Microphthalmia-associated transcription factor regulates the visual cycle genes Rlbp1 and Rdh5 in the retinal pigment epithelium

Regeneration of the visual pigment by cells of the retinal pigment epithelium (RPE) is fundamental to vision. Here we show that the microphthalmia-associated transcription factor, MITF, which plays a central role in the development and function of RPE cells, regulates the expression of two visual cycle genes, Rlbp1 which encodes retinaldehyde binding protein-1 (RLBP1), and Rdh5, which encodes retinol dehydrogenase-5 (RDH5). First, we found that Rlbp1 and Rdh5 are downregulated in optic cups and presumptive RPEs of Mitf-deficient mouse embryos. Second, experimental manipulation of MITF levels in human RPE cells in culture leads to corresponding modulations of the endogenous levels of RLBP1 and RDH5. Third, the retinal degeneration associated with the disruption of the visual cycle in Mitf-deficient mice can be partially corrected both structurally and functionally by an exogenous supply of 9-cis-retinal. We conclude that the expression of Rlbp1 and Rdh5 critically depends on functional Mitf in the RPE and suggest that MITF has an important role in controlling retinoid processing in the RPE.

microphthalmia-vitiligo (Mitf mi−vit ), however, do not lead to microphthalmia but to postnatal retinal degeneration 10,11 . MITF is well known to regulate the expression of tyrosinase and tyrosinase-related protein-1, whose mutations are associated with Oculocutaneous Albinism and Pigmentary Glaucoma 12,13 , and of BEST1, whose mutations cause Best Macular Dystrophy 14 . Nevertheless, little is known about MITF's potential involvement in the regulation of the visual cycle and whether the regulation of visual cycle genes by MITF contributes to retinal degeneration.
Here, we focus on Rlbp1 and Rdh5 and their regulation by MITF based on several observations. First, previous results have shown that in Mitf-vit mutant RPE retinyl ester levels are significantly elevated but that the accumulation of retinyl ester is not due to increased LRAT activity 15 , suggesting that dysregulation of other genes may be involved. Second, preliminary experiments indicated that visual cycle genes are down-regulated in Mitf-deficient mouse optic cups and are abundantly expressed in MITF-positive, fully differentiated RPE cells.
Our results show that the visual cycle genes Rlbp1 and Rdh5 expression are reduced in the RPE of P11 and optic cups of E12.5 of Mitf mutant mouse embryos, and Rlbp1 and Rdh5-expression is missing in Mitf mutant embryonic RPE cells. In addition, RLBP1 and RDH5 expression directly correlates with experimental manipulation of MITF levels in the human RPE cell line ARPE-19 cells. Finally, addition of exogenous 9-cis-retinal partially rescues retinal degeneration in Mitf-deficient mice, suggesting that Mitf regulation of visual cycle genes can be functionally relevant for the pathogenesis of retinal degeneration.

Impairment of visual cycle gene expression in Mitf-deficient retina.
It has been shown previously that RPE development is disrupted in Mitf mi-vga-9 homozygous embryos (hereafter called Mitf− /− ) 9,16 , but the ultrastructure of the mutant RPE has not been examined. Examination by transmission electron microscopy (TEM) revealed that the RPE of Mitf− /− P0 and P20 mice displayed apical microvilli and basal infoldings also seen in wild-type RPE but they were sparse and disorganized. In addition, Bruch's membrane was interrupted and melanosomes and melanin were absent ( Fig. 1A-F). These results indicated that although the RPE was retained in Mitf− /− mice, its ultrastructure was abnormal.
It has previously been shown that retinyl ester levels were significantly elevated in Mitf-vit mutant RPE, without, however, showing alterations in LRAT activity 15 . Hence, we first examined the expression of visual cycle genes in wild type and Mitf− /− retina/RPE preparations in comparison to the expression of pigmentation-related genes, which are known to be regulated by Mitf. At P11, the expression of visual cycle genes Rlbp1 and Rdh5 as well as Lrat, Rpe65, and Rgr were significantly decreased in Mitf− /− retina/RPE, as was Tyr which encodes the rate-limiting enzyme for melanin synthesis (Fig. 1G). Interestingly, further examination revealed that in wild type, though not in mutant, Rlbp1 and Rdh5 were expressed in eye vesicles already at E12.5 when the RPE develops (Fig. 1H,I). In contrast, Lrat, Rpe65, and Rgr were not expressed at these early stages (not shown). These results suggest that Mitf regulates Rlbp1 and Rdh5 gene expression both during development and after birth.

