The Birth and Death of Photoreceptors The Friedenwald Lecture

Rod and cone structure is intriguing. These cells have unique compartments called outer segments that capture light at one end, the synaptic terminals at the other, and inner segments in between. How does a rod or a cone cell do that? How are molecules synthesized in one site and selectively transported and retained at another so that the cell becomes polarized? As I began postdoctoral research in the Dreyer laboratory at CalTech (California Institute of Technology, Pasadena, CA), Heller showed that rhodopsin could be isolated in quantity. ¹ The early studies of the isolation of outer segments were quite helpful, but they had two defects: homogenization of whole retinas in blenders and contamination by other membranes. The first problem was resolved by George Wald’s Nobel Prize lecture, ² in which he noted that he isolated rod outer segments by gently stroking a camel hair’s brush over the top of a retina and dipping it in buffer to perform his microspectrophotometric studies on individual outer segments that documented the absorption spectra of photopigments. Thus, the connecting cilium was a fragile joint that could be used for low-shear homogenization. The second problem, contamination by other membranes in the retina, was resolved by using approaches for isolating membranes from cancer cells that Don Wallach developed and taught to me during a prior fellowship at Harvard. Wallach showed the importance of controlling shear forces as the first step in purifying membranes and that endoplasmic reticulum membranes had a propensity to collapse and extrude water in the presence of divalent cations so that they floated at a higher buoyant density. So, I gently shook off rods from bovine retinas in dense sucrose and floated them in simple step sucrose gradients containing magnesium. Purified rod outer segments floated above the ER and mitochondrial membranes. ³ The resultant membranes provided a clean starting material to assess rod membrane protein composition and also became the basis for many studies by my friends in the Dreyer laboratory, Paul Hargrave and Hermann Kuhn, who came the next year, and Bob Molday, who arrived as I left for Yale. 

The outer segments capture photons and transduce the energy into hyperpolarization that eventually turns off synaptic vesicle turnover in light. A nonmotile cilium is the only constant means of communication between them. This connection not only transmits the graded potential from the outer segment to the synapse, but also serves as the conduit for the transfer of huge numbers of protein and lipid molecules from the inner segment to the outer segment. With the possible exception of a diurnal movement of some cytosolic proteins back and forth between the outer segment and the inner segment, the outer segment membrane proteins all go toward it and stay there until they are discarded. Early studies by Richard Young and his colleagues Dean Bok and Michael Hall at the University of California Los Angeles (UCLA) in the 1960s and 1970s defined these pathways. They were elegantly reviewed by Young in his Friedenwald Award lecture of 1976. ⁴ They showed that protein synthesis is virtually confined to the rough endoplasmic reticulum of the inner segment and that the newly synthesized proteins traverse the inner segment and assemble into the disks of rod outer segments and the lamellae of cone outer segments. 

What Does Rhodopsin’s Structure Tell Us About Its Biosynthesis?

Knowing that rhodopsin is synthesized in the rough endoplasmic reticulum, rather than on free polysomes in the cytoplasm, suggests that it begins its life as a membrane protein. There were few membrane proteins sequenced in the early 1970s, and they all passed through the lipid bilayer once, with the N-terminal either on the extracellular surface or on the cytoplasmic surface. The known exception was bacteriorhodopsin, a bacterial photopigment that acted as a light-activated proton pump to provide energy to the halobacterium. When Carolyn Converse joined my new laboratory at Yale, we set out to determine when the association of rhodopsin and membranes begins. We found that rhodopsin is membrane bound as soon as it is synthesized. ⁵ Like many other proteins, rhodopsin is glycosylated. ¹ Michiko Fukuda, my first graduate student, took on the task of determining the number and structure of its oligosaccharides. Using peptide fragments of bovine rhodopsin prepared by Hargrave, she showed by simple cleavage with chymotrypsin that the N-terminal peptide can be cut in two and that each half has an N-linked hexasaccharide of remarkably simple structure. ⁶  

To be glycosylated on the N-terminal domain means that this portion of rhodopsin crosses the bilayer, because such posttranslational modifications are exclusively displayed on the extracellular (intradiscal) surface. But this finding raises another important issue: how does rhodopsin’s N-terminal cross the bilayer? With this question in mind, during a sabbatical year at the Weizmann Institute, I worked with Israel Schechter who had just succeeded in isolating insulin’s mRNA. Joining the late Ronald Zemell and others in Schechter’s group, we isolated rhodopsin’s mRNA by immunoprecipitation of polysomes with a goat antibody to opsin. We translated the mRNA in vitro and microsequenced radiolabeled rhodopsin. Hargrave generously shared unpublished sequence data on rhodopsin’s N-terminal sequence. Our sequence exactly paralleled Hargrave’s sequence of mature rhodopsin. ⁷ There was no cleaved hydrophobic “signal” peptide preceding the mature N-terminal methionine of rhodopsin. Gunter Blobel, one of the discoverers of the signal peptide hypothesis, immediately asked for the antibody and proceeded to confirm this result (Goldman and Blobel ⁸ ). Although subsequent studies by Friedlander and Blobel ⁹ indicated that later portions of rhodopsin’s sequence have properties that resemble signal peptides and stop-transfer peptides in succession, the question remains unanswered. We still do not understand how a hydrophilic glycosylated N-terminal peptide destined for the aqueous environment translocates across the bilayer, rather than staying on the cytoplasmic side. Comparative molecular analysis and the genome sequence indicate that rhodopsin is a member of a large family of hundreds of heptahelical receptors. The question, therefore, applies to all of them. 

