Pluripotent Stem Cell Model Reveals Essential Roles for miR-450b-5p and miR-184 in Embryonic Corneal Lineage Specification§


  • Author contributions: R.S.-F.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; L.S.: collection and assembly of data and data analysis and interpretation; S.D.L.F.D., I.P., E.A., L.C., and C.G.: collection and assembly of data; O.D., C.V., and A.S.: provision of materials; J.I.-E.: financial support and administrative support; S.A.: provision of materials and final approval of manuscript; D.A.: conception and design, data analysis and interpretation, manuscript writing, and final approval of manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS February 24, 2012.


Approximately 6 million people worldwide are suffering from severe visual impairments or blindness due to corneal diseases. Corneal allogeneic transplantation is often required to restore vision; however, shortage in corneal grafts and immunorejections remain major challenges. The molecular basis of corneal diseases is poorly understood largely due to lack of appropriate cellular models. Here, we described a robust differentiation of human-induced pluripotent stem cells (hiPSCs) derived from hair follicles or skin fibroblasts into corneal epithelial-like cells. We found that BMP4, coupled with corneal fibroblast-derived conditioned medium and collagen IV allowed efficient corneal epithelial commitment of hiPSCs in a manner that recapitulated corneal epithelial lineage development with high purity. Organotypic reconstitution assays suggested the ability of these cells to stratify into a corneal-like epithelium. This model allowed us identifying miR-450b-5p as a molecular switch of Pax6, a major regulator of eye development. miR-450b-5p and Pax6 were reciprocally distributed at the presumptive epidermis and ocular surface, respectively. miR-450b-5p inhibited Pax6 expression and corneal epithelial fate in vitro, altogether, suggesting that by repressing Pax6, miR-450b-5p triggers epidermal specification of the ectoderm, while its absence allows ocular epithelial development. Additionally, miR-184 was detectable in early eye development and corneal epithelial differentiation of hiPSCs. The knockdown of miR-184 resulted in a decrease in Pax6 and K3, in line with recent findings showing that a point mutation in miR-184 leads to corneal dystrophy. Altogether, these data indicate that hiPSCs are valuable for modeling corneal development and may pave the way for future cell-based therapy. STEM CELLS 2012;30:898–909


The cornea, which is the outermost tissue of the eye, plays an essential optic role and also serves as a barrier against external insults. Like the epidermis, the corneal epithelium originates from the ectoderm, which is hallmarked by the expression of cytokeratin pair K8/K18. At day 9.5 of mouse embryogenesis (E9.5), the expression of Pax6, a master regulator of eye development, becomes restricted to a narrow zone at the head surface ectoderm that is committed to eye development, while the adjacent surface ectoderm in which Pax6 expression is switched off will give rise to the epidermis. These Pax6-positive cells thicken and form the lens placode structure that invaginates and gives rise to the lens and corneal epithelium [1]. Further commitment of the ectoderm into the epithelial lineages (e.g., epidermal, oral, and corneal epithelium) is orchestrated by P63 and is hallmarked by the substitution of K8/K18 by cytokeratins of epithelial progenitors (K5/K14), at E11-12. Corneal epithelial-specific cytokeratins are K12 and K3, which are firstly detected around E15.5 and E17.5, respectively [2].

Pax6 deficiency results in no eye development [3], while Pax6 haploinsufficiency and ectopic expression of Pax6 result in severe eye and corneal defects [4, 5], reminiscent of the phenotype of aniridia patients that suffer from a mutation in the PAX6 allele [6, 7]. Interestingly, rabbit corneal epithelium transdifferentiated into a hair-forming epidermis when transplanted on mouse dermis, and this sequential process was accompanied by loss of Pax6 expression [8, 9]. Similarly, a reduction in Pax6 expression was observed in corneal diseases in which the cornea gains epidermal characteristics and becomes opaque [10]. Altogether, these accumulating data supported the hypothesis that Pax6 is directing the choice between epidermal and corneal fate; however, the mechanism of Pax6 deregulation remains elusive.

Eye development and homeostasis are regulated by microRNAs (miRNAs) [11–15], a group of small noncoding RNA molecules. However, no information is available regarding the expression or function of miRNAs during human corneal embryogenesis. Recently, a point mutation at the seed sequence of miR-184 was linked with eye syndromes affecting the cornea and the lens [16, 17]. However, the exact stage and mechanism by which miR-184 acts remain unknown.

Corneal diseases affect more than 10 million people worldwide [18]. The molecular basis for the initiation and progression of those diseases is largely unknown. New models are required for developing diagnostic tools and specific drugs for a variety of inherited and acquired corneal diseases. Corneal transplantation is common and on the whole successful. However, a shortage in corneal grafts limits treatment of corneal diseases, and as the graft is allogeneic, immune rejection always remains a risk [19]. Despite topical immunosuppression, corneal allograft rejection is common in the long-term [19]. Therefore, alternative cell sources for autologous therapy are being widely investigated [20–23], with only partial success to date.

Human pluripotent stem cells, either embryonic stem cells (hESCs) or reprogrammed induced pluripotent stem cells (hiPSCs), are powerful cellular models to recapitulate in vitro embryonic events and could provide cellular sources for therapy, with hiPSCs derived from patients as autologous alternatives. Previous studies have shown that hESCs can partially differentiate into corneal epithelial-like cells [24, 25]. Here, we designed a robust protocol that recapitulates the canonical steps of corneal epithelial embryogenesis from hair follicle (HF)-derived and skin fibroblasts-derived hiPSC lines. It gave rise to a relatively pure population of differentiated corneal epithelial-like cells. This cellular model was further used to identify miRNAs profiling during corneal epithelial differentiation and revealed new roles for two miRNAs in corneal embryonic lineage.


