Proteins that stabilize the structure of outer-segment discs

Rhodopsin is not only integral to the phototransduction cascade, it is also required for the formation and maintenance of OS discs; in the absence of rhodopsin, the OS structure is not formed.²² Peripherin and Rom-1^15,34,35 are two other disc membrane components that contribute to maintaining the flattened structure of the disc. Both of these proteins belong to the tetraspanin family of integral membrane proteins that form large multiprotein complexes known as the tetraspanin web or tetraspanin-enriched microdomains.³⁶ Peripherin/rds is normally found within the edge or “rim” of the disc. It self-associates to form higher-order complexes and also interacts with Rom-1.³⁷ Peripherin/rds and Rom-1 likely function as adhesion molecules that assist in keeping the discs vertically aligned, and may stabilize the disc stack with bridges to the overlying plasma membrane. Peripherin/rds is also thought to function in forming the curvature of the disc rim.³⁸

Deletion or disruption of peripherin/rds results in malformation of discs at the OS base. The naturally occurring retinal degeneration in the mouse named “retinal degeneration, slow” (rds) has been shown to be due to a defect in peripherin/rds. The absence of peripherin/rds in the rds/rds mouse prevents normal development of the photoreceptor OS and leads to photoreceptor cell death.^15,39 Transgenic mice that lack peripherin/rds fail to form an OS. When the level is reduced, as in the heterozygous mice, large whorls of disc membrane are formed instead of an organized stack of uniform discs.^15,39 Mutations that affect the quaternary structure of peripherin/rds also lead to malformed discs. These observations underscore the importance of peripherin/rds in the formation and maintenance of disc structure in the OS.

Although Rom-1 is analogous to peripherin in structure and function, its absence in rods shows a less severe phenotype than peripherin/rds knockout mice, suggesting that peripherin/rds is more critical for disc formation. Although the discs grow larger than normal sizes, photoreceptor function appears to be retained to a much higher level than when peripherin/rds is absent. Nevertheless, abnormal Rom-1 does lead to slow and progressive photoreceptor degeneration both in mice and in humans.⁴⁰

Functional differences of peripherin/rds and Rom-1 have been noted in rods and cones. Cones appear to have a lower ratio of Rom-1/peripherin,³⁶ and different peripherin/rds mutations appear to affect rods and cones differentially. For example, the C214S and N244K mutations have a greater effect on rods whereas R172W and N244H tend to cause cone-dominant diseases such as macular dystrophy.³⁶

Transmission electron microscopy has revealed structural elements between the lamellar discs as well as between the discs and the plasma membrane that are mainly localized to the rim region and incisures of discs.^41–44 A study of OS structure using cryoelectron tomography of vitrified OS showed previously unobserved spacers distributed throughout the discs.³⁰ These structural elements likely maintain the proper distance between adjacent discs and between discs and the plasma membrane. The protein identities of these spacers are not known, but peripherin/rds and Rom-1 may be candidates.

Another component of the rod disc is ABCR (photoreceptor cell-specific ATP-binding cassette transporter).^45–47 Mutations in ABCR are responsible for a large variety of retinal degenerations, including Stargardt macular dystrophy, fundus flavimaculatus, some forms of cone–rod degeneration, and retinitis pigmentosa (RP). Other ABCR mutations are thought to increase the risk of developing age-related macular degeneration (AMD). ABCR is a member of the ABC superfamily. This is a large family of transmembrane proteins involved in energy-dependent transport of many different substrates across membrane “barriers.” The localization of ABCR at the disc rim suggests that it is involved in the movement of molecules from the disc lumen into the cytosol. This hypothesis was supported by studies in ABCR knockout mice, which show delayed dark adaptation, increased all-trans retinal following light exposure, and elevated phosphatidylethanolamine (PE) in the rod OS.⁴⁸ Biochemical analysis of retinas from ABCR knockout mice revealed accumulation of a novel complex of all-trans retinal and PE, termed N-retinylidene-PE, which is not found in normal retina. Once in the cytosol, the all-trans retinol moves to the retinal pigment epithelium (RPE) for recycling and chromophore (11-cis retinal) regeneration (see Chapter 16, Cell biology of the retinal pigment epithelium). The primary pathologic defect in Stargardt disease, and also in ABCR knockout mice, is accumulation of toxic lipofuscin pigments, such as A2E, in RPE cells. Thus the phenotype is remarkably similar between mice and that observed in the fundus of Stargardt patients.

