Loss of N-Cadherin from the Endothelium Causes Stromal Edema and Epithelial Dysgenesis in the Mouse Cornea

Abstract

Purpose.: We analyzed the role of N-cadherin in maintaining proper architecture and function of corneal endothelium.

Methods.: To achieve specific removal of N-cadherin from corneal endothelium, we bred mice carrying a floxed N-cadherin gene with those expressing the Cre-recombinase gene under the control of P0 promoter. The corneal structure was analyzed by immunostaining for cell junction proteins as well as by electron microscopy. The apoptotic status was assessed by TUNEL staining. The permeability of corneal endothelium was evaluated using fluorescein dye.

Results.: Removal of endothelial N-cadherin led to the appearance of opacity in the adult corneas. All corneal layers exhibited histological defects: The apical junctional complex (AJC) in corneal endothelium was disorganized, losing the continuity in tight junctions. Collagen fibrils were rearranged in the stroma. The corneal epithelium showed decreased thickness and TUNEL staining revealed increased central areas of apoptosis. Fluorescein dye injection in the anterior chamber confirmed an increased permeability of the endothelial layer. Developmental analysis indicated that, although N-cadherin was lost during embryonic stages, the AJC was maintained normally until early postnatal stages, probably due to the presence of other cadherins at these developmental stages. The junctional defects in endothelial cells, however, became obvious by postnatal day 21 (P21), although stromal and epithelial phenotypes were clearly detectable only in the adult eyes.

Conclusions.: N-cadherin is essential for maintaining proper structure of corneal endothelial AJCs from late postnatal to adult stages. Its ablation leads to increased endothelial permeability and corneal edema in mature eyes.

Introduction

A unique quality of the eye cornea is its transparency. This property is determined by the quasi-ordered-quasi-random arrangement of its collagen fibrils. ¹ Any process leading to loss of this organization would generate corneal opacity. One of the most common reasons for the appearance of such opacity is corneal edema, which occurs due to an unbalanced influx of fluids into the stromal matrix from the anterior chamber of the eye. Corneal endothelial layers play a pivotal role in preventing such corneal swelling to occur. ²  

The corneal endothelium is a monolayer of hexagonally shaped cells separating the corneal stroma from the aqueous humor of the eye's anterior chamber. ³ One function of the corneal endothelium is to restrict fluid influx into the stromal matrix. ⁴ In addition, the endothelium layer must also secure an adequate supply of nutrients to the avascular tissue of the stromal compartment. The source of such nutrients is the metabolite-rich aqueous humor of the anterior chamber. To achieve a balance between these two opposing functions, the endothelial monolayer relies on a complex of highly specialized tight junctions (TJs) and adherens junctions (AJs) localized in the apical part of its cells. These junctions, together with the cytoskeleton tethered to them through adopter proteins, form the so-called apical junctional complex (AJC). 

The main function of the AJC in corneal endothelium is to selectively restrict the flow of water and solutes through the paracellular space, as well as to preserve the polarity of ionic channels including Na⁺/K⁺ ATPase. The latter function is particularly important, as it is fundamental for the “pump” function of the endothelial layer, the ability to actively transport excessive fluids and electrolytes from corneal stroma toward the anterior chamber. ⁵ Within the AJC, TJs are established by the homophilic interactions of specific transmembrane proteins such as claudins, occludins, and junction adhesion molecules (JAMs), and further supported by cytoplasmic components, including the adopter protein ZO-1. ^6,7 One characteristic of corneal endothelium TJs is their “leakiness,” or relatively high permeability for both fluids and macromolecules. ⁸ The AJs, which are localized below the TJs, function to stabilize cell-cell connections. The principal adhesion receptors constituting the AJs are “classical” cadherins, a family of transmembrane glycoproteins that bind in a homophilic manner with cadherin molecules on adjacent cells. ⁹ Inside the cytoplasm, cadherins interact with β-catenin and p120-catenin, among others. β-catenin binds αE-catenin, an adaptor protein known to interact with F-actin. As a consequence, cadherins accumulate actin filaments along the AJs, and in this way control cell shape and contractility. This also occurs in corneal endothelial cells; thick circumferential actin belts demarcate their cell-cell contacts. ⁶ Among the classical cadherins, N-cadherin is highly expressed in the corneal endothelium. ¹⁰ In mouse embryos, it is detected as early as stage E15 when endothelial cells start to differentiate from surrounding mesenchymal cells ^10,11 and its expression persists in adult corneas. 

