Structural approaches to understanding retinal proteins needed for vision

Abstract

The past decade has witnessed an impressive expansion of our knowledge of retinal photoreceptor signal transduction and the regulation of the visual cycle required for normal eyesight. Progress in human genetics and next generation sequencing technologies have revealed the complexity behind many inherited retinal diseases. Structural studies have markedly increased our understanding of the visual process. Moreover, technical innovations and improved methodologies in proteomics, macromolecular crystallization and high resolution imaging at different levels set the scene for even greater advances. Pharmacology combined with structural biology of membrane proteins holds great promise for developing innovative accessible therapies for millions robbed of their sight or progressing toward blindness.

Introduction

Recent elucidation of the complete transcriptome of the retina and accurate analysis of the retinal proteome by mass spectrometry continue to reveal protein products encoded by retinal transcripts and their modifications [1]. Many of these proteins contribute to the formation and function of photoreceptors, which are unique highly-differentiated light-sensing cells that comprise more than 80% of cells in the retina and initiate the process of vision [2]. Each of the genes encoding these proteins is susceptible to potential blinding mutations. A detailed representation of these protein structures and their assemblies will lead to improved understanding of retinal processes, pathologies, and rational development of drugs to combat retinal diseases.

Photoreceptor cells and specifically the sites of phototransduction, namely rod and cone outer segments (ROS and COS), contain many proteins that are integrated into plasma or disc membranes [3, 4]. Fortunately, the challenge of crystallizing such proteins is rapidly being overcome by novel methods that are becoming routine. Vision research has produced outstanding contributions to the field of structural biology and physiology. Notable examples include determining the first structure of a membrane-bound G protein-coupled receptor (GPCR), vertebrate rhodopsin [5–8]. This was followed by structures of an invertebrate homolog [9, 10], rhodopsin mutants [11], many rhodopsin photoactivated intermediates and other derivatives [12–19], structures of G proteins alone and in complex with other proteins, and the structure of the RPE65 isomerase involved in visual chromophore regeneration [20, 21]. As the breadth of our structural knowledge of phototransduction/visual cycle proteins expands from an estimated 10%, we should be able to translate genetic changes into visual pathology and use network pharmacology to design and develop novel therapeutics to halt the progression of blinding retinal diseases.

Rhodopsin and GPCRs

Determination of the X-ray structure of rhodopsin from a native source offered the first atomic view of an archetypical GPCR [5], revealing arrangements of the transmembrane helices around the 11-cis-retinyldene ligand and the unanticipated feature of an extracellular domain. Different crystallization conditions have trapped different conformations of rhodopsin’s intracellular loop-III [7, 12], which comprises part of the heterotrimeric G protein binding site. New insights into rhodopsin activation have been provided by X-ray structures for the activated forms (Meta II [11, 14]) and chromophore-free forms, called opsins [13, 14]).

Because most GPCRs have unique extracellular regions accessible to pharmacological agents, these new advances will improve future structure-based drug design. But what is truly remarkable is that the overall folding of these receptors, including their transmembrane and cytoplasmic regions, is highly analogous (Fig 1). Thermal motions of atoms are reflected in their B-factors. Although there could be parameters other than thermal motion that potentially influence B-factor values, it appears that flexibility (quantified by their thermal motion) differs among these receptors. This suggests that these structures are fine-tuned for their rates of activation, amplification of the light signal, and timing of desensitization.

Fig. 1. Phylogenetic tree of GPCRs with known X-ray structures.

