Inhibition of Vascular Endothelial Cell Morphogenesis in Cultures by Limbal Epithelial Cells

Abstract

purpose. To study the in vitro angiogenic activity of human conjunctival and limbal epithelial cells and conjunctival, limbal, and corneal fibroblasts in a three-cell-type coculture model.

methods. Human umbilical vein endothelial cells (EC) were cocultured with epithelial cells, fibroblasts, or epithelial cells and fibroblasts to test their effect on EC morphogenesis. Neutralizing antibodies to some known angiogenic factors were added to the culture to see whether the EC morphogenesis may be blocked by a particular antibody.

results. Conjunctival and limbal epithelial cells exhibited very little or no stimulatory effect on EC tube formation when examined in an EC–epithelial cell coculture system. In contrast, conjunctival, limbal, and corneal fibroblasts all promoted EC morphogenesis when examined under the same culture conditions. Fibroblast-induced EC morphogenesis was inhibited by addition of anti-vascular endothelial growth factor (VEGF) and/or anti-basic fibroblast growth factor (bFGF) antibodies to the culture medium. In the three-cell-type coculture system consisting of ECs, fibroblasts, and epithelial cells, limbal epithelial cells (but not conjunctival epithelial cells) exhibited a strong inhibitory effect on fibroblast-induced EC tube formation.

conclusions. The proangiogenic activity of ocular surface fibroblasts is probably mediated through a paracrine mechanism by VEGF and bFGF. Limbal epithelial cells, but not conjunctival epithelial cells, inhibit fibroblast-stimulated angiogenesis.

Angiogenesis plays important roles in normal tissue functions and in many pathogenic processes. In the ocular surface, vascularization is under strict control, to maintain the transparency and immune privilege of the cornea and the normal differentiation of ocular surface epithelial cells. ¹ Derangement of the angiogenic process is believed to have a role in certain external eye diseases. Excessive corneal neovascularization induced by persistent inflammation seen in chemical or thermal burns, infections, autoimmune diseases, such as Stevens–Johnson syndrome and ocular cicatricial pemphigoid, is deleterious to vision. Ischemia of the sclera caused by complications of pterygium excision or severe chemical injuries is also troublesome. Without vascularization, the sclera is subjected to melting and is prone to infection, which again is sight threatening. 

Recently, there have been extensive studies concerning corneal angiogenesis; however, most of the studies were focused on the effects of exogenous angiogenic cytokines on corneal neovascularization, ² and only a few have discussed the mechanism(s) underlying the normal vascularization pattern of the ocular surface. ^{3 4} Presumably, in a healthy cornea, antiangiogenic factor(s) is secreted to counteract the angiogenic stimuli from adjacent tissues. When such balance is tipped toward angiogenesis, corneal neovascularization may occur. However, the absence of angiogenic factors is an equally plausible explanation. 

The proliferative effect of ocular surface cells (i.e., conjunctival, limbal, and corneal epithelial cells ^{3 4 5} and fibroblasts ^{6 7} ) on cultured vascular endothelial cells (ECs) has been reported. However, there has been no report concerning their possible effects on EC differentiation. To see whether ocular surface epithelial cells and fibroblasts express differential angiogenic activities in vitro that may be correlated with the vascularization pattern of the ocular surface, we designed a three-cell-type coculture system to examine EC differentiation (tube formation) in the presence of various types of ocular surface cells. We found that all ocular surface fibroblasts promote the differentiation of ECs, and such promoting effect was suppressed by limbal epithelial cells (progenitors of corneal epithelial cells). In contrast, under the same culture conditions, conjunctival epithelial cells exerted very little suppressive effect. The promoting effect by fibroblasts is probably mediated by a paracrine mechanism through vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF). 

Methods

Tissue Procurement and Cell Culture

Human eye tissues were procured with a humane method that included proper consent and approval. The study was evaluated and approved by the Medical Ethics and Human Experiment Committee of Chang Gung Memorial Hospital and complied with the ARVO Statement for the Use of Animals in Research and the guidelines of the Declaration of Helsinki in Biomedical Research Involving Human Subjects. Cadaveric tissues were obtained from 22 donors with ages ranging from 2 to 79 years (mean ± SD = 37.3 ± 20.1 years). Conjunctival, limbal, and corneal fibroblasts were grown from tissue explants in HSE medium (25% Ham’s F-12, 75% Dulbecco’s modified Eagle’s medium, 5 × 10⁻⁹ M sodium selenite, 20 mM Hepes, 1.2 g/l sodium bicarbonate, 0.4 g/l histidine, and 5 × 10⁻⁵ M 2-aminoethanol) supplemented with 10% fetal bovine serum (FBS). Cells were used for experiment between passages two and three. 