The primary aim of our study was to identify the role of N-cadherin in maintaining proper corneal endothelium functions. To achieve this goal, we used highly specific genetic ablation of the N-cadherin gene in a neural crest cell population, which gives rise to the corneal endothelium. Our findings confirmed that corneal endothelial cells are predominantly of neural crest origin in mammals. We also identified N-cadherin as the major classical cadherin expressed in the adult mouse corneal endothelium. Removal of this molecule was sufficient to severely disrupt the formation of the AJC. In N-cadherin-null cells, there were no AJs and few TJs, and the actin cytoskeleton and cell shape underwent profound reorganization. These changes led to an increased permeability of the endothelial layer, accompanied by corneal stromal edema. Our results help clarify the role of cadherins in maintaining cornea transparency. The data presented here offer new insights into the pathogenesis of corneal edema. 

Methods

Animals

To analyze the expression pattern of P0-Cre recombinase in mouse cornea, we crossed Protein zero–Cre (P0-Cre) transgenic mice ¹² with mTomato-mEGFP mice. ¹⁶ The genotype of the mice was determined visually. floxN-cadh^+/+P0-Cre⁺ mice were generated by crossing P0-Cre mice with N-cadherin-floxed (floxN-cadh^+/+) mice (provided by Glenn L. Radice). ¹³ The genotypes of the mice were determined using allele-specific PCR, as described previously. ^12,13 The day of vaginal plug was considered to be embryonic day 0.5 (E0.5) and the day of birth was postnatal day 0 (P0). For analysis of “adult” corneas, we used 1.5- to 6-month-old animals, obtaining similar results. We handled mice in accordance with the ethics guidelines of the RIKEN Center for Developmental Biology. 

Antibodies and Stains

Mouse antibody against N-cadherin was obtained from BD Transduction Laboratories (610920; San Diego, CA), and rabbit N-cadherin antibody was obtained from Takara (M142; Shiga, Japan). Rabbit antibody against β-catenin was purchased from Sigma-Aldrich (C2206; St. Louis, MO), against ZO-1 from Invitrogen (mouse: 339100; and rabbit: 671300; Carlsbad, CA), against neuronal class III β-tubulin (Tuj-1) from Covance (MMS-435P; Princeton, NJ), against Na/K ATPase (ab7671-50) from Abcam (Cambridge, MA), and against p120-catenin from BD Transduction Laboratories (610133). For F-actin staining, a high-affinity F-actin probe (Alexa Fluor 594 phalloidin; Invitrogen) was purchased. For analysis of corneal endothelium permeability function, fluorescein dye (FLUORESCITE; Alcon, Tokyo, Japan) was used. 

Immunofluorescence

Mouse corneas were dissected and immediately fixed for 10 to 15 minutes in 4% paraformaldehyde at room temperature (RT). We used standard protocols for immunofluorescence as described previously, ¹⁴ with the following modifications: we used 0.5% Triton X-100 in TBS for 10-minute permeabilization of the samples. 

TUNEL Assay

For detection of apoptotic cell death in cornea, we used a cell death detection kit (In Situ Cell Death Detection Kit, TMR red; Roche Applied Science, Penzberg, Upper Bavaria, Germany), and followed the manufacturer's instructions. After mounting the samples, apoptotic cells were identified using a confocal microscope (Zeiss LSM780; Carl Zeiss MicroImaging GmbH, Jena, Germany). 