How Do Outer Segments Form?

After photoreceptor precursors complete their last cell division, the centriole is no longer used for separating chromosomes at mitosis. This small organelle then moves from its usual site in the center of the cell near the nucleus and Golgi apparatus to a position just below the apical plasma membrane of the primitive photoreceptor, where it adopts a new function. It becomes a basal body that organizes the extension of microtubules into a cilium. The cilium projects into the interphotoreceptor space. As it elongates, its tip swells, and small vesicles accumulate and gradually enlarge to form poorly oriented disks. Finally, the ciliary tip becomes well organized, and regular arrays of flattened disks form that are oriented perpendicular to the path of oncoming light. ¹⁰ This process does not require light, because it is completed in utero after approximately 6 months of gestation in humans and is thus an intrinsic feature of the differentiating photoreceptor. 

Young’s autoradiographic studies of protein synthesis in rods ⁴ revealed how the rod outer segment then elongates and begins its rhythm of shedding at the tip and renewal of new membranes at the base. Cones renew in a similar way, except that there is membrane continuity of the cone’s outer segment lamellae and plasma membrane that permit the newly synthesized protein to diffuse throughout the organelle. 

The process of disk formation has been only partially described. Steinberg et al. ¹¹ presented a useful model of disks evaginating until they reach full size and then pinching off by a zippering of the rim. Michael Chaitin’s immunocytochemical studies (Chaitin et al. ¹² ) localized actin next to the cilium at the site where new disks form, suggesting that actin may play a role in new disk formation. The proteins with which actin interacts at that site are still unidentified, however. Myosin VIIa localizes to the cilium interior and is mutated in patients with Usher type I retinal degeneration and deafness. ¹³ No other myosin family member has as yet been identified in the newly forming disks. Inhibition of actin function by cytochalasin alters disk formation and inner segment structure in two ways: the lateral calycal processes collapse, and the disks continue enlarging until they spill over the lateral border of the rod inner segment, further indicating that actin is important in disk morphogenesis. ¹⁴ A recent study of a family with a recessive retinal degeneration found that prominin, a protein previously associated with membrane extensions in bone marrow cells and renal epithelial brush borders was also localized to basal disks. ¹⁵ Its sequence suggests it might have an actin binding domain but on its intradiscal side. Thus an increasing number of proteins of newly forming disks are being identified, and it is likely that some of these, like prominin, when mutated interfere with photoreceptor function and longevity. 

The Proteins of the Rod Outer Segment

Once rod outer segments could be isolated free of contamination from other retinal compartments, we showed that rhodopsin is the major membrane protein in the disk and that there is a small amount of another large intrinsic membrane protein, now designated ABCR. ^{3 5 16} Techniques of isolation and analysis did not provide sufficient indication of other proteins that comigrate with rhodopsin in the one-dimensional SDS gels, but soon it was found that other proteins are obscured by the rhodopsin band including α- and β-transducin, peripherin, and Rom-1. ^{17 18 19 20 21 22} The discovery of enzymatic activity of some of the proteins associated with the disk membranes revealed how a photon captured by 11-cis-retinal in rhodopsin alters its conformation and transfers energy to the transducin subunits that eventually hyperpolarize the plasma membrane. Rhodopsin’s conformational change also initiates its phosphorylation by rhodopsin kinase, thereby increasing its affinity for arrestin and shutting down the transduction cascade. Uncovering each of these steps required major effort by dozens of laboratories until the picture emerged that is so widely understood today. These results were summarized by Molday ²³ and Hargrave ²⁴ in their recent Friedenwald addresses. When Bob and Laurie Molday ²⁵ developed a technique for isolating the plasma membrane of rods away from the disks, a new era of molecular understanding of the control of the rod’s permeability to ions was launched and correlated with the physiologic studies of rod behavior. 

Rod Disks Are Not Uniform

Early electron microscopic images of rod disks clearly show that the edges of disks differ from the interior. ^{26 27} The disk margins form a tight curve. Fibrils extending between disks and from the rim toward the plasma membrane were discovered by freeze-etch studies. ²⁸ Our antibody labeling showed that rhodopsin is distributed widely throughout the disk, but that it is excluded both from the edges of the disks on the outer rim and from the deep clefts of the disk, called incisures. ²⁹ By contrast, the large intrinsic membrane protein (ABCR) and peripherin and Rom-1 are exclusively localized only to the rims and incisures of disks. ^{20 30 31} Thus, the disk has two environments on its surface. The interior of the disk is specialized for transduction. The purpose of the proteins confined to the rim remains less clear. Are they solely structural, or do they also have important enzymatic functions, as recently proposed for ABCR? ^{32 33 34 35 36} Mutations of each of these membrane proteins on the rims and incisures are associated with devastating retinal degenerations. ³⁷  

How Do Rods Sort Outer Segment Proteins Exclusively to Their Sites of Function?

