Ambra1 haploinsufficiency in CD1 mice results in metabolic alterations and exacerbates age-associated retinal degeneration

ABSTRACT Macroautophagy/autophagy is a key process in the maintenance of cellular homeostasis. The age-dependent decline in retinal autophagy has been associated with photoreceptor degeneration. Retinal dysfunction can also result from damage to the retinal pigment epithelium (RPE), as the RPE-retina constitutes an important metabolic ecosystem that must be finely tuned to preserve visual function. While studies of mice lacking essential autophagy genes have revealed a predisposition to retinal degeneration, the consequences of a moderate reduction in autophagy, similar to that which occurs during physiological aging, remain unclear. Here, we described a retinal phenotype consistent with accelerated aging in mice carrying a haploinsufficiency for Ambra1, a pro-autophagic gene. These mice showed protein aggregation in the retina and RPE, metabolic underperformance, and premature vision loss. Moreover, Ambra1+/gt mice were more prone to retinal degeneration after RPE stress. These findings indicate that autophagy provides crucial support to RPE-retinal metabolism and protects the retina against stress and physiological aging. Abbreviations : 4-HNE: 4-hydroxynonenal; AMBRA1: autophagy and beclin 1 regulator 1, AMD: age-related macular degeneration;; GCL: ganglion cell layer; GFAP: glial fibrillary acidic protein; GLUL: glutamine synthetase/glutamate-ammonia ligase; HCL: hierarchical clustering; INL: inner nuclear layer; IPL: inner plexiform layer; LC/GC-MS: liquid chromatography/gas chromatography–mass spectrometry; MA: middle-aged; MTDR: MitoTracker Deep Red; MFI: mean fluorescence intensity; NL: NH4Cl and leupeptin; Nqo: NAD(P)H quinone dehydrogenase; ONL: outer nuclear layer; OPL: outer plexiform layer; OP: oscillatory potentials; OXPHOS: oxidative phosphorylation; PCR: polymerase chain reaction; PRKC/PKCα: protein kinase C; POS: photoreceptor outer segment; RGC: retinal ganglion cells; RPE: retinal pigment epithelium; SI: sodium iodate; TCA: tricarboxylic acid.


Introduction
The eye is a very metabolically active tissue with high levels of nutrient and oxygen consumption. The avascular outer retina relies on nutrient diffusion from the choroid vessels through the retinal pigment epithelium (RPE), which supplies the retina with nutrients such as glucose [1,2]. This glucose is then metabolized via the tricarboxylic acid (TCA) cycle, oxidative phosphorylation (OXPHOS), and aerobic glycolysis to produce glycerol and ATP, which in turn support phototransduction and phospholipid synthesis to renew photoreceptor outer segments (POS) [3]. β-oxidation of fatty acids and lactate are the main energy sources of the RPE, and are derived from the digestion of POS and photoreceptor glycolysis, respectively [4]. This arrangement results in a high degree of metabolic interdependence between the neural retina and RPE, whereby glucose is converted into lactate in the retina and then shunted to the RPE [5]. The metabolic ecosystem resulting from this interdependence between the two tissue types is key to the preservation of vision [5].
Autophagy is crucial to preserve cell homeostasis, serves as a recycling mechanism to extract nutrients and eliminate damaged cell components, and has important implications in neurological diseases [6,7]. While studies are beginning to unravel the role of autophagy in the eye and the consequences of autophagy alterations in ocular and retinal diseases, the degree of retinal degeneration depends on several factors including the protein that is altered, the cell type affected, and the age of the animal [8,9]. We previously demonstrated a progressive, age-dependent decrease in autophagic flux in the retina of C57/BL6 mice, beginning at 12 months of age [10]. In addition, complete autophagy deficiency, achieved by deleting Atg5 in neuronal precursors, results in severe retinal degeneration and vision loss at 7 weeks of age [10] while Atg5 deletion in only cones or rods results in much more subtle phenotypes [8,11,12]. Two mouse models that completely lack the essential autophagy proteins ATG5 or ATG7 in the RPE show increased predisposition to age-related retinal degeneration [13]. However, other studies in which these same autophagy regulators were eliminated in the mouse RPE reported no retinal degeneration [14,15], indicating that further research is needed to determine whether complete autophagy blockade in the RPE results in reduced visual function. More importantly, it is unclear how slight reductions in autophagy in both the RPE and retina, as occurs during physiological aging, affect retinal metabolism and visual function. AMBRA1 (autophagy and beclin 1 regulator 1) is a multifunctional scaffold protein that promotes autophagy initiation by triggering interaction between BECN1 and PIK3C3/VPS34 [16,17] and ULK1 activity in a MTOR-dependent manner [18]. Ambra1deficient (Ambra1 gt/gt ) mice display embryonic lethality and show massive alterations in the central nervous system (CNS) including exencephaly and spina bifida [16], as well as neurogenesis alterations in the olfactory bulb and sub-ventricular zone [19,20]. Moreover, decreased Ambra1 expression is linked to an attenuated autophagic response that may be implicated in neural disorders and neurodegenerative diseases such as autism, schizophrenia, and Alzheimer disease [21][22][23]. Recent findings indicate that AMBRA1 mediates an autophagy-mediated protective response to injury in retinal ganglion cells (RGC) [24,25]. It remains to be determined how Ambra1 deficiency affects the retina-RPE ecosystem in the context of aging.
Here, we showed that monoallelic deletion of Ambra1 resulted in diminished autophagic flux and induced protein aggregation, lipid peroxidation, and oxidative stress in the retina and RPE. Loss of autophagy-dependent proteostasis led to exacerbated RPE degeneration at 1 year of age and retinal neurodegeneration in 2-year-old mice, culminating in premature vision loss. Moreover, autophagy reduction was associated with metabolic imbalance in Ambra1 +/gt mice, including mitochondrial defects, altered glycolysis, and energetic underperformance. These data demonstrate for the first time that partial loss of autophagy results in marked metabolic alterations and increased susceptibility to RPE stress and age-dependent retinal degeneration.

