Retinopathy in Mice Induced by Disrupted All-trans-retinal Clearance*

The visual (retinoid) cycle is a fundamental metabolic process in vertebrate retina responsible for production of 11-cis-retinal, the chromophore of rhodopsin and cone pigments. 11-cis-Retinal is bound to opsins, forming visual pigments, and when the resulting visual chromophore 11-cis-retinylidene is photoisomerized to all-trans-retinylidene, all-trans-retinal is released from these receptors. Toxic byproducts of the visual cycle formed from all-trans-retinal often are associated with lipofuscin deposits in the retinal pigmented epithelium (RPE), but it is not clear whether aberrant reactions of the visual cycle participate in RPE atrophy, leading to a rapid onset of retinopathy. Here we report that mice lacking both the ATP-binding cassette transporter 4 (Abca4) and enzyme retinol dehydrogenase 8 (Rdh8), proteins critical for all-trans-retinal clearance from photoreceptors, developed severe RPE/photoreceptor dystrophy at an early age. This phenotype includes lipofuscin, drusen, and basal laminar deposits, Bruch's membrane thickening, and choroidal neovascularization. Importantly, the severity of visual dysfunction and retinopathy was exacerbated by light but attenuated by treatment with retinylamine, a visual cycle inhibitor that slows the flow of all-trans-retinal through the visual cycle. These findings provide direct evidence that aberrant production of toxic condensation byproducts of the visual cycle in mice can lead to rapid, progressive retinal degeneration.

mechanism-based pharmacological agents to prevent progressive retinal degeneration by aberrant reactions of the visual cycle.
All mice used in this study were housed in the animal facility at the School of Medicine, Case Western Reserve University, where they were maintained either under complete darkness or in a 12-h light (less than 10 lux)/12-h dark cycle environment. For acute light exposure, mice were dark-adapted for 24 h and then exposed to fluorescent light at 10,000 lux for 60 min. Retinal damage was analyzed 7 days after this light exposure. All other manipulations were done under dim red light transmitted through a Kodak No. 1 Safelight filter (transmittance Ͼ560 nm). All animal procedures and experiments were approved by the Case Western Reserve University Animal Care Committees and conformed to both the recommendations of the American Veterinary Medical Association Panel on Euthanasia and the Association of Research for Vision and Ophthalmology.
Synthesis and Analyses of RALdi-RALdi was extracted from six mouse eyes by homogenization with 3 ml of hexane/ethyl acetate (66%: 33%) and use of a glass/glass homogenizer. After evaporation of solvent, dried extracts were dissolved in 150 l of acetonitrile with 0.1% trifluoroacetic acid and passed through a Teflon syringe filter (National Scientific Co., Rockwood, TN). Samples (100 l) were loaded on C18 columns (Phenomenex, Torrance, CA) and analyzed with a gradient of isopropyl alcohol in acetonitrile (0 -25%) containing 0.1% formic acid for 20 min at a flow rate of 1 ml/min. Molecular masses and fragmentation patterns of an authentic standard synthesized by a published method (17) and RALdi extracted from mouse eyes were confirmed by liquid chromatography/mass spectrometry under chromatographic conditions described above with an LXQ mass spectrometer (Thermo Scientific, Waltham, MA) equipped with an APCI source and combined with a 1100 series HPLC (Agilent Technologies, Santa Clara, CA).
Retinylamine Treatment-Retinylamine was synthesized and administered by oral gavage as previously described (18).
Histology and Immunocytochemistry-The histological and immunocytochemical procedures employed were previously described (16).
Retinoid and A2E Analyses-Experimental procedures related to extraction, derivatization, and separation of retinoids from dissected mouse eyes were carried out as previously described (16). Quantification of A2E by HPLC was performed by comparison with known concentrations of pure synthetic A2E (15).
VEGF Quantification-Mouse eyes were enucleated and submerged immediately in ice-cold phosphate-buffered saline. Cornea, lens, and outside connective tissues were carefully removed, and eyes were placed in 500 l of lysis buffer containing 10 mM NaF, 300 mM NaCl, 50 mM Tris, pH 7.4, 1% Triton X-100, 10% glycerol, and 1 mM EDTA with a 1% volume (5 l) of phosphatase and protease inhibitor mixture (Sigma). Samples were sonicated and stored at Ϫ80°C. After two freeze-thaw cycles, they were centrifuged at 14,000 rpm for 10 min at 4°C. Levels of VEGF in the supernatant were determined with a mouse enzyme-linked immunosorbent assay kit (R&D Systems, Minneapolis, MN) according to the manufacturer's instructions. Protein concentrations in the supernatants were measured by the Bradford method (Bio-Rad). SCHEME 1. Retinoid flow in the visual cycle and condensation of all-trans-retinal. After 11-cis-retinal is bound to opsin forming rhodopsin, the resulting visual chromophore 11-cis-retinylidene is photoisomerized to all-trans-retinylidene, the precursor of all-trans-retinal. Most of the all-trans-retinal dissociates from opsin into the cytoplasm, where it is reduced to all-trans-retinol by RDHs, including RDH8. The fraction of all-trans-retinal that dissociates into disc lumens is transported by ABCA4 into the cytoplasm (4) before it is reduced. Thus, condensation products can be generated both within the disc lumens and the cytoplasm. Loss of ABCA4 and RDH8 exacerbates this condensation, which is reminiscent of an accelerated aging process.
