Pigment Epithelial-Derived Factor Peptide Facilitates the Regeneration of a Functional Limbus in Rabbit Partial Limbal Deficiency

Abstract

Purpose.: To investigate the potential of a pigment epithelial-derived factor (PEDF) peptide 44-mer to promote limbal regeneration in a rabbit partial limbal deficiency model.

Methods.: Limbal excision (180°) was created surgically, and topical application of 44-mer-containing ointment once a day for 2 weeks was started immediately after injury. Limbal barrier function was inspected at 2 and 6 months after treatment. Corneal neovascularization was observed under slit-lamp microscope. The presence of goblet cells on the corneal surface was examined using impression cytology. The resulting repair tissue was assessed by immunohistochemical staining with antibodies for putative limbal stem cell (LSC) markers ΔNp63α and ABCG2. Cells harvested from the regenerated tissue were analyzed for colony-forming capacity and expression of LSC markers by immunostaining assay and quantitative real-time PCR (qPCR).

Results.: Eyes treated with the 44-mer blocked vascularization and goblet cell migration onto the corneal surface. By means of immunohistochemical staining and cell isolation in the repair tissue, we showed that LSCs were widely distributed at the regenerated tissue after 44-mer treatment. The repaired limbus contributed robustly to corneal wound healing as effectively as undamaged limbus.

Conclusions.: We demonstrated that 44-mer regenerates a functional limbus-like structure on limbal excision wounds. Our finding suggests that the PEDF peptide derivative may be an innovative strategy for tissue engineering and repair therapy in partial LSC deficiency diseases.

The corneal epithelium is the outermost layer of the ocular surface, with active cell turnover and self-regenerating properties, that shields the cornea from exposure to the external environment. It is a unique tissue that not only maintains the corneal transparency, but also prevents the ingrowth of blood vessels from the adjacent conjunctiva. The homeostasis of the corneal epithelium is supported by stem cells that reside in the basal epithelium of the limbus, a transitional zone located between the cornea and conjunctiva.¹ Typically, limbal stem cell deficiency (LSCD) manifests signs of ocular surface insufficiency, characterized by recurrent or persistent corneal epithelial defects, conjunctivalization of the corneal surface, chronic stromal inflammation, scarring, and ultimately loss of corneal clarity.² 

Currently, limbal transplantation is the major option for treatment of LSCD. Limbal tissue autografts and allografts have been used extensively for clinical therapy with wide success.^3–5 Recently, cell-based therapy (transplantation of a limbal equivalent followed by ex vivo expansion of limbal stem cells [LSCs]) has showed encouraging results as an alternative to overcome limitations of limbal tissue transplantation. Not only is it a safer choice for living related or contralateral uninjured donor eyes by allowing smaller-size donor tissue biopsy; it also reduces the rejection rate and the necessity of systemic immunosuppression in allogeneic transplantation. Cell-based therapy has the potential to be an unlimited resource for LSC transplantation. However, the long-term survival of transplanted LSCs has been a major concern with this therapeutic modality.^2–8 In our study, we pursued a therapeutic modality for LSCD by in vivo expansion of the residual viable LSCs.^9,10 

Pigment epithelium-derived factor (PEDF) is a 50-kDa secreted glycoprotein with several biological effects on a variety of cell types. Pigment epithelium-derived factor targets vascular endothelial cells and is a potent inhibitor of angiogenesis. It is also a neurotrophic factor that protects neurons in the brain, eye, and spinal cord against a wide range of neurodegenerative insults.^11–14 The domain encompassing amino acids Val78-Thr121 of human PEDF, also termed the 44-mer, determines its neurotrophic activity and binds receptors on the surfaces of various types of neurons.¹⁴ 

