Light, which carries high energy, is an essential component for living organisms. However, it also generates various reactive oxidative materials that can damage cellular macromolecules, such as nucleic acids, proteins, and lipids (Putting et al., 1992). Light-induced damage is a particularly serious threat for vertebrate photoreceptors, which absorb light photons using the photopigments in their outer segments (OS) and transduce electrochemical signals to inner retinal neurons. A photo-damaged, non-functional photoreceptor OS is quickly eliminated by the retinal pigment epithelium (RPE) through phagocytic ingestion and replaced with a new OS to ensure the proper visual functions of photoreceptors (Garita-Hernandez et al., 2019). However, whereas OS undergo daily renewal, the RPE undergoes relatively minimal regeneration in mammals; over time, the accumulation of damaged photoreceptor components will, therefore, cause the RPE to degenerate. This can lead to retinal degenerative diseases, such as age-related macular degeneration (AMD), which is a leading cause of blindness in aged populations (Kang et al., 2009).

In degenerating RPE, many oxidative materials are accumulated. Lipofuscins, which are yellow-brown pigment granules composed of oxidized lipids, are frequently detectable in the aged RPE (Parish et al., 1998). Similar observations can be made in patients with Stargardt disease, who accumulate bis-retinoid A2E in their photoreceptors and lipofuscins in RPE cells that have phagocytized A2E-containing photoreceptor OS (Radu et al., 2004). Defective retinoid metabolism has also been reported in other retinal degenerative diseases. For example, Leber congenital amaurosis (LCA) is a retinal degenerative disease caused by mutations in lecithin-retinol acyltransferase (LRAT), which converts all-trans-retinol (at-ROL) received from the photoreceptors to all-trans-retinyl ester (at-RE) in the RPE (Saari et al., 1993) (Fig. 1A). The at-RE is then converted to 11-cis-ROL by RPE-specific 65-kDa protein (RPE65) for subsequent oxidation to 11-cis-retinaldehyde (11-cis-RAL), which is the functional photopigment used in human photoreceptors (Jin et al., 2005; Redmond et al., 1998). Mutations of RPE65 have been identified in some patients with LCA and retinitis pigmentosa (RP) (Bainbridge et al., 2008; Morimura et al., 1998). Therefore, at-ROL, which is the substrate of LRAT, and at-RE, which is the substrate of RPE65, accumulate in the RPE of patients carrying mutations in the LRAT and RPE65 genes, respectively, and could cause RPE degeneration (Imanishi et al., 2004; Sahu and Maeda, 2016).

Retinoids also contribute to many developmental and homeostatic events; for this purpose, they are converted to retinoic acid (RA), which binds its receptor to induce the expression of many important regulatory genes directly (Leid et al., 1992; Niederreither and Dolle, 2008). Thus, insufficient production of RA results in many health problems, including malformation of organs, growth of neoplastic tissue, and disturbance of the immune system (Cunningham and Duester, 2015).

Animals should take up ROL in their diet as vitamin A (VitA) or carotenoids like β-carotene, which can be broken into two RAL molecules for subsequent conversion to ROL by β-carotene dioxygenase 1 (BCO1) (Ferreira et al., 2020; von Lintig and Vogt, 2000). Insufficient intake of VitA and/or carotenoids therefore causes various health problems, including night blindness, xerophthalmia, hyperkeratosis, increased susceptibility to severe infection, and so on (Maden, 2002; Sommer, 2008). However, excessive ROL is also harmful to cells (DiPalma and Ritchie, 1977); thus, an optimal intake is needed.

In healthy cells, enzymes that convert ROL to RAL and then to RA are expressed to avoid of overaccumulation of ROL and RAL derivatives. The conversion of RAL to RA is carried out by retinaldehyde dehydrogenases (RALDHs), while the conversion of ROL to RAL is mediated by retinol dehydrogenases (RDHs) (Ferreira et al., 2020; Liden and Eriksson, 2006). These enzymes are especially high in hepatocytes and cells of the RPE, which actively metabolize ROL and RAL. Retinoid metabolism in the latter cell type plays a unique role in the visual system, whereas that in the former cell type plays a more general role in detoxification (Pares et al., 2008).

