DNA methyltransferase 1 regulates epithelial cell functions in corneal and eyelid development

Animals

Dnmt1^F/F/Le-Cre^TG/- females were crossed with Dnmt1^F/F males and the Dnmt1^F/F embryos/fetuses with or without a Le-Cre^TG allele, denoted as Dnmt1^ΔOSE or wild-type pups, respectively, were collected for analyses as described before [15]. Genotype was determined by PCR using tail-cut genomic DNA and the primers listed in Appendix 1. Animals were housed in a pathogen-free vivarium in accordance with institutional policies. All animal experiments were approved in advance by the Institutional Animal Care and Use Committee at the University of Cincinnati College of Medicine.

Keratinocyte and stable cell lines

Mouse keratinocytes derived from Dnmt1^F/F mice were cultured and maintained in Keratinocyte Serum Free Medium without Calcium Chloride (10,725-018, Gibco, Thermo Fisher Scientific, Waltham, MA) media, as described before [20]. To develop stable reporter cell lines, the Dnmt1^F/F cells were infected with lentivirus carrying the pathway luciferase reporters purchased from LipExoGen Biotech (Appendix 2); stable cells were obtained following Blasticidin (R21001, Thermo Fisher Scientific) selection. The cells were subsequently infected with adenoviruses expressing either green fluorescent protein (Ad-GFP) or GFP plus Cre recombinase (Ad-GFP-Cre, or Ad-Cre) at 5 IU for 48 h to generate wild-type and Dnmt1-knockout (deficient) keratinocytes, respectively.

Histology and immunofluorescent staining

The heads of appropriately euthanized dams were collected and fixed in 4% paraformaldehyde (PFA) at 4 °C overnight. Tissues were embedded in paraffin and sectioned at 5 µm with a microtome (MT-980, Research & Manufacturing Co., Inc., Tucson, AZ). Selected serial tissue sections were subjected to hematoxylin (ES701, Azer Scientific, Morgantown, PA) and eosin (ES709, Azer Scientific) staining, and images were captured using Zen 3.1 (Blue edition) under bright-field light microscopy (Zeiss AXIO Scope A1, Göttingen, Germany) at different magnifications.

For immunohistochemistry, tissue sections representing the center of the eye were deparaffinized and immersed in boiled 10 mM citrate buffer (pH 6.0) for 30 min for antigen retrieval. The sections were blocked with phosphate-buffered saline/tween (PBST) containing 5% normal goat serum (PCN5000, Gibco, Thermo Fisher Scientific) for 1 h at room temperature and incubated with a diluted primary antibody overnight at 4 °C followed by secondary antibodies and Hoechst 33,342 (H1399, Thermo Fisher Scientific) staining. The list of antibodies and dilutions for each antibody is listed in Appendix 3. Images were captured using a confocal laser scanning microscope (Zeiss AXIO Observer Z1 LSM 700).

For cultured cell immunofluorescent staining, the Dnmt^F/F keratinocytes grown on cover glass were infected with Ad-GFP or Ad-Cre in Defined Keratinocyte Serum Free Medium (10,785-012, Gibco, Thermo Fisher Scientific) as described before [20]. After washing with PBS, the cells were fixed with 4% PFA at 4 °C for 20 min followed by incubation with a blocking solution of PBST containing 0.25% bovine serum albumin (BSA; A2153, Sigma-Aldrich, St. Louis, MO) for one hour at room temperature. Immunostaining was performed as that described for tissue sections.

Apoptosis and proliferation

To determine apoptosis, the tissue sections were subjected to terminal deoxynucleotidyl transferase dUTP nick-end labeling assays using the ApopTag^® Plus in situ Apoptosis Fluorescein Detection Kit (S7111, Millipore Sigma, Temecula, CA) following the manufacturer’s protocol. Keratinocytes grown on cover glass with Ad-GFP and Ad-Cre infection as described above were examined by immunofluorescent staining with anti-Caspase-3 (CST 9663S, Cell Signaling Technology, Danvers, MA) as the primary antibody, Alexa Fluor™ 594 Goat anti-Rabbit IgG (H⁺L; A11012, Invitrogen) as the secondary antibody and Hoechst 33,342.

