Signal Transducing Membrane Complexes of Photoreceptor Outer Segments

Abstract

Signal transduction in outer segments of vertebrate photoreceptors is mediated by a series of reactions among multiple polypeptides that form protein-protein complexes within or on the surface of the disk and plasma membranes. The individual components in the activation reactions include the photon receptor rhodopsin and the products of its absorption of light, the three subunits of the G protein, transducin, the four subunits of the cGMP phosphodiesterase, PDE6 and the four subunits of the cGMP-gated cation channel. Recovery involves membrane complexes with additional polypeptides including the Na⁺/Ca²⁺, K⁺ exchanger, NCKX2, rhodopsin kinases RK1 and RK7, arrestin, guanylate cyclases, guanylate cyclase activating proteins GCAP1 and GCAP2, and the GTPase accelerating complex of RGS9-1, G_β5L, and membrane anchor R9AP. Modes of membrane binding by these polypeptides include transmembrane helices, fatty acyl or isoprenyl modifications, polar interactions with lipid head groups, non-polar interactions of hydrophobic sides chains with lipid hydrocarbon phase, and both polar and non-polar protein-protein interactions. In the course of signal transduction, complexes among these polypeptides form and dissociate, and undergo structural rearrangements that are coupled to their interactions with and catalysis of reactions by small molecules and ions, including guanine nucleotides, ATP, Ca²⁺, Mg²⁺, and lipids. The substantial progress that has been made in understanding the composition and function of these complexes is reviewed, along with the more preliminary state of our understanding of the structures of these complexes and the challenges and opportunities that present themselves for deepening our understanding of these complexes, and how they work together to convert a light signal into an electrical signal.

Introduction

The physiological function of rod and cone outer segments is the conversion of a light signal into an electrical signal. The biochemical cascade responsible for this conversion is known as the phototransduction cascade, and the major events in it are carried out by complexes of multiple polypeptides embedded in or attached to the surface of disk membranes and plasma membranes of the outer segments. Additional membrane complexes maintain the structures of these highly specialized membranes, and establish their proper spatial relationships.

Over the past two decades considerable progress has been made in identifying the major membrane proteins that make up these complexes and carry out their functions. Many of them have been purified from the retina, and all of their gene sequences are now known, allowing them to be studied in heterologous expression systems and in genetically engineered animals. Thus, despite the formidable challenges that continue to be faced in the biochemical characterization and structure determination of membrane proteins, much headway has been made in understanding the structure and function of membrane complexes important for phototransduction and outer segment structure. Understanding their structure-function relationships is important both for deepening our appreciation of the molecular mechanisms of vision, and for understanding the diseases that develop as consequences of disruption of the structures and functions of these complexes. The impact of progress in studying membrane complexes of the photoreceptors is felt well beyond the field of vision research, as these complexes have served as powerful models for understanding membrane complexes that mediate signaling pathways and membrane structures throughout the central nervous system and the rest of the body. One of the most striking examples is the enormous impact of the crystal structure of rhodopsin (Palczewski, et al., 2000) on the fields of G protein coupled receptors and membrane proteins.

An ongoing challenge and source of fascination is the role of the lipid milieu in which these complexes function. They have evolved to work optimally in an environment formed by a membrane bilayer with a highly specialized lipid composition. Most structural approaches and many biochemical studies begin by removing the complexes from the lipid bilayer, and one of the areas of research focus in the immediate future will be finding and exploiting creative approaches for determining structure and function in the membrane environment, and understanding the influence of the lipids on the behavior of the protein complexes.

