Vitamin A cycle byproducts explain retinal damage and molecular changes thought to initiate retinal degeneration

Atrophic lesions involving the photoreceptors, retinal pigment epithelium (RPE), choriocapillaris and loss of vision characterize end-stage Stargardt disease and age-related macular degeneration (AMD). According to histopathological analysis, atrophic lesions began with atrophic changes to the RPE, which spread to the choriocapillaris and Bruch membrane, and overlying neuroretina (Bertolotto et al., 2014; Bird, 2020; Birnbach et al., 1994; Eagle et al., 1980; Ferris et al., 1984; Hogan, 1972; Sarks et al., 1988; Sarks, 1976). In Stargardt disease and AMD, pre-symptomatic RPE changes can also be observed via optical coherence tomography (Curcio et al., 2017), along with atrophic changes to the ellipsoid zone (EZ, the photoreceptor inner segment/outer segment junction) (Cai et al., 2018; Toprak et al., 2017), and expansion (Lamin et al., 2019) and thinning (Whitmore et al., 2020) of the outer plexiform layer (OPL).

The above morphological changes are concurrent with various molecular derangements. The retina also exhibits signatures suggestive of altered states of immunity (Bora et al., 2014), autophagy (Kaarniranta et al., 2013), mitochondrial function (Kenney et al., 2013), and oxidative metabolism (Ugurlu et al., 2013), to name a few. A challenge in elucidating the pathophysiology of retinal disease is deconvoluting between events that drive the degenerative processes and events that are secondary symptoms.

Retinal poisons, such as ornithine and sodium iodate, can also cause several of the above morphological and molecular changes and result in atrophic lesions similar to those seen in late-stage Stargardt disease and AMD (Machalinska et al., 2010; Maeda et al., 1998). This work proposes that the atrophic lesions observed in retinal diseases such as Stargardt disease and AMD can be explained by a poisoning from the chronic formation of vitamin A byproducts during the vitamin A cycle.

During the usual process to enable vision, the vitamin A cycle, an unknown portion of vitamin A, is its retinaldehyde form, reacts with phosphatidylethanolamine or lysine (Eldred and Lasky, 1993; Fishkin et al., 2003, 2005; Katz et al., 1996; Sakai et al., 1996) and another molecule of retinaldehyde to generate dimers of retinaldehyde. These dimers of retinaldehyde can be oxidized, transformed into oxidative catabolites (Ben-Shabat et al., 2002a; Ueda et al., 2016; Wang et al., 2006; Washington et al., 2006, 2005; Wu et al., 2010; Yoon et al., 2012), and/or react with additional retinaldehyde molecules to generate higher-order oligomers (Murdaugh et al., 2011, 2010b). These dimers and oligomers of retinaldehyde can be found in the RPE (Eldred and Katz, 1988; Eldred and Lasky, 1993; Mihai and Washington, 2014; Sakai et al., 1996) and Bruch's membrane (Murdaugh et al., 2010a) of the eye. An obstacle in understanding diseases of the retina has been elucidating the significance of the formation of these vitamin A cycle byproducts. As a model byproduct, A2E is often used because it is characterized, can be synthesized in a lab, and extracted and quantified from eyes (Penn et al., 2014). However, there are presumably hundreds of waste byproducts and oligomer aggregates derived from vitamin A. Over the past three decades, several mechanisms have been proposed by which byproducts of the vitamin A cycle, particularly A2E, might contribute to retinal pathology (Anderson et al., 2013; Avalle et al., 2004; Ben-Shabat et al., 2002a,b; Bermann et al., 2001; De and Sakmar, 2002; Dontsov et al., 2009; Finnemann et al., 2002; Holz et al., 1999; Iriyama et al., 2009, 2008; Kanofsky et al., 2003; Lakkaraju et al., 2007; Lukiw et al., 2006; Ma et al., 2013; Moiseyev et al., 2010; Murdaugh et al., 2009, 2010a; Nowak, 2013; Radu et al., 2004; Roberts et al., 2002; Sokolov et al., 2007; Sparrow, 2010; Sparrow and Cai, 2001; Sparrow et al., 2000, 2003a, 1999, 2006, 2002, 2003b; Thao et al., 2013; van der Burght et al., 2013; Vives-Bauza et al., 2008; Wang et al., 2006; Washington et al., 2005; Wielgus et al., 2010; Wu et al., 2010; Yoon et al., 2011, 2012; Zhou et al., 2006). Nevertheless, the contribution of the vitamin A byproducts to retinal health is debated, and their formation has even been argued to be beneficial (Roberts et al., 2002).