The initial study linking PEDF to stem cells reported that murine PEDF is a niche factor located at the subventricular zone and is responsible for promoting neuronal stem cell renewal.^15,16 Pigment epithelial-derived factor also was reported to be a mitogen for other types of adult neural progenitor cells, such as hippocampal progenitor cells¹⁷ and retinal progenitor cells derived from the ciliary body.¹⁸ Recently, the effect of PEDF on stem cells was found to act beyond the neural epithelium. We reported that the PEDF 44-mer promotes LSC proliferation in vitro and in vivo.¹⁹ These findings suggest the potential of the 44-mer for the treatment of LSCD by expanding LSCs. Structural lesions confined to the limbus do not heal spontaneously, and this represents a serious clinical challenge for reconstruction of the ocular surface. We used a limbal excision animal model to reproduce the structural lesion and found that the 44-mer is able to promote limbal regeneration, accelerate corneal epithelial wound healing, and contribute to the restoration of a functional ocular surface. 

Materials and Methods

Chemicals and Antibodies

HEPES-buffered Dulbecco's modified Eagle's medium (DMEM), Ham's/F-12 medium, trypsin-EDTA, fetal bovine serum (FBS), antibiotic–antimicotic solutions, and trypsin were purchased from Invitrogen (Carlsbad, CA, USA). Hydrocortisone, dimethyl sulfoxide (DMSO), insulin-transferrin-sodium selenite (ITSE) media supplement, mitomycin C (MMC), bovine serum albumin (BSA), 5-bromo-2′-deoxyuridine (BrdU), Triton X-100, Hoechst 33258 dye, n-heptanol, and formalin were all from Sigma-Aldrich Corp. (St. Louis, MO, USA). Dispase II and epidermal growth factor (EGF) were obtained from Roche (Indianapolis, IN, USA). ΔNp63α polyclonal antibody and all the fluorescent dye-conjugated secondary antibodies were purchased from BioLegend (San Diego, CA, USA). Mouse anti-keratin-3 (clone AE5; CBL218) was purchased from Millipore Corporation (Bedford, MA, USA). Rat monoclonal anti-ABCG2 antibody (ab24115) was from Abcam (Cambridge, MA, USA). Vimentin polyclonal antibody (GTX100619) was from GeneTex (San Antonio, TX, USA). Pigment epithelial-derived factor peptide 44-mer was synthesized and modified with acetylation of the NH₂ termini and amidation of the COOH termini for stability and characterized by mass spectrometry (>95% purity) to order at GenScript (Piscataway, NJ, USA). 

Animals

Twenty New Zealand albino rabbits (3.0–3.5 kg, 6 months of age) were used in this study. All animals were handled in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and institutional guidelines for animal investigation. All procedures were performed with animals under general anesthesia induced by intramuscular injection of ketamine hydrochloride (35 mg/kg) and xylazine hydrochloride (5 mg/kg) and topical anesthesia with 0.5% proparacaine hydrochloride (Alcaine; Alcon, Fort Worth, TX, USA). The right eye of each animal was surgically injured and the left eye was left untouched. Partial limbal injury model was carried out surgically in all animals in the same manner by a single investigator. Animals were sorted randomly into the 44-mer treatment group (n = 10) or sham-operated vehicle control group (n = 8). Two rabbits were used as normal controls. 

Partial Limbal Injury Animal Model

Half-circumference (180°) limbal tissue excision was performed by lamellar dissection of the limbal zone, 2 mm into the cornea and 3 mm into the conjunctiva in the experimental eye. In order to ensure that the limbus was completely denuded of its epithelium, a layer of superficial corneoscleral stroma below was also removed (100–120 μm deep, confirmed by histology).²⁰ Preserved human amniotic membrane (AM) as a substrate was grafted over the limbal surgical wound with the epithelial side up and sutured onto the corneal, episcleral, and free conjunctival edges with interrupted 10-0 nylon sutures. The AM melted within 2 weeks, leaving a wound surface with minimal inflammation and scarring. Topical and subconjunctival dexamethasone were applied to the experimental eye on a daily and weekly basis, respectively, for inflammation control during the first 2 weeks post surgery. The limbal and corneal vascularization occurred within 2 weeks, was exacerbated at 3 to 5 weeks, and became stable at 2 months after surgery.²¹ 