MITF regulates the expression of RLBP1 and RDH5 in RPE cells.
To evaluate whether the regulation of Rlbp1 and Rdh5 was indeed controlled by Mitf, we tested whether the inhibition or overexpression of MITF can affect the endogenous expression of RLBP1 and RDH5 in a human RPE cell line. We first transfected cultures of ARPE-19 cells with si-Mitf, a pool of siRNAs directed against MITF or si-NC, a control siRNA. As shown in Fig. 2, knockdown of MITF reduced the expression of RLBP1 and RDH5 RNA and protein ( Fig. 2A,B). In contrast, overexpression of MITF using a lentiviral expression system increased the expression of RLBP1 and RDH5 proteins (Fig. 2C,D). These results indicate that MITF regulates the expression of RLBP1 and RDH5 in human RPE cells.
Rlbp1 and Rdh5 expression depends on MITF in vivo. Although Rlbp1 and Rdh5 are expressed in RPE, their spatial and temporal expression patterns during eye development are unclear. By in situ hybridization, the expression of Rlbp1 and Rdh5 was seen already at E9.5 when the optic vesicle is initially formed (Fig. 3A). At this time point, the expression of these two genes was confined to the dorsal part of the optic vesicle where by comparison Mitf expression was highest, but by E12.5 it was extended to the entire presumptive RPE. A comparison between WT and Mitf− /− embryos revealed that the expression of Rlbp1 and Rdh5 depended on intact Mitf while the RPE marker-Dct did not (Fig. 3B), confirming previous studies of Dct expression 16 . These results suggest that the loss of functional MITF leads to disruption of the expression of visual cycle genes and that Mitf is required for the expression of Rlbp1 and Rdh5 in mouse RPE. 9-cis-retinal treatment partially rescues Mitf −/− retinal morphology and function. In order to address the question of whether the retinal degeneration in postnatal Mitf mutant mice might result in part from impairment of the visual cycle, we made use of the fact that 9-cis-retinal, an analogue of 11-cis-retinal, can rescue retinal degeneration in mice with impaired visual cycle 3,17 . Hence, we injected 9-cis-retinal into Mitf− /− mice at P8 and P15 and examined their retinal structure and function at P22 (Fig. 4). When injected with solvent alone, the rod response and standard combined electroretinogramm (ERG) was lost in Mitf − /− mice (Fig. 4A,B), and the b wave amplitude of the rod response was 21.4 μ v and that of the standard combined ERG 11.9 μ V (Fig. 4C). Furthermore, the mice showed severe retinal degeneration, marked by loss of most of the outer segments of photoreceptors and loss of the outer and inner plexiform layers (Fig. 4D) and a general reduction in retinal thickness (Fig. 4E). When 9-cis-retinal was injected, however, Mitf− /− retinas were markedly improved in ERG and structure, and photoreceptor gene expression (Figs 4 and 5). As shown in Fig. 4C, the average b wave amplitude of rod response was increased to 44.3μ v and that of the standard combined ERG to 43.2μ V. Compared to untreated mutant, the retinal thickness was double, and the ONL, inner plexiform layer, and ganglion cell layer were partially spared from the degeneration seen in the absence of 9-cis-retinal (Fig. 4D,E).
Scientific RepoRts | 6:21208 | DOI: 10.1038/srep21208 Previous studies had shown that cone M/L-OPSIN expression was lost in visual cycle gene mutant mice, and that administration of 9-cis-retinal prevented this loss 18 Therefore, we also examined Opsin and Rhodopsin expression in WT, control Mitf− /− and 9-cis-retinal-injected Mitf− /− mice. As shown in the corresponding panels in Fig   cones, (Crx, Nrl, and Nr2e). Hence, we conclude that 9-cis-retinal can partially bypass the loss of visual cycle gene expression and prevent retinal degeneration in Mitf mutant mice.

