|Home | About | Journals | Submit | Contact Us | Français|
In the aging human eye, oxidative damage and accumulation of pro-oxidant lysosomal lipofuscin cause functional decline of the retinal pigment epithelium (RPE), which contributes to age-related macular degeneration. In mice with an RPE-specific phagocytosis defect due to lack of αvβ5 integrin receptors, RPE accumulation of lipofuscin suggests that the age-related blindness we previously described in this model may also result from oxidative stress. Cellular and molecular targets of oxidative stress in the eye remain poorly understood. Here we identify actin among 4-hydroxynonenal (HNE) adducts formed specifically in β5−/− RPE but not neural retina with age. HNE modification directly correlated with loss of resistance of actin to detergent extraction, suggesting cytoskeletal damage in aging RPE. Dietary enrichment with natural antioxidants grapes or marigold extract containing macular pigments lutein/zeaxanthin was sufficient to prevent HNE-adduct formation, actin solubility, lipofuscin accumulation, and age-related cone and rod photoreceptor dysfunction in β5−/− mice. Acute generation of HNE-adducts directly destabilized actin but not tubulin cytoskeletal elements of RPE cells. These findings identify destabilization of the actin cytoskeleton as a consequence of physiological, sublethal oxidative burden of RPE cells in vivo that is associated with age-related blindness and that can be prevented by consuming an antioxidant-rich diet.
Photoreceptor rod and cone neurons do not renew themselves in the adult human eye. Their long-term viability and function depends on lifelong support by neighboring retinal pigment epithelial (RPE) cells. Because of this critical and continuous dependence of photoreceptors on RPE cells, any functional deficiency of RPE cells ultimately harms photoreceptors and impairs vision.
Like photoreceptors, RPE cells are post-mitotic in the mature human eye and do not turn over. Therefore, they are themselves susceptible to cumulative damage acquired over time. The most profound visible change characteristic for human RPE cells with age is the enormous accumulation of lipofuscin, a complex mixture of partially digested, oxidized protein and lipid photoreceptor derivatives formed and trapped in RPE lysosomes [1–3]. In a harmful positive feedback loop, oxidative stress promotes lipofuscin formation and existing lipofuscin further exacerbates oxidative stress [4, 5]. Lipofuscin also acts as photosensitizer and may contribute to inflammation . Both inflammatory processes and oxidative damage contribute to dysfunction, distress, and death of RPE cells that cause atrophic age-related macular degeneration (AMD), the most common form of AMD that affects millions of the elderly worldwide [7, 8]. It is generally assumed that increasing, sublethal levels of endogenous oxidative stress directly contribute to development of AMD. However, it is unknown which cell type (RPE or photoreceptor cells) and which molecules or mechanisms are particularly vulnerable and damaged by increased oxidative burden in the aging eye.
A major support function of RPE cells is the diurnal phagocytosis and digestion of photoreceptor outer segment tips (POS) shed by photoreceptors in a circadian rhythm. Healthy RPE cells respond to POS shedding with a burst of phagocytosis and complete digestion of shed debris within hours. We previously showed that β5 integrin knockout mice (β5−/− mice) lack the diurnal burst of RPE phagocytosis . Like human patients with atrophic AMD, β5−/− mice lose visual function with age as a consequence of primary deficiencies of RPE cells, which accumulate pro-oxidant lipofuscin . Here, we quantify and specify the effects of oxidative damage to aging retina and RPE and determine its relevance for photoreceptor function. Our results show that protein oxidation destabilizes specifically the actin cytoskeleton of aging β5−/− RPE cells. Furthermore, reducing oxidative burden by consuming a natural-antioxidant-enriched diet is sufficient to prevent actin damage, lipofuscin buildup, and blindness.
Reagents were from Invitrogen (Carlsbad, CA, USA) or Sigma-Aldrich (St. Louis, MO, USA) unless otherwise indicated.
All procedures involving animals were performed according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and reviewed and approved by the Fordham University Institutional Animal Care and Use Committee. Animals were housed under cyclic 12h light:12h dark light conditions, and fed ad libitum. Long Evans wt rats were bred to generate litters for RPE isolation. β5−/− mice characterized in detail previously [9–11] and β5+/+ wt mice of the same genetic background (129T2/SvEmsJ) were bred to generate age-matched groups for aging studies. For dietary supplementation, littermates of β5−/− or wt mice were raised on standard Picolab Rodent Diet 20 5053 (#5053) (Purina Mills, Gray Summit, MO, USA) separated into feeding groups at 3 months of age and fed specific diets until being sacrificed for experiments. Feeding groups received one of the following: #5053 chow (control diet), #5053 supplemented with 0.68% glucose and 0.68% fructose (sugar diet), #5053 supplemented with freeze-dried grape powder containing a natural mix of resveratrol, flavans, flavonols, anthocyanins and simple phenolics supplied by the California Table Grape Commission (grape diet), #5053 supplemented with 1% FloraGlo® (Kemin Health, Des Moines, IA, USA), a patented marigold extract containing 5% lutein and 0.2% zeaxanthin xanthophylls (lutein diet). All diets were custom-mixed at room temperature followed by compression and heating for less than one second to 40°C and pelleting by the TestDiet division of Purina Mills. These procedures had previously been shown to maintain efficacy of compounds. Food intake and body weights of mice on different diets were monitored weekly and remained similar over the feeding periods. Our mice on control or lutein diet consumed 3.3 g of food daily and weighed 32 g, on average. Hence, lutein diet provided 1.65 mg per lutein and 66 μg zeaxanthin per mouse per day or 52 mg lutein/2 mg zeaxanthin per kg bodyweight. Similar lutein/zeaxanthin supplementation had shown earlier to be well tolerated by mice . Supplementation with 170 mg lutein/kg bodyweight was recently shown to prevent acute light-induced retinal oxidative damage in mice . As comparison, one serving of cooked kale provides 17 mg of lutein/zeaxanthin and commercially available supplements for human consumption commonly provide up to 30 mg lutein and 5 mg zeaxanthin per day. Our mice on sugar or grape diet consumed 2.9 g of food daily and weighed 35 g, on average. Hence, grape diet provided 43 mg freeze-dried grapes daily, which is equivalent to 236 mg fresh grapes per mouse or 6.7 g fresh grapes per kg body weight per day. For human consumption, one serving of table grapes is set as 126 g. Our mice on grape diet consumed ~3.5 servings of grapes per day.