We have observed that exposing cultured RPE cells to vitamin A byproducts triggered similar pathology to that observed in the aging in vivo RPE (Mihai and Washington, 2014). Further, we observed that exogenous delivery of the A2E byproduct to the in vivo rabbit eye triggered retinal degeneration in a similar sequence observed in humans (Penn et al., 2014). Iriyama et al. injected A2E into the subretinal space of the in vivo mouse eye and concluded that the vitamin A dimer was pro-angiogenic (Iriyama et al., 2009). Accordingly, we sought to determine the extent to which the A2E vitamin A byproduct, delivered exogenously to the living eye, triggered the functional, morphological, and molecular derangements observed in human retinal disease.

By delivering A2E into the rat eye, via intravitreal injection, we show that the vitamin A byproduct is likely detrimental. Data reveals that the formation the vitamin A cycle byproducts during the vitamin A cycle is sufficient alone to explain the functional, morphological, and molecular phenotypes observed during the most prevalent forms of retinal degeneration. The data suggests that retarding the formation of the vitamin A byproducts could result in a disease-modifying intervention.

Because intravitreally delivered materials tend to accumulate in the RPE (Katz et al., 1999; Tabatabay et al., 1987; Yasukawa et al., 2007), we injected A2E into rats’ vitreous chamber. Before administering A2E, the rat RPE and neuroretina contained about 50 and 20 pmols of A2E per eye, respectively, in accordance with literature reports (Reinboth et al., 1997). Six days after the intravitreal injection of A2E, the concentrations of A2E increased by 16 times in the RPE and 24 times in the neuroretina (Fig. 1A). The concentrations of A2E in both the RPE and neuroretina decreased and leveled off when measured 20 days after injection (Fig. 1A). The relative concentrations of the major ocular vitamin A congeners remained unchanged in the A2E-treated eyes, relative to control eyes injected with the vehicle only (Fig. 1B), when measured 2 weeks after injection, indicating normal retinal homeostasis in regards to vitamin A. From about 2 weeks after injection onward, the ratio of A2E relative to vitamin A (the sum of retinol, retinaldehyde, retinyl esters) in the A2E-treated eyes was similar to that found in humans after the seventh decade of life (Zhang et al., 2021).

Two months after A2E treatment, we evaluated retinal function via electroretinography. The ERG measures the electrical response of the eye elicited by flashes of light. ERG parameters, the amplitudes of the a- and b-waves, are used as indicators of retinal function. The average maximum a-wave was reduced by 44% in the A2E-treated animals compared to controls (−217±17 μV in control versus −121±8 μV in A2E-treated, P=<0.001) and the intensity of light needed to elicit half of the maximum a-wave was 33% greater in the A2E-treated animals (0.079 cd s/m² in control versus 0.053 cd s/m² in A2E-treated, P=<0.05) (Fig. 2A). Likewise, maximum b-wave of the A2E-treated animals was depressed by approximately 50% of that of control animals (653±38 μV in control versus 333±23 μV in A2E-treated, P=<0.001) and the intensity of light needed to elicit half of the maximum b-wave was 18% greater in the A2E-treated animals (0.0033 cd s/m² in control versus 0.0027 cd s/m² in A2E-treated, P=<0.05) (Fig. 2B). In contrast, depressed ERG a- and b-waves were not observed 7 days after injection, indicating that A2E causes a slow decline in retinal function instead of immediate retinal damage as observed for RPE poisons, such as sodium iodate and ornithine.

