G Protein-Coupled Receptor Rhodopsin: A Prospectus

Abstract

Rhodopsin is a retinal photoreceptor protein of bipartite structure consisting of the transmembrane protein opsin and a light-sensitive chromophore 11-cis-retinal, linked to opsin via a protonated Schiff base. Studies on rhodopsin have unveiled many structural and functional features that are common to a large and pharmacologically important group of proteins from the G protein-coupled receptor (GPCR) superfamily, of which rhodopsin is the best-studied member. In this work, we focus on structural features of rhodopsin as revealed by many biochemical and structural investigations. In particular, the high-resolution structure of bovine rhodopsin provides a template for understanding how GPCRs work. We describe the sensitivity and complexity of rhodopsin that lead to its important role in vision.

INTRODUCTION: HISTORICAL PERSPECTIVE

The visual pigment rhodopsin was originally extracted from bovine retina using bile salts by Kühne (1). He found that the color of rhodopsin resulted from a chromophore different from xanthophylls (egg yolk) or β-carotene (corpus luteum). It faded upon exposure to light, heating, or organic solvents and could be salted out from the extract by ammonium sulfate. Regeneration of bleached rhodopsin required interaction of the retina with adjacent retinal pigment epithelium cells (RPE). The need for vitamin A in vision triggered an interest in retinoids as a potential source of the chromophore. Wald provided evidence that visual pigment is composed of a protein conjugated with an unidentified retinoid (2). Vitamin A (all-trans-retinol) is a precursor of this retinoid and a product of visual pigment photobleaching. Successive work demonstrated that bleaching and regeneration of the visual pigment involves a cycle of stereoisomerization of all-trans-retinal/all-trans-retinol. The reddish color of rhodopsin results from the protonated Schiff base linkage of 11-cis-retinal to the membrane protein opsin (3, 4). Therefore, a metabolic cycle of production and recycling of the photoisomerized visual chromophore is essential for our vision. Bleaching of rhodopsin introduces a succession of quasi-stable intermediates that within milliseconds reach an equilibrium of metarhodopsin I (Meta I) and metarhodopsin II (Meta II) (5). Meta II is favorable in higher temperatures and pH, salt, and glycerol, and its formation is accompanied by a large increase in entropy. Subsequent work focused on bovine rhodopsin, where the naturally high abundance of this interesting protein was utilized in a number of biochemical and biophysical approaches [early work reviewed in (6)].

The first protein amino acid sequence of the G protein-coupled receptor (GPCR), opsin, was determined in the laboratories of Hargrave et al. and Ovchinnikov (7, 8), and the first models of the seven-transmembrane helix topology, based partly on proteolysis, antibody-recognition, and chemical and posttranslational (glycosylation) modifications, were introduced (9). The overall topology of rhodopsin is similar to another retinylidene-binding protein, bacteriorhodopsin (8). In parallel with the improved understanding of the properties of rhodopsin, functional studies provided a solid basis for the way the light signal is translated into biochemical reactions, a process termed phototransduction. Fung & Stryer demonstrated that light-activated rhodopsin catalyzes the nucleotide exchange of GDP to GTP in an interacting protein called transducin, a G protein, which in turn activates a photoreceptor cGMP-specific phosphodiesterase (10). Hundreds of transducin molecules are activated per photoactivated rhodopsin in this first amplification step during phototransduction (11). Deprotonation of Meta II's Schiff base between the chromophore and opsin appeared to be critical for the activating property of light-exposed rhodopsin (12). A fortuitous discovery, made in the early 1970s, that rhodopsin is phosphorylated (13-15), has been connected with observations that ATP mediates quenching of phosphodiesterase activity (16, 17). The responsible rhodopsin kinase has been purified (18) and cloned (19). For complete quenching of the light-activated phosphorylated rhodopsin activity, specific binding of a 48-kDa protein is critical (20). Arrestin-like molecules play a key role in the signaling and cellular processes of a large number of agonist-stimulated GPCRs (21).

Nathans & Hogness were the first to clone opsin from the bovine cDNA library (22). Subsequently, the generated bovine DNA sequence was essential in the determination of the bovine opsin genomic structure (22); the determination of the cDNA and genomic structures of human opsin and blue, green, and red visual pigments (23, 24); the cloning of the first invertebrate rhodopsin and its gene from Drosophila (25); and the total chemical synthesis of cDNA for bovine rhodopsin (26).

The crystal structure of rhodopsin isolated from bovine retinas has been solved (27) and refined (28), providing significant insight into the function of this important molecule. In this review, we describe the major structural properties of rhodopsin. As with any rich subject, not all of the nuances of rhodopsin structure and function can be covered here. Therefore, the authors encourage readers to peruse other reviews dedicated to rhodopsin and different aspects of phototransduction (6, 21, 29-36).

RHODOPSIN

Expression

Retinal rod cells are specialized neurons that function in capturing photons and communicating with secondary neurons about the presence of light. These events are initiated by conformational changes in the light-sensitive pigment, rhodopsin, followed by a biochemical cascade of reactions, termed phototransduction (30, 31). In mammalian retina there are ∼10⁸ photoreceptor cells. Rods are highly differentiated cells with outer segments (ROS) containing all components necessary for phototransduction. ROS are composed of stacks of 1000–2000 independent disk membranes surrounded by plasma membrane (Figure 1, see Supplemental Materials: Follow the Supplemental Material link on the Annual Reviews homepage at http://annualreviews.org/). The main component of the disk membranes is rhodopsin (>90% of the membrane's proteins). Opsin, the protein component of rhodopsin, is specifically expressed in retinal rod photoreceptors and in some cells of the pineal gland (37). In adult mice, 0.06% of the total retinal RNA encodes for opsin (photoreceptors constitute ∼80–90% of all cells in the retina) (38). This translates into the highest expression of any GPCR. About 0.5–1 mg of rhodopsin can be isolated from one bovine retina.

Figure 1.

Molecular packing of rhodopsin (A) and sensory rhodopsin (B) in their crystals. One asymmetric unit contains two rhodopsin molecules. The central dimer is shown as a continuous surface, whereas the rest are shown as ribbons. Views are from different crystallographic X-, Y-, Z-axes for rhodopsin (27), and X-, and Z- for sensory rhodopsin. The data for sensory rhodopsin were taken from Protein Data Base (1H68). Coordinates: Models of rhodopsin and the homology models of cone pigments have been deposited in the Protein Data Bank (PDB) (identifiers 1F88, 1HZX, 1KPN, 1KPW, 1KPX).

Purification

The absorption spectrum of rhodopsin in detergent solutions displays two maxima in the UV-Vis region: the protein band around 280 nm and the chromophore-related peak at 498 nm. The ratio of the absorption A_{280 nm}/A_{498 nm} for pure rhodopsin devoid of opsin is ∼1.6, with an extinction coefficient of 42,000 cm⁻¹ M⁻¹ at 498 nm. Rhodopsin is stable at room temperature for days when extracted into many detergents, including Chaps, dodecyl-β-maltoside, or Tween 80. The most frequent source of native rhodopsin is bovine eyes. Native rhodopsin is initially pre-purified by isolation of ROS from the retina using a sucrose gradient method (50–70% rhodopsin content in ROS) [reviewed in (6)]. The purification methods require separation of rhodopsin from contaminating proteins and from bleached opsin. All methods require detergent for rhodopsin solubilization. The first method takes advantage of rhodopsin glycosylation and employs concanavalin A affinity chromatography (39). A large percent of opsin binds irreversibly to the resin. Concanavalin A chromatography is scalable, and large amounts of material can be prepared, although the purified rhodopsin is contaminated with other glycoproteins and with concanavalin A, which leaches from the resin. The second method uses hydroxyapatite chromatography and yields only partially purified protein (40). Rhodopsin is also efficiently purified by immunoaffinity chromatography using Molday's 1D4 monoclonal antibody, which recognizes seven to nine C-terminal amino acids of bovine rhodopsin. This method is particularly useful for heterologously expressed rhodopsin and other visual pigments (26, 41). The fourth method takes advantage of highly enriched rhodopsin in ROS and the instability of opsin and contaminating protein during prolonged incubation in mild detergent in the presence of divalent metal ions (43). It is an effective and scalable method that does not use any chromatographic columns. This method also enriches rhodopsin preparations with native ROS lipids and produces highly concentrated rhodopsin with controllable concentrations of detergent suitable for crystallization studies. This method is not applicable to heterologous systems due to the low expression levels of rhodopsin.