Tissue Culture

Unless indicated otherwise, reagents were from Invitrogen. All cultures were in 37°C, 5% CO2, and humidified incubator. hiPSCs and hESCs (H9) were grown on mitomycin-treated (Sigma, 8 μg/ml, 3 hours) mouse embryonic fibroblasts (MEFs) (40,000 cells per square centimeter on gelatin [0.1%]) in hiPSC medium (85% Dulbecco's modified Eagle's medium [DMEM]/F12 [1:1], 15% serum replacement,1% nonessential amino acids, 0.1 mM β-mercaptoethanol, and 8 ng/ml basic fibroblast growth factor). Medium of human corneal epithelial (HCE) cell line contained DMEM/F12 (1:1), 5% fetal calf serum (FCS), 5 μg/ml insulin, 0.5% dimethyl sulfoxide, and 10 ng/ml epidermal growth factor (EGF). Epithelial media contained 60% DMEM+GlutMax, 30% F12, 10% fetal clone II, 5 μg/ml insulin, 0.5 μg/ml hydrocortisone, 10 ng/ml EGF, 0.2 mM adenine, and 10 mM choleratoxin. Other cells were grown in DMEM supplemented with 10% FCS. Keratinocytes were isolated by trypsin treatment of HFs and grown in coculture on mitomycin-treated NIH3T3 cells in epithelial medium. Fibroblasts were isolated by as previously described [26, 27].

Generation of hiPSCs

Lentiviral infections for hiPSC generation were performed as previously described [27]. Emerging colonies with pluripotent stem cell morphology were manually isolated and cocultured on mitomycin-treated MEF in 24-well plates. Some three to four colonies were amplified and characterized for the expression of pluripotent markers by real-time polymerase chain reaction (PCR) analysis and immunofluorescent staining, and their differentiation potential was tested by the formation of embryoid bodies followed by real-time PCR analysis of various markers of embryonic lineages [27].

Corneal Differentiation and Transfection During Differentiation

Conditioned medium (CM) was prepared by treating confluent corneal fibroblasts (CFs) or limbal fibroblasts (LFs) cultures with mitomycin (8 μg/ml, 3 hours), the next day washed with phosphate buffered saline (PBS), and epithelial media (15 ml) was added. CM (15 ml) was collected every day for 10 days and was kept in −20°C if necessary for up to 1–2 months or at 4°C up to 7 days before filtering and using for corneal differentiation. Corneal differentiation was induced by seeding hiPSCs/hESCs (1:2) on collagen IV (Sigma)-coated dishes [24], in CF-CM or LF-CM that was replaced every second day. Cells were trypsinized if necessary.

For transfections during differentiation, trypsinized cells were resuspended in CFs-CM and reseeded (1:2) on collagen IV-coated plates. Transfections were performed the next day or following cell adhesion (4–6 hours) by Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer recommendations using 2 μg K12-GFP plasmid or 50 μM pre-miR, or control oligos, or anti-miR (Ambion, Saint Aubin, France).

Western Blots

Cell were lysed in RIPA buffer (TrisHCl 10 mM, 10 mg/ml deoxycholate, 1% NP40, 1% SDS, 150 mM NaCl, and protease inhibitors cocktail [Roche, Paris, France]) and subjected to SDS-polyacrylamide gel electrophoresis, followed by immunoblotting. Antibodies used were mouse anti-K3 (1:300) (Millipore, Molsheim, France), mouse anti-K12 (1:20) (a kind gift from Danielle Dhouailly), rabbit anti-Pax6 (1:300) (Chemicon, Paris, France), rabbit anti-Connexin43 (1:500) (Thermo Scientific, Auburn, AL), rabbit anti-Oct4 (Santa Cruz), rabbit anti-Nanog (1:500) (R&D, Lille, France), and mouse anti-β-tubulin (1:1,000) (Santa Cruz, Heidelberg, Germany). Proteins were visualized by enhanced chemiluminescence (Amersham Pharmacia Biotech, Piscataway, NJ) and quantified by densitometry with Image Master VDS-CL using TINA 2.0 software (Ray Tests).

Quantitative Real-Time PCR and TaqMan Assay for miRNAs

cDNA was prepared from 1 μg RNA (Trizol extraction) using iScript Reverse Transcriptase kit (BioRad, Marnes-la-Coquette, France). Each reaction of real-time PCR contained 12.5 μl SYBR-Green PCR Master Mix (Applied Biosystems, Carlsbad, CA), 0.125 μl MultiScribe Reverse Transcriptase (Applied Biosystems, Carlsbad, CA), 5 μl cDNA, and 5 μl primer mix (0.5 μM, sequences available in Supporting Information Table S3), adjusted to 25 μl reaction volume. For TaqMan assays of miRNAs, 5 μl RNA 5 ng/μl was subjected to PCR using the reverse transcription kit and miRNA-specific primers (Applied Biosystems, Carlsbad, CA). Reaction products were further subjected to real-time PCR using TaqMan universal master mix and TaqMan miRNA-probes or U54 as control (Applied Biosystems, Carlsbad, CA). The relative amounts of each mRNA or miRNA was normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) or U54 reaction, respectively, and the relative expression of each reaction was calculated as a fold change relative to control sample. Wide miRNA profiling was performed by TaqMan low density miRNA arrays (Applied Biosystems, Carlsbad, CA), according to the manufacturer's instructions. The results were normalized using expression of small nucleolar RNA, according to the manufacturer's protocol.