Morphological changes induced by A2E, 2 months after an injection, were most prevalent in the RPE (Fig. 3). The average thickness of the RPE in A2E-treated animals, measured 1 mm from both sides of the optic nerve head, is shown in Fig. 3G and was 60% thinner compared to control animals (5±1 μm in controls versus 2±1 μm in A2E-treated, P=<0.01). Cellular debris and swelling were detected in and around the RPE in the A2E-treated animals (Fig. 2A), which were absent in control retinas. Signs of pathology were also observed in the photoreceptor inner and outer segments (PIOS) and outer plexiform layer (OPL). The average thickness of the PIOS in the A2E-treated animals was 42% thinner compared to control animals (26±6 μm in controls versus 15±5 μm in A2E-treated, P=<0.001), and the average thickness of the OPL, was 31% thinner compared to control animals (13±4 μm in controls versus 9±3 μm in A2E-treated, P=<0.05). Outer nuclear layer (ONL) nuclei were found in the OPL and PIOS in the A2E treated animals, relative to controls; this difference was not quantified (Gartner and Henkind, 1981). The thicknesses of the remaining retinal layers were statistically the same in the control and A2E-treated animals. As with ERG amplitudes, morphological changes were not observed 5 days after injection.

Because vitamin A byproducts are physically associated with melanin granules, we also compared histological changes in pigmented rats treated with intravitreal A2E with age-matched, vehicle-treated eyes (Fig. 4A). Similar morphological changes, such as RPE swelling (Fig. 4B), subretinal debris (Fig. 4C), and RPE degeneration (Fig. 4E), were seen in the pigmented rats. In addition areas of hypo-pigmentation were observed in the pigmented rats (Fig. 4F). The neuroretina was not significantly altered. All five A2E-treated eyes displayed some RPE pathology as described above, while none of the four control eyes displayed RPE pathology. A2E was significantly correlated RPE damage by histology [OR (95%CI)=0.01 (0.0002 to 0.6); P-value=0.0079].

To determine if the above electrophysiological and morphological changes correlated with molecular dysregulation, we used quantitative reverse transcription-polymerase chain reaction (qRT-PCR) to survey the expression of 167 genes involved in inflammation and autophagy. Out of the 167 genes surveyed, 27 transcripts were downregulated greater than twofold, and 9 transcripts were upregulated greater than twofold in the A2E-treated retinas relative to the controls, suggesting molecular dysregulation (Fig. 5). Of the 27 down-regulated transcripts, five were downregulated with statistical significance: interleukin 1 alpha (IL-1a) and nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha (Nfkbi-a), which are involved in innate immunity; the chemokine (C-C motif) receptor 5 (Ccr5), involved in adaptive immunity; interferon-gamma (Ifng, which is involved in both innate and adaptive immunity; and Transforming growth factor, beta 1 (Tgfb-1). One gene was significantly upregulated: chemokine (C-C motif) ligand 12 (CCL12).

Next, we quantified levels of nine cytokines via enzyme-linked immunosorbent assay (ELISA) (Fig. 6). A2E-treated animals had 56% less ciliary neurotrophic factor (CNTF), (2649±904 pg/ml in controls versus 1164±446 pg/ml in A2E-treated, P=<0.001) and 45% less intercellular adhesion molecule 1 (ICAM-1) (6644±1465 pg/ml in controls versus 3668±1237 pg/ml in A2E-treated, P=<0.001) compared to control animals. The concentrations of the other seven cytokines surveyed were statistically the same in both the control and A2E-treated animals (Fig. 5B).

We further surveyed total protein synthesis in the A2E-treated and control eyes using two-dimensional (2D) difference gel electrophoresis. We were able to detect over 2000 unique proteins in the homogenized eyecups. Of these proteins, 45 were downregulated, and 29 were upregulated in the A2E-treated eyes compared to the control eyes, in accord with an overall downregulation in transcription and the downregulation of CNTF and ICAM-1. We used mass spectrometry to identify 20 of the 74 differentially regulated proteins with greater than 99% confidence (Fig. 7C). When these 20 proteins were classified according to related gene ontology (GO), 9 of the 20 proteins were involved in metabolic processes (GO:0008152), 4 of the 20 proteins were involved in cellular processes (GO:0009987), biological regulation (GO:0065007), or multicellular organismal process (GO:0032501). Pathway analysis revealed that 6 canonical pathways might have been influenced by differential expression of 4 of the 20 identified proteins. These pathways with their pathway accession numbers and corresponding proteins were as follows: the pyrimidine ribonucleotides biosynthesis pathway (P02740) protein NDKA; the pyrimidine deoxyribonucleotide biosynthesis pathway (P02739) protein NDKA; the purine biosynthesis pathway (P02738) protein NDKA; the pentose phosphate pathway (P02762) protein TKT; the 5-hydroxytryptamine degradation pathway (P11883) protein AL3A1; and the pyruvate metabolism pathway (P02772) protein KPYM.

