Microphthalmia-associated Transcription Factor (MITF) Promotes Differentiation of Human Retinal Pigment Epithelium (RPE) by Regulating microRNAs-204/211 Expression*

Background: microRNAs 204/211 regulate retinal pigment epithelial cell phenotype. Results: In RPE, MITF regulates miR-204/211 expression and down-regulation of MITF results in loss of RPE phenotype, which can be prevented by overexpressing miR-204/211. Conclusion: MITF-mediated expression of miR-204/211 directs RPE differentiation. Significance: miR-204/211-based therapeutics may be effective treatments for diseases that involve loss of RPE phenotype. The retinal pigment epithelium (RPE) plays a fundamental role in maintaining visual function and dedifferentiation of RPE contributes to the pathophysiology of several ocular diseases. To identify microRNAs (miRNAs) that may be involved in RPE differentiation, we compared the miRNA expression profiles of differentiated primary human fetal RPE (hfRPE) cells to dedifferentiated hfRPE cells. We found that miR-204/211, the two most highly expressed miRNAs in the RPE, were significantly down-regulated in dedifferentiated hfRPE cells. Importantly, transfection of pre-miR-204/211 into hfRPE cells promoted differentiation whereas adding miR-204/211 inhibitors led to their dedifferentiation. Microphthalmia-associated transcription factor (MITF) is a key regulator of RPE differentiation that was also down-regulated in dedifferentiated hfRPE cells. MITF knockdown decreased miR-204/211 expression and caused hfRPE dedifferentiation. Significantly, co-transfection of MITF siRNA with pre-miR-204/211 rescued RPE phenotype. Collectively, our data show that miR-204/211 promote RPE differentiation, suggesting that miR-204/211-based therapeutics may be effective treatments for diseases that involve RPE dedifferentiation such as proliferative vitreoretinopathy.

The retinal pigment epithelium (RPE) 2 is a monolayer of cells that forms the outer blood retinal barrier and performs a num-ber of specialized functions that are critical for photoreceptor health and excitability (1). The RPE possesses long apical microvilli that ensheath the photoreceptor outer segments. These two structures are separated by a small volume of space (subretinal space Ϸ 10 l) that serves as a conduit for the transfer of nutrients and metabolic wastes between photoreceptors and the choroidal blood vessels (2). The RPE hosts a unique set of enzymes such as lecithin retinol acyltransferase (LRAT) and RPE65 that participates in the visual cycle by catalyzing the conversion of all-trans-retinol to 11-cis-retinal, the latter of which is critical for photoreceptor excitability (3). In addition, RPE cells phagocytose shed photoreceptor outer segments (4) and secrete growth factors to nourish the retina (5) and the choroidal blood vessels (6). Thus, loss of RPE functions, as occurs in age-related and proliferative diseases, invariably leads to photoreceptor degeneration and visual impairment (7).
Damage to the RPE or retinal detachment caused by trauma or intraocular diseases can trigger a repair process, in which RPE cells lose cell-cell contact and epithelial phenotype to become proliferative and motile fibroblast-like cells. In proliferative vitreoretinopathy (PVR), for example, unchecked proliferation of RPE cells and their migration into retinal layers and the vitreous result in formation of epiretinal membranes that can contract and cause retinal detachment and visual impairment (8). This switch from an epithelial to mesenchymal-like phenotype involves complex cellular reprogramming with significant alterations in core cellular functions (e.g. metabolism, cell-cell junctions, cell-cycle progression, cytoskeletal rearrangement) as well as gene and protein expression (9). In the search for potential regulators of this process, microRNAs (miRNAs) appeared to be excellent candidates because each miRNA potentially regulates expression of a large array of genes (ϳ300) that may be involved in a variety of cellular functions such as proliferation and metabolism (10). Recent studies in other biological systems have also established a role of miRNAs in cellular differentiation as positive and negative regulators of epithelial-to-mesenchymal transition (EMT) (11,12).
