MicroRNA-145 Regulates Human Corneal Epithelial Differentiation

Background Epigenetic factors, such as microRNAs, are important regulators in the self-renewal and differentiation of stem cells and progenies. Here we investigated the microRNAs expressed in human limbal-peripheral corneal (LPC) epithelia containing corneal epithelial progenitor cells (CEPCs) and early transit amplifying cells, and their role in corneal epithelium. Methodology/Principal Findings Human LPC epithelia was extracted for small RNAs or dissociated for CEPC culture. By Agilent Human microRNA Microarray V2 platform and GeneSpring GX11.0 analysis, we found differential expression of 18 microRNAs against central corneal (CC) epithelia, which were devoid of CEPCs. Among them, miR-184 was up-regulated in CC epithelia, similar to reported finding. Cluster miR-143/145 was expressed strongly in LPC but weakly in CC epithelia (P = 0.0004, Mann-Whitney U-test). This was validated by quantitative polymerase chain reaction (qPCR). Locked nucleic acid-based in situ hybridization on corneal rim cryosections showed miR-143/145 presence localized to the parabasal cells of limbal epithelium but negligible in basal and superficial epithelia. With holoclone forming ability, CEPCs transfected with lentiviral plasmid containing mature miR-145 sequence gave rise to defective epithelium in organotypic culture and had increased cytokeratin-3/12 and connexin-43 expressions and decreased ABCG2 and p63 compared with cells transfected with scrambled sequences. Global gene expression was analyzed using Agilent Whole Human Genome Oligo Microarray and GeneSpring GX11.0. With a 5-fold difference compared to cells with scrambled sequences, miR-145 up-regulated 324 genes (containing genes for immune response) and down-regulated 277 genes (containing genes for epithelial development and stem cell maintenance). As validated by qPCR and luciferase reporter assay, our results showed miR-145 suppressed integrin β8 (ITGB8) expression in both human corneal epithelial cells and primary CEPCs. Conclusion/Significance We found expression of miR-143/145 cluster in human corneal epithelium. Our results also showed that miR-145 regulated the corneal epithelium formation and maintenance of epithelial integrity, via ITGB8 targeting.


Introduction
In adult tissue, the renewal of epithelium relies on the population of stem cells. They generate transit-amplifying (TA) cells, which proliferate and differentiate to stratified squamous epithelium. Adult stem cells are usually slow-cycling in vivo whereas TA cells are frequently dividing with short cell cycles. When placed in culture, stem cells and TA cells generate holoclones and paraclones respectively [1]. Corneal epithelial progenitor cells (CEPCs) reside in the basal epithelium of limbus which is an annulus located at the vascularized junction between transparent cornea and opaque sclera [2]. They are characterized by a lack of cytokeratin-3/12 and connexin-43, which are corneal differentiation markers [3,4]. They undergo more frequent cell divisions than differentiated epithelial cells and can be cultured ex vivo from limbal tissues [5]. There has been persistent success in clinical application of limbal grafting or autologous limbal culture cells to restore damaged corneal epithelia [6,7].
Epigenetic factors, such as microRNAs, are known to affect stem cell biology, including the maintenance of pluritotency and differentiation [8,9]. MicroRNAs are small non-coding RNAs of 20 to 25 nucleotides in length and usually act as endogenous repressor of gene activity [10]. They bind to the 39 untranslated region (39UTR) of target mRNAs for translational repression or mRNA cleavage. More than 10,000 distinct microRNA sequences from genomes of viruses, worms and mammals have been identified through random cloning and sequencing or computational prediction (microRNA Registry, http://www.microrna. sanger.ac.uk/sequences). In human, more than 800 microRNAs, attributing to about 2% of known protein coding genes, are known to regulate various biological processes, although many of the target genes remain to be identified.

Microarray experiments
Total RNA was extracted by Trizol/chloroform and purified with the RNeasy kit (Qiagen, Valencia, CA, US). RNA quantity and integrity were examined by RNA 6000 Pico Chip Kit (Agilent, Santa Clara, CA, US) on Agilent 2100 Bioanalyzer. RNA samples with 28S/18S ratios in the range of 1.4 to 1.8 were used. MicroRNA profiling was performed using Agilent Human microRNA Microarray V2 platform, which screens for the expression of 723 human microRNAs from Sanger database v.10.1. For gene expression analysis, Agilent Whole Human Genome Oligo Microarray was used. RNA samples were labeled with cyanine-3 (Cy3) using the Agilent One-Color Labeling kit and hybridized to the array according to the manufacturer's protocol. The signal was detected with an Agilent DNA Microarray Scanner. The array data is MIAME compliant and the raw data was deposited in NCBI Gene Expression Omnibus (GSE24979 and GSE24980, http://www.ncbi.nlm.nih.gov/geo/).