Age-dependent vision loss is exacerbated in Ambra1 +/gt mice
We investigated how a moderate reduction in autophagy, as occurs during physiological aging, impacts retinal homeostasis. We used Ambra1 +/gt haploinsufficient mice [16], which are viable and display slight reductions in autophagic flux [25]. This outbred model, which has a CD1 background, provides a suitable means of studying age-dependent changes in a physiologically relevant context [26]. Autophagic flux decreased with age in this mouse strain, in agreement with our previous findings (Figures 1A and S1B) [10]. Retinas from Ambra1 +/gt mice displayed reduced Ambra1 gene expression ( Figure S1A) and a parallel reduction in autophagic flux compared with control (Ambra1 +/+ ) littermates at a young age (3-4 months old) ( Figure 1A). Compared with control littermates, gene expression of many essential autophagy regulators such as Atg5 and Sqstm1 was already diminished in middle-aged (MA, 12-15 months) Ambra1 +/gt mice, and further decreased in old (22-26 months) mice ( Figure 1B). At the protein level, ATG12-ATG5 conjugation or unconjugated ATG5 were decreased slightly in parallel with the ageassociated decline in autophagy. Ambra1 haploinsufficiency was not associated with any alterations in Atg5 forms ( Figure  S1D). Moreover, Ambra1 +/gt mice displayed a sustained reduction in mRNA expression of genes such as Tfeb and Wipi2 ( Figure 1B), suggesting exacerbation of autophagy deficiency in old Ambra1 +/gt mice. Histological analysis revealed a progressive reduction of LC3 + puncta with increasing age in all retinal layers, an effect that was more pronounced in Ambra1 +/gt mice ( Figure 1C, E and S1E). In agreement with the observed reduction in autophagic flux in young animals, Ambra1 +/gt mice showed a greater accumulation of protein aggregates as measured using the ProteoStat® protein aggregation assay ( Figure 1D, F). These differences were abolished with increasing age: in control Ambra1 +/+ littermates protein aggregation increased with age. Together, these results demonstrate an age-associated decline in autophagy function that correlates with the loss of proteostasis, a phenotype already evident in young Ambra1 +/gt mice.
Defective autophagy in the Ambra1 +/gt retina was accompanied by reduced visual function, an effect that was already significant in middle-aged animals compared with control littermates (Figure 2A-H). In low light (scotopic) conditions, we observed decreased amplitudes of b-scot (low intensity stimulus) and both a-and b-mixed (high intensity stimulus) waves, implying decreased photoreceptor and bipolar cell signal transduction, respectively, in Ambra1 +/gt versus Ambra1 +/+ mice (Figure 2A,B). In photopic conditions, decreased amplitudes of b-photopic and flicker waves indicated defective cone-dependent responses ( Figure 2F,G). More detailed analysis based on b:a wave ratio showed a significant reduction in old Ambra1 +/gt versus Ambra1 +/+ animals ( Figure 2D), suggesting poor performance of bipolar cells at this age due to a more pronounced reduction of bwave than a-wave amplitudes. Finally, diminished oscillatory potentials (OP) suggested poor performance of inner layer cells ( Figure 2E). Regression analyses revealed that monoallelic deletion of Ambra1 resulted in faster age-associated vision loss than in control mice, as evidenced by a significantly steeper slope for Ambra1 +/gt versus Ambra1 +/+ mice for all parameters studied ( Figure S2).
To examine the effects of the age-associated decline in autophagy on retinal morphology, we compared ONL (outer nuclear layer) nuclear density in young, middle-aged, and old Western blots corresponding to middle-aged and old mice are shown in Figure S1. Autophagic flux is determined as the ratio between NL and Control (n = 3-4 per group). (B) Decreased mRNA expression of Atg5, Sqstm1, Tfeb and Wipi2 as determined by qPCR of whole retina extracts from young, middle-aged, and old Ambra1 +/+ and Ambra1 +/gt mice (n = 3-4 per group). (C-F) Representative images and quantification of LC3 puncta in the outer nuclear layer (cyan) (C) and protein aggregates in the whole retina as measured by ProteoStat® protein aggregation assay (yellow) (D) in young, middle-aged, and old Ambra1 +/+ and Ambra1 +/gt mice. Nuclei were counterstained with DAPI (gray). Graphs show quantification of LC3 + puncta (E) and protein aggregates (F) (n = 3-9 per group). Data are presented as the mean ± SEM. *p < 0.05: two-tailed Student´s t-test or Mann-Whitney U-test (A, E, F); two-way ANOVA followed by post hoc LSD test for age (A and E) and genotype (B). Scale bars: 25 µm.
Ambra1 +/gt mice versus Ambra1 +/+ littermates. Old Ambra1 +/ gt mice showed reduced ONL volume density ( Figure 3A,B), but no significant increase in the number of apoptotic TUNEL-positive cells ( Figure S3A). These apoptotic cells may have been already removed by the more numerous microglial cells with amoeboid morphology, often associated to their phagocytic phenotype ( Figure S3C). The number of cones stained with either ARR3/Cone Arrestin (arrestin 3, retinal) or OPN1MW/red/green opsin (opsin 1 (cone pigments), medium-wave-sensitive (color blindness, deutan)) was reduced in aged Ambra1 +/gt retinas ( Figure 3C,D and S3B), indicating that Ambra1 haploinsufficiency results in an age-associated reduction in cone population. We next investigated whether monoallelic loss of Ambra1 results in alterations in the inner retina. Ambra1 +/gt mice showed a reduction in the thickness of the inner plexiform layer (IPL) ( Figure 3E, F) compared with Ambra1 +/+ littermates, suggesting reduced synapsis between retinal neurons. Moreover, in these mice, bipolar cells showed protrusions toward the ONL ( Figure 3E [insets], G), a phenotype associated with the bipolar cell response to photoreceptor retraction and connectivity loss [27]. In terms of photoreceptor and bipolar morphology and connectivity, the observed phenotypes match perfectly with the loss in visual function revealed by electroretinogram. Old Ambra1 +/gt and control Ambra1 +/+ mice showed no significant differences in the number of retinal ganglion cells stained with the specific transcription factor POU4F1/BRN3A ( Figure 3H). Müller cells are retinal glial cells that provide trophic factors and support the neuroretina. In response to damage stimuli, Müller cells activate and extend GFAP-positive projections toward the outer retina. A persistent inflammatory and gliotic response accompanying aging has been described in mice [28]. Although we did not observe major morphological changes in Müller glia stained with the Müller cell marker GLUL (glutamate-ammonia ligase (glutamine synthetase)) ( Figure S3D), we observed a massive increase in the number of GFAP projections toward the ONL in aged Ambra1 +/gt versus Ambra1 +/+ mice ( Figure 3I). This phenotype was accompanied by a larger area of GFAP + staining at the level of the ganglion cell layer in old Ambra1 +/gt retinal flat mounts ( Figure 3J), indicating astrocyte hypertrophy. In agreement with this gliosis response, we observed a slight increase in amoeboid microglia in old Ambra1 +/gt versus Ambra1 +/+ retinas ( Figure S3C) that could be linked to their activated state during inflammation [29,30]. Taken together, these data suggest that Ambra1 haploinsufficiency results in an accelerated retinal aging phenotype, accompanied by a proinflammatory state and vision defects.