Branches and sub-branches (detailed methods are described elsewhere) of the GPCR phylogenetic tree are shown (adapted from Katritch et al. [25]). Branches of varying thickness represent different subgroups. Known structures from class A (rhodopsin subfamily, light blue branches), class B (secretin subfamily, red branches), and class F (frizzled, green branches) GPCRs are shown, but other groups such as the adhesion and glutamate GPCRs are not included. Structures of known GPCRs obtained in the presence of different ligands are clustered together. Types of ligands are identified by geometrical shapes placed at the top left corner of each individual GPCR structure. Structures with partial agonists are denoted with open triangles, agonists are highlighted with closed triangles, antagonists are identified by open squares, partial inverse agonists are marked by closed squares and inverse agonists are shown with closed circles. GPCRs bound to different inhibitors are identified by their gene names. B-factors for each individual GPCR structure were used to generate the GPCR rainbow color coding (blue through red, for minimum and maximum B-factor values, respectively). The gene names of the clustered structures and their corresponding PDB codes are: RHO (1U19 and 3CAP); ADRB1 (2VT4, 2Y02, and 2Y01); ADRB2 (2R4R and 2RH1); HTR2B (4IB4); HTR1B (4IAQ); CHRM3 (4DAJ); CHRM2 (3UON); DRD3 (3PBL); HRH1 (3RZE); F2R (3VW7); CXCR4 (3ODU); OPRD1 (4EJ4); OPRM1 (4DKL); OPRK1 (4DJH); OPRL1 (4EA3); NTSR1 (4GRV); ADORA2A (3QAK; 3VGA), and 3EML); EDG1 (3V2W); CRF1R (4K5Y); GCGR (4L6R); SMOH (4JKV).

Although structures derived from X-ray crystallography have provided a wealth of knowledge, complementary biophysical techniques have also offered valuable insights into the structure/function of rhodopsin. For example, the activation mechanism that could involve an intricate molecular network of internal water molecules was first suggested based on X-ray structural analysis [7], and later supported by evidence obtained through complementary biophysical studies such as radiolytic footprinting and hydrogen-deuterium exchange [22–24].

To date, structures have been solved for roughly two dozen GPCRs from three different classes (Fig. 1). Due to crystallization difficulties, several methodologies have been employed to obtain well-diffracting crystals, including co-crystallization with a Fab fragment of a monoclonal antibody and fusion with T4 lysozyme (T4L) or cytochrome b₅₆₂RIL onto the N-terminus or into the third intracellular loop [25–28]. Other methods involved the use of a novel class of detergents, bicelles [29], lipid cubic phase [30–32] or mutagenic approaches to thermodynamically stabilize the target GPCRs [26]. This multitude of structures provides a valuable snapshot of GPCR conformations with a variety of ligands. Recently, members of class B (red branch [33, 34]) and class F (green branch [35]) have been added to the growing list of GPCRs with known structures from class A (light blue branch).

Photoactivation of rhodopsin

A number of excellent reviews devoted to various aspects of the photoactivation of rhodopsin have been written [36] based on high resolution structures determined by X-ray crystallography and probed by NMR or FTIR [5–10, 12, 13, 15–19, 37–40]. Major conclusions from these studies can be summarized as follows: 1) isomerization of the chromophore initially causes retention of all-trans-retinylidene in the active site within the space occupied by 11-cis-retinylidene; 2) relaxation of the protein and the photoisomerized chromophore lead to an inward movement of helix VI by about 6 Å, along with small adjustments of other helices; 3) photoactivation of rhodopsin produces fluctuations and relaxation of the overall GPCR structure; 4) internal embedded water molecules provide a key non-protein component of activation, transmitting activation-associated structural changes from the chromophore active site to the protein cytoplasmic surface [22, 41]. These changes were first probed with structures of rhodopsin and opsin (Fig. 2), but the complex of this GPCR with the G protein transducin (G_t) has not yet been determined at atomic resolution. Notably, multiple rhodopsin crystals represent a single conformation trapped under specific crystallization conditions, but in reality rhodopsin could switch between different conformations in native membranes [19, 39]. Ultimately, photoactivated rhodopsin relaxes prior to adopting a specific conformation as it forms a transient complex with G_t (Fig. 2) [24].

Fig. 2. Structures of rhodopsin, opsin and activated intermediates determined by X-ray crystallography.

