Protection of Human Corneal Epithelial Cells from Hypoxia-Induced Disruption of Barrier Function by Keratinocyte Growth Factor

Abstract

purpose. The possible detrimental effect of hypoxia on the barrier function of corneal epithelial cells and whether keratinocyte growth factor (KGF) might protect against such an effect were investigated.

methods. Simian virus 40–transformed human corneal epithelial (HCE) cells were cultured for 4 days to allow the establishment of barrier function. They were then deprived of serum for 24 hours before exposure to 1% (hypoxia) or 21% (normoxia) oxygen for 24 hours. Barrier function was evaluated by measurement of transepithelial electrical resistance (TER). The localization of ZO-1 and occludin was determined by immunofluorescence microscopy, and the expression of these tight junctional proteins as well as the phosphorylation of the mitogen-activated protein kinases ERK, p38, and JNK were examined by immunoblot analysis.

results. Hypoxia induced a decrease in the TER of HCE cells compared with that of cells maintained under normoxia. The localization of ZO-1 at cell–cell borders was disrupted by hypoxia, whereas the distribution of occludin was not affected. Hypoxia also induced the downregulation of ZO-1 and a decrease in the phosphorylation of ERK without affecting the phosphorylation of p38 or JNK. All these effects of hypoxia were inhibited by KGF. The effects of KGF on TER and ZO-1 localization in cells exposed to hypoxia were inhibited by PD98059, an inhibitor of ERK signaling. Neither hypoxia nor KGF exhibited mitogenic or cytotoxic effects in HCE cells.

conclusions. Hypoxia induces disruption of the barrier function of HCE cells by eliciting the redistribution and degradation of ZO-1, and this effect is inhibited by KGF in a manner dependent on ERK activation.

The corneal epithelium plays an important role in the maintenance of corneal transparency and provides a barrier that protects the cornea from harmful external agents, such as microbes and chemicals, and allows the establishment of a distinct internal environment. ¹ Hypoxia at the ocular surface promotes the initiation and progression of microbial infection and the development of corneal epithelial defects. Indeed, the hypoxia induced by extended wear of contact lenses increases the susceptibility of the eye to infectious keratitis. ^{2 3 4 5 6} The barrier function of human corneal epithelial cells is thus essential for the maintenance of ocular homeostasis. 

Tight junctions are present at the apical side of epithelia and play an important role in the establishment and maintenance of barrier function and cell polarity. ^{7 8 9} They are composed of three types of transmembrane proteins (occludin, claudin, and junctional adhesion molecules) and various associated cytoplasmic proteins including zonula occludens (ZO-1, ZO-2, ZO-3), cingulin, and 7H6 antigen. ⁸ Tight junctional proteins are also anchored to the actin cytoskeleton and thereby determine cell morphology. ^{8 10 11} Occludin, claudin, and ZO-1 have been localized to the apical cell–cell junctions of the corneal epithelium. ¹² These tight junctional proteins have been shown to be important for epithelial barrier function. ^{11 13 14}  

Keratinocyte growth factor (KGF) is a member of the fibroblast growth factor family of proteins and contributes to regulation of the proliferation and migration of corneal epithelial cells. ^{15 16} In the cornea, KGF is not expressed in corneal epithelial cells but is produced by stromal cells, whereas the KGF receptor is expressed only on epithelial cells. ¹⁵ Injury to the cornea triggers the release of various growth factors, including KGF. ^{17 18 19} KGF promotes the proliferation and migration of corneal epithelial cells during epithelial wound healing through the activation of mitogen-activated protein kinase (MAPK) signaling cascades. ^{20 21} These various observations suggest that KGF may play an important role in mesenchymal–epithelial interaction in the cornea. 

We have now examined the effects of hypoxia and KGF on the barrier function of corneal epithelial cells by measurement of transepithelial electrical resistance (TER). We also examined the expression and distribution of tight junctional proteins in such cells exposed to hypoxia or KGF. The possible contribution of MAPK signaling pathways to the effect of KGF on the barrier function of corneal epithelial cells exposed to hypoxia was also investigated. 