Monoallelic deletion of Ambra1 results in metabolic alterations in the retina associated with mitochondrial dysfunction
Recent findings by our group demonstrate that autophagydeficient animals display marked metabolic alterations in the embryonic retina [31], and that retinal degeneration is often associated with metabolic imbalances [32]. We next explored how deficient autophagy influences the retinal metabolome in the early stages of degeneration by comparing retinas from young and middle-aged Ambra1 +/+ and Ambra1 +/gt mice. Metabolite levels were determined by liquid chromatography/gas chromatography-mass spectrometry (LC/GC-MS) and concentrations represented as the area under the curve and metabolite ratios. Although sample correlation, principal component analysis, and unsupervised hierarchical clustering analysis revealed greater differences for age than for genotype ( Figure 4A, and S4A-B), middle-aged Ambra1 +/gt retinas showed the most pronounced changes in retinal metabolome.
To determine which metabolites were dysregulated, we performed a Skillings-Mack analysis to identify significant genotype-related effects, followed by an enrichment analysis. Of the 100 metabolites measured ( Figure S4A), 16 were significantly altered in middle-aged Ambra1 +/gt retinas ( Figure  4B). Those metabolites are implicated in purine metabolism, the pentose phosphate pathway, and the Warburg effect, among other processes ( Figure S4C). Nucleotide metabolite ratios revealed alterations in adenine and guanine metabolism in Ambra1 +/gt versus Ambra1 +/+ retinas ( Figure 4C). Furthermore, middle-aged Ambra1 +/gt retinas displayed marked reductions in total nucleotide phosphate levels and adenylate energy charge ( Figure 4D), a phenotype associated with energetic underperformance [33]. Interestingly, while amino acid levels were increased in young Ambra1 +/gt versus Ambra1 +/+ mice, levels of most amino acids analyzed and of many TCA cycle intermediates were reduced in middle-aged Ambra1 +/gt mice ( Figure 4E). Finally, the NADH:NAD and the NADPH:NADP ratios were altered and levels of ribose-5phosphate increased in middle-aged Ambra1 +/gt versus Ambra1 +/+ mice ( Figure 4F-G), indicating marked metabolic alterations suggestive of defective oxidative metabolism in Ambra1 +/gt mice.
We next investigated whether alterations in mitochondrial function in Ambra1-deficient retinas could explain the exacerbated metabolic dysregulation observed in middle-aged animals. Surprisingly, marked reductions in mitochondrial membrane potential as determined by flow cytometry with DiOC 6 were already detectable in young Ambra1 +/gt retinas ( Figure 5A). TOMM20 immunostaining revealed augmented mitochondrial mass in most retinal cell layers in aged Ambra1 +/gt animals ( Figure 5B,C). Increases in mitochondrial biogenesis can often compensate for mitochondrial alterations. However, Ambra1 +/gt animals displayed no such changes and old animals even showed reduced mRNA expression of the main mitochondrial biogenesis regulators Ppargc1a, Nrf1, Nfe2l2, and Tfam ( Figure 5D), as well as decreased expression of mitochondrial genes such as Timm23 and Cox4i1 ( Figure 5E). Thus, the increased mitochondrial mass in the absence of biogenesis suggests a defect in the removal of dysfunctional mitochondria via mitophagy in aged Ambra1 +/gt animals.
We further investigated whether these mitochondrial alterations result in oxidative damage to intracellular membranes. Photoreceptor outer segments (POS) are enriched in lipids that are susceptible to peroxidation upon oxidative stress, affecting their functionality [34]. Compared with control littermates, middle-aged Ambra1 +/gt mice showed increased levels of lipid peroxidation in all retinal layers, as    mitochondrial membrane potential in young animals that cannot be compensated for in later life by autophagy or mitochondrial biogenesis, potentially resulting in lipid peroxidation and oxidative damage to cellular membranes. Our enrichment analysis ( Figure S4C) revealed upregulation of the pathway underlying the Warburg effect. Because the pyruvate:lactate ratio was increased in young Ambra1 +/gt versus Ambra1 +/+ mice ( Figure 6A), we next examined mRNA expression of the main glycolytic enzymes. In all age groups, Ambra1 +/gt retinas showed reduced mRNA expression of many glycolytic enzymes, including Gapdh, Hk2, and Pkm ( Figure 6B), and a tendency toward reduced GAPDH and HK2 immunofluorescence ( Figure 6C-F). In agreement with our metabolic findings in middle-aged Ambra1 +/gt retinas, we observed decreased transcription of enzymes that help maintain cellular redox status. For example, we observed reduced expression of Nqo1 (NAD(P)H dehydrogenase, quinone 1) ( Figure 6G), an enzyme that oxidizes NADH to NAD in the electron transport chain, in agreement with the aforementioned alterations in NADH:NAD ratio ( Figure 4F). Furthermore, we detected lower mRNA levels of Pdha1 (pyruvate dehydrogenase E1subunit alpha 1). PDHA1 protein is crucial to allow entry of acetyl-CoA produced from pyruvate into the TCA cycle, where it serves as a source of carbon for anabolism ( Figure 6G). Together these data demonstrate that mitochondrial alterations already evident in young Ambra1 +/ gt retinas may underlie the metabolic changes, including decreased glycolysis and energetic underperformance, observed in later life.