For tissue harvest, mice were sacrificed by CO2 asphyxiation, a method approved by the Panel of Euthanasia of the American Veterinary Medical Association, ~2 hours after light onset. For tissue preservation for sectioning, eyeballs were immediately immersed in Davidson's fixative (32% ethanol, 11.5% acetic acid, 8% formaldehyde). Lens and cornea were removed from fixed eyes before dehydrating and paraffin embedding eyecups. For tissue lysis, lens and cornea were removed immediately after from each enucleated eyeball in chilled HEPES-buffered Hanks saline solution containing calcium and magnesium to generate whole eyecups containing retina and RPE. To separate neural retinas from RPE/choroid and remaining eyecup, we incubated eyecups in HEPES-buffered Hanks saline solution without calcium and magnesium for 5 min to loosen retinal adhesion . We then transferred eyecups to an empty plastic dish and performed a single radial cut toward the optic nerve, flattened the eyecup retina facing up and peeled off the neural retina with forceps. Eyecups, neural retinas and remaining eyecups containing RPE and choroid were used immediately for ELISA's or immunofluorescence assays, or flash frozen and stored at −80°C for up to two weeks for immunoblotting or lipid extraction assays.
Electroretinograms (ERG's) were recorded from age-matched groups of 4–5 male mice each as described previously . Briefly, mice were dark-adapted overnight before anesthesia by intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine. Topical anesthesia was induced with proparacaine hydrochloride. Pupils were dilated with phenylephrine hydrochloride and tropicamide. ERG's testing rod and cone functions in combination were recorded using a UTAS 2000 ERG recording system (LKC, Gaithersburg, MD, USA) with white flash stimuli of 1.5 cd-s/m2 attenuated to yield intensities from −1.8 to 0.2 log cd-s/m2. Stimuli were presented in order of increasing intensity. For each flash intensity, 3 to 4 recordings were averaged. ERG's testing rod and cone function separately were recorded using an LKC UTAS system. Rod responses were recorded by exposing dark adapted mice to −1.6 log cd-s/m2 white flashes. To specifically test cone activity, mice were subjected to a steady rod-desensitizing stimulus of 20 cd/m2 for 5 minutes followed by white flash stimulation at 0.4 log cd-s/m2. Six rod and six cone responses were averaged for each animal. For all recordings, a-wave amplitudes were measured from the baseline to the trough of the a-wave and b-wave amplitudes were measured from the trough of the a-wave to the peak of the b-wave.
Patches of RPE were manually isolated from eyes of adult β5−/− mice or of 9–11-day old Long Evans rat pups following a previously described procedure . Purified mouse RPE was processed for HNE ELISA immediately after isolation. Purified rat RPE cells were seeded in DMEM supplemented with 10% FBS in 96-well plates with or without glass coverslips and incubated at 37°C and 5% CO2 for 5 to 7 days. Antioxidants resveratrol, lutein, or trolox were used at 30 μM, 50 μM, and 20 μM, respectively, as these concentrations had previously been established to be effective but non-toxic for RPE cells in culture [15–17]. They were supplied in complete medium at day 3 and 5 or on day 6 before use of cells on day 7. To generate HNE-adducts, cells were incubated with a fresh solution of HNE (EMD, Gibbstown, NJ, USA) in DMEM for 1 hour before processing for lysis or fixation.
Freshly isolated individual eyes without lens, pooled two isolated neural retinae, two eyecups without neural retina, purified RPE from four eyes, or primary RPE from three wells of a 96-well plate were homogenized in 250 μl phosphate buffered saline on ice using a Tissue Tearor (Research Products International, Mount Prospect, IL, USA) for 15 s. Cleared homogenates obtained by centrifugation at 2000 × g for 3 min were used in series of two-fold dilutions as samples for HNE adduct or protein carbonyl ELISA kits (both Cell Biolabs, San Diego, CA, USA) according to the manufacturer's suggestions and using a Spectramax M2e plate reader (Molecular Devices, Sunnyvale, CA, USA). Protein concentration was quantified using Bradford reagent (Thermofisher, Waltham, MA, USA).
Following removal of the neural retina, eyecups were either fixed immediately with fresh 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) or pre-extracted for 1 min with cytoskeleton stabilization buffer (CSB, 50 mM MES, 5 mM MgCl2, 3 mM EGTA, 0.5% Triton X-100, pH 6.4) at room temperature followed by PFA fixation.