MiRNAs are small (ϳ23 nucleotides) regulatory RNAs that suppress gene expression by binding to specific sequences in the 3Ј-untranslated region of their target mRNA. Studies in various organ systems revealed that certain miRNAs are highly enriched in a tissue-specific pattern (13)(14)(15)(16). Furthermore, transfection of such miRNAs into stem cells (17,18) or even fibroblasts (19) can induce differentiation into the cell type that normally expresses the miRNA at high levels. These findings support the notion that specific miRNAs may direct cell specification and differentiation of cells that normally express them at high levels (reviewed in Ref. 20). Because down-regulation of tissue-specific miRNAs is commonly associated with disease, their restoration may slow or inhibit disease progression. For example, miR-145 directs smooth muscle differentiation, and its expression was down-regulated in vascular walls with neointimal lesions induced by arterial injury (15). Formation of these lesions was inhibited when injured arteries were transfected with miR-145.
In the RPE, miR-204 and 211 are the two most highly enriched miRNAs, and their expression is critical for maintaining barrier properties and function (16). miR-211 resides in the sixth intron of TRPM1 (melastatin, transient receptor potential cation channel subfamily M member 1), the transcription of which is regulated by microphthalmia-associated transcription factor (MITF) (21), a master regulator of melanocyte and RPE differentiation (22,23). Mice with homozygous null mutation in the MITF gene have white coats and microphthalmia (24). Furthermore, histological analysis of microphthalmia mouse eyes demonstrated that the absence of MITF prevented RPE differentiation (25). Because MITF and miR-204/211 are important regulators of RPE development and function, it was of interest to determine whether MITF regulates miR-204/211 expression in the RPE and whether expressing high levels of miR-204/211 alone is sufficient to direct RPE differentiation.
In this study, we used primary cultures of human fetal RPE cells (hfRPE) developed by Maminishkis et al. as a model system (26,27). These cells exhibit properties (morphology, physiology, protein and mRNA profiles) characteristic of native fetal or adult human RPE. In our experiments, we mimicked RPE detachment and dedifferentiation (as occurs in PVR) by subculturing cells at low cell density and found that this process resulted in significant down-regulation of MITF and miR-204/ 211. Using this in vitro model of RPE dedifferentiation, we found that introduction of pre-miR-204/211 promoted RPE differentiation and protected them from dedifferentiation. Our findings may help facilitate development of miR-204/211based therapies for human ocular diseases that involve RPE dedifferentiation such as age-related macular degeneration and PVR.

EXPERIMENTAL PROCEDURES
hfRPE Culture Model-hfRPE monolayers were cultured on T25 flasks (P0 hfRPE) as described previously (26). Briefly, hfRPE cells were trypsinized from a T25 flask and seeded onto 12-well Transwells at Ϸ1.25 ϫ 10 5 cells/well. P1 hfRPE cells were cultured for 3-4 weeks to reach maturity (transepithelial resistance (TER) Ͼ500 ohms⅐cm 2 ) prior to experimentation. TER was measured with an epithelial volt-ohm meter (EVOM) (WPI, Sarasota, FL) at room temperature. Media and Transwell resistances were taken into account by subtracting 122 ohms⅐cm 2 from the EVOM readout. To test for choroidal fibroblast contamination, hfRPE cells were stained with collagen type I/procollagen antibody (Cell Sciences; Canton, MA). Human fetal choroidal fibroblast cells were used as positive controls and were cultured in the same medium as hfRPE cells.
Total mRNA Extraction-Total mRNA of samples was extracted using mirVana miRNA extraction kit (Ambion, Austin, TX) according to the manufacturer's protocol. RNA bound in the column matrix was treated with RQ1 DNase (5 units/ sample; Promega) at 37°C for 30 min followed by multiple wash steps according to the manufacturer's protocol. RNA was eluted with diethylpyrocarbonate-treated water preheated to 85°C. Total RNA concentration was measuring using Qubit fluorometer (Invitrogen).
miRNA Microarray and Data Analysis-Total mRNA of differentiated and dedifferentiated hfRPE samples were prepared using TRIzol (Invitrogen) as described previously (28), and 100 ng of total mRNA from each sample was labeled and hybridized to a human miRNA microarray (V2) from Agilent Technologies (Santa Clara, CA) according to the manufacturer's protocol. The microarray was scanned with an Agilent Microarray Scanner, and the data were processed using Feature Extraction software v10.7.3.1 (Agilent). The microarray was normalized to miR-24 and miR-130a, whose expression levels were the least different between the two RPE cell phenotypes. The normalized array was analyzed using Significance Analysis of Microarrays (SAM 4.0 with R2.14.1) (29) for two-class unpaired statistical analysis with ⌬ ϭ 5.0 and -fold change Ͼ2. miRNAs with fluorescence Ͻ50 in both RPE sample types were eliminated. LOG2 fluorescence intensities of miRNAs were represented with a heatmap generated in MultiExperiment Viewer (MeV v4.8). The normalized version of the microarray data can be downloaded from NCBI GEO database (accession number GSE36137).
