Visual Rhodopsin Sees the Light: Structure and Mechanism of G Protein Signaling*

The availability of crystal structures for the dark, inactive, and several light-activated photointermediate states of vertebrate visual rhodopsin has provided important mechanistic and energetic insights into the transformations underlying agonist-dependent activation of this and other G protein-coupled receptors (GPCRs). The high natural abundance of rhodopsin in the vertebrate retina, together with its specific localization to the disk membranes of the rod cell, has also enabled direct imaging of rhodopsin in its native environment. These advances have provided compelling evidence that rhodopsin, like many other GPCRs, forms highly organized oligomeric structures that, in all likelihood, are important for receptor biosynthesis, optimal activation, and signaling.


G Protein-coupled Receptors
G protein-coupled receptors (GPCRs) 3 are by far the largest class of cell surface signal transducing receptors. These heptahelical integral membrane proteins are of immense importance in human physiology and health as evidenced by the fact that virtually every cell type expresses a subset of GPCRs. With the sequencing of the human genome complete, it is now abundantly clear that upwards of 950 genes encode GPCRs (1). The mammalian GPCRs are typically grouped by similar amino acid sequences into the three distinct families: A, B, and C. For most GPCRs, the external stimulus (agonist) leading to receptor activation is a small molecule that binds to a region of the transmembrane (TM) domain and triggers conformational changes that are transmitted to the cytoplasmic surface of the receptor to facilitate the activation of intracellular heterotrimeric guanine nucleotide-binding proteins (G proteins). Although it is widely believed that most GPCRs transduce their activating sig-nals via heterotrimeric G proteins, recent studies have also highlighted a role for G protein-independent signaling by GPCRs (2). Because GPCRs modulate an extremely wide range of physiological processes, it is not surprising that mutations in the genes encoding many of these receptors have been implicated in numerous disease states. As such, these receptors also form the largest class of therapeutic targets.
In vertebrate vision, rod cell rhodopsin, a Family A GPCR member, serves as the dim-light receptor. Remarkable advances in our understanding of rhodopsin structure, function, and signaling that remain at the forefront of GPCR research have been realized in recent years. Here, we review developments in rhodopsin research with regard to GPCR structure and activation, its propensity to form higher order oligomers, and the structural basis for the interaction with heterotrimeric G proteins.

Rod Outer Segment and Rhodopsin
Retinal rod cells are specialized neurons that detect photons and communicate with secondary neurons about the presence of light. They are so exquisitely sensitive that even a single photon can be detected (3). Rods have highly differentiated outer segments (ROS) connected to their inner segments. In the ROS, hundreds of stacked disk membranes are enveloped by the plasma membrane. Rhodopsin is the predominant membrane protein of the disk membranes and occupies ϳ50% of the disk surface area, the remainder of which is filled with phospholipids and cholesterol.
Rhodopsin is composed of a membrane-embedded chromophore, 11-cis-retinal, which is covalently bound to the opsin protein at Lys-296 in TM helix VII via a protonated Schiff base ( Fig. 1). Upon absorption of a photon, isomerization of the chromophore to an all-trans conformation induces changes in the opsin structure, ultimately converting it from an inactive to an activated signaling state ( Fig. 1 and supplemental Fig. S1). This form of the receptor, known as metarhodopsin II (MII) or R * , relays the activating changes to the retinal G protein, transducin (G t ), initiating a biochemical cascade of reactions in a process termed phototransduction. Numerous soluble and integral membrane proteins present in ROS are critical for amplification and conversion of the light signal (4).