All images were acquired on a Leica TSP5 laser scanning confocal microscopy system (Leica, Wetzlar, Germany). To visualize autofluorescent lipofuscin, PFA fixed samples were immediately mounted in Vectashield containing DAPI nuclei dye (Vectorlabs, Burlingame, CA, USA). Lipofuscin image acquisition was performed by using 488 nm excitation and recording emission from 580 to 670 nm. These imaging parameters are optimized to detect the lipofuscin chromophore A2E. They do not detect emission by all fluorophores known to be present in human lipofuscin. 3-μm stacks of x-y sections 0.24 μm apart were acquired to generate maximal projections representing the nuclear region of the RPE. DAPI nuclei counterstain was imaged subsequently. Numbers of vesicles per area were quantified by counting vesicles in lipofuscin projections as described above after conversion to gray scale. For each eyecup wholemount, vesicles in three non-contiguous 50 μm × 50 μm areas were counted and averaged. 5-μm thick central cross sections of paraffin-embedded eyecups were stained with hematoxylin and eosin, antibodies to rhodopsin (B6–30, ) or R/G cone opsin (Millipore, Billerica, MA, USA), or stained using the DeadEnd™ Fluorometric TUNEL kit (Promega, Madison, WI, USA) according to the manufacturer's protocol. For fluorescence microscopy of sections, nuclei were counterstained with DAPI. Images of sections show single x-y confocal scans. To visualize F-actin, eyecups or rat primary RPE cells were pre-extracted with CSB for 1 min, fixed with PFA, and incubated with phalloidin-AlexaFluor488 before mounting with Vectashield. Stacks of x-y sections 0.2 μm apart were acquired to generate maximal projections representing the entire F-actin content of the RPE.
To label HNE-adducts in RPE cells in culture, cells were fixed with PFA, permeabilized with 0.2% Triton-X100, incubated with primary antibody to HNE (EMD) and secondary antibody conjugated to AlexaFluor594. To label tubulin and ZO-1, primary RPE cells were fixed with ice-cold methanol followed by incubation with mouse anti tubulin (Abcam, Cambridge, MA, USA) and rabbit anti ZO-1 followed by AlexaFluor488-anti mouse antibody and AlexaFluor568-anti rabbit antibody.
Extraction of lipids from individual mouse eyes after removal of lens and cornea followed the method of Bligh and Dyer . Dried lipid extracts were dissolved in CHCl3/MeOH (2:1) and spotted onto thin layer chromatography (TLC) plates. Relative total autofluorescent lipid content was quantified by fluorescence excitation at 488 nm and quantification of fluorescence emission from 520 to 610 nm with a Typhoon Trio Imager and ImageQuant™ TL 7.0 (both GE Healthcare). A2E content was quantified following chromatography in CHCl3/MeOH/TFA (93:6:1) of samples and varying amounts of purified A2E as standards and fluorescence scanning as described previously .
To generate total protein extracts, samples were solubilized by vortexing in HNTG lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 1.5 mM MgCl2, 10% glycerol, 1% Triton X-100 freshly supplemented with 1% protease inhibitor cocktail). To extract proteins not associated with the cytoskeleton (fractions ex), samples were incubated with CSB for 1 min. Subsequent incubation of the remaining cell material with HNTG yielded insoluble proteins anchored in the cell by the cytoskeleton (fractions ins). Immunoprecipitations were performed by adding HNE antibody to pre-cleared total protein lysates for 2 hours with agitation followed by agitation for another 2 hours in the presence of Protein G agarose beads. Following 4 washes with lysis buffer, samples were eluted by boiling in reducing SDS sample buffer. Lysates, sample fractions or immunoprecipitates were separated by SDS-PAGE, blotted onto nitrocellulose membrane, incubated with primary antibodies and secondary antibodies conjugated to HRP, and developed by enhanced chemiluminescence detection (Perkin Elmer, Waltham, MA, USA) using autoradiography films. Films were scanned and bands quantified using ImageJ. Primary antibodies used for detection were: actin (Cell Signaling, Cambridge, MA, USA), α/β-tubulin (Abcam), αv integrin (BD Biosciences), β5 integrin (Santa Cruz), CD81 (Biolegend), CD36, MerTK (both R&D Systems, Minneapolis, MN, USA), HNE (Millipore), and RPE65 (Genetex).
Students t test was used to compare control and test groups. P values smaller than 0.05 were considered statistically significant.
Excess lipofuscin accumulation in the RPE suggested elevated oxidative stress in 1-year-old β5−/− mice . To directly test whether aging β5−/− retina accumulated oxidative damage, we measured levels of HNE-protein adducts in eyes from β5−/− mice at 6 months of age, before lipofuscin buildup, and at 13 months of age, when lipofuscin is excessive. Fig. 1A shows that whole eyes (eyes following removal of lens but containing all other eye tissues) obtained from β5−/− mice at 6 months of age contained on average the same levels of HNE-adducts as whole eyes obtained from age-matched wild-type (wt) mice. There was little increase in wt eyes by 13-months of age. In contrast, HNE-adducts in β5−/− eyes increased 6.7-fold from 6 to 13 months of age and significantly above levels of age-matched wt eyes. This increase was specifically due to higher levels of HNE-adducts present in eyecup tissue containing the RPE and the underlying choroid after removal of the retina (Fig. 1B, RPE/Ch), as HNE-adducts in β5−/− neural retina alone hardly differed from wt neural retina even at 13 months of age (Fig. 1B, NR). HNE adduct levels in brain and liver tissues did not differ between 13-month-old β5−/− mice and age-matched wt mice further confirming that abnormal protein oxidation occurred specifically in RPE/choroid tissues of β5−/− mice of age (data not shown).
