Loss of Extracellular Signal-Regulated Kinase 1/2 in the Retinal Pigment Epithelium Leads to RPE65 Decrease and Retinal Degeneration

ABSTRACT Recent work suggested that the activity of extracellular signal-regulated kinase 1/2 (ERK1/2) is increased in the retinal pigment epithelium (RPE) of age-related macular degeneration (ARMD) patients and therefore could be an attractive therapeutic target. Notably, ERK1/2 pathway inhibitors are used in cancer therapy, with severe and noncharacterized ocular side effects. To decipher the role of ERK1/2 in RPE cells, we conditionally disrupted the Erk1 and Erk2 genes in mouse RPE. The loss of ERK1/2 activity resulted in a significant decrease in the level of RPE65 expression, a decrease in ocular retinoid levels concomitant with low visual function, and a rapid disorganization of RPE cells, ultimately leading to retinal degeneration. Our results identify the ERK1/2 pathway as a direct regulator of the visual cycle and a critical component of the viability of RPE and photoreceptor cells. Moreover, our results caution about the need for a very fine adjustment of kinase inhibition in cancer or ARMD treatment in order to avoid ocular side effects.

by an analysis of RPE genomic DNA, as evidenced by the presence of the delta (Δ) band (Fig. 1D). Immunostaining showed a decreased ERK2 level in the RPE of RPE-DKO mice at 1 month in comparison to the level in control mice (Fig. 1E). In addition, Western blot analysis of RPE protein lysates clearly showed the absence of ERK1 in both RPE-DKO and CTL mice and a decreased ERK2 level only in RPE-DKO mice (Fig. 1F). Fundus analysis showed an apparent depigmentation of RPE-DKO eyes ( Fig. 2A). OCT analyses performed on RPE-DKO mice at 4 months revealed the degeneration of the retina with a significant reduction of all retinal and choroidal layers (Fig. 2B). In accordance with this result, ERG analyses revealed a significant impairment of vision in RPE-DKO mice at 2 months, as revealed by the lower amplitude of the b wave in both the scotopic and photopic stimulations (Fig. 2C). Scotopic patterns (50 mcd s/m 2 ) demonstrated the absence or severe reduction of a and b waves in RPE-DKO mice in comparison to CTL mice, while photopic patterns (10 mcd s/m 2 ) showed an absence of stimulation in RPE-DKO mice (Fig. 2C, right). This clearly indicates impairments in both the cone and rod photoreceptor responses to light stimuli in RPE-DKO animals.
The severe vision damage and retinal degeneration observed in RPE-DKO mice led us to measure the relative amounts of visual cycle intermediates. Mice that had been dark adapted for 12 h were enucleated under dim light. The retina was separated from the choroid and subjected to high-performance liquid chromatography (HPLC) analyses as previously described (32). The loss of ERK1/2 in the RPE led to a significant decrease in the level of ocular retinoids. Levels of retinyl esters in the choroid (dissected to include RPE cells) were significantly reduced. In the retinas of dark-adapted eyes, decreases of more than half of the amounts of 11-cis-retinal and all-trans-retinal were observed (Fig. 2D). We further intended to measure how light influences the ocular retinoid composition of these mice. Thus, we exposed the mice to bright light, where the turnover of the visual cycle is enhanced. As expected, we observed increases in the amounts of all-trans-retinol and retinyl ester(s) after light treatment. Notably, the levels of both retinoids were significantly lower in RPE-DKO mice than in controls, indicating an impairment of the visual cycle in both dark-adapted and light-exposed mice (Fig.  2D). We also noted, by immunostaining analysis, a decrease in the expression level of the stimulated by retinoic acid 6 (STRA6) protein, the membrane receptor for retinol binding protein found in the RPE in RPE-DKO mice at 1 month (data not shown). This decrease suggests that a reduction of retinoid uptake from the circulation could contribute to the diminished ocular retinoid levels in RPE-DKO mice.