Mice

Pten-flox mice, in which two loxP sequences are inserted into the intron 3 and 5 of Pten gene, were bred with TRP1-Cre mice, which express Cre recombinase under the promoter sequences of human tyrosinase related protein 1 (TRP1), to generate Pten^flox/flox;TRP1-Cre (Pten-cko) mice as it was described previously (Kim et al., 2008). Mice were fed with the VitA-enriched (ROL(+)) chow (ENVIGO, 2018S), the VitA-deficient (ROL(–)) chow (ENVIGO, TD.86143), or the VitA-deficient;β-carotene (100 mg/kg) chow after weaning at post-natal day 21 (P21). To supply RA, the mice were injected at-RA (Sigma, USA), which was dissolved in DMSO and mixed in corn oil (20 mg at-RA/kg body weight), once in every other day beginning at P21 and continuing for 1 month. All experiments using the mice were carried out according to the guidance of Institutional Animal Care and Use Committee (IACUC) of Korea Advanced Institute of Science and Technology (KAIST) (No. KA2011-37). All mice used in this study were maintained in a specific pathogen-free facility of KAIST Laboratory Animal Resource Center.

Western blot (WB) analysis

Mouse RPE and retina were isolated from the mice as it was described previously (Kim et al., 2008). The cell lysates (70 µg proteins) were analyzed by SDS-PAGE prior to the transfer onto polyvinylidene difluoride (PVDF) membrane for subsequent WB. Corresponding proteins on the PVDF membranes were detected by anti-ADH1 (1:500; Abcam, UK), anti-ADH7 (1:500; Abcam), anti-RDH5 (1:500; Abcam), anti-RDH10 (1:500; MyBioSource, USA), anti-PTEN (1:500; Cell Signaling Technology, USA), anti-RPE65 (1:500; Abcam), anti-FOXO1 (1:500; Cell Signaling Technology), anti-FOXO3a (1:500; Cell Signaling Technology), anti-FOXO4 (1:500; Cell Signaling Technology), anti-pFOXO1(Thr24)/pFOXO3a(Thr32) (1:500; Cell Signaling Technology), anti-pAKT(T308) (1:500; Cell Signaling Technology), and anti-β-actin (1:2,000; Santa Cruz Biotechnology, USA). The antibodies bound to the proteins on the blots were visualized by the incubation in SuperSignal™ West Pico Chemiluminescent Substrate or SuperSignal™ West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific, USA) after incubating with secondary antibodies, which were conjugated with horseradish peroxidase (HRP) and recognize the primary antibodies listed above.

Measurement of ROL and RA concentration

Mouse RPE and retina were collected from 20 mouse eyes and were immediately placed in brown tubes and immersed in liquid nitrogen to minimize the degeneration and modification of ROL and RA. After homogenization in ice-cold phosphate-buffered saline (PBS; pH 7.2) using Dounce homogenizer, the cell suspension was subjected to sonication prior to the centrifugation at 1,500g for 15 min. Concentrations of ROL and RA were measured by the Mouse Vitamin A ELISA Kit (MBS722364 [MyBioSource, USA]; mainly detect ROL) and Mouse Retinoic Acid ELISA Kit (MBS706971; MyBioSource) following the manufacturer’s instruction.

Cell culture and luciferase assay

Human embryonic kidney 293T were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS; Gibco, USA) and were transfected with pGL3 luciferase reporter plasmids by polyethylenimine transfection method. The cells were also co-transfected with pSV-β-galactosidase plasmid in the same amount of pGL3 reporter plasmids. The cell extracts were prepared after 24 h and assessed for luciferase activity followed by the normalization with β-galactosidase activity values.

Eye section and immunohistochemistry

After anaesthetizing by intraperitoneal injection of tribromoethanol (Avertin; Sigma), the mice were perfused with PBS (pH 7.5) containing 0.1% heparin (Millipore, USA) and then with 4% paraformaldehyde (PFA; Sigma) in PBS. The eyes were isolated from the mice and fixed further in 4% PFA/PBS solution at 4°C for 2 h. The samples were moved into 20% sucrose/PBS solution overnight at 4°C prior to embedding in the TissueTek OCT compound (Sakura Finetek, USA). The frozen eyes were then cryosectioned on the Superfrost glass slides (14 µm thickness) using Leica cryostat (CM1860; Leica, Germany).