We next sought to alleviate the oxidative burden of β5−/− mice through increase of their dietary antioxidant intake by enriching the regular diet either with grapes or with standardized marigold extract rich in lutein/zeaxanthin (for details on diets please see Materials and Methods). Grapes are rich sources of numerous natural polyphenols and resveratrol and have already shown anti-oxidant benefits in physiological settings [21, 22]. We supplemented the standard diet of a cohort of β5−/− mice with freeze-dried grapes (grape diet). Based on intake per body weight, mice on grape diet consumed about 3.5 servings of grapes per day. Control groups of age-matched mice received chow matching the sugar content of grape diet (sugar diet). Lutein/zeaxanthin are macular pigments that may serve as endogenous antioxidants in the human retina and whose decline may weaken retinal anti-oxidant defense with age. We supplemented the standard diet of a cohort of β5−/− mice with lutein/zeaxanthin (lutein diet). Feeding groups received different diets for nine months, from 4 to 13 months of age. All feeding groups tolerated the enriched chows well, consumed the same amount of food and had similar body weights on the different diets (Fig. S1). At 13 months of age, HNE adduct levels were similar in RPE/choroid tissues of mice on sugar diet as compared to standard chow (Fig. 1C). In contrast, either grape or lutein diet dramatically decreased HNE adduct content in RPE/choroid by 75% and 68% compared to the appropriate controls, respectively. The effects of the diets were not specific to HNE-adduct formation as protein carbonylation levels decreased to the same extent (Fig. 1D).
We next used fluorescence microscopy examination of RPE to test if increasing dietary antioxidants affected the buildup of autofluorescent lipofuscin-like granules we had found previously in β5−/− RPE of age. We set our acquisition parameters to detect long-wavelength emitting fluorophores such as the known lipofuscin component A2E because we previously found such fluorophores in β5−/− RPE tissue sections . It is important to note that human and likely mouse lipofuscin contains additional fluorophores that this imaging methodology does not detect. Flat mount preparations of the RPE of mice sacrificed at 13 months of age revealed more autofluorescent granules in β5−/− RPE as compared to wt RPE if mice were fed sugar diet (Fig. 2, compare B to A). We observed fewer autofluorescent granules in β5−/− RPE of mice on grape diet or on lutein diet than in β5−/− RPE of mice on sugar diet (Fig. 2, compare C and D to B). This difference was not due to a restoration of the daily rhythm of POS phagocytosis in β5−/− mice fed with anti-oxidant diets, as phagosome numbers did not increase at the time of the phagocytic peak 1 hour after light onset (Fig. S2). To support our microscopic observation, we employed three different methods to quantify lipofuscin load in the different aged mice. First, counting the number of autofluorescent granules per field of (50 μm)2 RPE demonstrated that aged β5−/− RPE harbors on average 5 times more granules than wt RPE (Fig. 2E). Grape and lutein diets reduced granule load significantly, by 65% and 43%, respectively, but kept it significantly above granule load of wt RPE (Fig. 2E). Second, quantifying total relative content of autofluorescent lipids extractable from eyes of the four groups of mice showed similar relative differences between mouse groups (Fig. 2F). Third, quantifying the best studied lipofuscin component, the pyridinium bis-retinoid known as A2E , showed significantly elevated A2E in β5−/− eyes on control diet compared to wt eyes (Fig. 2G). Grape diet reduced A2E to levels found in wt mice (Fig. 2G). Lutein diet reduced A2E by 32% such that eyes of β5−/− mice on lutein diet contained significantly less A2E than eyes from β5−/− mice on sugar diet but significantly more than wt mice on sugar diet (Fig. 2G). Altogether, these analyses support the conclusion that grape diet and to a lesser extent lutein diet reduce age-related accumulation of autofluorescent lipofuscin-like granules by β5−/− RPE.
Next, we monitored retinal light sensitivity of mice on different diets by recording electroretinograms (ERG's) from dark-adapted mice. The much larger light response of 13-month-old β5−/− mice fed grape diet or lutein diet from 4 to 13 months of age as compared to sugar diet or standard chow was immediately obvious (Fig. 3A). Quantifying ERG a-waves that indicate light responses specifically by photoreceptors from cohorts of mice on the different diets over the entire course of the feeding experiments demonstrated that either grape or lutein diets was sufficient to prevent the dramatic decline in photoreceptor function observed on both control and sugar diet with age (Figs. 3B and C). At 13 months of age, a-wave amplitudes recorded from β5−/− mice on lutein diet were on average 31% lower than wt a-waves, compared to a 74% reduction in β5−/− mice without protective diet (Fig. 3C). B-wave amplitudes, which reflect activities of retinal interneurons, were also higher if β5−/− mice had been fed grape or lutein diet, by 16% and 26% on average, respectively, but differences were less pronounced than for a-waves. To more precisely unravel the effects of the different diets on responses by different retinal cell types we conducted a second feeding experiment. We repeated the original regimen providing either sugar diet or grape diet to β5−/− mice for 9 months starting at 3 months of age. Additionally, we raised age-matched β5−/− mice that received grape diet for periods of 3 months only, starting at 3, 6, or 9 months of age. Given that age-related macular degeneration affects mainly cone photoreceptor function, we specifically tested cone photoreceptor function in animals of all five feeding groups at 12 months of age by performing photopic ERG's. Comparison of responses of mice fed with sugar diet and with grape diet for 9 months revealed dramatically better cone function in mice fed the grape diet (Fig. 4A). Strikingly, cone function was improved in comparison to mice fed a life-long sugar diet even if mice received grape diet only from 3 to 6 months or from 6 to 9 months of age (Fig. 4B). In contrast, grape diet consumption only from 9 to 12 months of age had no effect (Fig. 4B). Grape diet had similar effects on photopic ERG b-waves as for a-waves except that for all feeding groups differences to b-waves of mice on control diet were less pronounced (Fig. 4C). As a result, the effect on the photopic b-wave of feeding grape diet from 3 to 6 months of age was too small to be statistically significant. To determine whether the improvements in our original feeding experiment (Fig. 3) were solely due to effects of diet on cones, we repeated ERG recordings on the same cohorts of mice one week after cone recordings. To detect rod photoreceptor function with minimal contributions of cones, we recorded responses of dark-adapted mice to low intensity light flashes that are below the detection sensitivity of cones. This experiment revealed that rod function was improved -albeit to a lesser extent than cone function- in mice that received grape diet for 9 months, or from 3 to 6 months or 6 to 9 months of age (Fig. 4D). Like cones, rods did not benefit if mice consumed grape diet only late in life, from 9 to 12 months of age (Fig. 4D). All three diet groups with improved rod a-waves also showed significantly improved rod b-waves (Fig. 4E). However, as for cones, changes in b-wave amplitudes were considerably smaller than changes in a-wave amplitudes. While we cannot exclude a direct effect of grape diet on retinal interneuron function, we suspect that changes in b-waves we detected are mainly a consequence of altered photoreceptor function.
