Vitronectin: A Migration and Wound Healing Factor for Human Corneal Epithelial Cells

Abstract

Purpose.: The stratified squamous epithelial covering of the cornea provides one of the first physical and immunological lines of defense. A breach in its integrity results in a wound healing response that resolves quickly. However, under certain conditions—for example, when migrating cells are unable to adhere to the basement membrane—healing may be delayed. The aim of this study was to determine whether vitronectin (VN) promotes corneal epithelial wound healing.

Methods.: Primary human corneolimbal epithelial cells and the human corneal epithelial cell (HCEC) line were cultured and monitored to determine the rate of epithelial recovery after injury. Human corneas were placed in organ culture and the epithelium debrided. Therapeutic contact lenses (CLs) soaked in saline or coated in a solution of recombinant human VN were applied and the epithelium assessed histologically after 7 days.

Results.: Vitronectin (5 μg/mL) significantly enhanced the wound closure rate in cultured HCECs as well as in primary human corneal epithelial cells. Wound recovery was significantly blocked with a cyclic arginine-glycine-aspartate (RGD) peptide (10 μg/mL), a neutralizing antibody to VN (25 μg/mL) as well as after transiently silencing β5-integrin. Wound closure was not related to the effect of VN on cell proliferation. Finally, incubating epithelial-debrided human corneas with CLs loaded in VN resulted in effective re-epithelialization of the injured surface.

Conclusions.: Vitronectin is a key extracellular matrix protein that expedites corneal epithelial wound recovery in vitro and ex vivo. Delivery of VN on therapeutic CLs may be an effective and efficient way to treat patients with persistent epithelial defects of the ocular surface.

Introduction

The cornea serves a dual function: first, as a key tissue through which light is refracted and transmitted for exquisite vision, and second, as a protective barrier to the external environment. The very superficial layer of the cornea consists of a multilayered squamous epithelium and its integrity and transparency is critical for sight. A compromised ocular surface renders the cornea liable to vision-threatening infection, neovascularization, scarring, and melting, with the risk of perforation and loss of the eye. ¹ Resurfacing of the corneal epithelium after an injury occurs in three phases: migration of adjacent intact epithelial cells to cover the injured zone, proliferation of the migrated monolayer to reestablish epithelial thickness, and differentiation of the reconstruct epithelium to restore structure and function. ² The initial stage is characterized by cells attaching and spreading over the denuded substratum which typically occurs by epithelial “sliding.” ³ An important aspect of this process is cell attachment to the extracellular matrix (ECM), which is mediated by integrins. ⁴ Integrins and their ligands are components of structures known as focal contacts which facilitate cell-to-ECM interactions during wound-healing. ⁵ Extracellular matrix proteins involved in cell adhesion, motility, and corneal epithelial wound healing include laminin-5, fibronectin (FN), collagen IV, and vitronectin (VN). ^6–10  

Most corneal epithelial wounds are promptly repaired. However, in some individuals healing is retarded, wounds persevere, and a condition collectively known as persistent epithelial defects (PEDs) can develop; PEDs are characterized by ulcers with delayed (>2 weeks) healing ¹¹ and can recur once healed, particularly when the basement membrane (BM) is lost. They arise from a variety of insults, including trauma; tear-film disorders; neoplasia and its treatment; corneal graft surgery; chemical and thermal burns; or in neurotrophic, ¹² diabetic, ¹³ and herpetic ¹⁴ keratopathies, and can also accompany immunological disorders. ¹⁵  

Current therapies for PEDs are often unavailable for routine clinical use. To reduce the frequency of drug application, need for frequent specialist medical assessment and intervention, and to prevent complications, ^16,17 new methods of delivering therapeutics to facilitate rapid and reliable corneal repair are urgently required. Current therapies are largely based on a premise of patience and prevention of complications, and commonly include application of topical antibiotics and nonpreserved tear-substitutes. However, these agents are readily cleared from the ocular surface and require strict compliance to frequent application by the patient for success. A therapeutic contact lens (CL) may be inserted, essentially to provide a physical cover for the denuded corneal surface, but close medical supervision is necessary due to risks of infection. ¹⁸ Some PEDs are resistant to standard therapy, particularly if associated with comorbidities such as dry eye. ¹⁹ When standard therapy fails, a corneal graft may be required, ¹⁶ but this may be precluded if the cornea has vascularized. 