The G protein, transducin, and its multiple membrane complexes

The photon receptor protein, rhodopsin, is a G protein-coupled receptor, and phototransduction is a prototypical G protein-mediated signaling cascade. At the center of this cascade lies the heterotrimeric G protein, transducin, G_αtβ1γ1, (rods) a peripheral membrane protein (the similar but distinct subunits of the rod and cone subunits will be generically referred to here as G_αβγ, with distinctions between rods and cones noted as needed). As the first G protein and the first component of the phototransduction cascade to have its structure determined, its structure and function have been extensively reviewed (Arshavsky, et al., 2002, Birnbaumer, 2007, Bohm, et al., 1997, Chen, 2005, Coleman & Sprang, 1996, Downs, et al., 2006, Hargrave, et al., 1993, Hofmann, 1999, Shichida & Morizumi, 2007, Sprang, 1997a, Sprang, 1997b, Sprang, 2000, Sprang, et al., 2007). Much of this work has focused on soluble forms of transducin and its component subunits, whereas the focus here is on its membrane dependent complexes.

Lipid modifications & structure of membrane-bound heterotrimer in GDP state

The α subunit of transducin, G_αt, is the more dynamic half of the heterotrimer, moving rapidly among at least three conformational states, and shuttling back and forth between binding partners on the membrane surface. It has one of four different 12- or 14-carbon fatty acids (DeMar, et al., 1999, DeMar, et al., 1996, Kokame, et al., 1992, Neubert, et al., 1992, Yang & Wensel, 1992) attached in an amide linkage to its N-terminal glycine residue, and these provide for modest membrane binding affinity, which varies somewhat depending on the hydrophobicity of the fatty acid (Johnson, et al., 1994, Lobanova, et al., 2007, Neubert & Hurley, 1998, Neubert et al., 1992). However, it is its interactions with other membrane proteins that keep it tethered to the disk membrane in rods under dim light conditions (where it functions in signaling) and likely membrane-bound in cones over most illumination conditions (Coleman & Semple-Rowland, 2005, Kennedy, et al., 2004); however, see (Chen, et al., 2007). In its inactive GDP-bound form, which predominates in the dark, the arrangement of its “switch” domains favors binding to its partner subunits, G_βγ. G_β and G_γ bind to one another very tightly and have a mutual dependence for proper folding and stability. The intrinsic affinity of G_βγ for the disk membrane is higher than that of G_αt and is partly mediated by the presence of two hydrophobic modifications on G_γ: the cysteine residue which is the fourth residue from the carboxyl terminus in the initial translation product is methyl esterified after the last three residues are proteolytically cleaved, and a farnesyl group is attached in a thioether linkage to this same residue (Bigay, et al., 1994, Fukada, 1995, Fukada, et al., 1990, Lai, et al., 1990, Ohguro, et al., 1991). Both G_α and G_βγ bind membranes, with a higher affinity displayed by G_βγ than G_α-GDP (Bigay et al., 1994), which binds more tightly than G_α-GTP. However, for the heterotrimer, it is G_α-GDP that provides most of the membrane binding interactions (Seitz, et al., 1999, Zhang, et al., 2004b).

Studies with reconstituted vesicles or with spin-labeled lipids in disk membranes have revealed specificity in the interaction of transducin complexes with phospholipids (He, et al., 2004, Hessel, et al., 2003, Malinski & Wensel, 1992, Melia, et al., 2000, Melia, et al., 1999, Murray, et al., 2001).

A structure of the G_αβγ complex bound to GDP and a membrane bilayer was determined by cryo-electron microscopy of two-dimensional (helical) crystals of the complex bound to tubules of lipid bilayers (Melia et al., 1999, Zhang et al., 2004b). The structure reveals lipid interactions of both the amino-terminal and carboxyl terminal regions of G_α, and of the carboxyl terminus of G_γ , with no apparent lipid contact with G_β. Two caveats of this structure are that it reveals a static picture, whereas in reality, in the absence of crystal contacts there is likely considerable dynamic motion of the hydrophilic surface of the heterotrimer with respect to the membrane surface, and that the contacts shown may be biased in favor of those favored by electrostatic attraction to positively charged lipids used for crystallization. Studies of transducin complexes in micelles and vesicles suggest that a lipid-like milieu enhances the effective affinity of G_βγ and G_α for one another, likely as a result of their both having lipid moieties. The combination of interactions of the two lipid tails with the membranes and of the G_α and G_βγ polypeptides with one another, produces a cooperativity of membrane binding of the two subunits (Bigay et al., 1994).