Immunofluorescent Staining

Immunofluorescent staining were performed as previously described [28, 29]. The following primary antibodies were used: rabbit anti-K14 (1:200) (Covance, Paris, France), rabbit anti-K5 (1:200) (Covance, Paris, France), mouse anti-K18 (1:200) (Chemicon, Paris, France), rabbit anti-Pax6 (1:100) (Chemicon, Paris, France), mouse anti-K3 (1:100) (Millipore, Molsheim, France), mouse anti-P63 (1:100) (Santa Cruz, Heidelberg, Germany), and mouse anti-E-Cadherin (1:100) (R&D, Lille, France).

In Situ Hybridization

In situ was performed as detailed previously [30] with minor modifications. Frozen sections were fixed (15 minutes, 4% PFA [Electron Microscopy Sciences]), followed by PBS wash (all washes described below were 3 × 5 minutes at room temperature), acetylation (15 minutes), PBS wash, proteinase K treatment (5 minutes [Roche, Paris, France]), PBS wash, prehybridization (1 hour), and hybridization with locked nucleic acid (LNA) detection probes (5′/3′-Dig labeled (Exiqon), at 54°C and 44°C for miR-184 and miR-450b-5p, respectively). The next day, samples were washed (30 minutes, 54°C, 5× saline-sodium citrate (SSC) [BIO-LAB, Jerusalem, Israel]), washed (1 hour, 60°C, 0.2× SSC), washed (10 minutes, TBS [Bio-Rad, Marnes-la-Coquette, France]), incubated for 30 minutes (3% H2O2), washed (TBS supplemented with 0.1% Tween 20 [Sigma, Paris, France]), blocked (10% naïve goat serum [Biological Industries, Beit Ha-Emek, Israel]), incubated with anti-Dig antibody ([1:500], 1 hour, in blocking solution) followed by tris-buffered saline and Tween 20 (TBST) wash, tyramide signal amplification step (Perklin Elmer) for 10 minutes, TBST wash, and mounting in Dapi mounting gel (Sigma, Paris, France).

Cloning and Luciferase Assay

To clone K12 promoter region (1,390 bases upstream to K12 start codon) downstream to green fluorescent protein (GFP)-encoding sequence, mouse DNA was used as template for PCR using specific primers (5-cttacggaggcctgccatg-3, 5-gctggtatgccagaagggg-3). The amplified fragment was inserted into StuI and AgeI sites of pEGFP-1 plasmid (Clontech, Mountain View, CA). Cells were cotransfected with 1.5 μg K12-GFP plasmid and 50 μM pre-miR-450b-5p or control oligos. Transfection efficiency was determined separately by pEGFP transfection. Transfectants were subjected to fluorescent microscopy and data were normalized and K12-GFP expression was calculated as the ratio between the percentages of K12-GFP positive among transfectants (percentage of pEGFP positive).

For generating LUC-Pax6 plasmid, the 3′untranslated region (UTR) region of PAX6 gene was amplified (primers used: 5′-tatactcgagaaaggaaaggaaauauuguguuaauuc-3′, 5′-tatagcggccgcaaggttaaaaagaaaactgtataataaa-3′) and cloned downstream to luciferase-encoding sequence at xhoI/NotI sites of psiCHECH2 plasmid. Site-directed mutagenesis allowed the generation of LUC-ΔmiR-450b-5p, a mutated form of LUC-Pax6 plasmid in which the miR-450b-5p binding site is disrupted, by PCR (primers used: 5′-aacagatgaagtttatgtaacgaaaagggtaaga-3′, 5′-tcttacccttttcgttacataaacttcatctgtt-3′) using LUC-Pax6 plasmid as template. Luciferase activity was determined using luciferase kit (Promega, Paris, France) in lysates of 293 human embryonik kidney (HEK) cells that were cotransfected with 0.3 μg LUC-Pax6 or LUC-ΔmiR-450b-5p and pre-miR-450b-5p or control oligos (Ambion) by Fugene reagent (Invitrogen, Saint Aubin, France).

Organotypic Reconstitution of Corneal Equivalent Tissue In Vitro

Primary CFs were embedded in gels that were prepared in 24-well plates [31], incubated 4 hours at 37°C, followed by reseeding trypsinized hESCs that were differentiated for 7 days in six-well plates, on these gels. The next day, air-liquid interphase was induced for 8 days [31] followed by tissue sectioning and staining as detailed above.


Pluripotent Stem Cell Differentiation Recapitulated Corneal-Epithelial Commitment

HF keratinocytes and dermal fibroblasts (DFs) were isolated and reprogrammed by infection with a lentiviral vector containing a polycistronic cassette of OCT4, SOX2, CMYC, and KLF4 (Fig. 1A). As demonstrated by real-time PCR analysis (Fig. 1B) and immunofluorescent staining (Fig. 1C), we selected individual clones of reprogrammed cells that displayed typical hiPSC morphology and expressed markers of pluripotency. The ability of hiPSCs to differentiate into the three embryonic germline lineages was confirmed by the formation of embryoid bodies followed by inspection of the expression of various markers of the endoderm, ectoderm, and mesoderm (Fig. 1D and [27]).