Methods

Materials

Dulbecco modified Eagle medium–nutrient mixture F-12 (DMEM/F-12), phosphate-buffered saline (PBS), fetal bovine serum (FBS), trypsin-EDTA, and gentamicin were obtained from Invitrogen-Gibco (Carlsbad, CA). Bovine serum albumin (BSA), recombinant bovine insulin, cholera toxin, recombinant human epidermal growth factor, recombinant human KGF, a mouse monoclonal antibody to actin, and a protease inhibitor cocktail were obtained from Sigma-Aldrich (St. Louis, MO). Six- or 24-well transwell plates and 96-well culture plates were obtained from Corning (Corning, NY). PD98059 was from Merck (Whitehouse Station, NJ). Rabbit polyclonal antibodies to extracellular signal–regulated kinase 1 or 2 (ERK1/2), to phosphorylated ERK1/2, to p38 MAPK, to phosphorylated p38, to c-Jun NH₂-terminal kinase (JNK), or to phosphorylated JNK were obtained from Cell Signaling (Danvers, MA), whereas those to ZO-1 or to occludin were from Zymed (San Francisco, CA). Horseradish peroxidase–conjugated goat antibodies to rabbit or mouse immunoglobulin G reagents (ECL Plus) were obtained from Amersham (Piscataway, NJ). Cyanine dimer dyes (TOTO-3) and Alexa Fluor 488–labeled goat antibodies to rabbit immunoglobulin G were from Invitrogen. An MTS assay was obtained from Promega (CellTiter 96 AQ_ueous One Solution Cell Proliferation Assay; Madison, WI). 

Cells and Cell Culture

Simian virus 40–immortalized human corneal epithelial (HCE) cells ²² were obtained from RIKEN Biosource Center (Tokyo, Japan). They were passaged in supplemented hormonal epithelial medium (SHEM), which comprises DMEM/F-12 supplemented with 15% heat-inactivated FBS, bovine insulin (5 μg/mL), cholera toxin (0.1 μg/mL), human epidermal growth factor (10 ng/mL), and gentamicin (40 μg/mL). For experiments, HCE cells were plated at a density of 5 × 10⁴ or 2 × 10⁵ cells per well in 24- or 6-well transwell plates (filter pore size, 0.4 μm), respectively, or at 1 × 10⁴ cells per well in 96-well plates. The cells were cultured at 37°C under an atmosphere of 21% oxygen, 5% CO₂, and 74% nitrogen first for 4 days in DMEM/F-12 supplemented with 15% FBS and then for 24 hours in unsupplemented DMEM/F-12. They were then cultured in unsupplemented DMEM/F-12 under conditions of normoxia (21% oxygen, 5% CO₂, 74% nitrogen) or hypoxia (1% oxygen, 5% CO₂, 94% nitrogen) in the absence or presence of KGF for 24 hours. 

Measurement of TER

The TER of HCE cells cultured in 24-well transwell plates, as described, was measured with the use of an epithelial volt-ohm meter (EVOM; World Precision Instruments, Sarasota, FL). Background resistance from the filter alone was subtracted from the experimental values. 

Immunofluorescence Microscopy

HCE cells cultured in 24-well transwell plates were fixed with 100% methanol for 20 minutes at room temperature. All cells were then washed with PPG− and incubated with 1% BSA in Ca²⁺- and Mg²⁺-free PBS (PBS−) for 1 hour at room temperature. They were then incubated for 2 hours with antibodies to ZO-1 [1:100 dilution in PBS− containing 1% BSA] or to occludin (1:100 dilution), washed with PBS−, and incubated for 1 hour with Alexa Fluor 488–conjugated goat secondary antibodies (1:2000 dilution) and cyanine dimer dyes (TOTO-3; Invitrogen; 1:2000 dilution) in PBS− containing 1% BSA. The cells were washed again before examination with a laser confocal microscope (LSM5; Carl Zeiss, Oberkochen, Germany). 