The animal model and further biochemical investigations⁴⁹ suggest that ABCR functions as an outwardly directed flippase for N-retinylidene-PE. Delayed dark adaptation is likely due to accumulation (in discs) of the noncovalent complex between opsin and all-trans retinal. ABCR-mediated retinal degeneration in patients may result from “poisoning” of the RPE due to A2E accumulation, with secondary photoreceptor degeneration due to loss of the ABCR support role.

Though ABCR most definitely plays a role in the structure of discs, no mutations have been conclusively associated with abnormal structural features per se. In other words, it appears that mutations do not directly affect folding at the disc rim, and more likely affect the biochemical function of the transporter. In corroboration of these data, Radu and colleagues tested the effects of isotretinoin on lipofuscin accumulation in ABCR knockout mice,⁵⁰ a model of recessive Stargardt disease. They observed by electron microscopy that isotretinoin blocked both the formation of A2E biochemically, and the accumulation of lipofuscin pigments. Further, no significant visual loss was observed in ABCR null mice by electroretinography (when treated), and isotretinoin also blocked the slower, age-dependent accumulation of lipofuscin in wild-type mice. The results suggest that treatment with isotretinoin may inhibit lipofuscin accumulation and delay the onset of visual loss in patients with Stargardt disease and may be an effective treatment for other forms of retinal or macular degeneration associated with lipofuscin accumulation, though “normal” visual function may be somewhat compromised by such treatment.

The presence of an OS is clearly not optional, as there are many examples of retinal disease in which the photoreceptor cell dies shortly after loss of the OS.^51–53 Why loss of the OS triggers rod cell death is not obvious, since all of the required cellular organelles inhabit the IS and cell body; the OS appears devoid of organelles. The necessity to maintain an OS is further surprising since photoreceptors continually replace all the OS components. In fact, the renewal of rod cell OS components is perhaps more readily seen in the photoreceptor than in any other cell in the body. The disc and plasma membrane components of the OS are completely replaced within 2 weeks.^54,55 To support this turnover there is an influx of newly synthesized proteins and lipids from the IS. These components are transported to the base of the OS, and from there discs gradually migrate towards the RPE for phagocytosis and recycling. One idea is that the high oxygen tension coming from the choroidal vasculature may interfere with the cellular machinery in the IS and nucleus of the photoreceptor, and the presence of the OS puts a protective distance between the choroid and the photoreceptor nuclei.⁵⁶

Disc morphogenesis

Although it is known that new discs are formed at the base of the OS, the process by which these discs are formed is not fully resolved (Fig. 14.2). One model was described in the work of Anderson et al. (model 1).^57,58 Based on morphologic studies of adult monkey rods, it was observed that the folded membrane stacks at the base of the OS are continuous with the plasma membrane of the connecting cilium, whereas the older discs are closed off and separated from the plasma membrane. From this appearance a model was proposed that the newly arrived rhodopsin-bearing vesicles fuse with the plasma membrane, and the growing membrane evaginates to form open discs. Eventually these discs pinch off and form the mature closed discs. Another model (model 2), the vesicular targeting model, was recently proposed by Chuang and colleagues.⁵⁹ They provide molecular evidence that rhodopsin’s carboxyl terminus interacts with a protein called SARA (smad anchor for receptor activation), as well as phosphatidylinositol 3-phosphate (PI3P), and syntaxin 3. In their model, the new discs are already closed and enveloped by the OS plasma membrane. Their evidence suggests that the protein–protein and protein–lipid interactions organized by SARA regulate the vesicular targeting of axonemal rhodopsin-bearing vesicles to the newly formed discs. Thus the discs grow by fusing with the rhodopsin-bearing vesicles and not from the evaginated plasma membrane. This model is more consistent with the morphology of their experimental model system, the mouse rods, which do not display open discs at the base of the OS.