RPE-specific loss of ERK1/2 causes significant thinning of the ONL/INL. Visual impairment along with retinal dysfunction led us to examine the photoreceptor structure in RPE-DKO mice. Staining of various cone and rod markers was performed to analyze the integrity of the distinct retinal layers. Retinal degeneration was observed only in RPE-DKO mice. As controls, we used ERK1 Ϫ/Ϫ ; Erk2 f/f mice treated with PBS (CTL), ERK1 Ϫ/Ϫ ; Erk2 f/f mice treated with Dox, and VMD2-rtTA/TRE-Cre; ERK1 Ϫ/Ϫ ; Erk2 f/f mice treated with PBS; none of these mice showed retinal degeneration (data not shown). No toxic effect of Dox was observed in CTL (data not shown) or in Cre-tdTomato (Fig. 1A) mice. RPE-DKO mice exhibited reductions in the levels of cone markers as well as in the outer nuclear layer (ONL) and the inner nuclear layer (INL) thicknesses. Levels of cone arrestin and GNAT2 showed marked decreases in RPE-DKO mice at 2 months, and this effect was more pronounced at 4 months (Fig. 3A). The important loss of cone markers occurred before the disappearance of the respective rod markers, as shown by immunostaining of GNAT2 ( Fig. 3A and B). The levels of the rod markers rhodopsin and GNAT1 were slightly decreased in RPE-DKO mice, principally in the ONL. In addition, the outer segment (OS) length was already smaller in RPE-DKO mice at 2 months (Fig. 3A). Retinal degeneration could explain the absence of a marked reduction of the GNAT1 protein expression level shown by Western blot analysis, because both GNAT1 and tubulin levels were decreased, so the ratio was not changed. Measurement of the thicknesses of the ONL and INL confirmed clear decreases in both cases, underlining the devastating effect of the loss of ERK1/2 on RPE cells (Fig. 3C). In  [RAL]) measured in the retina and in the RPE-choroid from mice at 1 month, which were dark adapted for 16 h, or measured in whole eyes (containing the retina, RPE, and choroid) from mice (same conditions) exposed to 1,000 mA of light for 2 h. fact, ONL and INL thicknesses were reduced by 50% after only 2 months of ERK1/2 loss and by up to 70% after 4 months (Fig. 3C). We found staining of both GNAT1 and CONE-ARRESTIN in the RPE of RPE-DKO mice at 1 month, while no staining was observed in the RPE of CTL mice, as expected (Fig. 3D); this information supports the hypothesis of a defect in the phagocytosis process in these mice. Interestingly, ONL and INL degenerations correlated with the level of cell death, as demonstrated by terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) assays performed on RPE-DKO and CTL mice (Fig. 3E). The decreases of the ONL and OS thicknesses as well as the structural disorganization of the entire retina were confirmed by electron microscopy (EM) imaging, revealing the disruption of the inner segment (IS) and OS of photoreceptor cells (Fig. 4A). This disruption and loss of the OS in particular were quantified by measuring the ONL, OS, and IS lengths (Fig. 4B). A follow-up study of the mice for 1 year showed the disappearance of the ONL and a drastic thinning of the INL, with no staining for cones or rods. The only persistent staining was for glial fibrillary acidic protein (GFAP) in the ganglion cell layer (GCL) and glycine in the inner plexiform layer (IPL) (Fig. 4C).
RPE-specific loss of ERK1/2 causes specific reductions in levels of cone markers. Additional biochemical analyses were carried out in order to verify the loss of cone photoreceptors. Western blot analysis of retinas from CTL and RPE-DKO mice showed that the level of GNAT2, a specific cone marker, was significantly reduced at 2 months and was reduced by more than 60% at 4 months (Fig. 5A). However, at 4 months, we observed a marked decrease in both ONL and INL thicknesses, which could not be due only to cone photoreceptor loss. Indeed, at 2 months, the level of GNAT1 was already slightly affected, while at 4 months, the difference was visible even if not significant. At 4 months, the decreases in the levels of photoreceptor proteins were mirrored by reductions in their mRNA levels. Quantitative PCR (qPCR) showed that the levels of specific cone markers, including Gnat2, ConeArrestin, mOpsin, and sOpsin, were significantly decreased at 2 months and were decreased further by 60 to 80% at 4 months (Fig. 5B). In agreement with data from Western blot studies, the levels of the specific rod markers Gnat1 and Rho showed no difference at 2 months but declined at 4 months, although the reduction was not statistically significant. Retinal degeneration could explain the absence of a significant reduction in the level of GNAT1 protein expression shown by Western blotting, because both GNAT1 and TUBULIN levels were decreased, so the ratio was not changed.