Color coding of the shown structures was derived from B- factors of the αC atoms as noted in Fig. 1. Ground state rhodopsin (with bound 11-cis-retinal depicted as a red sphere) is activated by a photon of light. Prepared under different experimental conditions, presented GPCR structures were superimposed on ground state rhodopsin and generated using PDB codes: 1U19, 3C9l, 1GZM, 2I35, and 2I36. Photoisomerization of 11-cis-retinal to all-trans-retinal (shown as orange spheres) is accompanied by conformational changes in the transmembrane helices together with a variety of local environmental changes (such as rearrangement of the internal water molecular network). Some of these changes are reflected in B-factor values. The Meta II structure (PDB code 3PXO) contains the all-trans-retinal chromophore. A potential structure of the Meta II-G_t complex is shown in dark grey and light grey shades and indicated by red arrows. The structures of both opsin* and opsin* with bound G_t peptide lack a visual chromophore (PDB codes 3CAP (in gray cartoons because the B-factors for this structure are incorrect) and 3DQB, respectively) and both can be converted to a Meta II state [14]) in the presence of all-trans-retinal (orange sticks) (black arrows, PDB code 3PQR). Opsin structures are not apo-receptors, because they contained bound detergent molecules (depicted in red) [90]. The G_t peptide is represented as grey surface. However, addition of 11-cis-retinal (red sticks) to opsin* crystals cannot successfully regenerate ground state rhodopsin (red arrow). RPE65 (4F3D), a component of the visual cycle, is shown as a dark gray cartoon on a light gray background denoting the retinal pigmented epithelium (RPE). Recycling of all-trans-retinal to 11-cis-retinal through the visual cycle is indicated by gray arrows. Rhodopsin has propensity to oligomerize in both native and expression systems [91–93].

Molecular interactions stabilizing rhodopsin

The mechanism of rhodopsin folding is not well understood. However, a number of mutations in rhodopsin have been shown to associate with its structural destabilization and misfolding, thus leading to loss of visual function. Hence, understanding the molecular interactions critical for proper folding of this receptor in its native environment is of great interest but limited by available methodologies. In 2006, we first employed atomic force microscopy (AFM)-based single molecule force spectroscopy (SMFS) to map, for the first time, specific segments with highly conserved and functionally important residues in native bovine rhodopsin that stabilize its secondary structure [42]. Further studies comparing rhodopsin with G90D rhodopsin, a mutant associated with congenital stationary night blindness, detected decreased thermodynamic stability and increased mechanical rigidity of most structural segments in the mutant, providing insights into the nature of its pathology. Furthermore, SMFS was proven useful for detecting molecular interactions stabilizing rhodopsin in its different photo-states and various animal species [43, 44]. Further similar work followed with the β₂-adrenergic receptor [45]. The differences and similarities between two receptors were noted [46].

Rhodopsin and its G protein, transducin

Protein conformational dynamics represent the quintessential mechanism of GPCR-mediated signal transduction [47]. Following absorption of a photon of light, a series of activated conformational and dynamic changes are initiated in rhodopsin that culminate in its signaling in the Meta II state. This activation translates into movements of helixes V and VI [14]. But more generally, activation results in a general relaxation of the rhodopsin molecule [24]. This relaxation, and perhaps formation of several coexisting states [19], expose binding sites on Meta II rhodopsin that are required for optimal interaction with the phototransduction-specific G_t. This interaction involves a conformational change in the α5 helix of G_t. From a dynamic perspective, a similar sequence of events was found to take place for rhodopsin and the β₂-adrenergic receptor. For example, activation placed both receptors in a much more dynamic state, whereas binding to their corresponding G proteins reduced this flexibility [24]. The same trend was documented for the cognate G proteins; both free G_s and G_t exhibited increased dynamics upon binding to their activated receptors [24, 48]. Although a high resolution structure of the Meta II-G_t complex has not yet been determined, the available X-ray structure of the β₂-adrenergic receptor in complex with its G protein (G_s) [49] provides a glimpse of the domain organization for such a complex. The modified β₂-adrenergic receptor-bound α-helical domain of G_s exhibits a profound rigid body movement as compared with the X-ray structure of the free G protein [50], as well as with the structures of more than 40 heterotrimeric G protein structures determined to date. Notably, if this rigid structure is inserted into a simulated [51] phospholipid bilayer (Fig. 3), a conformational clash of the α-helical domain of G_s with the membrane becomes apparent. Therefore, the exact orientation of the α-domain of G_s relative to the lipid bilayer requires further evaluation in light of this apparent overlap.