We wondered if increased functionality of photoreceptors in β5−/− mice fed with antioxidant-enriched diet was a result of prevention of photoreceptor outer segment loss or cell death. To address this question, we carefully compared retinal tissue morphology between wt and β5−/− mice on sugar diet and β5−/− mice on grape diet. We did not observe obvious morphological changes between samples taken from the three groups of mice at either 6 months or 12 months of age (Fig. 5A and data not shown). Rhodopsin and cone R/G opsin immunofluorescence signals did not differ among groups either (Figs. 5B and C). Morphometry detected no significant difference in photoreceptor outer or inner segment length among the groups (data not shown). Finally, TUNEL labeling did not detect apoptotic nuclei in β5−/− mouse retina at either age (Fig. 5D). Taken together, reducing oxidative burden on RPE/choroid by increasing dietary antioxidant intake in early and/or midlife adulthood was sufficient to significantly delay age-related blindness by improving cone and rod photoreceptor functionality independently of detectable changes in outer segment structure or photoreceptor survival.
We next sought to specify the oxidative damage that occurred specifically in the RPE in aging β5−/− eyes and that was prevented by antioxidant supplementation. We chose to continue our experiments only on mice fed grape diet, as it was equally effective in preserving vision in our model as lutein diet and represented a healthy diet of several servings of fruit a day.
Quantitative immunoblotting using HNE antibodies confirmed that increased levels of HNE-protein adducts in eyes of aging β5−/− mice fed sugar diet were significantly reduced in mice fed grape diet (Fig. 6A). HNE-modified proteins were of various molecular sizes and were strongly enriched in the RPE/choroid eyecups as compared to the neural retina (Fig. 6B). Because preventing oxidative damage to β5−/− RPE proteins resulted in improved functionality of photoreceptor cells of the neural retina, which we found harboring little oxidative damage themselves, we reasoned that proteins relevant for photoreceptor support might be particularly damaged with age in β5−/− RPE. HNE adduct formation was not associated with altered total protein levels of any of the proteins we chose to test because of their importance for RPE structure (actin, tubulin), phagocytic function (αv and −5 integrins, MerTK, CD81 and CD36), or the visual cycle (RPE65) (Fig. 6C). In order to determine if these candidate proteins were modified by HNE, we performed HNE immunoprecipitations from whole eye lysates of 1-year-old mice followed by candidate protein immunoblotting. We did not detect any of our candidate proteins in HNE immunoprecipitates of wt eyes (Fig. 6D). Actin was the only one of our candidate proteins detectable in HNE immunoprecipitates of β5−/− eyes and its abundance in the HNE eluate was dramatically decreased, by ~9-fold, in samples obtained from mice on the grape diet (Fig. 6D). Notably, we did not detect HNE-modification of α-tubulin, which, like actin, is an abundant cytoskeletal protein of the RPE, and which is shown to be particularly susceptible to HNE adduct formation in cultured fibroblasts . To test if the actin cytoskeleton was also functionally altered in aging β5−/− RPE in situ, we next examined whole mount preparations of mouse RPE labeled with the F-actin binding reagent phalloidin by fluorescence microscopy. In comparison with age-matched wt RPE, 1-year-old β5−/− RPE showed overall reduced F-actin levels and obvious variations in F-actin labeling intensities among RPE cells in the same tissue (Fig. 7, compare B with A). In contrast, β5−/− RPE from mice fed the grape diet showed increased and largely even labeling of F-actin although less so than wt RPE (Fig. 7C). We did not observe obvious differences in F-actin labeling between RPE in situ from wt and β5−/− mice at 6 months of age (data not shown). To quantify differences in F-actin content, we subjected freshly isolated mouse RPE/choroid and neural retina tissue samples to differential detergent extraction. Extracts obtained from a 1-minute incubation of fresh tissues with cytoskeleton stabilizing buffer (CSB) that leaves cytoskeletal proteins intact contained 7-fold more actin from β5−/− eyes compared to wt eyes when mice were fed the control diet (Fig. 7D, compare actin bands in ex fractions of RPE/choroid, quantification in 7E). This increase in actin solubility was almost completely prevented by feeding β5−/− mice the grape diet. This destabilization was specific to actin in the RPE as tubulin solubility did not change with genotype and there was no change in actin or tubulin solubility in neural retina (Fig. 7, D and E).