Treatments

The 44-mer was reconstituted in DMSO as stock (5 mM) and mixed with TOBREX eye ointment (Alcon; containing 0.3% Tobramycin and 0.5% Chlorobutanol) to a 100 μM concentration.¹⁹ After the limbal wounding, rabbits sorted into the treatment group received 20 μL 44-mer-containing ointment once a day for 2 weeks. Corresponding control animals were treated with vehicle (DMSO-mixed) ointment after surgery. Clinical, histologic, and cell study characteristics of the healing limbal tissue from 44-mer-treated eyes and vehicle controls were compared to those of untreated eyes from an unused cohort of rabbits (normal controls). 

Capacity of Corneal Epithelial Wound Healing

Two months after 44-mer treatment, corneal epithelial debridement was performed to evaluate the capacity for wound healing.²² The central corneal surface, 10 mm in diameter, was treated with n-heptanol for 30 seconds, followed by irrigation with sterile saline and mechanical scraping of the epithelium.^23,24 Topical gentamicin ointment was given once a day until the wound had healed. The corneal epithelial defect was determined by 1% topical fluorescein staining (Fluor-I-Strip; Ayerst Laboratories, Philadelphia, PA, USA) and photographed with a surgical microscope (OPMI Pico I; Carl Zeiss Meditec, Jena, Germany). The area of the epithelial defect was quantified using a computer-assisted image analyzer (Adobe Photoshop CS3 10.0; San Jose, NM, USA) and calculated as the percentage of residual epithelial defect at each time point divided by initial wound area. 

Capacity of Limbal Barrier Function

The limbal barrier function is the capacity of a healthy limbus to prevent conjunctivalization. Repeated corneal epithelial injuries are an established approach to challenging limbal barrier function.²² Two months after limbal wounding and 44-mer treatment, repeated corneal debridement was carried out at 2-week intervals as described previously. 

Conjunctivalization was determined by identification of blood vessel ingrowths and goblet cells over the healing limbus and adjacent cornea. Digital photographs were taken with a Zeiss (OPMI Pico I; Carl Zeiss Meditec) surgical microscope, weekly in the first month after wounding and then monthly during the follow-up period of 6 months. Photographs were scored for limbal NV (neovascularization) by two ophthalmologists who were blinded to the treatment groups. The extent of limbal NV was scored from 1 to 12, according to the number of clock hours of limbus affected by NV (e.g., a score of 1 was 1 clock hour, and a score of 2 was 2 clock hours), and its ratio to the entire limbal circumference was determined as the percentage of limbal NV (PLNV). Centricity of NV was defined as the length (in millimeters) of the new vessel extending from the limbus toward the visual axis.²¹ 

Impression cytology (IC) of the healing limbus and peripheral cornea was collected and stained.²⁵ The mean goblet cell density (GCD) of each specimen was determined by averaging the total number of goblet cells in three consecutive visual fields of high magnification (×200). The mean GCD of each enrolled eye was calculated by averaging the GCD measured at nasal, superior, and temporal limbal zones and peripheral cornea. The average value of the specimen on the limbus was considered limbal GCD.²⁶ All limbal IC specimens were obtained and evaluated in a masked fashion by the same technician. Histologic sections of the corneoscleral transition zone were stained with hematoxylin and eosin (H&E) and periodic acid–Schiff (PAS) for microstructural analysis. 

Isolation and Culture of LSCs

Limbal stem cells were isolated from normal limbal epithelium and healing tissue 6 months after limbal excision. Isolation and cell-suspension culture of LSCs were performed using a method described previously.²⁷ Near-confluent P0 (Primary culture) cells were harvested with 0.25% trypsin, and then 1 × 10⁵ subcultured cells were cultured in DMEM/F-12 basal medium (10 mM HEPES, 5 ng/mL human EGF, 1% ITSE, 1% antibiotic–antimicotic solutions, 0.5% DMSO, and 0.5 μg/mL hydrocortisone) for 5 days (P1 cells). The P1 cells were used to determine colony-forming efficiency (CFE) and LSC marker expression. 