Fig. 3. The β₂-adrenergic-G_s complex modeled into a phospholipid membrane bilayer.

The β₂-adrenergic receptor in complex with G_s (PDB code: 3SN6) is inserted in a lipid bilayer. The phospholipid units (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) of the lipid bilayer are depicted as gray spheres and the model has been described in detail elsewhere [51]. The position of the β₂-adrenergic transmembrane domain (light blue cartoon) along the z axis of the phospholipid bilayer was chosen by superposition with rhodopsin (not shown). The position of rhodopsin (PDB code: 1U19) in turn is presented such that the palmitoyl fatty acyl chains attached to residues Cys322 and Cys323 are positioned at the same depth as the phospholipid fatty acyl chains of the membrane leaflet. Both the Nb35 nanobody and T4 lysozyme [49] are manifested as light gray structures. The G_s domains are represented as follows: G_sα and G_sγ are shown as cyan cartoons whereas G_sβ is pictured as a blue cartoon. Amino acid residues (Asn121 to Leu153) of the G_sα domain that overlap with the phospholipid head group region of the bilayer model are shown as a red cartoon. The face-on view of the transmembrane region is presented as a z-plane section of the phospholipid bilayer.

Lower resolution structures obtained by using negative staining coupled to electron microscopy (EM) have also been solved for both complexes [52–54]. Although EM models of the two receptors appear similar in shape, interpretations of the data differ. In the case of the β₂-adrenergic receptor, the extra electron density in the transmembrane region was attributed to detergent [54], whereas for rhodopsin, the extra space left after fitting in the monomeric receptor was attributed to the presence of a second rhodopsin molecule, in agreement with the previously determined 2:1 rhodopsin:G_t stoichiometry (Fig. 4) [52, 53, 55]. We attempted to resolve this inconsistency by using the same detergent as that in the β₂-adrenergic receptor study, but the presence of the second receptor molecule remained the only satisfactory explanation for the observed additional density [53, 56]. In contrast, EM single particle analysis of the T4L-β₂AR-G_s complex demonstrated only a single density corresponding to the T4L tag linked to the N-terminus of β₂-adrenergic receptor. Nevertheless, extensive biochemical analyses strongly suggest that the rhodopsin dimer is in fact asymmetric in nature [53, 55, 56]. This asymmetry allows the interaction of G_t with the C-terminus of only one of the two rhodopsin molecules, consistent with the model obtained from the EM study.

Fig. 4. Orientation of GPCR-G protein complexes in a phospholipid membrane.

Semi-empirical models of the complex formed between light-activated rhodopsin dimer and a G_t heterotrimer (A and B) [52] and the T4L-β₂AR-G_s-nanobody complex (C) [48] were fitted into a 3D molecular envelope calculated from projections of negatively stained, bis[succinimidyl] 2,2,4,4-glutarate (DSG) crosslinked rhodopsin*-G_t complexes purified in lauryl maltose neopentyl glycol (LMNG). Fitting of the rhodopsin*-G_t model generated with the structure of inactive G_t (PDB code: 1GOT) leaves a significant unoccupied density above one of the rhodopsin molecules (A), which becomes occupied after a 30^o hinge-like motion of the α-helical domain is applied (B). Though fitting the T4L-β₂-adrenergic receptor-G_s-nanobody structure into our EM 3D map leaves sufficient space to accommodate a second molecule of this receptor, conformation of the G_sα helical domain is inconsistent with our EM-density (C). Thus, the favored structure of the rhodopsin-G_t complex appears to be that shown in (B). Photoactivated rhodopsin (Rho*) that binds the C-terminal peptide derived from G_tα and the T4L-β₂-adrenergic receptor-nanobody molecule are both depicted in yellow. The second rhodopsin molecule in (A) and (B) is shown in gray. G_tα, G_tβ, G_tγ are colored pink, dark cyan and dark blue, respectively.