Immunoblot Analysis

HCE cells cultured in 6-well transwell plates, as described, were lysed in 300 μL of a solution containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 5 mM NaF, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM Na₃VO₄, and 1% protease inhibitor cocktail. The lysates were centrifuged at 15,000g for 10 minutes at 4°C, and the resultant supernatants were subjected to SDS-PAGE on a 7.5% gel. Separated proteins were transferred to a nitrocellulose membrane, which was then incubated at room temperature first for 1 hour with blocking buffer (20 mM Tris-HCl [pH 7.4], 5% dried skim milk, 0.1% Tween 20) and then for 2 hours with antibodies to ZO-1, occludin, or actin or to phosphorylated or total forms of ERK1/2, p38, or JNK, each at a 1:1000 dilution in blocking buffer. The membrane was washed with washing buffer (20 mM Tris-HCl [pH 7.4], 0.1% Tween 20) and incubated for 1 hour at room temperature with horseradish peroxidase–conjugated goat secondary antibodies (1:1000 dilution in washing buffer). Immune complexes were then visualized with enhanced chemiluminescence reagents. 

Assay of Cell Proliferation

The culture medium was removed from HCE cells cultured in 96-well plates, as described, and a mixture of 100 μL of DMEM/F-12 and 10 μL of MTS reagent was added to each well. The cells were then incubated for 2 hours at 37°C before measurement of absorbance at 490 nm with the use of a fluorescence plate reader (CytoFluorII; PerSeptive Biosystems, Framingham, MA). 

Cytotoxicity Assay

Culture supernatants of HCE cells cultured in 96-well plates as described were collected, diluted by a factor of 40, and analyzed for the activity of lactate dehydrogenase with a nonradioactive cytotoxicity assay. Absorbance at 490 nm was measured with a microplate reader (Bio-Tek, Richmond, VA). 

Statistical Analysis

Quantitative data are presented as mean ± SEM. Differences between two groups were analyzed by Student’s nonparametric t-test, and those among multiple groups were analyzed by analysis of variance (ANOVA) followed by Dunnett multiple comparison test. P < 0.05 was considered statistically significant. 

Results

To examine the effect of hypoxia on the barrier function of a monolayer of HCE cells, we measured TER. We previously showed that TER increased in a time-dependent manner during culture of HCE cells for 4 days and then reached a plateau, indicating that barrier function was established at this time. ²³ The cells were therefore cultured in serum-containing medium under an atmosphere containing 21% oxygen for 4 days after plating for TER to achieve a plateau. They were then cultured in serum-free medium first for 24 hours under 21% oxygen and for various times under 1% (hypoxia) or 21% (normoxia) oxygen. Whereas the TER of HCE cells maintained under the normoxic condition remained stable for up to 24 hours, that of cells exposed to hypoxia decreased gradually, with this effect already significant at 6 hours (Fig. 1) . 

We next investigated the effect of KGF on the hypoxia-induced decrease in the TER of HCE cells over 24 hours. Exposure of the cells to various concentrations of KGF (0–50 ng/mL) at the same time that the oxygen concentration was reduced from 21% to 1% resulted in a concentration-dependent inhibition of the effect of hypoxia on TER (Fig. 2) ; this effect of KGF was significant at concentrations of 10 to 50 ng/mL. 

To determine the effect of hypoxia on tight junctions in HCE cells, we examined the subcellular distribution of the tight junctional proteins ZO-1 and occludin by immunofluorescence analysis. Although ZO-1 was detected at the margin of each cell as continuous linear staining in the normoxic condition, exposure of the cells to hypoxia for 24 hours induced a loss of ZO-1 from the cell periphery (Fig. 3) . This effect of hypoxia was blocked by KGF (10 ng/mL), with the growth factor having no effect on ZO-1 localization in cells cultured under normoxia. The localization of occludin in HCE cells cultured under the normoxic condition was similar to that of ZO-1. However, in contrast to its effect on ZO-1 distribution, hypoxia did not affect the localization of occludin at the cell periphery (Fig. 4) . KGF also had no effect on occludin localization under normoxic or hypoxic conditions. 