Figure 1.

Generation of hiPSCs from HF keratinocytes and DFs. (A): Schematic representation of hiPSCs generation from HF keratinocytes and DFs. An example of real-time polymerase chain reaction (PCR) analysis (B) or immunofluorescent staining and AP assay (C) that were performed for screening the expression of the indicated markers of individual clones that were identified as hiPSCs. (D): To assess pluripotency of colonies, hiPSCs were subjected to differentiation through formation of embryoid bodies as described in Materials and Methods, followed by RNA extractions and real-time PCR analysis of endodermal (AFP), ectodermal (K14), and mesodermal (CD31) markers. Data were normalized to GAPDH and represents the relative fold change in gene expression as compared to undifferentiated hiPSCs. Abbreviations: AFP, alpha-fetoprotein (AFP); AP, alkaline phosphatase; DF, dermal fibroblast; hESCs, human embryonic stem cells; HF, hair follicle; hiPSCs, human-induced pluripotent stem cells.

Previously, corneal epithelial differentiation was demonstrated by seeding hESCs on collagen IV-coated dishes in LF-CM [24]. This method was relatively efficient, resulting in some 60% of the cells expressing K3, a marker of terminally differentiated corneal epithelial cells. However, the expression of limbal stem/progenitors cell markers (e.g., P63 and K14) was highly variable between the pluripotent stem cell lines used ([27] and not shown). We challenged human iPSCs to corneal fate by introducing two major modifications on the protocol designed for hESCs. Instead of using epithelia medium conditioned by LFs that are isolated from the limbus, a narrow band of tissue of limited amount, we used CFs isolated from the entire cornea including the corneal periphery. In addition, we tested the effect of bone morphogenetic factor 4 (BMP4), a major regulator of embryonic epithelial commitment [32, 33]. Accordingly, hiPSCs that were derived from HF cells (HF1-hiPSC) were seeded on collagen IV-coated dishes in the presence of CF-CM and were treated with BMP4 (0.5 nM) during restricted periods of differentiation (days 0–3 or 4–8 or 9–12 or for the whole period of 12 days of differentiation) (Fig. 2B, 2D). At day 12, cells were harvested and subjected to real-time PCR, immunofluorescent staining, or Western blotting. Among all conditions tested, addition of BMP4 during the first 3 days efficiently enhanced corneal epithelial differentiation, as illustrated by increased expression of markers of corneal progenitors, ΔNP63 and K14 (Fig. 2B, 2C), and markers of terminally differentiated corneal epithelial cells, K12 (Fig. 2D) and K3 (Fig. 2E), while constant BMP4 treatment seemed unnecessary. Corneal epithelial differentiation was completely abrogated in the presence of the BMP antagonist, LDN-193189 (LDN) (Fig. 2B, 2D), indicating that endogenous BMP signaling is essential for corneal epithelial commitment of hiPSCs. As shown in Supporting Information Figure S1, epithelial media that was conditioned by either LFs or CFs induced corneal epithelial lineage commitment of hiPSC to a similar manner.

Figure 2.

Differentiation of hiPSCs into corneal epithelial cells. (A): Schematic representation of culture method. hiPSCs were seeded on collagen IV-coated dishes in the presence of corneal fibroblast-conditioned media (CF-CM) and supplemented with BMP4 at days 0–3, as described in Materials and Methods. (B, D): Corneal differentiation was performed on collagen IV and CF-CM. BMP4 or LDN (a BMP inhibitor) (+) or vehicle (−) were supplemented during the indicated periods. Real-time polymerase chain reaction analysis of the indicated genes (B, D) was performed at day 12. Immunofluorescent staining (C) and Western blotting (E) of the indicated proteins were performed at day 12 of differentiation in the presence of BMP4 (supplemented at days 0–3) or vehicle (Ctl). Scale bar in (C) = 20 μm. Abbreviations: BMP, bone morphogenetic factor; LDN, LDN-193189; hiPS, human-induced pluripotent stem.

To follow corneal epithelial commitment, corneal epithelial lineage gene expression was recorded at different time points during HF1-hiPSCs differentiation by real-time PCR analysis (Fig. 3A). Elevation in ectodermal marker K18 (and K8, not shown) appeared already within 2–4 days, along with the expression of Pax6 (Fig. 2A), reminiscent of their coexpression in vivo in the embryonic lens placode cells. Markers of corneal epithelial progenitors emerged at day 8 by the expression of ΔNP63 and K14, while markers of terminally differentiated corneal epithelium (K3, K12) peaked at day 14 (Fig. 3A). This method was efficient for hESCs and for hiPSCs derived from HFs and DFs with relatively minor variability (Supporting Information Fig. S2A). At day 8, the vast majority of cells expressed K18, while few cells were already positive for K5, an epithelial progenitor cell marker (Fig. 3B). In addition, first sign of corneal epithelial differentiation was already evident at day 8 by the coexpression of K3 and K14. Increasing expression of K5, K14, and K3 was observed by day 14 (Fig. 3B). Typical corneal epithelial characteristic was also evident by the expression of E-Cadherin, K3, and Pax6 (Fig. 3C). This data were confirmed by Western blot analysis, showing that K3, K12, and Connexin 43 increased during differentiation, while Oct4 and Nanog declined and disappeared within 6–8 days of differentiation (Supporting Information Fig. S2B). The robustness of corneal epithelial-like cell production was quantified at day 14 by fluorescence-activated cell sorting (FACS) analysis. As shown in Figure 3D, some 20%–25% of the cell population was K14-positive progenitor cells, while most cells expressed K3 (>90%). We further generated a GFP-K12 construct carrying the corneal epithelial-specific promoter of K12 upstream to the green fluorescent protein (GFP) encoding sequence. Specificity of the construct was evaluated by transfecting different cell types with the K12-GFP construct followed by FACS analysis. As demonstrated in Figure 3E, the K12 promoter was silent in noncorneal cells (HeLa, HaCaT, and undifferentiated hESCs) but active in corneal epithelial cell line (HCE), and significantly expressed by hESCs, which were differentiated as described for 12 days.