RPE alterations in Ambra1 +/gt mice result in increased susceptibility to stress
Retinal metabolism relies heavily on the RPE, and vice versa. For example, photoreceptors mainly obtain energy through glycolysis from glucose provided from the outer retinal blood supply and transported through the RPE, and, in exchange, return lactate and lipid-containing outer segments that are used to fuel oxidative phosphorylation in the RPE [35]. Because alterations in RPE homeostasis have important consequences for retinal metabolism and function, we examined age-associated changes in RPE function in Ambra1 +/gt mice. While the RPE in young Ambra1 +/gt mice showed no alterations with respect to control littermates, phalloidin staining in middle-aged Ambra1 +/gt animals revealed important RPE lesions corresponding to monolayer disruption ( Figure 7A) [36]. Moreover, middle-aged Ambra1 +/gt mice displayed fewer ( Figure 7B) and larger ( Figure 7C) RPE cells, phenotypes associated with RPE cell death, whereby the remaining cells expand to cover the area left by dying cells [36]. ProteoStat protein aggregation assay showed that monoallelic loss of Ambra1 was associated with increased proteotoxicity, an effect that was already significant in young animals ( Figure 7D). Importantly, those differences were no longer observed in old animals, as protein aggregation was also increased in the old control animals. In agreement, autophagosome number determined by LC3 staining was reduced both in flat mounts and RPE cryosections from Ambra1 +/gt versus Ambra1 +/+ mice ( Figure S5A-D). The age-associated increase in lipid peroxidation was significantly augmented in the RPE of middle-aged Ambra1 +/gt mice ( Figure 7E). Chronic RPE dysfunction is associated with the appearance of autofluorescent lipofuscin aggregates [34]. Lambda-scan autofluorescence analysis showed specific accumulation of lipofuscin inside swollen lysosomes in old Ambra1 +/gt mice ( Figure 7F). In agreement, lysosomes in Ambra1 +/gt animals were similar in number to those of control littermates, but with greater individual volume ( Figure 7G,H). In summary, these data show that monoallelic loss of Ambra1 is associated with marked alterations in the RPE from a young age. These changes include a reduction in autophagy that was paralleled by increased protein aggregation, oxidative damage, lysosomal alterations, and lipofuscin accumulation, resulting in RPE cell death by middle age.
As previously stated, mitochondrial function is required to sustain ATP synthesis in the RPE. RPE flat mounts from middle-aged Ambra1 +/gt animals displayed numerous cells with markedly reduced mitochondria mass, as determined by TOMM20 staining (Figure 7I,J). Those mitochondriapoor cells were enlarged and showed alterations in cytoskeleton architecture, as assessed by phalloidin staining ( Figure 7K, L), suggesting a correlation between reduced mitochondrial mass and RPE hypertrophy. Western blot for TOMM20 and TIMM23 also revealed a trend toward reduced mitochondria protein levels ( Figure S5E). Given the high degree of metabolic interdependence between the retina and RPE, we hypothesize that alterations in mitochondrial function may make the RPE much more dependent on glucose for sustained ATP synthesis. In agreement, our data show a tendency toward increased SLC2A1/GLUT1 glucose transporter expression in the RPE of Ambra1 +/gt animals, and a tendency toward increased glycolytic GAPDH expression in the RPE of young and middle-aged Ambra1 +/gt mice ( Figure S5F, G), pointing to increased glucose demand in the RPE. Finally, the accumulation of activated phagocytic cells was evidenced by an increase in the number of amoeboid cells positive for AIF1/IBA1 (allograft inflammatory factor 1) staining in the RPE in middle-aged Ambra1 +/gt mice ( Figure S5H). Taken together, these data indicate that the autophagy defect with Ambra1 haploinsufficiency results in reduced mitochondrial mass, and metabolic reprogramming that favors glycolysis to fuel ATP synthesis.
In age-related diseases there is usually an interplay between physiological aging and both genetic and environmental factors. Since the alterations in the RPE of Ambra1 +/gt mice occur earlier and are more severe than those seen in the neuroretina, we used a pharmacological model of RPE degeneration to assess vulnerability to external stressors and agerelated defects in these mice [37,38]. Middle-aged mice were treated with sodium iodate (SI), a well-studied model of primary RPE damage that leads to secondary retinal degeneration. Middle-aged Ambra1 +/gt mice and control littermates were intraperitoneally injected with 50 mg/kg SI or vehicle and sacrificed 1 week later. No differences in retinal morphology or retinal layer thickness were observed in the vehicletreated groups ( Figure 8A-B). However, compared with control littermates, Ambra1 +/gt mice were significantly more sensitive to SI-induced damage, as evidenced by a decrease in ONL thickness ( Figure 8B). Surprisingly, SI-treated Ambra1 +/ gt mice also displayed a significant decrease in inner nuclear layer (INL) thickness ( Figure 8B) that was not observed in SItreated control littermates. This suggests an increased vulnerability of both retinal photoreceptors and interneurons to stress, correlating with loss of visual function in old animals. Supporting increased stress vulnerability in middle-aged Ambra1 +/gt mice, these animals showed increased  photoreceptor apoptotic cell death, as determined by TUNEL staining (Figure 8C,E), and decreased cone number after SIinduced damage ( Figure 8D,F). SI treatment exerts a degenerative effect following a central to peripheral gradient [37]. Assessment of RPE monolayer morphology using phalloidin staining showed severe RPE disruption, even in the peripheral region of the retina, in SI-treated Ambra1 +/gt mice ( Figure  8G). Given the retinal phenotype observed in Ambra1 +/gt mice, and the fact that SI exerts a time-and region-dependent effect [38], this finding demonstrates increased susceptibility of the Ambra1 +/gt RPE to stress, resulting in exacerbated retinal degeneration.