Fluorescein Analysis of Corneal Endothelium Permeability

Anesthetized mice were injected with fluorescein dye in the anterior chamber using a Hamilton syringe. After 30 to 40 minutes of incubation, the eyes were dissected in 4% paraformaldehyde and isolated corneas were further incubated in fixative for 30 minutes. Then the corneas were washed in PBS and immediately mounted for confocal imaging. Corneas collected from two wild-type and three mutant animals were used in the experiments. From each cornea, six to eight z-stacks were randomly acquired using a confocal microscope (LSM 710; Carl Zeiss MicroImaging GmbH). The images were further processed using microscopy imaging software (ZEN Imaging Software; Carl Zeiss MicroImaging GmbH) and the intensity of signal quantified with an imaging software package (ImageJ; National Institutes of Health, Bethesda, MD). A graphics editing program (Photoshop; Adobe, Mountain View, CA) was used to generate the artificial green-red gradient. 

Electron Microscopy

For electron microscopy imaging, corneas were isolated in 4% formaldehyde/0.01% glutaraldehyde in 0.1 M HEPES, pH 7.5. The samples were then processed as described previously. ¹⁵  

Statistical Analysis

For calculating the nonparametric Mann-Whitney Test, we used the online program at http://faculty.vassar.edu/lowry/utest.html (provided in the public domain by Vassar College, NY). A box-plot was generated using Smith's Statistical Package (SSP) from http://economics-files.pomona.edu/StatSite/SSP.html (provided in the public domain by Gary Smith, Pomona College, Claremont, CA). An unequal variance t-test was performed using Excel software package (Microsoft, Redmond, WA). 

Results

Assessment of P0Cre Expression in Cornea

In order to specifically ablate the N-cadherin gene in the corneal endothelium, we used the neural crest-specific P0 promoter. ¹² To verify the specificity of P0-driven Cre-recombinase expression, we used mTomato-mEGFP reporter mice. ¹⁶ In these mice, cells express a membrane-targeted tandem dimer Tomato (mT) fluorophore, which is replaced by a membrane-targeted green fluorescent protein (mEGFP) upon Cre-recombinase gene activation. In accordance with previously published data, ³ the green fluorescent protein reporter was expressed in the endothelial and stromal compartments of cornea, but not in its epithelial layer (see Supplementary Material and Supplementary Fig. S1). Green fluorescence was also detected in nerves traversing the corneal stroma, which is in accordance with the neural crest origin of the ophthalmic branch of the trigeminal ganglion, the main source of cornea innervation. ^17,18 Few stromal cells preserved red mT expression in mutant eyes. 

N-Cadherin Downregulation in the Mutated Corneal Endothelium

To confirm the original distribution of N-cadherin as well as its removal from the neural crest-derived compartments of corneal tissues, we performed N-cadherin immunostaining on both sectioned and whole-mount corneas of wild-type or mutant adult eyes. In the wild-type corneas, N-cadherin was detected almost exclusively at cell-cell boundaries in the endothelial layer (Figs. 1A, 1B). In mutant corneas, these N-cadherin signals disappeared (Figs. 1C, 1E). Upon careful examination, however, clusters of N-cadherin–expressing cells were discovered even in mutants (Fig. 1D). Approximately 0.5% to 20% of the endothelial cells were N-cadherin positive in mutant eyes. The number of cells per cluster ranged from several dozen to several hundred. This is in accordance with the previous finding of a mixed neural crest and mesoderm origin of corneal endothelial cells in mammals. ³ In addition, we detected local N-cadherin signals in both stroma and epithelial layer of wild-type corneas, which appeared to be of a neural origin, because of their colocalization with the neuronal marker Tuj-1 (data not shown). 

Figure 1. 

 

Expression of N-cadherin (N-cad) and β-catenin (β-cat). (A, B, F–H) Wild type endothelium. (C–E, I–K) Mutant endothelium. (A, C, D, F–K) Confocal images of whole mount cornea. Confocal plane was set to apical parts of corneal endothelial layer. (B, E) Confocal images of sections. Scale bars, 20 μm for (A, C, D, F–K); 12 μm for (B) and (E).