Rhodopsin, rhodopsin kinase (GRK1), the capping protein arrestin and its splice variant p⁴⁴

Photoactivated rhodopsin is phosphorylated by rhodopsin kinase and binds a 48 kDa capping protein called arrestin. Lack of either GRK1 or arrestin causes slowly progressive Oguchi disease characterized by profound delayed dark adaptation [57]. This visual phenotype is largely mimicked by knockout mice [58, 59]. Detailed analysis of a GRK1 Oguchi patient revealed that light adaptation was normal in low light backgrounds but abnormal in higher light environments indicating GRK1 involvement in rhodopsin phosphorylation at high light intensity [60]. A high resolution structure of GRK1 has been elucidated [61] as well as bipartite structures of rod- [62, 63] and cone-specific arrestins [64]. Similar to the (rhodopsin)₂-G_t complex, at least in some experimental conditions, one arrestin molecule binds to an asymmetric dimer of rhodopsin [65]. Palczewski and colleagues isolated a truncated form of arrestin called p⁴⁴ from bovine retina wherein the 15-amino acid-long C-terminal region was replaced by a single amino acid. As shown in subsequent work, p⁴⁴ was a novel splice form of the arrestin transcript. Significantly, a physiological assay revealed that p⁴⁴ binds to photoactivated rhodopsin even without phosphorylation, indicating a mechanism for full-length arrestin binding that involves displacement of the C-terminal region of arrestin by the C-terminal region of phosphorylated photoactivated rhodopsin. The structure of p⁴⁴ was determined under two different crystallization conditions that resulted in different conformations of this protein [66, 67]. A hysteresis model was proposed for the more recent of these two structures that assumed a stable “active” conformation because opsin was present during the crystallization [66, 67]. However, this hysteresis model would be extremely rare for a protein without chemical modification in any biological system. Neither rhodopsin kinase nor an arrestin/p⁴⁴ structure combined with any form of rhodopsin is currently available. Moreover, the structure of β-arrestin-1 with a phospho-peptide derived from vasopressin receptor 2 in complex with a Fab fragment shows additional C-terminal region conformational differences in β-arrestin-1 as compared to its inactive conformation [68].

Enzymatic production of 11-cis-retinal: the key step of rhodopsin regeneration

To maintain vertebrate vision, the spent all-trans-retinal chromophore released from rhodopsin in photoreceptor cell outer segments must be converted back to 11-cis-retinal, a process largely accomplished in an adjacent layer of the retina called the retinal pigmented epithelium (RPE) [69]. The key enzyme responsible for this cis-trans isomerization is retinoid isomerase (RPE65), with its bovine orthologous structure determined by Kiser and colleagues [20, 70]. RPE65 is a membrane-associated Fe²⁺-containing metalloenzyme. Importantly, disabling mutations of RPE65 cause a childhood blinding conditions termed Leber congenital amaurosis (LCA) and early-onset retinitis pigmentosa [71]. The overall structure of RPE65 is that of a seven-bladed β-propeller with single-strand extensions on blades VI and VII and a two-strand extension on blade III that contributes to dimerization of this protein. The Fe²⁺- ion is coordinated by four His residues and three second sphere Glu residues with each blade of the propeller contributing a single residue to the metal ion coordination. There is also a hydrophobic tunnel that allows the retinyl ester substrate to diffuse to the iron-containing catalytic site [20]. More recently crystals were obtained from enzymatically active native protein embedded in a lipid-detergent sheet. Based on these structures and complementary studies, we proposed that the Lewis acidity of iron could be used to promote ester dissociation and generation of a carbocation intermediate required for retinoid isomerization [42].