For immunohistochemistry, the sections were blocked with 5% normal donkey serum and 5% normal goat serum in PBS containing 0.2% Triton X-100 before incubating with primary antibodies at room temperature for 2 h. Mouse monoclonal antibodies against Ezrin (1:200; Invitrogen, USA) and Rhodopsin (1:200; Millipore); rabbit polyclonal antibodies against FOXO3a (1:200; Cell Signaling Technology), GFAP (1:200; Abcam), Otx2 (1:200; Abcam), GS (1:200; Sigma), and β-catenin (1:200; Cell Signaling Technology); goat polyclonal antibodies against Sox2 (1:200; Santa Cruz Biotechnology) were incubated at 4°C for overnight. Fluorescent images were obtained with a confocal microscope (Fluoview FV1000 and FV3000; Olympus, Japan) after staining the sections with Cy3 and Alexa488-conjugated secondary antibodies at room temperature for 1 h.

Chromatin immunoprecipitation (ChIP)

Mouse RPE and retina were isolated from P60 mice and protein-DNA complexes were immunoprecipitated with rabbit preimmune IgG or rabbit anti-FOXO3a IgG in the procedures reported previously (Moon et al., 2018). Quantitative polymerase chain reaction (qPCR) was performed with the immunoprecipitated DNA using SYBR Green (Applied Biosystems, USA) with designed primers to upstream region containing FOXO binding sequences of Adh1 (forward 5’-CCCAACAGCACTCGTCCA-3’, reverse 5’-CTTCCCTCCTCTAGTGAACCG-3’), Adh7 (forward 5’-CCCATCTCCATCGCCATTACT-3’, reverse 5’-ATCCGGTCTCCGTTCTTAGC-3’), and Rdh10 (forward 5’-CTAACTCGCGGCTGTCACC-3’, reverse 5’-CGATCCGGACGCGAAGG-3’).

Identification of evolutionary conserved regions (ECRs)

The analysis of ECRs in genomes of various species was performed in the ECR browser website (https://ecrbrowser.dcode.org/ecrInstructions/ecrInstructions.html). A base genome was mouse and parameter set in default value (minimum length of ECRs, 100 bp; identity of ECRs, 70%). Possible consensus sequences site recognized by FOXO were marked with yellow stars: 5′-GTAAA(T/C)AA-3′, known as the Daf-16 family member-binding element (DBE) and 5′-(C/A)(A/C)AAA(C/T)AA-3′, known as the insulin-responsive sequence (IRE). All FOXO-family members recognize the core sequence 5’-(A/C)AA(C/T)A-3’ (Obsil and Obsilova, 2011).

Statistical analysis

Statistical analysis was performed by Prism 9 software (GraphPad, USA). Data from statistical analysis are presented as the mean ± SD. Student’s t-test was used to determine the significant difference between two groups. P values were calculated using a two-tailed unpaired t-test. P value < 0.05 was considered statistically significant.

Retinoid metabolism enzyme expression is reduced in Pten-deficient mouse RPE

We previously reported that the mouse RPE is degenerated upon genetic and physiological inactivation of Pten (Kim et al., 2008; Lee et al., 2011), which is a negative regulator of the phosphoinositide 3-kinase (PI3K) and Akt pathway (Stambolic et al., 1998). The genes whose mRNA levels exhibited significant decreases in Pten-deficient mouse RPE included those encoding both classical retinol dehydrogenases (Rdh5 and Rdh10) and non-classical retinol dehydrogenases (short chain alcohol dehydrogenases, Adh1 and Adh7) (Kim et al., 2008). Here, to experimentally validate these changes, we compared the mRNA and protein levels of these enzymes in P21 TRP1-Cre and Pten-cko littermate mouse RPE by real-time PCR (RT-PCR) and WB, respectively (see the information of the mice in Materials and Methods section). We found that the mRNA and protein levels of Adh and Rdh in Pten-cko mouse RPE were significantly lower than those in TRP1-Cre littermate mouse RPE (Figs. 1A-1C). We also found that the level of Lrat, which converts at-ROL obtained from the photoreceptor OS to at-RE in the RPE (Saari and Bredberg, 1989), was also decreased in the Pten-cko mouse RPE, whereas the expression of Rpe65, which is responsible for converting at-RE to 11-cis-ROL in the RPE (Jin et al., 2005), was not changed significantly (Figs. 1A-1C). The results suggest that the expression of those retinoid metabolism-related enzymes in mouse RPE is regulated by a mechanism that is sensitive to Pten.