The degree of translational and posttranslational control in the processes of disc formation remains an open question. The processes necessary to manage and partition the correct ratio of rhodopsin to the other proteins of the phototransduction cascade (transducin: PDE: GC: GCAP: cGMP-gated channel) in the discs are unknown. Additionally, the insertion of the structural proteins peripherin, Rom-1, and ABCR into the forming disc rim is likely to require precise instruction. How these processes combine to generate a functional OS will require extensive new investigation and methodology. It is unlikely that these complex morphogenic processes will be elucidated using static two-dimensional microscopy methods with fixed materials, the method that has yielded most information to date. These problems may require the application of three-dimensional microscopy, in conjunction with immunocytochemical tags and fluorescent labels such as green fluorescent protein, to follow the assembling proteins in preparations of living photoreceptors.

Outer-segment plasma membrane

It is clear that, in addition to its obvious role in phototransduction, the OS plasma membrane contains a rich array of specialized proteins, many of which regulate the movement of ions into and out of the OS. The best-characterized proteins of this type are the retinal cGMP-gated (CNG) cation channels. In the dark-adapted rods and cones, Na⁺ and Ca²⁺ flow into the OS through these channels in the plasma membrane. Calcium comprises about 10% of the dark current carried by these channels in rods,⁶⁰ and perhaps 20% or more in cones.^61,62 In both rods and cones Na⁺ is extruded from the IS through the Na⁺/K⁺ pump. This flow of ions sets up the circulating dark current, of which the vast majority is carried by the Na⁺. The probability of the opening of the CNG channel, which in turn determines the size of the circulating current, depends on the amount of free [cGMP], which in the dark is estimated to be 3–4 µM.⁶³ At this concentration, the probability of channel opening is estimated to be only 0.1–0.2.^64,65 This underscores the impact that elevated [cGMP] can have on the number of open channels in the diseased state. The influx of Ca²⁺ through the channel is balanced by an efflux of Ca²⁺ by the Na⁺/Ca²⁺-K⁺ exchanger (NCKX) in the rod OS plasma membrane, thereby maintaining the intracellular level of Ca²⁺ at a relatively constant level.⁶⁶ Kaupp and colleagues⁶⁷ cloned the cGMP channel from bovine retina, while Pittler and colleagues⁶⁸ determined the primary structures of the human and mouse retinal rod cGMP-gated cation channel. These studies found that the sequence of the cGMP channel has significant similarity (59%) to the olfactory cAMP-gated channel. The retinal rod CNG channel is a hetero-oligomer composed of three alpha (CNGA1) and one beta (CNGB1) subunits,^69,70 each with cytoplasmic amino (N)- and carboxyl (C)-termini, six putative transmembrane domains, and a pore region.^71,72 Mutations in CNGA1 and CNGB1 have been linked to disease. A point mutation in CNGB1⁷³ and several mutations in CNGA1⁷⁴ have been identified and linked to recessive RP in humans. The CNG channel in cones consists of two CNGA3 and two CNGB3 subunits.⁷⁵ Together, mutations in either of these genes account for ~75% of complete acromatopsia.^76,77

In addition to the CNG channel, there are several other channels and transporters in the plasma membrane that serve to regulate the intracellular contents of the OS. The best studied of these is the NCKX. Rods express NCKX1 whereas cones express NCKX2.⁷⁸ The exchanger moves with every cycle 4 Na⁺ into the OS, and in exchange, moves one Ca²⁺ and one K⁺ into the subretinal space; exchange is thus electrogenic. Other transmembrane proteins, such as the GLUT-1 glucose transporter, have been shown to be present on both rod and cone OS.⁷⁹