To determine proliferation, 2 h after intraperitoneal injection/incubation with 5-ethynyl 2´-deoxyuridine (EdU; Component A, 5 mg/kg b.w. or 10 µM in growth media), embryos/fetuses were collected from pregnant mice and cultured cells harvested. Tissue sections and keratinocytes were infected with Ad-GFP/Ad-Cre and subjected to an iClick™ EdU Andy Fluor™ 594 Imaging Kit (A005, ABP Biosciences, Beltsville, MD) following the manufacturer's protocol.

Cell apoptosis and proliferation were observed, and images were captured with a Zeiss AXIO Observer Z1 LSM 700 microscope. The staining positive cells were quantified and the percentage of positive staining was calculated versus total cells. The results represent data from at least three sections per embryo and three embryos/genotypes of different litters; data from cultured keratinocytes were quantified in three samples, and only green cells positive for Ad-GFP or Ad-Cre infection were included in quantification.

RNA isolation, reverse transcription and quantitative real-time PCR

Total RNA was extracted from cultured keratinocytes with a PureLink^® RNA Mini kit (121,830, Invitrogen, Waltham, MA) following the manufacturer’s protocol. A total of 1 μg RNA was subjected to reverse transcription using a SuperScript™ IV First-Strand Synthesis System (18091050, Invitrogen). The cDNA was mixed with gene-specific primers and SYBR Green QPCR Master Mix, and subjected to quantitative real-time PCR an MX3000p thermal cycler system (Agilent Technologies, St. Clara, CA). Relative gene expression was calculated by the comparative ΔΔCt-method normalized to the constitutively expressed housekeeping gene, Gapdh. Data represented duplicates of at least three independent samples. Primer sequences are listed in Appendix 4.

Western blotting

Cells grown on a culture dish were lysed with RIPA buffers (0.5M EDTA pH 8, 0.1% SDS, 1% NP-40, 5M NaCl, 1M Tris-HCl pH 7.6, 2.5 g NaDOC, protease inhibitor). A total of 50 µg cell lysates were subjected to SDS PAGE, followed by western blotting using anti-DNMT1 and anti-β-actin. Sources and dilutions of the primary and secondary antibodies are listed in Appendix 3. Immunoblotting was detected by chemiluminescence and images were captured with a Gel Imaging System (UVP Transilluminator, Analytik Jena Company, Upland, CA). Signals were quantified by ImageJ software and the signal intensity of DNMT1 was compared with that of β-actin as the loading control.

Luciferase activities

Ad-GFP– or Ad-Cre–infected Dnmt1^F/F reporter cells in 24-well plates were lysed and the activity of luciferase was measured using a luciferase assay system (Rapid detection of Firefly Luciferase activity, E1500, Promega, Madison WI). Relative luciferase activities were quantified after normalization with the cell's total protein. The results represent duplicate data from four biologic samples of each adenovirus transfection, Ad-GFP and Ad-Cre.

Statistical analysis

At least three independent samples were used in all experiments for data analysis. Statistical analysis was performed using a Student’s two-tailed paired t test for comparison between groups. The exact number of biologic replicates and statistical analysis used for each experiment are included in the figure legends. Values of p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***) were considered statistically significant.

Abnormal corneal and eyelid development in Dnmt1 ocular surface epithelium knockout mice

We crossed Dnmt1^F/F mice with mice carrying one Le-Cre allele that expressed Cre recombinase in the OSE [21]. The resultant Dnmt1^F/F/Le-Cre mice, denoted as Dnmt1^ΔOSE, should have Dnmt1 specifically knocked out in the OSE, whereas the Dnmt1^F/F littermates were considered as wild-type controls. We collected E15.5 embryos to determine DNMT1 expression. In the wild-type embryos, DNMT1 was expressed ubiquitously in the epithelial and stromal cells of the ocular surface tissues, including the lens, cornea and eyelids. In Dnmt1^ΔOSE embryos, while DNMT1 expression was still abundant in stromal cells, it was largely absent in the OSE (arrowheads, Figure 1A). Concurrently, DNMT1-mediated 5-methylcytosine (5meC) was abundant in the OSE of the wild-type embryos but undetectable in that of the knockout embryos (arrowheads, Figure 1B).