No model can reproduce every aspect of human retinal degeneration: as retinal degenerations are heterogeneous, neither does any one human. To model the formation of the vitamin A byproducts, we delivered the vitamin A byproduct A2E via intravitreal injection into the rat. With the use of the whole eye, biological cascades involved in inflammation, apoptosis, and necrosis, which occur during human retinal degeneration, can potentially be modeled.

In vivo, the vitamin A byproducts form in the RPE and the outer nuclear layer, where retinaldehyde enters or exits opsin-binding sites. In contrast, intravitreal delivery exposes the entire neural retina to A2E. Despite the whole retina being exposed, marked morphological changes were mainly observed in the RPE, with thinning of the PIOS and OPL. The histological changes to the RPE occurred over months and were not acute, in accord with the slow histological initiation and progression of Stargardt disease and AMD. The similar sequence of pathology and slow timing of the pathology suggest some utility for this intravitreal injection model.

The cause of the PIOS and OPL thinning is not apparent. The thinning may be in response to the A2E-damaged RPE or a direct A2E-induced damage to the neuroretina. As degeneration of the neural retina does not necessarily lead to deterioration of the PRE, as exemplified by the RCS rat (Ryals et al., 2017), the observed RPE damage is unlikely to be a result of A2E-induced damaged to the neural retina. We did not investigate the choriocapillaris. Early damage to choriocapillaris has been argued to drive retinal degeneration (Biesemeier et al., 2014). The choriocapillaris and RPE are mutually dependent: damage to one leads to damage in the other. Previous work has shown that vitamin A cycle byproducts, delivered to the retina, damage the choriocapillaris (Iriyama et al., 2009; Penn et al., 2015). At this stage, it is not clear whether the delivered A2E preferentially damaged the choriocapillaris or RPE, which resulted in subsequent damage of the other, or if both structures were equally damaged primarily.

We interrogated the retina for molecular signatures 2 months after delivery of A2E when ERG amplitudes were reduced by approximately 50%. Several proteins that were dysregulated in this model are also dysregulated during human retinal degeneration. For example, CNTF, a survival factor for neurons and oligodendrocytes, was downregulated here and downregulated in AMD (Harris et al., 2013). Glutathione S-transferase (GST) showed the largest protein expression changes, both up- and downregulated, upon treatment with A2E, spots 42, 43, and 49 in Fig. 7A. A single protein can resolve into multiple spots, reflecting functionally distinct subpopulations due to proteolytic processing, isoforms, or changes in charge resulting from posttranslational modifications (phosphorylation, deamidation, acetylation, glycation, and glutathionylation), indication that treatment induced different subpopulations of the protein. GST expression is regulated by retinoid X receptor (RXRA) (Dai et al., 2005), which is in turn modulated by A2E (Iriyama et al., 2008). The downregulation of GST and the upregulation of pyruvate kinase observed here also occur during AMD's progressive stages (Nordgaard et al., 2006). The second-most upregulated protein was hemoglobin, which is synthesized by the RPE (Tezel et al., 2009). Deposition of hemoglobin within the matrix of Bruch's membrane is significantly associated with retinal senescence (Beattie et al., 2010). Furthermore, hemoglobin and serum albumin ­– upregulated here – are both found in human drusen, a hallmark of AMD eyes (Crabb et al., 2002). As hemoglobin is used for retinal oxygenation, an upregulation of hemoglobin is consistent with a hypoxic environment. Along these lines, fibromodulin, a stimulator of angiogenesis, was also upregulated in A2E-treated eyes and is upregulated in the eyes of people with AMD (Adini et al., 2014). The upregulation of fibromodulin and hemoglobin is consistent with A2E promoting an angiogenic environment responsible for wet-AMD and consistent with previous work suggesting that exogenous A2E can lead to vesicle-like structures in the rabbit retina. Lastly, quantification of mRNA (Fig. 4) and ICAM-1, a cell surface protein that is upregulated upon cytokine stimulation, suggested that A2E triggered a general downregulation of transcripts involved in immunity. Dysregulation in both the innate and adaptive immune pathways is thought to be involved in the etiology of AMD (Bora et al., 2014; Camelo, 2014). Although excessive transcription of immunity-related proteins has been considered harmful, a reduction in transcription may represent an immunodeficiency. Suppression of specific immune pathways could trigger a subsequent upregulation of alternative immune pathways (Ertel et al., 1995). The observation that the majority (9 out of 20) of the proteome changes involved metabolic pathways indicates that the highly metabolic photoreceptors may be affected.