Rhodopsin translates and rotates rapidly in the plane of the disk membrane and in the plasma membrane of the rod and cone outer segments. The highly unsaturated lipids of the disk make it one of the most liquid of all membranes. ^{38 39} Despite this lack of constraint within its home territory in the disk, it is not permitted to leave the yard. What is the molecular equivalent of a fence? Because rhodopsin is not made in the outer segment, how does it get there and stay there? Why are the proteins of rod disk membranes not distributed randomly over the surface of the cell or, as George Palade put it, “Why is the invitation to randomize declined?” ⁴⁰  

As the unique composition of the rod outer segment membrane proteins was defined, it became obvious that the rod could confine rhodopsin and its associated proteins to the outer segment. Immunocytochemistry illustrated this feature, because anti-opsin antibodies bound at high density over the disks and plasma membrane of the outer segment, but barely labeled the inner segment plasma membrane. ²⁹ This result supported earlier observations by Besharse and Pfenninger ⁴¹ that the distribution of uniquely sized intramembrane particles revealed by freeze fracture were uniformly present at high density in outer segment disks but were present at low density in the inner segment plasma membrane. Some of the early antibodies also cross-reacted with cone photopigments and revealed similar compartmentalization of the photopigment of cones. ²⁹ Thus, a clearly demarcated boundary was identified at the junction of the inner and outer segments of both photoreceptors. 

Newly synthesized rhodopsin is membrane bound during the synthesis of the polypeptide chain. The components of the cell transporting newly synthesized rhodopsin and the large protein are small post-Golgi tubulovesicular membranes. ^{5 16} Beth Burnside, learning of this result kindly provided me electron microscopic images of specially fixed retinas showing abundant vesicles between mitochondria in monkey retinas (Burnside, written communication, 1975). Besharse and Pfenninger ⁴¹ also found vesicles near the base of the connecting cilium that had intramembraneous particles of the right size to be rhodopsin. Were these vesicles transporting rhodopsin? We already knew from Young’s autoradiographic studies ⁴ that proteins that were synthesized in the rough endoplasmic reticulum of the inner segment passed through the Golgi apparatus before assembling at the base of the outer segment. What mechanism was adopted to accomplish this task of vectorial transport? How was rhodopsin sent north but not south? 

We began a search for a post-Golgi pathway of vesicular transport of rhodopsin to the base of the connecting cilium. To reveal the molecules in the vesicles in the intact cell, Barbara Schneider ⁴² made important improvements in techniques of electron microscopic immunocytochemistry initially invented by Singer (Painter et al. ⁴³ ) and Kraehenbuhl (Kraehenbuhl and Jamieson ⁴⁴ ) and their colleagues. These involved embedding retinas in hydrophilic media such as glutaraldehyde cross-linked bovine serum albumin to preserve the antigenic epitopes of the embedded tissues ^{29 30 42 45 46} Hydrophilic polymethacrylate plastics subsequently simplified electron microscopic immunocytochemistry so that it could be readily adopted by many laboratories. ⁴⁷  

The Golgi apparatus and outer segments are prominently labeled by anti-opsin. Remarkably, the adjacent inner segment plasma membrane is nearly free of detectable rhodopsin, in agreement with earlier studies. ^{29 30 42 45 46} The invention of a new scanning electron microscope capable of very-high-resolution analysis and Klaus Peters’ development of new techniques of thin metal coating of exposed cell surfaces provided us an opportunity to look at the plasma membranes of the inner and outer segment with new clarity. In collaboration with Peters and Palade, Schneider and I found a new domain of the plasma membrane of photoreceptor cells surrounding the base of the connecting cilium of frog rods and cones, the periciliary ridge complex—the PRC. ⁴⁸ This membrane folds into an array of nine ridges and grooves that bend at right angles as the membrane changes from a flat surface to a deep invagination surrounding the base of the connecting cilium. The nine-fold symmetry suggests that some organizing feature arising from the basal body deforms the plasma membrane into this special domain (Fig. 1) . What was its purpose? Immunocytochemistry and freeze-fracture electron microscope studies showed that the vesicles transporting rhodopsin were clustered beneath the PRC. ^{42 48} Was the PRC the destination for the vesicles before the organization of the rhodopsin into the disk membranes at the other end of the cilium? 

With Joseph Besharse and Dennis Defoe who conducted quantitative electron microscopic autoradiographic analysis on the radiolabeled retinas and freeze-fracture immunocytochemistry, Schneider and I found that newly synthesized rhodopsin leaves the Golgi compartment and traverses the large distance to the base of the connecting cilium in the form of tubulovesicular membranes that fuse with the base of the grooves of the PRC. ^{49 50} These vesicles are the “envelope” carrying the rhodopsin “letter,” and the PRC groove is the “mailbox” for the deposit of rhodopsin after it leaves the Golgi “post office.” 

