Short-Term Ultraviolet A Irradiation Leads to Dysfunction of the Limbal Niche Cells and an Antilymphangiogenic and Anti-inflammatory Micromilieu

Abstract

Purpose: We analyzed the effects of short-term ultraviolet A (UVA) irradiation on the putative limbal stem cell phenotype, limbal fibroblasts, corneal inflammation, and corneal (lymph)angiogenic privilege.

Methods: Primary human limbal epithelial cells and fibroblasts were irradiated with 5.2 J/cm² of UVA. The limbal epithelial cell phenotype was assessed using P63a, cytokeratin 15, integrin b1 (marking stem and transient amplifying cells), and cytokeratin 3 (a differentiation marker) as well as by a colony-forming efficiency (CFE) assay. An epithelial-fibroblast coculture model was used to compare the ability of irradiated and nonirradiated fibroblasts to support the putative limbal stem cell phenotype. The effects of the conditioned media of irradiated and nonirradiated cells on proliferation and tube formation of human lymphatic and blood endothelial cells also were tested. The levels of factors related to angiogenesis and inflammation were assessed in a protein array and using ELISA.

Results: Ultraviolet A induced phenotypical changes of limbal epithelial cells, as their CFE and putative stem cell/transient amplifying marker expression decreased. Limbal epithelial cells cocultured with UVA-irradiated limbal fibroblasts also exhibited differentiation and CFE decrease. Conditioned media from irradiated limbal epithelial cells and fibroblasts inhibited lymphatic endothelial cell proliferation and tube network complexity. Levels of monocyte chemoattractant protein 1 (MCP1) were reduced following UVA irradiation of both cell populations, while levels of IFN-γ increased in irradiated limbal epithelial cells.

Conclusions: These data imply a key role of cellular components of the limbal niche following short-term UVA irradiation. Overall, UVA irradiation leads to dysfunction of these cells and a anti(lymph)angiogenic and anti-inflammatory micromilieu.

The cornea protects the eye from solar ultraviolet (UV, 100–400 nm) irradiation by blocking approximately 95% of it while still allowing visible light (400–750 nm) to reach the retina and allow visual function.¹ The eye's exposure to UV radiation often is linked to acute clinical effects resulting in reduced vision. The cornea is particularly susceptible to UV due to its natural transparency and shape, which contributes to a peripheral light-focusing effect affecting the nasal limbus, where UV irradiation is 20-fold stronger.^2,3 Since the limbus is the habitat of the limbal epithelial stem cells (LESC), which maintain the transparency and homeostasis of the cornea by replenishing its outermost layer, the epithelium, any insult to this area may be detrimental for sight.⁴ When these stem cells become depleted by injury or disease, the neighboring conjunctival epithelium encroaches upon the corneal surface, inducing neovascularization, persistent epithelial breakdown, severe pain, and loss of vision.⁵ Although a limbal stem cell specific marker has not yet been identified, the expression of a battery of markers has been linked to the limbal stem cell phenotype, including P63α, ABCG2, cytokeratin 15, and, more recently, ABCB5,⁶ in combination with high levels of colony-forming efficiency (CFE).⁷ 

Previous research efforts have focused mainly on the impact of UVB radiation (∼295–315 nm) on human health, which is connected to chronic skin inflammation, skin aging, and melanoma.⁸ Ultraviolet B-related ocular pathologies include cataract, conjunctival melanoma, macular degeneration,^9–11 and pterygium, a benign tumor of the conjunctiva.¹² In contrast, UVA radiation (315–400 nm) carries lower energy and, therefore, is considered to have a less damaging effect. Nonetheless, recent reports demonstrate that UVA, too, is a major inducer of corneal photo-aging by causing changes to the stromal extracellular matrix.¹³ Also, UVA is a regulator of immunity, since it may induce systemic immunomodulation by reducing the numbers of epidermal Langerhans cells as well as upregulating mucosal immunoglobulin A in the intestine.¹⁴ Furthermore, UVA has been shown to modulate contact hypersensitivity of the skin and intestines by affecting mast cells.¹⁵ Ultraviolet A is used clinically in corneal cross-linking for the treatment of keratoconus and other corneal-ecstatic disorders. In these treatments, adverse effects in the limbal epithelial cell compartment have been reported. Specifically, it has been found that UVA treatment promoted the upregulation of genes connected to apoptosis,¹⁶ while signs of oxidative damage were found in basal limbal epithelial cells positive for the putative stem cell marker P63α, thus suggesting a link of UVA to an impairment of the limbal stem cell population.¹⁷ 

It has been speculated that UV damage to LESC or niche accessory cells, such as limbal fibroblasts, may affect the niche function and stem cell phenotype but the exact mechanisms remain elusive. In general, (photo)-aging of the LESC niche features, a change in the physical size and appearance of the putative stem cell niche structures located at the palisades of Vogt (which become limited to the light-protected, lid-covered parts of the superior and inferior limbus), as well as a decline in the proliferative efficiency of the LESCs. As previously mentioned, it furthermore has been linked to effects of UV radiation.¹⁸ 

While the main concern of most studies has been the assessment of the impact of UVB in the occurrence of pterygium, cataract, and skin tumors, the impact of UVA so far has been poorly studied. Epidemiologic studies have shown that the occurrence of pterygium was associated significantly with UVA and UVB.^19,20 However, the specific effect of UVA has yet to be defined. Ultraviolet B-induced cornea inflammation essentially is mediated by the increase of proinflammatory cytokines, such as IL-1,²¹ IL-6,²² IL-8,²² and TNF-α,²³ which relate to the increased inflammatory cell infiltrate associated with acute keratitis and pterygium. What currently is known about UVA radiation, however, is that it, next to inducing inflammation, is associated furthermore with the oxidative stress in cells, gene mutations, and extracellular matrix remodeling that contribute to the development of skin tumors and pterygium.^24–26 

Growth factors that have been observed to increase in pterygium specimens include VEGF-A,^27,28 VEGF-C, the master regulator of lymphangiogenesis,²⁹ and its receptor VEGF-R3.²⁷ This increase correlates with the higher density of the lymphatic network associated with pterygium recurrence and staging, and is considered a long-term effect of UV irradiation.^30,31 These changes, combined with a disruption of limbal stem cell function, may lead to corneal neovascularization. In this study, we use an in vitro approach to analyze the effect of UVA on the phenotype and functionality of limbal epithelial cells, on their accessory limbal fibroblasts, as well as on their paracrine-signaling, which regulates inflammation and (lymph)angiogenesis. 