Finally, we set out to determine if HNE modification was sufficient to destabilize actin in polarized RPE cells. To this end, we incubated primary RPE cells with synthetic HNE for 1 hour before analyzing their cytoskeletal systems. HNE immunofluorescence microscopy demonstrated HNE adduct formation (Fig. 8, A and B). Immunoblotting revealed HNE adduct formation on RPE proteins including protein/s of the molecular size of actin that were extractable in CSB (Fig. 8I). Phalloidin labeling after pre-extraction and fixation of cells showed dramatic decrease in F-actin content upon HNE treatment (Fig. 8, compare D with C). Quantification of HNE's effect on actin stability using differential extraction revealed that this short-term treatment with HNE caused a 5.4-fold increase in soluble actin (Fig. 8, I and J). Importantly, our treatment with 20 μM HNE for 1 hour had no effect on appearance or stability of the microtubule cytoskeleton (Fig. 8, E, F and I) or on the tight junction localization of ZO-1 (Fig. 8, G and H). We found that disruption of microtubules and cell junctions required 1-hour incubation with at least 5-fold more concentrated HNE instead (Fig. S3). These results indicate that availability of the oxidative stress product HNE is sufficient to directly destabilize the actin cytoskeleton in RPE cells. Moreover, the actin cytoskeleton is more sensitive to HNE than the microtubule cytoskeleton in RPE cells. We wondered if our HNE incubation yielded a similar overall HNE-adduct load as the load building up in β5−/− RPE in vivo with age. We tested this directly by measuring HNE-adducts relative to total RPE protein in RPE purified from 12-month-old β5−/− mice and in primary RPE in culture harvested immediately after treatment for 1 hour with HNE at concentrations between 1 and 20 μM. HNE-ELISA showed that incubation with 20 and 10 μM HNE generated 50% and 26%, respectively, more HNE-adducts than present in aged β5−/− mouse RPE in situ (Fig. S4A). Treatment with 5 μM HNE yielded HNE-adduct load that was on average 72% of load of RPE in situ, which was not a statistically significant difference. Treatment with 1 μM HNE generated negligible HNE-adduct load. Notably, 10 or 5 μM HNE (but not 2 or 1 μM HNE) destabilized F-actin significantly and to the same extent as 20 μM HNE (Fig. S4B). Thus, generation of HNE-adducts at the level present in the aging β5−/− RPE is sufficient to destabilize F-actin.
Finally, we tested if antioxidants may benefit F-actin stability directly by preventing HNE-actin formation or by promoting clearance of destabilized actin. We chose to determine efficacy of three different antioxidant compounds with different biochemical properties, purified resveratrol and lutein, in analogy to our enriched diets, or trolox, a water-soluble Vitamin E analogue we had previously found both non-toxic and effective in reducing oxidative stress exerted by A2E on polarized RPE cultures . Pre-incubation of primary RPE in culture with any of the antioxidants overnight or for 4 days did not prevent HNE-adduct formation (data not shown) or the solubilization of actin by either 5 or 20 μM HNE (Fig. 9A). Either HNE concentration solubilized 28 to 41% of actin regardless of antioxidant regimen, which was not significantly different (n = 3 independent experiments, with duplicate samples in each). Following a 1-hour HNE pulse, actin remained soluble to a similar extent for all time points tested, up to 24 hours (Fig. 9B). Significant actin destabilization persisted regardless whether RPE cells received 5 or 20 μM HNE or whether any of the three antioxidants previously tested were added during a 24-hour recovery period (Fig. 9C, quantification of three independent experiments not shown). All effects shown were specific to the actin cytoskeleton as tubulin remained insoluble in all conditions (Fig. 9, A to C, panels tubulin). Taken together, providing antioxidants to RPE cells as tested did not prevent acute HNE destabilization of actin or detectably promote actin re-stabilization.
Aging of the retina and RPE is associated with increased levels of oxidative damage [24, 25]. It is generally assumed that oxidative stress also plays a role in the development or progression of AMD [26, 27]. This is supported by studies demonstrating that increasing oxidative insult in experimental animal models results in visual impairment and in pathology as seen in atrophic AMD [28–31]. However, it is poorly understood precisely how physiological levels of chronic sublethal oxidative stress affect retina and/or RPE functionality. Here, we demonstrate oxidative protein modification specific to the RPE that coincides with age-related photoreceptor dysfunction in β5−/− mice. In these mice, a primary defect in phagocytic rhythm directly due to lack of the outer segment recognition receptor αvβ5 integrin causes gradual accumulation of autofluorescent lipofuscin-like granules . While the exact mechanism triggering lipofuscin accumulation in β5−/− RPE is still under investigation, it is likely that it involves impaired phagolysosomal processing. We detect both lipofuscin buildup and increased levels of oxidized lipids and protein starting at about 6 month of age and we have not yet found experimental conditions in which one occurs without the other. Thus, we do not know at this time if these changes are independent or if lipofuscin causes oxidative modifications or vice versa. Gradually, β5−/− photoreceptors also become dysfunctional. We did not detect elevated levels of oxidative modifications in neural retina of β5−/− mice even in mice at a high age when ERG's showed that their photoreceptor cells largely failed to respond to light. This suggests that increased levels of oxidative damage of the RPE may result in photoreceptor dysfunction even if it does not directly harm photoreceptor themselves.