What serves as the fence to restrain rhodopsin from rapidly diffusing out of the grooves of the PRC over the inner segment surface? One hint comes from studies of lipids in the plasma membrane. A collar of cholesterol surrounds the cilium. ⁵¹ The PRC was not yet described at that time, but in retrospect, that cholesterol collar lies just outside the PRC. How is it organized and what keeps it intact? Is it perturbed in injured rods? These questions deserve study, because the boundary function of the environment of the PRC and whatever is its equivalent in mammalian rods is likely to be a crucial function for photoreceptor health. Hints that this function is important come from the analysis of the developing rod, which does not polarize its rhodopsin until the cilium is fully organized. Rhodopsin distributes all over the inner segment and synaptic terminal in the immature rod. ⁵² Similarly, as rods become injured in a variety of forms of inherited retinal degeneration, they loose the capacity to keep rhodopsin confined to the outer segment. Before their death rhodopsin delocalizes in rods as if the cell were immature. ^{53 54 55} Is there a link between delocalization of rhodopsin and rod viability, or is the delocalization simply a secondary reflection of an injured rod that is about to die? 

What Intracellular Molecules Participate in Sorting Rhodopsin to the Outer Segment?

To address this question, biochemical techniques were needed to assess the steps of transport from the Golgi to the PRC. I thought it would be impossible to separate the post-Golgi vesicles from a retinal homogenate by simple sucrose-gradient techniques and spent nearly a decade trying to isolate them by immunoaffinity partitioning, using streptavidin-tagged antibodies and polyethylene glycol-dextran two-phase gradients, ^{56 57} but that effort collapsed when the manufacturer of dextran changed its synthesis and destroyed its useful properties. My only consolation was that we had developed a simple procedure for making streptavidin in large amounts, showed the utility of streptavidin as an immunocytochemical reagent for biotinylated sites, and found its advantage over avidin, because it had a neutral pI (isoelectric point) and therefore did not bind nonspecifically to molecules without biotin. ⁵⁸ We transferred the technology to several companies that made it a useful reagent for electron microscopy, molecular biology, and protein chemistry. It may be the most useful result of my career. 

Forced by circumstance to try simple approaches, I suggested to Dusanka Deretic, who joined us in Texas, that we try sucrose gradients, because studies of Golgi and post-Golgi pathways in the liver were making progress with that approach. ⁵⁹ She masterfully developed reproducible high-resolution gradients that recapitulated the cell’s structure in the centrifuge tube. ⁶⁰ In Deretic’s sucrose gradients, the densest gradient fractions contain the rough endoplasmic reticulum, the intermediate fractions contain the Golgi and plasma membranes, and, to our great delight and very good fortune, the lightest fractions contain the post-Golgi vesicles well separated from all the rest of the homogenized membranes in the gradient. These low-density membranes become radiolabeled at just the same time of incubation as electron microscopic autoradiographic studies had said they should (i.e., ∼1 to 1.5 hours after the onset of incubation with [³⁵S]-methionine and cysteine), indicating that they were kinetically past the Golgi. Incubation of the retinas with brefeldin A, a fungal antibiotic that blocks vesicle recycling between the endoplasmic reticulum and Golgi blocks their formation. They constitute only 0.03% of a retinal homogenate, and yet they contain 85% of the radiolabeled rhodopsin at the appropriate time point. ⁶⁰  

With the post-Golgi vesicles now isolated we could learn what else was in their membranes. Deretic noted several small proteins approximately 20 to 25 kDa in size. One set are newly synthesized αA- and αB crystallins (Deretic et al. ⁶¹ ). This was a surprise, because these proteins were thought to be expressed only in lens. Phosphorylated crystallins are in the cytosolic fraction, but only unphosphorylated crystallins are membrane bound to the post-Golgi vesicles transporting rhodopsin. The crystallins are members of the heat shock protein family and may act as chaperones, but their low molar ratio to rhodopsin in the vesicles eliminates the possibility of frequent interaction with rhodopsin. Might they bind the vesicles to the motor proteins transporting them to the PRC? 

Would some of the other small proteins bind GTP, like the ras family of small GTP binding proteins? Several bands are labeled by [³²P]-GTP and their identification again benefited from good fortune. These proteins are so highly conserved that antibodies to the similar canine small G-proteins prepared in Kai Simon’s laboratory cross-react with the proteins of the frog retina. Moreover, the proteins migrate in two-dimensional SDS-isoelectric focusing gels with similar size and charge. In this way, we could identify these proteins to be members of the rab subfamily and at least four members—rabs 3A, 6, 8, and 11—are bound to the vesicles. ^{62 63 64} Using the anti-canine-rab antibodies for immunocytochemical studies in frog retinas we found that rab6 localizes mostly to the Golgi and the post-Golgi vesicles near the Golgi. Anti-rab8 labels the distal vesicles nearest the PRC and also labels the distal end of the connecting cilium where new disks form. ^{62 63 64}  