“Active” therapies for PEDs include human amniotic membrane (HAM) grafts ¹⁷ ; HAM grafts have anti-inflammatory, antiangiogenic, and antiscarring properties. However, this is a foreign biomaterial that can integrate and remain opaque if not adequately remodeled, or spontaneously dislodge from the cornea. Topical autologous serum eye drops have also been trialed with varied success. ^20–22 The rationale for using serum is that it contains beneficial growth factors and mitogens, some of which are found in tears. However, serum also contains numerous proteolytic enzymes and proteins that can potentially interfere with corneal wound healing. ²³ Furthermore, this medication requires daily application that may result in overusage, risk of microbial contamination, and is medically contraindicated in some patients. Investigators have isolated and topically applied selective serum-derived growth and adhesion factors including nerve growth factor and FN. ²⁴ While no side-effects were noted with FN, erosions reoccurred in 30% of patients, ²⁵ and others recorded no benefit. ^26,27 Kabata and colleagues ²⁸ showed that topically instilled VN, an ECM protein with functional properties similar to FN, rapidly resolved chemically and mechanically wounded rabbit corneas. To our knowledge, this factor has not been used to treat corneal epithelial defects in man. 

Certainly, the importance of VN should not be underestimated as delayed wound healing ²⁹ and a defective inflammatory response ³⁰ has been documented in VN knockout mice. Vitronectin is also involved in the early stages of periodontal repair ³¹ and, coupled with growth factor complexes, promotes wound closure in a porcine burns model. ³² The precise mechanism by which VN promotes wound healing has not been elucidated and there is ongoing conjecture about whether it partakes directly with other ECM molecules or it merely provides a temporary scaffold to micromanage the local microenvironment and orchestrate wound closure. ³³  

We recently reported that contact lenses (CLs) submerged in human serum adsorb a spectrum of proteins. Proteins on surfaces of commercial CLs (Lotrafilcon A; CIBA Vision, Fort Worth, TX, USA) were digested with trypsin. In peptides analyzed by mass spectrometry and among the many proteins, VN featured prominently. ³⁴ This finding was not surprising, as epithelial or tear-derived proteins can be deposited on CLs. ³⁵ However, these recent findings prompted us to investigate the activity of VN as a corneal wound healing factor as it has been implicated in cell attachment, ^36,37 spreading, migration, ⁸ and wound healing. ^28–33 Using an in vitro assay system, we discovered that VN specifically promoted corneal epithelial wound healing through its migratory activity mediated in part via integrin-ECM engagement. Moreover, using an ex vivo organ culture model, we demonstrated that VN-loaded CLs were able to repair mechanically debrided human corneas. These exciting results may be used in the future to inform preclinical studies. 

Materials and Methods

Culturing Human Corneal Epithelial Cells

Rims were excised from donor human corneas (Lions NSW Eye Bank, Sydney, Australia) after central buttons had been punched out for transplantation, and digested with an amino peptidase (Dispase II 2.4 U/mL; Roche Diagnostics, Indianapolis, IN, USA) at 37°C for 90 minutes. Dislodged corneolimbal epithelial sheets and single cells were washed twice with sterile PBS (sPBS), plated and cultured in CnT-50 medium (CELLnTEC; Advanced Cell Systems, Bern, Switzerland) supplemented with 10 U/mL penicillin and streptomycin (Invitrogen, Eugene, OR, USA) at 37°C in a humidified 5% CO₂ environment. ³⁴ Primary cells were subcultured weekly and only used between passages 1 and 3. The SV-40 immortalized human corneal epithelial cell (HCEC) line was cultured under the same conditions as primary cells. Notably, CnT-50 is a serum-free undefined media, but importantly, it does not contain detectable VN as assessed by Western blotting (data not shown). This study was carried out in accordance with the tenets of the Declaration of Helsinki. The protocol for using primary cells and tissue was approved by the institutional Human Research Ethics Committee (HREC-11190). 