We also examined the effects of hypoxia and KGF on the expression of ZO-1 and occludin in HCE cells. Immunoblot analysis revealed that exposure of the cells to hypoxia for 24 hours resulted in a marked decrease in the abundance of ZO-1 and that this effect was inhibited by KGF (10 ng/mL; Fig. 5 ). In contrast, the amount of occludin was not affected by hypoxia in the absence or presence of KGF (Fig. 5) . KGF had no effect on the expression of either ZO-1 or occludin under the normoxic condition. 

We next examined the possible role of the MAPKs ERK, p38, and JNK in the effects of hypoxia and KGF in HCE cells. Immunoblot analysis showed that exposure of the cells to hypoxia induced a time-dependent decrease in the level of ERK phosphorylation (activation) that was apparent between 3 and 24 hours compared with that observed under the normoxic condition and that this effect of hypoxia was blocked by KGF (10 ng/mL; Fig. 6 ). The phosphorylation level of p38 or JNK was not affected by hypoxia or KGF (Fig. 6) . The total amounts of ERK, p38, and JNK were also unaffected by hypoxia or KGF. 

To determine whether ERK activation is required for the ability of KGF to maintain the barrier function of HCE cells exposed to hypoxia, we treated the cells with PD98059 (10 μM), a specific inhibitor of MEK1 (an upstream kinase that activates ERK), for 1 hour before exposure to hypoxia or normoxia in the absence or presence of KGF (10 ng/mL) for 24 hours. PD98059 inhibited the effect of KGF on the hypoxia-induced decrease in TER. It also potentiated the hypoxia-induced decrease in TER and itself decreased TER in cells maintained under the normoxic condition in the absence or presence of KGF (Fig. 7) . 

In addition, we investigated the effect of PD98059 on the ability of KGF to maintain the localization of ZO-1 at the cell periphery during exposure of HCE cells to hypoxia. Immunofluorescence analysis revealed that PD98059 (10 μM) indeed slightly inhibited the effect of KGF (10 ng/mL) on the distribution of ZO-1 in cells subjected to hypoxia (Fig. 8A) . In contrast, the distribution of occludin at the margins of cells exposed to hypoxia was not affected by KGF (10 ng/mL) in the absence or presence of PD98059 (10 μM; Fig. 8B ). 

Finally, we investigated the effects of hypoxia and KGF on HCE cell proliferation and viability. MTS assay revealed that exposure of the cells to hypoxia for 24 hours in the absence or presence of KGF (10 ng/mL) had no effect on cell proliferation (Fig. 9A) . Similarly, measurement of the release of lactate dehydrogenase into the culture medium revealed that hypoxia did not exhibit a cytotoxic effect in HCE cells in the absence or presence of KGF (10 ng/mL; Fig. 9B ). KGF also had no effect on cell proliferation or viability under the normoxic condition. 

Discussion

We have shown that hypoxia (1% oxygen) disrupted the barrier function of a monolayer culture of HCE cells, as revealed by measurement of TER. Moreover, immunofluorescence analysis revealed that hypoxia induced the disappearance of the tight junctional protein ZO-1 from the interfaces of adjacent HCE cells. Immunoblot analysis showed that hypoxia also downregulated the expression of ZO-1. Neither the distribution nor the expression of occludin was affected by hypoxia. KGF inhibited the effects of hypoxia on the barrier function of HCE cells and on the subcellular localization and expression of ZO-1. Hypoxia reduced the level of ERK phosphorylation in HCE cells in a manner sensitive to KGF, and the preservation of both barrier function and distribution of ZO-1 by KGF in cells exposed to hypoxia was blocked by the MEK1 inhibitor PD98059. Neither hypoxia nor KGF exhibited a mitogenic or cytotoxic effect in HCE cells under our experimental conditions. Our results thus suggest that KGF inhibits the negative effect of hypoxia on the barrier function of HCE cells by maintaining the expression and interfacial distribution of ZO-1 in a manner dependent on the ERK signaling pathway. 