Ambra1 is upregulated in RPE cells after SI treatment and in the macular region of age-related macular degeneration (AMD) patients
ARPE-19 cells are an RPE-derived human cell line [39] that has been widely used to study the effects of SI in vitro. Treatment of ARPE-19 cells with SI for 24 h resulted in loss of cell viability and mitochondria membrane potential, decreased mitochondrial mass, and increased lipid peroxidation ( Figure S6A, B), characteristics that are phenotypically similar to that observed in the RPE of old Ambra1 +/gt mice. Next, we analyzed a dataset derived from transcriptionallyprofiled ARPE-19 cells treated for 24 h with SI (GSE142591) [40], and found that mRNA expression of AMBRA1 as well as many genes of the autophagy-initiation complex, including BECN1, ULK1, and RB1CC1, were significantly upregulated in response to SI-induced damage ( Figure S6C, D). Minor increases in mRNA levels were observed for autophagy core genes such as MAP1LC3B, ATG5, ATG7, and ATG16L1 ( Figure S6D). These results suggest upregulation of autophagy machinery is a mechanism to protect against SI-induced damage in RPE cells. Consistent with this idea, we also observe an increase in AMBRA1 expression in the retina of SI treated mice ( Figure S6E). Together, these data indicate that Ambra1-associated autophagy in RPE may play a key protective role in preventing age-or stress-dependent RPE dysfunction.
As defects in autophagy have been implicated in AMD pathology [41][42][43], we sought to determine what characteristics of our animal model with autophagy deficiency replicate features of AMD. Our data shown previously indicate that Ambra1 haploinsufficiency results in RPE hypertrophy ( Figure 7A-C) [36,44], increased oxidative stress (Figures 5F-G and 7E) [45], and inflammation ( Figure 3I-J, S3C-D), features previously linked to AMD [46]. Using a publicly available dataset of RNA sequencing (RNA-seq) profiles of macular and non-macular regions of the human retina and RPE from healthy controls and AMD patients (GSE135092) [47], we investigated the expression of the main autophagy regulators ( Figure 9A). While AMBRA1 mRNA expression was increased in the macular region of AMD donors, no changes were observed in the non-macular region ( Figure  9B). A similar trend was observed for RB1CC1 and ULK1 ( Figure 9C). This gene expression pattern demonstrates that in AMD the autophagy initiation machinery is specifically induced in the macular region, but not in the non-macular region, and that this may constitute an important mechanism to combat the disease.
Metabolic underperformance in the RPE-retina induced by Ambra1 deficiency was associated with RPE degeneration (Figure 7A,B) and loss of visual function (Figure 2). We therefore investigated whether the aforementioned RNA-seq dataset [47] revealed metabolic alterations in the retina and RPE in AMD patients. In line with our findings in Ambra1deficient retinas, mRNA expression of many glycolytic enzymes was significantly decreased in both macular and non-macular regions of the retina from AMD patients. However, no glycolytic changes were observed in the RPE ( Figure 9E-H). These observations, together with the observed decrease in RPE mRNA expression of IDH2 ( Figure 9J), the limiting enzyme in the reductive carboxylation pathway essential for RPE metabolism, suggest decreased mitochondrial oxidative function in the RPE in AMD patients. Taken together, these findings support the view that defective mitochondrial function in the RPE results in a glucose metabolism imbalance, leading to decreased glycolysis in the retina. Moreover, the transcriptional upregulation of autophagy machinery, including AMBRA1, observed in the affected macular region in AMD patients suggests a key role of Ambra1-dependent autophagy in protecting the retina in AMD patients.
Taken together, these data demonstrate that Ambra1 deficiency alters the retina-RPE ecosystem leading to RPE cell death, metabolic underperformance, loss of visual function, and increased susceptibility to stress. Our data also show that Ambra1 expression is selectively upregulated in the macular, but not the non-macular, region of the RPE in AMD patients. Overall, our study demonstrates that AMBRA1-mediated autophagy may be a key process for preventing age-and stress-dependent RPE dysfunction and retinal degeneration.