For isolation of epithelial cells, conjunctival tissues and corneoscleral buttons were treated with 0.5% Dispase II (Boehringer–Mannheim, Indianapolis, IN) in phosphate-buffered saline (PBS) at 37°C for 2 to 3 hours, epithelial cells were then removed and dispersed by gentle scraping and repeated pipetting. Cells were collected by low-speed centrifugation and resuspended in basal medium supplemented with 0.5% dimethyl sulfoxide, 2 ng/ml mouse epithelial growth factor, 1 μg/ml bovine insulin, 0.1 μg/ml cholera toxin, and 5% FBS (SHEM with 5% FBS) and were used immediately in experiments. ^{8 9} Limbal and corneal epithelial cells are of the same lineage, but epithelial cells of human central cornea are terminally differentiated cells and did not proliferate in culture. ¹⁰  

Vascular ECs were isolated from human umbilical cord vein by type V collagenase (Sigma, St. Louis, MO) digestion. Cells were cultured in MCDB-107 medium (JRH Biosciences, Lenexa, KS) supplemented with 2% FBS, and a fibroblast growth factor–enriched fraction of porcine brain extract at 1 μg/ml. Cells were used below the fifth passage. 

Cocultures

Type I collagen was extracted from rat tail tendons, as previously described, ¹¹ and was stored as a 3 mg/ml solution in 0.3% acetic acid at 4°C. 

Endothelial cells were cocultured with fibroblasts, epithelial cells, or fibroblasts and epithelial cells to test their individual or combined effect on EC tube formation. When included, 8 × 10⁵ conjunctival or limbal epithelial cells were plated on the top of 3 ml type I collagen gel (1 mg/ml) in a six-well tissue culture plate in 2 ml SHEM with 5% FBS. The collagen gel, on which epithelial cells were plated, contained either nothing or were suspended with 8 × 10⁵ human conjunctival, limbal, or corneal fibroblasts. To make fibroblast-suspended collagen gel, 1 ml collagen solution (3 mg/ml) was mixed with 1 ml 2 × HSE medium and was neutralized with appropriate amount of 5 N NaOH. The mixture was then mixed with 1 ml 1 × HSE medium containing 8 × 10⁵ fibroblasts. Three milliliters of collagen-cell mixture was poured per well in a six-well plate and incubated at 37°C, where it gelled in a few minutes. 

The collagen gels contained with or without fibroblasts and with or without epithelial cell overlaying were cultured for 24 hours, and the medium was replaced with the same medium. The ECs were then harvested from subconfluent monolayer cultures and suspended in collagen solution at 5 × 10⁵ cells/ml, as described, and 1 ml was poured per Falcon cell culture insert (Becton Dickinson, Franklin Lakes, NJ). After the collagen–EC mixture had gelled, the inserts were placed to allow the interactions of EC with fibroblasts, epithelial cells, fibroblasts and epithelial cells, or acellular collagen gel. Cultures were maintained for 3 days and were fixed with 4% formaldehyde in PBS. Endothelial cell tube formation was then quantified by a computerized image analysis system connected to a phase-contrast microscope (IMT-2;Olympus, Lake Success, NY). At least three duplicates were performed in each experiment. Tube length in five random fields per insert was computed and was expressed as millimeters per well. The EC gels and the epithelial cell–fibroblast gels were then embedded in paraffin, and were processed for hematoxylin-eosin (HE) staining. In addition, the latter were also stained with periodic acid–Shiff (PAS) and Alcian blue to detect conjunctiva-specific goblet cells. 