To determine whether defective DNMT1 expression and DNA methylation affected eye development, we examined the histology of wild-type and Dnmt1^ΔOSE eyes at E14.5–E16.5, encompassing the late embryogenic to early fetal developmental phases. Consistent with a previous report [15], OSE DNMT1 ablation caused lens degeneration. The knockout lens lost capsule epithelial cells and was smaller in size compared with those of wild-type embryos (Figure 2). The knockout lens was also frequently associated with anterior cell clusters (Figure 2, hash #) fused to the corneal endothelium as well as posterior debris located between the lens and the retina.

Unlike the Dnmt1^ΔOSE lens, in which degeneration was evident in the E14.5 embryos, the knockout cornea and eyelids at this stage appeared to be normal; however, they exhibited clear abnormalities from E15.5 onward. The knockout corneas protruded anteriorly and became thicker and the knockout eyelids failed to fuse, leaving the corneal surface exposed at E16.5. In contrast, the wild-type had thinner corneas and their upper and lower eyelids were fused over the cornea. Corneal and eyelid abnormalities were observed only in Dnmt1^F/F/Le-Cre but not Dnmt1^+/F/Le-Cre fetuses, indicating that Dnmt1 deletion was autosomal recessive for corneal and eyelid defects (Appendix 4). These observations also suggest that Le-Cre expression is not responsible for the defective phenotypes observed in Dnmt1^F/F/Le-Cre pups. Thus, DNMT1 expression in the OSE is required for corneal and eyelid development at late embryogenesis in addition to its role in lens development.

DNA methyltransferase 1 is required for corneal epithelial organization and differentiation

During embryonic development, the corneal epithelium is comprised of uniformed epithelial cells aligned in two layers above the basement membrane. We detected well-organized epithelial layers in the corneas of wild-type fetuses; however, we found irregular-shaped epithelial cells arranged in a distorted pattern with large intercellular spaces in the corneas of knockout fetuses (arrows, Figure 3A). Some of the basal Dnmt1^ΔOSE corneal epithelial cells were detached from the basement membrane (asterisks, Figure 3A). The wild-type and Dnmt1^ΔOSE corneal stroma was also strikingly different. The corneal stroma of wild-type fetuses was packed with residential keratocytes aligned parallel to the basement membrane, whereas the corneal stroma of Dnmt1^ΔOSE fetuses, filled with irregular-shaped keratocytes, was nearly twice the thickness of that in wild-type (Figure 3A,B). Similar observations were made in both E15.5 and E16.5 embryos/fetuses (Figure 3 and Appendix 4).

The stromal thickness in Dnmt1^ΔOSE embryos was not due to an increased number of keratocytes, but corresponded to larger intercellular spaces (Figures 3A,B). These observations led to the hypothesis that DNMT1 loss impairs the barrier function of the corneal epithelium, which in turn causes fluid accumulation and edema of the corneas. To test the hypothesis, we determined cell–cell junctions that constituted the epithelial barrier by establishing tight cell connections. One of the cell junctions is mediated by E-cadherin (E-Cad), a transmembrane protein and a key component of the adherens junctions to establish cell–cell contacts [22]. While E-Cad was detected on the plasma membrane of corneal epithelial cells, apparent differences were found between wild-type and Dnmt1^ΔOSE (Figure 3C). E-Cad was evenly distributed on the plasma membrane at the cell–cell and cell–basement membrane junctions in wild-type corneas. It was also located at the cell–cell junction in Dnmt1^ΔOSE corneas but was missing on ectopic spots between the epithelial and basement membranes (arrows, Figure 3C). Another key cell junction is mediated by Zonula Occludens-1 (ZO-1), also a transmembrane protein and a component of the tight junction complexes that regulate paracellular movement of ions and solutes between cells [23]. ZO-1 proteins were located predominantly on the plasma membrane of the corneal epithelium in wild-type embryos; however, they displayed diffused distribution throughout the cytosol in the knockout corneal epithelium (Figure 3D).