Figure 3.

Human-induced pluripotent stem cell (hiPSC) differentiation recapitulated major steps of corneal epithelial embryogenesis. (A--D): Corneal epithelial differentiation of hair follicle-derived hiPSCs was performed as in Figure 1A, cells were harvested at the indicated days and subjected to real-time polymerase chain reaction analysis (A), immunofluorescent staining (B, C), or flow cytometer analysis (D) of the indicated markers. (E): The indicated cell lines or hiPSCs that were differentiated for 12 days (Day 12) were transfected with K12-GFP plasmid followed by flow cytometry, as described in Materials and Methods. Data represent the percentage of K12-GFP-positive cells among transfectants (calculated as described in Materials and Methods). (F): Organotypic reconstitution of pluripotent stem cell-derived cells followed by immunofluorescent staining of Pax6 and K3 (F). Inset shows enlargement of the annotated region. Scale bar in B, C, F = 20 μm. Abbreviations: GFP, green fluorescent protein; HCE, human corneal epithelial.

To test the ability of these cells to stratify, pluripotent cells were differentiated into corneal epithelial fate for a week and then trypsinized and seeded on the top of a stroma equivalent made of CF embedded into collagen gel for a three-dimensional organotypic reconstitution assay. As illustrated in Figure 3F, differentiated cells were able to stratify and form two to three layers of cells coexpressing K3 and Pax-6.

miRNA Profiling of hESCs During Corneal Epithelial Differentiation

Although miRNAs are essential for eye development and corneal integrity [3, 14, 17], the distribution and function of human miRNAs during embryonic commitment into the corneal epithelial lineage are unknown. We thus performed a wide miRNA profiling during corneal epithelial differentiation of pluripotent stem cells at days 2, 6, and 10 (Fig. 4A). The array data were confirmed by TaqMan assays for four miRNAs using the same RNA samples that were used for the array (Fig. 4B) and also by examining RNA preparations from two additional independent experiments (not shown). In agreement with the array data, the levels of miR-24, miR-27b, and miR-184 gradually increased during corneal epithelial differentiation, while the levels of miR-450a decreased at days 2 and 6 and then increased at day 10 (Fig. 4B), thus confirming the reliability of the array data. Comparable expression profiles were shown for these four miRNAs in corneal epithelial differentiation of two clones of hiPSCs derived from HFs (Supporting Information Fig. S3 and not shown). Next, we compared our data with a list of 21 miRNAs that were previously identified as miRNAs with the highest expression levels in adult mouse corneal epithelium [34]. Notably, 86% of these miRNAs was significantly induced (>2) at day 10 of corneal epithelial differentiation (Fig. 4A), suggesting that miRNA profile of pluripotent stem cell-derived corneal epithelial-like cells shares a high degree of similarity with that of corneal epithelial cells in vivo.

Figure 4.

MicroRNA (miRNA) profiling during human embryonic stem cell (hESC) differentiation into corneal epithelial lineage. (A): miRNA array was performed using RNA samples obtained from hESC differentiation for the indicated days. Heatmap presentation shows the fold change in the expression of 21 miRNAs that were previously found to have the highest levels in the adult mouse corneal epithelium [34], listed in a descending order according to their expression. (B): TaqMan assay using specific probes for the indicated miRNAs. Data represent the relative fold change of each miRNA as compared to the levels in undifferentiated hESCs (A, B).

miR-450b-5p Represses Pax6 Expression by Binding to Its 3′-Untranslated Region

Pax6 is a major corneal transcription factor and mutations in the PAX6 gene are associated with aniridia, a rare eye dystrophy affecting the iris and cornea, leading to blindness [35]. Interestingly, inductions of Pax6 protein (Fig. 5A) and mRNA (Fig. 5B) levels were shown at day 2 of differentiation, peaked around day 6, and then moderately dropped down at day 14. Surprisingly, miRNAs related to Pax6 regulation have not been described yet. We thus used the cellular model described here to identify miRNAs that may potentially inhibit Pax6 expression. In silico analysis of the 3′-untranslated region (3′UTR) of Pax6 messenger RNA by “TargetScan” software ( allowed the prediction of 51 potential binding sites for different miRNAs. We next screened the expression pattern of the 51 candidates in our array data to identify miRNAs that display reciprocal expression profile relative to Pax6. miR-450b-5p was clearly positive for this criterion (Fig. 5B). The match between the seed sequence of miR-450b-5p and positions 570-576 of Pax6 3′UTR and to upstream sequences is illustrated in Figure 5C. The sequence of the 3′UTR of Pax6 gene is highly evolutionarily conserved among vertebrates, and the seed sequence of miR-450b-5p is conserved between mouse and human but not all mammals (Supporting Information Table S1).