Complex with rhodopsin and photoexcited rhodopsin (R*)- progress and challenges

The affinity of transducin for disk membranes is not likely to be due entirely to its interactions with lipids. Although it can diffuse freely between photoreceptor outer and inner segments on a timescale of minutes, in the dark all three subunits of rod transducin are found almost exclusively in the outer segment. When a substantial portion of rhodopsin is bleached (i.e., to a level well beyond the point of saturation for rod vision), all three subunits translocate passively to the inner segment, most likely as separate G_α-GTP and G_βγ units(Lobanova et al., 2007, Rosenzweig, et al., 2007). One possible explanation for these results is a higher affinity of G_α for dark disk membranes than for partially bleached membranes, which would suggest that the relatively low affinity of the heterotrimeric form of transducin for the dark state of rhodopsin (Alves, et al., 2005) is sufficient to allow sequestration of the G protein to the outer segments in the dark. This rhodopsin-transducin-GDP complex has received relatively little attention, and is deserving of more thorough characterization. Alternatively, the lower membrane affinities of G_α and G_βγ separately for membranes, as compared to the higher membrane affinity of the heterotrimer, may be important in the net translocation.

The most attention has been focused on the complex between photoexcited rhodopsin, metarhodopsin II or R*, and the transducin heterotrimer (Fig. 1). Likely there are multiple states of this complex, and currently there are high resolution structures for none of them. The highest affinity form seems to involve the nucleotide-free form of G_αβγ , which is a key intermediate in the nucleotide exchange reaction (release of GDP and binding of GTP) catalyzed by R* as its central role in phototransduction. However, even this complex may exist in multiple conformations, and the complexes involving bound GDP or GTP must be considered as well. The structural and kinetic characterization of all of these complexes will be an important area of research in the coming years. Significant insights into their properties have been obtained by electron paramagnetic resonance (EPR) studies of the related protein G_i bound to rhodopsin in detergent, and by NMR and fluorescence studies of complex formation(Abdulaev, et al., 2006, Abdulaev, et al., 2005, Brabazon, et al., 2003, Knierim, et al., 2007, Medkova, et al., 2002, Oldham, et al., 2006, Ridge, et al., 2006). Kinetic studies using light-scattering and other methods (Bayburt, et al., 2007, Ernst, et al., 2007, Heck & Hofmann, 2001, Herrmann, et al., 2004, Herrmann, et al., 2006a, Herrmann, et al., 2006b, Oprian, 1992) have also shed light on the mechanisms of transducin activation by R*, but much remains to be done. There is reason for optimism that structures derived from three-dimensional or two-dimensional crystals of R*-transducin complexes will be forthcoming in the near future.

Figure 1.

Proposed model for first complex formed in photoactivation. Following absorption of light and photoisomerization by one subunit of the rhodopsin dimer, a heterodimer of rhodopsin and Metarhodopsin II (R*) is formed, which quickly complexes with the G protein, transducin, in its heterotrimeric form. A conformational change within G_αt allows release of bound GDP, which then allows binding of GTP and dissociation of activated G_αt–GTP. The representation of the complex is purely schematic, as only indirect low resolution information is available on the structures of the actual complexes. The major contacts with R* are provided by G_αt, with both its carboxyl and fatty acylated amino termini known to be involved. G_βγ are also important for R* binding, with potential interactions with the membrane hydrocarbon phase provided by the farnesyl group on G_γ. For making the structure images shown, PDB files 1U19.pdb(Okada, et al., 2004) and 1GOT.pdb(Lambright, et al., 1996) were used with UCSF Chimera (Meng, et al., 2006, Pettersen, et al., 2004).