Reverse Transcription and Real-time Quantitative PCR (qPCR)-RNA (1 g/sample) was reverse transcribed using oligo(dT) 20 primers and SuperScript III (Invitrogen). qPCRs for gene expression studies were performed using ITaq SYBR Green Supermix with ROX (Bio-Rad) in 20-l reactions (10 ng of cDNA/RxN). qPCR was performed using Eppendorf Mastercycler ep realplex 2 . Primers were designed according to guidelines set by Dieffenbach et al. (30). Custom oligonucleotides were purchased from Eurofins MGW Operon (Huntsville, AL). Sequences for all primers used in this study are listed in supplemental Table 1.
qPCR Data Analysis-For SYBR Green qRT-PCR, ribosomal protein S18 (RPS18) gene was used as reference gene because the 2 -Ct values of RPS18 from differentiated versus dedifferentiated hfRPE samples were statistically insignificant. For Taq-Man assays, U18 snoRNA was used as reference gene because U18 lies within the intron of RPL4 and the mean Ct values (2 -Ct ) of RPL4 of dedifferentiated versus differentiated hfRPE samples were statistically insignificant. 2 -⌬Ct of treated versus control samples was analyzed for statistical significance using Student's t test (two-tailed; unpaired samples, unequal variances). p values of Ͻ 0.05 were considered statistically significant.
siRNA, Pre-miRNA, and Anti-miRNA Transfection-Ambion pre-miR miRNA precursors and anti-miRNA were purchased from Applied Biosystems. All siRNAs were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). In all experiments, pre-miRNA were transfected upon seeding and on the 3rd day after seeding. Dharmafect 4 was used as transfection reagent (0.2%) in antibiotic-free complete MEM containing 5% serum.
Western Blotting-hfRPE cells on Transwell filters were lysed and homogenized as described previously (28). 15 g of total protein lysates was loaded onto a NuPAGE 4 -12% Tris-acetate gel (Invitrogen) for electrophoresis. Proteins were subsequently transferred onto PVDF membranes using XCell II TM Blot Module (Invitrogen). Nonspecific binding sites were blocked with TBS (ϩ0.1% Tween 20) containing 5% w/v powdered milk. Antibodies used in this study are listed in supplemental Table 2.
Immunofluorescence and Imaging-hfRPE cells on Transwell filters were fixed with 4% formaldehyde in 1ϫ PBS for 5 min at room temperature followed by 20 min at 4°C. Samples were permeabilized for 5 min with 0.3% Triton X-100 and blocked with PBS ϩ 0.1% Tween 20 (PBST) containing BSA (5% w/v). Samples were incubated overnight with antibodies against MCT3 and ZO-1 (clone ZO1-1A12; Invitrogen). Samples were washed with PBST and incubated for 1 h in secondary antibodies (Invitrogen). After washing with PBST, samples were stained with phalloidin (1 h at 1:100; Invitrogen) and DAPI (5 min at 1:1000) prior to mounting with gelvatol onto microscope slides. Confocal images on Figs. 3G and 5G were taken with Zeiss LSM 510 confocal microscope at ϫ40 (Plan-Neofluar ϫ40/1.3 oil differential interference contrast) with ϫ2 scanner zoom (ϫ80 final) and 0.5-m Z-stack intervals. All other images were taken using Nikon A1R confocal microscope at ϫ60 (Plan Apo VC ϫ60 WI differential interference contrast N2) with ϫ1.33 scanner zoom (ϫ80 final) and 0.5-m Z-stack intervals. Images were extracted from NIS-Elements and analyzed using Adobe Photoshop 7.0.
miRNA Target Prediction Analysis-Putative miR-204/211 targets were obtained from TargetScan, miRanda, PicTar, and miRDB. Because predictions made by TargetScan (and PicTar), among other algorithms, have been shown to be among the most accurate (as analyzed by proteomics) (32), miRNA target genes ranked in the "top 100" list of TargetScan were scored higher. Each gene was annotated with their respective ontology profiles and pathway profile (GenMAPP and KEGG data bases) that were extracted from the annotation files of Affymetrix human microarray chipset (HG-U133 Plus 2). This list of miR-204/211 targets is available in supplemental Table 3, and a selection of these targets was categorized and is listed in Fig. 6.