Materials and Methods

Culturing of 3T3 Mouse Fibroblasts

A 3T3 mouse fibroblast cell line, a gift from the lab of Professor Nischt (Department of Dermatology, University of Cologne, Cologne, Germany) was cultured in Dulbecco's modified Eagle medium (DMEM; Life Technologies, Darmstadt, Germany), supplemented with 10% fetal bovine serum (Gibco, Darmstadt, Germany) and 1% penicillin/streptomycin/amphotericin (Life Technologies). The culture medium was refreshed three times per week and the cells were subcultured upon reaching 60% to 70% confluence at a ratio of 1:10. The cultures were kept at 37°C and at 5% CO₂ in air. To be used as a feeder layer for the culture of corneal epithelial cells, the 3T3 cells were growth-arrested in a culture medium containing 6 μg/mL mitomycin C (Sigma-Aldrich Corp., Munich, Germany) for 3 hours. 

Primary Human Limbal Epithelial Cell Harvesting and Maintenance

Human corneoscleral rims donated for research purposes and subjected to ethical approval procedures, as well as buttons, a surplus of surgery, were used for cell isolation in accordance with the Declaration of Helsinki. Human limbal epithelial (HLE) cells were cultured in a medium containing DMEM-F12 (1:1; Life Technologies) with added 10% fetal bovine serum, 1% penicillin/streptomycin/amphotericin (Life Technologies), 5 μg/mL human recombinant insulin (Sigma-Aldrich Corp.), 0.1 nM cholera toxin B (Sigma-Aldrich Corp.), 0.05 mM hydrocortisone (Sigma-Aldrich Corp.) and 10 ng/mL epidermal growth factor (Life Technologies). The culture medium was refreshed three times a week. Human limbal epithelial cells were isolated from corneas donated for research purposes supplied by the Cornea Bank of the Department of Ophthalmology at University of Cologne, Cologne, Germany. Whole corneoscleral buttons or limbal rims were immersed in a 1.2 U/mL dispase II solution (Sigma-Aldrich Corp.) for 2 hours at 37°C or overnight at 4°C. After the enzymatic treatment, the tissue segments were transferred into a 10 cm petri dish. The epithelial cells were scraped gently by using a feathered scalpel aiming at the limbal border to achieve an enriched LESC/progenitor population. The cells were collected using a 5 mL epithelial culture medium and then placed into a T-25 tissue culture flask (Nunc, Schwerte, Germany) containing a feeder layer of growth-arrested 3T3 fibroblasts at a cell density of 2.4 × 10⁴ cells/cm². The cultures were kept at 37°C and 5% CO₂ in the air. Epithelial colonies appeared after 3 to 5 days. 

Isolation and Culture of Human Limbal Fibroblasts

Human corneoscleral rims donated for research purposes and subjected to ethical approval procedures, as well as buttons, a surplus of surgery, were used for cell isolation in accordance with the Declaration of Helsinki. After isolating epithelial cells for culture, as described previously,^18,19 the scleral and corneal part of the rims were cut off to leave approximately 1 mm on either side of the limbus. Subsequently, the tissue was dissected further into smaller fragments. These pieces were allowed to adhere, epithelial side down, onto 10 cm petri dishes. The explants subsequently were cultured in DMEM (Life Technologies) plus 10% FBS and 1% penicillin–streptomycin (Life Technologies) until fibroblasts emerged. The cells were subcultured after approximately 2 to 3 weeks. The cells were passaged at a ratio of 1:2 and the medium was changed three times per week. 

Serum-Free Epithelial Fibroblast Coculture Model

Limbal fibroblast (not growth arrested) and limbal epithelial cells were cultured in a 1 to 3 ratio at 1.2 × 10⁴ cells/cm² and 4 × 10³/cm² respectively in the corneal epithelium medium described previously, omitting the serum as reported.³² The cultures were grown over 5 days, during which time the limbal epithelial cells prevailed. Due to the difference in the time of establishing the primary respective cultures, limbal epithelial and limbal fibroblast cells from different donors were used. The experiments were repeated a minimum of three times, every time with cells from different donors. 

Maintenance of Human Lymphatic and Blood Endothelial Cells

Primary human dermal microvascular lymphatic endothelial cells (LEC; catalogue number C12217, lot number 2010909.1) and blood endothelial cells (BEC, catalogue number C12225, lot number 0100505) where purchased from PromoCell (Heidelberg, Germany) and were maintained in a supplemented ECGM MV2 culture medium according to the manufacturer's instructions. Cells from a single donor were used in each case. The cells were passaged once, reaching 80% confluence by using a Trypsin/EDTA (0.04%/0.03%) solution for 2 minutes, followed by a trypsin neutralizing solution (0.05% Trypsin Inhibitor in 0.1% BSA; both by PromoCell). The cells were expanded up to passage 8. In each experiment, cells from 3 different passages were used. 

UVA Irradiation of Cells and Collection of Conditioned Medium

Limbal epithelial and limbal fibroblasts were plated in 2 × 10 cm Petri dishes per type, one to be subjected to UVA irradiation (from this point onwards referred to as AHLE and AHLF, respectively) and the other one to be used as a control. To avoid the use of feeder cells, HLE cells were expanded as described in the sections above and were separated from their feeder 3T3 cells using differential trypsinization. Subsequently, they were plated in a commercially available serum-free corneal epithelial culture media that does not require the use of feeders, CNT-50 (CellTech, Bern, Switzerland), while the HLF cells were placed in their normal culture media (DMEM supplemented with 10% FBS and 1% pen-strep). The cultures were left to reach approximately 90% confluence. Just before the irradiation, the culture media was replaced with PBS to avoid formation of toxic photoproducts from media components, serum, and phenol red. A VilberLourmat (Eberhardzell, Germany) Bio-Sun UV irradiator set at 265 nm was used to irradiate the cultures at 5.2 J/cm² (17 minutes). Following the irradiation, the PBS was replaced with their respective culture medium and the cultures were left for 24 hours to settle. Then, cells were plated in 6-well plates at a seeding density of 10⁵ cells per well. One day later, the culture medium was replaced with MV2 basal endothelial medium with added 2% FBS (basal medium; BM). The produced conditioned media were collected after 24 hours, centrifuged at 1500g to clear them of dead cells and debris, aliquoted, and stored at −80°C for a maximum of 2 months before use. The experiments were repeated a minimum of three times, every time with cells from different donors. 