PDE6, and complexes with G_αt-GTP

The only known physiological function of activated GTP-bound G_α, is activation of its downstream effector enzyme, the cGMP-specific phosphodiesterase, PDE6. PDE6 is yet another lipidated peripheral membrane protein of the phototransduction cascade. It consists of four subunits of three kinds with a stoichiometry of PDE6_αβγγ. PDE6_α and PDE6_β are catalytic subunits with similar structures; in cones there are two identical copies of a single type of catalytic subunit, PDE6_α’, most closely related to PDE6_β. These catalytic subunits have modifications identical to those described above for G_γt, except that PDE6_α is farnesylated, whereas PDE6_β is geranylgeranylated in mammals (Anant, et al., 1992, Catty & Deterre, 1991,Ong, et al., 1989, Qin, et al., 1992, Qin N, 1992), while the situation is reversed in frogs (Yamazaki, et al., 2002). These modifications at the carboxyl termini are critical for positioning PDE6 on the membrane surface where it interacts with activated GTP-for G_αt, and proteolytic removal of the carboxyl termini, or binding of a prenyl binding protein(Zhang, et al., 2004a), also known, somewhat misleadingly, as PDE6_δ, releases PDE6 from disk membranes (Cook, etal., 2000, Cook, et al., 2001, Deterre, et al., 1988, Florio, et al., 1996, Li, et al., 1998, Norton, etal., 2005, Wensel & Stryer, 1986). The PDE6_γ subunit is a 9.7 kDa inhibitory polypeptide that keeps PDE6 catalytic activity at a very low level in the dark. Much of the action of G_α–GTP on the activity of PDE6 is mediated through its interactions with the PDE6_γ subunit, and complexes of PDE6_γ with either GDP-form or GTPγS form PDE6_γ have been studied in solution(Antonny, et al., 1993, Artemyev, et al., 1993, Skiba, et al., 1995, Slepak, et al., 1995). Indirect evidence suggests that two molecules of G_α–GTP bind to each PDE6 heterotetramer. In solution, the affinity of G_α–GTP for holo-PDE6 is relatively low, but phospholipid surfaces enhance their interactions dramatically, leading to nearly stoichiometric complex formation when both are membrane bound. This lipid-mediated interaction is enhanced by either positively charged lipids, which are not found in rods, or by phosphoinositides, especially phosphatidylinositol (4,5) bisphosphate (PIP₂) which is present in disk membranes at low levels (He et al., 2004, Melia etal., 2000, Womack, et al., 2000). The physiological significance of the PIP₂ interactions remains to be determined.

Low resolution structures of PDE6 have been determined by electron microscopy of the complex in heavy metal negative stain(Kajimura, et al., 2002, Kameni Tcheudji, et al., 2001). Electron microscope images of PDE6 in frozen solution or bound to vesicles have been recently obtained, as have images of quasi-crystalline arrays of PD6 bound to GTPγS-form G_αt. (Z. Zhang, F. He and T. Wensel, unpublished observations), so that low- to medium resolution structures of these membrane complexes should be forthcoming in the near future.

RGS9-1-G _β5L-R9AP and its complex with G _αt· –GTP

The other important membrane-associated complex formed by G_αt in the course of phototransduction is its complex with the GTPase accelerating complex consisting of the GTPase accelerating protein (GAP) RGS9-1, G_β5L, and the membrane anchor, R9AP (Fig. 2). Indirect evidence suggests that under physiological conditions this complex may also be bound to PDE6, so that assuming a 2:1 stoichiometric ratio of G_α and its bound GAP complex to PDE6, one such complex could have a minimum of 12 membrane-associated polypeptides. Although high resolution structures have been determined by x-ray crystallography for a complex of G_α with a C-terminal fragment of PDE6_γ and the catalytic core of RGS9-1 (Slep, et al., 2001, Sowa, et al., 2001), as well as for a nearly full-length complex of RGS9-1 with G_β5L, (Cheever, et al., 2008), we are far from knowing how this large multi-subunit membrane complex is organized and how it is positioned with respect to the membrane surface. It is clear that association of RGS9-1 with its membrane anchor, R9AP, a member of the syntaxin super-family with a single transmembrane helix(Hu & Wensel, 2002, Hu & Wensel, 2004, Hu, et al., 2003), not only dramatically enhances its affinity for the membrane, but also is critical for its catalytic GAP activity and for its stability in rod cells(Baker, et al., 2006, Keresztes, et al., 2004, Keresztes, etal., 2003, Krispel, et al., 2006, Nishiguchi, et al., 2004).