Loss of ERK1/2 in the RPE causes RPE ultrastructural damage. The loss of vision as well as the pronounced retinal degeneration in these mice prompted us to examine the state of the RPE ultrastructure in RPE-DKO mice at 1 month by EM imaging. An overview of the RPE-retina cell layers showed massive shrinkage of the RPE cell layer, a disruption of Bruch's membrane (BM), and PR degeneration (Fig. 6A). In order to characterize the damage to the RPE-retinal layer in more detail, different layers were observed separately and quantified. The BM ultrastructure was clearly disrupted, and the thickness of the membrane was significantly reduced (Fig. 6B). The RPE cell layer exhibited massive shrinkage, as evidenced by the significant reduction of its length (Fig.  6C). The ultrastructural damage to the RPE cell layer was accompanied by the accumulation of mitochondria at the basolateral layer membrane, as demonstrated by the significant increase of the organelle area when normalized to the area of the cytosol (Fig. 6C). The RPE cell layer of RPE-DKO mice contained structures enriched in membranes that resembled phagolysosomes of the OS of the PR (Fig. 6D). This indicates that phagocytosis of OS, one of the major functions of RPE cells, is impaired in RPE-DKO mice. Confirmation was obtained by the presence of GNAT1 and CONE-ARRESTIN in the RPE of RPE-DKO mice at 1 month (Fig. 3D). Ultrastructural analysis demonstrated that the RPE-specific loss of ERK1/2 caused massive alterations to the ultrastructures of not only the RPE cell layer but also the BM and the retinal cell layer. The ultrastructural RPE damage in RPE-DKO mice was confirmed by flat-mount RPE-choroid analysis. Phalloidin staining revealed a disorganized RPE structure along with an increased RPE cell size, including multinucleated cells (Fig. 6E). A total loss of RPE cells was observed in parts of the retinas of RPE-DKO mice at 2 months (Fig. 6F). Quantification graphs for qPCR performed on retinal extracts from CTL and RPE-DKO mice at 1, 2, or 4 months. RL8 was used as an internal control to normalize RNA expression. Results are expressed as a percentage of the value for CTL mice and as means Ϯ standard errors of the means of results from 4 independent experiments (*, P Ͻ 0.05; **, P Ͻ 0.003).

ERK1/2 directly regulates RPE65 levels.
We previously showed that the inhibition of the ERK1/2 pathway in ARPE19 cells as well as in isolated mouse RPE cells leads to a reduction of Rpe65 mRNA expression (9). This decrease was confirmed by RPE65 immunostaining in RPE-DKO mice at 1 month and at 2 weeks, respectively ( Fig. 7A and  B). In addition, we observed a mislocalization of LRAT, the enzyme upstream of RPE65  in the visual cycle, to the apical side of RPE cells in RPE-DKO mice, instead of the basolateral side as observed in CTL mice ( Fig. 7A and B). qPCR analysis confirmed the results of immunostaining and revealed that RPE65 but not LRAT expression was significantly decreased in RPE cells of RPE-DKO mice at 1 month (Fig. 7C). The RPE65 protein expression level, normalized to the TUBULIN expression level, was significantly reduced in RPE-DKO mice at 1 month (Fig. 7D). Notably, the promoter of RPE65 has been shown to contain an AP-1 site, which is under the transcriptional control of the AP-1 family of proteins (33). We thus carried out an in vitro luciferase reporter assay using RPE65 (33) or LRAT (34) promoter constructs. Luciferase activity was detected with both promoters transfected into the HEK293 cell line (Fig. 7E). Treatment with the MEK1/2 inhibitor U0126 caused a significant decrease of the activity of the RPE65 promoter, while no change in activity was observed when the luciferase reporter was driven by the LRAT promoter. The addition of U0126 did not significantly affect luciferase activity when the AP-1 site in the RPE65 promoter reporter construct was deleted, indicating that it is mandatory for regulation by ERK1/2. Next, we set out to identify the downstream effectors of ERK1/2 that regulate this AP-1 site. Nuclear proteins of ARPE19 cells showed strong binding to the AP-1 sequence, which was reduced in the presence of a competitor AP-1 oligonucleotide as well as by treatment with U0126 (Fig. 7F). In order to identify which factors regulated by ERK1/2 are part of the AP-1 complex, nuclear extracts from ARPE19 cells were subjected to an AP-1 binding enzyme-linked immunosorbent assay (ELISA) that included individual AP-1 family transcription factors. The inhibition of the ERK1/2 pathway in ARPE19 cells, using either U0126 or PD0325901, led to decreased C-FOS and FRA-1 protein binding to the AP-1 sequence (Fig. 7F), whereas the inhibitors had no effect on c-JUN, JUN-B, and JUN-D binding (data not shown). In order to investigate whether this effect is maintained in RPE-DKO mice, whole-cell protein lysates from isolated RPE cells were subjected to an AP-1 DNA binding assay. Indeed, the binding of C-FOS and FRA-1 to AP-1 was significantly reduced (Fig. 7G), and C-FOS and FRA-1 protein expression levels were decreased in the RPE of RPE-DKO mice at 1 month (Fig. 7G). Taken together, data from analyses of cell cultures and the mouse model argue in favor of a direct regulation of RPE65 gene expression by the ERK1/2 pathway via the binding of C-FOS and FRA-1 complexes to the AP-1 site in the upstream promoter region.