Figure 5.

Inverse correlation in the expression of miR-450b-5p and Pax6. hESCs were differentiated into corneal epithelial fate for the indicated periods followed by Western blotting of Pax6 and β-tubulin (β-tub) (A), real-time PCR analysis of Pax6 (B), or TaqMan assay of miR-450b-5p (B). (C) In silico analysis (by TargetScan ( showing the predicted binding of miR-450b-5p (seed sequence in red) and Pax6-3′UTR. Below is the mutated form of Pax6-3′UTR that is described in Figure 6A, 6B (mutated sequence is underlined). (D): In situ hybridization of miR-450b-5p using cryosections of mouse embryos at E14.5. Basement membrane is annotated by a dashed line. Right inset is an enlargement of the dashed region, showing a staining of miR-450b-5p in the corneal stroma but not epithelium. Below (left) is immunofluorescent staining of Pax6 in an adjacent section showing that Pax6 expression is restricted to the presumptive conjunctiva and cornea. White arrows indicate positive staining of miR-450b-5p and no staining of Pax6 in the epidermis. Red arrows demonstrate positive staining of Pax6 and negative staining of miR-450b-5p at the conjunctiva and ocular surface. Scale bar = 50 μm. Abbreviations: el, eye lid; ep, corneal epithelium; st, stroma.

At E14.5, Pax6 was expectedly detected in the ocular surface epithelium and absent in the adjacent epidermis of the eye lids (Fig. 5D, left). Remarkably, in situ hybridization revealed that miR-450b-5p was distributed throughout the presumptive epidermis, at the surface ectoderm that is negative for Pax6, while it was absent in the ocular regions that are Pax6 positive (Fig. 5D). It was difficult to specifically detect miR-450b-5p at earlier embryonic stages. This reciprocal expression could suggest that Pax6 repression by miR-450b-5p in the presumptive epidermis prevents ocular epithelial commitment of the ectoderm.

To further test this hypothesis, we examined whether miR-450b-5p can bind to and repress Pax6. We cloned the 3′UTR region of Pax6 (1,023 bases) downstream to a luciferase gene as illustrated in Figure 6A. Additionally, by site-directed mutagenesis, we generated a mutated construct (ΔmiR-450b-5p) in which the binding site of miR-450b-5p was disturbed (illustrated in Fig. 6A and underlined mutated sequence in Fig. 5C). 293HEK cells were cotransfected with the wild-type construct (LUC-Pax6) or the mutated construct (LUC-ΔmiR-450b-5p) along with increasing concentrations of pre-miR-450b-5p or with scrambled oligonucleotide as control. As shown in Figure 6B, a significant and dose-dependent decrease in the luciferase activity was observed in the presence of pre-miR-450b-5p. This effect was specific to pre-miR-450b-5p as cotransfection with pre-miR-450a, which is clustered with miR-450b-5p and shares high sequence similarities with miR-450b-5p but yet contains altered seed sequence, had no such effect (Fig. 6B). Moreover, this inhibitory effect of miR-450b-5p on luciferase activity was dependent on the integrity of the predicted binding site, as miR-450b-5p had almost no effect on the mutated luciferase construct (Fig. 6B). This data indicated that miR-450b-5p can specifically bind to the 3′UTR of Pax6 and may inhibit Pax6 expression.

Figure 6.

miR-450b-5p represses Pax6 by binding to the 3′UTR. (A): Schematic representation of cloning into luciferase vector of the wild-type Pax6-3′UTR (1,023 bases length) and a mutated form (referred as ΔmiR-450b-5p, sequence indicated in Fig. 5C) in which the miR-450b-5p binding site (positions 570-576) was disrupted. (B): Luciferase assay of 293HEK cells cotransfected with a luciferase construct containing wild-type Pax6-3′UTR (LUC-Pax6) or its mutated form (LUC-ΔmiR-450b-5p) and increasing concentrations (25–50 nM) of pre-miR-450b-5p or pre-miR-450a (50 nM) or scrambled oligos as control. Data represent the relative fold change in luciferase activity as compared to control transfectant. (C): Western blot analysis of Pax6 and β-tubulin (β-tub) in HCE cells transfected with pre-miR-450b-5p (PM) or anti-miR-450b-5p (AM) or control sequence (Ctl). hESCs were transfected at day 1 of corneal differentiation, harvested at day 5, and subjected to Western blot analysis of Pax6 and β-tubulin (D). (E): hESCs were differentiated for 7 days and cotransfected with K12-GFP plasmid and pre-miR-450b-5p or control oligos followed by flow cytometry at day 10. Data represent the percentage of K12-GFP-positive cells among transfectants (calculated as described in Materials and Methods). Abbreviations: HCE, human corneal epithelial; hESC, human embryonic stem cell; UTR, 3′-untranslated region (3′UTR).