Statistical Analyses-Data representing the means Ϯ S.E. for the results of at least three independent experiments were compared by the one-way analysis of variance test.
Retinal Degeneration in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ Mice-Threeweek-old Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice did not display any retinal dystrophy, but dramatic regional retinal degeneration was evident by 4 -6 weeks of age (Fig. 2, A and B). By 3 months of age, retinal changes were advanced as manifested by reduced thickness of the photoreceptor layer, fewer and more scattered photoreceptor nuclei (Fig. 2C), and formation of retinal rosettes by 6 weeks of age ( Fig. 2A). Further progression of retinal degeneration was detected at 6 months of age (Fig. 2, C and D). Cone photoreceptor atrophy, first noted in 4 -6-week-old mice, gradually increased with age (Fig. 3, A and B). Cone and rod degeneration was more severe in the inferior central area than in the superior area (Fig. 3C). 11-cis-Retinal, which quantitatively reflects the level of rhodopsin, decreased with age in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice (Fig. 3C). Retinal abnormalities were not detected in age-matched Rdh8 Ϫ/Ϫ or Abca4 Ϫ/Ϫ mice (Fig.  3A) nor were they found in 3-month-old Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice kept in the dark (data not shown). Retinal degeneration in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice was accompanied by complement deposition (Fig. 3D). ERGs obtained to assess retinal function showed that amplitudes of a-and b-waves under scotopic conditions were significantly attenuated in 3-month-old Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice (Fig. 3E, left) that also displayed reduced flicker ERG responses at 20 and 30 Hz (Fig. 3E, right). These data clearly indicate cone/rod dystrophy in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice. Severely delayed dark adaptation rates similar to those found in Rdh8 Ϫ/Ϫ mice also were observed in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice of the same age (data not shown). The incidence of retinal degeneration in the different knock-out mice is summarized in Fig. 3F. Mice carry two different alleles of Rpe65 that have amino acid variations at position 450 (21), with the Leu variant known to accumulate more A2E than the Met variant (22). Interestingly, Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ 3-monthold mice with the Met variation showed a 43.8% incidence of retinal degeneration, whereas identically aged Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice with Leu exhibited a 97.5% incidence. These rates of retinal degeneration incidence were well correlated with the amounts of A2E found (Met versus Leu; 21.4 Ϯ 2.7 versus 38.8 Ϯ 3.9 pmol/eye at 3 months of age). Therefore, the incidence of retinal degeneration correlates well with the higher visual cycle rates found in mice that carried the Leu 450 variant of Rpe65.
Ultrastructure of Degenerated RPE in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ Mice-The RPE of affected 3-month-old Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice revealed increased thickness with more pigmented granules as compared with the RPE of WT mice (Fig. 4, A-D). We also observed changes in the Bruch's membrane of these double knock-out mice (Fig. 4, E and F). Interestingly, the RPE of Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice exhibited accumulation of lipofuscin, lipid granules, and vacuolarization (Fig. 4G). Some cells in the RPE had died, and their remnants were more obvious in the inferior area of the retina near the optic nerve head (Fig. 4H). Dead cells in the RPE contained undigested ROS/COS and basal laminar deposits that invaded Bruch's membrane (Fig. 4I). Lipofuscin accumulation and vacuolarization around lipofuscin in the RPE also was detected by electron microscopy of these mice at 3 months of age (Fig. 4J), indicating the toxicity of A2E as a cationic detergent (23). Lipofuscin deposits occurred between the RPE and the ROS (Fig. 4K), and drusen deposits were detected under the RPE (Fig. 4L). These findings correspond with abnormalities found in the retinas of humans with age-related macular degeneration (AMD), a leading cause of blindness in developed counties (24,25).
