19, 20), a model for the transition of R*·G_t·GDP into R*·G_t[empty] could be derived. The proposed “helix switch” mechanism supports and elaborates the idea of the α5 helix acting as a transmission rod (4, 13, 15, 16). It describes its role as a key element in the propagation of the conformational change from the receptor-G protein interface to the nucleotide binding site of the G protein and the release of GDP.

Results

Kinetic Analysis: Intermediary Interaction Between G_t·GDP and R*.

To analyze whether a R*·G_t·GDP complex does really exist, the initial rate of R* catalyzed G_t activation as a function of the concentration of G_t·GDP, GTP and GDP was evaluated (18). The classical Ping Pong (or double displacement) reaction sequence contains an intermediate R*·G_t·GDP complex with finite lifetime (Scheme 1) (18). In an alternative model in which the Ping Pong sequence is combined with the “hit and run” feature of the Theorell-Chance mechanism (21), the lifetime of R*·G_t·GDP is so short that the concentration of this complex is essentially 0 (Scheme 2).

Scheme 1.

Scheme 2.

The 2 rivaling mechanisms are clearly distinct when the concentration of GDP is not 0. As Fig. 1A shows, the maximum rate of R* catalyzed G_t·GDP activation is not approached in the presence of GDP even at infinite G_t·GDP concentrations, indicating that GDP is not acting competitively. The noncompetitive inhibitory effect of GDP is clearly apparent in a replot of the data by using the Hanes-Woolf method (Fig. 1B). With increasing concentration of GDP, the slopes of the curves increase, which is in contrast to the parallel product inhibition pattern predicted for reaction Scheme 2 (21). This demonstrates that the binding of R* to G_t·GDP and formation of R*·G_t[empty] with concomitant GDP release is separated by an intermediary R*·G_t·GDP complex with finite lifetime.

Fig. 1.

Kinetic analysis of R* catalyzed activation of G_t·GDP by nucleotide exchange. (A) Steady state activation rate of G_t·GDP (ν) as a function of initial concentration of membrane-bound G_t·GDP (selected data taken from (18) and normalized to R* concentration). Titrations of G_t·GDP in the presence of 200 μM GTP and without added GDP (black symbols), 750 μM GDP (red symbols), and 2 mM GDP (green symbols), respectively. Different symbols identify different sets of experiments. Black solid lines are best fit of the data to rate Eq. 1 (Ping Pong model); dotted line plots Eq. 1 with the same kinetic constants but with 2.38 mM GDP. Blue dashed lines are best fit of the data to rate Eq. 2 (Theorell–Chance model (see SI Appendix). (B) Hanes-Woolf plot of the data shown in A. Symbols, line style, and color code as in A.

Structural Analysis: Molecular Modeling of the Interaction Between the α5 Helix of G_t and R*.

The starting point for structural modeling of the R*/G_t interface was the R* conformation present in the crystal structure of active ligand-free opsin in complex with the synthetic 11mer GαCT peptide (G_tα_340–350) (4). The GαCT structure was used to reconstruct a short variant (G_tα_332–350) of the full native G_tα C-terminal α5 helix (G_tα_325–350) for docking studies to R* (see ref. 4, Methods, and SI Appendix). The partially flexible docking of α5 to R* by means of the program GOLD (22) resulted in 2 distinct docking modes obtained from cluster analysis (Fig. S1). These 2 modes were termed (stable) S-interaction and (intermediary) I-interaction, respectively (Fig. 2). The S-interaction was termed so because it matches the binding of the GαCT peptide in the R*·GαCT crystal structure. This close similarity also includes the potential hydrogen bonding networks with the conserved E(D)RY (Arg-135 to main chain O of α5-Cys-347) and NPxxY(x)_5,6F (Gln-312 to main chain O of α5-Lys-345) motifs of the receptor (Fig. 2B Right). The tilt angle of docked α5 relative to the membrane plane is 43.4° (see SI Appendix) and thus <0.5° different from the angle of the shorter GαCT peptide in the crystal structure (Fig. 2A Right). An alternative fit of α5 to R* is found in the I-interaction, which relative to the S-interaction is 90.0° axially rotated and 42.4° tilted such that α5 is oriented nearly parallel to the membrane (1.0° tilt angle between α5 and membrane) (Fig. 2A Left). The C-cap of α5 forms a hydrogen bonding network to R*, which includes the NPxxY(x)_5,6F motif (Gln-312 to main chain O of α5-Cys-347) and the loop connecting TM7/helix8 of R* (Lys-311 to main chain O of α5-Phe-350) (Fig. 2B Left).

Fig. 2.