In Vitro Wound Closure Assays

Recombinant human vitronectin (rhVN; R&D Systems, Inc., Minneapolis, MN, USA) at concentrations up to 10 μg/mL was used to coat 12-well culture plates. After 24 hours, unbound rhVN was washed off with ample sPBS. Primary limbal epithelial cells or HCECs (2 × 10⁵) were seeded into each well and allowed to grow to confluence over 3 to 5 days. Confluent cells were mechanically wounded by passing a yellow pipette tip through the monolayer with one stroke. The rate of repair (% wound recovery) was monitored using an inverted phase-contrast microscope and wound recovery was compared with cells grown on uncoated dishes over 30 hours. Percentage wound recovery was calculated by averaging the rate from four individual wells each with the same concentration of rhVN. The formula used was; percentage (%) of wound recovery = (X_{T 0} − X_Tx )/X_{T 0} × 100%, where X_{T 0} is wound width at time zero and X_Tx is wound width at a specified time point. 

Next, a fenestrated/barrier assay was employed. In these experiments, HCECs were resuspended in CnT-50 media at a density of 5 × 10⁵ cells/mL and dispensed into adjacent chambers within a μ-Dish (ibidi, Munich, Germany). When cells reached confluence, the insert was removed to reveal a gap of 500 μm. Cells were incubated in the presence or absence rhVN (5 μg/mL) and monitored in a live cell imager (Biostation IM-Q; Nikon Instruments Europe BV, Amsterdam, The Netherlands) for cell movement across the clear zone over 24 hours. 

As an alternative to imaging and assessing cell behavior in the Biostation, HCECs were grown to confluence in 96-well plates (ImageLock; Essen Bioscience, Ann Arbor, MI, USA) in a standard cell culture incubator. The 96-pin wound-making tool (WoundMaker; Essen Bioscience) was used to simultaneously create a precise and reproducible wound in each well. Immediately after wounding, media was removed and plates washed twice with sPBS to prevent dislodged cells from settling and reattaching. Cells were next incubated with or without rhVN (5 μg/mL) in the presence or absence of cyclic RGD (10 μg/mL) peptide (Merck, Darmstadt, Germany). In parallel experiments, 25 μg/mL of a neutralizing anti-VN antibody (Clone HV23; CSIRO, North Ryde, Sydney) was preincubated with 5 μg/mL rhVN at 37°C for 1 hour before this solution was added to the wounded monolayer. A potent inhibitor of integrin-mediated cell adhesion, HV23 binds to the somatomedin B region of the VN molecule near the RGD binding site. Plates were placed inside a live cell viewing platform (IncuCyte; Essen Bioscience) and imaged at 2 hours intervals for up to 72 hours. The data was analyzed by integrated metrics and reported as percentage of relative wound density. For this analysis, the algorithm relies on the initial scratch wound to differentiate between cell-occupied and cell-free regions within a selected image. The formula used was; relative wound density (%) (t) = [w (t) − w (0)] / [c (t) − w (0)] * 100%, where w (t) = density of wound region at time t, and c (t) = density of cell region at time, t. 

Cell Proliferation

The ability of rhVN to promote cell proliferation was assessed using a colorimetric 5-bromo-2-deoxyuridine (BrdU) ELISA (Roche Diagnostics, Mannheim, Germany). In brief, a single cell suspension of HCECs or primary corneo-limbal epithelial cells (15 × 10³ cells/200 μL) was seeded on rhVN pre-coated (0-5 μg/mL) wells of 96-well plates and incubated at 37°C in 5% CO₂ for 48 hours. The pyrimidine analogue BrdU (20 μL) was added to each well and incubated for a further 2 hours at 37°C. After removing the culture media, cells were fixed, the DNA was denatured, and an anti-BrdU-POD added (100 μL/well) for 90 minutes at 25°C. Immune complexes were detected by adding 100 μL substrate solution and the reaction product quantified by measuring absorbance at 370 nm on a scanning multiwell spectrophotometer (SpectraMax M3 Plate Reader; Molecular Devices, Sunnyvale, CA, USA). As a confirmatory assay, the rate of cell proliferation was also assessed using the live-cell imaging system (Essen Bioscience) over the same period. The formula used was: % cell proliferation = (% confluence at 48 hours − % confluence at time zero) ÷ % confluence at 48 hours. 