Composition

Rhodopsin is composed of a transmembrane apoprotein, opsin, and 11-cis-retinal bound to the protein through a Schiff base linkage to a lysine side-chain. Bovine opsin (Swiss-Prot: P02699) is composed of 348 amino acids with a molecular mass of 39,007. The total molecular mass increases to 42,002 when posttranslational modifications are included (palmitoylation, acetylation of N terminus, and glycosylation). The most prevalent amino acids are Phe (8.9%), Val (8.9), Ala (8.3), and Leu (8.0), suggesting a major hydrophobic character for this protein.

Rhodopsin is extensively modified by posttranslational modifications. The chromophore is attached through a protonated Schiff base to Lys²⁹⁶ (44). The N-terminal Met is acetylated (45), as is found frequently for other eukaryotic proteins at their initiation Met residue. Using ex vivo [³H]-labeling, it was shown that opsin is palmitoylated (46), a modification that is frequently observed among GPCRs. Two Cys residues, Cys³²² and Cys³²³ located at the C terminus, are palmitoylated (47). Two Cys residues from helix III and E III (the loop connecting helices IV and V) are cross-linked by a disulfide bond (48). Two (Man)₃(GlcNAc)₃ groups modify opsin through an asparagine-linkage (Figure 2, see Supplemental Materials: Follow the Supplemental Material link on the Annual Reviews homepage at http://annualreviews.org/) at the N terminus (Asn² and Asn¹⁵) (49). Glycosylation is not homogenous and other, but minor, compositions of the carbohydrate moieties were identified (50). Rhodopsin undergoes a light-dependent phosphorylation on six to seven Ser/Thr residues at the C-terminal end (51). In vivo, there are three sites, Ser³³⁶, Ser³³⁸, and Ser³⁴³, that are phosphorylated by direct and quantitative methods after 20–40% bleaching of the protein (52, 53). The phosphorylation processes, at lower bleaches such as 100–10,000 photons/s per rod, are not yet accessible to currently available technologies. At these bleach levels, our rods are functional before they saturate at more intense light. Heterogeneity and multiple rhodopsin phosphorylation have been observed in vitro [reviewed in (54)].

Figure 2.

Ribbon drawings of rhodopsin. Helices I–VIII are colored as a spectrum of visible light from blue (helix I) to red (helix VIII), and two orientations are shown. Palmitoyl chains and oligosaccharide groups shown using ball-and-stick models.

Rhodopsin, when isolated from ROS, contains variable amounts of tightly bound phospholipids. These lipids possibly stabilize rhodopsin and coat the hydrophobic transmembrane regions of the protein. We found, using radioactive phospholipids, that they can be removed completely only under very harsh conditions, such as 1% SDS in formic acid and trifluoroethanol (1:1) (X. Zhou & K. Palczewski, unpublished data). It is reasonable to speculate that the integrity of rhodopsin and other membrane proteins may be lost by removal of all tightly associated lipids.

Regeneration and Photobleaching Pathway

The key reaction of visual excitation is the ultra-fast (femtoseconds) photochemical reaction of rhodopsin's chromophore, 11-cis-retinal forming all-trans-retinal (55). From a theoretical organic chemistry point of view, this isomerization most likely involves the “Hula-Twist” mechanism that preserves, at first approximation, the positions of theβ-ionone ring and the Schiff base (56). A conventional “one-bond-flip” mechanism would predict a large rotation of the β-ionone ring within the rhodopsin molecules (Figure 3, see Supplemental Materials: Follow the Supplemental Material link on the Annual Reviews homepage at http://annualreviews.org/). In milliseconds, the signaling Meta II is established and this catalytically active form of rhodopsin binds and activates transducin (T, G_t).

Figure 3.

Two-dimensional model of bovine rhodopsin. Buried residues are shown in gray. Orange, red, and violet denote surface residues. Residues in contact with polar headgroups of the membrane on the cytoplasmic side are orange. Similar residues on the extracellular side are red. All other surface residues are violet.

The photoisomerization of 11-cis-retinylidene to all-trans-retinylidene triggers conformational changes of opsin through multiple intermediates, such as photorhodopsin, bathorhodopsin, lumirhodopsin, Meta I, Meta II, and Meta III, before the chromophore hydrolyzes and leaves the binding pocket (34). These intermediates are short-lived, but can be trapped by low temperature and distinguished by specific absorption maxima in the range of visible light (57). The final steps, namely hydrolysis of all-trans-retinylidene and its relation to Meta III, are poorly understood at the mechanistic level. The hydrolysis of the Schiff base appears to be the rate-limiting step in the release of the chromophore (58). In addition, two forms of Meta II were identified, Meta IIa and Meta IIb, that differ in the protonation state at the cytoplasmic surface, whereas the Schiff base is deprotonated with a characteristicλ_max at 380 nm (59). The transducin-activating form is Meta IIb (34).

Free all-trans-retinal also binds to opsin and forms partially active receptors toward activation of transducin [reviewed in (60)]. This activity is enhanced compared with the activity of free opsin. From the crystallographic model, a potential binding site for the retinal has been identified in the hydrophobic domain of rhodopsin (D.C. Teller, unpublished data). For 11-cis-retinal, binding to this site could be the first step in the regeneration of rhodopsin, before the chromophore binds non-covalently to the retinylidene cavity and prior to stable formation of the Schiff base (61). In the Meta II-transducin complex, all-trans-retinylidene can be photoisomerized back to 11-cis-retinylidene without transducin dissociation, suggesting that the G protein imposes conformational changes on the rhodopsin surface that cannot be reversed even upon photoisomerization (62). However, when Meta II is photolyzed by blue light, a product is formed that has absorption properties similar to that of rhodopsin with λ_max at 500 nm, but with all-trans-retinal bound (63). These data indicate that all-trans-retinylidene can be bound in the transmembrane segment of rhodopsin in two distinct conformations: one that activates the opsin moiety and another that inactivates it.

Mutagenesis

Availability of rhodopsin sequences and development of molecular biology techniques have had tremendous impacts in probing rhodopsin structure using muta-genesis and biochemical approaches. Findings from this research are discussed in the subsequent paragraphs, where they strengthen observations and interpretation of rhodopsin's structure and function.

A general method for rhodopsin expression and purification from monkey (or human) kidney cells (64), insect cells (65), and different yeast strains (66, 67) was introduced. In particular, we benefited from these studies in understanding the spectral tuning of rhodopsin. Retinals in their Schiff base forms absorb in the near ultraviolet range (360–380 nm, compared with ∼320 nm for retinols). The predicted counterion for the protonated Schiff base has been identified as Glu¹¹³, which is highly conserved among all known vertebrate visual pigments (68-70). The addition of a proton to the Schiff base, together with critical placement of intramolecular negative charges, results in a bathochromic shift in the absorption spectrum to that characteristic of native rhodopsin (500 nm). The counterion in the hydrophobic environment causes a ∼sevenfold increase in the pK_a of the protonated Schiff base. Therefore, rhodopsin in its native environment will have its chromophore in the protonated Schiff base linkage (29).

Glycosylation was shown to be dispensable for the proper folding and function of rhodopsin. Mutant rhodopsins lacking Asn¹⁵-glycosylation exhibited poor folding and were defective in transport to the cell surface. They were also poor transducin activators, perhaps owing to their intrinsic instability (71). These studies have not yet been extended to mouse animal models.