KGF has previously been shown to inhibit the disruption of barrier function in other types of epithelial cells, including retinal pigment epithelial cells and airway epithelial cells. ^{24 25} These previous observations, together with our present results, suggest that KGF may play an important role in the regulation of epithelial barrier function. 

ZO-1 associates with various integral membrane and cytosolic proteins at tight junctions, including occludin, claudin, ZO-2, and ZO-3. ^{26 27} Tight junctional proteins are also structurally and functionally associated with perijunctional actin filaments. ²⁸ ZO-1 binds directly to actin filaments and regulates tight junction assembly. ¹¹ Tight junctional proteins contribute to the formation of the corneal epithelial barrier and to maintenance of the internal environment of the cornea. ^{12 29 30} KGF has been shown to regulate the organization and stability of the actin cytoskeleton. ^{31 32} We have now shown that KGF maintained the localization of ZO-1 at the borders of HCE cells exposed to hypoxia. These various observations suggest that KGF may protect HCE cells from hypoxia-induced barrier disruption through an effect on the actin cytoskeleton in the region of tight junctions and the consequent maintenance of the junctional localization of ZO-1. 

We found that hypoxia also induced the downregulation of ZO-1 expression. The ubiquitin–proteasome pathway of protein degradation is activated in response to environmental stressors such as hypoxia and nutrient deprivation. ³³ The tight junctional protein occludin is degraded by this pathway in kidney epithelial cells. ³⁴ Our results suggest that hypoxia may induce the degradation of ZO-1 in HCE cells and that KGF may inhibit this degradation and thereby maintain the expression of ZO-1. In contrast to its effects on ZO-1, hypoxia did not substantially affect the localization or abundance of occludin in HCE cells. Hypoxia was previously found to have no effect on occludin expression in alveolar epithelial cells or endothelial cells, ³⁵ but it did induce the mislocalization of occludin to the cytoplasm, together with a decrease in ZO-1 expression in alveolar epithelial cells. Interaction of occludin with ZO-1 is thought to be required for proper localization of occludin to tight junctions. ³⁶ The reason for the apparent discrepancy in the effects of hypoxia on occludin localization between HCE and alveolar epithelial cells remains unclear. 

Hypoxia reduced the level of ERK phosphorylation in HCE cells without affecting that of p38 or JNK, and this effect was inhibited by KGF. ERK and p38 signaling pathways have been shown to contribute to regulation of the proliferation, survival, or migration of corneal epithelial cells. ^{37 38} The ERK pathway also plays a role in transcriptional regulation, ³⁹ and KGF simulates the transcription of various genes in epithelial cells. ^{40 41} These observations suggest that KGF might also maintain the expression of ZO-1 in HCE cells exposed to hypoxia through transcriptional regulation of the gene for ZO-1 or ZO-1-associated proteins through the ERK signaling pathway. 

The human corneal epithelium is exposed to low concentrations of oxygen as a result of the overnight wearing of rigid contact lenses or prolonged eyelid closure. ^{6 42} Such hypoxic conditions may induce disruption of the barrier function of the corneal epithelium and thereby increase the risk for corneal infection. Our present results suggest that the local application of KGF might serve to prevent such barrier dysfunction and to lower the risk for corneal infection under certain clinical conditions. 

 Submitted for publication November 14, 2007; revised January 19, 2008; accepted April 21, 2008.

 Disclosure: S. Teranishi, None; K. Kimura, None; K. Kawamoto, None; T. Nishida, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Hypoxia-induced loss of barrier function in HCE cells. Cells were cultured in transwell plates for 4 days under 21% oxygen. They were then deprived of serum for 24 hours before exposure to 21% oxygen (normoxia •) or 1% oxygen (hypoxia ○) and measurement of TER at the indicated times. Data are mean ± SEM from four independent experiments. *P < 0.01 versus the corresponding value for cells cultured under normoxia (Student’s nonparametric t-test).

Figure 2.