To assess the possibility that classical cadherins other than N-cadherin might be expressed in the adult mouse corneal endothelium, we performed immunostaining for β-catenin. This catenin is known to interact with any of the classical cadherins via their cytoplasmic domain in a 1:1 ratio. ¹⁹ In wild-type samples, we observed intense β-catenin immunostaining signals that colocalized with N-cadherin at the AJC (Figs. 1F–1552H). In mutant endothelial cells, however, β-catenin was hardly detectable at cell-cell contacts (Figs. 1I–1552K). We also looked at expression of p120-catenin, another catenin that ubiquitously binds the classic cadherins, and found its disappearance from endothelial cell-cell contacts in mutant corneas (see Supplementary Material and Supplementary Fig. S2). We concluded that N-cadherin is the major classical cadherin expressed in the adult mouse corneal endothelium, and its removal resulted in an almost complete absence of classic cadherins in this tissue. 

The Overall Defects of Mutant Cornea

All mutant eyes exhibited a varying degree of corneal opacity, which began around postnatal day 20 (Figs. 2A, 2B). Careful examination of dissected adult eyes showed the presence of anterior synechia in some of the mutants (data not shown). Hematoxylin-eosin (HE) staining of mutant cornea revealed a significant decrease in the thickness of the epithelial layer (Figs. 2C, 2D), combined with an increase in that of stroma (Fig. 2G), compared with wild-type. αE-catenin and DAPI staining showed that in the mutant epithelium, the basal layer remained relatively intact, whereas the superficial epithelial layers became thinner (Figs. 2E, 2F), suggesting that keratinocyte organization is impaired. Thus, although corneal epithelial and stromal cells do not display N-cadherin expression, these two compartments of the cornea were significantly affected by the removal of this molecule. 

Figure 2. 

 

Defects of cornea in mutant mice. (A) Dissected wild-type eyes. (B) Mutant eyes. (C, E) Sections of wild-type corneas. (D, F) Sections of mutant corneas. (C, D) HE staining. (E, F) αE-catenin (red) and DAPI (blue) immunostaining. (G) Comparison of thickness of wild-type (WT) and mutant (M) cornea. Black bar, whole cornea; white bar, corneal stroma only. **P < 0.005 (Student's t-test), n = 8. Scale bar, 20 μm for (C, D); 10 μm for (E, F).

Changes in Endothelial Junctional Architecture and Function

We performed detailed analysis of the changes in mutant endothelial layers that were the direct target for N-cadherin knockout. Classical cadherins interact with actin microfilaments through αE-catenin. ^20,21 In wild-type endothelial cells, F-actin was localized along cell-cell boundaries, forming a circumferential ring (Figs. 3A–1552C). From these rings, fine “spikes” protruded radially. In mutant cells, the actin ring and radial spikes were almost completely lost. Instead, stress fiber–like actin bundles became predominant (Figs. 3D–1552F). Next, we examined whether the TJs were normal or not by immunostaining for ZO-1, a TJ component (Figs. 3G, 3H). In wild-type endothelial cells, ZO-1 stains were organized into a linear circumferential ring like F-actin, whereas in mutants, these stains were severely fragmented, causing ZO-1–deficient areas of cell-cell boundaries in individual cells (Figs. 3G, 3H). 

Figure 3. 

 

Changes in actin cytoskeleton structure in mutant endothelium. Confocal microscopy (z-stacks encompassing the entire depth of endothelial layer). (A–C, G) Wild-type. (D–F, H) Mutant endothelium layer. F-actin was stained with phalloidin (white), and nuclei with DAPI (blue). (G, H) ZO-1 in red. (I) Average distance between nuclei of wild-type (WT) and mutant (M) endothelial cells (*P < 0.05 by Mann-Whitney test, n = 124 for wild-type; n = 90 for mutant cornea). Scale bar, 10 μm for (A–F); 6 μm for (G, H).