Discussion
Here we provide experimental evidence that the expression of two visual cycle genes, Rlbp1 and Rdh5, depends on functional MITF both in vitro in a human RPE cell line and in vivo in Mitf mutant mice. The finding helps explain the previous observation that retinyl ester levels were elevated significantly in the RPE of Mitf-vit mutant 15 . In these mutants, the activity of LRAT was not changed so that the increase in retinyl ester was unlikely due to increased production. Rather, it might be due to reduced levels of Rlbp1 and Rdh5 as the products of these two genes are involved in processes of the visual cycle downstream of retinyl ester: RLBP1 binds 11-cis-retinol and RDH5 catalyzes the reaction to 11-cis retinal. In Mitf− /− embryos, however, Lrat expression was reduced, suggesting that retinyl ester levels might not be increased, and hence that the visual cycle was disrupted at even more levels than in Mitf-vit mutants.
Our findings clearly show that by regulating visual cycle genes, Mitf controls the visual cycle and consequently the reconstitution of functional rhodopsin. They do not, however, answer the question of whether it is the mere interruption of the visual cycle, or a more general damage to the RPE, that leads to the associated retinal degeneration. The RPE, after all, has multiple functions in photoreceptor and retinal physiology through its supply of neurotrophic factors, phagocytosis of photoreceptor outer segments, blood retinal barrier maintenance, etc, and its damage may well interfere with any of these functions.
An experimental approach to distinguish between the role of visual cycle interruption and general RPE damage in the cause of retinal pathophysiology was partially provided by the application of 9-cis-retinal, which previously had been shown to substitute for 11-cis-retinal as the final product of the visual cycle and to rescue retinal degeneration in models of isolated visual cycle interruptions. We found, indeed, that postnatal injection of 9-cis-retinal can partially rescue Mitf− /− retinae because gene expression, structure and function were all improved. This was in a way surprising as one might expect that the lack of Mitf in the RPE may have so profound developmental effects that the adjacent retina might be permanently damaged. Nevertheless, it is clear that the rescue was only partial. Several explanations may be offered for this observation. First, the supply of 9-cis-retinal may not have reached high enough levels, or may have occurred too late, to fully rescue photoreceptors and retina. Second, as Mitf is not only expressed in the future RPE but also in the future retina, albeit at low levels, it was conceivable that the retina might be developmentally impaired even if RPE and the visual cycle were normal. Third, as mentioned above, the restoration of only one function, the visual cycle, might just not be enough to fully normalize the structure and function of a retina that is likely severely damaged at multiple levels by the absence of a normal RPE. Following this latter thought, we would predict that the retinal phenotype associated with certain milder Mitf mutations, which might damage the visual cycle but leave other RPE functions largely intact, might more readily be corrected with 9-cis-retinal alone. In sum, then, our work provides evidence that MITF affects the visual cycle at least through the regulation of RLBP1 and RDH5. Recently, it has also been shown that two other transcription factors present in RPE, SOX9 and OTX2, act synergistically to activate the RLBP1 and RPE65 promoters 19 . Future work will delineate whether this regulation is direct, operating through enhancers and promoters of the respective genes, or indirect through interspersed factors, what co-operating factors might be involved, and whether the whole set of visual cycle genes is regulated concordantly by MITF alone or in conjunction with SOX9 and OTX2.

Materials and Methods
Animal. All animal experiments were carried out in accordance with the approved guidelines of the Wenzhou Medical University Institutional Animal Care and Use Committee. C57BL/6J, FVB, and Mitf mi-vga-9 mutant mice were used for this study. To collect mouse embryos at the indicated time points, the noon of the day a vaginal plug was found was defined as embryonic day 0.5 (E0.5). Conventional PCR and quantitative real-time PCR. For PCR analysis, mouse retinas were collected at postnatal day 11, and optic cups from mouse embryos were isolated at E12.5. The respective tissues were immediately homogenized in Trizol for RNA extraction. Conventional PCR was first conducted to analyze a set of pigmentation-related genes and visual cycle genes in P11 retina and E12.5 optic cups of wild-type and Mitf mi-vga-9 homozygous mice, and quantitative real-time PCR was performed to confirm expression level alterations of the selected genes. Total RNA was isolated and reverse transcribed into cDNA from mouse retinas, optic cups, or cultured cells as previously described 20 . Cell culture, lentivirus preparation and infection. The human retinal pigment epithelial cell line (ARPE-19 cell) was obtained from American Type Cell Culture (ATCC) and maintained in DMEM/F12 medium supplemented with 10% fetal calf serum (Gibco-BRL, Invitrogen). D407 and 293T cells were cultured in DMEM/ F12 medium supplemented with 10% fetal calf serum. For lentivirus preparation, 7.5 μ g FUW-Mitf, 5.625 μ g Pax2 and 1.875 μ g PMD plasmids were cotransfected into 293T cells by polyjet transfection reagent according to the instructions of manufacturer (Signagen). After 48h, the supernatant were concentrated by Lenti-X Concentrator (Clontech). Briefly, lentivirus-containing supernatants were mixed with Lenti-X Concentrator at 3:1 ratio, then the mixture was incubated at 4 °C for 30 minutes to overnight. Samples were centrifuged at 1,500 X g for 45 minutes at 4 °C, then off-white pellets were resuspended in DMEM. For lentivirus infection, ARPE19 or D407 cells were seeded in 6-well plate at 1 × 10 5 cells per well. After 24 hrs, lentivirus with 8 μ g/ml polybrene was added to the culture medium. Six hours after infection, fresh medium was added for 1 hour, and a second round of infection was conducted. siRNA transfection. ARPE19 cells were seeded in 12-well plates at 4 × 10 4 cells/well before transfection.
When the cell density reached 30% confluence, cells were transfected with appropriate dilutions of 20 μ M siMitf pool (Thermo scientific) containing 4 siRNAs targeting different regions of MITF RNA, each at 5μ M, using lipojet reagent (Signagen) at the indicated concentrations following the company's instructions. Briefly, for transfection at a final concentration of 20 nM, 20 pmoles siRNA were diluted into 100 μ l working solution of LipoJet ™ Transfection Buffer, and 1.5 μ l LipoJet ™ reagent was added and mixed by pipetting up and down.
The transfection complex was incubated for 15 min at RT and added to 1 ml medium in each well. Seventy-two hours post-transfection, cells were collected in Trizol or protein lysis buffer. In all experiments, control siRNA (CONTROL Non-Targeting siRNA; Thermo scientific) was included.