Deretic then developed a technique for forming the post-Golgi vesicles in vitro from a homogenized retina (Deretic et al. ^{65 66} ). When the retinas are homogenized after 60 minutes of incubation with radiolabeled amino acids, the labeled proteins are in the Golgi fractions of the sucrose gradient used to analyze their distribution. By breaking the cells open at that time but not diluting the homogenate significantly, the intrinsic capacity to form post-Golgi vesicles is retained. During subsequent incubation with unlabeled amino acids, the labeled proteins move into the post-Golgi vesicles. This powerful technique permitted Deretic to add macromolecules to the homogenate to determine their effect on post-Golgi vesicle formation. She found that addition of antibodies to the C-terminal domain of rhodopsin, but not antibodies to rhodopsin’s N-terminal, inhibit vesicle formation (Deretic et al. ^{65 66} ). Because the N-terminal is inside the vesicles and the C-terminal is exposed on the cytoplasmic surface of the vesicle, this result showed that the inhibition is a specific property of the specificity of the antibody and the appropriate orientation of rhodopsin in the lipid bilayer of the vesicle. Furthermore, the function of rab6 is inhibited by GDI, a protein that inhibits the dissociation of GDP from the active site of rab proteins so that they cannot recycle back into the active GTP-bound form. ⁶⁷ A synthetic C-terminal peptide could compete for the binding of the rhodopsin in the forming post-Golgi vesicle and prevent their formation as well. ⁶⁶ Thus, the C-terminal domain of rhodopsin and rab6 are required to move rhodopsin out of the Golgi and into the post-Golgi vesicle in preparation for transport. 

In the Golgi, rhodopsin’s oligosaccharides are posttranslationally modified. ^{6 29} The vesicles are sorted in the trans-Golgi network, and they separate from vesicles heading to the opposite end of the cell, the synapse. ⁶⁸ One candidate for moving the rhodopsin-laden vesicle along microtubules has been identified: a small subunit of a dynein light chain called Tctex-1. ⁶⁹  

At this point, we seemed to be getting insight into the composition of the post-Golgi vesicles, the envelope carrying the rhodopsin letter, and some of the elements of the sorting machinery (i.e., rab6 and rhodopsin’s C-terminal), but the mailbox at the PRC was still only described morphologically. We could not isolate the cytoplasmic surface of the PRC for a comparable series of biochemical studies—only in vivo studies would work. This required yet another invention: the procedure of introducing genes into the frog to evaluate their effects on cell function—a procedure to make transgenic frogs. 

Transgenic Frogs: the Breakthrough

Frankly, I was at a loss, unable to imagine an approach that would reveal functions of the PRC in vivo, because the PRC was only found in lower vertebrates (amphibians and reptiles). The solution was invented by a graduate student, Kris Kroll, and a postdoctoral fellow, Enrique Amaya, in John Gerhart’s laboratory at the University of California at Berkeley. ⁷⁰ Permeabilized sperm are incubated with linearized DNA containing the promoter of choice fused to the cDNA of jellyfish green fluorescent protein (GFP) and the gene to be studied. This DNA expression cassette integrates into the sperm’s genomic DNA. By fertilizing eggs with the transgenic sperm nuclei, they could easily produce transgenic tadpoles expressing GFP. They used the cardiac actin promoter to drive GFP expression and made tadpoles with a green beating heart. After reading their paper, I called Joe Besharse and found that he was already involved in such studies to investigate circadian rhythms in Xenopus. Together with Barry Knox’s group, who had isolated the Xenopus rhodopsin promoter (Batni et al. ⁷¹ ), they had expressed GFP in rods. ⁷² Knox gave me the promoter-GFP construct. Kroll generously taught me the technique. 

Orson Moritz and Beatrice Tam joined my new laboratory at the University of Connecticut Health Center (UCHC) in 1997, and we began making various transgenic frogs. Soon, they improved the technique, increased the yield of surviving tadpoles, and started making large numbers of constructs and tagging the expressed molecules with GFP. What follows is largely their work, which involves transgenic tadpoles’ expressing over 70 different transgenes in just 4 years, an unprecedented accomplishment that would have been impossible in transgenic mammals, because their eyes take so much longer to develop, and they are so much more expensive to generate and house. 

The eyes of Xenopus laevis tadpoles begin to form a few days post fertilization (dpf) and begin to make rhodopsin at 5 dpf. By 14 dpf their eyes are quite mature. ^{73 74} Because the transgene is inserted into the sperm, the tadpoles are not chimeras, and no breeding is necessary, as it is in transgenic mice. After 8 to 12 months, transgenic frogs can be bred and generate hundreds to thousands of offspring frequently. Thus, the scale of the research is both accelerated and expanded to large numbers and is quite inexpensive compared with studies of mammals. Moreover, unlike rodents, the Xenopus retina is rich in both rods and cones, ⁷⁵ so that the interactions of these two photoreceptor classes could be readily evaluated. We planned our studies to evaluate rhodopsin transport but expected that such studies, by introducing mutations, would also give us insights into retinal degeneration. We were not disappointed. 

Rhodopsin Transport in Transgenic Frogs: What Is the Address?

Fragments of the rhodopsin sequence might contain the information for correct sorting to the outer segment. Alternatively, the address could be generated by the complex folding of the entire molecule, bringing it in contact with partners on either or both sides of the lipid bilayer of the vesicle and with partners on its lateral borders in the membrane. Some hints as to the answer to this issue are provided by humans with mutations of the rhodopsin gene that induce their rods to die in a complex set of disorders grouped together as retinitis pigmentosa. Of particular interest were those with mutations of the penultimate proline or deletion of the last five amino acids of the rhodopsin sequence. These patients have particularly severe forms of the disease, often losing their sight by midlife. ⁷⁶ Transgenic mice expressing these mutations display one of the hallmarks of a sick rod: The rhodopsin is delocalized and spreads over the entire surface of the rod from the outer segment to the synaptic plasma membrane. ^{77 78}  