Changes were also observed in nuclear shape and spacing (Fig. 3). Nuclei in N-cadherin-expressing cells had a relatively uniform shape. In mutant endothelial layers, however, they often exhibited irregular shapes. The even spacing of nuclei in wild-type samples was also lost in mutants, that is, the distance between two adjacent nuclei showed a much broader distribution in the mutant versus wild-type samples (Fig. 3I), indicating that nuclear positioning became irregular in the absence of N-cadherin. 

Electron microscopy images of corneal sections shed additional light on the changes in mutant endothelia (Fig. 4). The smooth and even layer of endothelial cells in wild-type corneas was often replaced by “swollen” cells in mutant corneas (Figs. 4A, 4D). Areas with large vacuoles, which presumably represent open intercellular spaces, were also identified (Fig. 4G). This phenotype is consistent with partial loss of TJs. ² Excessive formation of microvilli was also observed in localized regions of mutant samples. Descemet's membrane, the basal membrane separating the endothelial cells from the underlying stroma, was relatively intact in mutant samples but showed a variation in thickness. Most importantly, high-magnification electron microscopy revealed that TJs, characterized by the close apposition of adjacent cell membranes, were missing in mutant endothelia (compare Fig. 4C and Figs. 4F, 4I). 

Figure 4. 

 

Electron microscopy. (A–C) Wild-type. (D–I) Mutant corneal endothelium. (C, F, I) Magnified area of apical cell-cell contacts (dotted squares from [B, E, H], respectively). Scale bar, 2 μm for (A, D, G); 0.5 μm for (B, E, H); 0.2 μm for (C, F, I). White arrows point to putative tight junctions.

The loss of TJs in N-cadherin–ablated endothelia suggested that their permeability barrier function was impaired. To test this possibility, we injected fluorescein dye into the anterior chamber of wild-type and mutant eyes, and assessed its penetration into the corneal stroma with the help of confocal microscopy (Fig. 5A). The results showed that mutant corneas consistently displayed a stronger overall labeling with fluorescein than wild-type ones (Figs. 5B, 5C). This finding supports the idea that N-cadherin gene mutation in the endothelial layers disrupted their permeability barrier function. In addition to these observations, we also found that the distribution of Na⁺/K⁺ ATPase in corneal endothelium was severely disorganized in the mutant eyes. This sodium pump was concentrated around cell-cell contacts in wild-type cells, whereas it diffused over wider areas of the cell membranes in mutant cells (see Supplementary Material and Supplementary Fig. S3). This aberrant redistribution of Na⁺/K⁺ ATPase might also have contributed to the abnormal permeability of the N-cadherin-deficient endothelium. 

Figure 5. 

 

Analysis of corneal endothelium permeability. (A) Schematic representation of experimental approach. Fluorescein was injected in the anterior chamber of the eye. (B) Confocal images of wild-type and mutant cornea labeled with fluorescein. Scale bar, 10 μm. (C) Box plot of fluorescence intensity in randomly chosen areas from z-stacks of wild-type (WT) and mutant (M) corneal stroma. Mann-Whitney test, ***P < 0.001. a, an approximate position of cornea analyzed in (B); AU, arbitrary units; ep, epithelial layer; end, endothelial layer.

Developmental Processes of Junction Disorganization

To assess when the endothelial anomalies appear during development, we observed corneas at early developmental stages. Corneal endothelial cells start to differentiate at around E15 in mouse embryonic eyes. It is reported that N-cadherin expression occurs already at E15. ^10,11 As the P0-driven Cre-recombinase expression is activated in neural crest cells at around E9.5, it is expected that N-cadherin expression is blocked from the beginning of endothelial development. This was confirmed by immunostaining E17.5 corneas for N-cadherin (data not shown). However, we did not find any sign of abnormalities in corneal architecture at embryonic stages. Therefore, we examined corneal endothelium at postnatal stages. Surprisingly, immunostaining for ZO-1 at postnatal day 4 (P4) showed that this TJ protein normally accumulates along cell-cell borders in the endothelium, despite the absence of N-cadherin (Figs. 6A–1552F). At P21, however, we now found that ZO-1 signals were significantly downregulated and disorganized in mutant endothelial cells (Figs. 6K, 6N). This was consistent with our observation that the corneal opacity became detectable around this developmental stage. 