Ridge et al. pioneered studies of rhodopsin using expressed polypeptide fragments in COS cells. Splitting rhodopsin in the second and third cytoplasmic loops led to production of two-fragment rhodopsins that show properties similar to the wild-type (72, 73). These studies, and earlier limited proteolysis studies [reviewed in (6)], revealed several important aspects of rhodopsin function. For example, connecting subsequent loops II and III have only a small stabilizing effect on rhodopsin, whereas the chromophore is critical for the stabilization of the tertiary structure. It appears that helices I–III and V–VII, with helix IV connected to either of these two fragments, form active rhodopsin, suggesting the existence of two independent intradomain interactions. These data are in agreement with another set of experiments. Rhodopsin retains its spectrum after extensive proteolysis of its exposed loops by pronase in native membranes, or in detergent solutions, albeit, digested rhodopsin is more temperature-sensitive to denaturation (K. Palczewski, unpublished data).

CRYSTAL STRUCTURE OF RHODOPSIN

Crystallization and Data Analysis

Rhodopsin was selectively extracted from highly purified bovine ROS using a mild detergent (such as nonyl-β-glucoside) and high concentrations of divalent metal ions (for example, 80 mM Zn²⁺) in the absence of Cl⁻ ions. This method, developed by a postdoctoral fellow, T. Okada, therefore does not involve any column chromatography and maintains high concentrations of endogenous lipids in the extract (43). Importantly, pre-incubation of ROS at room temperature in detergent causes opsin and other membranous proteins to precipitate and be efficiently separated from rhodopsin. Opsin is present up to 15% of total opsin and rhodopsin in cow eyes obtained from a local slaughterhouse. The purified rhodopsin had an A_{280 nm}/A_{500 nm} ratio of 1.6, close to the theoretical minimum of 1.5, suggesting high purity of the protein and a lack of bleached rhodopsin, opsin (43, 74). Vapor diffusion crystallization techniques were used to crystallize rhodopsin, using ammonium sulfate as the precipitant and heptane-1,2,3-triol as an additive in nonyl-β-glucoside detergent solutions. Diffraction quality crystals were obtained and all were merohedrally twinned to different degrees (43). The crystals were also sensitive to visible light (75). To avoid bleaching, all procedures were performed under dim red-light illumination. Initial diffraction data for the native protein were collected at Stanford Synchrotron Radiation Laboratory and Advanced Photon Source synchrotrons; however, attempts to solve the structure using bacteriorhodopsin as a molecular replacement model failed. Hg-derivatives of rhodopsin were generated for multiple isomorphous replacement (MIR) and multiwave length anomolous dispersion (MAD) phasing experiments. The structure was solved using a six-wavelength mercury MAD data set obtained at SPring8 on a crystal with a 10% twinning ratio by Miyano and colleagues (27). Crystallographic refinement of the two molecules in the asymmetric unit generated the current model of rhodopsin at 2.8 Å resolution (28).

Dimers and Crystallographic Lattice

Rhodopsin molecules, crystallized in space group P4₁, are packed in the crystal lattice to form an array of tube-like structures (Figure 1A). Eight rhodopsin molecules fill the unit cell, with ∼70% of the space occupied by solvent. The dimensions of the unit cell are: a = b = 97.3 Å, c = 149.5 Å. Two copies of rhodopsin (denoted molecules A and B) make up the asymmetric unit and are related by a non-crystallographic twofold axis between helices IA and IB. The crystallographic lattice is formed by remarkably few contacts between the molecules when compared with an unrelated, but also seven-transmembrane helical protein, sensory rhodopsin, crystallized by a lipid cubic phase method (Figure 1B) (76).

The dimer of rhodopsin, as seen in the crystal structure, cannot be physiologically relevant, as one molecule is rotated by ∼160° in relation to the second. The two cytoplasmic surfaces (Figure 4, see Supplemental Materials: Follow Supplemental Materials link on Annual Reviews homepage at http://annualreviews.org/) would project toward opposite sides of the membrane if this arrangement were found in the bilayer. One of these two rhodopsins makes much fewer contacts than the other in the crystal. It appears that a strong hydrophobic interaction between rhodopsin molecules allows elongation of the dimers in the crystal, with only one molecule promoting the tube-like structure while the other piggybacks in the dimer. The two sets of interacting residues found for molecules A and B are not identical because of asymmetry in the dimer and slightly different protein conformations for the two crystallographically independent molecules (Figure 4A, Suppl. Figure). These interactions may possibly deform the structure of rhodopsin to some degree; however, preservation of rhodopsin's spectral properties suggests that any disorder would be localized and not affect the binding site of the chromophore. The contacts between dimers also occur with different residues in molecules A and B (Figure 4A, Suppl. Figure). In addition, these contacts involve the oligosaccharides bound to the protein (Figure 4A–C, Suppl. Figure).

Figure 4.

Hydrophobic and hydrophilic interactions between rhodopsin helices (loops not included) viewed from the intracellular side. Hydrophobic interactions are shown as orange fuzzy bars together with numbers denoting their strength. Hydrophilic interactions are shown as red lines. List of hydrophilic interactions between helices include helices I–II, Asn⁵⁵-Asp⁸³; helices I–VII, Tyr⁴³-Phe²⁹³, Asn⁵⁵-Ala²⁹⁹; helices I–VIII, Gln⁶⁴-Thr³²⁰; helices II–III, Asn⁷⁸-Ser¹²⁷; helices II–IV: Tyr⁷⁴-Glu¹⁵⁰, Asn⁷⁸-Thr¹⁶⁰, Asn⁷⁸-Trp¹⁶¹; helices III–V, Cys¹⁴⁰-Thr²²⁹, Glu¹²²-Trp¹²⁶-His²¹¹-Tyr²⁰⁶ network; helices III–VI, Arg¹³⁵-Glu²⁴⁷; helices III–VII, Glu113-Lys²⁹⁶ (retinal); helices IV–V, Ala¹⁶⁶-Tyr²⁰⁶; helices VI–VII, Cys²⁶⁴-Thr²⁹⁷; and helices VII–VIII, Ile³⁰⁷-Arg³¹⁴.

Three-Dimensional Model

The current model of bovine rhodopsin (Figure 2) includes >95% of its amino acid residues. Missing residues in the structure are Gln²³⁶GlnGlnGluSer²⁴⁰ and Asp³³¹GluAla³³³ in molecule A; and Lys¹⁴¹ProMetSerAsnPheArgPhe¹⁴⁸, Val²²⁷PheThrValLysGluAlaAlaAlaGlnGlnGlnGluSerAlaThrThr²⁴³, and Pro³²⁷LeuGlyAspAspGluAlaSerThrThrValSerLysThrGluThrSerGlnValAlaProAla³⁴⁸ in molecule B. The shape of the transmembrane domain is that of an elliptic cylinder of width ∼41 Å, height ∼75 Å. The length and width of the elliptic footprint on the plane at the middle of the membrane is roughly 45 by 37 Å. Rhodopsin, when projected on the membrane, occupies a surface close to 1500 Ų (28). Two palmitoyl chains have been identified and two oligosaccharide units have been modeled. Rhodopsin folds into seven transmembrane helices, I–VII, varying in length from 22 to 34 residues, and one cytoplasmic helix (Figure 2; helix VIII in red running parallel to imaginary membranes). These helices also differ on the level of irregularities and are tilted at various angles with respect to the expected membrane surface (28). The chromophore occupies a cavity approximately two thirds from the top or cytoplasmic side of the transmembrane segment, a location that had been previously predicted from biophysical studies (77).