Data normalization and analysis
The scanned images were extracted with Feature Extraction Software 9.5.3.1 (Agilent). Background intensity and feature nonuniform outliers were removed by standard procedures. The processed data were imported to GeneSpring GX10.0.2 for log 2 transformation. Signal cut-off measurements were less than 0.01, and normalized to 75 th percentile of signal intensity to standardize each chip for cross-array comparison. Differentially expressed miRNAs were identified using unpaired Student's t test with P values cut off by 0.01 and fold change more than 2.0. MicroRNA target gene prediction was performed by TargetScan algorithm. The gene expression list was uploaded to web-based tool DAVID (Database for Annotation, Visualization and Integrated Discovery) V6.7 for enriched Gene Ontology terms and significant pathway analysis. The statistically significant terms calculated by Fishers Exact T test associated with the biological processes were selected for comparison.

Real-time quantitative PCR
Complementary DNA (cDNA) was prepared using High Capacity cDNA Reverse Transcription kit (Applied Biosystems, Calsbad, CA, US) from total RNA (1 mg) using Taqman miRNAspecific RT primer (Exiqon, Vedbaek, Denmark, Table S1) and a cocktail containing reverse transcription buffer, dNTP mix, recombinant RNase inhibitor, Multiscribe reverse transcriptase at standard procedure. Real-time PCR was performed using Taqman Universal PCR Master Mix on PRISM 7900HT Sequence Detection System (Applied Biosystems). Amplification of a single fragment was confirmed by a dissociation curve with good correlation with standards and threshold-cycle values. Gene expression was performed by Sybr Green assay or semiquantitative RT-PCR and primers were listed in Table S2.

Corneal epithelium organotypic culture
Human amniotic membrane (AM) obtained by elective Caesarean with written consent was preserved sterile in 50% glycerol at 280uC. Prior to culture, AM epithelium was removed by treatment with 5 mg/ml dispase. The remaining basement membrane and stroma were placed in a culture insert (Corning, Corning, NY, US) with the epithelial side facing up. CEPCs transfected with pre-miRs or scrambled sequences were seeded at a density of 5610 4 cells/cm 2 in serum-free SHEM supplemented with EGF and bFGF and cultured until confluence. The cell monolayer was air-lifted with basal side nourished by culture medium for 21 days to induce multilayer formation. The constructed epithelium was fixed with 10% neutral buffered formaldehyde, paraffin embedded and sectioned for histological examination. Sections were used for immunohistochemical staining with anti-human ITGB8 antibody and horseradish peroxidase-DAB (3,39-diaminobenzidine) reaction.

Luciferase reporter assay
A 5734 base-pair XhoI/Not1 fragment encompassing the fulllength 39UTR of the human ITGB8 gene was amplified by PCR, forward primer: 59-GAAGCTCGAGCTTTCGGTGCAACT T-CTAAA, and reverse primer: 59-ATTAGCGGCCGCGATTAA-CACCTACTACTAAACAG. The fragment was ligated into the XhoI/Not1 sites of psiCHECK TM -2 vector (Promega, Madison, WI, US) with Renilla luciferase as the primary reporter gene. The wildtype pCHECK-ITGB8_39UTR construct was used as template to generate specific substitution (AACT to TTCT) of miR-145 target site using QuikChange II Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA, US) and oligonucleotides (28-34 th site: GATTTTTAAACACTTAATGGGATTCT GGAATTGTTA-ATAATTGC; and 4421-4427 th site: TCTCACTTTTAAACA-AAATTTTCT GGAAAAATATTACATGG). Wildtype and mutant constructs were verified by direct sequencing. For the luciferase assay, HeLa cells in 24-well plates were transfected with 0.5 mg construct and 50 pmol of pre-miR-145 or scrambled sequences using Lipofectamine 2000. At 24 hours post-transfection, cells were collected for measuring the luciferase activity by the Dual Luciferase Reporter Assay (Promega). The experiment was repeated 5 times. Mean activities and standard deviation were calculated and samples compared for statistical significance using paired Student's t-test.