Effect of Neutralizing Antibodies to VEGF and bFGF on Fibroblast-Induced Capillary Tube Formation

To explore the potential role of major angiogenic cytokines in the regulation of fibroblast-induced EC differentiation, 2 × 10⁵ limbal fibroblast in 1 ml/well type I collagen solution (1 mg/ml) was poured into a 24-well culture plate. After incubation at 37°C for 24 hours, 2.5 × 10⁵ EC in 0.5 ml collagen solution was overlaid on the fibroblast-suspended collagen gel. After the EC–collagen layer gelled, 0.5 ml MCDB-107 medium with 2% FBS was added per well. Neutralizing antibodies to VEGF (monoclonal, ranging from 0.1 to 1.0μ g/ml) and/or bFGF (polyclonal, ranging from 1.0 to 10 μg/ml, R&D Systems, Minneapolis, MN) were added. Irrelevant rabbit IgG (Sigma) was added at the same concentration range as control. Cells were cultured for 3 days, and the total capillary tube length per well was compared and expressed as a percentage of control. 

Statistics

Total capillary tube length between two groups were compared using two-sample unpaired t-test or the Mann–Whitney test when appropriate. All P determinations were two sided and were considered significant when P < 0.05. 

Results

After coculturing for 3 days, the fibroblast-suspended collagen gels usually contracted to a diameter half the original size. The overlaying epithelial cells formed a confluent cell sheet. Histologic examination showed that the conjunctival epithelial cell layer formed on fibroblast-dispersed gel was mostly one cell in depth with occasional patches that were two cells thick, and the limbal epithelial cell layer was a uniform monolayer of cells (not shown). The conjunctival epithelial cell layer contained PAS- and Alcian blue–positive cells, indicating the presence of glycoconjugate-expressing cells. In contrast, limbal epithelial cell layers were devoid of PAS- and Alcian blue–positive cells (not shown). This observation is similar to previous reports of cocultures of fibroblast with conjunctival epithelial cells or limbal epithelial cells under the submerged condition ^{9 12} and is consistent with the in vivo observation that conjunctival and limbal epithelial cells belong to separate lineages. ¹³  

When cocultured with acellular collagen gel, the ECs remained rounded, and there was no sign of EC morphogenesis. When cocultured with either conjunctival epithelial cells (Fig. 1 A or limbal epithelial cells (Fig. 2 A), most of the ECs still remained rounded; however, a few reorganized ECs were observed. On coculture with either conjunctival fibroblasts (Fig. 1B) , limbal fibroblasts (Fig. 2B) , or corneal fibroblasts (Fig. 3 A), prominent EC morphogenesis was observed. The ECs migrated and organized into capillary-like networks. 

To see whether the paracrine interactions between epithelial cells and fibroblasts have any influence on EC morphogenesis, conjunctival and limbal epithelial cells were cocultured with corresponding fibroblasts. In such three-cell-type cultures, the capillary tube formation by the ECs was reduced compared with those cocultured with fibroblasts alone. In three-cell-type cultures containing conjunctival epithelial cells and conjunctival fibroblast (Fig. 1C) or conjunctival epithelial cells and corneal fibroblasts (Fig. 3B) , the total length of the capillary network was slightly decreased. Figure 1 shows that the total capillary tube length in EC–conjunctival fibroblast coculture was 2974.0 ± 765.6 (mean ± SD) mm, and tube length in EC–conjunctival epithelial cell–conjunctival fibroblast coculture was 2757.8 ± 876.7 mm. (7.3% decrease; P = 0.585). Figure 3 shows that the total capillary tube length in the EC-corneal fibroblast cell coculture was 3846.0 ± 279.3 mm, and tube length in EC–conjunctival epithelial cell–corneal fibroblast coculture was 3535.9 ± 422.3 mm (8.1% decrease; P = 0.302). As is shown in Figure 1C , some cords were formed by clusters of EC instead of by well-aligned, elongated cells similar to those seen in conjunctival fibroblast-induced cultures. 

When conjunctival epithelial cells were replaced by limbal epithelial cells in the three-cell-type cocultures, EC morphogenesis was reduced significantly. As is shown in Figures 2B 2C and 2D , total capillary tube length in EC–limbal fibroblast coculture was 4789.9 ± 1104.2 mm, and that in EC–limbal epithelial cell–limbal fibroblast coculture was 2566.7 ± 566.3 mm (46.4% decrease; P < 0.001). A similar effect was observed when limbal fibroblasts were replaced by corneal fibroblasts. There was a 44.0% (P < 0.001) reduction in the total capillary tube length when limbal epithelial cells were included in the EC–corneal fibroblast coculture system (Fig. 3) . In EC–limbal epithelial cell coculture, ECs stayed rounded or were slightly elongated, but did not reorganize. 