RPE Dedifferentiation Involves Loss of RPE-specific Genes and miR-204/211-
The RPE is normally quiescent and nonmigratory, but in disease conditions such as PVR, it can undergo dedifferentiation into fibroblast-like cells that are proliferative and motile. This phenomenon was observed in vitro as cells at the free edge of differentiated hfRPE monolayer dedifferentiate, migrate, and establish a new population of sparsely pigmented fibroblast-like cells (28). To study the role of miRNAs in RPE differentiation, we compared the miRNA expression profile of dedifferentiated versus differentiated hfRPE cells using miRNA microarray analysis (Fig. 1A). In this experiment, we also included a sample of partially differentiated hfRPE cells (pigmented but lost epithelial morphology) to represent RPE cells in an earlier stage of dedifferentiation. From this array, we found that the three most highly expressed miRNAs (miR-204, miR-211, and miR-125b) in the RPE (16) were significantly down-regulated in dedifferentiated hfRPE cells. miR-200a and miR-200b, which suppress EMT by targeting Zeb1 and Zeb2 transcription factors (33), were also down-regulated in dedifferentiated RPE cells. In addition, expression of miRNAs that are commonly down-regulated (let-7 family) or up-regulated (miR-21 and miR-31) in cancer was also altered. This model of RPE dedifferentiation, however, was difficult to manipulate because dedifferentiation and migration of cells from the edge of the RPE monolayer occur sporadically and therefore cannot be experimentally induced and controlled. Thus, we developed an alternative in vitro model of RPE dedifferentiation in which we passaged P1 hfRPE cells at low density (P2 at 1%, P3 at 30%) twice to produce a homogeneous population of dedifferentiated hfRPE cells (Fig. 1B). However, such seeding conditions also promote growth of choroidal fibroblast contaminants that may be present in the hfRPE culture. To address this concern, we examined the purity of our hfRPE cultures and showed that our dedifferentiated hfRPE cells did not express collagen type I (gene and protein), which was highly expressed in choroidal fibroblast cells (supplemental Fig. 2). Using this model of RPE dedifferentiation, we validated our microarray data with Taq-Man qRT-PCR miRNA assay (Fig. 1C). In agreement with the microarray data, we found that miR-204/211 were among the most significantly down-regulated miRNAs in dedifferentiated RPE cells.
The study of miR-204/211 in RPE differentiation necessitates the use of a model system of RPE differentiation and dedifferentiation. We chose to vary seeding density (P1 to P2 at 7.5, 15, 30, or 60%) and found that hfRPE cells seeded at 30 and 60% densities developed characteristic RPE morphology (supplemental Fig. 1A) and expressed high levels of RPE-specific proteins such as MCT3 and CRALBP (supplemental Fig. 1B). On the other hand, hfRPE cells seeded at 7.5 and 15% densities exhibited fibroblast-like morphology and expressed low levels of MCT3 and CRALBP. In addition, qRT-PCR results show that expression of miR-204/211 (supplemental Fig. 1D) and RPEspecific genes (BEST1, CLDN19, CRALBP, MCT3, and RPE65) (supplemental Fig. 1E) were generally higher in RPE cells seeded at 60 versus 15% density. This trend was also observed at all three time points tested (3, 7, and 15 days after seeding). By 15 days, the expression of several RPE-specific genes (BEST1, CLDN19, and MCT3) in hfRPE cells seeded at 60% density reached levels comparable with that of P1 hfRPE cells, whereas RPE65 took significantly longer to achieve high levels of expression. These data showed that seeding hfRPE cells at 15% or lower led to RPE dedifferentiation whereas seeding at 30% or higher resulted in RPE differentiation.