Western blotting. Total protein from Mitf or Control siRNA-treated ARPE19 cells, and EGFP or
Mitf-lentivirus-infected ARPE19 cells were subjected to Western blotting with anti-MITF (Abcam, 12039), anti-CRALBP (Thermo scientific, MA1-813), anti-RDH5 (Santa Cruz, sc-98348), anti-TUBULIN (Santa Cruz, sc-5286) antibodies. Western blots were prepared and performed according to a published procedure 20 . In Situ hybridization. Embryos at various embryonic days as indicated in the results were collected in PBS and fixed overnight at 4 °C using 4% paraformaldehyde (PFA). Sectioning and whole mount in situ hybridization were performed according to a published procedure 21 . The Mitf and Dct riboprobes were described previously 22,23 . The Rlbp1 riboprobe was kindly provided by Dr Connie Cepko (Department of Genetics, Harvard Medical School, Boston, MA). The Rdh5 riboprobe was obtained by amplifying the 864 bp fragment of the murine Rdh5 coding region from E12.5 C57BL/6J mouse optic cup cDNA using the sense primers 5-GCC GGC CAG TGA TGC TTT C and the antisense primer 5-CGG GGA AGG ATC CAG GTG AG. The PCR product was cloned into TOPO-TA vector (Invitrogen), and verified by sequencing. 9-cis-retinal experiment. For 9-cis-retinal treated experiments, 9-cis-retinal (Sigma) was dissolved in 100% ethanol to generate a 125 mg/ml stock solution. For each experiment, 2 μ l of 9-cis-retinal stock was combined with 18 μ l ethanol and 180 μ l matrigel at 4 °C for a final volume of 200 μ l. The mixture was injected subcutaneously into the dorsal torso region as previously described 18 . 9-cis retinal was omitted from the mixture for all control groups. Mitf mi-vga-9 homozygous and heterozygous littermates were injected subcutaneously at P8 and P15, and reared in darkness. Eye balls were collected at P22 for further examinations as described in the results. Electroretinography (ERG). At P22 following solvent or 9-cis-retinal administration, a RETIport system with a custom-built Ganzfeld dome (Roland Consult, Wiesbaden, Germany) was utilized for ERG recording. Overnight dark-adapted mice were anesthetized under dim red light by intraperitoneal injection of 7 μ L/g body weight of 6 mg/mL ketamine and 0.9 mg/mL xylazine diluted with 10 mM sodium phosphate, pH 7.2, containing 100 mM NaCl. Pupils were dilated with 1% tropicamide, and a heating pad was used to keep the body temperature at 38 °C. The corneal electrode was a gold wire loop; a reference electrode was placed on the forehead and a ground electrode was applied subcutaneously near the tail. For dark-adapted ERGs, the intensities of flashes were − 5.0, − 15, − 25,and − 35 log scotopic candela sec/m2 (cd. s/m2). Twelve ERG scans were averaged for dark-adapted ERGs. Ganzfeld illumination with white light stimulus was applied for a duration of 2 milliseconds. B-wave amplitudes were defined as the difference between the trough and peak of each waveform.
Immunostaining. For immunostaining, eyes were fixed in 4% PFA at 4 °C for overnight, cryopreserved in 30% sucrose, and embedded in optimal cutting temperature compound. Sections (14 μ m) were collected on a cryostat, dried at RT for 30 min, and fixed in 2% PFA for 10 min. After rinsing, the sections were permeabilized with 0.2% Triton-X-100 for 10 minutes and blocked with 10% goat serum for 1 hour at RT. Anti-Rhodopsin (milipore MAB5316) and anti-Opsin (milipore AB5405) were used as primary antibodies, and goat anti-mouse IgG alexa Fluor ® 488 (life technologies, A11029 ) and goat anti-rabbit IgG alexa Fluor ® 594 (life technologies, A11037) Scientific RepoRts | 6:21208 | DOI: 10.1038/srep21208 were used as secondary antibodies. Each staining was performed on slides from at least three animals per condition. Immunostaining results were observed and photographed using a Zeiss confocal microscope. Statistical Analysis. Statistical comparisons between two groups were performed by an unpaired Student's t-test. P < 0.05 was considered to be significant. Significant differences between groups are noted by *or **.