Rhodopsin expression is essential for the formation of ROS, which are absent in knock-out (rhodopsin Ϫ/Ϫ ) mice (5). The ROS of heterozygous mice for the rhodopsin gene deletion (rhodopsin ϩ/Ϫ ) have a similar density of rhodopsin, but their rhodopsin volume is reduced by ϳ60% when compared with wild-type mice (6). Synthesis of opsin begins in the inner segments of photoreceptors, where it undergoes maturation in the endoplasmic reticulum and Golgi membranes before being vectorially transported to the ROS. The formation of rhodopsin from opsin and its 11-cis-retinal chromophore does not appear to be essential for transport to the ROS as mice deficient in chromophore production still develop ROS (7).

Structure of Rhodopsin
A three-dimensional structure of rhodopsin was first determined from diffraction-quality crystals of detergent-solubilized bovine ROS rhodopsin (8). A refined (2.8 Å) model of rhodopsin included greater than 95% of the amino acid residues as well as post-translational modifications (9). Alternate crystal forms (10) and improvements in crystallization conditions (11,12) have since yielded a 2.2-Å structure for rhodopsin that resolves the complete polypeptide chain ( Fig. 2A). One of the initial surprises from the crystal structure was a compact extracellular (intradiscal) arrangement, parts of which fold inward to enclose the 11-cis-retinal chromophore. This "plug" over the retinal binding pocket involves the second extracellular loop. Cys-187 of this loop forms a disulfide bond with Cys-110 near the extracellular surface of TM helix III. The amino terminus contains two asparagine-linked (N-linked) oligosaccharides attached to Asn-2 and Asn-15 ( Fig. 2A). Neither of these oligosaccharide chains appears to make structural contacts with the protein.
The cytoplasmic domain, in contrast to the extracellular domain, is not as compact and highly organized. In fact, the crystallographic B-factors for this region, which reflect freedom of movement of atoms, vary considerably among the various inactive, dark-state crystal structures (13). This observation can be related to the apparent plasticity of the region likely required for function. One exception is helix 8 (H8), an amphipathic helix that lies almost perpendicular to the carboxyl-terminal end of TM helix VII ( Fig. 2A). The palmitoyl chains attached to Cys-322 and Cys-323 ( Fig. 2A), which anchor H8 to the membrane, also display resolvable density.
The TM helices are the sites where a majority of the conserved amino acid residues are found in rhodopsin and other GPCRs. For example, Asn-55 of TM helix I, Asn-78 and Asp-83 of TM helix II, and Asn-302, Pro-303, and Tyr-306 of TM helix VII (part of the NPXXY(X) 5,6 F motif) are all highly conserved amino acids across the GPCR family that appear to form important contacts for maintaining the inactive state of rhodopsin. In the dark, rhodopsin contains an 11-cis-retinal chromophore that is attached to Lys-296 of TM helix VII in a protonated Schiff base linkage. Upon absorption of a photon, the chromophore is isomerized to an all-trans conformation. The first structurally characterized photointermediate, bathorhodopsin, thermally relaxes to the BSI, followed by lumirhodopsin, and then MI. At MI, the all-trans-retinal chromophore remains bound in a protonated Schiff base linkage. During the MI to MII transition, the all-trans-retinylidene Schiff base becomes deprotonated. MII, the signaling state capable of G protein activation, ultimately decays to free all-trans-retinal and opsin. The max for rhodopsin, MI, MII, free all-transretinal, and opsin, as well as the time scales for the various transformations, are indicated. Note that some photointermediates have not been included for simplicity. For example, conserved residues from the NPXXY(X) 5,6 F motif form a set of interactions that help stabilize the cytoplasmic domain ( Fig. 2B and Refs. 8, 9, and 14). Another important conserved element is the (E/D)RY motif at the cytoplasmic end of TM helix III. Arg-135 forms a network of interactions with Glu-134 of this motif and Glu-247 and Thr-251 at the cytoplasmic end of TM helix VI to provide additional stability to rhodopsin (Fig. 2C). However, the electrostatic interaction between the protonated Schiff base counterion, Glu-113, and Lys-296 (Fig. 2D), residues that are conserved only among visual pigments, is considered to be the main factor in maintaining the inactive, dark-state conformation of rhodopsin (15).