Fluorescence Activated Cell Sorting (FACS) Analysis

To demonstrate that cell viability numbers were not affected due to the irradiation process, an Annexin V test was performed. This test assesses the number of apoptotic, dead, and live cells. The HLE and HLF cells were plated at a concentration of 2 × 10⁵ cells per 6-well plate well and in three replicates. Three different plates were used for the respective time points: 0 hours (no irradiation), 4 hours post-irradiation, and 24 hours post-irradiation (time of plating cultures for supernatant collection). The plates were irradiated as described in the section above. Following cell trypsinization, cells were labelled with a FITC-conjugated Annexin V system according to the manufacturer's instructions (Biolegend, Fell, Germany). The stained cells were resuspended in 600 μl PBS, 2% FCS, 10 mM HEPES, and 4 μl 7 AAD (to detect dead cells; Biolegend), and then passed through a 40-μm mesh before analysis in a Guava 12HT (Merck Millipore, Darmstadt, Germany). Data processing was performed using FlowJo (Tree Star; Ashland, OR, USA). These experiments were done with cells from 3 different donors. 

Immunocytochemistry

To identify changes in the phenotype of limbal epithelial cells following different treatments a panel of markers associated with different levels of cellular differentiation within the limbal epithelium has been used, namely integrin β1, p63α, cytokeratin (K)15, and cytokeratin (K)3. β1-Integrin, a marker normally observed in the basal and some of the suprabasal limbal epithelium,³³ the transcription factor p63α,³⁴ and K15,³⁵ are associated with limbal stem and progenitor cells and are absent in the superficial layers where the epithelial cells are considered terminally differentiated. On the other hand, K3 is expressed only in the superficial layers and is completely absent from basal and suprabasal cells.³⁶ 

The cells were irradiated and left to settle for 24 hours in their respective full media as described in section previous section. Subsequently, they were plated in eight-well Permanox chambered slides (Labtek, Nunc, Schwerte, Germany). In one set of experiments, HLE or AHLE were plated alone in CNT50 at a cell density of 2 × 10⁴cells/cm². In a second set of experiments, HLE were plated in a coculture with HLF or AHLF (not growth arrested) in a 3:1 ratio at 4 × 10³/cm² and 1.2 × 10⁴ cells/cm², respectively in the corneal epithelium medium described previously, omitting the serum (serum-free epithelial fibroblast co-culture model described above). After 5 days, the cells were washed three times with PBS, fixed for 10 minutes at room temperature in 4% (wt/vol) paraformaldehyde and, in case of not performing the staining immediately; they were treated with 20% (wt/vol) sucrose before storage at −20°C. The samples were blocked for 1 hour in PBS supplemented with 5% goat serum (Sigma-Aldrich Corp.) and 0.5% Triton X-100 (Sigma-Aldrich Corp.) followed by the mouse monoclonal K15 (LHK15) from Santa Cruz Biotechnologies (Heidelberg, Germany), rabbit polyclonal integrin β1 antibody from Abcam (Cambridge, UK), mouse monoclonal antibody for K3 from Merck Millipore, and rabbit polyclonal antibody for P63α from New England Biolabs (Frankfurt am Main, Germany) or blocking reagent only (negative control) overnight at 4°C. It should be noted that due to the lack of a commercially available antibody of the ΔNP63α isomer, which is more specific to putative stem cells,³⁴ an antibody recognizing the ΔN and TA variants has been selected and, therefore, may recognize stem cells and transient amplifying cells. 

Subsequently, the cells were incubated with their respective secondary antibody (goat anti-rabbit Alexa 488, goat anti-mouse Alexa 647; both from Life Technologies), and washed and counterstained with DAPI (Sigma-Aldrich Corp.). All incubations apart from the primary antibody incubation were performed at room temperature, and each step was intermittent with 3 × 5 minute rinses with PBS containing 0.1% Tween-20 (Sigma-Aldrich Corp.). Negative controls were treated in the same way, except for omitting the primary antibody step. A minimum of 3 random fields of each stained sample were photographed and the percentages of marker-positive cells were quantified using the plugin “cell count” of Image J (4′,6-diamidino-2-phenylendole [DAPI] was used to count the total number of cells). Stainings were repeated with cells from at least 3 different donors. 

CFE Assay

The HLE or HLF cells were irradiated and left to settle for 24 hours in their respective full media as described in the previous section. Subsequently, they were plated in 6-well plates. In one set of experiments, HLE or AHLE were plated alone in CNT50 at a cell density of 5 × 10⁵ cells/well. In a second set of experiments, HLE were plated in a coculture with HLF or AHLF (not growth arrested) in a 3:1 ratio at 4 × 10³/cm² and 1.2 × 10⁴ cells/cm², respectively in the corneal epithelium medium described previously, omitting the serum (serum-free epithelial fibroblast coculture model described above). The cells were grown in these conditions for 5 days and then the CFE assay was performed. 

For CFE assay,^18,32 3T3 fibroblasts were used as a feeder layer. The cells were treated with mitomycin C (Sigma-Aldrich Corp.) as above and plated at a cell density of 4.8 × 10⁵ cells in each well of a six-well plate and were allowed to attach overnight. Human limbal epithelial cells were seeded at a cell density of 1000 cells per well of the six-well plate. The cultures were fixed with cold methanol for 20 minutes at −20°C on day 12. Subsequently, the cells were stained with a solution of 1% rhodamine B (Sigma-Aldrich Corp.) and 1% Toluidine Blue (Sigma-Aldrich Corp.) for 30 minutes at 37°C. Finally, the plates were photographed and Image J software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA) was used to count the number of colonies that measured greater than 2 mm diameter. The percentage of CFE was calculated by using the equation:  

 

 

The experiments were performed with cells from three different donors (n = 6). 