Outer-segment lipids

Docosahexaenoic acid (DHA, 22:6n-6) is the major fatty acid found in the retinal rod OS, and rod photoreceptors have higher levels of DHA than in any other membrane system examined.⁸⁰ Numerous studies have established that the high level of DHA in rod OS membranes provides an optimal microenvironment for rhodopsin. DHA belongs to the n-3 family of essential polyunsaturated fatty acids. These fatty acids cannot be synthesized by vertebrates and they, or their shorter-chain precursors, must therefore be obtained in the diet.⁸¹

Humans with RP and dogs with progressive rod–cone degeneration (prcd) have lower than normal blood levels of long-chain polyunsaturated fatty acids, including DHA. In addition, prcd-affected dogs have lower levels of DHA in their rod OS than control animals.⁸² The reason for the reduced level of DHA in rod OS of animals with inherited retinal degeneration is not known. However, the fatty acid composition of the rod OS of these animals suggests that the synthesis of DHA-containing glycerolipids is downregulated in retinas of animals with inherited retinal degenerations.

Signal activation and amplification

A hallmark of rod phototransduction is its high sensitivity. Fully dark-adapted rods achieve sensitivity that reaches the theoretical maximum, the ability to detect individual quanta of light.^1,2 This high sensitivity is generated through several stages of amplification, resulting in a tremendous increase in the signal gain (Fig. 14.4). This cascade of amplification begins at the light-activated G-protein-coupled receptor rhodopsin (R). Absorption of a photon by R leads to a conformation change (R*), allowing it to interact with the heterotrimeric G-protein transducin (Tα, β, and γ), thereby promoting the exchange of bound GDP for GTP. The complex then dissociates into Tα-GTP and Tβγ, and R* is free to activate many other transducin molecules during its catalytic lifetime.⁸³ In the subsequent step, Tα-GTP binds and removes the inhibitory γ-subunit of the cGMP-PDE so that it can now hydrolyze cGMP to 5’-GMP; the enzymatic activity of PDE (PDE*) in turn degrades thousands of cGMP molecules. The decrease in cytoplasmic cGMP concentration then leads to closure of the cGMP-gated cation conductance channel in the rod plasma membrane, resulting in a reduction in the influx of ~ 1 000 000 Na⁺ ions.⁸⁴ This is a graded effect: a slight lowering in the cGMP concentration leads to closure of some channels, while a large decrease will eventually lead to closure of all channels. In this instance, the rod cell is said to be saturated.

Fig.14.4 (A) The phototransduction cascade in the rod and cone outer segment is initiated when a photon (hν) strikes the visual pigment, rhodopsin in rods, and promotes it to an active state (R*). R* interacts with a heterotrimeric G-protein (transducin) and promotes the exchange of guanosine triphosphate (GTP) for guanosine diphosphate (GDP) on the α-subunit (Tα). Tα in turn disinhibits a cyclic guanosine monophosphate (cGMP) phosphodiesterase (PDE), thereby allowing the hydrolysis of cGMP and the closure of cyclic nucleotide gated (CNG) channels. The closure of CNG channels interrupts the flow of Na⁺ and Ca²⁺ into the photoreceptor and hyperpolarizes the membrane potential. (B) Deactivation of R* is required to quench the phototransduction cascade. R* deactivation is initiated by the phosphorylation Ser and Thr residues at its C-terminus by rhodopsin kinase (RK). RK is inhibited by recoverin (Rv), a process which is controlled by Ca²⁺. Ca²⁺ concentration falls as CNG channels close during the activation of phototransduction, allowing the dissociation of RK from Rv, and subsequently the phosphorylation of R*. The R* catalytic activity is further quenched by the binding of visual arrestin (Arr) to phosphorylated R*. (C) Deactivation of activated PDE (PDE*) requires the hydrolysis of GTP to GDP on Tα. To reopen CNG channels, guanylyl cyclase synthesizes cGMP from GTP, a process that is regulated by the Ca²⁺-binding protein, guanylyl cyclase-activating protein 1 and 2 (GCAPs).