The extracellular, doubly glycosylated domain is formed by 33 N-terminal amino acids and by interhelical loops between helices II and III (residues 101–105), IV and V (residues 173–198), and VI and VII (277–285). The region displays a noticeably compact structure. Four distorted short β-strands form a plug that limits accessibility to the chromophore-binding site from this extracellular side. The Gly³ to Pro¹² region forms the first β-hairpin that runs parallel to the expected plane of the membrane. The third and fourth β-strands, Arg¹⁷⁷ to Asp¹⁹⁰, form a second twisted β-hairpin. This region contains the Cys¹⁸⁷ portion of the disulfide bond. The fourth β-strand is below the 11-cis-retinal and is a part of the chromophore-binding pocket with the carbonyl oxygen atom of Cys¹⁸⁷ being only 3.0 Å from C₁₂ of the retinal. An additional consequence of the plug is that Glu¹⁸¹, highly conserved in rhodopsins and in short- to mid-wavelength visual pigments (29), is in close proximity (4.4 Å) to retinal at the C₁₂ position. Glu¹⁸¹ modulates the stability of Meta II and its accessibility to hydroxylamine (78), suggesting that the plug region prevents rapid dissociation to the intradiscal space. In red visual pigments, a His residue occupies this position, forming a part of the binding site for a Cl⁻ ion (79). A Cl⁻ ion modulates the λ_max of the absorption spectrum, causing a bathochromic shift by ∼40 nm for these cone visual pigments, compared with pigments in mutants lacking this binding site (80).

The C-terminal structure of rhodopsin does not have any secondary structural elements. From Gly³²⁴ to Asp³³⁰, the polypeptide folds back over cytoplasmic helix VIII. Residues 331–333 have no discernable electron density; therefore, the conformation of this region is uncertain. Functionally, rhodopsin's C terminus plays two roles. It is necessary for the vectorial transport of rhodopsin from the site of synthesis to the rod outer segment (81-83) and for desensitization of photoactivated rhodopsin (35, 84).

The cytoplasmic surface contains two short loops connecting helices I and II (residues 65–70), and helices III and IV (residues 141–149). A third cytoplasmic loop is between helices V and VI (residues 226–243). In the current model this loop is incomplete, lacking electron density for residues 236–240. The C terminus is unstructured in solution (85) but likely folds over a loop between helices I and II (86). At the exit of transmembrane helix VII is helix VIII, which is fully cytoplasmic, formed by residues 311–323, and 16 Å in length.

Two-Dimensional Model

A two-dimensional model of rhodopsin (Figure 3) is based on a similar earlier model (9). Fourteen positively charged residues are present on the cytoplasmic surface of rhodopsin, whereas only three are on the extracellular face, in agreement with the positive inside rule for multi-spanning eukaryotic membrane proteins (87). In particular, three positive charges in the first cytoplasmic loop of rhodopsin may be critical for proper insertion into the membranes (88). The negatively charged residues are similar on both sides of the membrane. Nominally, rhodopsin has four charged residues buried in the transmembrane region of rhodopsin: Asp⁸³ in helix II, Glu¹¹³ and Glu¹²² in helix III, and Lys²⁹⁶ in helix VII. Two of these residues are engaged in chromophore-binding, one by donating an amino group (Lys²⁹⁶) to form a protonated Schiff base and the other, Glu¹¹³, by acting as a counterion for this linkage. Two other charged residues, along with positive and negative residues at the ends of the transmembrane helices, could be involved in the formation of water channels and water polarization within the hydrophobic interior of the molecule. This inference has yet to be substantiated by strong experimental proof.

A set of residues facing the outside of the molecule are all highly hydrophobic, as expected, either lacking side-chains (Gly), or having a small side chain such as Ala. Surprisingly, five imino-acid residues, Pro, are also facing the lipids (Figure 3; the three-dimensional model is shown in Figure 5, see Supplemental Materials: Follow the Supplemental Material link on the Annual Reviews homepage at http://physiol.annualreviews.org/). The composition of the lipid-facing residues could be important for the formation of an organized oligomeric structure of this transmembrane protein in the lipid bilayer (89).

Figure 5.

Vicinity of retinal in the binding site of rhodopsin. (A) Chromophore-binding site of bovine rhodopsin. The chromophore is in purple, and colors of helices are as in Figure 2. (B) The structure of 11-cis-retinal bovine opsin (27) using space-filling model. In blue are nitrogen atoms of the peptide bond and the Schiff base, with the hydrogen between Lys²⁹⁶ and the retinal in green. In red, two acidic residues in the binding site, Glu¹¹³ and Glu¹²², which is close to the β-ionone ring.

Helices of Rhodopsin and Interhelical Interactions

Helix I, (encompassing residues 34 through 64 and having a length of 45 Å, is tilted ∼25° from the vector perpendicular to the plane of membranes and is bent by 12° around Pro⁵³. The C-terminal tail of rhodopsin occupies the space over part of the helical bundle, helix I and helix VII. Helix I forms multiple hydrogen bond interactions with helices II, VII, and VIII (Figure 4). One of the residues that exhibits the highest conservation among related proteins is Asn⁵⁵ in helix I. Its side chain is responsible for two interhelical hydrogen bonds to Asp⁸³ in helix II and to the peptide carbonyl of Ala²⁹⁹. Additional interactions involve Tyr⁴³-Phe²⁹³ and Asn⁵⁵-Ala²⁹⁹ with helix VII and Gln⁶⁴-Thr³²⁰ with helix VIII (Figure 4). These and other hydrophobic interactions are depicted in Figure 6A (see Supplemental Materials: Follow the Supplemental Material link on the Annual Reviews homepage at http://physiol.annualreviews.org/).

Figure 6.

The DRY region (A) and NPXXY region (B) are believed to be involved in the activation and transformation of photoactivated rhodopsin. Note that in rhodopsin, the sequence of the DRY region is ERY.

Helix II, formed by residues 71 through 100, has a length of 40 Å, is tilted ∼25° from the vector perpendicular to the plane of membranes, and is bent by 30° around tandem Gly residues, Gly⁸⁹ and Gly⁹⁰. As a consequence of this bending, Gly⁹⁰ is in the vicinity of Glu¹¹³, the counterion for the protonated Schiff base of the chromophore. This helix forms multiple hydrogen bond interactions with helix I (see above), helix III (Asn⁷⁸–Ser¹²⁷), and helix IV (Tyr⁷⁴-Glu¹⁵⁰, Asn⁷⁸-Thr¹⁶⁰, Asn⁷⁸-Trp¹⁶¹) (Figure 4). It appears that it forms a tight pair with helix I owing to extensive hydrophobic interactions, in addition to the ionic interaction between Asn⁵⁵ and Asp⁸³ (Figure 4). The cytoplasmic ends of helices II and IV are close to each other.

Helix III, extraordinarily important for rhodopsin, encompasses residues 106 to 139. It is the longest helix at 48 Å in length, tilted ∼33° from the vector perpendicular to the plane of membranes, and is bent twice, once by 12° around Gly¹²⁰-Gly¹²¹ and the second by 11° at Ser¹²⁷. The large tilt of this helix results in the helix running across the helical bundle and making contacts with five other helices (II, IV, V, VI, and VII) (Figure 4; Figure 6A, see Suppl. Figure). The ionic/hydrogen bond interactions include Asn⁷⁸-Ser¹²⁷, Glu¹²²-Trp¹²⁶-His²¹¹, Cys¹⁴⁰-Thr²²⁹, Arg¹³⁵-Glu²⁴⁷, and Glu¹¹³-Lys²⁹⁶ (retinal). At the extracellular end of helix III, Cys¹¹⁰ is engaged in a disulfide bond with Cys¹⁸⁷ of the β-sheet. This disulfide bond is highly conserved among GPCRs. In the middle of this helix, Glu¹¹³, Gly¹¹⁴, Ala¹¹⁷, Thr¹¹⁸, Gly¹²⁰, Gly¹²¹, and Gly¹²² run below the β-ionone ring to C₁₅ of the chromophore. Glu¹¹³ is the counterion for the protonated Schiff base formed between retinal and Lys²⁹⁶. The C-terminal end of helix III at the cytoplasmic surface contains the E(D)RY motif located in an extremely hydrophobic environment formed among residues from helices II, IV, V and VI. The E(D)RY residues (although for rhodopsin this motif is ERY, in a majority of GPCRs it is DRY and therefore referred to as the DRY motif) have been implicated in the regulation of the receptor's interaction with its G protein. In this motif, there are salt bridges between Glu¹³⁴, Glu²⁴⁷, and Arg¹³⁵. This organization of ionic interactions within the hydrophobic pocket could be one of the critical constraints in keeping rhodopsin in the inactive conformation. The crystal structure in this region has high atomic displacement parameters (B-values), however, and the side chains may assume different orientations. The DRY motif is highly conserved among GPCRs.