Demarcation of human limbal-peripheral versus central corneal epithelia
To assess the quality of corneal specimens, one-eighth of each cornea rim with intact limbus was paraformaldehyde-fixed. Cryosections were obtained for immunofluorescence of corneal progenitor and differentiation markers, including p63a, ATP binding cassette glycoprotein member 2 (ABCG2), cytokeratin-15 (CK15), cytokeratin-3/12 (CK3/12), Cnx43 and epidermal growth factor receptor (EGFR) [20][21][22][23]. Those specimens exhibiting correct expression and localization of these markers were used for further experiments ( Figure S2). The cell membrane staining of ABCG2 was observed in limbal basal epithelial cells but not in any of the peripheral and central corneal epithelial cells. The nuclear p63a staining was strong in limbal basal and suprabasal cells, gradually restricting to the basal epithelia of peripheral and central cornea. Undifferentiated CK15 was prominent in limbal basal, limbal suprabasal and peripheral corneal basal epithelia but absent in central corneal basal epithelium. Corneal differentiation marker CK3/12 was absent in limbal basal epithelium and weak in the peripheral and central corneal basal epithelia. Gap junction protein Cnx43 was not expressed in most basal cells of limbus but was positive in central cornea. Cell membranous EGFR was found in basal cells of both limbus and cornea. The exclusive existence of p63a strong ABCG2 + CK3/12 2 Cnx43 2 cells in basal LPC epithelia demonstrated the existence of CEPCs. Holoclone formation was obtained from cells dissociated from LPC but not from CC, substantiating the presence of CEPCs in LPC (Fig. S1B).

MicroRNA profiling in human corneal epithelium
From 10 ìm thick cryosections, LPC and CC epithelia were dissected out as pairs and processed immediately for small RNA extraction to preserve the in situ microRNA expression, which could be disrupted due to in vitro manipulation. Both LPC and CC samples were analyzed for ABCG2 and p63a expression by qPCR and western blotting. The sample pairs with positive ABCG2 expression in LPC but negative in CC were selected for microRNA analysis (Figs. S1C, D).
Since CEPCs are present in LPC epithelia but not in CC and they could be regulated by microRNAs, we predicted that LPC epithelia might contain microRNAs distinguishable from those in CC epithelia. For microRNA profiling using microarray investigation, we studied 4 pairs of LPC and CC samples showing similar miR-205 expression levels (Fig. 1C) Fig. 1B). Among them, miR-184, as previously reported [16], was the most significantly up-regulated microRNA in CC epithelia (4.9 folds, P = 0.00005, unpaired Student's t test). Elevated expression of miR-143 and miR-145 in LPC epithelia was validated by qPCR on additional 11 pairs of human LPC and CC epithelia. After normalization with the respective U6, DCT of miR-143 was 5.960.8 in LPC and 11.160.9 in CC epithelia (P = 0.0006, Mann Whitney U-test) (Fig. 2A). Similarly, DCT of miR-145 was 4.560.7 in LPC and 10.260.7 in CC epithelia (P = 0.0004, Mann Whitney U-test) (Fig. 2B). The smaller the DCT values relative to U6, the higher the expression. With about 88% efficiency in our PCR amplification system, LPC had miR-143 and miR-145 expressions 26.6-fold and 36.5-fold higher than CC epithelia respectively. Such levels were comparable to the array results described earlier.
Localization of miR-143 and miR-145 in human corneal rim specimens was shown by LNA-based in situ hybridization. We hybridized the corneal rim cryosections with denatured DIGlabeled LNA-miRCURY oligo probes for miR-143 (Fig. 2C), miR-145 (Fig. 2E), scrambled sequences (Fig. 2G) and U6 (Fig. 2H). In contrast to U6 as the positive control and scrambled sequence as the negative control, miR-143 and miR-145 were more intensively detected in the limbal epithelium, but low to negligible in the corneal epithelium. Similar results were found in 4 repeated experiments. As shown in Figures 2D and F under higher magnification, miR-143 and miR-145 were present predominantly in the parabasal layers, with the intensity reducing towards the superficial layers. They were not strongly expressed in the basal layer, which contains CEPCs.