As we hypothesized that the altered expression levels of those Rdh genes would change the contents of retinoids in the RPE, we examined the levels of ROL and RA in the mouse RPE. Consistent with the decreased expression of Rdh and Lrat, the level of ROL, a precursor of RAL and RE, was significantly increased in P21 Pten-deficient RPE compared to TRP1-Cre littermate mouse RPE (Fig. 1D). On the contrary, the level of RA, a metabolic product of RAL, was decreased in the Pten-deficient mouse RPE (Fig. 1E). These results suggest that the metabolic flow of ROL-RAL-RA is ineffective in the Pten-deficient mouse RPE due to the reduced Rdh expression.

Degeneration of Pten-deficient mouse RPE does not arise from a decrease of RA

Given the importance of RA-dependent gene expression for cell survival (Michalik and Wahli, 2007), we speculated that a decrease of RA could be correlated with the degeneration of the Pten-deficient RPE. We, thus, tested this by increasing RA in the mouse RPE. In practical, we supplemented TRP1-Cre and Pten-cko littermate mice with at-RA via intraperitoneal injection once in every other day beginning at P21 and continuing for 1 month, and investigated whether this protected the RPE from degeneration in Pten-cko mouse eyes. However, we found that the numbers of Otx2-positive RPE cells in P50 at-RA-injected Pten-cko mouse eyes were not significantly different from those observed in the vehicle (i.e., DMSO)-injected groups (Figs. 2A and 2B). Instead, we found that the numbers of rhodopsin-positive photoreceptors in the outer nuclear layer (ONL) were increased in Pten-cko mouse retinas without any significant change in the numbers of the cells in the inner nuclear layer (INL) (Figs. 2A and 2B). These results suggest that ectopic at-RA does not act as a protective factor for the RPE; instead, consistent with previous reports (Kelley et al., 1994), this treatment appears to protect photoreceptors.

Unexpectedly, we found that the ONL structures of at-RA-injected mice were disrupted and formed multiple rosettes regardless of genotype (Fig. 2A). These retinal rosettes have been suggested to arise from the degeneration of Müller glia (MG) (Willbold et al., 2000). Supporting this, Sox2-positive MG cell bodies failed to align properly the INL of rosette areas in at-RA-injected TRP1-Cre mouse retinas (Fig. 2A, arrowheads), whereas they formed a sharp layer in the middle INL of DMSO-injected retinas. We also found that the RPE of at-RA-injected TRP1-Cre mice failed to exhibit polarized features, which are necessary for the ability of the RPE to interact with photoreceptor OS and contribute to maintaining the retina-blood barrier function of the RPE (Finnemann and Chang, 2008). β-Catenin, which was enriched in the basolateral sides of the RPE from DMSO-injected TRP1-Cre mice, was detected throughout the RPE of at-RA-injected TRP1-Cre mice with a distribution similar to that seen in Pten-cko mouse RPE (Fig. 2C). Ezrin was detected in the apical side of RPE in TRP1-Cre mice injected with at-RA, but it failed to show the well-organized brush-like distribution pattern seen in DMSO-injected TRP1-Cre mouse RPE (Fig. 2C).

Protection of Pten-deficient mouse RPE by feeding of a ROL-free diet

As previous work showed that high-dose ROL is toxic to cells (DiPalma and Ritchie, 1977), we next investigated whether increase of ROL was correlated with the degeneration of Pten-deficient mouse RPE. To address this question, we fed mice ad libitum from P21 with a chow lacking VitA, which is a natural source of ROL. In contrast to the severely reduced RPE and photoreceptor cell numbers in P50 Pten-cko mice fed a ROL-containing normal diet, these numbers in P50 Pten-cko mice fed the ROL-free diet were rescued to levels comparable to those observed in P50 TRP1-Cre littermate mice (Fig. 3A, Supplementary Fig. S1A). Moreover, the polarized distributions of β-catenin and ezrin to the basolateral and apical sides (Le et al., 2021), respectively, were recovered in the RPE of Pten-deficient mice fed the ROL-free diet (Figs. 3A and 3B). These results suggest that a ROL-free diet alleviates the RPE degeneration in Pten-cko mice from P21 to P50.

However, the rescued RPE and retinal phenotypes were not maintained in Pten-cko mice fed the ROL-free diet for an additional 30 days (to P80) (Figs. 3C and 3D, Supplementary Fig. S1B). Dietary supplement of β-carotene, which can be broken into two RAL molecules for further conversion to ROL and RA (Thompson et al., 2019), was necessary to maintain RPE and retinas intact in Pten-cko mice fed a ROL-free chow by P80 (Figs. 3C and 3D, Supplementary Fig. S1B). Together, these results suggest that an elevation of ROL in Pten-deficient RPE is likely to be a cause of the observed degeneration.