To evaluate the role of miR-204/211 in establishing barrier functions of the RPE, we transfected hfRPE cells (15% density) with pre-miR-204/211 (25 nM each) or control miRNA (50 nM) on Transwell filters and measured TER after 21 days in culture. In this experiment, pre-miRNAs were transfected only twice (once upon seeding and another on the 3rd day). Fig. 2F shows that at the end of the 3rd week, hfRPE cells transfected with miR-204/211 had higher TER (79 Ϯ 23 ohms⅐cm 2 ) compared with control (35 Ϯ 8 ohms⅐cm 2 ; n ϭ 4; p ϭ 0.03). Immunostaining of these cells revealed that hfRPE cells transfected with control miRNA formed multiple layered fibroblast-like cells with stress fibers that did not express MCT3 (Fig. 2G, upper). In contrast, hfRPE cells transfected with pre-miR-204/211 formed a monolayer of hexagonally packed cells with circumferential bundles of actin filaments at the lateral junctions and apical microvilli as revealed by phalloidin staining. Furthermore, these cells reestablished proper epithelial polarity as MCT3 labeling was restricted to the basolateral membrane as observed in a mature and polarized RPE in situ. Taken together, our data indicate that miR-204/211 can prevent RPE dedifferentiation.
To test the effect of anti-miR-204/211 on RPE barrier function, we measured the TER of hfRPE cells cultured at 30% density transfected with anti-miR-204, -211, or both anti-miR-204/ 211 (two transfections over 14 days). After 14 days in culture, control hfRPE cells established a resistance of Ϸ250 ohms⅐cm 2 , whereas cells transfected with anti-miR-204, -211, or both 204/ 211 had significantly lower resistances (Ϸ120, 180, and 80 ohms⅐cm 2 , respectively) (Fig. 3F). The morphology of cells transfected with either anti-miR-204 or -211 alone was not sig-FIGURE 1. RPE dedifferentiation is characterized by loss of epithelial phenotype, changes in miRNA profile, and significant alterations in mRNA and protein expression. Differentiated hfRPE cells (passage 1 (P1)) were grown on Transwell filters over 4 weeks to obtain a differentiated RPE monolayer. A, dedifferentiated RPE cells that migrated from the free edge of confluent hfRPE monolayer were isolated, and microRNA microarray was performed to compare their miRNA expression levels with that of differentiated RPE cells. B, in a different model, dedifferentiated RPE cells were obtained by passaging P1 hfRPE cells at low cell density twice (P1 to P2 at 1%, P2 to P3 at 30%) on 100-mm culture dishes. C, the expression of several miRNAs that were significantly altered in the microarray analysis was verified using TaqMan microRNA assay. D-F, to compare mRNA expression of differentiated versus dedifferentiated RPE cells, qRT-PCR was performed to evaluate the expression of miR-204/211 targets (D), genes involved in barrier, nutrient/ion transport, and RPE-specific functions (E), and genes that are commonly up-regulated in EMT (F). G, Western blot shows that RPE dedifferentiation involves a loss of RPE-specific proteins and an increase in EMT-associated proteins. Statistically significant changes (p Ͻ 0.05) are marked with asterisks. Error bars, S.D.  nificantly different from control anti-miRNA-transfected cells (data not shown), but cells transfected with both anti-miR-204 and -211 exhibited dramatic loss of RPE phenotype as characterized by the complete loss of MCT3 and ZO-1 and the formation of multilayered cells with stress fibers (Fig. 3, G and H). Collectively, our data indicate that inhibition of both miR-204 and -211 is required to induce RPE dedifferentiation.
Next, we examined whether hfRPE cells transfected with both MITF siRNA and miR-204/211 could reestablish barrier functions. TER was measured on the 14th and 21st day, and we observed that hfRPE cells with MITF KD had no detectable resistance at either time point (Fig. 5F). However, hfRPE cells transfected with both MITF siRNA and pre-miR-204/211 had resistances of Ϸ240 ohms⅐cm 2 on the 21st day, demonstrating that miR-204/211 can prevent loss of RPE barrier function caused by MITF KD. Immunofluorescence staining of these samples showed that MITF KD resulted in loss of MCT3 and ZO-1 (Fig. 5G), whereas co-transfecting miR-204/211 with MITF siRNA maintained expression and polarized distribution of MCT3 and ZO-1 to the basolateral membrane and tight junction region, respectively. Phalloidin staining revealed that knockdown of MITF in hfRPE resulted in the formation of mul-tilayered fibroblast-like cells with stress fibers. Co-transfection of hfRPE cells with MITF siRNA and pre-miR-204/211 rescued the RPE phenotype (Fig. 5H). Taken together, our data strongly suggest that loss of MITF led to miR-204/211 down-regulation and subsequent loss of RPE phenotype and function.