Discussion
In this study we show that a moderate reduction in autophagy, achieved through monoallelic deletion of Ambra1 (Ambra1 +/gt mice), exacerbates the retinal degeneration and visual function impairment associated with the aging process. We also describe metabolic alterations associated with mitochondrial dysfunction in Ambra1 +/gt retinas that exacerbate and Ambra1 +/gt littermates and corresponding quantification of mean fluorescence intensity (MFI) (n = 3-5 per group). (F) Measurement of autofluorescence in the RPE using Lambda-scan to capture the entire emission spectrum upon UV excitation in old Ambra1 +/+ and Ambra1 +/gt littermates (n = 5-8 per group). (G) RPE flat mounts from middle-aged Ambra1 +/+ and Ambra1 +/gt immunostained with the lysosomal membrane marker LAMP2 (nuclei are counterstained with DAPI [blue]), and (H) corresponding quantification of the number and individual volume of puncta (n = 3-5 per group). (I-K) Immunostaining of mitochondria (TOMM20, green) in RPE flat mounts from middle-aged Ambra1 +/+ and Ambra1 +/gt littermates. Nuclei were counterstained with DAPI (blue) and the F-actin marker phalloidin (red). (J-L) Quantification of TOMM20 staining in I-K. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01; ***p < 0.001; ****p < 0.0001: two-tailed Student's t-test (B, D-F), Mann- Whitney U-test (A, H, J and L), or Kruskal-Wallis test followed by Dunn's post hoc test (C). Scale bars: 25 µm.
age-dependent photoreceptor degeneration and vision loss. In the RPE, proteostasis alterations were already evident in the young Ambra1 +/gt mice, resulting in RPE atrophy, increased lipid peroxidation, and the formation of lipofuscin aggregates later in life. Upregulation of Ambra1 and of genes of the autophagy initiation machinery was also observed in the macular, but not the non-macular, region in samples from AMD patients, highlighting the protective role of autophagy  in combatting disease. Lastly, Ambra1 haploinsufficiency sensitized mice to RPE-targeted damage and subsequent photoreceptor death caused by in vivo sodium iodate administration, a classical model of RPE degeneration. Ambra1 +/gt mice provide a useful means of investigating how a slight reduction in autophagy, similar to which accompanies aging, impacts the metabolism and function of the retina and the RPE. As a scaffold protein, AMBRA1 is involved in a wide-range of cellular processes, including the cell cycle [48][49][50][51], tumor growth and invasion [52], cell death [53], and gene expression regulation [54]. However, most of these AMBRA1 functions occur in mitotically active tissues or cultured cells. In the retina, which is a post-mitotic tissue, AMBRA1ʹs contribution to the autophagic pathway may be of greater relevance. In line with this view, we observed no differences in MYC/c-MYC, CCND1 (cyclin D1), FOXO3 or FOXP3 expression between Ambra1 +/gt mice and control littermates at any age ( Figure S7), suggesting that the observed phenotype may be specifically attributed to the pro-autophagy role of Ambra1.
In this study, we use an outbred strain and heterozygous animals, in which some level of autophagic flux is retained. Partial loss of autophagic flux ensures a physiologically relevant context that cannot be achieved using other autophagy-deficient models, which are either perinatal lethal or, in the case of Atg5 heterozygous animals, retain normal autophagy levels. Also, because AMBRA1 is only implicated in canonical autophagy, in contrast to ATG5 and ATG7 [55], our experimental setting eliminates the confounding effect of LC3-associated phagocytosis in photoreceptor outer segment digestion [14] and of general lysosomal deficiency [56]. Our mouse model therefore provides the unique context in which to examine how the retina-RPE ecosystem copes with the slight reduction in canonical autophagy observed during physiological aging.
The retina's energy consumption is among the highest of all tissues, even exceeding that of the brain. The retina's metabolic requirements are fulfilled by glucose, 80% of which is oxidized to lactate via aerobic glycolysis. These reactions produce ATP but also supply many of the building blocks, such as glyceraldehyde 3-phosphate, needed to fuel lipid biosynthesis. This is key to preserve photoreceptor and retinal function [57], as approximately 10% of POS are shed, phagocytized, and degraded daily by the RPE. Carbons from glucose also are completely oxidized to ATP in the mitochondria, as evidenced by the retina's high level of oxygen consumption [57], and TCA cycle intermediates are also used to support anabolic reactions. These high metabolic requirements of photoreceptors are sustained by the choroidal vasculature, which delivers glucose via the RPE. The RPE is specialized in a form of energy metabolism called reductive carboxylation, which can support redox homeostasis and spares glucose for delivery to photoreceptors [58]. Our data show decreased glycolytic metabolism in the Ambra1 +/gt retina, as demonstrated by reduced mRNA expression of Hk2, Pkm, and Gapdh and an increased pyruvate:lactate ratio. Reduced lactate levels increase the glucose demand of the RPE, as evidenced by the slight increase in SLC2A1/ GLUT1 expression, thereby decreasing glucose delivery to the retina. Making RPE cells more glycolytic induces degeneration of neighboring photoreceptors [32,59]. In normal conditions, the retina would manage this metabolic stress by activating autophagy [60] to cope with starvation and preserve visual function, as observed in mouse models in which autophagy is selectively blocked in the RPE alone [13][14][15]. However, when autophagy is defective, as in Ambra1 +/gt mice, the retina cannot appropriately respond to starvation by activating autophagy, thus resulting in retinal degeneration and vision defects in the long term. Mitochondrial damage and reduced glucose oxidation also result in a reduction in TCA cycle intermediates and impaired visual function [61], consistent with the marked reductions in amino acids and TCA cycle intermediates that we observed in middle-aged Ambra1 +/gt retinas.
Our findings indicate that marked metabolic alterations in the retinas of middle-aged Ambra1 +/gt mice impair retinal function. Purine metabolism was one of the processes in which the most significant alterations were observed. Although ATP levels were only slightly decreased, we observed a reduction in the total energy charge, in agreement with findings in cancer cells in which autophagy provides metabolic substrates to maintain energy charge and nucleotide pools [33]. Both de novo purine biosynthesis and the purine salvage pathway require phosphoribosyl pyrophosphate (PRPP) synthesis from ribose-5-phosphate via the pentose phosphate pathway (PPP) [62], which is highly ATP-consuming. Increases in non-phosphorylated purines and ribose-5-phosphate in old Ambra1 +/gt retinas suggest that altered recycling of nucleotides in autophagy-deficient conditions may enable a compensatory increase in purine biosynthesis, which is reduced due to energy deficit. The PPP is also a source of reducing power in the form of NADPH, and we detected an increase in the NADPH:NADP ratio in old Ambra1 +/gt retinas. Another possible explanation is a reduction in the consumption of NADPH for POS lipid synthesis [63,64]. Our observations suggest a pronounced bioenergetic imbalance in Ambra1 +/gt retinas, in which we also observed an increase in the NADH:NAD ratio. Because NADH is mainly consumed through OXPHOS, we speculated that this finding could reflect dysfunction at the level of the mitochondrial electron transport chain. Notably, mitochondrial mass was unchanged in young Ambra1 +/gt mice, although mitochondrial membrane potential was reduced. In middle-aged Ambra1 +/gt retinas, we observed downregulation of the Nqo1 gene, and of the transcription factors NRF1 and NFE2L2/Nrf2, which drive transcription of OXPHOS genes, resulting in lower mitochondrial metabolism. Overall, these results suggest alterations in the electron transport chain in the Ambra1 +/gt retina. Recently, it was reported that Ambra1 mediates PINK1 stabilization during mitophagy [65], leading to the speculation that mitochondrial alterations observed in Ambra1 +/gt mice could be the consequence of impaired mitophagy.
Our findings in Ambra1 +/gt retinas are also consistent with RPE damage together with alterations in proteostasis that are already evident in young animals. Indeed, deletion of other autophagy-related genes has been previously reported to induce RPE degeneration [13,43,56,66]. The RPE is a postmitotic tissue that requires efficient quality control mechanisms. Because the RPE is exposed throughout life to oxidative stress as a consequence of the visual cycle [45], autophagy plays an essential housekeeping role and facilitates the removal of intracellular debris produced by chromophore recycling [67]. Decreased autophagy has important consequences, resulting in cell enlargement and disruption of the RPE monolayer. These features can be observed in physiological aging in some geriatric mice [36], but are also seen in RPE samples from aged human patients and those with AMD [44]. We propose that Ambra1 haploinsufficiency gives rise to RPE dysfunction that renders the RPE unable to sustain its metabolic load, ultimately leading to accelerated aging in the neural retina and vision loss. Furthermore, we show that middle-aged Ambra1 +/gt mice are more sensitive to SI, which is commonly used to model RPE degeneration [37].
Our analysis of a publicly available RNA-seq dataset of retinal and RPE tissues from control and AMD patients revealed that genes of the autophagy initiation machinery, including AMBRA1, were markedly upregulated in the macular region of the RPE compared with the peripheral retinal region. This suggests that the AMBRA1-dependent autophagy initiation complex may be induced to counteract the stress that causes RPE dysfunction and cell loss, a key feature of macular degeneration. Disrupted metabolism was also evident. Glycolytic enzymes that were highly expressed in the retina were significantly downregulated in AMD. In RPE, the decrease in IDH2 implicates decreased mitochondrial function with AMD, which is consistent with reduced mitochondrial function reported in primary RPE cultures from human donors with AMD [42,68]. While our animal model does not fully replicate the cardinal features of AMD, it does inform about the retinal consequences associated with a reduction in autophagy, which has been reported in RPE tissue and primary cultures from human donors with AMD [42,43]. Taken together, these results reveal the metabolic alterations that occur in the RPE-retina in retinal degeneration and highlight the potential role of autophagy induction in preserving retinal homeostasis.
In conclusion, our findings showing that Ambra1 haploinsufficiency exacerbates age-associated retinal degeneration indicate that maintaining autophagy is crucial to preserve metabolism and retinal function and that slight alterations in autophagy similar to those observed during normal aging promote early RPE damage. Disruption of retinal-RPE homeostasis has important consequences for visual function later in life. Therefore, interventions designed to preserve autophagy could be of therapeutic benefit to prevent blindness in the elderly as well as age-dependent retinal diseases.