DISCUSSION
In order to evaluate the potential use of ERK1/2 pathway inhibitors in ARMD treatment and to better understand the ocular side effects observed in cancer patients treated with such inhibitors, we disrupted ERK1/2 expression specifically in RPE cells of mice. We observed two major consequences. The first consequence is the downregulation of RPE65, the mislocalization of LRAT, and the depletion of the retinoid content. This already occurs 1 month after ERK1/2 depletion, when retinal degeneration is just beginning to be detected, as shown by chromatin compaction in ONL cells as well as the shortening of the photoreceptor OS. The second consequence includes changes in the morphology of the RPE and in the RPE cell death process, which appears later, 2 months after Cre induction, when most of the hallmarks of retinal degeneration are visible in the eyes of RPE-DKO mice.
RPE65 is an enzyme involved in vitamin A metabolism in RPE cells (35). This key protein of the visual cycle catalyzes the transformation of all-trans-retinyl to 11-cisretinol. Mutations in RPE65 are associated with several retinal disorders, including retinitis pigmentosa (RP), Leber congenital amaurosis (LCA) (36), and early-onset retinal dystrophy (RD) in children (37). Much effort has been devoted to deciphering the roles of various genetic mutations that are linked to these diseases (38). However, little is known about how the activity of RPE65 is regulated at the RPE cell level. Several studies analyzed the promoter region of the RPE65 gene (33,39), but little is known about the factors involved in its transcriptional regulation. Retinoic acid has been suggested to participate in the downregulation of RPE65 (40) and of fatty acid transporter protein 4 (FATP4) and elongation of very-long-chain fatty acid protein 1 (ELOVL1) (41). Recently, Masuda and colleagues provided evidence for an involvement of sex-determining region Y box-containing gene 9 (SOX9) in the regulation of the transcription of visual cycle genes, including RPE65, retinaldehyde binding protein 1 (RBP-1) and retinal G-protein-coupled receptor (RGR). They showed that SOX9 acts synergistically with orthodenticle homeobox 2 (OTX2) to activate RPE65 gene expression (42). Here we identify another critical upstream regulatory factor for RPE65. We demonstrate that the disruption of ERK1/2 specifically in RPE cells leads to a marked decrease of RPE65 expression that occurs through an AP-1 site present in the promoter region of the RPE65 gene. The inhibitory effect of U0126 on reporter gene expression was completely abrogated when the AP-1 site was removed, even if we cannot exclude that other regulatory elements close to the AP-1 sequence were also deleted from the construct; this could explain the low promoter activity observed in the absence of the AP-1 sequence. The key roles of C-FOS and FRA-1 were confirmed by Western blot analysis and by an AP-1 DNA binding assay. Interestingly, ERK1/2 and SOX9 signaling pathways were recently shown to interact in urothelial carcinoma (43) and zebrafish sex determination (44). Moreover, the activation of ERK1/2 is associated with the activation of the Wnt/␤-catenin pathway (45,46), which plays a key role in the expression of the RPE-specific transcription factors microphthalmia-associated transcription factor (MITF) and OTX2 (47). In conjunction with protein paired box 6 (PAX6), MITF and OTX2 are key factors in RPE development (48).