Figure 6. 

 

Expression of cell junction proteins in corneal endothelium. (A–H) P4. (I–N) P21. (A–C, I–K) Wild-type corneal endothelium (z-stacks encompassing the entire depth of endothelial layer). (D–H, L–N) Mutant corneal endothelium (z-stacks encompassing the entire depth of endothelial layer). In (G, H), dashed line demarcates the mEGFP-positive (N-cadherin–deleted) and mEGFP-negative (N-cadherin–expressing) areas. Ncad, N-cadherin; β-cat, β-catenin. Scale bar, 10 μm for (A–F); 12 μm for (G, H); 18 μm for (I–N).

The absence of TJ defects in P4 endothelium suggested that other classical cadherins might be expressed in the endothelium to compensate for the loss of N-cadherin. To test this idea, we examined the expression of β-catenin, which is known to consistently bind the classical cadherins. As predicted, clear β-catenin immunostaining was detected at cell-cell boundaries in P4 mutant endothelium, although its expression level was slightly reduced in the N-cadherin-null zone, compared with the N-cadherin expressing areas (Figs. 6G, 6H). In P21 mutant samples, however, β-catenin signals were further downregulated (Figs. 6J, 6M), suggesting that N-cadherin is becoming dominant at this stage. These data are consistent with the idea that yet unidentified classical cadherins are expressed in corneal endothelial layer during embryonic and early postnatal stages, and they disappear afterwards. On the other hand, we failed to identify any morphological changes in corneal epithelium (see Supplementary Material and Supplementary Fig. S4) and stroma (data not shown) during these postnatal stages. This suggests that the histological changes in these two layers, which are already visible in 1.5-month-old animals, take place between 21 and 45 days after birth. 

Defects in Stromal and Epidermal Architecture

To investigate the structural background of the opacity, we looked at the organization of collagen fibrils in the stroma. Transmission electron microscopy (TEM) images revealed an irregular spacing of collagen fibrils in mutant corneas, contrasted with the relatively even spacing of wild-type collagen fibrils (Figs. 7A–1552F, 7K). On the other hand, no apparent difference was observed in the diameter of collagen fibrils between wild-type and mutant samples. These findings in N-cadherin-negative corneas were consistent with the changes observed during swelling of corneal stroma. ¹ Then, we examined the epithelial layer. As found by confocal microscopy, TEM revealed a smaller number of layers in mutant epithelium (Figs. 7G–1552J). The superficial layers in mutant samples showed signs of excessive delamination, especially in peripheral regions. Additionally, the cells of the basal layer appeared swollen in mutant eyes. 

Figure 7. 

 

Changes in collagen fibrils organization. (A–F, K) Corneal stroma. (G–J) Morphological changes in corneal epidermis. (A–C, G–H) Wild type samples. (D–F, I–J) Mutant samples. (C, F) Magnified area of dotted squares from (B, E), respectively. Scale bar, 2 μm for (A, D); 0.3 μm for (B, E); 0.06 μm for (C, F); 5 μm for (G–J). (K) Average distance between collagen fibrils in WT and M corneal stroma. Mann-Whitney test, **P < 0.005.

To further investigate the reason for these changes, we performed a TUNEL labeling assay on adult cornea. In normal cornea, there is a limited number of apoptotic cells localized mainly in central regions. ²² As expected from such observations, in wild-type samples, apoptosis was observed only in restricted areas of the cornea (Figs. 8A, 8B). In some mutant corneas, however, the apoptotic areas were greatly expanded; four out of six mutant corneas exhibited this phenotype (Figs. 8C, 8D). The apoptotic areas often spread from the central region to the very periphery of the cornea. The shape of the affected region was irregular and varied significantly from one sample to another. Confocal microscopy of whole-mount corneal samples confirmed that apoptotic cells were exclusively located in the epithelial layer (Fig. 8E). We failed to identify such apoptotic areas in postnatal corneas (P1–P21; data not shown). 