Cell Metabolic Activity

Cell metabolic activity was assessed using the Alamar blue assay (Thermo Fisher Scientific, Schwerte, Germany). Limbal epithelial cells and limbal fibroblasts were cultured in 10 cm petri dishes and were UVA-irradiated as described in the section above. Then, 24 hours after the UVA irradiation, the cells were plated in 96-well plates at a cell density of 5 × 10³ cells per well in a minimum of 5 replicates. The Alamar blue assay was performed the following day. 

For the LEC and BEC cells, these too were plated in 96-well plates at cell density of 5000 cells/well and were left to settle overnight. Then, the endothelial culture medium (CM) was replaced with the various CM and the cells were left for another 24 hours before carrying out the Alamar blue assay. 

To perform the assay, the cultures were incubated for 1 hour in 150 μL/well Alamar blue reagent diluted 10 times in PBS (with n = 5 at minimum). Cell-free wells with added Alamar blue reagent were used as blanks. After incubation, the plates were analyzed in an Epoch plate reader (BioTek, Bad Friedrichshall, Germany) in absorbance mode at 570 and 600 nm and the percentage of reduction of the Alamar blue reagent was calculated as suggested in the manufacturer's instructions. These experiments were repeated with cells from at least 6 different donors. 

Scratch Wound Assay

Scratch wound assays were performed as described previously.²³ Briefly, LEC or BEC were plated to complete confluence in a 96-well plate, serum-starved for 2 hours and scratch-wounded using a 10 μL pipette tip (n = 5). Then, the cells were treated with the various CM (produced as described in a previous section). Here, too, as a control, LEC or BEC CM was used, respectively (spent medium). The wounds were photographed at 0, 2, and 6 hours for BEC and 0, 8, and 16 hours for LEC to accommodate for the slower wound closure rate of the latter. The wound surface areas at each time-point were measured using Image J software. The data of each replicate were presented as a percentage of the healed wound area compared to the original wound area at 0 hours. The experiments were performed a minimum of five times with cells LEC and BEC cells of three different passages using supernatants from different donors. 

Tube Formation Assay

The tube formation assays were performed on Matrigel (Corning, Wiesbaden, Germany) in μ-Slide angiogenesis assay (Ibidi, Planegg/Martinsried, Germany) according to the manufacturer's instructions. Then, BEC or LEC were seeded at a cell density of 1 × 10⁴/well in a complete endothelial cell medium. One hour later, the cells fully adhered on the Matrigel and the full medium was replaced with the supernatants and control media (n = 5). The tube networks formed were photographed after 16 hours using a Zeiss Primo Vert inverted microscope fitted with an AxioCam ERc5s camera (Zeiss, Munich, Germany). The number of branches, loops, and branching points were quantified using the Lymphatic Vessel Analysis Protocol (LVAP) plugin²⁸ in Image J²⁹ software. The experiments were performed a minimum of three times with cells LEC and BEC cells of three different passages using supernatants from different donors. 

Protein Microarray

Culture media from all groups were collected as described above and subsequently analyzed in a Proteome Profiler Human Angiogenesis Array (R&D Biosciences, Wiesbaden-Nordenstadt, Germany) according to the manufacturer's instructions. The arrays were visualized using a ChemiDocXRS+ System. 

Enzyme-Linked Immunosorbent Assay

Enyme-linked immunosorbent assay kits from R&D Biosciences were used for protein analysis according to the manufacturer's instructions. Angiogenin, IGBP-3, TNFα, MCP1, and IFNγ were quantified using the respective human Quantikine ELISA, while VEGFA and VEGFC were analyzed by using the corresponding human DuoSet. Each sample was analyzed in triplicate. The proteins were quantified in a conditioned medium from at least three donors. 

Statistical Analysis

Statistical analysis of results was done using Prism 6.0 software (GraphPad). One-way ANOVA with Tuckey's multiple comparisons test was used or a t-test with Mann-Whitney posttest as appropriate. Sets of data producing P < 0.05 were considered statistically significant. As explained in the sections above, the experiments were performed using a minimum of 3 experimental triplicates and repeated at least three times (using cells from a minimum of three different donors to reflect biological variability (primary HLE and HLF cells) or with 3 different cell passages (LEC and BEC cells). All error bars represent standard deviation values. 

Results

UVA Irradiation Induced Partial Loss of Putative Stem Cell Markers and a Reduction in the CFE of Limbal Epithelial Cells

Together with causing a transient reduction in HLE cell proliferation, UVA affected their putative stem cell phenotype. Specifically, the CFE of the irradiated limbal epithelial cells (AHLE) significantly declined compared to the one of the nonirradiated HLE (an approximately 2-fold decrease; P < 0.001, Fig. 1A). This also applied for colonies with a diameter larger than 2 mm, which are attributed to clones with higher potency (P < 0.05, Fig. 1A). At the same time, the AHLE cultures featured areas with a differentiated phenotype consisting of enlarged squamous cells, which were negative for integrin β1, (Figs. 1B, 1C, where negative regions are indicated by white arrows) as well as the P63a and K15 (Figs. 1D, 1E, white arrows highlight the negative regions and yellow arrows highlight brightly p63α stained cells). In addition, the AHLE cultures exhibited an increase in cells positive for K3, which were enlarged (Figs. 1F, 1G, respectively, white arrows indicate the K3 positive regions). The quantified results of the immunocytochemistry are illustrated in graph H, where it is shown that in the AHLE cultures, the number of cells positive for β1 integrin, P63α, and K15 are significantly lower (P < 0.01, P < 0.001, and P < 0.01, respectively), while the number of cells expressing the differentiation marker K3 is significantly higher (P < 0.05). 