Helix IV, the shortest transmembrane helix, encompasses residues 150 through 172 and is 33 Å in length. It runs almost perpendicular to the plane of the membrane and is bent by 30° around two neighboring proline residues, Pro¹⁷⁰ and Pro¹⁷¹, near the C-terminal end of the helix. Furthermore, additional residues near the Pro residues are involved in accommodating the perturbation of the helix by these imino acids. The short nature of this helix results from the presence of the plug, which edges up close to the chromophore. The plug also includes a disulfide bond formed by Cys¹¹⁰ and Cys¹⁸⁷. From helix IV, Cys¹⁶⁷ participates in the binding pocket of the chromophore. This helix is tightly coupled through ionic/hydrogen bond interactions with helix II (Tyr⁷⁴-Glu¹⁵⁰, Asn⁷⁸-Thr¹⁶⁰, Asn⁷⁸-Trp¹⁶¹), interacts through hydrophobic interactions with helix III, and has an additional contact site with helix V (Ala¹⁶⁶-Tyr²⁰⁶) (Figure 4; Figure 6A, see Suppl. Figure). Interestingly, the disulfide bond formed between Cys¹¹⁰ and Cys¹⁸⁵ cannot substitute for the correct disulfide Cys¹¹⁰-Cys¹⁸⁷ (90, 91). It is important to keep in mind that the Cys¹¹⁰-Cys¹⁸⁷ bridge is not absolutely essential for retinal-binding because Ala mutants form visual pigments, although with modified bleaching properties (92).

Helix V, formed by residues 200–225 and having a length of 36 Å, is highly tilted by ∼26° from the vector perpendicular to the plane of the membrane, and is bent by 25° and 15° at Phe²⁰³ and His²¹¹, respectively, surrounded by aromatic residues at the bends. The bend at His²¹¹ is caused by Pro²¹⁵. Helix V forms ionic/hydrogen bond interactions with helix III (Cys¹⁴⁰-Thr²²⁹, Glu¹²²-Trp¹²⁶-His²¹¹-Tyr²⁰⁶ network clustered about a cation or water molecule), helix IV (Ala¹⁶⁶-Tyr²⁰⁶), and through hydrophobic interaction with helix VI (Figure 4; Figure 6A, see Suppl. Figure). At the cytoplasmic end, helix V breaks around Leu²²⁶, followed by a polypeptide chain connecting helix V and VI. This loop is critical in the binding of transducin, arrestin and rhodopsin kinase. Furthermore, this helix provides residues for the binding site for the β-ionone ring of the chromophore (Met²⁰⁷, His²¹¹, and Phe²¹²).

Helix VI, the second longest helix in rhodopsin, is composed of residues 244 to 276, 47 Å in length, and only slightly tilted by ∼5° from the vector perpendicular to the plane of membranes. It is strongly bent by 30° by Pro²⁶⁷. Helix VI appears to be most loosely associated with other helices in rhodopsin, a feature that could be very important during the activation process. Helix VI interacts through ionic/hydrogen bond interactions with helix III (Arg¹³⁵-Glu²⁴⁷, important DRY region) and helix VII (Cys²⁶⁴-Thr²⁹⁷) (Figure 4; Figure 6A, see Suppl. Figure). Arg¹³⁵ forms a salt bridge with the previous residue, Glu¹³⁴, but is also connected to Glu²⁴⁷ and Thr²⁵¹ in helix VI. These interactions could be critical constraints in keeping rhodopsin in its inactive conformation. Interaction of helix VI with other helices is discussed further below. Hydrophobic residues from this helix (Val²⁵⁰, Met²⁵³) form a hydrophobic pocket for the ionic interaction within the DRY region. At the extracellular end of helix VI, Lys²⁴⁵, Lys²⁴⁸, and Arg²⁵² form a positively charged cluster, whereas Tyr²⁶⁸, Phe²⁶¹, Trp²⁶⁵, and Ala²⁶⁹ surround the β-ionone ring. Lys²⁴⁵ and Lys²⁴⁸ residues are 4 and 6 Å, respectively, from the inner limit of the transmembrane domain. The indole ring of Trp²⁶⁵ is very close to the β-ionone ring and the C₁₃-methyl group of the chromophore.

Helix VII is critical for rhodopsin function because it contains Lys²⁹⁶, which forms the retinylidene linkage with the chromophore within the transmembrane region. Helix VII is comprised of residues 286–309 and forms a 37 Å-long helix that is minimally tilted with respect to the plane of the membrane (9°), and is significantly bent by 24° and 21° to accommodate linkage with the chromophore around Pro²⁹¹and Pro³⁰³. It interacts with helix I (Tyr⁴³-Phe²⁹³, Asn⁵⁵-Ala²⁹⁹), helix III [Glu¹¹³-Lys²⁹⁶ (retinal)], and helix VI (Cys²⁶⁴-Thr²⁹⁷) through ionic and hydrogen bonds, as well as hydrophobic interactions, and with helix II via hydrophobic interactions (Figure 4; Figure 6A, see Suppl. Figure). Helix VII also stabilizes cytoplasmic helix VIII (Ile³⁰⁷-Arg³¹⁴) and provides residues that surround the protonated Schiff base. The C-terminal end of this helix contains a highly conserved NPXXY motif in GPCRs. The side chain of a polar residue in the NPXXY region, Asn³⁰², projects toward Asp⁸³. The interaction between these residues may involve water molecules (93). Tyr³⁰⁶ also projects inside the rhodopsin molecule and interacts with the conserved Asn⁷³ of helix II. In addition, there is a hydrophobic interaction between helices VII and I through Phe²⁹³ (helix VII) and Leu⁴⁰ (helix I). Residues Lys²⁹⁶-Ala²⁹⁹ form a helix in the 3₁₀ conformation (three residues per helical turn, and ten atoms in the ring closed by the hydrogen bond). All other helices in rhodopsin are regular α-helices. An additional kink in helix VII is at Val³⁰⁰. This distinct part of the helix (one helical turn) gives more freedom to Lys²⁹⁶ because this residue has to accommodate big changes in retinal's structure during the first stages of isomerization. This accommodation greatly decreases the energy barrier of isomerization.

The crystal structure of rhodopsin also revealed one cytoplasmic α-helix. Helix VIII, residues from 311 to 323, spans 16 Å, and forms a straight, short amphiphilic helix connected through ionic/hydrogen bond interactions with helix I (Gln⁶⁴-Thr³²⁰) and helix VII (Ile³⁰⁷-Arg³¹⁴) (Figure 4; Figure 6A, see Suppl. Figure). Elongation of the cytoplasmic surface by helix VIII may allow docking of the transducin Gt trimer (α-, β-, γ-subunits) on the surface of a single photoactivated rhodopsin (94). The C-terminal end of this protein contains Cys³²² and Cys³²³ with the attached palmitoyl moieties (Figure 3). The presence of these palmitoyl groups, attached enzymatically or spontaneously (46, 95), and the loop between helix VII and VIII may affect rhodopsin regeneration (72, 96). In COS cells, Cys³²² and Cys³²³ mutants showed properties similar to those of wild-type rhodopsin (97). It has been proposed that this region of helix VIII and palmitoylation is the site of non-covalent binding of all-trans-retinal to opsin (98), in agreement with rhodopsin modeling (D.C. Teller, unpublished data).