In collaboration with Tomoko Nakayama, Moritz and Tam studied bovine rhodopsin with GFP tucked into the middle of the C-terminal domain (rho-GFP-CT). ⁷⁹ This fusion protein could be expressed in COS cells and forms a photopigment that binds 11-cis-retinal and absorbs light with the proper rhodopsin spectrum. It activates transducin at rates approximately 50% of rhodopsin and is only partially phosphorylated, presumably because of steric hindrance of rhodopsin kinase by GFP. Transgenic frogs expressing rho-GFP-CT transport the protein entirely to the outer segment as if it were not modified. We could readily see this by simple fluorescence microscopy. They also made a second rhodopsin-GFP fusion protein without the C-terminal 15 amino acids of rhodopsin (rho-GFPΔCT). This fusion protein goes to the outer segment but also spreads throughout the lateral plasma membrane of the inner segment and the synapse. This result suggests that the address of rhodopsin is the C-terminal of the protein. 

Next, Tam prepared over 15 different constructs to evaluate this interpretation. I will illustrate only a few of the key pieces of evidence, because her work, involving 15 different transgenes has been fully described (Tam et al. ⁸⁰ ). The C-terminal 44 amino acids of Xenopus rhodopsin are very conserved when compared with the sequence of mammalian rhodopsins. Two cysteines in the C-terminal domain are modified by palmitoyl groups that insert into the lipid bilayer. The remainder is exposed on the cytoplasmic surface. ^{81 82 83 84} These lipid acylated sites provide a convenient membrane anchor for a fusion protein in which GFP is fused to the C-terminal peptide of rhodopsin (GFP-CT44). GFP-CT44 is faithfully transported to the outer segment in contrast to GFP, which fills the cytoplasm and nucleoplasm of inner segments (Figs. 2A 2B) . By contrast, deletion of five amino acids from the C terminus or substitution of other amino acids for the penultimate proline leads to the distribution of the fusion protein to the lateral plasma membranes and synaptic membranes, as well as the outer segment. This provides the same sort of loss-of-function evidence previously shown in transgenic mice that expressed rhodopsin with mutations similar to those described earlier. To evaluate the generality of this finding, Tam next made a fusion protein of GFP with the C-terminal peptide of the α-adrenergic receptor (GFP-AAR), a heptahelical G-protein coupled receptor that is a member of the large multigene family that includes rhodopsin. GFP-AAR distributes to the synaptic terminal of the rod. When its terminal eight amino acids are replaced with the last eight amino acids of rhodopsin, its destination changes to the opposite end of the cell: It goes exclusively to the outer segment (Figs. 2C 2D) . Thus, the address for rhodopsin transport in rods resides in its carboxyl-terminal, within the last eight amino acids or less. ⁸⁰  

This result can be compared with the targeting of rhodopsin in polarized epithelial cells. Using epitope-tagged rhodopsin and rhodopsin fragments, rhodopsin was delivered to the apical surface of well-polarized Madin-Darby canine kidney (MDCK) cells in culture. Similarly, the C-terminal domain also directed the epitope tag to the apical membrane, but deletion of the distal C-terminal sequence had no impact on localization. ⁸⁵ This result is the exact opposite of the subdomains of the C-terminal that are required for directed transport of rhodopsin or GFP-CT fusion proteins to the outer segments of rods. ⁸⁰ Thus the differing results are a consequence of the cellular context of the expression. More such experiments are needed to clarify these issues in vivo. 

What Is the Role of the Small G-Protein Rab8 in the Transport of Rhodopsin?

This question arises from the immunocytochemical studies by Deretic et al., ⁶³ showing that frog rab8 localizes near the base of the connecting cilium. Does rab8 play a role in docking to the PRC? Because we could not design a biochemical approach to isolate the PRC with its cytoplasmic surface exposed to macromolecular manipulation, we needed an in vivo approach. Key regents were provided by Johan Peranen, who studied two mutant forms of rab8 in tissue cultures of MDCK cells and hippocampal neurons (Peranen et al. ⁸⁶ ). One mutation (rab8-Q67L) blocked the GTPase capacity of rab8 and held the protein in the GTP-bound form (on). The other mutation (rab8-T22N) could hydrolyze GTP to GDP (off) but could not release its bound GDP so that it could recycle back to the active (on) form by acquiring a fresh molecule of GTP. Peranen gave us the mutant constructs of canine rab8 as well as the normal (wild-type [wt]) form fused to the C terminus of GFP. Fluorescence microscopy of transgenic tadpole retinas reveals that GFP-wt-rab8 distributes essentially to the same internal membrane sites as the native frog protein previously localized with anti-rab8 antibodies (Fig. 3) . The GFP-rab8-Q67L mutant localizes similarly, but gradually induces the rods to die. By contrast, GFP-rab8-T22N has a devastating impact on the rods from the onset of its expression at 5dpf. Immature rods die quickly by apoptosis, leaving behind a nearly all-cone retina devoid of rods except for the few surviving newer rods added at the periphery as the retina grows in size (Fig. 4) . ⁸⁷  