Figure 1

 

Short-term UVA irradiation induced loss of putative stem cell markers causing differentiation of limbal epithelial cells. Colony-forming efficiency of irradiated limbal epithelial cells was significantly lower compared to their nonirradiated counterparts, indicating loss of proliferative potential as a result of UVA treatment (A). Immunocytochemistry of putative stem cell marker expression, including the basal markers integrin β1 ([B, C]; alexa488), p63α, and K15 ([D, E]; alexa488 and alexa647, respectively), and of the mature corneal epithelial marker K3 ([F, G]; alexa 647). In irradiated limbal epithelial cells, the markers integrin β1 (B, C) as well as p63α and K15 (D, E) were partially lost (areas indicated by white arrows), while areas expressing the differentiation marker K3 increased ([F, G], white arrows accentuated K3-positive regions). The immunocytochemistry data are quantified and summarized in (H). For CFE assay n = 6, and for immunocytochemistry n ≥ 3. *P < 0.05, **P < 0.01, ***P < 0.001. Scale bar: 50 μm.

UVA Irradiation of Limbal Fibroblasts Reduced Their Ability to Inhibit Limbal Epithelial Cell Differentiation in a Coculture Model

A similar, differentiated phenotype was displayed in parts of HLE cultures that were grown in coculture with AHLF cells. The cell morphology observations corresponded to CFE and marker expression data. Specifically, HLE cultured on irradiated HLF featured a significantly lower CFE compared to control cultures (2-fold decrease, P < 0.001, Fig. 2A). This also applies for colonies with a diameter larger than 2 mm (P < 0.05, Fig. 2A). Here, too, more areas of the cultures were observed where cells were enlarged and K3, while the presence of K3-positive cells increased (Figs. 2F, 2G). The immunocytochemistry results are quantified and summarized in Figure 2H. There, it is shown that in the HLE cells cocultured with irradiated HLF cells, the number of cells positive for β1 integrin, P63α, and K15 is significantly lower (P < 0.01, P < 0.05, and P < 0.001, respectively), while the number of cells expressing the differentiation marker K3 is significantly higher (P < 0.05). 

Figure 2

 

Short-term UVA irradiation of limbal fibroblasts induced loss of the putative stem cell phenotype of cocultured LECs. Limbal epithelial cells cocultured with irradiated limbal fibroblasts exhibited significantly lower CFE compared to the one of LECs cocultured with nonirradiated fibroblasts. This indicated loss of the ability of limbal fibroblasts to support limbal epithelial proliferative potential as a result of UVA treatment (A). Immunocytochemistry of putative stem cell marker expression, including the basal markers integrin β1 ([B, C]; alexa488), p63α and K15 ([D, E]; alexa488 and alexa647, respectively), and of the mature corneal epithelial marker K3 ([F, G]; alexa 647). In epithelial cells cocultured with irradiated limbal fibroblasts, the markers integrin β1 (B, C) as well as p63α and K15 (D, E) were partially lost (areas indicated by white arrows), while areas expressing the differentiation marker K3 increased ([F, G]; white arrows highlighted K3-positive regions The immunocytochemistry data are quantified and summarized in (H). For CFE assay: n = 6 and for immunocytochemistry n ≥ 3. *P < 0.05, **P < 0.01, ***P < 0.001. Scale bar: 50 μm.

UVA Altered the Paracrine Action of Limbal Epithelial Cells and Limbal Fibroblasts on LECs and BECs

To investigate the effect of secreted factors produced by irradiated and nonirradiated limbal epithelial cells and fibroblasts, their conditioned media was used to culture LEC and BEC. Their metabolic activity, wound-healing, and tube formation ability subsequently were evaluated. It should be noted that to avoid discrepancies due to possible cell death following UVA irradiation treatment, the cells have been left to settle for 24 hours after treatment before plating any of the experiments presented in this study. In parallel, an Annexin V FACS assay, which measures apoptosis and used 7AAD to label dead cells, was implemented to confirm that there were no significant differences in terms of the number of apoptotic, dead, and viable cells between nonirradiated cells and cells at 4 or 24 hours after irradiation. That latter time-point is when equal numbers of cells from all groups are plated to carry out experiments and collect supernatants. Therefore, the concern that any changes observed in the effect and protein-profiling of the supernatants is due to UVA-induced apoptosis/cell death is eliminated (Supplementary Fig. S1). 

It was demonstrated that CM from HLE stimulated more LEC metabolic activity, compared to CM from AHLE and the LEC's own CM (used as “spent” control; Fig. 3A; P < 0.01 and P < 0.001, respectively). Similarly CM from AHLF cells also induced significantly lower levels of LEC metabolic activity compared to CM from HLF and LEC CM (Fig. 3B; P < 0.01 and P < 0.0001). 

Figure 3

 

The paracrine activity of limbal epithelial cells and fibroblasts on LECs was affected by UVA irradiation. Conditioned media from limbal epithelial cells and limbal fibroblasts induced an increase in metabolic activity (A, B) and wound healing ([C, D]; 8 and 16 hours) of ***LECs, while this stimulatory effect diminished with UVA irradiation. A-tube formation assay (E–H) illustrate bright field microphotographs of LECs cultured in CM from limbal epithelial cells, irradiated limbal epithelial cells, limbal fibroblasts, and irradiated limbal fibroblasts, respectively) was used to compare the effect of soluble factors secreted by each group on the complexity of the formed vessel networks. (for each assay, n = 5, *P < 0.05, **P < 0.01 and ****P < 0.0001).

The metabolic activity observations concurred with the results of the wound-healing assay. In fact, the stimulating effect of CM from HLE compared to the one from AHLE and the control was consistent at the 8-hour time-point (Fig. 3C; P < 0.001 and P < 0.05), but not at the 16-hour one. Comparably, CM from HLF stimulated LEC wound closure more, as compared to AHLF and LEC at 8 hours (Fig. 3D; P < 0.01 in both cases), while AHLF also induced a higher percentage of wound closure than LEC (Fig. 3D; P < 0.05). A similar trend was observed at 16 hours with CM from HLF stimulating LEC more than the one of AHLF and LEC (Fig. 3D; P < 0.01 in both cases). 

A tube formation assay, illustrated in bright-field photos, was carried out to reproduce the ability of LEC cells to form vessels in vitro while cultured in CM from HLE, ALE, HLF, and ALF (Figs. 3E–H). A quantification of key morphologic features of the depicted tube networks made is included in supplementary data illustrating that UVA treatment of HLE cells did not impact branch number, loop number, and branching point CM of LECs cultured in the respective culture media (Supplementary Fig. S2). 