To further assay the effect of miR-450b-5p on PAX6 gene expression, we transfected HCE cells by 50 nM anti-miR-450b-5p (AM), pre-miR-450b-5p (PM), or control oligos (Ctl) followed by Western blotting (Fig. 6C). Transfection was confirmed by TaqMan assay coupled with real-time PCR analysis (Supporting Information Fig. S3A). As shown in Figure 6C, Pax6 expression significantly decreased in the presence of pre-miR-450b-5p, and an inverse effect was shown in the presence of specific antago-miR.

We next addressed the question of whether the decrease in miR-450b5p during early corneal epithelial differentiation (Fig. 5B) was directly linked to an increase in Pax6 protein (Fig. 5A, 5B). To prevent the decline of miR-450b-5p during commitment, hESCs were transfected by pre-miR-450b-5p at day 1 of corneal epithelial commitment (transfection validation is shown in Supporting Information Fig. S3B), and Pax6 expression was determined by Western blot analysis. As expected, Pax6 expression was attenuated by the presence of pre-miR-450b-5p (Fig. 6D), confirming that Pax6 is a target of miR-450b-5p, and indicating that the decrease in the levels of miR-450b-5p is contributing to the induction in Pax6 protein expression during early neuroectodermal commitment of hESCs [36].

The in situ hybridization analysis shown in Figure 5D could suggest that by repressing Pax6 at the presumptive epidermis, miR-450b-5p was preventing the ocular epithelial fate. To further test this hypothesis, cotransfection of K12-GFP and pre-miR-450b-5p or control oligos was performed at day 7 of corneal epithelial differentiation, and cells were analyzed at day 10 of differentiation. As shown in Figure 6E, transfection of pre-miR-450b-5p significantly reduced K12 promoter activity. Altogether, these data indicate that miR-450b-5p inhibits Pax6 expression and regulates corneal epithelial fate.

miR-184 Is Essential for Corneal-Epithelial Lineage Commitment

miR-184 showed the highest hybridization signal among all adult mouse corneal epithelial miRNAs listed in vivo [34]. In silico analysis revealed that the mature sequence of miR-184 is extremely conserved in vertebrate's evolution (Supporting Information Table S2). miR-184 expression, which gradually elevated concomitantly with markers of epithelial progenitors at days 6–8, markedly increased (>100-fold increase) during terminal differentiation stages (Fig. 4A, 4B). Previously, miR-184 was detected in the corneal epithelium at postnatal stages [13]. In situ hybridization performed at E11.5 revealed that miR-184 is detectable at low levels in the head surface ectoderm and at higher levels in the newly formed lens vesicle. The expression of miR-184 in advanced embryonic stages was punctuated or variable (Fig. 7A). At day 8 after birth (P8), prior to eye lid opening and first corneal stratification, miR-184 was clearly detected in the corneal epithelium, which comprises one to two layers, and became restricted to the corneal epithelial basal layer following stratification (at P14, not shown). Intense miR-184-signal was observed in the lens, while unclear signals were occasionally found in the corneal stroma or endothelium at P8 (Fig. 7A). Remarkably, miR-184 was also expressed in the epidermis and in late anagen-staged HFs (Fig. 7A). No significant staining was shown with scrambled probes (not shown). Altogether, this expression profile suggested that miR-184 plays a role in early eye development in the lens and the cornea, in line with the early onset of severe corneal and lens abnormalities in patients with miR-184 mutation [16, 17]. In addition, the specific expression of miR-184 in the skin and HF suggests that it plays a general role in ectodermal derivatives.

Figure 7.

Contribution of miR-184 to corneal epithelial differentiation of human embryonic stem cells (hESCs). (A): In situ hybridization of miR-184 was performed using cryosections of mouse tissues of the indicated stages. Positive staining at the developing ocular surface is highlighted by arrows. Scale bar = 50 μm. (B): hESCs were transfected at day 7 of corneal differentiation with anti-miR-184 oligonucleotides (AM) or scrambled sequences (Ctl), and cells were harvested at day 14 of differentiation (B--D). Transfection validation by TaqMan assay is presented in (B). Immunofluorescent staining (C) and flow cytometer analysis (D) of K3 and Pax6 is shown. Scale bar = 20 μm. Abbreviations: el, eye lid; le, lens; re, retina; se, surface ectoderm; st, stroma.

We next addressed the question of whether the increase in the level of miR-184 contributes to embryonic commitment into the corneal epithelial lineage. hESCs were transfected at day 7 of corneal epithelial differentiation (as described in Materials and Methods) with a synthetic antagonizing anti-miR-184 oligonucleotides or scrambled sequences as control and analysis was performed at day 14. The marked elevation of miR-184 expression during differentiation was significantly attenuated in the presence of antago-miR-184 (AM) (Fig. 7B). Importantly, the knockdown of miR-184 resulted in reduced levels of K3 and Pax6, as demonstrated by immunofluorescent staining (Fig. 7C). In agreement, FACS analysis confirmed that inhibition of miR-184 expression prevented the corneal epithelial fate of committed hESCs (Fig. 7C). This data suggest that miR-184 is essential for corneal epithelial embryogenesis. The temporal concomitant appearance of miR-184 during early epithelial commitment in vitro and in vivo implies that it may first act at this particular developmental stage.


We demonstrated here that pluripotent stem cells are useful for modeling corneal epithelial embryogenesis. By seeding hiPSCs on collagen IV-coated dishes in the presence of CF-CM allowed corneal differentiation that was augmented by the presence of BMP4. Indeed, the major steps of corneal development were recapitulated in vitro and miRNA profiling during embryonic commitment into the corneal epithelial fate was obtained, and allowed the identification of miR-450b-5p and miR-184 as essential for this lineage.