Because increasing oxidative damage in aging β5−/− RPE directly correlates with photoreceptor dysfunction, we decided to check for alterations in RPE proteins that are particularly important for photoreceptor support. We did not find abnormalities in steady state protein levels of any of the candidate proteins we tested. We conclude that oxidative stress in the eyes of aging β5−/− mice does not reach a level that would cause non-specific cell damage. Moreover, none of the RPE's known phagocytic receptors nor the critical visual cycle isomerase RPE65 were modified by HNE to detectable extent. However, we detected an age-dependent increase in HNE-adducts of the microfilament constituent actin in the RPE of β5−/− mice but not in the RPE of age-matched wt mice nor in β5−/− neural retina. Furthermore, stability of the actin cytoskeleton was dramatically altered solely in the RPE of β5−/− mice of age. Thus, HNE-actin content directly correlates with microfilament destabilization in the RPE in situ. Interestingly, a recent proteomic study reported higher levels of HNE-modified actin (among numerous other proteins) in the brain of Alzheimer's Disease patients relative to brain of control subjects, although it was not determined if this was associated with F-actin destabilization . Proteomic analysis has detected increased HNE-load of numerous proteins in human retina with age that include HNE-actin . Tanito and colleagues identified a number of specific HNE-adducts in rat neural retina subjected to damaging light . Most of them were functionally related to energy metabolism and neither actin nor tubulin was among them but their study did not investigate modifications in the RPE. Like ours, these studies show that the overall increase in HNE-adduct load is due to considerable modification of select proteins rather than due to low level modifications of proteins in general. Interestingly, we found obvious variability in autofluorescent granule load and in the extent of actin destabilization even among RPE in individual eyes and especially in eyes of mice fed with antioxidant-enriched diet. This observation leads us to speculate that RPE damage and prevention thereof occurs at the single cell level and is influenced by differences in individual cell properties that may include differences in actin dependent structures and in abundance of melanin- and/or lipofuscin-containing granules (for a review of RPE mosaicism, please see ).
We complemented our studies of the RPE in vivo with cell culture experiments proving that short-term incubation with HNE is sufficient to solubilize actin in RPE cells. F-actin associated with the apical surface of the RPE and its microvilli was more obviously disrupted than circumferential microfilaments both in β5−/− RPE in situ and in HNE-treated primary RPE cells. As the apical actin cytoskeleton undergoes frequent dramatic re-organization e.g. during diurnal outer segment phagocytosis, it is possible that dynamic apical actin is more vulnerable to HNE-modification than lateral actin. RPE microfilaments were vulnerable even at HNE concentrations low enough not to disassemble tight junctions or microtubules. The specific sensitivity of F-actin to HNE destabilization may be characteristic to RPE. Other studies testing HNE incubation of cells in culture (albeit at higher concentrations than ours) have shown that fibroblasts and neuronal cells mainly disassemble microtubules while endothelial cells lose focal adhesions and lateral junctions [23, 35]. Changes in monolayer permeability suggesting altered cell-cell junctions have also been reported to occur in the ARPE-19 cell line upon HNE treatment . We did not observe obvious changes in these vital cell structures in our assays investigating unpassaged primary RPE but we did not directly test epithelial barrier properties. As our HNE immunoblots show a number of HNE-proteins of molecular sizes other than actin in β5−/− RPE with age, it is clear that the selection of candidate proteins we examined was not comprehensive. However, the identification of RPE actin as specifically vulnerable to HNE modification provides important new insight into the functional alterations of RPE cells affected by physiological levels of oxidative stress. Like aging μ5−/− RPE, human RPE/retina accumulate lipofuscin and HNE-adducts over a lifetime . Our results predict that human RPE with age may acquire actin cytoskeletal abnormalities that could impair their support for photoreceptor cells. The same likely holds true for other experimental animal models of retinal dysfunction that involve oxidative damage [28–31, 33].
Based on the presence of oxidative modifications in human RPE, it has long been hypothesized that antioxidant supplements and/or altering dietary habits to increase natural antioxidant consumption may be beneficial to prevent or delay progression of AMD (summarized in ). Here, we tested if increasing dietary antioxidant content by enriching an otherwise unchanged diet with grapes or with marigold extract providing lutein/zeaxanthin reduced oxidative damage and preserved vision in our animal model. The results were dramatic. Increasing dietary intake of either grapes or lutein/zeaxanthin was sufficient to prevent RPE oxidation, cytoskeletal damage and vision loss. This suggests that photoreceptor loss of function in this model occurs as a consequence of oxidative damage to the RPE. Beneficial effect of grape diet for both rod and cone photoreceptor function even if only consumed during young adulthood (from 3 to 6 months of age) or during midlife (from 6 to 9 months of age) suggests that age-related vision loss is a result of cumulative, lifelong oxidative insult. ERG recordings from 1-year-old wt mice did not detect differences in light responses between wt mice that had consumed control or grape diet from 4 months of age (Fig. S5). This lack of benefit for normal retinal function suggests that dietary antioxidants prevent the vision loss specifically caused by pathological oxidative stress rather than enhancing visual function independently. Unlike the RPE in our wt mice kept under standard vivarium conditions, human RPE accumulates oxidative damage with age and individuals with high oxidative burden, such as smokers, are at increased risk for AMD. We speculate therefore that long-term increase in dietary antioxidant intake will likely reduce RPE oxidative damage in the human eye and may delay onset of age-related visual impairment. We do not know at this time whether the antioxidants exert their effects locally by acting directly on the RPE or whether they induce systemic changes that secondarily reduce oxidative burden on the RPE. Little is known about availability of phenolic compounds as included in grapes to specific tissues following dietary intake. Intake of lutein-rich diet has been shown to increase lutein levels in the macula of human patients . This suggests that dietary lutein reaches the retina/RPE. However, as the mouse eye differs from the human eye in that its retina does not possess a macular region, mechanisms accumulating dietary lutein available to the human macula may not exist in the murine eye. Considering that antioxidants did not alter degree or duration of actin destabilization by added HNE at least within the time period we tested in our cell culture experiments, we propose that dietary antioxidants reduce availability of HNE to the RPE in our mice rather than its efficacy.