Colony-Forming Efficiency

Limbal cells (1 × 10³) were seeded in a 3.8-mm² dish and cocultured with MMC-treated NIH-3T3 feeder cells located within the transwell. The medium was changed every 2 to 3 days. Colonies were fixed by 4% paraformaldehyde (room temperature [RT] for 1 hour) 14 days later for immunostaining or crystal violet staining and photographed. The CFE (%) was calculated using the following formula: number of colonies formed/number of cells plated × 100. 

Immunoblot Analysis

Cell lysis, SDS-PAGE, and antibodies used for immunoblotting were as described previously.²⁸ The band intensity in immunoblots was assessed with a model GS-700 imaging densitometer (Bio-Rad Laboratories, Hercules, CA, USA) and analyzed using Labworks 4.0 software (UVP, Inc., Cambridge, UK). 

Immunofluorescence

Deparaffinized tissue sections or 4% paraformaldehyde-fixed LSCs were blocked with 10% goat serum and 5% BSA in PBS containing 0.1% Tween 20 for 1 hour. Staining was performed using primary antibodies against ΔNp63α (1:150 dilution; BioLegend), keratin-3 (1:250 dilution; Millipore Corporation), ABCG2 (1:100 dilution; Abcam), and vimentin (1:100 dilution; GeneTex) at 37°C for 2 hours, followed by incubation with the appropriate rhodamine- or FITC-conjugated donkey IgG (1:500 dilution) for 1 hour at RT. Images were captured using a Zeiss epifluorescence microscope with a charge-coupled device camera and photographs taken using the Zeiss Axiovision version 3.1 software (Carl Zeiss MicroImaging GmbH, Jena, Germany). 

RNA Extraction and Quantitative Real-Time PCR

Experiments were performed as described previously.²⁸ The PCR primers were rabbit ABCG2 sense, 5′-gtgagccctacaacaaccct-3′, anti-sense, 5′-cttcagtcctgttggcttca-3′ (accession number: XM_002716965; PCR product: 111 bp); rabbit integrin, alpha 6 sense, 5′- gaggccatgagaagtttggt-3′, anti-sense, 5′-cgaacaatccctttccagtt-3′ (XM_002712183; 115 bp); and rabbit GAPDH sense, 5′-tctggcaaagtggatgttgt-3′, anti-sense, 5′-gtgggtggaatcatactgga-3′ (NM_001082253; 87 bp). The cycle threshold (Ct) values of the PCR product and a GAPDH (Glyceraldehyde 3-phosphate dehydrogenase) control mRNA were used to calculate relative quantities of mRNA. 

Statistical Analysis

Statistical analyses were performed using GraphPad Software (San Diego, CA, USA). The results are expressed as mean ± SEM of (n) independent experiments. Comparisons were analyzed using the Student's t-test, Wilcoxon nonparametric paired test, and one-way analysis of variance (ANOVA). Differences were considered significant when P < 0.05. 