Two docking modes of the reconstructed α5 helix of G_tα to the active conformation of rhodopsin (R*). The Left row shows α5 in the intermediary or I-interaction, the Right row shows the stable or S-interaction. All R* structures are shown in orange. (A) Overall structure of R* is shown in cartoon/surface representation. C-terminal α5 helix with its C-cap is shown as cartoon/stick in green in the I-interaction and in blue in the S-interaction. (B) Close up view of the cytoplasmic domain of R* with potential hydrogen bonds to α5 helix; color codes as in A. (C) Close-up of the cytoplasmic domain of R* with potential hydrophobic contacts to α5. Contacts are shown in stick/dot representation in yellow, and the side chains of α5 are omitted. The R* conformation of rhodopsin was obtained from the crystal structure of active, peptide-bound opsin (PDB entry 3DQB). Energetically minimized docking modes were found by means of the docking software GOLD (see ref. 22 and Methods).

The hydrogen bonding network at the far C terminus of α5 exclusively involves main chain hydrogen bonds in both interaction modes (Figs. S2 and S3). The precise geometry of the C-cap is mandatory for the recognition and binding of α5 to R* (4). Here, we report additional hydrogen bonds between the extended moiety of α5 and R*, which may further specify the interactions. In the S-interaction, a hydrogen bond is formed between Asn-343 at the α5 C terminus (α5-Asn-343) to main chain O of Val-138 in loop C2 (connecting TM3 and TM4). Toward the N terminus of α5, a side chain hydrogen bonding network extends from α5-Asp-337 and α5-Lys-341 to Ser-240 and Thr-243 from the mitt-like structure formed by loop C3, respectively (Fig. 2B Right and Figs. S2 and Figs. S4). In the I-interaction, a hydrogen bonding network is formed from the N-terminal α5-Asp-333 to Glu-232 in loop C3 and Thr-229 in TM5 of R* (Fig. 2B Left and Figs. S3 and S5).

In the S-interaction, 40% (1130 Ų) of the solvent accessible surface of α5 is buried. Hydrophobic contacts are formed between α5 (α5-Asp-333, α5-Asp-337, α5-Ile-340, α5-Asn-343, α5-Leu-344-Phe-350) and loops C1 (Leu-72) and C2 (Val-138, Val-139), with the hydrophobic belt of the mitt-like structure formed by TM5, TM6 and loop C3 (Leu-226, Thr-229, Val-230, Ala-233, Gln-236, Gln-237, Thr-243, Lys-245, Ala-246, Val-250, Met-253), Arg-135 of TM3, and Asn-310 of the loop connecting TM7/helix8 (Fig. 2C Right). This type of interaction is consistent with experimental results that have already revealed a hydrophobic contribution to R*-GαCT interaction (23). In the I-interaction, only 31% (892 Ų) of the solvent accessible surface of α5 is excluded from the solvent. Hydrophobic contacts are formed between the α5 helix (α5-Asp-333, α5-Thr-336, α5-Asp-337, α5-Ile-338, α5-Ile-340, α5-Lys-341, α5-Gly-348-Phe-350) and cytoplasmic loops C2 (Val-139, Lys-141) and C3 (Val-230, Ala-233, Gln-237, Ala-246), Thr-229 of TM5, and Gln-312 of the loop connecting TM7/helix8 (Fig. 2C Left). The interfaces of both interactions are well packed (see SI Appendix), although local packing defects were found by means of Voronoia (24). Such defects, which may arise from interfacial water molecules, are preferentially found between the floor of the binding crevice and the C-cap of α5 (Figs. S2B and S3 B and C), and between the C3 loop and the central part of α5 (Fig. S2C).

Discussion

We have combined kinetic and structural modeling approaches to study the steps of interaction between activated rhodopsin (R*) and transducin (G_t) that are necessary to trigger the release of GDP from the nucleotide binding site in G_t. The kinetic analysis has shown that an intermediary complex between the proteins is formed, in which the G protein has not yet lost its GDP, i.e., both the receptor and the GDP nucleotide are bound to the G protein. We term this complex R*·G_t·GDP, to distinguish it from the stable R*·G_t[empty] complex, in which the GDP is no longer bound to its binding site. Consistently, the molecular docking experiments have identified 2 stable modes of interaction between R* and the C-terminal α5 helix of G_tα, termed S- and I-interaction, respectively. The S-interaction corresponds to that seen in the crystal structure of opsin-GαCT and was assigned to the R*·G_t[empty] complex (4). In agreement with previous results from EPR studies (13), we now correlate the I-interaction with the R*·G_t·GDP complex. We will discuss that this interaction is a key element in the dynamic R*/G_t interface, which determines the structural changes that eventually lead to GDP release.

The Dynamic Receptor-G Protein Interface in R*·G_t·GDP.