These surviving peripheral rods provided us with a telling insight into the function of rab8. The surviving peripheral rods expressing GFP-rab8T22N accumulate rhodopsin-laden tubulovesicular membranes that fill the space beneath the PRC (Fig. 5) . By contrast, the rods expressing GFP-wt-rab8 look nearly normal. They have fewer vesicles near the base of the PRC. Curiously, rod outer segments of the surviving peripheral rods appear relatively normal, which may indicate that the obstruction is only partial or that there may be compensating molecules that can serve redundantly for the partial loss of function caused by the expression of the mutant GFP-rab8-T22N. The mutant partially blocks the docking of the vesicles with the grooves of the PRC in the peripheral rods. The letter has been successfully brought to the neighborhood, but the mailbox is closed, and undelivered mail piles up around it. Despite this failure of the post-Golgi vesicles to fuse properly, the inner segment plasma membrane is still unlabeled by anti-rhodopsin. Thus, the cell does not adopt another pathway or an alternative destination for delivery of rhodopsin when the vesicle’s fusion with the PRC is obstructed and rhodopsin’s targeting signal is intact. Contrasting this result with the delocalized distribution of mutant rhodopsins suggests that they enter vesicles targeted for multiple subcellular domains and appear in those diverse membranes rather than the outer segment exclusively. The details of these experiments have been described recently. ⁸⁷  

Can the Results of Studies of Rhodopsin Transport Be Applied to Other Proteins of Photoreceptors or to Other Cells?

Given that the frog rod synthesizes 50,000 molecules of rhodopsin each minute and that no other outer segment protein exists in a molar ratio greater than 1:10, it is that much more difficult to study the biosynthesis of proteins made in small amounts. We have been able to show that the large protein of the disk rim, now identified as ABCR, traverses the photoreceptor cell inner segment and arrives in the outer segment with similar kinetics. ¹⁶ Does that mean that this protein is cotransported with rhodopsin? It is difficult to conduct immunocytochemical studies with the high degree of certainty that the high levels of rhodopsin permit. Nonetheless, one study claims to identify alternative post-Golgi pathways for rhodopsin and peripherin transport. ⁸⁸ We are now beginning numerous studies of intrinsic membrane proteins of rod outer segments using transgenic frogs to resolve these questions. 

How Do Animal Models of Retinal Degeneration Aid in Understanding Human Blindness?

Once it was found that inherited and environmentally induced retinal degenerations in mice and rats activate the apoptotic cell death pathway, ⁸⁹ a large number of studies were initiated to determine whether manipulation of the many genes involved in apoptosis could ameliorate the rate or extent of degeneration. This has been a challenging decade of research, so far, with only partial and idiosyncratic results, confounding attempts to generalize their interpretation. In a wide variety of tissue culture and in vivo studies, elevation of expression of Bcl-2 and its homologous proteins reduces apoptosis. Similarly, loss of Bax function reduces apoptosis. ⁹⁰  

Ganglion cells seem to be most protected by these mechanisms but photoreceptor survival has been more difficult to interpret. One extensively studied transgenic mouse line that overexpresses Bcl-2 in rods under the control of the mouse rhodopsin promoter shows partial reduction of rod death after light exposure. It also induces a transient delay in cell death of rd and rds mutant mice. ^{91 92} Other transgenic mice expressing the same gene under control of the same promoter ⁹³ or under the control of the IRBP promoter ⁹⁴ do not show such survival, despite clear evidence that the product is produced and that the product has appropriate anti-apoptotic properties in other cell systems. The issue of where the transgene inserts into the genome of the transgenic mouse and its phenotypic consequences becomes important when such discrepancies arise and will be more easily resolved now that the mouse genome sequence is nearly completed. 

Clinical studies of retinal degenerations clearly document the failure of attempts to follow the syllogism that one mutation equals one disease. Numerous previously distinguished retinal disorders can be generated by single mutations of rhodopsin, peripherin, or ABCR. Moreover, the variation in penetrance within a kindred with autosomal dominant retinitis pigmentosa highlights the role of other genes or environmental stimuli that can either accelerate or reduce degeneration. For example, in the original family expressing a mutant rhodopsin (P23H), the propositus, a 49-year-old woman was blind from loss of both rod and cone function. Her older sister drove a truck at night as her occupation and was startled to learn she had the same mutation and no rod function but sufficient surviving cone function. She was not sick—that is, in her opinion she had no disease—but that impression was solely the consequence of survival of a sufficient fraction of her cones and her use of lights at night. Why did her sister’s cones die? They did not express rhodopsin. Cones use related genes coding for color vision. What genes in the sighted sister kept her cones alive? If we understood this, the “disease” of blindness could be eliminated, even if the rods died. Similarly, in a man bearing a rhodopsin T17M mutation, the top half of his retina survived, whereas the lower half and the fovea (all cones) died, so that he had profound loss of useful vision, yet could see objects below the equator. If mutations of rhodopsin are lethal to rods, what saved the rods of the top of his retina? 

These fascinating cases illustrate that there are cures, if we can find them, and that the genes do not necessarily have to be replaced by gene therapy, if only we can keep the cells from dying, because surviving rods with mutations function to old age. ⁹⁵ So far, this hypothesis has been tested fully only in flies. Expression of anti-apoptotic genes in Drosophila photoreceptors cause the cells to survive, even when they express a mutant rhodopsin that otherwise kills them. ⁹⁶ Thus, the search for sight-preserving genes and anti-apoptotic agents is justified by the clinical and experimental studies and encourages all of us involved in such studies that the search will not be in vain. 