MiR-145 regulated corneal epithelial differentiation
Freshly isolated human CEPCs were cultured in CnT20 medium and cells from primary holoclones were pooled and plated as passage 2 (P2) for transfection with lenti-miR expression plasmid pMIRH143PA-1 or pMIRH145PA-1 at a ratio of 3 ml Lipofectamine 2000 per mg DNA. Over-expressions of miR-143 and miR-145 were shown by GFP live imaging (Figs. 3A-B, miR-143; 3C-D, miR-145) and qPCR (Fig. 3E). The transfected cells were kept in CnT50 medium with low bovine pituitary extract (15 mg/ml) for optimal corneal epithelial cell growth for 2 days followed by expression analysis. By immunofluorescence, both miR-143 and 145-transfected cells had increased CK3/12 expression (Figs. 3G and H), when compared to cells transfected with scrambled sequences (Fig. 3F). This was also detected by western blotting (Fig. 3N, second panel). On the other hand, miR-145-transfected cells showed relatively stronger Cnx-43 expression (Fig. 3K), which was mild in cells transfected with scrambled sequences (Fig. 3I) or miR-143 (Fig. 3J). Western blot analysis showed Cnx-43 upregulation in CEPCs transfected with miR-145 by about 15 folds more than those with scrambled sequences (Fig. 3L). In addition, these cells had reduced ABCG2 (Fig. 3M, bottom panel) and p63a expression (Fig. 3N) as revealed by RT-PCR and immunofluorescence, respectively. This expression pattern of corneal differentiation markers, i.e., reduced ABCG2 and p63a expression, and its expression in parabasal layers of limbal epithelium, indicated that miR-145 might be involved in corneal epithelial differentiation. We corroborated this supposition by a three-dimensional corneal epithelial organotypic assay. Human P2 CEPCs transfected with miR-143, miR-145 or scrambled sequences were expanded to monolayer cell sheet on denuded AM in submerged culture, followed by air-lifting to induce cell stratification. The composites were harvested for morphological examination. The number of epithelial layer was quantified in 15 random sites along the composite to obtain the epithelium forming efficiency. CEPCs without transfection (Fig. 4D), with lipofectamine only (Fig. 4E) or transfected with scrambled sequences (Fig. 4A), generated thicker epithelia. They had typical epithelial morphology with basal cuboidal-like cells next to the basement membrane. The cells were packed and appeared squamous in shape at the superficial layers (non-transfected: 12.561.5 layers; lipofectamine-only; 9.962.3 layers; transfected with scrambled sequence: 10.262.4 layers) (Fig. 4F). However, this was not observed in the epithelia generated from miR-145transfected CEPCs (Fig. 4B). The epithelium was degraded, thin (5.661.3 layers) and loosened with reduced cell density. Few cuboidal cells were found in the basal layer and cells were tends to be flatten or squamous in shape. The epithelium generated from miR-143-transfected CEPCs had morphology and compactness intermediate between control and miR-145 epithelia (8.361.6 layers) (Fig. 4C). The same results were obtained in duplicated experiments.

Transcriptional regulation by miR-145
We compared the transcription profile of human corneal epithelial HCE cells transfected with lenti-miR-145 or with scrambled sequences using the Agilent Whole Human Genome Oligo Microarray platform, which screens for 41K human genes and transcripts. In two separate array experiments, miR-145 upregulated 324 genes and down-regulated 277 genes by a five-fold difference compared to cells transfected with scrambled sequences (Tables S3A-B). Significant Gene Ontology terms enriched in the deregulated gene sets represented immune response (P,10 27 ), process (P,10 25 ), regulation (P,10 23 ), inflammatory response (P,10 24 ), cell defence (P,10 25 ), cell apoptosis (P,10 23 ), cell differentiation (P,10 23 ) and development (P,10 23 ). In addition, differentially regulated genes with .2 folds difference could be associated with epithelium development, cell proliferation and differentiation (Table 2, a full list is shown in Table S4). Real-time PCR analysis in 4 transfection experiments consistently showed that miR-145 markedly down-regulated integrin b8, ITGB8 (P = 0.00024, paired Student's t test) and up-regulated interferon b1 (IFNB1) (P,0.005) but not other candidate genes, such as Wnt7A, SOCS7 and Klf4 (Fig. 5A). Similar reduction of ITGB8 expression was observed in primary human CEPCs transfected with miR-145 (Fig. 3M). Predicted by TargetScan Human version 5.1 (http://www.targetscan.org), 2 conserved sites for miR-145 binding: 28-34 th and 4421-4427 th was found in the 39UTR, equivalent to 3043-3049 th and 7436-7442 nd of human ITGB8 (NM_002214) (Fig. 5B). Notably, the first site is conserved in primates only whereas the second is found among primates, rodents and avian. Hence, influence of miR-145 on ITGB8 expression could be species-specific. To confirm miR-145 regulated ITGB8 expression by direct targeting the 39UTR, we cloned the full-length wildtype ITGB8 39UTR fragment downstream of psiCHECK-2 luciferase reporter gene and introduced mutated miR-145 target sites: 28-34 th and 4421-4427 th , by sitedirected mutagenesis. Luciferase expression, which represented promoter activity, was examined in HeLa cells co-transfected with the vectors and pre-miR-145 or scrambled sequences. The cells co-transfected with wildtype 39UTR and miR-145 had reduced luminescence when compared to cells with wildtype 39UTR and scrambled sequences (P,0.005, one-way ANOVA) (Fig. 5C). Reduced luminescence was also found in cells co-transfected with miR-145 and mutated 39UTR at site 28-34 th but not with mutated site at 4421-4427 th . Immunofluorescence showed membranous staining of ITGB8 in human limbal epithelium, particularly the basement membrane in contact with basal cells and superficial cell layers (Fig. 5D). Negative staining of the parabasal layers was coincident with positive miR-145 expression as shown by in situ hybridization (Figs. 2E and F). ITGB8 was strongly detected in cultured epithelia generated from CEPCs transfected with scrambled sequences but mild in those from pre-miR-145transfected CEPCs (Fig. 5E).