Consistent with the assignment of R*·G_t·GDP to the I-interaction, superposition of GDP-bound G_t (PDB accession: 1GOT) to α5 in the I-interaction does not cause major clashes or distortions of the proteins. G_t lies parallel to the lipid bilayer such that the 2 lipid anchors – a myristoyl moiety at the G_tα N terminus and a farnesyl moiety at the G_tγ C terminus – interact properly with the membrane (see Methods, Fig. 3C, and ref. 3). Only the highly flexible (25, 26) loop C2 of R* needs to be rearranged to avoid a clash with G_t. We applied a search routine to compute low energy C2 loop conformations taken from known protein structures (see SI Appendix) (53). The best scored loop structure displays a conformation between the C2 structures of rhodopsin or opsin (4, 27, 28) on the one side and the β₁-and β₂-adrenergic receptors (29–31) or the A_2A-adenosine receptor (32) on the other side (Fig. 3A, SI Appendix, and Fig. S6). With this C2 conformation, a double-sandwich structure forms with alternating elements from G_tα and R*, namely β1-β3 half-barrel of G_tα / C2 loop of R* / α5 helix of G_tα / mitt-like structure of R* provided by loop C3 (Fig. 3A).

Fig. 3.

Dynamics of R*·G_t·GDP and R*·G_t[empty] complexes. The complexes modeled by superposition of G_t·GDP (green or blue) (PDB accession 1GOT) and R* (red) with bound reconstructed C-terminal α5 helix (green or blue). (A and B) Close up view of the R*/G_t interface in the I-interaction (A) and in the S-interaction modes (B), respectively. Helices are depicted as tubes. Yellow arrows and the Inset illustrate the transition from I- to the S-interaction. (C) Conceptual model showing the transition from the R*·G_t·GDP to the R*·G_t[empty] complex. The G_t body, which is fixed to the lipid bilayer by its lipid anchors, undergoes a rotation on the membrane of 40–50° (arrow) during the transition from the I- to S-interaction. The resulting distortion within the β6–α5 loop, which is involved in the binding of the GDP guanine-ring, leads to reduced GDP affinity. Proteins are depicted in cartoon representation with color codes as in A and B. GDP and lipid anchors of G_t are shown space filled.

With the double-sandwich structure, an interface between receptor and G protein within the intermediary R*·G_t·GDP complex is provided. The question arises of how this complex structure can form. Electrostatics govern the attraction between the negatively charged G_tα C terminus and the positively charged active receptor surface (4, 28) (Fig. S3D). The last helix turn and the C-cap of α5 are induced on receptor contact (4, 6, 7). Concomitant with progressive desolvation, the curved α5 helix will straighten (33). This straightening, together with the intercalation of C2, facilitates α5 release from the β1-β3 half-barrel, e.g., by disconnecting the hydrogen bond between Lys-188 in the β1-β3 half-barrel and α5-Asp-333 and α5-Asp-337 in α5. The intermediary interaction between the α5 C-cap and the kink between TM7 and cytoplasmic helix 8 would then serve to hold α5 for its further transition into the S-interaction mode (Fig. 3A).

Coupling Between the R*/G_t Interface and the Nucleotide Binding Site.

Based on comparison of the modeled R*·G_t·GDP and R*·G_t[empty] complexes, we can now propose a mechanism (helix switch) of how α5 undergoes a rotational and translational movement, which eventually distorts the GDP binding site to cause GDP release. Starting from the detachment of α5 from the β1-β3 half-barrel by loop C2, α5 rotates counterclockwise around its axis by 90° and tilts relative to the membrane normal by 42° (Fig. 3 A and B). This movement is supported and directed by the formation of new hydrogen bonds between α5 and R*. Such interactions arise from the backbone carbonyl of α5-Cys-347 at the C-cap to Arg-135 at the polar floor of the R* binding crevice and from α5-Asp-337 and α5-Lys-341 to Thr-243 and Ser-240 in the mitt-like structure of R*, respectively. Our analysis further indicates that these hydrogen bonding networks are extended by interfacial water (Figs. S2 and S3), which is known to facilitate the transition between different conformational states of proteins (34–36). Thus, interfacial water may allow the necessary sliding motion of α5 in its crevice. The final conversion into R*·G_t[empty] is then stabilized by optimized well packed hydrophobic interactions of α5 with the mitt-like structure of R* as concluded from our packing analysis (see SI Appendix).

When the α5 C-cap approaches the floor of its binding crevice in R*, the β1-β3 half-barrel is forced away from R* loop C2 and thus tilted sideward. As a possible consequence, the G_t body, which is fixed to the lipid bilayer by its lipid anchors, undergoes a rotation on the membrane away from the C2 loop by ≈40–50° (Fig. 3C). These structural changes may result in a distortion of the β6–α5 loop that has properties of a flexible hinge (see Methods and ref. 13) and is involved in the binding of the GDP guanine-ring (37). Fig. 3B shows the resulting geometry in the R*·G_t[empty] complex. The model is in agreement with photo cross-linking experiments in which a photoactivated reagent attached at Ser-240 in loop C3 of R* cross-linked predominantly to G_tα sequence 342–345 (38). At the protein concentrations used in the experiment, a small fraction of the R*·G_t·GDP intermediate complex may have been present in equilibrium, explaining the second weaker cross-link to G_tα sequence 310–313 in these experiments.