DISCUSSION
Dedifferentiation of RPE cells is a major contributing factor to the pathophysiology of proliferative ocular diseases such as PVR (8). Thus, we sought to understand the molecular mechanisms underlying RPE dedifferentiation and identify potential therapeutics that could inhibit this process. We focused our search to microRNAs because they are important regulators of gene expression and have well established roles in many biological processes including development and differentiation (17, 34 -37). Previously, we and others demonstrated that RPE cells at the free edge of an intact monolayer can proliferate and migrate, giving rise to mesenchymal cells that express low levels of RPE-specific proteins and increased levels of EMT-associated proteins (28,38). Microarray analysis comparing the miRNA profile of these samples with that of differentiated RPE cells revealed that miR-204 and miR-211 are among the most significantly down-regulated miRNAs in RPE dedifferentiation. Because different tissues have unique miRNA profiles that reflect their state of differentiation and functional activity, this finding is consistent with an early study by Wang et al., who demonstrated that miR-204/211 are the two most highly expressed miRNAs in the RPE and are also critical for maintaining its epithelial phenotype and function (16). Here, we extend upon the previously established role of miR-204/211 in maintaining RPE function by demonstrating that miR-204/211 could also direct RPE differentiation. Furthermore, we demonstrate that MITF regulates the transcription of miR-204/211 in the RPE and show for the first time that miR-204/211 act downstream of MITF to promote RPE differentiation.
In addition to miR-204/211, our microarray analysis revealed 49 additional miRNAs that were down-regulated by Ͼ2-fold in dedifferentiated RPE cells (Fig. 1A). Although any one of these miRNAs could potentially have a role in RPE differentiation, the let-7 family of miRNA was of particular interest as many of its members (isoforms a, b, c, d, e, f, and g) were significantly down-regulated in dedifferentiated RPE cells. Let-7 is a marker of cellular differentiation (39) that also has well established functions as a tumor suppressor (40). Earlier studies showed that let-7 inhibits tumor growth by suppressing the expression of high mobility group A2 (41)(42)(43), which induces transcription of two well established regulators of EMT, SNAIL, and TWIST (44,45). Therefore, down-regulation of let-7 and the resultant increase in SNAIL and TWIST expression in RPE cells may contribute to the loss of RPE phenotype. In addition to let-7, miR-26a/b were also down-regulated in dedifferentiated RPE cells. Because miR-26a/b regulate cell cycle progression by targeting genes such as cyclin D2, D3, E1, and E2, and cyclindependent kinases (CDK4 and 6) (46,47), down-regulation of miR-26a/b may also contribute to the increased proliferative potential that is characteristic of dedifferentiated RPE cells. MiR-204 and -211, the two most highly enriched miRNAs in the RPE, were most significantly down-regulated in dedifferentiated hfRPE cells. Because miR-204/211 target EMT-associated genes (SNAI2 and TGFBR2) and are necessary for maintaining RPE function (16), we asked whether they could also direct RPE differentiation. To test this idea, we developed a new model in which we can induce RPE dedifferentiation by subculturing hfRPE cells at low cell density and test whether overexpressing miR-204/211 in these cells could rescue the RPE phenotype.