Discussion
In this study, we found differential expressions of 18 microRNAs between human limbal-peripheral and central corneal epithelia. Among them, miR-143 and miR-145 were expressed predominantly in the limbal epithelium but at very low levels in the central corneal epithelium. In primary human CEPCs, transfection of miR-145 up-regulated cytokeratin-3/12 and connexin-43 and concomitantly suppressed p63 and ABCG2 expression. We hypothesized that miR-145 could be an important regulatory molecule for human corneal epithelial differentiation. These cells developed to thinner and defective epithelium in vitro. This morphological alteration could be caused by miR-145 via the direct targeting on ITGB8. Meanwhile, disruption of binding site in ITGB8 39UTR by site-directed mutagenesis eradicated the inhibition caused by miR-145. To our knowledge, this is the first report of microRNA regulation on human ocular cell differentiation and an example that microRNA can interfere with tissue development.
Located in human chromosome 5 and rodent chromosome 18, miR-143/145 are co-transcribed as one microRNA cluster from the same microRNA precursor. In a mouse model, both are initially expressed in the developing embryonic heart, followed by migration to smooth muscle cells of the aorta, intersomitic arteries, esophagus, lung, colon bladder and umbilical cord at later stages [24]. In adult mice, they are expressed in lung, skeletal muscle, heart and skin, and most abundantly in aorta and fat. MiR-145 is faintly expressed in self-renewing human ES cells but up-regulated during differentiation, indicating an inductive role on ES cell differentiation [25]. It directly targets on core plurpotency factors, repressing Oct4, Sox2 and Klf4, which are crucial to maintain the self-renewal and pluripotency capacity of ES cells and promote developments of the mesoderm and ectoderm lineages [24,[26][27][28]. A feedback mechanism has been proposed between miR-145 and Oct4 regulation [25]. Oct4 down-regulates miR-145 expression through repressive binding to its promoter. Hence, in ES cells, high Oct4 level suppresses miR-145 and the cells are capable of self-renewing. When entering into specific lineages, the differentiated cells express miR-145, which promotes differentiation and targets on Oct4 to suppress the self-renewal capability. In this study, miR-143/145 were expressed in the limbal epithelium, in particular the parabasal wing cell layers, but not in the central corneal and basal limbal epithelia. The parabasal region is enriched with proliferative TA cells with no self-renewal ability. The absence of Oct4 facilitates miR-145 expression. Furthermore, in normal adult CEPCs, Oct4, Sox2 and Klf4, as anticipated, are not detectable (Fig. 5A). This restricted the proof-of-concept study of miR-145 targets by knock-down experiments. Instead, overexpressing miR-145 in human CEPCs promoted CK3/12 and Cnx43 expression, which indicated the onset of corneal epithelial differentiation. Concurrently, these cells had reduced ABCG2 and p63 expressions, indicating their exit from stem cell proliferation state. As shown by in vitro corneal epithelium organotypic assay, a thinner and loosened epithelium was generated from miR-145transfected human CEPCs. These findings strongly indicated that  miR-145 suppressed the progenitor cell pool, and the cells were prone to differentiate, resulting in an underdeveloped epithelium with fewer cuboidal basal cells. We have attempted to transfect primary CEPCs by lenti-miR plasmid and observed about 70-80% transfection efficiency according to live GFP expression. We selected these cells for organotypic culture, which had demonstrated the influence of miR-145 on corneal epithelium development. Together with the reduced ITGB8 expression in the resulting epithelium, this likely showed that some cells could maintain miR-145 expression after culture for a month. However, we could not determine if this was due to transfection or endogenous expression in cells. Further experiments using lentivirus to obtain a long-term over-expression of miR-145 or specific