The results clearly showed that conjunctival, limbal ,and corneal fibroblasts all induced prominent capillary tube formation. It would be interesting, therefore, to see what soluble factor(s) may be involved. ¹⁴ Various neutralizing antibodies to various selective EC mitogens were tested for their ability to block fibroblast-induced EC morphogenesis. ¹⁵ Although irrelevant antibody had no effect on EC morphogenesis, we found that neutralizing antibody to VEGF at 0.1 and 1.0 μg/ml inhibited capillary tube formation by 62.5% and 86.4%, respectively. Moreover, neutralizing antibody to bFGF was also effective; at 1.0 and 10 μg/ml, it exerted, respectively, a 46.7% and 60.1% inhibition. The inhibitory effect of both antibodies appeared to be additive. Simultaneous presence of both antibodies at appropriate concentrations almost completely inhibited fibroblast-induced capillary tube formation. 

Discussion

In vitro models of angiogenesis are usually incomplete, because only some early cellular events of the processes are investigated. However, it offers a simplified system to analyze the roles of cell–cell and cell–matrix interactions, and to avoid confounding factors such as inflammatory cells and factors from circulation. It is generally believed that corneal neovascularization is a manifestation of inflammation and that infiltrating leukocytes play a major role. ¹⁶ However, an earlier experiment showed that even in leukopenic mice, neovascularization still occurs in response to chemical injury, although to a much lesser degree. ¹⁷ It therefore raises the possibility that cells other than leukocytes or noncellular constituents of inflammation may also contribute to the angiogenic signals. Fibroblasts and epithelial cells from other tissues have been shown to express VEGF, ^{18 19 20 21 22} and to induce capillary formation by ECs. ^{23 24 25} It is, therefore, reasonable to hypothesize that ocular surface epithelial cells and fibroblasts also express angiogenic activities and participate in the regulation of ocular surface vascularization. 

It has been suggested that active cytokine cross talk between ocular surface epithelium and fibroblasts is present and may have a regulatory role in important cellular processes such as epithelial differentiation or corneal wound healing ²⁶ and, as proposed in this article, the angiogenic phenotype of the ocular surface. It is difficult, if not impossible, to analyze the effects of cytokine cross talk on cell behavior in the in vivo system. Thus, development of culture systems with closer mimicry of the in vivo tissue may be necessary if this problem is to be approached. In this regard, we recently reported the fabrication of in vitro human conjunctival equivalents from conjunctival epithelial cells, fibroblasts, and type I collagen and clearly showed the regulatory effect of stromal fibroblasts on epithelial cell development and differentiation. ^{8 9} In this study, we extended the coculture system to contain vascular ECs and examined the effect of epithelial–fibroblast interactions on EC differentiation. 

Studies of the proliferative effect (on ECs) and cytokine production by others all indicate that conjunctival epithelial cells are angiogenic, whereas limbal and corneal epithelial cells are antiangiogenic. ^{3 4 5 27} However, the role of stromal fibroblasts and epithelial–fibroblast interactions is less defined. In this study, we clearly showed that conjunctival, limbal, and corneal fibroblasts all promoted prominent capillary tube formation by ECs, indicating the expression of proangiogenic activity by these cells. The observation is supported by enzyme-linked immunosorbent assay and reverse transcription–polymerase chain reaction results (not shown) showing that all these cells express comparable amounts of VEGF protein and mRNA in culture. Although it has been reported that bFGF mRNA was differentially expressed among ocular surface fibroblasts ²⁶ and that interleukin-1β upregulates bFGF more in limbal than in corneal fibroblasts, ²⁸ the present study suggests that different ocular surface fibroblasts express comparable angiogenic activity in vitro. In our study, fibroblast-suspended collagen gels without overlaying epithelial cells may be viewed as a mimicry of the in vivo fibroblast activation during epithelial defect (wounding). With reepithelialization, fibroblast activity is then downregulated. Compared with conjunctival epithelial cells, limbal epithelial cells significantly downregulated corneal fibroblast–induced EC morphogenesis, indicating the antiangiogenic property of the limbal epithelial cell. 