Figure 2.

A multi-subunit complex essential for normal photoresponse recovery kinetics. Sub-second GTP hydrolysis by activated G_αt–GTP is catalyzed by the GTPase accelerating complex of RGS9-1, G_β5L, and single-pass trans-membrane anchor protein, R9AP. Indirect evidence suggests that formation of this complex occurs while G_αt–GTP is bound to its effector PDE6, largely through interactions with PDE6_γ, whose C-terminal fragment, PDE6_γ’, (Slep et al., 2001) is shown in red space-filling representation. For making the structure images shown, the RGS domains from PDB files FQJ.pdb(Slep et al., 2001) and 2PBI(Cheever et al., 2008) were ligned in UCSF Chimera (Meng et al., 2006, Pettersen et al., 2004) and the resulting assembly of models positioned next to a model of holo PDE6, based on unpublished cryo-electron microscopy data kindly provided by Dr. Zhixian Zhang. The schematic representation of R9AP was loosely derived from a model presented previously(Cheever et al., 2008) The spatial relationships of PDE6_γ to holo PDE6, and of all of the polypeptides with respect to the membrane are not intended to be accurate; however, the attachment of the complex to the membrane via insertion of the isoprenyl tails of PDE6 and of the transmembrane segment of R9AP into the membrane is based on substantial biochemical evidence.

Additional Complexes important for recovery

There are several membrane associated complexes that are critical for the recovery phase of phototransduction. As their structures remain to be determined, and they have been recently reviewed, they will only be discussed briefly here.

R*-Rhodopsin Kinase

An essential step for a return to the dark state of the phototransduction cascade is phosphorylation of R* by rhodopsin kinase, (GRK1; in some species a related kinase, GRK7 is found in cones), which belong to a family of serine/threonine kinases specific for activated states of G protein-coupled receptors (Arshavsky, 2002, Hurley, et al., 1998, Maeda, et al., 2003, Penn, et al., 2000, Pitcher, et al., 1998, Premont & Gainetdinov, 2007). The phosphorylation sites are found in the carboxyl-terminal tail of rhodopsin, which lies near the surface of the lipid membrane, and the transient complex between rhodopsin kinase and rhodopsin is a membrane-dependent kinase. Both GRK1 and GRK7 are isoprenylated, and also have the other covalent modifications at their C-termini described for G_γt above, with GRK1 being predominantly farnesylated (C15) and GRK7 having a C-terminal sequence directing geranylgeranylation (C20). An intriguing possibility is that the faster inactivation of cone pigments by phosphorylation as compared to rhodopsin (Tachibanaki, et al., 2005, Wada, et al., 2006) is related to differences in membrane binding between GRK1 and GRK7. An unresolved question in photoresponse termination is the role of the calcium binding protein recoverin, of the neuronal calcium binding protein family and calmodulin superfamily(Chen, 2002, Chen, et al., 1995,Kawamura & Tachibanaki, 2002, Klenchin & Bounds, 1995, Otto-Bruc, et al., 1998). It has been proposed that recoverin plays a key role in allowing lowered intracellular calcium concentration to serve as a feedback signal for activation of rhodopsin kinase in response to photoactivation. However, results of biochemical studies with permeabilized rods challenge this role(Otto-Bruc et al., 1998), and the effects of a knockout of the recoverin gene on photoresponse recovery kinetics, while in the right direction to support this hypothesis, are rather subtle(Makino, et al., 2004, Sampath, et al., 2005). One feature of recoverin’s interactions with Ca²⁺ is clear: binding of Ca²⁺ induces a conformational change termed a “myristoyl switch” which causes extrusion of an N-terminal fatty acyl group from a pocket within the protein structure to an exposed state which would be expected to maximize interactions with the hydrocarbon phase of the phospholipids membrane. (Ames, et al., 1996, Zozulya & Stryer, 1992).