Higher Order Organization of Rhodopsin
Oligomerization of GPCRs is now a widely accepted phenomenon. For most GPCRs, evidence for oligomerization has been obtained from indirect immunoprecipitation, cross-linking, size-exclusion chromatography, and fluorescence or bioluminescence resonance energy transfer studies (16). For rhodopsin, however, direct evidence for oligomerization has been obtained by both atomic force microscopy (AFM) and electron microscopy (17,18). The paracrystalline arrangement of rhodopsin dimers in the disk membrane provides evidence for how rhodopsin can be self-organizing and provides a platform for signal propagation and/or desensitization (18,19). A current model for the rhodopsin dimer, the IV-V model, proposes a dimeric interface between helices IV and V (Fig. S2, I). Experimental evidence supporting such an arrangement for the inactive state of Family A GPCRs has been shown from cross-linking studies on the dopamine D2 receptor (20). In a new crystal form that remains stable upon light activation (21), rhodopsin forms a potentially physiologically relevant dimer interface that involves TM helices I and II, and H8 (Fig. S2, II and III), and when taken with the prior work that implicates TM helices IV and V as the physiological dimer, this can account for one of the interfaces of the oligomeric structure of rhodopsin seen in the membrane by AFM. This mode of dimerization can be responsible for the stability of these crystals.

Rhodopsin Photointermediate Energetics and Structure
Light-induced isomerization of 11-cis-retinal to its all-trans conformation (Fig. 1) is the primary step in vision. This ultrafast (femtosecond) photochemical process is followed by much slower events culminating in increased accessibility of the alltrans chromophore to the aqueous environment with subsequent hydrolysis of the Schiff base linkage (22). These two events in rhodopsin photolysis are separated by several thermally stable dark processes resulting in photoproducts with defined spectral, kinetic, and functional properties that can be trapped at different temperatures (23).
The first structurally characterized stable photointermediate detected is bathorhodopsin, which thermally relaxes to the blue-shifted intermediate (BSI), followed by lumirhodopsin, metarhodopsin I (MI), and MII (Fig. 1). Of particular interest in this sequence is MII, the active form of rhodopsin with its bound all-trans-retinal chromophore. Two forms of MII appear to exist, MIIa and MIIb, in a pH-dependent equilibrium, and only the latter is capable of activating G t (24). This equilib-rium is regulated by proton uptake at Glu-134 in the (E/D)RY motif (Fig. 2C). Similarly, evidence indicates that there are two forms of MI, MIa and MIb, and that the latter is also capable of binding but not activating G t (25). The positive enthalpy change (⌬H) accompanying MII formation (ϳ35 kcal/mol, Fig. S1) suggests that molecular interactions that exist in MI are lost upon the transition to MII. In the two-step reaction scheme for the MI-MII conversion (24), the transition of MI to MIIa is accompanied by translocation of the Schiff base proton to the Glu-113 counterion. Here, the all-trans-retinylidene chromophore acts as an agonist to facilitate the MI 3 MIIa transition (26). Subsequent uptake of a proton from the cytoplasm leads to the formation of MIIb. Alternatively, it is also possible that relaxation of the chromophore after photoactivation might change the geometry of the chromophore-binding site and lead to minor changes in the structure culminating in deprotonation of the Schiff base and subsequent changes that lead to specific protonation of side chain(s) on the cytoplasmic surface. Although the specific mechanism of further transformations for MII are not fully understood, two products typically detected are metarhodopsin III, which contains a protonated all-trans-retinylidene Schiff base, and opsin plus free all-transretinal (Fig. 1). Formation of these photoproducts appears to be influenced by the phosphorylation of MII by rhodopsin kinase and the subsequent binding of arrestin (27).