Silencing Vitronectin Receptor Subunits

On the day before transfection, HCECs were seeded at 4 × 10⁴ or 4 × 10⁵ cells/well in either 96- or 6-well plates, respectively. Upon reaching ~60% confluence, cells were transfected with vectors harboring three different siRNA constructs (Eurogentec, Angers, France) at 100 nM final over 6 hours in the presence of a transfection reagent (Lipofectamine RNAiMAX; Invitrogen, Carlsbad, CA, USA). They included a random scrambled sequence (siScramble; 5′-GACGUGGGACUGAAGGGGU-dTdT-3′) and two siRNAs targeting human integrin subunits β3 (5′-CAAGCCUGUGUCACCAUAC-dTdT-3′) and β5 (5′-GCUCGCAGGUCUCAACAUA-dTdT-3′). We used a commercial control siRNA (Block-iT Alex Fluor Red Fluorescent Oligo, Invitrogen) to assess transfection efficiency. Cells grown in 6-well plates were used to gauge integrin knocked down by flow cytometry, while those in 96-well plates were used to monitor wound healing in the presence or absence of rhVN (5 μg/mL). 

Flow Cytometry

Human corneal epithelial cells were dislodged from their plastic substratum by brief exposure to recombinant trypsin (TryPLE; Invitrogen), then incubated for 30 minutes at 4°C with 25 μg/mL of mouse anti-human αvβ3, αvβ5 integrin (R&D Systems) or a matching isotype control immunoglobulin G (IgG; eBioscience, San Diego, CA, USA). Cells were washed in sPBS and a goat anti-mouse FITC-conjugated secondary antibody (Thermo Fisher Scientific, Inc., Rockford, IL, USA) added for 30 minutes at 4°C. Cells were fixed in 2% paraformaldehyde then passed through a flow cytometer (FACS Calibur; Becton Dickinson, San Jose, CA, USA) and the data processed with the flow cytometry analysis software (Cell Quest; Becton Dickinson). ³⁸  

Ex Vivo Corneal Wound Healing

Cadaver human corneas (left and right from three donors) were mechanically debrided with a circular trephine and a scalpel blade under a dissecting microscope, leaving the basement membrane intact. Corneas were rinsed in sPBS then rested epithelium-side up in wells of a 24-well plate (Costar; Sigma Aldrich, Sydney, Australia). One day prior, CLs (CIBA Vision) were soaked in 2 μg/mL rhVN in sPBS overnight at 4°C. ³⁴ Contact lenses in VN or PBS (control) were rested over each cornea and 2 mL CnT-50 media added to each well, allowing specimens to be semisubmerged. Epithelial recovery was allowed to progress over 7 days and 25% of the media (i.e., 500 μL) was exchanged twice during this period. Corneas were removed from each well, extensively washed in sPBS then placed in 10% buffered-formalin, bisected through the center, paraffin-embedded and sectioned for H&E staining. Histology was performed on the tissue and epithelial cells were counted over the intact Bowman's layer (i.e., from limbus to limbus) on both halves of the cornea from each specimen. 

Statistical Analysis

Statistical analysis was performed using graphing and statistics software (Prism, version 6.01; GraphPad Software, Inc., San Diego, CA, USA). A two-tailed unpaired t-test was applied to ascertain significance, which was set at P < 0.05. 

Results

In Vitro Epithelial Wound Closure

Initial experiments to test the effects of VN as a wound-healing factor were performed by manually scratching a monolayer of HCECs that were grown on VN-coated tissue culture plastic. Vitronectin dose dependently increased wound recovery, reaching statistically significant levels with 5 μg/mL when compared with uncoated controls (78.1% ± 7.3% vs. 49.0% ± 3.3%, P = 0.0112) at 24 hours, and with 5 μg/mL (94.6% ± 5.4% vs. 56.9% ± 3.7%, P = 0.0012) and 10 μg/mL (94.1% ± 6.0% vs. 56.9% ± 3.7%, P = 0.0018), respectively, at 30 hours (Fig. 1). 