Animal procedures
All animal experiments were performed following European Union guidelines and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Animal procedures were authorized by the CSIC ethics committee and the Comunidad de Madrid (PROEX244/17). Ambra1 mutant mice (Ambra1 +/gt ) bred on a CD1 background were kindly provided by F. Cecconi [16]. In this study we used male and female Ambra1 +/gt mice and Ambra1 +/+ littermates aged 3-4 (young), 12-15 (middle-aged), and 22-26 (old) months. The ambra1 gt/gt mice were not used as these display embryonic lethality [16]. Mice were maintained at the CIB animal facility in a temperature-controlled barrier facility on a 12-h light/ dark cycle, with free access to water and food. All animals were sacrificed by cervical dislocation between 9:00 a.m. and 10:00 a.m. to avoid the confounding effects of circadian rhythm changes on autophagy and retinal metabolism.
Tail bud tissue was processed to genotype Ambra1 +/gt animals. Total genomic DNA was isolated using the NZY Tissue gDNA Isolation kit (Nzytech, MB13503). Polymerase chain reaction (PCR) was performed using the primers 5'-CCCAGTCACGACGTTGTAAAA-3' (primer A) and 5'-TCCCGAAAACCAAAGAAGA-3' (primer B), mapping downstream and upstream of the gene-trap insertion site corresponding to the lacZ reporter sequence. PCR conditions were as follows, for a total of 35 cycles: unfolding temperature, 95°C for 2 min; annealing temperature, 58°C for 54s; elongation temperature, 72°C for 45s. For the SI model, 6.25 mg/mL of sterile sodium iodate (SI, [NaIO 3 ]) solution was freshly prepared from solid SI (Sigma-Aldrich, S4007) diluted in PBS (Sigma-Aldrich, D8537). Mice received a single intraperitoneal injection of 50 mg/kg of SI solution (experimental group) or vehicle (control group) and were sacrificed 1 week after injection.

Electroretinogram recordings
For ERG recordings mice were first allowed to adapt to darkness overnight (o/n), and then subsequent manipulations were performed in dim red light. Mice were anesthetized by i.p. injection with ketamine (95 mg/kg) and xylazine (5 mg/ kg) solution while placed on a heating pad set at 37°C. Pupils were dilated with a drop of 1% tropicamide (Colircusi Tropicamida; Alcon Cusi, 57.050). A drop of 2% Methocel (Hetlingen, 8442) was placed in each eye immediately before placement of the corneal electrode to optimize electrical recording.
Flash-induced ERG responses to light stimuli produced with Ganzfeld stimulator were recorded in the right eye. A photometer was used to measure light intensity at the level of the eye (Mavo Monitor USB). In scotopic conditions, 4-64 consecutive stimuli were averaged with an interval of 10s for dim flashes and 60s for the highest intensity flashes. In photopic conditions, the interval between light flashes was set at 1 s. ERG signals were band-filtered between 0.3 and 1000 Hz and amplified with an amplifier (Grass Instruments, CP511 AC amplifier). A power laboratory data acquisition board (AD Instruments) was used to digitize electrical signals at 20 kHz. An electrode (Burian-Allen electrode; Hansen Ophthalmic Development Laboratory) was fixed on the corneal lens and a reference electrode was placed in the mouth, with a ground electrode attached to the tail to perform bipolar recordings. In dark-adapted conditions, the following responses were recorded: rod responses to light flashes of −2 log cd·s·m −2 ; and mixed responses to light flashes of 1.5 log cd·s·m −2 . White flashes of −1.5 log cd·s·m −2 were used in a recording frequency range of 100-10,000 Hz to isolate the oscillatory potential (OP). In light-adapted conditions, cone-mediated responses to light flashes of 1.5 log cd·s·m −2 were recorded on a rod-saturating background of 30 cd·m −2 . Wave amplitudes of scotopic rod responses (b-rod), OP, mixed responses (a-mixed and b-mixed), and photopic cone responses (b-phot and flicker) were measured off-line by an observer blind to the experimental condition of the animal. Also, the b:a wave ratio was calculated to better understand the electroretinographic responses.

Tissue preparation for imaging
For retina and RPE/choroid flat mounts, eyes were enucleated and briefly washed in ice-cold PBS. The optic nerve, cornea, and lens were gently removed and the resulting posterior eyecup was fixed in freshly prepared ice-cold 4% (w:v) PFA (EMS, 171,010) in PBS for 30 min. Four perpendicular incisions were made and the retina and RPE/choroid were gently separated and fixed for an extra 1.5 h at room temperature (RT). Tissues were kept in 0.01% azide in PBS at 4°C.

Immunofluorescence
Retinal or RPE/choroid samples were permeabilized with 2% or 0.2% (v:v) Triton X-100 (Sigma, T9284) and blocked for The ProteoStat® assay (Enzo, ENZ-51023-KP050) was performed following the manufacturer's instructions. Cryosections were incubated with the dye at 1/500 dilution for 1 h at RT in a wet chamber before DAPI counterstaining. TUNEL assay (DeadEnd™ Fluorometric TUNEL System; Promega, G3250) was used to detect apoptotic cells in cryosections following the manufacturer's directions. Briefly, once the primary antibody was washed, cryosections were incubated for 30 min with TUNEL buffer, after which the TUNEL reaction (1.9% TdT, 9.8% dNTPs, and 88.3% TUNEL buffer) was performed in darkness for 1 h at 37°C. Finally, saline-sodium citrate (SSC) provided in the kit (20X) was diluted to a concentration of 2X and added to cryosections to stop the TUNEL reaction.
Confocal z-stacks were obtained with Leica TCS-SP5-A0BS or Leica TCS SP8 STED 3X microscopes (Leica Microsystems). To assess retinal layer thickness, images of DAPI-stained cryosections were acquired at 40X with a Leica DMI6000B fluorescence microscope coupled to a Leica AF6000 LX multidimensional system (Leica Microsystems).