Human Diseases Associated with Mutation in the Rhodopsin Gene

Mutations in the rhodopsin gene, mostly missense mutations and some frame-shift mutations, correlate with the autosomal-dominant form of the blinding disease termed retinitis pigmentosa (RP) and account for ∼30–40% of all cases of this disease (99, 100). The most relevant mutation is Pro²³His, which accounts for about 10% of autosomal-dominant RP, but ∼100 other mutations were identified throughout rhodopsin. The sites of mutation were previously presented in a two-dimensional model of rhodopsin (34). A null mutation in the rhodopsin gene produces the recessive phenotype, but even heterozygotes have impaired rod vision (101, 102), suggesting the importance of a proper dosage of rhodopsin expression for healthy rod cells. These results are in an agreement with data obtained from rhodopsin knockout mice (103-105). Mice lacking rhodopsin expression do not develop ROS, suggesting that it is also a structural protein necessary for the proper formation of ROS in addition to being a receptor. Many mouse models generated in recent years allow an understanding of how mutations in rhodopsin cause the defect in rod cell biology.

Using the heterologous expression system, these mutants can be classified into three classes: Class I is defective in photoactivation rearrangements, class II remains in the endoplasmic reticulum and does not bind the chromophore, and class III is expressed at low levels and poorly binds chromophore, suggesting a problem with folding and stability (82, 106, 107). Here, we analyze some of these mutants considering the crystal structure of rhodopsin. Mutations in rhodopsin's extracelullar domain interfere with formation of the plug, destabilizing proper folding of rhodopsin and binding of the chromophore. The C-terminal mutations are most likely associated with the breakdown of vectorial transport to rod outer segments (82). Owing to massive amounts of produced protein, inhibition of this process would lead to blocking the inner segment as a consequence of inefficient disposal of mutated opsin. Mutations in the transmembrane region of rhodopsin may also lead to constitutive activity (108), metabolically overwhelming the rod photoreceptor cells. Most of these mutations are centered on the chromophore-binding site and intense intramolecular stabilizing interactions are involved (Figure 6B, see Suppl. Figure). However, there are also mutations at the lipid-facing residues that most likely prevent proper packing of rhodopsin in ROS disks and/or formation of the organized structure of this protein within the disk.

Chromophore-Binding Site

The retinylidene-binding site is largely hydrophobic, but several charged residues are also present around the chromophore and are necessary for the proper functioning of rhodopsin. The chromophore of rhodopsin, 11-cis-retinal, is linked to opsin through a Schiff base (Figure 5). There is clear electron density from the crystal structure analysis that corresponds to the Schiff base and the β-ionone portion of the chromophore (27). The helical nature of the transmembrane segment in the vicinity of Lys²⁹⁶ is greatly distorted by Pro²⁹¹ and Pro³⁰³. Lys²⁹⁶ is located just below the NPXXY region, which forms the C-terminal region of helix VII. The chromophore is in a twisted 6-s-cis-conformation within the binding pocket of rhodopsin (Figure 5). The torsion angles for the chromophore are given in a previous publication (28). The Schiff base linkage is protonated, and the counterion on helix III, Glu¹¹³, is 3.26 Å away. The vicinity of the Schiff base is formed by residues from helix II (Phe⁹¹, Thr⁹⁴), helix III (Glu¹¹³, Ala¹¹⁷), the loop connecting helices IV and V that forms an antiparallel β-sheet and runs under the chromophore (Ser¹⁸⁶, Cys¹⁸⁷), and helix VII (Ala²⁹², Phe²⁹³, Ala²⁹⁵). The kink in the chromophore at C₁₁ (the cis conformation) is close to Gly¹¹⁴, Ala¹¹⁷, and Thr¹¹⁸ of helix III, and to Cys¹⁸⁷, and Gly¹⁸⁸ in the loop connecting helices IV and V. The β-ionone is surrounded by residues from helix III (Thr¹¹⁸, Gly¹²¹, Glu¹²², Leu¹²⁵), helix IV (Cys¹⁶⁷), helix V (Met²⁰⁷, Phe²⁰⁸, His²¹¹, Phe²¹²), and helix VI (Phe²⁶¹, Trp²⁶⁵, Tyr²⁶⁸, Ala²⁶⁹). Thr¹¹⁸, Gly¹⁸⁸, Ile¹⁸⁹, Tyr¹⁹¹, Tyr²⁶⁸ are critical for the proper conformation around the C₉-methyl of the chromophore. Lack of the C₉ methyl group impedes photoactivation (109). Furthermore, Glu¹²² has been proposed to determine the rate of Meta II decay (110) (Figure 5B). Water also plays an important part in the binding site, although its location and role are not fully defined based on the crystallographic studies. The only well-defined water molecule is near the retinal chromophore near C₁₃ and is hydrogen-bonded to Glu¹⁸¹ and Ser¹⁸⁶ (28). The localization of the chromophore is in reasonable agreement with the cross-linking studies of retinal analogs and identification of Phe¹¹⁵, Ala¹¹⁷, Glu¹²², Trp¹²⁶, Ser¹²⁷, and Trp²⁶⁵ as residues in the binding pocket of the β-ionone ring (111, 112)¹.

Over the years, retinoid analogs have provided useful information on various photochemical steps during photoactivation, mapping of the chromophore-binding site, and the conformation of the chromophore within rhodopsin [reviewed in (113)]. Opsin reacts within minutes with 11-cis-retinal to form rhodopsin. Similarly, 9-cis-retinal (also called isorhodopsin) and 7-cis-retinal form visual pigments with λ_max at 483 nm (114) and 450 nm (115), respectively. When illuminated, 7-cis-, 9-cis- and 11-cis-rhodopsin converge to the same signaling Meta II intermediate (116, 117) with quantum efficiencies one seventh, one third, and equal to that of native rhodopsin, respectively. Double and triple cis-retinals and a number of analogs are known to reconstitute with opsin. Conversely, all-trans-retinal and 13-cis-retinal do not regenerate the visual pigment. For rhodopsin, it is generally accepted that 11-cis-retinal keeps this receptor in an inactive conformation. The mechanism of these structural constraints imposed by 11-cis-retinal may involve multiple hydrophobic and ionic interactions, as described above, that bring helices together, sequester Arg¹³⁵ from the DRY region in a hydrophobic environment, and tighten up the interaction between helices VII and VIII. Light induces the cis-trans isomerization and rhodopsin activation. However, the active site contains sufficient plasticity to accommodate 7-cis- and 9-cis-retinal and multiple double cis-isomers. This fact suggests either that each of these retinoids induces a different conformation of the protein or that the active site has sufficient fluidity to allow isomers to fit identical or similar binding sites. From the crystal structure of rhodopsin, it appears that the binding site has well-defined binding properties. One involves the fixed position of the Schiff base and the second, the binding of β-ionone in the active site. These two constraints on the chromophore determine the length of the chromophore that will fit into the binding site; any shortening or extending of the chromophore through addition or removal of carbon atoms is non-productive and does not lead to pigment formation (118). Furthermore, rhodopsin regenerated with a retinal with the 11-cis-bond constrained by cyclohexanen (6-lock-retinal) or cycloheptanen rings undergoes photoisomerization around other double bonds; however, rhodopsin remains marginally active (119; K. Palczewski, unpublished data). We believe that the isomerization occurs without the chromophore-induced conformational change of the opsin moiety because the β-ionone ring rotates but is largely confined within the binding site of the natural 11-cis-retinal chromophore. Isomerization of 6-lock-rhodopsin opens up the chromophore-binding site to nucleophiles, leading to hydrolysis of the chromophore (120, 121). Mutation of Trp²⁶⁵ to a smaller Phe and regeneration with the six-membered ring of 11-cis-retinal leads to formation of truly photostable rhodopsin (122). Whether isomerization of the chromophore is still taking place when bound to the mutant has not been explored. The covalent linkage of the chromophore to opsin is not necessary for the characteristic visible spectrum of rhodopsin or for photoactivation of rhodopsin, as has been elegantly shown using opsin mutants and diffusible 11-cis-retinylidene alkylamine analogs (123).