Retinylamine-attenuated A2E Production and Retinal Degeneration-Retinylamine acts on the retinoid cycle by inhibiting 11-cis-retinal production (18,29,31). By slowing regeneration of visual pigments, in turn, production of alltrans-retinal caused by bleaching of these receptors is atten-uated (1). To determine if reduced accumulation of condensation products protects against retinal degeneration, we treated Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice with retinylamine. Treatment with 1 mg of retinylamine by oral gavage was given to 1-month-old mice either once every week or every other week for 3 months. These treatment regimens were chosen, because retinylamine is amidated and stored, providing long lasting efficacy after a single dose (18,29,30). This dose suppressed 11-cis-retinal regeneration up to 1 week in WT mice (30). Under 12-h light/12-h dark conditions, mice undergoing either treatment regimen exhibited reduced A2E levels to near those found in animals kept in darkness (Fig.  7A). Degeneration was not detected anywhere in the retinas of retinylamine-treated mice, whereas vehicle-treated animals developed severe pathology in the inferior cone-rich retina (Fig. 7, B and C). Because 1 mg of retinylamine given every 2 weeks by gavage was effective in preventing marked A2E accumulation and retinal damage, we tested the effects of administering the alternate week regimen to Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice over different time periods. As expected, 7-month-old Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice treated with only vehicle for 6 months showed severe retinal degeneration with the greatest A2E accumulation (Fig. 7, D and E). Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice gavaged at an early age (1 month of age treated for 3 months) showed amounts of A2E similar to those of older mice gavaged for the same period (4 months of age treated for 3 months). A2E levels in both of these groups were significantly reduced as compared with vehicle-treated controls, but they also were higher than levels found in dark-adapted animals. Levels of A2E in young 1-month-old Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice treated for 6 months were only marginally increased as compared with those found in dark-reared mice, and retinal histology showed minimal changes in only 50% of these treated animals. However, untreated 7-month-old dark-reared Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice did have increased amounts of A2E as compared with same age (7 months) WT animals raised under regular 12-h light/12-h dark conditions (Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ versus WT; 17.5 Ϯ 2.0 versus 5.4 Ϯ 0.7 pmol/eye of A2E). Only minimal retinal changes were observed in 67% of these darkreared Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice (Fig. 7, D and E; WT not shown).

DISCUSSION
Herein, we describe a retinopathy in mice induced by disrupted all-trans-retinal clearance. Double knock-out mice lacking ABCA4 and RDH8 manifested retinal abnormalities with hallmark features that include A2E/RALdi/lipofuscin accumulation, RPE/photoreceptor atrophy, drusen, basal laminar deposits, thickened Bruch's membrane, and choroidal neovascularization. Most of these changes were evident in the first 2-3 months of life and were exacerbated by exposure to bright light. Mice lacking either ABCA4 or RDH8 alone failed to show much retinal pathology despite accumulation of A2E (Abca4 Ϫ/Ϫ mice) (13) or delayed clearance of all-trans-retinal (Rdh8 Ϫ/Ϫ mice) (16).
Accumulation of All-trans-retinal Condensation Products in the RPE-A2E accumulation in the RPE was greater in mice lacking the Abca4 gene than in those without the Rdh8 gene. The reason for this difference could be that residual all-trans-retinal remains in the disc lumen longer than in the cytoplasm of Abca4 Ϫ/Ϫ mice, thereby allowing condensation to take place before ROS/COS are phagocytized by the RPE, which are completely renewed every week or so. Thus, all-trans-retinal condensation reactions in Abca4 Ϫ/Ϫ mice would not be linked to the clearance of all-trans-retinal that occurs within 1 h after intense bleach. Recently, RALdi was shown to be more toxic to the RPE than A2E (20). Accumulation of all-trans-retinal leads to formation of RALdi in Rdh8 Ϫ/Ϫ mice and results directly from impaired clearance of all-trans-retinal in the cytoplasm of ROS or COS. Indeed, more RALdi was detected in RDH8-deficient mice than in Abca4 Ϫ/Ϫ mice (Fig. 1F), together with a higher incidence of light-induced retinal changes (100% in Rdh8 Ϫ/Ϫ versus 41.7% in Abca4 Ϫ/Ϫ mice) (Fig. 6). Greater accumulation of all-trans-retinal and RALdi in Rdh8 Ϫ/Ϫ mice implies an important role of these reactive molecules in retinal degeneration. Significantly, Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice accumulated substantially more A2E and RALdi than all of the other tested mutant mice (Fig. 1, C and F).
Retinals and Degeneration of the Retina-Although retinals are essential to vision because of their ability to regenerate visual chromophores, they also are chemically reactive and toxic, so cells have developed specific mechanisms to protect against them. For example, 11-cis-retinal is largely coupled to opsin or CRALBP, whereas all-trans-retinal is quickly removed from internal discs and reduced to the far less reactive all-trans-retinol. Visual cycle retinoids are sheltered from light of wavelengths shorter than ϳ400 nm, because the anterior eye filters it out. But this is no longer true when all-trans-retinal molecules condense to form products that absorb in the range of visible light. Upon excitation, their conjugated double bonds may be efficiently oxidized to form a variety of oxirane derivatives that trigger further radical reactions (32)(33)(34). If all-trans-retinal/Nretinylidene-phosphatidylethanolamine transfer by ABCA4 is slowed, A2PE and RALdi-PE conjugates would first accumulate inside the discs. Then with shedding and phagocytosis of ROS, the RPE could acquire excessive amounts of these products, as evidenced by accumulation of A2PE and RALdi-PE in the RPE from Abca4 Ϫ/Ϫ mice (13) and humans with Stargardt disease (7).