This model system is based upon the finding that primary RPE cells have a limited number of divisions within which they can return to a differentiated state (48). Thus hfRPE cells seeded above a "threshold" density will differentiate whereas cells seeded below the threshold will dedifferentiate. By varying cell seeding density, we found that hfRPE cells seeded at 30% or higher achieved differentiation whereas hfRPE cells seeded at 15% density or lower resulted in dedifferentiation (supplemental Fig. 1). However, concerns arise when primary cells were seeded at low densities to induce dedifferentiation because these conditions favor the overgrowth of contaminating fibroblasts, which may be a confounding factor in our analysis. To address this issue, we first demonstrated that collagen type I/procollagen is a suitable fibroblast marker by showing that fibroblasts derived from human fetal choroid (the most likely source of contaminating cells) stained positive for collagen type I/procollagen whereas P1 RPE cells on Transwells do not (supplemental Fig. 2, A and B). However, we did find an average of 31 Ϯ 7 randomly scattered collagen I-positive fibroblast cells embedded underneath the RPE monolayer (Ϸ600,000 RPE cells/Transwell) (n ϭ 9). RPE cells seeded at 15% density on Transwells (for 3 days; n ϭ 3 each) had ϳ2-3 collagen I-positive cells (supplemental Fig. 2C). Because fibroblast cells have a doubling time of Ϸ24 h, a small starting number of fibroblasts (1:20,000 RPE cells) could not have overtaken the RPE culture. Consistent with these observations, RT-PCR and Western blot analysis showed that choroid-derived fibroblasts express collagen type I, whereas dedifferentiated RPE cells (from RPE seeded at 1% density) do not (supplemental Fig. 2, D and E), thus confirming that dedifferentiated RPE cells were of RPE origin and that our model system is valid for the study of RPE dedifferentiation. Using this model, we show that hfRPE dedifferentiation caused by seeding at 15% density can be prevented by transfecting with pre-miR-204, -211, or both -204/211, as evaluated by increases in RPE-specific gene and protein expressions, increase in TER, and formation of characteristic RPE morphology (Fig. 2). Of particular importance is the observation that transfecting pre-miR-204 or -211 individually had the same effect on RPE differentiation (mRNA, protein, morphology) as transfecting both pre-miR-204 and -211. In addition, anti-miR-induced loss of RPE phenotype (mRNA, proteins, and morphology) occurred only when cells were transfected simultaneously with both anti-miR-204 and -211, but not individually (Fig. 3). Collectively, these results suggest that miR-204 and -211 are functionally redundant in RPE cells, consistent with the fact that miR-204 and -211 possess an identical seed sequence and therefore have the same target genes. Using in silico computational programs, we obtained a list of potential miR-204/211 target genes. Among them are genes that are commonly associated with EMT, cancer, cytoskeletal rearrangement, proliferation, and survival (Fig. 6). However, the role of these genes in RPE differentiation is largely unknown. Future work will include using high throughput miRNA target validation methods such as RIP-ChIP (49) to verify these miR-204/211 targets and using a customized siRNA library to determine the functional role of these genes in RPE differentiation.
To understand how RPE dedifferentiation can lead to downregulation of miR-204/211, we investigated upstream mechanisms that regulate miR-204/211 expression. MITF plays a key role in the differentiation of melanocytes and pigmented epithelial cells by regulating transcription of genes involved in melanogenesis such as tyrosinase (TYR) and tyrosinase-related protein 1 (TYRP1) (50,51). Mice homozygous for mutant MITF are completely white and have underdeveloped eyes (microphthalmia) in which the RPE transdifferentiates into neural retina (24,25). In addition to TYR and TYRP1, MITF also regulates expression of TRPM1, which hosts primary miR-211 (pri-miR) within its sixth intron. Coincidentally, the primary miR-204 hairpin sequence lies within the sixth intron of TRPM3. Earlier studies showed that miR-204 is co-expressed with TRPM3 mRNA in the choroid plexus (52), a cerebral spinal fluid secreting tissue in the brain that, like the RPE, is derived from the neural ectoderm. In hfRPE cells, we showed that MITF KD resulted in significant decreases in TRPM1, TRPM3, and miR-204/211. This effect was accompanied by a significant downregulation of RPE-specific genes and a dramatic change in morphology from a polarized epithelial monolayer to multilayered mesenchymal cells. Importantly, co-transfection of pre-miR-204/211 into MITF KD cells prevented RPE dedifferentiation, indicating that MITF-mediated regulation of miR-204/211 expression is critical for RPE differentiation.
In conclusion, our results suggest targeted expression of miR-204/211 in RPE cells may be an effective preventive strategy for diseases that involve degeneration and dedifferentiation of RPE cells, such as age-related macular degeneration and proliferative vitreoretinopathy. With advances in adeno-associated virus (AAV) vector-based transgene delivery to specific tissues in the eye (for review, see Ref. 53) and the recent successes in AAV-mediated therapy for patients with Leber congenital amaurosis (54 -57), developing an AAV-based miR-204/211 expression vector that specifically target RPE cells may be a viable strategy against ocular diseases that involve RPE dedifferentiation and loss of epithelial phenotype and function.  . miR-204/211 target genes are involved in various cellular functions. miR-204/211 targets were obtained from TargetScan, miRanda, PicTar, and miRDB. These targets were classified into genes that are known to be involved in Wnt signaling, proliferation and survival, cytoskeletal rearrangement, metabolism, cancer, and EMT. Experimentally confirmed miR-204/211 targets are marked with an asterisk.