Figure 8. 

 

TUNEL staining of corneas. (A–B) Wild-type. (C–E) Mutant corneas. (E) Confocal microscopy z-stack of mutant cornea. cen, central cornea; pe, peripheral cornea. Scale bar, 100 μm for (A–D); 10 μm for (E).

Changes in the Nerve Plexus in Mutant Cornea

The cornea is innervated by the ophthalmic branch of the trigeminal ganglion, which derives from the neural crest. We investigated the innervation status in mutant corneas. For this purpose, we performed immunostaining for a neuronal marker, Tuj-1, on sections of wild-type and mutant cornea (see Supplementary Material and Supplementary Figs. S5A, S5B). Immunolabeling revealed relatively well-developed neural plexuses in mutant epithelium. Upon careful examination, however, we found areas missing innervation in N-cadherin-negative eyes (see Supplementary Material and Supplementary Fig. S5B). These areas were highly localized, not exceeding more than 20 to 40 μm, and were surrounded by regions with neural densities similar to those in the wild-type corneas. To assess the possible contribution of the observed neuronal phenotype to the increased apoptosis in epithelial layers, we performed double TUNEL and Tuj-1 staining (see Supplementary Material and Supplementary Fig. S5C). No clear correlation was, however, observed between the localization of apoptotic cells and neural plexus deficiencies: Apoptotic cells were often discovered in the immediate vicinity of nerve fibers. 

Discussion

In the present study, we analyzed the role of N-cadherin in maintaining the architecture and functions of corneal endothelium. To this end, we generated transgenic mice in which N-cadherin was selectively ablated in the neural crest lineage. We found that the genetic removal of N-cadherin from corneal endothelium severely impaired the formation of not only AJs but also TJs in this cell layer. These defects led to an increased endothelial permeability, as revealed by fluorescence dye injection experiments. Our results also indicated that Na⁺/K⁺ ATPase, which plays an important role in the directional flux of ions across the cell membranes, is mis-distributed in mutant endothelial cells, even though its expression level was not affected. Such changes in the distribution of Na⁺/K⁺ ATPase should have contributed to the impairment of electrochemical gradients across the mutant endothelial layer, and further deteriorated its “pump” function. Thus, N-cadherin loss seems to have affected the proper function of corneal endothelium via multiple mechanisms. The effects of N-cadherin removal on endothelial architecture were only evident in late-postnatal and adult mice. Our results suggested that other classical cadherins are expressed in embryonic and early postnatal endothelium to compensate for the loss of N-cadherin. 

Recent studies demonstrated that genetic removal of p120-catenin from the neural crest lineage induces disorganization of corneal architecture as well as its opacity, ²³ as seen in our N-cadherin-deleted eyes. These defects involved a downregulation of N-cadherin expression in corneal endothelium, as expected from the finding that p120-catenin is important for stabilizing classical cadherins in the cell membranes. ⁹ Thus, the observations of cornea's phenotypes in p120-catenin and N-cadherin knockout corneas are consistent. On the other hand, the p120-catenin deletion affected broader tissues in the eye than that of N-cadherin; this was most likely due to the ubiquitous expression of this catenin, compared with the tissue-specific expression of N-cadherin. 