FoxO transcription factors regulate Adh and Rdh expression in the mouse RPE

To begin identifying the transcription factors responsible for regulating the expression levels of Adh and Rdh genes exhibited decreased expression the Pten-deficient mouse RPE (Figs. 1A-1C), we compared the DNA sequences of various vertebrate ADH1, ADH7, and RDH10 genes. We identified conserved DNA sequences located upstream of the transcription initiation sites of these genes (Fig. 4A, yellow highlights). We next isolated the DNA sequences upstream of mouse Adh1, Adh7, and Rdh10 genes, cloned them into the pGL3 luciferase reporter vector, and assessed for potential promoter and enhancer activities of the sequences. All of the tested upstream sequences induced meaningful luciferase expression in human embryonic kidney 293T cells (Supplementary Fig. S2A). Given that the Adh1, Adh7, and Rdh10 mRNA were down-regulated in the Pten-deficient RPE, which exhibited hyperactive Akt (Figs. 1A and 1B), we next tested whether the luciferase expression driven by the upstream sequences was sensitive to Akt. Indeed, we found the luciferase expression was significantly decreased by co-expression of constitutively active (CA) AKT1 (AKT1(CA)) but not kinase-inactive AKT1^K179M (AKT1-KM) (Fig. 4B). These results suggest that the expression levels of the tested Rdh genes are induced by transcription factors that are negatively regulated by Akt.

Further analyses of the DNA sequences revealed that they commonly included consensus binding sequences for forkhead box class O (FOXO) transcription factors (Fig. 4A, Supplementary Fig. S2B), which are known to be negatively regulated by Akt-mediated phosphorylation (Brunet et al., 2001). Thus, we performed chromatin immunoprecipitation (ChIP) with an anti-FOXO3a antibody to examine whether FoxO transcription factors could bind the putative forkhead response elements (FHRE) in the upstream sequences of Adh1, Adh7, and Rdh10 in the mouse RPE. The obtained results suggested that FoxO3a bound to the FHRE sequences upstream of the analyzed genes in the RPE with an efficiency comparable to that of the well-characterized FHRE of growth arrest and DNA-damage-inducible-45 α (Gadd45a) (Tran et al., 2002) (Fig. 4C).

Next, we tested whether luciferase expression driven by the upstream sequences could be positively or negatively regulated by FOXO3A. We found that luciferase expression was elevated by co-expression of human FOXO3A (Fig. 4D), and further observed that co-expression of AKT1-CA abolished this effect in the cells expressing FOXO3A but not FOXO3A (T32M) (FOXO3-TM), which is resistant to phosphorylation by AKT (Fig. 4D). These results suggest that FOXO3A positively regulates the expression levels of the analyzed genes and AKT antagonizes this effect.

We found that the mouse RPE expresses all three FoxO isotypes: FoxO1, FoxO3a, and FoxO4 (Fig. 4E). The levels of phosphorylated FoxO1 and FoxO3a were significantly elevated in P21 Pten-deficient mouse RPE in comparison to TRP1-Cre littermate mouse RPE (Fig. 4E), suggesting that FoxO1 and FoxO3a undergo Akt-mediated phosphorylational inactivation in the Pten-deficient mouse RPE. Furthermore, the cytoplasmic retention of FoxO3a, which is induced by Akt-mediated phosphorylation (Brunet et al., 1999), was enhanced in the Pten-deficient mouse RPE (Fig. 4F). These results suggest that hyperactive Akt in the Pten-deficient RPE might inhibit the expression of Adh and Rdh genes via the phosphorylation and remobilization of FoxO transcription factors from the nucleus to the cytoplasm. This could lead to the toxic accumulation of ROL in the RPE, which could be rescued by dietary replacement of VitA and β-carotene (Fig. 5).