The corneal epithelium comprises stratified squamous epithelium that express distinct keratin intermediate filaments during embryonic development [24]. Keratin 14 (K14) is the basal epithelial cell marker expressed in all epithelial cells as early as E9.5 [25], whereas K12, a corneal-specific keratin essential for epithelial integrity and function, starts to express at E15.5 [26]. Examination of these keratins in wild-type and Dnmt1^ΔOSE E15.5 embryos showed that K14 was evenly distributed on the plasma membrane of wild-type corneal epithelial cells but exhibited an uneven and sporadic distribution and was largely absent in basal cells above the basement membrane of Dnmt1^ΔOSE corneas (Figure 3E). Moreover, K12, located on the plasma membrane of the suprabasal corneal epithelial cells in wild-type embryos, was completely undetectable within the corneas of Dnmt1^ΔOSE embryos (Figure 3F). Together, our data indicate that DNMT1 is required to maintain the integrity and differentiation of the corneal epithelium, and that its deficiency causes a disorganized epithelium as well as stroma and corneal edema.

The role of DNA methyltransferase 1 in eyelid morphogenesis

A massive morphogenetic event in the late embryonic development of the mammalian eye is closure of the upper and lower eyelids, resulting in an enclosed eyelid covering the corneal surface [27,28]. Mouse eyelid closure occurs between E15.5 and E16.5, and the eyelids remain fused at birth. Thus, mice are normally born with their eyes closed, whereas newborn pups display an “eye-open” at birth (EOB) phenotype when the eyelids fail to fuse prenatally [29]. The observations that eyelids fail to fuse in Dnmt1^ΔOSE E16.5 fetuses has led us to ask whether this defect resulted ultimately in the EOB phenotype. To address the question, we examined E17.5, E18.5 and postnatal day 1 (P1) pups and found that, unlike wild-type mice, which had fully closed eyelids, Dnmt1^ΔOSE pups had partially open eyelids (Figure 4).

Prior to eyelid closure, the peridermal cells at the eyelid tip undergo morphological changes, leading to the formation of a protrusion that presumably initiates the centripetal epithelial movement for lid closure [30]. To understand how Dnmt1 knockout impairs closure, we examined the tips of wild-type and Dnmt1^ΔOSE eyelids at E15.5, a developmental stage at the onset of lid closure. Eyelid tip epithelial protrusion was evident in the wild-type but not the knockout embryos (arrowhead, Figure 5A). The knockout eyelid tip instead appeared to fold posteriorly and became thicker (asterisk, Figure 5A); its epithelial cells were less compact with larger intercellular spaces, and some of the suprabasal cells had condensed chromatin (arrows, Figure 5A). Similar abnormalities were observed in the eyelid tip of E14.5 and E16.5 Dnmt1^ΔOSE embryos (Appendix 4), indicating that DNMT1 ablation prevented the normal formation of eyelid tip protrusion.

To determine if DNMT1 ablation impaired the structural integrity of the eyelid tip like that seen in the cornea, we assessed E-Cad and ZO-1 via immunohistochemistry. Both ZO-1 and E-Cad were detected in the knockout eyelids, but their distributions were somewhat different from that of wild-type eyelids (Figure 5B). In the wild-type eyelid tips, ZO-1 was evenly distributed on the plasma membrane of basal and suprabasal epithelial cells, but in Dnmt1^ΔOSE it was scattered throughout the cytosol in the suprabasal cells (asterisk, Figure 5B). E-Cad, on the other hand, was located on the plasma membrane of the eyelid tip epithelial cells in both genotypes, although it appeared to be less abundant in the basal epithelial cells of the knockout eyelids.

The embryonic eyelid is composed of stroma covered by epidermal layers at the anterior surface and conjunctival epithelium on the posterior side [31]. To assess the roles of DNMT1 in anterior and posterior eyelid epithelial differentiation, we determined the expression of K14, a marker of basal proliferative keratinocytes, and K10, which is expressed only in terminally differentiated suprabasal keratinocytes [25]. K14 expression was detected in the basal epithelial cells on both anterior and posterior surfaces of the eyelids, with no major differences observed between wild-type and Dnmt1^ΔOSE E14.5–E16.5 embryos (Appendix 5). K10 was detected in the suprabasal layers of the anterior epithelium, and the level of expression gradually increased as the embryos grew older (arrowheads, Figure 5C). Compared with wild-type eyelids, Dnmt1^ΔOSE eyelids showed increased K10 expression, both in intensity and in cell layers. In particular, eyelid tip cells were K10-negative in wild-type embryos but became K10-positive in Dnmt1^ΔOSE embryos. Thus, DNMT1 loss impairs lid closure through potentiating differentiation and blocking the protrusion of lid tip epithelial cells.