Other proteins of the phototransduction and retinoid cycle

Research discoveries over the past several years have significantly enhanced our understanding of the structural regulation of visual processes in the eye. In ROS, 11-cis-retinal binds to the apo-protein opsin forming rhodopsin, a functional light receptor. Consequently, a number of proteins involved in signal transduction are activated. Light-induced structural changes in rhodopsin lead to activation of G_t, resulting in formation of the G_α-GTP complex which activates retinal cGMP phosphodiesterase (PDE6). It is also known that transducin translocates between photoreceptor compartments (from ROS to inner segments) during the dark-light cycle [72]. This movement could be supported in part by helper proteins. One of these proteins, UNC119, was co-crystalized with an acylated α-subunit of transducin N-terminal peptide. This structure revealed that the lipid chain is buried deeply in UNC119’s hydrophobic cavity [73]. The recently determined EM structure of PDE6γ in complex with prenyl-binding protein delta (PrBP/delta), revealed probable locations of its isoprenylation sites, protein subunits, and catalytic sites [74]. Activated PDE6 catalyzes the hydrolysis of cGMP followed by closure of the cyclic nucleotide gated (CNG) channels and photoreceptor hyperpolarization, in effect attenuating synaptic glutamate release. Native CNG channels from rod cells are composed of CNGA1 and CNGB1 subunits in a 3:1 stoichiometry. The stoichiometry is not known for cone CNG channels. However, recently solved crystal structures of the parallel 3-helix coiled-coil domains of CNGA1 and the cone-localized CNGA3 (domains that regulate subunit assembly) were similar, suggesting a shared mechanism controlling the stoichiometry of rod and cone CNG channels.

Absorption of a photon of light causes isomerization of 11-cis-retinal to its all-trans-configuration and subsequent release of all-trans-retinal into the cytoplasm. However, a small fraction of all-trans-retinal is released into the luminal space of ROS and must be transferred to the cytoplasm to prevent accumulation of toxic condensation products of retinal and phosphatidylethanolamine. The ABCA4 transporter is involved in this process and EM and hydrogen-deuterium exchange studies have provided the first direct structural information about the membrane topology of this transporter and mechanistic insights into its function [75]. Released all-trans-retinal is reduced to all-trans-retinol by NADPH-dependent all-trans-retinol dehydrogenases (primarily RDH12 and RDH8) in the ROS and then all-trans-retinol diffuses into RPE cells co-chaperoned by inter-photoreceptor retinoid-binding protein (IRBP). Upon entry into the RPE, all-trans-retinol is bound by cellular retinol-binding protein (CRBP1) and then esterified by lecithin-retinol acyltransferase (LRAT) [76]. Besides LRAT, two other subfamilies belong to the LRAT-like family of proteins: the largely uncharacterized neuronal sensitive proteins (NSE1-2) and H-Ras-like tumor suppressor proteins (HRASLS1-5). Although the structure of LRAT has yet to be determined, a recently solved structure of HRASLS3 sheds light on the catalytic properties of LRAT [77]. Formed retinyl esters can be either stored in specialized compartments of RPE cells called retinosomes [78, 79], or used as substrates for the retinoid isomerase RPE65, which catalyzes the formation of 11-cis-retinol [80]. The last retinoid is rapidly bound by cellular retinal-binding protein (CRALBP) and its subsequent oxidation to 11-cis-retinal is catalyzed by NAD(P)⁺-dependent 11-cis-retinol RDHs (RDH5, RDH11 and RDH12) [81]. Structures of human CRALBP and of its mutant R234W which causes Bothnia dystrophy, with the endogenous ligand 11-cis-retinal were determined at 3.0 and 1.7 Å resolution, respectively [82]. The high redundancy of RDHs highlights the importance of maintaining low cellular levels of retinoids due to their high toxicity. 11-cis-Retinal is released from CRALBP by membrane acidic lipids [83] into the interphotoreceptor matrix where IRBP shuttles and releases it to photoreceptors, allowing the chromophore to regenerate accumulated opsin. A novel hydrophobic cavity that binds 11-cis-retinal and all-trans-retinal and is distinct from the long-chain fatty acid-binding site has recently been identified in the IRBP structure [84].