Image analysis and data quantification
Unless stated otherwise, maximum projections of all z-stacks are displayed in representative images (z-step: 0.5 and 1 µm for RPE and retina, respectively; 1 μm for cryosections). For retinal thickness measurements, each retinal layer was manually measured using the straight-line tool in ImageJ. Quantification of mean fluorescence intensity (MFI) was performed using unprocessed images at maximal projection. Positive cells for a given marker were determined by manual counting plane by plane in a given z-stack. LC3 puncta were quantified using a manually designed Fiji-based plugin, which takes into account the 3D component. Briefly, a fluorescence intensity threshold is set to discriminate positive signal from background. Next, using a minimum voxel size threshold, the 3D objects counter tool is used to detect the number of objects in a 3D confocal z-stack. Finally, the number of puncta obtained is corrected to the thickness of the retina based on the z-stack size.
All confocal images from the same experiment were acquired using the same laser intensity and photomultiplier settings to avoid any variability or bias. Moreover, only nuclear DAPI staining was used to select the field to photograph. At least 3 sections per mice and 4 retinal regions were included in each analysis.
Total protein extract (15 &#x03BC;g) was mixed with 10 mM DTT and bromophenol blue and loaded into Any kD Criterion TGX Precast Gels (567-1124, Bio-Rad). PVDF membranes (170-4157, Bio-Rad) were then activated with 100% methanol and total protein extracts were transferred for 14 min at 25 V using a Trans-Blot Turbo Transfer System (Bio-Rad). Membrane-bound proteins were visualized with Ponceau S (78,376; Sigma) in 5% acetic acid After washing with PBS-T, membranes were incubated with the secondary antibody coupled with horseradish peroxidase (1:2000). Finally, membranes were revealed using the HRP Pierce ECL Western Blotting substrate (Thermo Scientific, 32,106) and film (AGFA, XDAOG) after exposure for varying durations. Band densitometry analysis was performed using Quantity One 1-D Analysis Software (Bio-Rad). To normalize across different gels and to encompass all 3 ages and genotypes, a sample from the ARPE-19 cell line was included in all 3 gels as an internal loading control.

Flow cytometry of dissociated retinas
Retinas were dissected and incubated with 1% (w:v) trypsin (Gibco, 25,300,054) in DMEM for 5 min at 37°C. Each sample was centrifuged at 200 × g for 5 min and pellets were incubated with 40 nM DiOC 6 (Invitrogen, D273) for 15 min at 37° C and 5% CO 2 to assess mitochondrial membrane potential. Finally, dissociated retinas were resuspended in 300 μL of DMEM and analyzed in an XL flow cytometer (Beckman Coulter).

Mass spectrometry and bioinformatics analysis
Metabolomic analyses were conducted using mass spectrometry as previously described [31]. Briefly, retinal tissue (20 mg) was obtained from 2 pools of 4 retinas from 4-and 12month-old Ambra1 +/gt mice and Ambra1 +/+ littermates. For each condition, tissue was weighed and solubilized with 500 μL of cold (−20°C) lysate buffer (MeOH:water:chloroform [9:1:1] with internal standards) in 1.5-mL polypropylene tubes (Precellys lysis tubes; Bertin corp., P000911-LYSK0-A). Tissue was homogenized 3 times for 20s using a Precellys 24 tissue homogenizer (Bertin Technologies). The homogenized tissue was centrifuged (10 min at 15,000 × g and 4°C) and the upper phase of the supernatant was collected. A second round of extraction was performed using 500 μL of lysate buffer, followed by homogenization and centrifugation to collect the supernatant, which was added to the first supernatant. All supernatant collected was then evaporated in microcentrifuge tubes at 40°C in a pneumatically assisted concentrator (Techne DB3). Methanol (100%, 300 μL) was added to the dried extract and the resulting solution split into 2 volumes of 150 μL each. The first part was used for the GC-MS (gas chromatography-mass spectrometry), and the second for LC-MS (liquid chromatography-mass spectrometry) analysis. The GC-MS aliquots were solubilized in methanol and transferred to a glass tube. Next, 50 μL of methoxyamine (20 mg/ mL in pyridine) was added to the dried extracts, which were stored in darkness for 16 h at RT. The following day, 80 μL of N-methyl-N-(trimethylsilyl)trifluoroacetamide/MSTFA was added to perform the final derivatization at 40°C for 30 min. Samples were then transferred to glass vials and directly injected into the GC-MS system. After the second evaporation of the LC-MS aliquots, the LC-MS-dried extracts were solubilized in 50 μL of Milli-Q water, centrifuged (10 min at 15,000 × g at 4°C), transferred to glass vials, and injected into the UHPLC/MS (ultra-high-performance liquid chromatography/mass spectrometry) system. Target analysis was performed as previously described [70]. Manual verifications and QC protocols were followed to select 83 metabolites for GC targeted analysis and 17 metabolites for LC targeted analysis.
After the metabolomic analyses, a peak area proportional to concentration was obtained for each metabolite. Based on these data, metabolite ratios were calculated as the coefficient between arbitrarily chosen metabolites. Principal Component Analysis (PCA) was performed using prcomp function in R. Raw data were processed and analyzed using MultiExperiment Viewer (MEV) software. Skillings-Mack test was performed with significance set at 0.05 to compare metabolite data between Ambra1 +/ gt and Ambra1 +/+ littermate replicates. Next, unsupervised hierarchical clustering (HCL) was performed to cluster significantly different metabolites and construct heatmap representations. Finally, the online MetaboAnalyst 4.0 tool (http://www.metaboa nalyst.ca/) for metabolomic data analysis and interpretation was used to perform enrichment analysis of differentially obtained metabolites between Ambra1 +/gt and Ambra1 +/+ duplicates.
Processed datasets GSE142591 [40], and GSE135092 [47], were downloaded from the GEO database. FPKM (fragments per kilobase of transcript per million mapped reads) values were obtained from the already deposited data. From all detected genes in the available datasets, manually-curated autophagy or glycolysis gene lists were subtracted using R programming. Next, using pheatmap in the R package, heatmaps were generated and hierarchical clustering performed for both samples and genes. Finally, statistical analysis and plotting of selected genes was performed using GraphPad Prism software (GraphPad Software, Inc.).

Statistical analysis
All numeric results are presented as mean values ± SEM. Statistical analyses were performed using GraphPad Prism software (GraphPad Software, Inc.). To test for differences between genotypes, two-tailed Student's t-tests were used. To test for differences between genotypes and age class twoway analyses of variance (ANOVA) were applied and in the case of a significant interaction, differences between genotypes within each age class were assessed using Fisher's least significant difference (LSD) post hoc tests. If assumptions of normality and homoscedasticity were not met, we applied non-parametric tests such as Mann-Whitney U-test for twogroup comparisons and Kruskal-Wallis when more than two groups were compared. Electroretinographic responses were analyzed with general linear models using the lmer function implemented in R. The statistical model contained the electroretinographic response as response variable, genotype and sex as factors, the exact age in number of days as a covariate, the interaction between genotype and number of days, and the experimental group as random factor. Model assumptions were tested. If the normality assumption was violated, transformations were applied and in the presence of heteroscedasticity, weighted least square regressions were performed. For all tests the significance level was set at p < 0.05 (two-tailed). The number of animals used per experiment is stated in the corresponding figure legend.