Given that the mammalian visual system uses 11-cis-RAL as a functional photopigment, proper retinoid metabolism in the photoreceptor and RPE is necessary for proper vision. In support of this, many forms of congenital blindness and early onset retinal degenerative disease, including LCA, Stargardt disease, cone-rod dystrophy, and RP, are associated with mutations in retinoid metabolism genes (Morimura et al., 1998). Among the enzymes involved in retinoid metabolism, LRAT (which converts at-ROL to at-RE) and RPE65 (which isomerizes at-ROL to 11-cis-ROL) are expressed exclusively in the RPE, suggesting that the RPE plays crucial roles in retinoid metabolism (Redmond et al., 1998; Saari et al., 1993). Consequently, the substrates of those enzymes (at-ROL and at-RE) are accumulated in the RPE of Lrat- and Rpe65-deficient mice, respectively (Imanishi et al., 2004). Dietary supply of at-ROL to Lrat-deficient mice, therefore, should elevate at-ROL in the mouse RPE to a toxic level. Fortunately, at-ROL can be metabolized to at-RAL by RDHs, which are also enriched in the RPE (Figs. 1A and 1B). This enables the RPE to persist in mice lacking Lrat (Ruiz et al., 2007). Therefore, combinatorial mutations of retinoid metabolic enzymes expressed in the RPE might be needed to induce RPE degeneration and the associated retinal degenerative diseases.

Here, we suggest that Pten deficiency elevates ROL via the reduced expression of Rdh genes in mouse RPE, and that this appears to trigger degeneration of the RPE (Figs. 1D and 3A). Excessive ROL was reported to activate monoamine oxidase enzyme (MAO) and subsequently increase the level of hydrogen peroxide (H₂O₂) to disrupt mitochondrial redox maintenance (de Oliveira et al., 2012; Richter et al., 1995). Thus, the elevation of ROL in the Pten-deficient mouse RPE might also induce mitochondrial dysfunction, which was shown to induce RPE degeneration (Brown et al., 2019). However, additional studies are needed to more precisely define the mechanisms of ROL toxicity in the Pten-deficient mouse RPE.

Our results indicate that dietary restriction of ROL can delay the degeneration of the Pten-deficient mouse RPE in the short term (Figs. 3A and 3B). However, the RPE did not persist under longer-term feeding of the ROL-free diet (Figs. 3C and 3D), suggesting that an alternative source of ROL should be supplied. ROL and RA can be produced from β-carotene-derived RAL by RDH and RALDH, respectively, and could be utilized by mice in the absence of a dietary supply of ROL (Kim et al., 2011). Thus, we supplied β-carotene to Pten-cko mice being fed the ROL-free chow, and found that this protected the RPE over the longer term (Figs. 3C and 3D).

Dietary supplementation of β-carotene (β,β-carotene) and other carotenoids, such as lutein/zeaxanthin (β,ε-carotene-3,3'-diol) and lycopene (ψ,ψ-carotene), has long been recommended to support healthy vision (Krinsky and Johnson, 2005). β-Carotene is also known to have anti-oxidative, anti-cancer, and anti-obesity activities (Nagao, 2009). However, excessive intake of β-carotene can cause other health problems, including a yellowish discoloration of the skin (Hess and Myers, 1919). Interventional trials also found that high doses of β-carotene could have negative effects on the incidence of lung cancer and cardiovascular diseases in smokers (Alpha-Tocopherol, Beta Carotene Cancer Prevention Study Group, 1994; Omenn et al., 1996). Thus, it should be provided at an optimal concentration.

Previous reports indicated that FOXO1 regulates Rdh1, Rdh8, and Rdh10 but the physiological relevance of FOXO-dependent Rdh gene expression remained unknown. Here, we show that FoxO3a binds to well-conserved FOXO-binding sequences upstream of Adh1, Adh7, and Rdh10 to induce the expression of these transcripts in the mouse RPE (Fig. 4C). We further identified that FoxO-dependent regulation of Rdh gene expression is sensitive to Pten-Akt signaling (Figs. 4D-4G). Similar to the pathogenic output of Pten inactivation in the mouse RPE (Kim et al., 2008; Lee et al., 2011), ROL toxicity could account for the degeneration of other tissues in which FoxO activity is reduced by Akt or other negative regulators (Fig. 5). Additional studies are therefore needed to investigate whether FoxO transcription factors are major regulators of Rdh expression in vivo.

This work was supported by the National Research Foundation of Korea (NRF) grants funded by Korean Ministry of Science and ICT (MSIT) (2017R1A2B3002862 and 2018R1A5A1024261; J.W.K.); the grant funded by Samsung Foundation of Science and Technology (SSTF-BA1802-10; J.W.K.).

Y.J.K. conceived and performed experiments, and wrote the manuscript. S.P., T.H., S.K., S.L., and H.Y. performed experiments. J.W.K. conceived and supervised the experiments, wrote the manuscript, and secured funding.

The authors have no potential conflicts of interest to disclose.