 Submitted for publication October 11, 2001; accepted November 2, 2001.

 Commercial relationships policy: N.

 

Figure 1.

 

The Periciliary Ridge Complex (PRC) is formed by 9 ridges and grooves surrounding the base of the connecting cilium. Arrow points to an opsin laden vesicle fusing with the groove of the PRC. Reprinted, with permission, from Papermaster DS, Schneider BG, Besharse JC. Vesicular transport of newly synthesized opsin from the Golgi apparatus toward the rod outer segment: ultrastructural immunocytochemical and autoradiographic evidence in Xenopus retinas. Invest Ophthalmol Vis Sci. 1985;26:1386–1404. © Association for Research in Vision and Ophthalmology.

Figure 2.

 

Comparison of the distribution of GFP and GFP fusion proteins in the rods of transgenic frogs. (A) GFP fills the cytoplasm and nucleoplasm of the inner segment and synaptic-terminal. Only a small amount is found in the rod outer segment (ROS) because the impenetrable disks occupy most of the volume there. (B) GFP-CT44 is a fusion protein of GFP with the last 44 amino acids of Xenopus rhodopsin. It is membrane bound because its cysteines are palmitoylated and it localizes nearly exclusively to the ROS. (C) GFP-AAR is a fusion protein of GFP with the last 39 amino acids of the C-terminal of the α-adrenergic receptor (AAR), a heptahelical receptor resembling rhodopsin in overall structure. In contrast to GFP-CT44, GFP-AAR localizes predominantly to the synaptic terminal. (D) GFP-AAR-CC-CT8 is a fusion protein of GFP, the C-terminal of AAR which has had its last 8 amino acids replaced by the last 8 amino acids of Xenopus rhodopsin. The switch of this small domain changes the distribution of the fusion protein from the synaptic terminal (C) to the ROS to resemble GFP-CT-44 (B). Thus the localization signal of rhodopsin in rods resides within the last 8 amino acids of its sequence. Reprinted, with permission, from Tam BM, Moritz OL, Hurd LB, Papermaster DS. Identification of an outer segment targeting signal in the COOH terminus of rhodopsin using transgenic Xenopus laevis. J Cell Biol. 2000;151:1369–1380. © Rockefeller University Press.

Figure 3.

 

GFP-wt-rab8 is a fusion protein of GFP and the normal sequence of canine rab8. The green fluorescence of the protein is contrasted with that of membrane glycoproteins labeled with rhodamine-wheat germ agglutinin (red). The nucleus (N) is stained blue with Hoechst dye. The yellow color of the Golgi (G) and the post-Golgi pathway is a result of the overlap of the GFP-rab8 and the glycosylated membranes of the Golgi, and post-Golgi compartments. Note the bright spot near the junction of the inner and outer segments between the mitochondria (M). This likely represents the vesicles that are docking at the PRC. Some GFP-rab8 is seen at the synapse (S). Reprinted, with permission, from Moritz OL, Tam BM, Hurd LL, Peranen J, Deretic D, Papermaster DS. Mutant rab8 impairs docking and fusion of rhodopsin-bearing post-Golgi membranes and causes cell death of transgenic Xenopus rods. Mol Biol Cell. 2001;12:2341–2351. © The American Society for Cell Biology.

Figure 4.

 

Retinas of transgenic X. laevis expressing GFP-rab-T822N or GFP-rab8-wt. GFP-rab-T822N is an “off” mutant form of rab8 fused to the C-terminal of GFP. This protein is toxic to newly forming rods. Its expression is turned on just as the cells differentiate. (A) GFP-rab8-T22N Most of the rods have died by 14 days leaving behind a nearly all cone retina. Frog cones have refractile oil droplets in the inner segment. Only a few newly added rods survive at the periphery. (B) GFP-wt-rab 8. The rods survive and the retinal organization is normal. H&E stained sections of 14 dpf tadpole retinas. From Moritz et al. submitted for publication.

Figure 5.

 

Electron micrograph of the junction of the inner and outer segments of a peripheral rod expressing GFP-rab8-T22N. The mutant induces the accumulation of tubulo-vesicular membranes adjacent to the PRC. Controls expressing GFP-wt-rab8 have only a few vesicles at that site (not shown). Reprinted, with permission, from Moritz OL, Tam BM, Hurd LB, Peranen J, Deretic D, Papermaster DS. A rab8 mutant causes accumulation of rhodopsin-bearing membranes and cell death in transgenic Xenopus rods. Mol Biol Cell. 2001;12:2341–2351. © American Society for Cell Biology.

I am indebted to the American Cancer Society and the Anna Fuller Fund for my first research support. The National Eye Institute has generously supported my studies since 1973 with Grants EY-845, EY-3239, EY-6891, EY-6892, and EY-10992. The early immunocytochemical studies were supported by National Institute of General Medical Sciences Grant GM-21714 and the Veterans Administration. Our work on transgenic frogs has been supported by the Foundation Fighting Blindness and National Eye Institute grants. I am most grateful for the enthusiasm, insight, and creativity of my students, fellows, research associates, and collaborators whose work is highlighted in this address. They continue to teach me daily.