Pax6 must be tightly regulated for appropriate eye development [4–6], while abnormal expression of Pax6 is associated with eye diseases [7, 10]. Therefore, Pax6 regulators have significant importance. In this study, we revealed the first described role for miR-450b-5p, as a Pax6-repressing miRNA. The sharp reciprocal correlation between Pax6 and miR-450b-5p at E14.5 suggested that miR-450b-5p acts as a molecular switch that directs cell fate decision of the ectoderm. Accordingly, miR-450b-5p is repressing Pax6 in the developing epidermis and thereby allowing initiation of epidermal morphogenesis, while the absence of miR-450b-5p is required for Pax6 expression in the developing ocular surface. When so, the knockout of miR-450b-5p at the surface ectoderm could possibly result in ectopic eye formation, while the uncontrolled expression of miR-450b-5p in the ectoderm is expected to interfere with lens placode formation and/or eye development.

Given that Pax6 also regulates early neuroectodermal commitment [36], embryonic and adult neurogenesis [37], and pancreatic β-cell function [38], this regulatory mechanism may have a broader relevance in various tissues. Future studies using genetically modified mice should allow tracing the expression and assessing the roles of miR-450b-5p in different tissues in which it may modulate Pax6. Strikingly, mutations in two rare genetic syndromes [39, 40] were linked with a defined region at chromosome X, which encompasses the locus of MIR450B (Xq26.3). Cerebello-trigeminal-dermal dysplasia is a rare genetic disorder that is accompanied by several defects including mental retardation, corneal opacity, hypoplasia of the cerebellum [39]. The X-linked split-head/split-foot malformation 2 is associated with developmental defects in ectodermal structures and is characterized by the absence of medial digital rays, syndactyly, and median cleft of the hands and feet [40]. It would be therefore essential to examine whether these phenotypes, which could hypothetically be owing to aberrant modulation of Pax6, involve mutation in miR-450b-5p.

Recently, it was shown that Pax6 is involved in early ectodermal cell fate determination [36] in line with the observed increase in Pax6 at early differentiation stages of hiPSCs. Our data indicated that the decrease in miR-450b-5p is an essential molecular switch of Pax6 at early ectodermal commitment of pluripotent stem cells. It is possible that Pax6-repression by miR-450b-5p contributes to the pluripotent state of hiPSCs and its decrease allows the exit from this state, similar to the role of miR-302 that has been shown to be a guardian of pluripotency [41] and ameliorates the efficiency of somatic cell reprogramming [42]. Accordingly, it would be interesting to examine whether pre-miR-450b-5p has similar roles.

Very recently, a point mutation in miR-184 was linked with early onset of eye defects involving the corneal epithelium, corneal endothelium, and the lens [16, 17], in line with the expression pattern of miR-184 by in situ hybridization. Corneal diseases such as keratoconus and limbal stem cell deficiency lack appropriate cellular models to allow the characterization of the etiology and the underlying molecular defects. Our model suggests that mutations in miR-184 in patients could be a primary embryonic defect, affecting normal corneal epithelial commitment. The knockdown of miR-184 resulted in a decrease in Pax6 and K3, which are both important for corneal integrity and their deregulation was associated with corneal diseases [10, 43]. Since the 3′UTRs of K3 and Pax6 do not contain predictable binding sites for miR-184, this effect was most probably indirect. Recent studies have identified genes that are repressed by miR-184. However, the miR-184-binding sites of these four genes identified in the fly [44] and mouse [45]) are not evolutionarily conserved. The stringent conservation of miR-184 among vertebrates suggests that the function of miR-184 in the eye is indispensable for the viability and/or fitness of these species. Additionally, miR-184 regulates neural stem cells [45], germline cells [44], and epidermal cells (our unpublished data). These functions need to be examined in mouse models or human diseases.


We have demonstrated that a highly enriched population of corneal epithelial-like cells could be obtained from pluripotent stem cells of different origins. Corneal therapy by means of hiPSC-derived autologous transplantation may circumvent the major problems of current allogeneic therapies namely graft rejection and a shortage of corneal donors. This remarkable therapeutic potential of hiPSCs must be further investigated in vivo. Moreover, derivation of hiPSCs from somatic cells of patients with defined genetic alterations may provide new models for identifying the abnormal molecular circuitry of a given disease and developing appropriate treatments. Indeed, we have recently generated hiPSCs from patients that have a mutation in P63 gene leading to ectodermal dysplasia syndrome, an epithelial disease affecting the skin and the cornea. This model allowed us to mimic the congenital pathology and to partially rescue the altered phenotype with a small chemical compound (Shalom-Feuerstein et al., manuscript submitted for publication).


We acknowledge Thierry Virolle and Virginie Virolle for the help in miRnome experiment. This work was partially supported by the Agence Nationale de Recherche (ANR-08-GENOPAT-024) and the Israel Ministry of Science and Technology (MOST 3-6494 for DA) to D.A., J.I.E., and A.S., a “Poste Vert” fellowship of INSERM, a “Chateaubriand” fellowship of the Embassy of France in Israel to R.S.F., and a Ph.D. fellowship from the Israeli Ministry of Integration to L.S.


The authors declare no conflict of interest.