Thus, ocular surface fibroblasts may not be a major determinant in the regional difference of ocular surface vascularization. Moreover, in normal in vivo conditions, fibroblasts are quiescent and exhibit very little cytokine or extracellular matrix production. It has been suggested that ocular surface cells may shift to an angiogenic phenotype in response to external stimuli such as hypoxia ²⁹ or inflammation ^{30 31} and may thus have implications in the pathogenesis of pterygium recurrence or pannus formation. In addition, fibroblasts may produce chemokines such as interleukin-8 in response to bacterial products. Interleukin-8 has been shown to induce corneal neovascularization ³² and regulates inflammatory reactions. ³³  

The in vitro observations of antiproliferative effect on ECs, ³ and the inhibition of EC morphogenesis by limbal epithelial cells in our study may not be readily interpreted to have antiangiogenic activity in vivo. However, some clinical observations are supportive of this notion. The stem cells of corneal epithelium reside in the basal layer of limbal epithelium. ³⁴ Clinical observations show that after severe chemical injuries, infection, or autoimmune diseases with significant loss of limbal stem cells, the corneal surface is repopulated with conjunctival epithelium accompanied with chronic inflammation and prominent fibrovascular ingrowth. ³⁵ Current treatment for this vision-threatening dilemma is to perform limbal transplantation. By providing healthy stem cells of corneal epithelium from the contralateral eye ^{36 37} or cadaveric eyes, ^{38 39} the surgery not only restores corneal epithelial phenotype, but also effectively reduces corneal neovascularization. Because limbal fibroblasts are proangiogenic in phenotype, it is thus possible that transplanted limbocorneal epithelial cells exert antiangiogenic activity. 

 Reprint requests: Jan-Kan Chen, Department of Physiology, College of Medicine, Chang Gung University, 259, Wen-Hwa First Road, Kwei-San, Tao-Yuan, Taiwan, Republic of China.

 Supported by Chang Gung Memorial Hospital Grant CMRP-645E (1996) and National Science Council of Taiwan Grant NSC 87-2314-B-182A-042 (1997).

 Submitted for publication April 28, 1998; revised December 7, 1998; accepted January 15, 1999.

 Proprietary interest category: N.

 

Figure 1.

 

Effect of endothelial cell tube formation by human conjunctival epithelial cells and/or human conjunctival fibroblasts. Endothelial cells (5 × 10⁵/ml) were cultured in collagen gel in the culture insert with conjunctival epithelial cells (A), conjunctival fibroblasts (B), or conjunctival epithelial cells and conjunctival fibroblasts (C). Three days later, cultures were fixed and photographed, and total capillary tube length per dish was calculated and expressed as mean ± SD (D). Original magnification, ×100.

Figure 2.

 

Effect of endothelial cell tube formation by human limbal epithelial cells and/or human limbal fibroblasts. Endothelial cells (5 × 10⁵/ml) were cultured in collagen gel in the culture insert with limbal epithelial cells (A), with limbal fibroblasts (B), or with limbal epithelial cells and limbal fibroblasts (C). Three days later, cultures were fixed and photographed, and total capillary tube length per dish was calculated and expressed as mean ± SD (D). Original magnification, ×100.

Figure 3.

 

Effect of human corneal fibroblasts on endothelial cell tube formation in the absence or presence of human conjunctival epithelial cells or human limbal epithelial cells. Endothelial cells (5 × 10⁵/ ml) were cultured in collagen gel in the culture insert either with corneal fibroblasts (A), with corneal fibroblasts and conjunctival epithelial cells (B), or with corneal fibroblasts and limbal epithelial cells (C). Three days later, cultures were fixed and photographed, and total capillary tube length per dish was calculated and expressed as mean ± SD (D). Original magnification, ×100.

The authors thank Shuen-Kuei Liao and Ying-Tung Lau, College of Medicine, Chang-Gung University, for initiating the present study; Lai-Chu See and Shu-Mei Lin, for statistical analysis; Li-Lin Lu, for technical support; I-Lin Lee and Kuei-Ching Lee, for graphing; and Sheffer C. G. Tseng and Andrew J. W. Huang, Bascom Palmer Eye Institute, University of Miami, for helpful discussions.