ERG analysis combined with the measurement of ocular retinoid levels revealed the impaired vision of RPE-DKO mice. The significant decrease in the level of retinoids and histological evidence of retinal degeneration explain the absence of an ERG response in these mice. The RPE plays critical roles in providing nutrients to the adjacent retina and in the recycling of the visual chromophore. Upon absorption of light, 11-cis-retinal isomerizes to all-trans-retinal in the outer segments of the photoreceptors, and the RPE is necessary for the subsequent regeneration of the chromophore throughout the visual cycle (49). The loss of ERK1/2 leads to a decrease in the level of RPE65, and therefore, one would expect retinyl ester(s) to accumulate in RPE cells, as reported previously for RPE65 mutant mice (50). However, for RPE-DKO mice, we observed a significant decrease in the level of retinyl ester(s) in both dark-and light-exposed eyes. Therefore, in addition to RPE65 downregulation, which could explain the decrease in the level of 11-cis-retinal, other pathological alterations in the RPE must account for the decreased levels of ocular retinoids. We found a mislocalization of LRAT, which is expressed mostly at the apical region of RPE cells in RPE-DKO mice, in comparison to its normal basolateral localization (51). This mislocalization could affect LRAT activity and explain the decreased levels of all-trans-retinyl esters. Moreover, we also measured a lower expression level of STRA6, an RPE membrane receptor for the retinol binding protein responsible for retinyl ester transport. Because STRA6 and LRAT work together to retrieve retinoids from the blood circulation (52), such diminution may explain the reduction of the retinoid content observed in RPE-DKO mice. These results phenocopy some of the phenotypes observed for Stra6 Ϫ/Ϫ mice, including low retinoid content, the lack of an ERG response, decreases of BM and RPE lengths, and altered morphology of RPE cells (53).
The alteration of the RPE morphology in RPE-DKO mice was already visible at 1 month. Whole mounts of RPE from CTL mice show mono-and binucleated hexagonal RPE cells with an apex shared by three cells, while RPE-DKO mice exhibit an unusual RPE morphology, with a total disappearance of the RPE in some parts of the retina 2 months after ERK1/2 depletion, whereas other parts display large cells with an irregular shape and multinucleation (Fig. 6). RPE multinucleation, which has been reported for humans (54), could result from a cytokinesis defect. The involvement of cytokinesis defects as well as RPE dysfunctions and RPE cell death in aging eyes was recently described (55). As the ERK1/2 MAPK pathway plays a key role in cell proliferation, we can envisage that these kinases might be involved in the process. The sustained activation of ERK1/2 is necessary for the progression from G 1 phase to S phase and is associated with the upregulation of proliferation-associated genes and the downregulation of antiprolif-erative genes (56). Moreover, ERK1/2 signaling is necessary to allow the entry of RPE cells into cell cycling and RPE cell proliferation (57). In addition to its role in cell cycle control, ERK1/2 is also implicated in centrosome orientation (58), which is crucial for cytokinesis. Thus, we postulate that a disruption of ERK1/2 in the RPE induces defects in cell cycle progression, leading first to the multinucleation of the RPE and then to RPE cell death. Whereas central RPE cells are senescent, peripheral cells proliferate, and ERK1/2 could be important for their maintenance (59,60). Retinal regeneration in the chicken embryo was shown to be dependent on the FGF/FGFR/MEK/ERK-dependent upregulation of PAX6 (25). The activation of the ERK1/2 pathway was also responsible for the 15-deoxy-Δ 12,14 -prostaglandin J 2 (dPGJ 2 )-dependent protection of RPE cells from oxidative injury (61). These data point not only to a protective role of the ERK1/2 pathway in RPE cells but also to its participation in the maintenance and proliferation of these cells, at least in the peripheral region. Our data indicate that the ablation of ERK1/2 in 2-month-old adult RPE cells almost completely abrogated vision by initiating early cone-specific degeneration, followed by rod impairment. Surprisingly, we observed a decrease of GNAT1 immunostaining in Cre-DKO mice at 2 and 4 months, which was not observed by either qPCR or Western blot analysis. At the same time, we clearly detected a decrease in the ONL length, and thus, even if the total GNAT1 retinal content decreases, the tubulin abundance decreases in parallel so that the GNAT1/tubulin ratio is not modified. One outstanding question is why cones are more affected than rods in this model. The dependence of cone PR survival on RPE65 was reported previously (62); therefore, the direct regulation of RPE65 by ERK1/2 could account, at least in part, for the severe cone dystrophy found in RPE-DKO mice. Moreover, inadequate 11-cis-retinal production has been shown to be associated with cone degeneration, although the mechanisms are currently not well understood (63)(64)(65). Measurement of retinoid levels specifically in the retina of RPE-DKO mice revealed a significant decrease in the level of 11-cis-retinal, which could contribute to rapid cone degeneration. This finding suggests that the cone-specific effect seen at 2 months could be directly attributed to the role of ERK1/2 in regulating RPE65 activity and 11-cis-retinal production that are critical for cone PR survival.