The ability of rhodopsin kinase to effect R* inactivation depends to a large degree on the action of the capping protein, visual arrestin(Vishnivetskiy, et al., 2007). Although there has been relatively little attention paid to interactions of arrestin with membrane lipids, an intriguing observation is that incorporation of acidic phospholipids, which are relatively abundant in disk membranes, enhances the interactions between arrestin and phosphorylated R* in detergent micelles (Sommer, et al., 2006).

Guanylate cyclase-GCAPs

Another membrane complex between transmembrane and peripheral membrane proteins is the complex between the single pass transmembrane protein, photoreceptor guanylate cyclase (GC; GC1 and GC2), and its associated guanylate cyclase-activating proteins or GCAPs (Baehr, et al., 2007, Koch, 2002, Koch, et al., 2002, Palczewski, et al., 1994, Pugh, et al., 1997, Yu, etal., 1999). Unlike recoverin, deletion of the GCAP genes has a dramatic effect on photoresponse recovery kinetics, as well as on the peak amplitudes of dim flash photoresponses (Burns, et al.,2002, Howes, et al., 2002, Mendez, et al., 2001). GCAP complexes are necessarily formed at the membrane-cytoplasm interface with GC, which is an integral membrane protein. Like recoverin, they have an N-terminal sequence directing fatty acylation, but do not seem to undergo a “Ca²⁺-myristoyl switch” mechanism. Structural studies by x-ray crystallography and NMR have revealed that the myristoyl group is embedded in the protein structure of GCAP1, but that the fatty acyl group of GCAP2 inserts into the membrane bilayer.(Stephen, et al., 2007, Vogel, et al.,2007) Myristoylation influences the affinity of GCAP1 for GC and its Ca²⁺ sensitivity, but not that of GCAP2 (Hwang & Koch, 2002a, Hwang & Koch, 2002b, Olshevskaya, et al., 1997). In photoreceptor membranes, guanylate cyclase dimerizes with itself as well as binding GCAPs, and these interactions are critical to its activation when intracellular calcium concentrations fall in response to light (Hwang, et al., 2003, Olshevskaya, et al., 1999, Ramamurthy, et al., 2001, Tucker, et al., 1999, Yu et al., 1999).

Plasma membrane complexes

Multi-subunit protein complexes mediate the light regulated ionic fluxes through the plasma membrane of photoreceptor outer segments which are integral to light responses (Kaupp& Altenhofen, 1992, Molday, et al., 1999). The cyclic GMP-gated cation channel is a heterotetrameric complex of three α (CNGA1) and one β (CNGCB1) subunits(Trudeau & Zagotta, 2002, Weitz, et al., 2002, Zheng, et al., 2002, Zhong, et al., 2002); cones have a similar but apparently symmetric assembly of two CNGA3 and two CNGB3 subunits (Peng, et al., 2004). These complexes are in turn associated with the major calcium extrusion protein of outer segments, the Na⁺/Ca²⁺,K⁺ exchanger (Bauer, 2002, Schnetkamp, 1989). Although progress has been made in establishing the stoichiometry of the participants in this complex, and in the functional roles of specific domains and residues within them (Bradley, et al., 2005, Kaupp & Seifert, 2002, Matulef & Zagotta, 2003, Shibukawa, et al., 2007), little is known about their structural arrangement. Electron microscopy and single-particle analysis have been used to determine a low resolution structure of the CNG channel(Higgins, et al., 2002), and there is a high resolution structure of a cyclic nucleotide-binding domain similar to that of the photoreceptor CNG channel(Zagotta, et al., 2003).