Interrogation of molecular signatures at earlier or later time points would presumably have presented different molecular states. Further, different species, including humans, may respond differently to A2E-induced toxicity. In Stargardt disease, the vitamin A cycle byproducts form abnormally in the young retina. In AMD, the byproducts form, to a similar degree, in the aged retina. The proposed work would conclude that the difference in the two diseases is explained by the age of the eye in which the vitamin A cycle byproducts form. The major relevance of the presented transcriptional and proteomic dysregulation, at this time, is that the vitamin byproduct, A2E, was sufficient alone to cause them.

Data are consistent with the theory that the formation of the vitamin A byproducts is neither benign nor beneficial. This work demonstrates that vitamin A byproducts are capable of initiating changes observed in the prodromal phase of Stargardt and AMD. It presumes, like the aforementioned retinal poisons, that these changes would progress to atrophic lesions, given adequate time, a greater concentration of the vitamin byproducts, or a different biological system. Accordingly, we propose that the morphological and molecular derangements that characterize retinal degenerations, marked by the formation of vitamin byproducts, can be explained by a chronic poisoning from the byproducts. If so, therapeutic interventions that target downstream reactions to the byproducts, without addressing their initial formation, will be limited in therapeutic value. The continued poising will eventually overwhelm any therapeutic block or divert the system to alternative pathways towards degeneration. A therapeutic that retards the formation of the vitamin A cycle byproducts may be necessary to significantly alter the course of these diseases.

All experiments were conducted in accordance with the current Guide for the Care and Use of Laboratory Animals, published by the US National Academies Press. Columbia University's Institutional Animal Care and Use Committee (IACUC) approved animal protocols. We used 3-month-old Fischer crossbred Wistar rats for the albino rats and Long Evans rats (Charles River) for the pigmented animals. Roughly an equal number of male and female animals were used. Animals were allocated to experimental groups at random. No animals were excluded from analysis.

A2E was prepared as previously described (Penn et al., 2014). A2E was dissolved in dimethyl sulfoxide (DMSO). This stock solution was added to a solution of phosphate-buffered saline (PBS) that was probe sonicated during the addition to give a solution of A2E (1 mg A2E/ml PBS) containing 1% DMSO. Ten microliters of the solutions were injected intravitreally into each eye. For controls, a solution of PBS with 1% DMSO in PBS was injected.

Eyes were enucleated and stored at −20°C in the dark before analysis. Either one or two eyes were placed into 1.5 ml of homogenization tube and minced 5-8 times with scissors. Added to the tube were: 500 μl of brine, 10 μl of 37% paraformaldehyde, 400 μl of ethyl acetate (containing 1 mg butylated hydroxytoluene per ml of ethyl acetate) and 0.5 g of stainless-steel beads, 0.9-2 mm in diameter. The mixture was homogenized with a bead mill homogenizer (BBY24M Bullet Blender^® Storm, Next Advance, Inc., Averill Park, NY, USA; setting speed of 8 for 15 min) and centrifuged for 5 min at 14,000 rpm. We removed 200 μl of the butanol phase and added it to a 300 μl capacity autosampler vial (Microsolv Technology Corp., Eatontown, NJ, USA). Samples were held at room temperature in an autosampler until they were injected in the ultra-performance liquid chromatography (UPLC) system.

For UPLC analysis, we injected 30 μl of the above eye extracts into a Flexar FX-15 system with a PDA Detector (PerkinElmer, Waltham, MA, USA). The system contained a 2.1 mm×50 mm, 2.1 μm particle size, C18 column (Brownlee SSP, PerkinElmer) with a 2.1 mm×5 mm guard column with the same packing material (Waters Corp., Milford, MA, USA). The column oven was set to 40°C and samples were eluted at 1 ml per minute, starting with a mixture of acetonitrile containing 0.01% trifluoroacetic acid (solvent A) and 50% type I water containing 10% methanol and 0.01% trifluoroacetic acid (solvent A) for 1 min. The gradient was changed to 100% solvent A over 3 min, and further eluted for 5 min with 100% solvent A. We detected A2E and vitamin A dimers at 445 nm.