Previously reported observations demonstrated that PR degeneration in Rpe65 Ϫ/Ϫ mice caused a massive activation of ERK1/2 in the GCL of the retina (23), probably in order to protect the retina against stress induced by the absence of RPE65. However, we recently showed that in polarized ARPE19 cells and isolated mouse RPE, A2E treatment decreases phospho-ERK1/2 significantly, along with a reduction of the RPE65 level. In addition, the inhibition of the ERK1/2 pathway by U0126 also induces a significant decrease in the level of RPE65 expression (9). The role of ERK1/2 in RPE cells has not been the subject of detailed investigation. Our study showing that the loss of ERK1/2 leads to RPE cell death, retinal atrophy, and degeneration strongly supports a significant role for ERK1/2 in the maintenance and survival of RPE cells. Defects of RPE cells and RPE cell death will lead to PR degeneration. Consistently, these features were observed in a mouse model of genetic RPE ablation using mice that expressed inducible diphtheria toxin A (DTA) specifically in the RPE. In this RPE CreER /DTA model, functional analysis showed very low scotopic and photopic ERG responses as well as PR degeneration (66). Accordingly, interactions of PRs with the RPE are essential for PR survival (67).
The very rapid INL degeneration observed in RPE-DKO mice is more difficult to explain. Even if Cre expression is observed specifically in RPE cells, as shown for Cre-tdTomato mice (Fig. 1) and as previously described (30), we also noted the ectopic expression of Cre in IPL vessels (but not in choroidal vessels) in certain mice. Therefore, we cannot totally exclude that ERK1/2 was also depleted in some IPL vessels, which may impact INL degeneration. We also hypothesize that the absence of key factors normally secreted by the RPE may influence INL degeneration. Further analyses are required in order to decipher the mechanisms leading to INL degeneration in ERK1/2-depleted mice.
Several MEK1/2 inhibitors have been clinically evaluated for cancer therapy, but some of these early-phase trials were stopped prematurely because of toxicity issues, including various ocular adverse effects (21,(68)(69)(70). Treatments last approximately 3 weeks and are repeated every month (71,72). Such long-term treatment with multiple doses of highly potent MEK inhibitors lead to a marked and sustained inhibition of ERK1/2 activity, phenocopying in part the impact of the double knockout. We observed a significant decrease of the RPE65 expression level already at 2 weeks (Fig. 7B), while mRNA and protein levels of this RPE marker were decreased by about 50% at 1 month ( Fig. 7C and D). Our study provides evidence for a key role of ERK1/2 signaling in the eye and, more specifically, in RPE cells. The ablation of ERK1/2 signaling reduces RPE65 expression levels, leads to decreased retinoid levels, affects the RPE and the retinal structure, and induces retinal degeneration. Therefore, the use of ERK1/2 pathway inhibitors in ARMD treatment, as recently suggested (26), has to be reevaluated, taking into consideration the findings reported here. Fine-tuning of ERK1/2 inhibition will be necessary in order to block only the negative effects of high kinase activity and restore low ERK1/2 activity without impacting the role of this kinase in RPE cells.

Animals. Studies involving mice adhered to the Association for Research in Vision and
Ophthalmology (ARVO) statement for the use of animals in ophthalmic and vision research (https://www.arvo .org/content-detail/level2-about2/policies2/statement-for-the-use-of-animals-in-ophthalmic-and-visual -research/) and were approved (permit number VD3023) by the Veterinary Service of the State of Vaud (Switzerland). Animals were maintained in a 12-h-light/12-h-dark cycle and had unlimited access to food and water. The generation of Erk1 Ϫ/Ϫ /Erk2 fl/fl mice was reported previously (29). The RPE-specific loss of ERK1/2 (VMD2-rtTA/TRE-Cre; Erk1 Ϫ/Ϫ ; Erk2 Δ/Δ [RPE-DKO]) was created by crossbreeding an Erk1 Ϫ/Ϫ /Erk2 f/f mouse with a mouse expressing the Cre recombinase driven by the human VMD2 (RPE-Cre) promoter and inducible by doxycycline (30) to first obtain VMD2-rtTA/TRE-Cre; Erk1 Ϫ/Ϫ ; Erk2 f/f mice. Two months after birth, VMD2-rtTA/TRE-Cre; Erk1 Ϫ/Ϫ ; Erk2 f/f mice were then injected twice, 1 week apart, with 10 g of Dox in 500 l PBS in order to induce the Cre recombinase; we call these mice RPE-DKO mice. As a control, Erk1 Ϫ/Ϫ /Erk2 f/f mice were injected in a similar way with PBS; we call these mice CTL mice. Every time period mentioned in the experiments refers to the amount of time after Dox injection (e.g., an RPE-DKO mouse at 2 months is a 4-month-old mouse treated after 2 months and analyzed 2 months later, after Cre induction and ERK1/2 depletion). Cre-tdTomato mice were created by crossbreeding VMD2-rtTA/TRE-Cre mice with tdTomato mice (31) to set up the Dox injection protocol and visualize Cre expression. Genotyping of mice was carried out by using PCR analysis with genomic DNA isolated from ear punches (Direct PCR [Ear]; Viagen). All the mice were verified to have the mutant Rd1-negative genotype. The mice were processed for fundus, OCT, and ERG analyses or sacrificed for other functional assays 2 weeks, 1 month, 2 month, 4 months, or 1 year after the second injection.