Results

The 44-mer Promotes the Regeneration of a Stable Limbal Structure

The eyes of rabbits given the vehicle control ointment exhibited prominent inflammation, with vascularization of the limbal wound and adjacent corneal surface that peaked at 3 weeks and became stationary after 2 months. In the 44-mer-treated group (n = 10), the limbal wound healed without vascularization in seven eyes and minimal focal vascularization of less than one score extension in three eyes, suggesting the recovery of limbal anatomy (Fig. 1). Impression cytology collected at 2 and 6 months post treatment confirmed conjunctivalization of the limbal wound in the vehicle-treated eyes characterized by the presence of conjunctival goblet cells. This feature was not observed in the 44-mer-treated eyes. The mean GCD measured at 2- and 6-month follow-ups were, respectively, 50.46 ± 11.35 and 40.83 ± 6.92 cells/mm² in vehicle controls, which were clearly higher than in the 44-mer-treated eyes with 1.97 ± 1.17 and 1.13 ± 0.75 cells/mm² (P < 0.001) (Fig. 2). Histologic studies at 6 months post wounding revealed that 44-mer treatment gave rise to a cohesive stratified epithelium consisting of a basal layer of cuboidal cells and two to three suprabasal layers of rounded cells. Expression of ΔNp63α and ABCG2 indicated that LSCs were preserved. In contrast, vehicle-treated eyes contained only stratified columnar cells, resembling conjunctival epithelium, and the expression of LSC markers was not detected (Fig. 3). Collectively, 44-mer treatment apparently re-established a limbus-like structure on the limbal excision wound. 

Figure 1

 

External photographs of the limbal wound healing pattern in the 44-mer and vehicle control treatment groups. The photos were taken at 1, 3, 5, and 7 weeks after limbal injury. The eyes were treated with topical ointment containing the 44-mer or vehicle once a day for 2 weeks immediately after limbal injury. All vehicle control eyes (n = 8) showed prominent inflammation with fibrovascular pannus formation over the excision wound peaking at the first 3 weeks and gradually became stationary after 2 months. As the limbal wound healed, the fibrovascular pannus turned into mature fine blood vessels that grew across the limbal wound and invaded the peripheral corneal surface. In 7 of 10 eyes treated with 44-mer, the limbal wound healed smoothly without vascularization. *Three 44-mer-treated eyes showed minimal vascularization.

Figure 2

 

Representative photographs of healed limbal wounds (delimited by dotted lines) and corresponding impression cytology (goblet cell stained purple) 2 and 6 months after treatment with the 44-mer or vehicle mixed ointment (A). (B) Schematic drawing of the limbal area where nitrocellulose filter papers were applied for impression cytology evaluation. (C) Mean goblet cell density of the healing limbus assessed by impression cytology evidencing the effect of 44-mer treatment on the limbal wound healing. Data represent four independent experiments and four rabbits per group. *P < 0.001 versus vehicle-treated group.

Figure 3

 

The structure of the regenerated limbus examined at 6 months. Upper: Hematoxylin and eosin (H&E) staining (original magnification ×400). Lower: Immunofluorescence analysis of LSC markers ΔNp63α and ABCG2. Stromal tissue and nuclei were visualized with vimentin and Hoechst 33258 staining, respectively. Representative results from four separate experiments are shown. Normal controls were untreated eyes from a nonoperative group of rabbits.

Colony-Forming Cells Can Be Isolated From Regenerated Limbal Epithelium Induced by the 44-mer

The potential to divide continuously enables LSCs isolated from the limbus to form colonies in culture and to express stem cell marker proteins. We sought to determine whether such cells can be isolated from limbal tissue regenerated by the 44-mer. Colonies formed by cells harvested from the 44-mer-regenerated limbus have the same density and size as those from a normal, control limbus (Fig. 4A). Microscopically, these colonies contain small, polygon-shaped cells, and immunostaining reveals that the majority of these cells are ΔNp63α-positive, LSC-like cells. Colonies from the regenerated limbus contain a few more keratin 3 (K3; a differentiated corneal marker)-positive cells than normal controls (Fig. 4B). Immunoblot analysis revealed similar levels of ΔNp63α in cells derived from the normal control and the regenerated limbus, but cells derived from the regenerated limbus contained a higher level of K3 (Fig. 4C). The expression of LSC markers, including ABCG2 and integrin α6, was analyzed by quantitative real-time PCR (qPCR). As shown in Figure 4D, although not statistically significant, the expression of LSC markers in 44-mer-regenerated tissue was lower than in the normal limbus but 10-fold higher than in vehicle-treated wound tissue (P < 0.05). These findings indicate that the cells in the 44-mer-regenerated limbus have the same proliferation potential in culture as cells from an intact, native limbus. 