ERG recordings were performed as previously described (Ma et al., 2011; Washington et al., 2007). For the dose-response protocol, we used flashes provided by a xenon lamp ranging from 0.0001 to 2.25 cd s/m², scotopic units, in nine steps. For each animal, ERG potentials were recorded from both eyes. Traces judged excessively noisy were rejected. At least two ERG curves were averaged for each step and eye. Traces from both eyes of each animal were then averaged to give an ERG response for each probe flash for each step for each animal. Using GraphPad Prism (GraphPad Software, Inc., La Jolla, CA, USA), we then fit and averaged the a- and b-wave responses for each light intensity and for each animal to a sigmoidal curve (Low, 1987) described by: Y=Bottom+(Top-Bottom)/{1+10 ^ [(LogEC50-X)×HillSlope]}, where Top is the maximum a- or b-wave, in the same units as Y; LogEC50 is the intensity needed to achieve a half-maximum a- or b-wave, in the same units as X; HillSlope is a parameter that describes the steepness of the dose-response curve. P values were calculated with an F-Test.

For the single flash screening protocol, a single flash of 2.25 cd s/m² was presented 5-10 times to each eye. Traces judged excessively noisy were rejected. For each animal, a- and b-waves for the left, A2E-treated eye were normalized to the right, control eye. We then averaged the differences between the right and left eye for each animal.

Eyes were fixed in Excalibur's fixative (Excalibur Pathology, Inc., Norman, OK, USA), embedded in paraffin, sectioned and stained with hematoxylin and eosin using standard procedures. Eyes were sliced through the optic nerve. Four of the eight slices were then imaged 1 mm from either side of the optic nerve head at 400× and 1000× magnifications. The retinal layers were measured using ImageJ (Wayne Rasband, US National Institutes of Health, Bethesda, MD). For each eye, measurements from the four slices were averaged to give the retinal layer thickness per eye. Measurements for each eye were then averaged to give the final retinal layer thickness per experimental cohort. P-values were calculated with a t-test. The investigator was blinded to the group allocation when assessing the outcome.

Profiling of messenger RNA was done as previously described (Ma et al., 2011). We isolated total RNA from eyecups and retinas. For the control group, we isolated RNA from four biological replicates consisting of a total of ten eyes pooled into four groups (i.e. 2 eyes, 2 eyes, 3 eyes, and 3 eyes). For the A2E-treated group, a total of six eyes were used, pooled into four biological replicates (i.e. 1 eye, 1 eye, 2 eyes and 2 eyes) before isolation of total RNA.