Cell culture. HEK293T cells were cultured in high-glucose Dulbecco's modified Eagle's medium (DMEM) containing 25 mM HEPES. The cells were passaged every 2 to 3 days. A total of 600,000 cells were seeded into a P60 plate and transfected after 24 h by using a calcium-phosphate transfection kit (GE Healthcare) with the indicated plasmids, according to the manufacturer's instructions. Noninvasive experiments. Mice were anesthetized by i.p. injection of 10 mg of xylazine/kg of body weight plus 80 mg/kg ketamine. Mydriatic agents (0.5% tropicamide for 2 min and 10% phenylephrine hydrochloride for 1 min) were applied on the eye for pupil dilation and to block eyebrow movement. The fundi of age-matched CTL and RPE-DKO mice were photographed. OCT was carried out on age-matched CTL and RPE-DKO mice by using an OCT system (Micron III; Phoenix Research Laboratories), according to the manufacturer's recommendations, at 4 months and 1 year.
For scotopic ERGs, age-matched CTL and RPE-DKO mice were dark adapted overnight before anesthesia and the application of the mydriatic agents as described above. For photopic ERGs, mice were exposed to bright light of 3 cd for 10 min. Full-field ERGs were recorded by using the HMsERG system (Ocuscience).
For flat-mount choroid-RPE structures, dissected choroid-RPE sections were mounted onto a coverslip, fixed in 4% PFA-PBS for 45 min, and processed for immunolabeling with Oregon Green 488 as described above. A TUNEL assay (Roche Life Science) was carried out as described previously (23).
Tissue isolation, mRNA extraction, and protein preparation. Enucleated eyes from CTL and RPE-DKO mice were dissected under a microscope to exclude extraocular tissues. The cornea, lens, iris, and vitreous body were removed, and the retina was extracted. Either the retina, the RPE, or the RPE-choroid was processed for protein/mRNA isolation.
For mRNA extraction, either the retina or the RPE-choroid was resuspended in TRIzol reagent (ThermoFisher) and stored at Ϫ80°C until further handling. RPE mRNA was extracted by using a previously described protocol (73), while retinas and RPE-choroids were extracted by using the following protocol. Briefly, 0.1 volumes of sodium acetate, 1 volume of phenol, and 0.2 volumes of chloroform were added to the tube and mixed, and phase separation was allowed for 15 min on ice. One volume of isopropanol and 2 l of glycogen (5 g/l) were added to the upper aqueous phase and left to precipitate at Ϫ20°C overnight. Following centrifugation at 10,000 ϫ g for 10 min at 4°C, the precipitate was washed with 75% ethanol (EtOH) and allowed to dry at RT before being resuspended in diethyl pyrocarbonate (DEPC)-treated double-distilled water (ddH 2 O). Quantitative PCR was carried out as described previously (23), using 100 nM primers. The sets of primers used for genotyping (Fig. 1B) and reverse transcription-PCR (RT-PCR) can be found in Table 1.
An AP-1 binding assay (TransAM AP-1 kit; Active Motif) was carried out, as described previously (24), on nuclear extracts from ARPE19 cells. Briefly, cells were resuspended in 1 ml cold buffer A (10 mM HEPES [pH 7.9], 10 mM KCl, 1 mM DTT, and protease inhibitors) and placed on ice for 15 min. Next, 12.5 l of 10% NP-40 was added, and the supernatant was discarded after centrifugation. Nuclei were lysed in buffer B (20 mM HEPES [pH 7.9], 400 mM NaCl, 1 mM DTT, and protease inhibitors) and quantified for protein content. Eight micrograms of the nuclear protein extract or 15 g of whole-cell protein extracts was used for the AP-1 binding assay according to the manufacturer's instructions.