Signal Transfer from Receptor to G Protein.

Several models of signal transfer from the activated receptor to the G protein have been designed and are reviewed in refs. 3 and 39. The “gear shift” (40) and “lever arm” (41) models have in common that the Gα- and Gβγ-subunits anchor simultaneously to the receptor and the membrane and build up a force flow to operate a switch for the release of GDP. In the alternative “sequential fit” model, the 2 spatially distant binding sites on G_t, namely the Gγ C terminus with its farnesyl anchor and the Gα C terminus, act sequentially (42). The Gγ C terminus has a function in the initial encounter interaction, which is thought to depend on proper membrane anchoring of the G protein heterotrimer by electrostatics and lipid modifications (12, 43, 44). The actual catalytic interaction, which leads to nucleotide exchange, occurs between the Gα C terminus and R*.

Independent of specific overall models of receptor-G protein signal transfer, the α5 helix has been identified as a key transmission element (4, 13, 15, 16, 45). The helix switch mechanism described in this study refines the α5-transmission rod concept and assigns 2 basic modes of receptor interaction. The helix switch affords the integration of experimental data from studies of signal transduction into a structural concept. The helix switch mechanism also explains the crucial role of the TM7/helix 8 region (46–49) and why both loops C2 (I-interaction) and C3 (I- and S-interaction) are required to activate the G protein. It was found that impairment in either loop allows binding of the G protein but impedes GDP release and thus G protein activation (19, 20). Our model (Fig. 3) is also consistent with the evidence that the reduction of the affinity of G_t for GDP depends on the integrity of the αN/β1-loop region in G_tα (42). Further experimental work, which may include structural information on the metarhodopsin II photointermediate or analogous fully active conformations in other G protein coupled receptors, will be needed to elaborate the model and to find out how the functional core that it provides is linked to structural rearrangements within other domains of the G protein.

Materials and Methods

Using the steady state approach (18), rate equations were derived for the initial rate of R* catalyzed G_t·GDP activation (v). These equations explicitly account for the concentrations of G_t·GDP, GTP, and GDP (for details see SI Appendix). Eq. 1 refers to the Ping Pong scheme (see Results, Scheme 1), and Eq. 2 refers to the Theorell–Chance scheme (Results, Scheme 2).

[G_t GDP] is the initial surface density of membrane bound G_t·GDP, and [GDP] and [GTP] denote the volume concentrations of the respective nucleotide (see SI Appendix for definition of the kinetic constants).

The α5 helix (G_tα_325–350) was fully reconstructed by an N-terminal elongation of the 11 aa synthetic GαCT peptide (G_tα_340–350/K341L) of the R*·GαCT crystal structure (see SI Appendix). An analogous reconstruction was done with the native G_tα_340–350 peptide. . For docking experiments, a shortened version of the α5 helix (G_tα_332–350) was taken in its R*-interacting conformation, with all backbone atoms including the fully induced C-cap kept fixed (see SI Appendix). Up to that length α5 specifically interacts with the receptor over its entire length (see Fig. 2B and Figs. S1–S3). The side chains were allowed to adapt to the receptor during the docking process. No major conformational changes of the receptor were allowed during the entire docking process except for Lys-141, Glu-232, Gln-236, Gln-237, Glu-239, Ser-240, Thr-242, Thr-243, Lys-311, and Gln-312, located along the rim of the GαCT binding pocket.

The partial complexes R*·G_tα_332–350 were selected after a hierarchical cluster analysis implemented in GOLD (Fig. S1). The initial R*·G_t complexes were obtained by the superposition of backbone atoms of G_t (PDB accession: 1GOT) with a fully reconstructed and straightened α5 helix (see ref. 33) to the partial complexes R*·G_tα_325–350. The complexes were further modified as described in Discussion and energetically minimized with help of the GROMOS 43B1 force field to avoid distorted geometries (50). Properties of the β6–α5 loop as a flexible hinge region for movements within G_t was identified with help of FlexOracle (51). Membrane anchors were attached to the fully reconstructed N and C termini (in extended conformation) of G_tα and G_tγ, respectively. The calculations of helix axis, angles, and packing densities were performed as described in refs. 34 and 52 and in SI Appendix. The figures were drawn by using the PyMOL visualization software (http://www.pymol.org).

SI.

Further information, including Table S1 and Table S2, can be found in the Supporting Information.