Figure 4

 

The regenerated limbus contains colony-forming cells. (A) Cells were harvested from 44-mer-regenerated and normal limbal epithelium for colony formation assays, as described in Methods. Four independent experiments were performed. (B) ΔNp63α and K3 immunostaining assay and representative phase-contrast micrographs of the colonies described above (original magnification, ×400). Representative images from three independent experiments. (C) Intensities of ΔNp63α and K3 in the immunoblots were determined by densitometry and normalized to β-actin. Representative immunoblots from three independent experiments are shown. (D) Quantitative real-time PCR analysis of mRNA isolated from limbal tissue after the treatments indicated, as described in Methods. Data represent three independent experiments. *P < 0.05 versus vehicle-treated cells.

The Regenerated Limbus Confers Barrier Function and Withstands Repeated Corneal Challenges

Removal of the 44-mer-regenerated limbal epithelium (n = 2) by soaking with n-heptanol and mechanical scraping resulted in prompt loss of barrier function as evidenced from the limbal and corneal conjunctivalization (Figs. 5A, 5B). Histologic sections stained with PAS demonstrated invasion of goblet cells (in purple) into the healing limbus and peripheral cornea. The results confirmed that the preserved barrier function is derived from the regenerated limbus. The stability of the 44-mer-regenerated limbal barrier function was demonstrated by its capacity to withstand repeated removal of the corneal epithelium without conjunctivalization (at intervals of 1 month). Nearly total conjunctivalization of the half-circumference limbal wound occurred in the vehicle control group (pre challenge PLNV of 40.00 ± 0.11%). The limbal NV was exacerbated by each repetition of corneal epithelial wounding, more in centricity (P < 0.001 after each challenge) than in extension (P < 0.05 only after the third challenge). Impression cytology confirmed ingrowth of goblet cell–containing epithelium in the corneas. On the other hand, the 44-mer-treated group remained free of neovascularization and goblet cells even after repeated epithelial debridement challenges (Table; Fig. 6). These data demonstrate that 44-mer treatment apparently enables the regeneration of a limbus with structural and functional similarities to the normal limbus. 

Figure 5

 

(A) Schematic representation of n-heptanol removal of the 44-mer-induced regenerated limbal epithelium and experiment design rationale confirmed that the preserved barrier function is derived from the regenerated limbus. (B) Two months after 180° limbal excision and 44-mer treatment, an absorbant gelatin sponge (GelFoam; Pfizer Pharmaceutical, Kalamazoo, MI, USA) was moistened with n-heptanol and applied to the healed limbal wound. The ocular surface was then rinsed with sterile saline and the limbal epithelium removed by mechanical scraping. In control eyes, the 44-mer-regenerated limbus received saline only and no mechanical debridement. External eye photographs of corneal epithelial wound healing stained with topical fluorescein revealed delayed wound healing in the eyes treated with n-heptanol. Histologic sections from the regenerated limbus and cornea were stained with PAS to document the invasion of goblet cells (purple). Data represent two independent experiments and two rabbits per group.

Table.

 

Distribution of Limbal NV After Serial Challenges to Limbal Barrier Function

 

Figure 6

 

Photographs at high magnification of the limbal excision area from rabbit eyes following two consecutive 10-mm corneal epithelial removal challenges, starting 2 months after 180° limbal excision and 44-mer treatment. The interval between the challenges was 1 month. Exacerbation of fibrovascular ingrowth into the cornea was observed with each scraping in the vehicle-treated limbus. Dotted lines demarcate the area of limbal resection. Impression cytology was performed after the corneal epithelial wound had healed completely, with goblet cells stained purple. Representative images are from three independent experiments.