Culture media from HLF, on the other hand, stimulated the formation of a more complex network compared to CM of AHLF and LEC. Specifically, CM from HLF induced the formation of higher branch number (Supplementary Fig. S2D; P < 0.01 and P < 0.05), loop number (Supplementary Fig. S2E, P < 0.01 and P < 0.05) and branching points (Supplementary Fig. S2F, P < 0.01) for AHLF, there was no significant difference compared to the LEC control. 

In contrast to LEC, BEC did not display significantly different levels of proliferation or wound closure in response to the different CM (Figs. 4A–D, respectively). Also, quantification of the tube formation assay microphotographs (representative photos shown in Figs. 4E–H) indicated that CM from the different cell types had no significant effect on BEC cells in terms of branches, loops, and branching points (Supplementary Fig. S3). 

Figure 4

 

The paracrine activity of limbal epithelial cells and fibroblasts on blood endothelial cells remained unaffected by UVA irradiation. No significant differences were observed in the metabolic activity (A, B) and wound healing ([C, D]; 24 and 48 hours) of BECs cultured in the CM from irradiated and nonirradiated limbal epithelial cells and fibroblasts. Microphotographs from a tube formation assay are displayed in (E–H) where BECs were cultured in CM from limbal epithelial cells, irradiated limbal epithelial cells, limbal fibroblasts, and irradiated limbal fibroblasts, respectively.

Ultraviolet A Irradiation of Limbal Epithelial Cells and Limbal Fibroblasts Impacts on Soluble Mediators Regulating Inflammation and Lymphangiogenesis

To understand the potential effects of the various CM on BEC and LEC functionality, lymphangiogenic and inflammatory signals were investigated by first using a human angiogenesis protein array as a scanning tool. Then, we performed ELISA analyses to further assess key proangiogenic proteins, such as VEGF-A and -C, which were not included in the protein array panel. 

Semiquantitative densitometry analysis of a human angiogenesis protein array on the respective blots of CM from HLE and AHLE (in Fig. 5A) as well as HLF and AHLF (Fig. 5B) is illustrated in heat-maps including upregulations and downregulations of protein expression levels as well as the corresponding relative intensity fold changes. Major relative intensity fold changes (>2-fold) are highlighted with bold letters. A bar graph format of the densitometry data is included in Supplementary Fig. S4. 

Figure 5

 

Short-term UVA irradiation affected angiogenesis-related proteins produced by limbal epithelial cells and limbal fibroblasts. Protein array semiquantification analysis by densitometry is displayed in heatmap A for irradiated and nonirradiated epithelial cells, and in heatmap B for irradiated and nonirradiated fibroblasts. In heatmap A, limbal epithelial cells are compared to irradiated limbal epithelial cells, and in heatmap B, limbal fibroblasts are compared to irradiated limbal fibroblasts. Relative intensity changes higher than 2-fold are noted in bold letters (n = 2).

The first heat-map highlights protein level changes in CM of HLE cells and AHLE cells (Fig. 5A). A downregulation of IGFBP-1 (5.31-fold), TSP-1 (3.11-fold), endothelin1 (2.49-fold), amphiregulin (2.47-fold), IL-8 (2.39-fold), and IGFBP-2 (2.24-fold) levels was observed in CM from AHLE, while TSP1, uPA, SERPIN1, MCP1, and TIMP1 were to a lesser extent downregulated (all by less than 1.8-fold). Only CXCL16 exhibited a minor increase by1.42-fold. 

The second heat-map illustrates protein concentration differences between CM from cultured HLF and AHLF (Fig. 5B). Here, a substantial decrease in HLF-secreted amphiregulin, uPA and CXCL16 was noted (20.77-, 14.57-, 8.28-fold, correspondingly). SERPIN1, TIMP1, and TSP1 were mostly unchanged (each reduced by 1.26-, 1.23-, and 1.20-fold). On the other hand, MCP-1 was increased 2.16-fold, while smaller increases were observed for IL-8, IGFBP-1, IGFBP-2, and endothelin1 (1.54-, 1.47-, 1.4-, and 1.18-fold increases, respectively). 

To supplement the protein array, absolute quantification via ELISA analyses of more significant factors of angiogenesis were done (Figs. 6A–H). Here, the CM of LEC and BEC cells also were analyzed to assess their autocrine proangiogenic action as well as that of the BM. VEGF-A and VEGF-C, which have a fundamental role in hem- and lymphangiogenesis, were not represented in the array. The proteins angiogenin and IGFBP-3 were selected for further investigation. While differences in the protein array blots were visually observable, it was not possible to assess them reliably with densitometry as a result of the higher exposure times needed to create a background. Here, the CM of LEC and BEC cells also were analyzed to assess their autocrine proangiogenic action as well as that of the BM. 

Figure 6

 

Short-term UVA-irradiation modified the expression of key angiogenesis proteins produced by limbal epithelial cells and limbal fibroblasts. ELISA analysis of conditioned media from irradiated and nonirradiated limbal epithelial cells and limbal fibroblasts for VEGF-A (A, B), VEGF-C (C, D), angiogenin (E, F), and IGFBP-3 (G, H) was done. Limbal fibroblasts produced significantly higher amounts of VEGF-A, VEGF-C, angiogenin and IGFBP-3 compared to irradiated limbal fibroblasts (B, D, F, H). Moreover, there were significantly lower levels of VEGF-C in irradiated limbal epithelial cells compared to their nonirradiated counterpart ([C]; n = 3, *P < 0.05, **P < 0.01, and ***P < 0.001 and ****P < 0.0001). Four asterisks (****) situated above bars without brackets correspond to significance in comparison to the HLF group. (• denotes that the protein concentration was under the detectable levels of the assay.)

Specifically, VEGF-A levels were downregulated 1.5-fold in AHLF compared to HLF (P < 0.0001, Fig. 6B). The protein levels were similar in the CM from HLE and AHLE, while they were not detectable in the ones from LEC, BEC, and BM. Regarding VEGF-C, AHLE featured a (1.6-fold) downregulation compared to HLE, LEC, BEC, and the BM (Fig. 6C; P < 0.0001 in all cases VEGF-C in HLF-CM was significantly higher compared to AHLF, LEC, BEC, and BM (Fig. 6D; P < 0.0001 in all cases). 