miR-145 knockdown will be carried out to study the biological mechanism how miR-145 regulates corneal cell proliferation, migration and differentiation. Previous study using a zebrafish platform illustrated that miR-145 knockdown resulted in underdeveloped gut and heart [29]. We observed miR-145 down-regulated ITGB8 in human corneal epithelial cells and this might influence epithelium development and formation. In human limbal epithelium, ITGB8 was predominantly detectable in cell-cell boundary of superficial layers and also found scattered between cells in the basal layer. Interestingly, it was negligibly expressed in the parabasal region where miR-145 was detected. ITGB8, with its binding partner aV, is expressed in normal epithelial and neuronal cells in vivo and regulates transforming growth factor b (TGFb) activation in various events, including cell growth, matrix modeling, epithelialmesenchymal homeostasis, immune regulation and vasculogenesis [30]. Binding of Sp1, Sp3 and AP-1 transcription factors to its core promoter regulates ITGB8 expression in a p38-dependent manner [31]. TGFbactivation could lead to autocrine and paracrine signaling on cell growth and matrix production, which are important for epithelial cell adhesion and motility [32]. b8 interaction with Rho guanine nucleotide dissociation inhibitor-1 selectively stimulates Rac1, which regulates actin cytoskeleton arrangement, an important event in cell proliferation and differentiation [33,34]. Also, aVb8 integrin facilitates Fas induction [35], which is crucial for cell migration, production of inflammatory cytokines and corneal wound healing.
In addition, miR-145-transfected HCE cells had up-regulated IFNB1, which is known with anti-inflammatory activity. The cornea is described as ''immune privilege'', characterized by suppression of systemic immunity after infection. This can be associated with low vascularization, presence of Fas ligand, which is a target of aVb8 integrin, and TRAIL molecules and inhibitory substances in the aqueous humor [36]. As a consequence, corneal allografts usually survive longer than allografts in other body parts [37]. The immunologically protective mechanism in cornea can be associated with the production of nitric oxide, which intoxicates various pathogens [38]. Production of interleukins, interferon and TGFb via aVb8 integrin pathway) are pathogen-induced immunologically protective mechanism in corneal cells [39,40].
Up-regulation of IFNB1 may contribute to enhancement of antiinflammatory capability of corneal cells.
In conclusion, we discovered differential expression of micro-RNAs in human limbal and corneal epithelia. MiR-145 could be an important regulatory molecule for human corneal epithelial progenitor cell proliferation and differentiation. It is also crucial for the integrity of corneal epithelium, likely via ITGB8 targeting, which will be further investigated with ITGB8 knockdown mice. Our findings provide the first identification of microRNAs expressed in adult tissue-specific site with regulatory effect on tissue cell differentiation.    pCHECK-ITGB8_39UTR and pre-miR-145 showed lower luciferase reporter activities when compared with cells transfected with scrambled sequences (n = 5) (red labels). Disruption of binding site 4421-4427 th region resulted in higher luciferase activity (blue labels) whereas mutation at 28-34 th region had no influence and the reduced luciferase activity levels (green labels) were similar as WT. * P,0.005, as compared to scrambled control (one-way ANOVA). (D) Immunofluorescence of ITGB8 in human limbal epithelium. Positive immunoreactivity was observed scattered in basal layer and continuous in superficial layers. No observable expression was noted in the parabasal layers. (E) Immunoperoxidase staining of ITGB8 in organotypic generated epithelia. Distinct ITGB8 expression was observed in epithelial layers generated from CEPCs transfected with scrambled sequences whereas reduced expression was found in epithelia from pre-miR-145-transfected CEPCs. Scale bars (D, E): 50 mm. doi:10.1371/journal.pone.0021249.g005