The Regenerated Limbus Supplies Epithelial Cells for Corneal Wound Healing

In order to investigate the ability of the regenerated limbus to sustain corneal epithelial wound healing, 2 months after 180° limbal excision and 44-mer (n = 8) or vehicle (n = 8) ointment treatment, rabbits underwent corneal epithelial wounding as described in Methods. The area of epithelial defect showing fluorescein staining was comparable in the normal controls and the 44-mer-treated group that exhibited complete healing within 7 days (Fig. 7A). Epithelial healing was significantly delayed in the eyes of vehicle control rabbits compared to the 44-mer group at all time points (P < 0.001) with the exception of day 1, as was wound closure after 9 to 12 days (Fig. 7). As expected, the vehicle control group showed an asymmetric, drop-shaped healing pattern attributable to delayed wound closure of the cornea adjacent to the injured limbus (marked by dotted line; Fig. 7A, upper).²⁹ Overall, the 44-mer-regenerated limbus can support corneal wound healing as efficiently as the normal limbus. 

Figure 7

 

The regenerated limbus supplies epithelial cells for corneal wound healing. (A) Serial external eye photographs of corneal epithelial wound healing. The 10-mm corneal wound was stained with topical fluorescein and followed for approximately 7 to 9 days. Dotted lines demarcate the area of limbal resection. Treatments of the limbal wounds were as indicated. Normal indicates animals that did not receive limbal excision. Representative images are from three independent experiments and three rabbits per group. (B) Comparison of the healing curve between normal control, vehicle, and 44-mer-treated PLD corneas in response to one large central epithelial debridement (10-mm diameter). The healing curves were derived as described in Methods. Statistically significant delay of wound healing had been noted ever since day 2 after debridement (P < 0.05, n = 6 for each group).

Discussion

A variety of pathological conditions may lead to the damage of the limbus, which do not regenerate spontaneously. In this study, we sought to determine whether in vivo application of the 44-mer can stimulate limbal regeneration and to develop new insights into ocular surface reconstitution. We showed that the 44-mer can regenerate limbal tissue following limbal excision wounding. The regenerated limbus has many structural similarities to the native limbus, such as cells expressing the LSC markers ΔNp63α and ABCG2, and remains stable throughout many months of observation. Clinically, it retains the capacity for corneal wound healing and the limbal barrier function that prevents the invasion of conjunctival epithelium and fibrovascular tissue into the cornea. 

In our study, the 44-mer promoted limbal regeneration and corneal wound healing. We excised the full thickness of the limbal epithelium and a layer of stroma beneath to exclude the possibility of limbal regeneration from residual LSCs in the wound.²² The failing of the ocular surface when the regenerated limbus was destroyed with n-heptanol indicated that it was the source of the progenitor cells that sustained the corneal epithelium. In our study, subconjunctival and topical steroids were used at the early stage of the limbal wound healing with the aim of potentiating control of scarring and inflammation, which would compromise normal epithelialization that occurred within 2 weeks. Scarring and inflammation control are critical during limbal wound healing because these risk the viability of either residual or transplanted LSCs by interfering with their capacity to proliferate and differentiate.^30,31 The utility of the 44-mer under such conditions may be limited and awaits exploration. Clinically, the possibility of limbal regeneration can benefit autologous transplantation by permitting smaller donor limbal grafts and minimizing the risk to healthy donor eyes. Also, in partial limbal deficiency (PLD), numbers of LSCs inadequate to sustain a healthy corneal surface may be addressed by using the 44-mer for LSC proliferation and tissue engineering techniques to improve stromal integrity. 