The discontinuities in ZO-1 localization and the absence of TJ ultrastructure in N-cadherin–deficient endothelial cells suggest that this cadherin is required for the maintenance of TJs in these cells. Our findings partially contradict the results of studies on cell lines, ²⁴ in which the E-cadherin–catenin complex was shown to be necessary for the formation of TJs but not for their maintenance. Similarly, experiments using hepatocyte cells have shown that TJs can be formed and maintained even in the absence of the cadherin-catenin complex, albeit with some delay. ²⁵ Notably both experiments were done in vitro and the assembly of TJ complexes was observed on cell monolayers that were not subjected to external physical stress. In contrast, endothelial cells are exposed to various mechanical stresses in vivo. It is well documented that with age, there is a constant decrease in cell density of the endothelial layer. ^26,27 This observation, together with the very limited proliferative potential of adult corneal endothelial cells, ^28,29 suggests a mechanism in which cells migrate from more peripheral endothelium and cover more central parts. During these processes, individual endothelial cells should be subjected to significant mechanical stress. Moreover, the fluid of the anterior chamber undergoes a constant flow generated by temperature gradients, ³⁰ producing another type of force. In the absence of AJs, TJs become less stable, as discussed before. ³¹ In the absence of AJs, therefore, it is expected that mechanical stresses in vivo significantly enhance the disruption of TJs. 

In N-cadherin–deficient corneas, the stroma swelled and the epithelial layer became thinner. Since these corneal compartments do not express N-cadherin, it is assumed that their defects were indirectly produced as a result of dysfunction in other parts of the cornea. We speculate that the ionic or other physiological conditions were altered in the stroma as well as in epithelial layers due to the defective endothelial functions and these changes led to the structural abnormalities observed in these regions. Consistent with this idea, the defective ZO-1 localization in endothelial cells and the appearance of corneal opacity became evident at similar postnatal stages (at P20–21). The disturbance of collagen fibril arrangement in the stroma can explain the opacity of the mutant cornea. On the other hand, the structural swelling of the stroma was not detectable yet at these postnatal stages, suggesting that the molecular disorganization preceded the histological changes in this tissue. In the epithelial layer, most striking was the extended area of apoptosis, nearly reaching the corneal limbus in some mutant eyes. It is well known that mechanical stresses induce apoptosis, as studied in a variety of model systems. ^32,33 A similar process could occur in epithelial cells of N-cadherin–deficient corneas, which must be subjected to a significant increase in mechanical stress resulting from the swollen stroma. In support of this idea, epithelial cell death increases when rabbit corneas are subjected to shear stresses. ²²  

Another possible explanation for the increased apoptosis in the epithelial layer of mutant eyes is impaired innervation. The trophic function of corneal nerves is well documented. ^34,35 Our analysis of mutant cornea identified the presence of regions where innervation was missing. This is not surprising because the ophthalmic branch of the trigeminal ganglion is of neural crest origin ³⁶ and could be affected by the N-cadherin ablation in the current experiments. It must be noted, however, that these areas were highly localized and the overall nerve fiber density in mutant corneas was comparable to that seen in wild-type samples. We also found no correlation between apoptosis and the position of defective innervation. Furthermore, while endothelial abnormalities become detectable at postnatal stages, epithelial defects appear only in matured animals. Thus, the neurological component, if present, might be a secondary contributor to the changes of epithelial layers in our N-cadherin mutant mice. A precise mapping of corneal plexus nerves using whole-mount samples combined with apoptosis analysis in epithelial layers should give a more definitive answer to the question of neurological contribution to the observed mutant phenotype. In conclusion, our findings demonstrate that the N-cadherin–dependent AJC functions in the corneal endothelium play a crucial role in maintaining normal physiological states for the entire cornea. 

Supplementary Materials

pdf S1 

pdf S2 

pdf S3 

pdf S4 

pdf S5 

Acknowledgments

We are grateful to Hiver Sylvain and Saitou Hiroko for technical support, and LARGE in the RIKEN Center for Developmental Biology for mice. 

Footnotes

 Supported by the program Grants-in-Aid for Specially Promoted Research of the Ministry of Education, Science, Sports, and Culture of Japan (to MT).

Footnotes

 Disclosure: V.S. Vassilev, None; M. Mandai, None; S. Yonemura, None; M. Takeichi, None

Footnotes

 © ARVO