AP-1 EMSA. An AP-1 electrophoretic mobility shift assay (EMSA) was carried out on nuclear extracts from HEK293 cells that were either left untreated or treated with U0126 according to the provider's protocol (Odyssey Infrared EMSA kit; Licor Biosciences). In brief, the nuclear extract was incubated with a solution containing 25 mM DTT-2.5% Tween 20, 1 g/l poly(dI·dC), 1% NP-40, 100 mM MgCl 2 , and the IRDye 700 AP-1 consensus oligonucleotide (5=-CGCTTGATGACTCAGCCGGAA-3=) in either the presence or the absence of the AP-1 (in italics) competitor oligonucleotide. Binding was performed for 20 min, after which the sample was loaded onto a 5% native acrylamide gel in the presence of a loading dye. Fluorescence was detected on a Licor Odyssey scanner.
Retinoid measurement. Age-matched CTL and RPE-DKO mice were either dark adapted overnight or exposed to bright light of 1,000 mAH for 2 h. For dark conditions, the retina was separated from the RPE-choroid layer. For bright-light conditions, the whole eye was processed for measurement. Ocular tissues were transferred to a 2-ml reaction vial, and 200 l of 2 M hydroxylamine (pH 6.8) and 200 l of methanol were added. All steps were conducted under red safety light (Ͼ600 nm) to avoid retinoid isomerization. Tissues were mechanically ground by using a Bio-Gen PRO200 homogenizer (Fisher Scientific). The homogenized extracts were allowed to stand for 10 min for oxime formation. Next, 400 l of acetone and 500 l of hexane were added. The samples were vortexed, and the aqueous and organic phases were separated by centrifugation at 5,000 ϫ g (Minispin Plus; Eppendorf). The organic phase was removed, and extraction was repeated with 500 l of hexane. Collected organic phases were pooled, dried with a SpeedVac (Eppendorf) at 30°C, and redissolved in HPLC mobile-phase solvent. HPLC analysis was carried out with an Agilent 1260 Infinity Quaternary HPLC system (Santa Clara) equipped with a pump (catalog number G1312C) with an integrated degasser (catalog number G1322A), a thermostated column compartment (catalog number G1316A), an autosampler (catalog number G1329B), a diode array detector (catalog number G1315D), and online analysis software (Chemstation). The analyses were carried out at 25°C by using a normal-phase Zorbax Sil (5-m, 4.6-by 150-mm) column (Agilent Technologies) protected with a guard column with the same stationary phase. For retinoid separation, the column was developed with 90% hexane and 10% ethyl acetate at an isocratic flow rate of 1.4 ml · min Ϫ1 . For molar quantification of retinoids, the HPLC system was scaled with authentic retinoid standards.
Electron microscopy. Enucleated whole mouse eyes were fixed in a 2.5% glutaraldehyde solution (Electron Microscopy Sciences) in phosphate buffer (PB) (0.1 M, pH 7.4) (Sigma-Aldrich) for 90 min at RT. Ocular tissues were then postfixed with a fresh mixture of 1% osmium tetroxide (Electron Microscopy Sciences) with 1.5% potassium ferrocyanide (Sigma-Aldrich) in PB during 1 h 30 min at RT. The samples were then washed three times in distilled water and dehydrated in graded concentrations of an acetone solution (Sigma-Aldrich) (30% for 20 min, 70% for 20 min, 100% for 1 h, and 100% for 2 h). This was followed by infiltration in graded concentrations of Epon (Sigma-Aldrich) (Epon-acetone for 2 h at a 1:3 dilution, 2 h at a 3:1 dilution, and 4 h and 12 h at a 1:1 dilution) and, finally, polymerization for 48 h at 60°C in an oven. Ultrathin sections of 50 nm were cut transversally on a Leica Ultracut microtome (Leica Microsystems) and picked up on a 2-by 1-mm nickel slot grid (Electron Microscopy Sciences) coated with polystyrene film (Sigma-Aldrich). Sections were poststained with 4% uranyl acetate (Sigma-Aldrich) in H 2 O during 10 min, rinsed several times with H 2 O followed by 0.4% Reynolds' lead citrate (75) in H 2 O (Sigma-Aldrich) during 10 min, and rinsed several times with H 2 O.
Micrographs were taken with a Philips (now FEI Company) CM100 transmission electron microscope at an acceleration voltage of 80 kV with a TVIPS TemCam-F416 digital camera (TVIPS GmbH).