In our experimental design we chose to assess efficacy of dietary antioxidants from natural sources. Mice consumed antioxidants as part of a healthy diet rather than receiving periodic supplements. Our lutein diet provided about 1.65 mg lutein per mouse per day, which had shown health benefits to mice earlier . Taking into account their body weight and food intake, mice on our grape diet consumed about 3.5 servings of grapes per day. This is not excessive given that 5 servings of fruit and vegetables are recommended to be a part of a healthy diet. We did not determine the minimum amount of grapes that would still be protective in our animal model. Furthermore, we did not test if a diet combining lutein/zeaxanthin and grape additives may be more effective in preventing oxidative damage and vision loss, which is an intriguing possibility especially since the antioxidants prevalent in the two dietary additives are chemically different.
Mice differ from humans not just in body size but in many aspects of physiology as well as in life-style and compliance. Yet, we conclude from our study that a lifelong diet enriched in natural antioxidants is directly beneficial for RPE and retinal health and function.
Natural-antioxidant-enriched diet prevents age-related blindness due to RPE damage. Antioxidant-enriched diet prevents HNE actin destabilization in the RPE with age. Pathophysiological levels of oxidative stress generate HNE-modified actin in the RPE. HNE-modification directly and specifically destabilizes F-actin in the RPE in vivo.
Different diets do not affect mouse body weights.
Wt mice (gray bars) or β5−/− mice (black bars) were weighed at 13 months of age after fed different diets as indicated from 4 months of age. Bars show mean body weight in g ± SD, n = 7–8. Differences among groups were not significant (P > 0.07).
Grape diet does not restore the RPE phagocytic rhythm in β5−/− mice.
β5−/− and wt mice were fed diets as indicated from 4 to 13 months of age before sacrifice at 1 hour or at 8 hours after light onset and paraffin embedding of fixed eyes for microtome sectioning. Opsin-labeled POS phagosomes in the RPE were quantified in retinal cross sections exactly as described previously . Phagosome counts were similar at both time points for β5−/− RPE regardless of diet (compare striped bars and black bars), while wt RPE showed the characteristic peak of POS phagosomes 1 hour after light onset (white bars). For each time point, 4 eyes from 4 different mice of each genotype were analyzed. For each eye, phagosome numbers were established by counting phagosomes in 5 sections cut ˜100 μm apart. Bars show mean number of phagosomes of ± SD, n = 4 eyes.
Short-term incubation with HNE at concentrations above 50 μM causes microtubule disassembly in primary rat RPE.
Microscopy images show representative fields of three independent experiments comparing RPE cells incubated for 1 hour with HNE at concentrations as indicated. Images shown are from the same experiment. Scale bar, 10 μm. The scale is identical for all fields. Upper row shows immunolabeling for tubulin, lower row shows tubulin (green) merged with nuclei stain (red) in the same fields.
Generating HNE-adducts at levels similar to those accumulating in β5−/− RPE in vivo with age is sufficient to destabilize actin.
(A) ELISA quantification of HNE-adducts in extracts of RPE freshly isolated from β5−/− mice on control diet (RPE in situ, black bar) and of primary RPE cells in culture treated with HEN as indicated for 1 hour before harvest and immediate analysis. Asterisks indicate significant difference to RPE in situ. (B) Sequential detection of tubulin (top panel) and actin (bottom panel) of rat primary RPE on the same blot membrane showing fractionation among soluble proteins (ex) obtained by 1 min incubation with CSB extraction buffer and among insoluble proteins (ins) obtained by extraction of remaining proteins with HNTG. Panels show a representative result of three independent experiments.
Consuming grape diet does not affect wt visual function.
Bars show a-wave and b-wave amplitudes obtained by recording scotopic ERG's from 12- month-old wt mice fed for 8 months with sugar diet (black bars) or grape diet (gray bars). Bars show mean ± SD, n = 4.
We thank Ms. Kathryn Silva for excellent technical support. We thank Dr. Zoraida Freitas from Kemin Health for generously supplying the FloraGlo® lutein/zeaxanthin mix. This work was supported by National Institutes of Health Grant EY013295 from the National Eye Institute (to S.C.F.) and a Research Award by The California Table Grape Commission (to S.C.F.). E.F.N. was supported by Institut National de la Santé et de la Recherche Médicale (INSERM), Université Pierre et Marie Curie-Paris6, Centre National de la Recherche Scientifique (CNRS) and Départment de Paris (to Institut de la Vision), and Young Investigator Grants from Fondation Voir et Entendre and Fondation Bettencourt Schueller.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Supplementary Material Fig. S1 shows that mice gained similar body weight regardless of diet. Fig. S2 shows that grape diet does not restore the phagocytic rhythm that is lacking in β5−/− mice. Fig. S3 shows that HNE disrupts the microtubule cytoskeleton of primary RPE cells concentrations only if applied at a concentration above 50 μM. Fig. S4 compares HNE-adduct load of aged β5−/− RPE with that achieved in primary RPE cells in culture by 1-hour incubation of HNE and determines actin solubility in cells harboring HNE-adducts at levels similar to those present in vivo. Fig. S5 shows that grape diet does not alter vision of wt mice.