In addition to its association with the Na/Ca exchanger, the CNG channel also binds calmodulin (Molday, 1996, Trudeau & Zagotta, 2002) at a site between the N-terminal GARP (glutamic acid-rich protein) domain and the transmembrane domain of the β subunit (CNGB1) , and may be involved in interactions with an alternative protein product of the channel β subunit gene in which the glutamic acid rich protein or GARP (Korschen, et al., 1999, Molday & Molday, 1998, Pentia, et al., 2006) is expressed as a soluble protein without the transmembrane and cyclic nucleotide-binding domians. GARP (there are two splice variants of GARP, a minor form, GARP1, and a major form, GARP2) has been proposed to bind to multiple membrane proteins of both the plasma and disk membranes and may be involved in forming some of the connections between these non-continuous membranes (see below).

Membrane Complexes of the disk rims

An additional membrane compartment, in addition to the planar surface of the disk membranes or the gently curved surface of the plasma membrane is the rim region of the disk and the corresponding region of the cone plasma membranes where there is an extraordinarily low radius of curvature and a unique set of membrane protein complexes. The tetraspanin proteins peripherin/rds and ROM1 are important for maintaining the unusual membrane structure in this region (Goldberg, 2006, Molday, et al., 1987, Molday et al., 1999). These proteins do not act as monomers, but rather as multimers, with a non-covalent heterotetramer of peripherin/rds and ROM1forming higher order multimers through covalent disulfide linkages. These disulfide linkages may form in the disk lumen and serve as molecular “staples” to hold together the closely apposed bilayers of either side of each disk membrane. While several studies have described the interactions among these proteins and their ability to induce sharp membrane curvature into heterologous membranes (Wrigley, et al., 2000) little is known about the three dimensional structure of these important membrane protein complexes. Also confined to the disk rims is a photoreceptor-specific member of the ATP binding cassette family known as ABCR or ABCA4(Allikmets, et al., 1997, Azarian & Travis, 1997, Illing, et al., 1997, Shroyer, et al., 2001, Sun & Nathans, 1997, Wiszniewski, et al., 2005). Rather than playing a structural role, this transport protein has been proposed to serve as a lipid-flippase, possibly for the covalent Schiff’s base adduct between all-trans-retinaldehyde and the amino group of phosphatidylethanolamine (APE). It seems likely that localization to the disk rims is mediated by complex formation between ABCR and other rim-specific proteins such as peripherin/RDS and ROM-1. Although some structural information has been obtained for other members of the ATP binding cassette family which ABCR likely resembles in many of its structural features, its three-dimensional structure and the structure or even existence of its complexes with other rim proteins remain to be determined. This determination may be facilitated by the existence of experimental protocols for expression and purification of ABCR in functional form (Ahn & Molday, 2000, Sun, et al., 1999).

Mystery Proteins Controlling Membrane Structure

There may be additional membrane protein complexes important for photoreceptor structure and function whose components have yet to be identified. Freeze-etch electron microscopy has revealed connections, likely formed by membrane-associated proteins, between the disks and the plasma membranes, and between the rims of adjacent disks(Roof & Heuser, 1982, Roof, et al., 1982). A recent study of mouse disk membranes by cryo-electron tomography revealed very large protein complexes connecting adjacent disks, randomly distributed over the disk plane(Nickell, et al., 2007). Identifying the constituents of all these intermembrane complexes remains a fascinating challenge to be overcome in the next few years.