Angiogenin ELISA data, as illustrated in Figures 6E and 6F, demonstrated no significant difference between HLE and AHLE, while LEC seemed to produce significantly lower protein levels compare to all groups (Fig. 6E; P < 0.0001). At the same time, CM from HLF contained significantly higher levels of angiogenin compared to AHLF (Fig. 6F; P < 0.05) and all other groups (P < 0.0001). 

Similarly, IGFBP-3 levels were comparable in HLE and AHLE, while both were significantly higher than LEC, BEC, and BM (Fig. 6G; P < 0.0001). A reduction of IGFBP-3 in AHLF compared to HLF was observed (Fig. 6H; P < 0.01), while both groups featured significantly higher protein levels compared to LEC, BEC, and BM (Fig. 6H; P < 0.0001). 

To explore potential changes in the proinflammatory factors produced by each cell type, MCP1, TNF-α, and IFN-γ were analyzed using ELISA (Figs. 7A–E). The respective analyses showed that MCP1 levels in AHLE-CM were significantly lower compared to the one of HLE cells (Fig. 7A; 4-fold, P < 0.0001), while both these groups featured significantly higher protein concentrations compared to LEC, BEC, and BM (Fig. 7A; P < 0.0001). The same reducing effect was observed in AHLF compared to HLF (Fig. 7B; P < 0.0001), while again CM from HLF and HLF exhibited higher MCP-1 concentrations compared to LEC, BEC, and BM (Fig. 7B; P < 0.0001). 

Figure 7

 

Short-term UVA irradiation modified the expression of key inflammation-related proteins produced by limbal epithelial cells and limbal fibroblasts. ELISA analysis of conditioned media from irradiated and nonirradiated limbal epithelial cells and limbal fibroblasts for MCP1 (A, B), TNF-α (C), and IFN-γ (D, E). Irradiated limbal epithelial cells and limbal fibroblasts exhibited lower levels of MCP1 compared to the nonirradiated ones (A, B), while irradiated limbal epithelial cells produced significantly more IFN-γ (D). Tumor necrosis factor-α levels remained unchanged in limbal epithelial cells (C). (n = 3, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). Four asterisks (****) situated above bars without brackets correspond to significance in comparison to the HLF group. (• denotes that the protein concentration was under the detectable levels of the assay.)

Tumor necrosis factor-α expression remained unchanged between HLE and AHLE, while the protein was not detected by the assay in any other group (Fig. 7C). Interferon-γ levels, on the other hand, increased significantly in AHLE compared to HLE (Fig. 7D, 2-fold; P < 0.05), while they remained unchanged in AHLF compared to HLF (Fig. 7E). Interestingly, CM from BM featured the highest IFN-γ levels compared to all the other groups, while BEC cells produced more of the protein compared to LEC cells (Figs. 7D, 7E). 

Discussion

We have used an in vitro approach to assess UVA-induced alteration in the phenotype, functionality and (lymph)angiogenic effect of human limbal epithelial cells and fibroblasts. It should be noted that this tissue culture approach cannot reflect with accuracy the function of these cells in situ; it still may provide an insight on the effects of UVA treatment on individual cell types which are key components of the limbal niche. The main finding of our study was that cells tapes assessed have a key role in the response to acute UVA radiation and subsequent (lymph)angiogenic and inflammatory events. Overall, acute UVA irradiation leads to a dysfunction of these cells and to a micromilieu that generally is anti(lymph)angiogenic and anti-inflammatory. 

While UVA used to be considered as relatively innocuous, it now is emerging as a factor to mutagenesis³⁷ and damage to DNA,³⁸ proteins, and lipids,^39,40 thus associating it with cancer⁴¹ and even dysregulation of immunity.¹⁴ In ophthalmology, even though it is acceptable to use UVA for the treatment of keratoconus via collagen cross-linking,⁴² studies have demonstrated that this technique becomes safer with the use of a protective ring around the limbal area to prevent damage to the stem cell compartment.⁴³ Previous work has shown alterations in the phenotype of the basal epithelium of UVA-irradiated eyes, as well as in corneal epithelial cell lines, including evidence of oxidative damage.¹⁷ Here, we demonstrated that primary LECs that did not receive UVA exhibited a cobblestone-like morphology. The cells were small and organized in tightly packed colonies, which expressed the markers P63a³⁴ and cytokeratin 15³⁵ which are associated with putative limbal stem cells and transient amplifying cells. The nonirradiated cells also were positive for the basal epithelial marker β1 integrin.³³ In short-term UVA-irradiated limbal epithelial cells, these markers were partially lost and they exhibited a differentiated morphology and a significant drop in CFE. These changes point toward loss of the putative stem cell phenotype as a result of even short-term UVA irradiation. 

Interestingly, similar changes in putative stem cell marker expression, CFE percentages, and cell morphology were observed in limbal epithelial cells that did not receive UVA irradiation, but which were cocultured with irradiated limbal fibroblasts. The serum-free coculture system used to coculture limbal epithelial cells and fibroblasts allows the limbal epithelial cells to prevail in culture while exhibiting an improved putative stem cell marker expression, colony phenotype, and CFE, the latter also confirmed in the present study.³² Well-functioning limbal fibroblasts are essential due to their facilitation of stem cell maintenance by providing the paracrine-signaling cues, including cytokines and growth factors that are necessary to maintain their phenotype.³² Here, we illustrated a UVA effect on the ability of limbal fibroblasts to allow expansion of limbal epithelial cells while successfully preserving a nondifferentiated phenotype, an effect that to our knowledge we report for the first time. 