The 167 genes screened mediating inflammation and autophagy were: Chemokine genes: RGD1561905_predicted (C5), Ccl11, Ccl12, Ccl17, Ccl19, Ccl2, Ccl20, Ccl21, Ccl22, Ccl24, Ccl25, Ccl3, Ccl4, Ccl5, Ccl6, Ccl7, Ccl9, Cx3cl1, Cxcl1, Cxcl10, Cxcl11, Cxcl12 (Sdf1), Cxcl2, Cxcl5, Cxcl9, Il13, Pf4. Chemokine receptors: Ccr1, Ccr2, Ccr3, Ccr4, Ccr5, Ccr6, Ccr7, Ccr8, Ccr9, Cx3cr1, Cxcr3, Ccr10, Il8ra, Cxcr2, Xcr1. Cytokine genes: Ifng (IFNγ), Il10, Il11, Il13, Il15, Il16, Il17b, Il18, Il1a, Il1b, Il1f5, Il1f6, Il3, Il4, Il5, Itgam, Itgb2, Lta, Ltb, Mif, Aimp1, Spp1, Tgfb1, Tnf, Cd40lg (Tnfsf5). Cytokine receptors: Ifng (IFNγ), Il10ra, Il13, Il13ra1, Il1r1, Il1r2, Il2rb, Il2rg, Il5ra, Il6r, Il6st, Tnfrsf1a (TNFR1), Tnfrsf1b (TNFR2). Other genes involved in the inflammatory response: Abcf1, Bcl6, Cxcr5, C3, Casp1, Crp, Il1r1, Cxcr2, Tollip. Genes involved in autophagic vacuole formation: Ambra1, Atg12, Atg16l1, Atg4b, Atg4c, Atg5, Atg9a, Becn1, Ctsd, Gabarap, Gabarapl2, Irgm, Map1lc3a, Map1lc3b, Psen1, Rb1cc1, RGD1359310, Rgs19, Tm9sf1, Ulk1, Wipi1. Genes responsible for protein targeting to membrane/vacuole: Atg4b, Atg4c, Gabarap. Genes responsible for protein transport: Atg16l1, Atg16l2, Atg3, Atg4b, Atg4c, Atg7, Atg9a, Gabarap, Gabarapl2, Rab24. Genes involved in protein ubiquitination: Atg3, Atg7, Park2, Park7, Pink1. Genes with protease activity: Atg4b, Atg4c. Co-regulators of autophagy and apoptosis: Akt1, App, Atg12, Atg5, Bad, Bak1, Bax, Bcl2, Bcl2l1 (Bcl-x), Becn1, Bid, Bnip3, Casp3, Casp8 (Flice), Npc1, Cdkn1b (p27Kip1), Cdkn2a (p16Ink4), Cln3, Ctsb, Cxcr4, Dapk1, Eif2ak3, Fadd, Fas (Tnfrsf6), Hdac1, Htt, Ifng, Igf1, Ins2, Mapk8 (Jnk1), Mtor, Nfkb1, Park2, Pik3cg, Prkaa1, Pten, Snca, Sqstm1, Tgfb1, Tgm2, Tnf, Tnfsf10 (Trail), Tp53. Co-regulators of autophagy and the cell cycle: Bax, Cdkn1b (p27Kip1), Cdkn2a (p16Ink4), Mapt, Pim2, Pten, Rb1, Tgfb1, Tp53. Autophagy induction by intracellular pathogens: Eif2ak3, Ifng, Lamp1. Autophagy in response to other intracellular signals: Arsa, Ctss, Dram2 (Tmem77), Eif4g1, Esr1 (ERa), Gaa, Hdac6, Hgs, Mapk14 (p38 Mapk), Pik3c3 (Vps34), Pik3r4, Rps6kb1. Chaperone-mediated autophagy: Hsp90aa1.

To determine associations and visualize networks, we used Gene Network Central [SABiosciences (QIAGEN), Venlo, The Netherlands] (http://www.sabiosciences.com/genenetwork/genenetworkcentral.php).

All procedures were performed at 4°C. Eyes were enucleated and eyecups were prepared. Each eyecup was placed into a 1.5 ml centrifuge tube containing 250 mg of 0.1 mm glass beads and 250 mg of 0.5 mm ceria stabilized zirconium oxide beads, 200 μL of ice-cold lysate buffer (1% Triton-X-100, 20 mM HEPES, 2 mM EDTA, pH 7.4) and 2 μl of 100×protease inhibitor cocktail (Cat. P8340, Sigma-Aldrich, St. Louis, MO, USA). Samples were homogenized in a bead mill homogenizer (Bullet Blender homogenizer, Next Advance, Inc.; setting speed of 8 for 10 min) then centrifuged at 20,000 g for 15 min. The supernatant containing the protein was transferred to a new tube and protein concretion was determined by measuring the absorbance at 260 and 280 nm (NanoVue Plus, GE Healthcare Bio-Sciences Corp., Piscataway, NJ, USA).

We used the Quantibody^® Rat Cytokine Array-3 (Cat. QAR-CYT-3-1, RayBiotech, Inc., Norcross, GA, USA) according to the manufacturer's instructions.

The tissue lysate buffer was exchanged into the 2D cell lysis buffer (30 mM Tris-HCl, pH 8.8, containing 7 M urea, 2 M thiourea and 4% CHAPS {3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate}. Protein concentration was measured using the Bio-Rad (Bio-Rad Laboratories, Inc., Hercules, CA, USA) protein assay method. The internal standard was made by mixing an equal amount of protein from each sample.