Three-dimensional crystal structures for bathorhodopsin and lumirhodopsin obtained after trapping these photolyzed states at low temperature, as well as a structural model for MI based on electron crystallography of two-dimensional crystals, have recently been determined (28 -30). Although these structural models are of moderate to low resolution (2.7-5.5 Å), a comparison of the inactive, dark-state rhodopsin structure with bathorhodopsin shows that the C11ϭC12 bond in retinal (see Fig. 2D) has adopted a trans configuration upon illumination. From bathorhodopsin to lumirhodopsin, a positional/conformational change in the ␤-ionone ring region of retinal (see Fig.  2D) is apparent but is minimal within the resolution. In the structure for MI that has been determined to lower resolution and corresponds to ϳ60% occupancy of this photointermediate state, there is not much difference in the position of the ␤-ionone ring relative to the ground-state structure. In addition, no large structural rearrangements in the protein are apparent in the structural model for MI relative to dark-state rhodopsin, raising the prospect that only modest conformational changes can accompany the MI 3 MII conversion (positive ⌬H).
A model for the structure of an intermediate containing a deprotonated Schiff base based on ϳ4.1 Å diffraction data obtained from illuminated rhodopsin crystals (21) reveals that the photoactivated structure determined from these yellow crystals is similar to that of the ground-state structure but has several significant alterations. Specifically, portions of the second and third cytoplasmic loops are found to be disordered in the photoactivated rhodopsin, and the density that can correspond to the ␤-ionone ring of the chromophore is close to residues Phe-212, Trp-265, Leu-266, and Tyr-268 in TM helices V and VI. These residues make direct interactions with the ␤-ionone ring as observed by solid state NMR studies of changes within the chromophore and surrounding residues upon acti-vation of rhodopsin to MII (31,32). Despite the low resolution of this "MII-like" photoactivated structure, relatively large displacements of individual TM helices, as predicted from EPR studies on spin-labeled cysteine mutants of rhodopsin (33,34), are not readily apparent and yet would be evident at this resolution. Although this lack of major structural rearrangements is consistent with thermodynamic values for the MI 3 MII transition ( Fig. S1 and Ref. 23) in conjunction with functional studies on disulfide constrained rhodopsin mutants (35), it is unknown how much the crystal lattice limits the magnitude of the changes observed in this and the other light-activated rhodopsin structures.

Mechanism of Rhodopsin Activation and Signal Transfer
Light activation of rhodopsin appears to rely on the disruption of inactive, dark-state contacts and replacement with a new set of interactions (36). This process is clearly directed toward creation of a highly specific and effective binding site for G t . Because opsin exhibits negligible activity at neutral pH (37), the isomerized retinal chromophore is considered to play the primary role in establishing these new interactions. Studies with retinal analogs indicate that both the C19 methyl group and ␤-ionone ring of retinal (see Fig. 2D) make important steric contributions to rhodopsin activation (26,38). The exact mechanism of activation is not fully understood, but the general view is that there is an isomerization-induced separation of the cytoplasmic ends of TM helix III and TM helix VI relative to each other (33,34). This departure is thought to be a consequence of charge separation and proton transfer from the all-trans-retinylidene Schiff base to the Glu-113 counterion (Fig. 2D), although a counterion switching mechanism at MI involving Glu-181 has also been proposed (39). The latter observation is at variance with solid state NMR and crystallographic results, indicating no major structural reorganization around the chromophore-binding site up to the MI stage (28,40). Nonetheless, local conformational changes arising from retinal isomerization appear to be propagated to the cytoplasmic surface causing greater structural changes coupled to the reprotonation and rearrangement around the (E/D)RY motif as well as coordinated changes in and around the NPXXY(X) 5,6 F motif (Fig. 2, A and B, and Refs. 15 and 41). In this context, light-induced changes along the dimeric interface(s) of rhodopsin, which remain to be evaluated, can also contribute to the mechanism of activation as has been proposed upon agonist activation of GPCRs (20,42).