Figure 1

 

Wound closure following a manual scratch. Human corneal epithelial cells were seeded on VN-coated, 12-well tissue culture plates and confluent monolayers wounded by passing a sterile yellow micropipette tip through the culture in a single motion. Cells cultured under control conditions (A, B) or with 10 μg/mL rhVN (C, D) were imaged by phase contrast microscopy (A–D). Percentage wound-healing across a range of VN concentrations is displayed in the line graph (E). Measurements at each time point represent mean percentage recovery ± standard deviations from four independent wells. *Indicated significance (P < 0.05) between 0 and 5 μg/mL rhVN after 24 hours, and between 0 and 5 μg/mL and 10 μg/mL rhVN after 30 hours. These data are representative of three independent experiments.

Although the manual scratch assay is widely accepted and utilized as an in vitro model, we regard this as a primitive assay for recording the activity of a potential wound-healing factor, mainly because the wound generated varies in size, and secondly because it is difficult to monitor the same area of cells long-term. Given these disadvantages, a time-lapse, live-cell imaging system (Nikon Instruments Europe BV) was used to record wound-healing in real-time. This is regarded a fenestrated-type assay as a gap between two adjacent cell populations is created by lifting a culture insert (Fig. 2). After 24 hours of monitoring, relatively little cell migration was noted in control cultures (Figs. 2A–D). However, after adding rhVN (5 μg/mL), migration was accentuated between 7 and 24 hours post treatment as noted by the number of cells that began to move into the gap (Figs. 2E–H). 

Figure 2

 

Cell migration in a fenestrated assay. Human corneal epithelial cells were seeded in adjacent chambers separated by a 500-μm insert. When cells reached confluence, the insert was removed resulting in a cell-free zone (area between the hatched white lines). Cells remained untreated (A–D) or incubated with a solution of rhVN (5 μg/mL; [E–H]). Phase contrast still images were acquired from a video recording at 0, 7, 12, and 24 hours (A–H). Original magnification: ×20. This data is representative of three independent experiments.

Based on the data displayed in Figures 1 and 2, we decided to monitor wound-closure over 24 hours with 5 μg/mL rhVN. Moreover, because the images from the live-cell imaging system (Nikon Instruments Europe BV) screens only one vessel at any given time, a more robust method was sought—that is, one where many replicates could be screened in real-time. Therefore, all subsequent experiments related to in vitro wound healing were conducted in a 96-well plate setup using a wound-maker tool (Essen Bioscience) to etch a standard and reproducible wound in a single motion in multiple wells. Using these parameters and hardware, the effect of VN on accelerating the wound-healing response was confirmed on primary human corneolimbal epithelial cells, where it was found that it significantly repaired the defect over 24 hours (Supplementary Fig. S1). Notably, however, the wound closure rate in primary cells was faster than in HCECs; nonetheless, a similar profile was observed. From this point, it was decided that all subsequent assays be performed on HCECs as we anticipated more consistent and reliable results by eliminating donor-to-donor variations that would arise from using primary cells. 

Using this assay, rhVN repaired the injured HCEC monolayer. For example, at 16 hours post wounding, 5 μg/mL rhVN significantly increased wound closure compared with cells without VN (57.8% ± 5.3% vs. 44.5% ± 5.7%; P < 0.001). Furthermore, when cRGD was added after injury, wound healing was significantly suppressed (57.8% ± 5.3% vs. 43.5% ± 7.6%; P < 0.001; Figs. 3A, 3B). Because the RGD peptide is not specific for VN—that is, it interacts with other proteins such as fibronectin and laminin—an additional experiment was conducted with a neutralizing antibody to VN. When the HV23 antibody was added in conjunction with rhVN, wound closure was significantly attenuated when compared with cells treated with an appropriate isotype control antibody (31.3% ± 5.1% vs. 54.6% ± 12.5%; P < 0.0001; Figs. 3C, 3D). 

Figure 3

 

Inhibition of wound closure with VN antagonists. We seeded HCECs in 96-well plates, scratched the monolayer using the wound-maker tool (Essen Bioscience), and monitored on the live-cell imager (Essen Bioscience). Replicate wells (n = 6–12) were treated with basal media (control) or media containing the synthetic cRGD peptide (A, B). In other experiments, cells were incubated with a VN-neutralizing antibody or an isotype control IgG antibody (C, D). The line graphs displayed in panels (A, C) were generated by the live-cell system software (Essen Bioscience). The bar graphs in panels (B, D) represent mean percentage relative wound density ± standard deviation and were derived from a 16-hour time point. ***P < 0.001. ****P < 0.0001. Data are representative of at least three independent experiments.