For each sample, 30 μg of protein was mixed with 1.0 μl of diluted CyDye (GE Healthcare Life Sciences, Piscataway, NJ, USA), and kept in the dark on ice for 30 min. Samples were labeled with Cy3 and Cy5, respectively. The labeling reaction was stopped by adding 1.0 μl of 10 mM lysine to each sample and incubating in the dark on ice for an additional 15 min. The labeled samples were then mixed together. The 2× 2D sample buffer (8 M urea, 4% CHAPS, 20 mg/ml dithiothreitol [DTT], 2% pharmalytes and a trace amount of bromophenol blue), 100 μl DeStreak rehydration solution (7 M urea, 2 M thiourea, 4% CHAPS, 20 mg/ml DTT, 1% pharmalytes and a trace amount of bromophenol blue) were added to the labeling mix to make a total volume of 250 μl for the 13 cm IPG strip. It was mixed well and spun before loading the labeled samples into the strip holder.

After loading the labeled samples, isoelectric focusing (IEF) (pH 3-10 Linear) was run following the protocol provided by GE Healthcare. Upon finishing the IEF, the IPG strips were incubated in the freshly made equilibration buffer-1 (50 mM Tris-HCl, pH 8.8, containing 6 M urea, 30% glycerol, 2% SDS, a trace amount of bromophenol blue and 10 mg/ml DTT) for 15 min with gentle shaking. The strips were then rinsed in the freshly made equilibration buffer-2 (50 mM Tris-HCl, pH 8.8, containing 6 M urea, 30% glycerol, 2% SDS, a trace amount of bromophenol blue and 45 mg/ml DTT) for 10 min with gentle shaking. Next, the IPG strips were rinsed in the SDS gel running buffer before transferring into 12% SDS gels. The SDS gels were run at 15°C.

Gel images were scanned immediately following the SDS-PAGE using Typhoon TRIO (GE Healthcare). The scanned images were then analyzed by Image Quant software (version 6.0, GE Healthcare), followed by in-gel analysis using DeCyder software (version 6.5, GE Healthcare).

The spots of interest were picked up by an Ettan Spot Picker (GE Healthcare Life Sciences) based on the in-gel analysis and spot picking design by DeCyder software. The gel spots were washed a few times then digested in-gel with modified porcine trypsin protease (Trypsin Gold, Promega, Madison, WI, USA). The digested tryptic peptides were desalted by ZipTip^® C18 (EMD Millipore, Billerica, MA, USA). Peptides were eluted from the ZipTip with 0.5 μl of matrix solution (cyano-4-hydroxycinnamic acid 5 mg/ml in 50% acetonitrile, 0.1% trifluoroacetic acid, 25 mM ammonium bicarbonate) and spotted on the AB SCIEX MALDI plate (Opti-TOF^TM 384 Well Insert).

MALDI-TOF MS and TOF/TOF tandem MS/MS were performed on an AB SCIEX TOF/TOF™ 5800 System (AB SCIEX, Framingham, MA, USA). MALDI-TOF mass spectra were acquired in reflectron positive ion mode, averaging 4000 laser shots per spectrum. TOF/TOF tandem MS fragmentation spectra were acquired for each sample, averaging 4000 laser shots per fragmentation spectrum on each of the ten most abundant ions present in each sample (excluding trypsin autolytic peptides and other known background ions).

Both of the resulting peptide mass and the associated fragmentation spectra were submitted to the GPS Explorer workstation equipped Sodium iodate with the MASCOT search engine (Matrix Science, Boston, MA, USA) to search the database of the National Center for Biotechnology Information non-redundant (NCBInr). Searches were performed without constraining protein molecular weight or isoelectric point, with variable carbamidomethylation of cysteine and oxidation of methionine residues, and with one missed cleavage also allowed in the search parameters. Candidates with either protein score C.I.% or Ion C.I.% greater than 95 were considered significant.

Statistical analyses were performed with GraphPad Prism. Values are shown as mean±s.e.m. or s.d. A P-value of less than 0.05 was considered significant as calculated by a two-sided, unpaired, T- or F-test, where appropriate. For population proportions, we used two-tailed, Fisher's Exact Test, using 2×2 contingency tables and the method of summing small P-values.