The structure of the MII-like intermediate implies that activation can involve relaxation of the more rigid structure found in the inactive dark-state and suggests a rhodopsin-G t -induced fit model of interaction (21). The affinity of G t for the dark state of rhodopsin is high (ϳ50 nM) and does not change appreciably after photoactivation of rhodopsin (ϳ1 nM) (43). The fundamental change in the light-activated complex is that guanine nucleotide exchange can take place within G t but not upon interaction of the G protein with the inactive dark state of rhodopsin. Although much of the biochemical and biophysical work that has focused on light-activated rhodopsin/G t interactions for the past several years has assumed a 1:1 stoichiometry for the signaling complex (44), recent models based on the available surface areas for interaction suggest that complex formation can also be accommodated by an activating rearrangement of an oligomeric structure ( Fig. S3 and Refs. 18 and 19). Experimental evidence to support such a model comes from studies showing that the G t -activating potential of detergentsolubilized rhodopsin dimers exceeds that of monomer preparations (45) and trans-activation studies with other Family A GPCRs that employ fusion proteins between active and inactive receptors and G protein ␣-subunit (G ␣ ) subunits (46). The isolation of a pentameric complex between a dimeric leukotriene B4 receptor BLT1 and the G protein heterotrimer is also consistent with this view (47). Interestingly, the B4 receptors in the dimer are not equivalent, but rather each monomer in the complex assumes a unique conformation, further supporting the notion of an asymmetric arrangement for GPCRs in the dimer (48). This second hypothesis is also consistent with a model for the Family C metabotropic Glu receptor 1␣ for which ligand binding does not significantly change the structure of each monomer but does change the dimeric location of the cytoplasmic regions (49).
High-resolution NMR studies on isotope-labeled G ␣ -reconstituted heterotrimers have revealed that functionally important rhodopsin-induced changes in G ␣ structure and dynamics can be detected and characterized, enabling the generation of models for the global and local structural changes accompanying signal transfer from rhodopsin to the G protein. Specifically, correlated changes at the carboxyl terminus of G ␣ , a well known receptor-interacting site, and in the conformational flexibility of switch II in the guanine nucleotide binding pocket are evident during the course of light-activated rhodopsin-catalyzed GDP/GTP exchange (50). In addition, the trapping and interrogation of discrete activated receptor-bound conformations of G ␣ have not only revealed that GTP uptake can be uncoupled from heterotrimer release and subunit dissociation but that receptor-mediated changes in the interacting regions of G ␣ , and not the absence of bound guanine nucleotide, are the predominant factors that dictate G ␣ conformation and dynamics in the nucleotide-released receptor-bound "empty pocket" state of the heterotrimer (50,51). Subsequent EPR studies employing spin labels incorporated at specific sites in cysteine mutants of G ␣ are consistent with this more dynamic view of G ␣ upon interaction with light-activated rhodopsin and after nucleotide exchange (52,53). The NMR results also suggest that G ␣ in the activated rhodopsin-bound "empty pocket" heterotrimer state, as well as in the guanine nucleotide-exchanged state, do not exist as single, discrete conformations but rather as an ensemble of states. These findings may have functional implications for optimal signaling in phototransduction.

Future Directions
Progress in understanding the structure and function of rhodopsin has greatly benefited the fields of vision and GPCR research. The availability of various structures for the inactive and photointermediate states of rhodopsin and an appreciation for how rhodopsin can be highly organized in the native membrane cannot be understated given the preponderance of naturally occurring mutations that are manifested in such visual disorders as retinitis pigmentosa and congenital stationary night blindness. Although the determination of higher resolution structures for the photointermediate states of rhodopsin remains an active area of investigation, an immediate challenge is a highresolution structural model for the complex between MII and G t . A clear and comprehensive understanding of the mechanism underlying information transfer from an activated GPCR to a heterotrimeric G protein would undoubtedly provide valuable insights into the molecular basis for this key signaling event.