Finally, structures of the globular domains of C1q and tumor necrosis factor related protein 5 (C1QTNF5), a membrane associated protein involved in adhesion of RPE cells, provide insights into the structural basis of autosomal-dominant late-onset retinal macular degeneration (L-ORMD) [85].

Supramolecular organization of visual signaling machinery

Photoreceptor rod and cone cells are highly differentiated neurons with clearly defined substructures. In particular, the ROS house all components of the phototransduction signaling machinery in disc membranes, well-defined compartments containing about 50% internal membranes by volume [3]. Thus, almost all chemical reactions occurring in discs are interfacial, including those involving enzymes and other proteins. Animal models of many human retinal diseases provide powerful tools for studying the mechanisms of vision and their impairment by genetic mutations. The mouse has several advantages as the species of choice, including a non-redundant set of genes involved in vision, relative ease of genetic manipulation, and a close mechanistic resemblance to human vision. Segregation of nearly all involved proteins in a single cellular compartment renders the visual signalling system very suited to modern structural methodologies developed for studying biologically relevant complexes in their native environments. Cryo-EM and tomography, in particular, yield a wealth of information about large protein complexes, subcellular substructures and internal segments of the cell. Nickel et. al. first used cryo-EM tomography to obtain three-dimensional morphological information about a vitrified ROS structure from murine retina [86]. As one of the largest structures imaged by this technique, this allowed distance measurements among the various membrane components of the ROS to define the space available for phototransduction and provide a glimpse into the unfixed three-dimensional architecture of this highly differentiated neuron. Wensel and colleagues extended these cryo-EM tomography studies to obtain 3D maps of the connecting cilium and adjacent cellular structures of the ROS, a modified primary cilium, from wild-type and genetically defective mice [4]. After finding that the ciliary rootlet is involved in cellular transport and stabilizes the axoneme, they proposed a model for disc morphogenesis in which basal discs are enveloped by the plasma membrane.

Imaging studies of the retina remain at the forefront of progress in reconstruction analyses to visualize multiple cells simultaneously at high resolution. Novel applications include scanning EM (SEM) coupled with serial ion ablation (SIA) technology [87], electron microscopic high resolution imaging reconstruction such as serial block-face electron microscopy (SBEM) [88], and two-photon microscopy [89]. These continuously evolving techniques provide indispensable information about the higher order organization of the retina at subcellular resolution.

Conclusions

Progress in the structural biology of individual photoreceptor proteins must be followed with studies to assign their position within the supramolecular assemblies that underlie visual function. Hybrid microscopy techniques such as cryo-EM at sub-nanometer resolution, hybrid serial sectioning combined with high resolution transmission EM and correlative microscopy already allow imaging of retinal layers in unprecedented detail. Moreover, serial sectioning-coupled scanning and transmission EM permit three dimensional reconstructions that expedite understanding the global pathological features associated with genotypic disorders. The final challenge is to understand at an atomic level what therapies can be employed to prevent devastating blinding diseases.

Highlights.

• Recent structural advances related to vertebrate visual transduction proteins are described.

• Progress pertaining to rhodopsin and its complex with rod G protein is summarized.

• The RPE65 structure reveals a novel mechanism for retinoid isomerization.

• Structural analyses of other proteins involved in visual processes are reviewed.

• New approaches to study relevant cellular complexes are discussed.