Effect of Vitronectin on Cell Proliferation

Next, we assessed whether wound closure was related to the ability of VN to induce cell proliferation. For this study, two assay systems were employed. In the first instance, we demonstrated that incorporation of BrdU was unaltered irrespective of the VN concentration used (Fig. 4A). A second confirmatory assay measured cell proliferation on the live-cell viewing platform (Essen Bioscience) where similar results were obtained (Fig. 4B). These data suggest that VN does not modulate proliferation, but is principally involved in cell migration. Same results were generated with primary cells (data not shown). 

Figure 4

 

Effect of VN on cell proliferation. Human corneal epithelial cells were seeded in 96-well plates, allowed to reach 60% confluence, then treated with varying concentrations of rhVN. Cell proliferation was measured at 24 hours using a BrdU ELISA (A) or after monitoring live cells in the live-cell imager (Essen Bioscience [B]). Bars represent mean data from 6 to 12 replicate wells ± standard deviation.

Effect of Vitronectin Receptor Silencing on Wound Closure

In order to further prove that VN promoted wound-healing by mediating corneal epithelial cell migration, HCECs were transfected with siRNAs targeting the VN receptor; that is, β3 and β5 integrin subunits. In general, the average transfection efficiency was estimated at 83% with a fluorescent oligo-reporter (Fig. 5A). Transfecting HCECs with siβ5 dramatically reduced cell surface-associated αvβ5 on HCECs to negligible levels (Fig. 5C, green histogram overlapping blue histogram). Notably, transfection with the siβ3 and siScr (Figs. 5D, 5E, respectively) did not affect αvβ5 on HCECs. The mean fluorescence shift (%) relative to the IgG₁ isotype control was 11.1% ± 8.9%, 85.3% ± 31.8%, and 89.6% ± 13.1% for silencing with siβ5, siβ3, and siScr RNAs (Figs. 5C–E). The difference in the mean fluorescence shift (%) between siβ5 and siScr was 78.5%, which indicated a significant knockdown of β5 subunit in these cells (P < 0.0001; Fig. 5F). Notably, HCECs (data not shown) or primary human corneal epithelial cells do not express αvβ3, ³⁸ hence cell movement through this receptor was not anticipated. Three days post transfection, HCECs were wounded in the presence (5 μg/mL) or absence of VN. The results demonstrate that wound closure in the presence of VN was significantly inhibited by 17% (P = 0.04) in siβ5-transfected cells (Fig. 5G). 

Figure 5

 

Effect of β5-Integrin silencing. Before HCECs were transfected with integrin-silencing constructs, transfection efficiency was established with a red fluorescent oligo reporter (A). The corresponding phase-contrast image is also displayed ([B], ×20 original magnification). Human corneal epithelial cells transfected with silencing constructs for β5 (C), β3 (D), or the scrambled (Scr) control (E) were monitored for αvβ5 integrin expression by flow cytometry ([C–E], green histograms). Purple histograms represent background staining with an isotope control antibody. Results were quantified and represented as a bar graph using mean fluorescent intensity (F). The same cells were wounded then treated with 5 μg/mL (black bars) or without (gray bars) rhVN for 16 hours. Bars in (G) represent mean relative wound density ± standard deviation from 6 to 12 replicates. *P < 0.05. ****P < 0.0001. These data are representative of two to four independent experiments.

Ex Vivo Corneal Epithelial Wound Healing

In a previous study, we showed that CLs (CIBA Vision) absorb and release VN. ³⁴ Therefore, we used this as a delivery device for VN in an ex vivo organ culture model of corneal wound healing. Mechanically wounded human cadaver corneas were covered with either a VN-coated or a sPBS-soaked CL. In corneas covered with a VN-loaded lens, a continuous corneal epithelium covered the defect within 7 days (Figs. 6A, 6C). In contrast, discontinuous islands, and indeed less epithelial cells were counted in corneas covered with control lenses (Figs. 6B, 6C). 