In darkness, glutamate, the transmitter used by both rods and cones, is released at a steady rate because the photoreceptor’s membrane potential is in a relatively depolarized state. The closure of CNG channels results in a graded hyperpolarization of the cell, leading to a decrease in glutamate release at the synapse. In this manner, the first sensation of light perception is transmitted from the photoreceptor cell to second-order cells of the retina, where signals are further processed and ultimately conveyed to the retinal ganglion cells (see Chapter 15, Functional anatomy of the mammalian retina).

Signal deactivation

A reversal of the activation steps is required ultimately for the photoreceptor to return to its resting state. First, the catalytically active components of the phototransduction cascade, R* and PDE*, must be quenched, then cGMP needs to be resynthesized (Fig. 14.4). These events must be rapid and reproducible for the photoreceptor to maintain its high sensitivity and respond to subsequent stimuli. Our current understanding of these processes is detailed below.

Deactivating PDE: control of transducin’s GTPase activity

Quenching of the phototransduction cascade also requires the shutoff of PDE*, which occurs through the deactivation of Tα. The GTPase activity of Tα allows the reassociation of the inhibitory γ-subunit to the catalytic PDE subunits,⁹⁶ a process which is speeded through the participation of a photoreceptor-specific RGS protein, RGS-9, and its binding partner, Gβ5.⁹⁷ These proteins act sequentially on Tα to facilitate the hydrolysis of bound GTP. The physiological role of all these proteins in the shutoff of the rod photoresponse was evaluated with transgenic mice that had alterations in each of these components. Rod photoresponses from transgenic mice expressing a PDE-γ with an impaired ability to stimulate GTP hydrolysis⁹⁸ showed abnormally slowed recovery, demonstrating the necessary role of PDE-γ in Tα deactivation.⁹⁹ Similarly, mice lacking RGS-9 showed profoundly slowed response recovery,¹⁰⁰ as do mice lacking Gβ5.¹⁰¹ RGS-9 is anchored to photoreceptor membranes by the adaptor protein, R9AP.^102–104 Interestingly, human patients with mutations in R9AP have difficulties adapting to sudden changes in light levels that are mediated by cones,¹⁰⁵ consistent with the higher concentration of RGS-9 complex in cones compared to rods.¹⁰⁶ Together, the current data support an important role of PDE-γ, RGS-9/Gβ5, and R9AP in Tα deactivation, and consequently, response recovery, in both rods and cones.

Light adaptation

In a normal cycle of day and night, the illumination at the Earth’s surface varies over 11 orders of magnitude, making photoreceptor adaptation fundamentally important to the normal functioning of the vertebrate visual system. Rods cells are able to adjust their sensitivity over 2–3 log units of light intensities, while cone cells exhibit no response saturation over 6–7 log units.^3,4 The switch between rod and cone function from night to bright daylight covers a large portion of vision’s functional range, and downstream circuitry accounts for the remainder of the range extension. While the molecular events underlying the amplification cascade following photon capture by rhodopsin are well delineated, the mechanisms underlying photoreceptor adaptation are less well understood; however, Ca²⁺ is known to play a feedback role in the adaptation process.^111,112

Adaptation mediated by Ca²⁺ feedback to retinal guanylyl cyclase

It has long been recognized that feedback controlling accelerated recovery contributes to light adaptation.¹⁰⁸ Recovery of the light response requires the resynthesis of cGMP, which is needed to reopen the CNG channels and re-establish the circulating current. The control of cGMP synthesis by RetGC/GCAPs is central to this process. The contribution of accelerated cGMP synthesis to adaptation in rod photoreceptors was investigated in GCAP knockout mice.^23,110 This work showed that the presence of GCAPs extends the sensitivity to background light in rod photoreceptors ~10-fold (Fig. 14.5), and can account for about half of the rod’s capacity to adapt to background light. Recordings from cones in GCAPs knockout mice also show that GCAP contributes to background adaptation, but to a lesser extent than it does in mouse rods.¹¹⁹