Influence of membranes on kinetics and thermodynamics of protein-protein interactions

A recurring theme that has emerged in the study of membrane protein complexes in phototransduction is that the membranes do much more than simply organize the protein components. By greatly increasing the local concentrations of protein binding partners and by reducing their conformational and orientational freedom they dramatically increase their effective affinities for one another. These effects are evident not only for the heteromeric complexes, but likely also for the rhodopsin homo-dimer, since the form it takes in disk membranes is likely incompatible with those it assumes in detergent micelles and reconstituted membranes, where it can be found either as dimers or monomers, (Bayburt et al., 2007, Fotiadis, et al., 2003, Jastrzebska, et al., 2004, Li, et al., 2004, Liang, et al., 2003, Schertler, et al., 1993). It is worth noting that the existence and/or functional importance of rhodopsin dimers remains a subject of some controversy (Bayburt et al., 2007, Chabre, et al., 2003, Chabre & le Maire, 2005, Fotiadis et al., 2003, Hanson, et al., 2007).

In principle membranes can have a similar effect on kinetics both by the concentration effect and by the reduction in dimensionality from three to two dimensions. At this point the diffusion kinetics in photoreceptors have been reported only for rhodopsin (Cone, 1972, Liebman, et al., 1982, Montal, 1976, Poo & Cone, 1973, Poo & Cone, 1974, Wey, et al., 1981), so it will be important to document the dynamic behavior of the other membrane protein components of the phototransduction cascade within intact photoreceptors. It seems likely that the varying lipid environments presented by lipid microdomains within disk membranes will exert varying effects on behavior of the protein complexes (Boesze-Battaglia, et al., 2002, Martin, et al., 2005, Nair, et al., 2002, Nickell et al., 2007, Senin, et al., 2004, Seno, et al., 2001). Extensive evidence exists for the modulation of rhodopsin’s behavior, as well as that of the downstream phototransduction complexes by lipid composition(Alves et al., 2005, Botelho, et al., 2006, Brown, 1994, He et al., 2004, Koenig, et al., 2002, Litman, et al., 2001, Malinski & Wensel, 1992, Melia et al., 2000, Mitchell, et al., 2001, Mitchell, et al., 2003a, Mitchell, et al., 2003b, Niu, et al., 2001, Niu, et al., 2002, Womack et al., 2000)

Remaining challenges and future prospects

In reviewing the current state of our understanding of membrane protein complexes of photoreceptor outer segments it is clear that the protein composition of most of the major complexes has been determined, much is known about the biochemistry of the polypeptides involved, and through spontaneous or engineered mutations in animal models or humans, much is known about their physiological functions. The weakest link in our knowledge is in the structures of these complexes, especially as they exist in vivo, embedded in or attached to the surface of the photoreceptor membrane. Part of the reason is clearly that multi-subunit membrane complexes remain the most challenging subjects for x-ray crystallography, which has been by far the most commonly used approach for determination of protein structure. New approaches, or new applications of existing alternative approaches, are needed if progress on this front in the next decade is to exceed the progress made over the past decade. Some of these approaches, involving cryo-electron microscopy and electron cryo-tomography are especially well suited to determining membrane structure in the environment of a membrane bilayer. Spectroscopic techniques, including solid-state NMR, electron paramagnetic spectroscopy of spin-labeled proteins, and fluorescence energy transfer techniques are emerging as powerful methods for extracting structural information about membrane protein complexes over a range of spatial resolutions. Also promising in this regard is the development of miniature bilayer membranes stabilized by lipoproteins, known as nano-disks or bicelles (Bayburt, et al., 2006, Boldog, et al., 2007, De Angelis & Opella, 2007, Leitz, et al., 2006, McKibbin, et al., 2007, Prosser, et al., 2006, Struppe, et al., 1998) These have begun to be exploited for the study of photoreceptor proteins (Bayburt et al., 2007, McKibbin et al., 2007) and may offer a route to application of high resolution structural techniques to bilayer-embedded proteins. However, these alternative approaches are still in the process of development, and are being pursued by a relatively small number of laboratories. We can only hope that an appreciation of the importance of establishing and extending new approaches to membrane protein complexes will motivate sufficient support from funding agencies, scientific journals, research institutions and others to keep these efforts going through the awkward stages faced by all truly innovative scientific endeavors.