Figure 6

 

Wound-healing in donor human corneas. Cadaveric human corneas were mechanically wounded and VN (A) or uncoated (B) CLs placed over the defect. After 7 days, corneas were removed from organ culture, fixed in formalin, sectioned and stained with H&E (A, B). The number of cells (mean ± SD) that stretched from limbus to limbus over Bowman's layer was counted in three pairs of specimens (C). The histology of an unwounded donor corneas is also displayed ([D], H&E staining). Original magnification ([A, B, D]; ×1000 oil immersion). **P < 0.01.

Discussion

Our recent report on the adsorptive (loading) and desorptive (slow-release) capacity of CLs toward serum-derived and rhVN, ³⁴ coupled with the knowledge that VN is an attachment, ^36,37 spreading, migration-promoting protein ⁸ and wound-healing factor for the cornea, ²⁸ skin, ³² and other tissues ^29–31 prompted us to perform additional mechanistic investigations. We showed that VN heightens cell migration and wound closure in a corneal epithelial cell line (Figs. 1 1552–3) and in primary human corneal epithelial cells (Supplementary Fig. S1) after an injury. We also determined that the activity of VN was not directed toward cell proliferation (Fig. 4) and confirmed the specificity of our results by observing partial blockade of wound closure with an RGD peptide (Fig. 3), a neutralizing antibody (Fig. 3), and transfection with silencing RNAs to β5 (Fig. 5). Finally, in an organ culture model we demonstrated significant re-epithelialization of injured human corneas after delivering VN on a therapeutic CL (Fig. 6). 

Vitronectin has been localized to the corneal epithelial BM, implying this may be an endogenous source. ³⁹ However, Xiao's data do not concur with ours ³⁴ and those of others ⁴⁰ who localized this protein to the peripheral corneal BM under homeostatic conditions, more specifically within the stem cell harboring limbal transition zone. Because the central cornea is devoid of a vasculature, VN is most likely deposited in this region from the adjacent limbal stromal blood vessels. Its levels in plasma and serum measure approximately 0.3 mg/mL ⁴¹ and this is a likely source after an injury to the central cornea. Another reservoir of VN that may contribute to ocular protection and corneal healing is the tear film. ⁴² The activity of VN is predominantly motogenic with little or no metabolic ⁴³ or mitotic-modulating activity (Fig. 4). Interestingly, its migratory activity for corneal epithelial cells is amplified by insulin-like growth factor 1 (IGF-1) ^43,44 and IGF-1-derived peptides, ⁴⁵ thought to be mediated through its ability to sequester these and others factors including epidermal growth factor, FGF, VEGF, and TGF-beta 1 and 2. ⁴⁶ However, the mitogenic effects of VN-bound growth factors may be deleterious for patients with PEDs, particular during the initial phase of corneal epithelial healing which predominantly requires cell motility. ^2,3 Certainly Lee and associates ⁴⁴ addressed this potential adverse effect by truncating IGF-1 to eliminate its mitogenic activity. 

There are known regional differences within the cornea regarding ECM composition. ⁴⁰ Of particular interest are the proteins that comprise the BM of the ocular surface epithelium as these factors are implicated in supporting basal cells, and in modifying their cellular function including differentiation. Therefore, injury not only removes corneal epithelial cells, but can also fragment their BM substratum. One of the first processes initiated after an injury is epithelial cell sliding ³ and the synthesis of new matrix proteins including FN, which becomes a temporary scaffold for epithelial cells to adhere to and migrate on. ⁴⁷ Wounding not only results in deposition of a FN-containing matrix, but also upregulation of its integrin receptors in the corneal epithelium. ⁷ This was further noted through the use of a corneal epithelial cell line cultured over different ECM proteins, which resulted in FN inducing the greatest cell attachment, spreading and motility; processes that were elaborated through integrin receptor expression. ² Interestingly, FN like VN copurifies with mitogens such as EGF and this could also potentiate its activity as a wound-healing factor. ⁴⁸ To our knowledge, a comprehensive and comparative assessment of the wound healing capacity for FN and VN in the cornea has not been performed. However, both are required for wound contraction, a process associated with corneal tissue repair, ⁴⁹ and both seem to be interchangeable during this process. ⁵⁰ In addition, partially degraded species of both proteins are found in tears from patients with chronic corneal ⁵¹ and venous stasis ⁵² ulcers, suggesting a functional defect in their wound healing activity under pathological conditions. 