 

Effect of KGF on the hypoxia-induced loss of barrier function in HCE cells. Cells were cultured as described in Figure 1 , with the exception that the cells exposed to hypoxia were incubated in the presence of the indicated concentrations of KGF. TER was measured 24 hours after the switch to hypoxia. Data are mean ± SEM from four independent experiments. *P < 0.05; †P < 0.01 versus the value for cells cultured under hypoxia without KGF (ANOVA followed by Dunnett test).

Figure 3.

 

Effects of hypoxia and KGF on ZO-1 distribution in HCE cells. Cells were cultured for 4 days under 21% oxygen. They were then deprived of serum for 24 hours before exposure to normoxia or hypoxia for 24 hours in the absence or presence of KGF (10 ng/mL). The cells were then fixed, stained with antibodies to ZO-1 (green) and with cyanine dimer dyes (TOTO-3; Invitrogen; blue), and examined by fluorescence microscopy. Scale bar, 20 μm. Data are representative of three independent experiments.

Figure 4.

 

Effects of hypoxia and KGF on occludin distribution in HCE cells. Cells were cultured as in Figure 3 , fixed, stained with antibodies to occludin (green) and with cyanine dimer dyes (TOTO-3; Invitrogen; blue), and examined by fluorescence microscopy. Scale bar, 20 μm. Data are representative of three independent experiments.

Figure 5.

 

Effects of hypoxia and KGF on ZO-1 and occludin expression in HCE cells. Cells were cultured under the conditions described in Figure 3 . Cell lysates were then prepared and subjected to immunoblot analysis with antibodies to ZO-1, occludin, and actin (loading control). Data are representative of three independent experiments.

Figure 6.

 

Effects of hypoxia and KGF on the phosphorylation level of MAPKs in HCE cells. Cells were cultured under the conditions described in Figure 3 . Cell lysates were prepared at the indicated times after the switch to hypoxia (or normoxia) and were subjected to immunoblot analysis with antibodies to phosphorylated (p-) or total forms of ERK1/2, p38, or JNK. Data are representative of three independent experiments.

Figure 7.

 

Inhibitory effect of the MEK1 inhibitor PD98059 on the ability of KGF to maintain the TER of HCE cells exposed to hypoxia. Cells were cultured in transwell plates for 4 days under 21% oxygen. They were then deprived of serum for 24 hours, incubated in the absence or presence of PD98059 (10 μM) for 1 hour, and exposed to hypoxia or normoxia for 24 hours in the additional absence or presence of KGF (10 ng/mL) before measurement of TER. Data are mean ± SEM from four independent experiments. *P < 0.01 for the indicated comparisons (ANOVA followed by Dunnett test).

Figure 8.

 

Effects of the MEK1 inhibitor PD98059 on the distribution of tight junctional proteins in HCE cells exposed to hypoxia in the absence or presence of KGF. Cells were cultured for 4 days under 21% oxygen. They were then deprived of serum for 24 hours, incubated in the absence or presence of 10 μM PD98059 for 1 hour, and exposed to hypoxia for 24 hours in the additional absence or presence of KGF (10 ng/mL). The cells were then fixed, stained with antibodies to (A) ZO-1 or (B) occludin (green) and with cyanine dimer dyes (TOTO-3; Invitrogen; blue), and examined by fluorescence microscopy. Scale bars, 20 μm. Data are representative of three independent experiments.

Figure 9.

 

Lack of mitogenic and cytotoxic effects of hypoxia and KGF in HCE cells. Cells were cultured in 96-well plates under the conditions described in Figure 3 . They were then subjected to MTS assay for evaluation of cell proliferation (A), or culture supernatants were assayed for the release of lactate dehydrogenase for evaluation of cytotoxicity (B). All data are expressed as absorbance at 490 nm and are the mean ± SEM from four independent experiments.

The authors thank Makoto Asashima and Satoshi Ebina (ICORP Organ Regeneration Project, Japan Science and Technology Agency) for helpful discussion on the manuscript, and Yasumiko Akamatsu and the staff of Yamaguchi University Center for Gene Research for technical assistance.