Structural Constraints and Functional Regions

The protein structure must impose the stringent properties that are required for rhodopsin to be a single-photon detector. In a milieu of 10⁸ rhodopsin molecules, detection of a single photon requires that noise from all other non-photoactivated rhodopsin be exceedingly low. Therefore, the rhodopsin structure evolved to fulfill this requirement with multiple structural constraints. (a) The extracellular plug is highly structured with multiple intradomain interactions fixed by seven transmembrane helices that likely will not allow significant spontaneous conformational changes within this region to take place in the dark or during activation. However, it was noted by others that Cys¹⁸⁵ is labeled only upon illumination (91, 124), suggesting that some conformational changes are possible. This domain is further stabilized by a disulfide bond within this plug (48, 90, 92, 125). (b) The chromophore tightens up the structure through multiple hydrophobic interactions, as well as ionic interactions between the protonated Schiff base (helix VII) and its counterion, Glu¹¹³ (helix III). Mutation of Lys²⁹⁶ or Glu¹¹³ results in opsins that display constitutive activity toward transducin in accordance with the idea that opsin is constrained in an inactive conformation by a salt bridge between Lys²⁹⁶ and Glu¹¹³ (108). (c) Asp⁸³, with other residues and in addition to hydrophobic interactions that involve neighboring residues, forms a tight network of interhelical interactions with helix I (Asn⁵⁵ in helix I) and helix VII in the NPXXY region (Asn³⁰² via a putative water molecule). (d) Highly tilted and central to the bundle of helices, the long helix III interacts with most of the other helices (with the exception of helix I, which is a tightly paired with helix II), especially with helix V (Trp¹²⁶-His²¹¹, Cys¹⁴⁰-Thr²²⁹, Glu¹²²-Trp¹²⁶-His²¹¹-Tyr²⁰⁶). (e) Sequestered Arg¹³⁵ within the hydrophobic pocket is stabilized by ionic interaction with Glu¹³⁴ (the same helix III) and Glu²⁴⁷ (helix VI), and further changes the Glu²²⁵ and Tyr¹³⁶ hydrogen bonding (Figure 6A). It appears that Arg¹³⁵ is critical for the interaction of photoactivated rhodopsin with transducin (126, 127); therefore, sequestering these residues may eliminate one of the key interactions between these two proteins. Arg¹³⁵ is highly conserved among GPCRs in the so-called DRY motif. (f) The NPXXY region interacts with the double-palmitoylated helix VIII and brings it close to the rest of the protein. The movement of helix VIII is observed during rhodopsin activation (128) (Figure 6B). Furthermore, an anti-Meta II antibody that recognizes residues 304–311 (a hinge between helix VII and helix VIII) does not react with rhodopsin or opsin (129).

Previous Models: Prediction, Bacteriorhodopsin and Cryo-Electron Microscopy

For years, the rhodopsin structure has been compared with that of bacteriorhodopsin, also a retinylidene seven-transmembrane protein. However, the arrangements of the seven helices were found to be different on the basis of earlier low-resolution studies on rhodopsin using cryo-electron microscopy at 7.5 Å resolution in the plane of the membrane and at 16.5 Å resolution perpendicular to the membrane (130-132) [see structural comparison in (28)]. The projection map of invertebrate rhodopsin is also similar to maps previously determined for bovine and frog rhodopsins (133). In rhodopsin, the helices are slightly longer than those of bacteriorhodopsin and differently arranged. Helices IV and V of the two proteins do not superimpose (whereas helices I, II, and III are similar). The helices in rhodopsin are more tilted and kinked, and its extramembrane regions are larger and more organized. Rhodopsin also contains an extra helix (helix VIII). The retinal chromophore is in the twisted 6-s-cis conformation in rhodopsin, whereas in bacteriorhodopsin it is 6-s-trans-conformation. In addition, the retinylidene groups in the two proteins are in different positions relative to the membrane plane, although the ionone rings are placed in similar positions. Because G protein receptors were not amenable for crystallographic studies, three-dimensional models based on the modeling studies were proposed. A particular model of rhodopsin was computed that closely resembles the transmembrane region structure found from the actual crystal structure (134). Similar models were also generated for other GPCRs (135).

A superimposition of the crystal structure of bovine rhodopsin and a model of frog rhodopsin from cryo-electron microscopy (gray balls), obtained under different resolutions and in different packing environments, are generally in agreement (Figure 7, see Supplemental Materials: Follow the Supplemental Material link on the Annual Reviews homepage at http://annualreviews.org/). However, the largest differences are seen within helices I, III, and VII. Remarkably, several of the kinked helices are observed in both structures. In addition to these structural studies, UV, FTIR, NMR, and EPR spectroscopic investigations of rhodopsin and its mutants in inactive and photoactivated states provide additional validation of the crystallographic model and further significant insights into rhodopsin structure and activation. Owing to space limitation, these spectroscopic studies cannot be fully evaluated herein.

Figure 7.

Interactions of helix VI. Hydrophobic contacts of helix VI shown as white residues and hydrophilic interactions represented by lines. (A) Interaction of helix VI with helices II and III. (B) Interaction of helix VI with helices V and VII.

ACTIVATION MECHANISM OF RHODOPSIN

Unified Model

Although the activation of rhodopsin is still roughly sketched, the crystal structure of rhodopsin provides the opportunity to summarize our current understanding of this process. The present view is inevitably oversimplified because of a vast amount of information existing in the literature. Therefore, the objective is to present a simplified, but coherent, hypothesis that unites major findings on rhodopsin activation. The reader should also consult recent publications (34, 75, 136, 137). One should also reflect on the complexity of the interactions with the transmembrane segment of rhodopsin (Figure 6A, see Suppl. Figure). Our activation model discusses only the major determinants of the activation process.

A photon of visible light absorbed by rhodopsin provides sufficient energy for the structural transformation from quiescence into a signaling protein. First, photoisomerization causes transformation of 11-cis-retinylidene to all-trans-retinylidene. The quantum yield for this reaction is ∼0.67 (138). The reaction is too fast for any significant conformational change in opsin (139) (Figure 8, see Supplemental Materials: Follow the Supplemental Material link on the Annual Reviews homepage at http://annualreviews.org/). Twisting of all-trans-retinylidene within the original 11-cis-retinylidene binding site of opsin is a way by which two thirds of the 55 kcal/mol energy of the photon is stored in the all-trans-retinylidene-opsin complex (140). This energy is then released during relaxation of earlier intermediates to the signaling stage of the receptor, Meta II. Formation of Meta II leads to relocation of the β-ionone ring to a new position, as proposed by cross-linking studies (141)² (Figure 8, see Suppl. Figure). Photoisomerization of the chromophore may initially involve reorganization of the 3₁₀ helical region of helix VII. The consequence of this movement is the breakage of a salt bridge between Glu¹¹³ and the protonated Schiff base between Lys²⁹⁶ and the retinal (108). This involves movement of a proton from the donor to the acceptor (142) in conjunction with the motion of the transmembrane helices (136, 143). The movement of helix VII away from helix I by 2–4 Å and the pushing of helix VI (64) ∼3 Å by the photolyzed chromophore from the first cytoplasmic loop lead to removal of Glu²⁴⁷ from the ionic interaction in the DRY region (Figure 6) [summarized in (136)]. Subsequent movement of Arg¹³⁵ to the cytosol allows photoactivated rhodopsin to attract and interact with transducin. Protonation of Glu¹³⁴, which is left in a very hydrophobic milieu, completes this transformation (34). This would reflect changes attributed to the central helix III (144). It is also important to note that helix VI is only loosely associated with the remaining bundle of helices (28) (Figures 4 and 7). The residues in helix VI, particularly Pro²⁶⁷, affect the assembly of rhodopsin from two fragments and the conformation of the cytoplasmic loop III (72). For the α_1B-adrenergic receptor, this region has been identified as key to the coupling process with its G protein (145). Therefore, the charge-neutralizing mutation Glu¹³⁴ to Gln¹³⁴ produces hyperactivity in the activated state and produces constitutive activity in opsin (144). The consequences of photoactivation are also transmitted toward a second region, the NPXXY cluster in helix VII. Structural changes connected with activation have been detected using spin labels at residues 306, 313, and 316. The changes at side-chain 313 can be accounted for by movements in the adjacent helix VII (146). Movement of this helix is critical for efficient coupling with transducin (147, 148). Methyl groups, in particular the C₉-methyl group on retinal, are critical for efficient movement of this helix (O.P. Ernst, K.P. Hofmann & K. Palczewski, unpublished data). The separation of helices on the cytoplasmic side of Meta II opens up a cavity into which transducin can dock (149). The formation of Meta II is transiently stabilized by rearrangement of the Glu¹²²-Trp¹²⁶-His²¹¹ triad (helices III–V) (150) before Meta II decays to opsin. Therefore, the activation process could be considered a removal of structural constraints in rhodopsin, and it is not surprising that a soluble cytoplasmic domain grafted into soluble proteins is very effective in the stimulation of transducin (151). It is important to note that certain phospholipids may impede these conformational changes (152).