After investigating the direct impact of UVA on limbal epithelial cells and limbal fibroblasts, other effects pertaining to their pro(lymph)angiogenic role were assessed. Conditioned media from the cells with and without UVA irradiation was collected to be used in functional assays of LECs and BECs. We investigated the synergistic effect of soluble factors produced by the cells on (lymph)angiogenesis. The Table illustrates the key proteins detected by protein array and ELISA and their known functions in relation to (lymph)angiogenesis and inflammation. As shown by the protein array assessment, all except one of the detected proteins were downregulated following UVA irradiation of HLE cells. The downregulated proteins included endothelin 1, IGFBP-1, IGFBP-2, IL-8 and TSP1. ELISA analysis of VEGF-C demonstrated a small but significant decrease in UVA-irradiated limbal epithelial cells, while VEGF-A levels remained unchanged. Given these data, especially considering the assessments of the master regulators of lymphangiogenesis and angiogenesis respectively VEGF-C and VEGF-A, it is possible to conclude that this protein expression profile is tipping the balance toward a reduced prolymphangiogenic effect of irradiated HLE cells compared to their nonirradiated counterpart and, to a lesser extent, a weaker prohemangiogenic one. These data correlate with the LEC and BEC proliferation data, which show that the supernatant of irradiated limbal epithelial cells seems to have a reduced stimulatory effect on LECs, while no significant difference is observed for blood endothelial cells. A comparable effect is reflected at the 8-hour time-point of the wound closure assay in LECs, while a similar but not significant trend is observed in the tube formation assay data; blood endothelial cells were again unaffected. The surprising antilymphangiogenic effect observed here may at least in part be due to the short duration of the UVA exposure. Future studies will need to analyze the effect of prolonged exposure and also putative direct effects of UVA on LECs. 

Table

 

Angiogenesis-Regulating and Proinflammatory Cytokines/Growth Factors and Their Roles

 

In the case of irradiated limbal fibroblasts, most of the detected proangiogenic factors appear to be reduced. There is a most substantial 100-fold drop of uPA, a strong promoter of angiogenesis.⁴⁴ Given this major reduction in uPA, the 2-fold reduction of Serpine 1 (plasminogen activator inhibitor 1) also observed here is less likely to affect the system, since its main function is the regulation of uPA. Amphiregulin and CXCL16 also were strongly downregulated in UVA-irradiated limbal fibroblasts while IGFBP-1 also was reduced. On the other hand, TIMP1 is linked to proangiogenic activity and, therefore, its downregulation, observed in irradiated limbal fibroblasts, would lead to increased neovascularization. 

Data from ELISA demonstrated that VEGF-A, VEGF-C, angiogenin, and IGBFP-3 also were significantly decreased following UVA treatment of limbal fibroblasts. These results suggest that the proangiogenic effect of HLF cells is revoked after short-term UVA irradiation, which corresponds to results for LEC proliferation, wound healing, and tube formation assay and BEC tube formation assay. 

While investigating the impact of UVA treatment on the angiogenic effect of HLE and HLF cells, differences in the produced inflammatory and macrophage-recruiting cytokines also were evaluated. We chose three key cytokines which regulate corneal inflammation, namely MCP1, TNF-α, and IFN-γ. Analysis by ELISA showed that UVA irradiation caused a downregulation of the macrophage recruiting cytokine MCP-1⁴⁵ in HLE and HLF cells, while TNF-α levels remained unchanged. Conversely, IFN-γ marginally increased in HLE cells only while remaining unchanged in HLF cells. These data demonstrated that UVA irradiation causes the cellular components of the limbal niche to partially reduce the proinflammatory effect. 

Taken together, these observations suggested that limbal epithelial cells and fibroblasts may respond to short-term UVA irradiation in two ways. Firstly, limbal epithelial cells and to a greater extent limbal fibroblasts, reduce their prolymphangiogenic activity by downregulating expression of cytokines, such as VEGF-A and VEGF-C, which have a vital role in corneal neovascularization. It is possible that the limbal cell populations apply an “acute phase” defense mechanism to prevent pathologic (lymph)angiogenesis. At the same time, these cells reduce their proinflammatory signals, thus helping in tissue repair, but also lessening the inflammatory conditions that propagate neovascularization by encouraging the intrusion of immune cells known to produce proangiogenic mediators.^46,47 It is especially the reduction of MCP-1 that may prevent the recruitment of macrophages, which are key indirect regulators of immune-amplification cascades and which induce hem- and lymphangiogenesis via producing VEGF-A, -C, and -D.^46,48 However, the UVA-induced changes in the limbal epithelial cell phenotype as well as in the limbal fibroblast function may threaten the limbal barrier and gradually affect corneal epithelial integrity, thus causing an indirect pro(lymph)angiogenic shift in the limbus. Therefore, long-term UVA exposure could contribute to pathologic neovascularization. The long-term effects of UVA have not been studied here and will be the subject of future studies. This study highlighted the changes that short-term UVA treatment induces to the limbal stem cell niche phenotype as well as the functions of its cellular components while preventing (lymph)angiogenesis via downregulation of macrophage-recruiting cytokines. Our results bring attention to the damaging effects of UVA to limbal niche phenotype and function, while at the same time demonstrating the beneficial effect of reducing prolymphangiogenic and proinflammatory conditions. Provided there is sufficient protection of the limbus, there may be promise in the further investigation of the therapeutic use of UVA in reducing corneal neovascularization and inflammation. 

Conclusions

Short-term UVA irradiation had a direct effect on the limbal epithelial cell phenotype as well on the ability of limbal fibroblasts to support limbal stem cell maintenance. Moreover, while limbal fibroblasts normally produce soluble mediators that promote blood- and lymphatic-endothelial cell activity, this effect is reversed following UVA irradiation. At the same time, short-term UVA also induces a downregulation of proinflammatory and macrophage-recruiting cytokines by limbal epithelial cells and fibroblasts. These data make evident the quintessential role of limbal epithelial cells and fibroblasts in the response of the niche to short-term UVA as a defense mechanism against subsequent inflammatory and (lymph)angiogenic events. Ultraviolet A irradiation alleviates the pro(lymph)angiogenic milieu at least partially via downregulation of macrophage-recruiting cytokines. Furthermore, these results underscore the importance of protecting limbal stem cells during corneal cross-linking in keratoconus patients. 

Acknowledgments

The authors thank Hans Gunter Simons and Sabine Hackbarth from the Cornea Bank of the University of Cologne, Cologne, for their help in this project. 

Supported by the DFG-funded Research Unit FOR2240 (www.for2240.de), and by Grants EU COST BM1302 (MN, FB, CC), DFG FOR 2240 (FB, PS, CC), and Brunner Foundation Cologne (MN). 

Disclosure: M. Notara, None; N. Refaian, None; G. Braun, None; P. Steven, None; F. Bock, None; C. Cursiefen, None