Amniotic membrane is widely used as a carrier for ex vivo expansion and transplantation of corneal LSCs. Several clinical studies supported the finding that AMT may facilitate in vivo expansion of residual viable LSCs in ocular surface diseases: Partial and even nearly total LSCD can be treated by AMT alone^32–34; smaller-size limbal autograft can be used to treat unilateral total LSCD,^35–37 and early intervention of AMT can prevent LSCD in severe ocular surface disorders such as chemical burns.^8,38–40 The beneficial effect of AMT in our study may be in part due to the restoration of an intact basement membrane that is known to support epithelial cell adhesion, differentiation, and migration⁴¹ and to suppress apoptosis.⁴² Moreover, the stromal matrix of the AM is known to contain growth factors and several forms of protease inhibitors that may reduce stromal inflammation and fibrosis during corneal and conjunctival surface wound healing as suggested in previous reports.^32,43 Our experiments failed in the absence of AMT due to severe inflammation and fibrovascular proliferation at the limbal excisional wound as reported in previous studies.²¹ We used AMT as a biological bandage over the surgical wound not only with the aim of reducing wound inflammation and scarring that per se may impede normal epithelialization, but also with the idea that it would serve as a carrier for in vivo expansion of LSC induced by 44-mer treatment. 

In eyes with PLD, corneal epithelialization can be achieved due to the existing residual LSC function.^44,45 Although many reports have described the successful application of AMT for treatment of ocular surface diseases associated with PLD and proposed it as an alternative to limbal transplantation, recurrence and failure of the procedure were also reported in several cases, especially in extensive PLD.^32,33 It seems that to avoid recurrence and achieve stable corneal epithelialization, not only the basement membrane but also a source of stem cells and a niche are important. Follow-up studies showed that AMT was insufficient to maintain long-term stability of the ocular surface and could provide only a temporary treatment option for patients with PLD before the application of other treatment modalities such as limbal transplantation.³² Although several biological effects have been attributed to PEDF, including antiangiogenic, antioxidant, anti-inflammatory, and neuroprotective properties, the 44-mer is known for its neurotrophic and mitogenic effect on adult neural and limbal progenitor cells.^19,46,47 Our study demonstrated successful limbal regeneration in PLD eyes by combining AMT and topical 44-mer treatment. 

In order to regenerate the limbus in patients with total LSCD, several surgical transplantation methods, such as conjunctival limbal autograft (CLAU), keratolimbal allograft (KLAL), and transplantation of LSC after ex vivo expansion, have been employed with variable success.^3–5,7 The prolific approaches for the treatment of LSCD show the difficulty in regenerating limbal function. The possibility of limbal regeneration presented in this report may have an impact on both the exploration of the true nature of limbus and future therapy of ocular surface diseases. The mechanisms governing the activation, migration, and relocation of LSCs can now be studied in this new system. There is also much we can learn about the interaction between LSCs and stromal mesenchymal cells in the limbal niche by using this system. Clinically, the possibility of limbal regeneration can benefit autologous transplantation by using a smaller donor limbal graft in order to minimize the risk to a healthy donor eye. For partial limbal damage such as by chemical injury, the number of LSCs may be inadequate to sustain a healthy cornea. With the help of PEDF peptide and tissue engineering techniques to improve stromal integrity, these patients may regain a healthy ocular surface. More extended studies and further structural analysis of the regenerated limbus with more specific markers are needed to provide more convincing proof of the presence of LSCs. Our next step is to determine whether the 44-mer can regenerate the limbus after total native limbal excision and to search for the molecular mechanisms involved in the mitogenic effect of PEDF peptide on LSCs. We are optimistic about the therapeutic potential because there are many possible improvements of our current formulation. 

Acknowledgments

We thank Chu-Ping Ho for support, Chin-Min Wang for assistance with animal experiments, and Ju-Yun Wu for assistance with cell culture experiments. 

Supported by grants from the National Science Council, Taiwan (NSC 101-2314-B-195-006-MY3), and Mackay Memorial Hospital (MMH-E-101-006). The authors alone are responsible for the content and writing of the paper. 

Disclosure: S.-I. Yeh, None; T.-C. Ho, None; S.-L. Chen, None; C.-P. Chen, None; H.-C. Cheng, None; Y.-W. Lan, None; J.-W. Hsieh, None; C.-T. Wang, None; Y.-P. Tsao, None