From our previous investigation ³⁸ and from the experiments conducted in the current study (Fig. 5, and data not shown), corneal epithelia lack expression of the VN receptor αvβ3-integrin ⁵³ but express the alternative VN receptor αvβ5. ⁵⁴ We showed that silencing β5 abrogated the expression of this integrin heterodimer on HCECs (Fig. 5C) and as a consequence, wound healing was significantly retarded in the same cells, although the effect was partial (Fig. 5G). This implies that migration was not exclusively mediated by a VN/integrin axis and that other protein/adhesion molecule combinations are also involved. Certainly this is reflected by the incomplete but nonetheless significant suppression of wound closure in our culture model, following the addition of the RGD peptide and a neutralizing antibody to VN (Fig. 3). However, the precise signaling mechanism(s) by which VN enhances epithelial cell motility has not been defined in the cornea. Studies using primary coronary artery smooth muscle cells have shown that VN dose dependently increases cell migration with a concomitant enhancement of tyrosine phosphorylation of focal adhesion kinase and matrix metalloproteinase-2 production and that this is negated by EMD121974, a selective pharmacological inhibitor of αv-integrin. ⁵⁵ Likewise, VN induced matrix metalloproteinase-9 in human umbilical vein endothelial cells and this was found to be mediated through protein kinase B and mitogen activated protein kinase signaling cascades. ⁵⁶  

The current investigation focused on VN and not its functionally-related protein FN. Our results are congruent with others that have demonstrated a heightened corneal wound healing response to VN in rabbits. ²⁸ The vast majority of corneal wound healing factors are applied topically in eye drop formulations. This delivery route can be problematic as the time in contact with the ocular surface can vary due to blinking and dilution with the tear film. Additionally, the risk of microbial contamination and patient compliance are also major disadvantages of this delivery route. Our recent discovery of the loading and slow-release capacity of therapeutic CLs for VN renders this an attractive therapeutic strategy that is simple, easy to standardize through the application of a recombinant protein species, and the ability to provide the treatment as a sterile, single-use, short-term application for patients with PEDs and related corneal disorders. One limitation of the current investigation was that we did not corroborate our in vitro and ex vivo proof-of-concept findings in a preclinical animal model. In a previous study, ³⁴ we monitored the fate of human limbal epithelial cells grown on a CL in the presence or absence of fetal bovine serum and showed that cells transferred from this device to a de-epithelialized organ cultured cornea. In addition, we demonstrated that VN of bovine origin transferred from the CL and bound to the host tissue as determined with a species-specific antibody against bovine VN. These data provide solid evidence that VN diffuses from the CL polymer, is adsorbed onto recipient tissue, and is likely to play a role in cell attachment and migration to promote ocular surface repair. 

There is precedence for treating external eye disease with CL-loaded factors, including hyaluronic acid for dry eye, ⁵⁷ antibiotics for bacterial keratitis, ⁵⁸ and anti-inflammatory, ⁵⁹ and immunosuppressive ⁶⁰ agents. Recently, epidermal growth factor–coated CLs were applied to wounded rabbit corneas, but the regenerating epithelium was only loosely connected to its BM, suggesting that proteins important for cell adhesion may be required for effective healing. ⁶¹ A second limitation of the current study is that a variety of commercial CL polymers were not investigated and it is possible that there are materials with better VN loading/releasing capacity than those described in this report. Finally, we highlight the additional importance of VN not only as a promoter of wound-healing, but also as a stem/progenitor cells support factor for several organs, ^62–64 including the cornea. ³⁸  

Acknowledgments

The authors thank Raj Devasahayam from the Lions NSW Eye Bank (Sydney, Australia) for providing fresh donor human corneas and Penny Bean (CSIRO, North Ryde, Sydney, Australia) for providing the anti-HV23 antibody clone. 

Supported by the Brian Kirby Foundation and the Ophthalmic Research Institute of Australia. 

Disclosure: S. Chow, None; N. Di Girolamo, None