Figure 8.

Different regions of cone pigments. (A) DRY region and the chromophore-binding sites of blue and red/green cone opsins. (B) The vicinity of the chromophore from left to right: blue, green, and red pigments. The numbering of corresponding regions is based on the bovine rhodopsin primary sequence.

Cavities and the Second Chromophore-Binding Site

The resolution of the rhodopsin structure is 2.8 Å and does not allow precise localization of all water molecules associated with the receptor. We have identified several regions around the chromophore and intracellular face that could be occupied by water molecules in the dark state, as well as large cavities suitable for protein-protein interactions (Figure 9, see Supplemental Materials: Follow the Supplemental Material link on the Annual Reviews homepage at http://annualreviews.org/). Water molecules could be involved in the activation process similar to the role of water in the bacteriorhodopsin photocycle. The reorganization of water molecules during activation has been detected by FTIR spectroscopy (153). Furthermore, one of these cavities could be involved in the non-covalent binding of all-trans-retinal, which activates opsin in transducin, arrestin, and rhodopsin kinase assays [reviewed in (60)]. One such binding site may be present around helix VIII and the palmitoylation sites (D.C. Teller, unpublished data). A recent refinement of rhodopsin at 2.6 Å resolution has provided an updated view of the contributions of water molecules to the activity of the protein (93).

Figure 9.

Functional domains of rhodopsin that are highly conserved among members of the GPCR superfamily. (A) Location of these domains in the three-dimensional structure of rhodopsin. (B) Close-up of the critical regions: DRY region (a, panel A, in rhodopsin ERY, blue); palmitoyl groups (b, panel A, red); NPXXY region (c, panel A, light blue); chromophore; Pro kink in helix VI, Lys²⁹⁶ (d, panel A, violet); disulfide bridge (e, panel A, yellow); and oligosaccharide moieties (f, panel A, brown).

CONE VISUAL PIGMENTS

Although a full analysis of spectral tuning and the properties of the cone pigments is outside the scope of this review, a few points are worth mentioning. Rhodopsin appears to be an ideal structural template for cone pigments. We carried out homology modeling for cone pigments that revealed several interesting features related to the DRY region (Figure 8A) and the chromophore-binding sites (Figure 8B) (154). In contrast, the NPXXY region is highly homologous among cone pigments and rhodopsin (data not shown). The only difference in the chromophore-binding site between green and red pigments is in the immediate vicinity of retinal, where Phe²⁶¹ is replaced by Tyr²⁶¹ (red), as predicted from studies on the spectral tuning of primates aimed at understanding the molecular properties underlying red-green color vision (155, 156). These studies were extended to identify residues responsible in in vitro experiments [157, 158; reviewed in (159)]. Furthermore, it has been convincingly documented that these in vitro analyses correlate with the in vivo data based on psychophysical measurements (160). The difference in λ_max between the green and red pigments has been proposed to result from other hydroxyl-bearing amino acids (Ser¹⁶⁴ and Thr²⁶⁹) that are farther away from the chromophore [(161); reviewed in (159)]. These earlier findings were confirmed by more recent studies (162). The difference between rhodopsin and green pigments (30 nm blue shift) could be accounted for by the increased hydrogen bonding of the protonated Schiff base group in rhodopsin (163). In the chromophore-binding site, Ser²⁰² is shown to hydrogen bond with Glu¹¹³ in the blue pigment, but points toward a Cl⁻ ion in the red/green pigment. Through mutagenesis studies, two residues, His¹⁹⁷ and Lys²⁰⁰, were identified in the Cl⁻-binding site of red and green visual pigments (79). In our model, Lys²⁰⁰ does not participate in the binding of this ion.

OTHER G PROTEIN-COUPLED RECEPTORS

An enormous set of information must flow across the cell surface to the cell to ensure survival, and transmembrane proteins play a key role in this process. Membrane proteins, among them rhodopsin-like proteins, GPCRs, represent a large and versatile group of protein sensors involved in nearly all physiological processes in vertebrates. In all organisms, from the most primitive to mammals, but perhaps not in plants, GPCRs are present in multiple molecular forms. For example, 5–10% of the Homo sapiens genome encodes GPCRs, which translates into >600 non-sensory and 1000s of sensory GPCRs. These molecules are involved in decoding information through the olfaction, vision, and taste systems, and comprise one of the largest families of proteins encoded by our genome. Understanding how rhodopsin, the best-studied member of this family, works reaches far beyond its key role in vision. Parallel information that can be learned from phototransduction is directly implemented into a better picture of other signal transduction via GPCRs, one of the most important topics in current neurobiology/pharmacology.

GPCRs transduce an extracellular signal by coupling with G proteins on the cytoplasmic side of the cell; the binding site for the activating ligand is either within the hydrophobic bundle of helices or at an extracellular domain of these receptors. There is a growing body of evidence that these receptors may also modulate the intracellular processes independently of G proteins (164). On the basis of the conservation pattern of the primary amino acid sequence, three major families of GPCRs have been described (165). Family 1 (or A), the largest of the three, is defined by receptors whose sequences are homologous to rhodopsin, including opiate, adrenergic, cannabinoid, and muscarinic receptors. Family 2 (or B) consists of GPCRs homologous in sequence to the secretin receptor, and Family 3 (or C) includes GPCRs homologous in sequence to the GABA receptor. The overall homology between receptors even within the same family is low (frequently <35%). However, some regions are highly conserved, suggesting that their function is either structural or related to the molecular mechanism of receptor activation. Some of these conserved residues, clustered in space, define conserved three-dimensional motifs referred to as functional microdomains (167, 169). The functional microdomains include the DRY region (Figure 9B, in rhodopsin ERY), helix VIII and palmitoyl moieties, the NPXXY region, a ligand-binding site, a Pro kink in helix VI, a disulfide bond, and oligosaccharide moieties. Therefore, rhodopsin is a very useful template for all GPCRs, at least until more high-resolution structures become available, especially for Family 1 (163).

The structure and organization of rhodopsin and other GPCRs in native membranes is not known. For many GPCRs, including adenosine, adrenergic, dopamine, opioid, and other receptors, dimeric structures have been more or less stringently documented (166).

FINAL REMARKS

In coming years, further advances will be made by understanding all of the proteins and complexes in different signaling states on a structural level and by giving reliable and logical chemical/physical models of visual transduction at an atomic resolution that will be coherent and consistent with thermodynamic laws. The structures of G proteins, arrestins, and now rhodopsin are the first steps toward these goals. Similar to research on the metabolic pathways in the past 10–20 years, signal transduction by GPCRs has become a very mature field. However, many fundamental questions remain in relation to the activation mechanism of GPCRs, the specificity and recognition of interacting proteins and their mechanisms of activation by GPCRs, and the structure of receptors and interacting proteins in the native setting of phospholipids in in vivo conditions. The great interest in these topics from academic and industrial laboratories makes it certain that further progress will be made.