Based on our results, it is reasonable to speculate that retinal degeneration in mice with delayed clearance of free all-transretinal is caused by several factors. Most likely RPE/photoreceptor cells tolerate high levels A2E and RALdi, but the additional presence of reactive compounds, most likely free alltrans-retinal, is detrimental for the eye. In such cases, profound and acute degeneration takes place. Additional animal models will be needed to confirm this hypothesis.
But is the retinopathy observed in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice causally related to abnormal retinoid metabolism or just an associated phenomenon? The dramatic effects of the retinoid cycle inhibitor retinylamine in protecting the retina against bright light-induced damage (Fig. 6) and delaying and ameliorating both A2E accumulation and retinal pathology (Fig. 7) argue strongly for a causal effect involving disrupted all-trans-retinal clearance. Importantly, the best therapeutic effects were obtained by pretreatment with retinylamine, and optimal protection was noted when therapy was initiated in the youngest animals (Fig. 7E). It should be noted that significant therapeutic effects also were observed when retinylamine was administered to mice with other etiologies of retinal degeneration (29,30,35).
Steps in Retinal Degeneration-Retinal degeneration in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice started as a punctuate process at the age of 4 -6 weeks. These local changes correspond to the areas supported by single RPE cells, suggesting that retinal degeneration is promoted by death or impaired function of the RPE (Fig. 2B). By the age of 3 months, degeneration had progressed to the entire inferior retina and was accompanied by massive RPE death. Since retinal degeneration localized to the inferior region of the eye is characteristic of animal models with light-induced retinal damage and genetic alterations within the light-induced signaling cascade (15,27,30,36), the unihemispherical distribution of retinal damage in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice strongly indicates that it also is associated with light exposure, and indeed, such damage was increased by light in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice (Fig. 6). Although pseudorosette formation might be promoted by different degenerative changes in a narrow area, acute induction of retinal degeneration by light exposure in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice well mimics the murine model of NaIO 3 -induced punctuate RPE death and rosette formation (Fig. S1). These findings also indicate that the origin of this retinopathy primarily relates to the RPE, so we propose that aberrant reactions of the visual cycle are responsible for the initial insult to the retina.
Inhibition of the Visual Cycle and AMD-From the positive results in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice, we speculate that visual cycle inhibitors might not only provide optimal therapy for a juvenile macular degeneration called Stargardt disease, but also could be used to prevent disease progression in AMD. Recent FIGURE 6. Light exposure accelerates retinal degeneration. Light exposure (10,000 lux for 60 min) accelerated retinal degeneration in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice as seen by DAPI nuclear staining. Prior to examination, mice were maintained in the dark for 7 days after light exposure. Top, inferior retina of a light-exposed 4-weekold Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mouse shows widely distributed pathology, including severe photoreceptor loss with rosette formation. Bottom, mild changes occurred in light-exposed retinas from 3-month-old Rdh8 Ϫ/Ϫ and Abca4 Ϫ/Ϫ mice as compared with normal retina from light-exposed 3-month-old WT mice. Retinylamine (Ret-NH 2 ) treatment (1-mg gavage 6 h prior to the light exposure) prevented the severe degeneration observed in retinas of 4-week-old light-exposed Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice. OS, outer segment; IS, inner segment; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. Bars, 20 m. research indicates that the risk of AMD development is enhanced by multiple factors, including genetic background, environmental effects, and immunological reactivity (37,38). However, this work indicates that biochemical aberrations in cellular homeostasis can trigger changes leading to AMD. Evidence presented here indicates that combined deficiencies of RDH8 and ABCA4, resulting in delayed all-trans-retinal clearance and A2E accumulation, successfully reproduce key features of human AMD in a murine model. Although the ABCA4 gene has been reported to modify AMD susceptibility in humans (39,40), Abca4 Ϫ/Ϫ mice failed to mimic human AMD (13). The Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mouse model, in contrast, presents several advantages over previously reported murine models. First, it reveals all of the major features of AMD, providing genetic proof that abnormal reactions of the visual cycle can cause progressive changes similar to AMD. Moreover, these changes occurred quite early (i.e. in the first 3 months of life). Different clinical phenotypes of dry type (atrophic) and wet type (exudative) human AMD might suggest that the pathogenesis of atrophic RPE/retina and choroidal neovascularization formation have different origins. However, our findings in Rdh8 Ϫ/Ϫ Abca4 Ϫ/Ϫ mice provide strong evidence that dry and wet type phenotypes can both be induced by the same genetic defect.