DNA methyltransferase 1 regulates cell proliferation and apoptosis

Because DNMT1 has been shown to regulate proliferation and apoptosis in many organ-formation processes [11,32], we determined these roles in the OSE of E14.5 to E16.5 embryos. An EdU labeling assay was used to assess proliferation and a terminal deoxynucleotidyl transferase dUTP nick-end labeling assay was used to determine apoptosis. Quantification of the labeling data showed that DNMT1 ablation did not change proliferation but increased apoptosis in the OSE (Figure 6A–D). The number of apoptotic cells in all ocular surface epithelia, including the eyelid (EL), cornea (CO) and lens (LN), was significantly higher in Dnmt1^ΔOSE than in wild-type embryos. The eyelid tip of the knockout, but not wild-type, embryos showed abundant apoptotic cells, corroborated with histological data showing cells with condensed nuclei in this region (Figure 5A and Figure 6B). The knockout corneas, on the other hand, displayed increased apoptosis of basal epithelial cells located between the epithelial layer and the stroma (Appendix 6). Neither proliferation nor apoptosis was altered in the stroma of the Dnmt1^ΔOSE ocular surface (Appendix 7). These observations suggest that DNMT1 is required for the survival of corneal and eyelid epithelia, like that established previously in the lens epithelium [15].

DNA methyltransferase 1 regulates molecular and cellular activities in cultured keratinocytes

To investigate the roles of DNMT1 in epithelial development in vitro, we isolated epithelial cells from the skin of Dnmt1^F/F mice and developed keratinocyte lines using a protocol we have recently developed to obtain epidermal stem cell-like characteristics and extensive subculture capacity [20]. Infection of the cells with Ad-Cre resulted in a marked reduction in DNMT1 mRNA and protein, whereas infection with Ad-GFP did not have such an effect (Figure 7A,B). When placed in differentiation media, Ad-GFP– and Ad-Cre–infected Dnmt1^F/F cells showed the same level of K14 expression; however, Ad-Cre–infected cells, but not Ad-GFP–infected cells, exhibited higher expression of K1 and K10, indicating terminal epidermal differentiation (Figure 7C). Additionally, compared with Ad-GFP–infected cells, Ad-Cre infection significantly increased active Caspase-3 positive cells, suggesting increased apoptosis, without an apparent change in EdU positivity for proliferation (Figure 7D).

To address whether DNMT1 ablation disrupted cell–cell junctions in vitro, as seen in the OSE, we infected Dnmt1^F/F cells with Ad-GFP and Ad-Cre, and examined E-Cad expression and localization. Clear E-Cad localization was detected on the plasma membrane in Ad-GFP–infected cells, but it disappeared in Ad-Cre–infected cells (Figure 7E). Additionally, in contrast to neighboring Ad-Cre–negative cells displaying regular epithelial morphology with polygonal shapes, Ad-Cre–positive cells lost their epithelial shape and became either round or stretched.

Taking advantage of the easily manipulatable feature of cultured keratinocytes, we further explored the signaling pathways affected by DNMT1 loss. To achieve this, we made stable Dnmt1^F/F cells carrying luciferase reporters for specific signaling pathways. Infection of the stable cells with Ad-Cre significantly increased SRE-luc activities for MAPK signaling, TCF-luc for WNT signaling, RORγ for retinoic acid signaling and Gli-luc for Hedgehog signaling compared with those infected with Ad-GFP (Figure 7F). On the other hand, neither Ad-Cre nor Ad-GFP infection changed the activities of XRE-luc for AhR signaling, suggesting that DNMT1 loss led to the activation of specific signaling and transcription programs.