Fig. 14.5 Sensitivity as a function of background light was measured from mouse rod photoreceptors using suction electrodes (Fig. 14.3). Sensitivity was calculated from dim flash responses as the peak response amplitude divided by the flash strength. Flash sensitivity (S_F) in the presence of increasing background light, background light (I_B) was normalized against the flash sensitivity in darkness (S_F^D) and is plotted for the rods of several transgenic mice with particular elements of the phototransduction cascade altered. These mice include wild-type (•), mice lacking recoverin (), mice where the calmodulin-binding site on the cyclic nucleotide gated (CNG) channels has been eliminated (), mice lacking guanylyl cyclase-activating proteins 1 and 2 (), and mice where both the calmodulin-binding site on the CNG channel and guanylyl cyclase-activating proteins 1 and 2 have been eliminated (). The red line shows the predicted decline in sensitivity as a function of background light intensity if rods displayed no light adaptation. The blue line is best-fitting Weber–Fechner function for wild-type rods, and shows that light adaptation extends the sensitivity of rod photoreceptors to 100-fold higher background light levels. Rods from mice lacking recoverin, or the calmodulin-binding site on the CNG channels, show little deviation from wild-type rods, indicating that these proteins do not play a significant role in light adaptation. However, in mice lacking guanylyl cyclase-activating protein 1 and 2, light adaptation is impaired, but not absent, even when the calmodulin-binding site on the CNG channel is also eliminated (red dashed line). The remaining mechanisms that control light adaptation in rod photoreceptors remain unidentified.

(Data replotted from Chen J, Woodruff ML, Wang T, et al. Channel modulation and the mechanism of light adaptation in mouse rods. J Neurosci 2010;30:16232–40.)

Recoverin and control of rhodopsin kinase

The accelerated response recovery that mediates light adaptation is also achieved through the speeded deactivation of R*. The first step in the quenching of R* activity is the phosphorylation of its C-terminus by rhodopsin kinase, a process that is believed to be regulated in a Ca²⁺-dependent manner by the protein recoverin. Recoverin is a highly conserved Ca²⁺-binding protein found in both rod and cone photoreceptors. In vitro, recoverin binds and inhibits rhodopsin kinase when it is Ca²⁺-bound, and accelerates rhodopsin kinase activity as Ca²⁺ is removed.^120,121 Since [Ca²⁺]_i is high in darkness, it has been suggested that, under dark-adapted conditions, recoverin prolongs the catalytic activity of R* by inhibiting rhodopsin phosphorylation. Early evidence in experiments that introduce recombinant recoverin into gecko and salamander rods indicated that recoverin prolongs the light response.^122,123 However, recoverin’s effect was most prominent at Ca²⁺ concentrations outside the rod’s physiological range (>1 µM). The Ca²⁺ requirement for rhodopsin kinase inhibition by recoverin in reconstituted biochemical assays was similarly high.¹²⁰ These results placed into doubt the role of recoverin in mediating light adaptation. Suction electrode recordings from the rods of transgenic mice lacking recoverin, however, reveal a response speeding.¹²⁴ This speeding effect is exacerbated in rods that lack recoverin, but also have a reduced expression of rhodopsin kinase.¹²⁵ Thus, in the physiological range of Ca²⁺ concentrations, recoverin can influence R* lifetime. Despite this effect on R* lifetime, recoverin appears to play a relatively minor role in rod light adaptation (Fig. 14.5).¹²⁴ It remains to be seen in cone photoreceptors what is the efficacy of recoverin’s action. However, one might expect a greater contribution to adaptation as experiments from salamander L cones show that R* lifetime dominates the light response.¹²⁶