Cell Biology of Spontaneous Persistent Epithelial Defects After Photorefractive Keratectomy in Rabbits

Abstract

Purpose: To evaluate wound healing in rabbit corneas that developed a spontaneous persistent epithelial defect (PED) after photorefractive keratectomy (PRK).

Methods: Forty-eight 10- to 15-week-old female New Zealand White rabbits weighing 2.5 to 3.0 kg underwent either –3 diopter (D) or –9 D PRK to generate a series of corneas to study wound healing after injury. During that series, seven corneas developed a PED detected with 1% fluorescein staining at a slit lamp that either did not have epithelial closure by 1 week after surgery or subsequently had the closed epithelium break down to form a PED 2 to 3 weeks after surgery. The corneas had slit-lamp photography, with and without 1% fluorescein, and were removed from the normal PRK series. Each PED cornea was evaluated using immunohistochemistry for the myofibroblast marker α–smooth muscle actin (α-SMA), keratocyte marker keratocan, and mesenchymal cell marker vimentin, as well as basement membrane components perlecan and collagen type IV.

Results: All seven corneas that had PRK with a PED, even the two evaluated at only 1 week after PRK, had α-SMA–positive myofibroblasts populating the anterior stroma within the PED, along with comingled α-SMA–negative cells that were likely corneal fibroblasts and possibly bone marrow–derived fibrocytes. Both perlecan and collagen type IV accumulated in the anterior stroma of the epithelial defects without an epithelial basement membrane, likely produced by corneal fibroblasts to modulate transforming growth factor–β entering the stroma from the tears and peripheral epithelium.

Conclusions: Corneas with a PED that occurred following PRK (a procedure that produces a transient neurotropic state in the cornea) had myofibroblasts populating the superficial stroma within the epithelial defect as early as 1 week after the surgery.

Translational Relevance: Pharmacologic treatments that trigger myofibroblast apoptosis, including topical losartan, could facilitate decreased scarring fibrosis in corneas with a PED.

Introduction

A persistent epithelial defect (PED) in a cornea is commonly associated with diseases such as herpes simplex keratitis, herpes zoster ophthalmicus, or diabetes mellitus.^1,2 A PED often leads to severe corneal scarring and vision loss, even if the epithelial defect eventually heals.^1,2 A PED occurs infrequently in humans after photorefractive keratectomy (PRK) but can become a serious problem leading to severe corneal scarring fibrosis. One study of eyes that had PRK found delayed healing of the epithelium in 3.5% of eyes that had PRK with mitomycin C (MMC) intraoperative treatment but only 0.8% of eyes that had PRK without MMC treatment.³ All of those eyes had healing of the PED by 14 days after surgery with treatments that included therapeutic bandage contact lenses and epithelial scrape to freshen the edges of the PED.³ 

It is not surprising there are occasional issues with epithelial healing after PRK, with or without MMC treatment, because these corneas have injury to the corneal nerves that trigger retrograde degeneration of the nerves beyond the excimer laser ablation zone.⁴ The corneal nerves at least partially regenerate by 3 months after PRK in rabbits.⁴ In that rabbit study, MMC had a small additive toxic effect on the corneal nerves to the excimer laser ablation in rabbits, but that effect was only significant prior to 1 month after the surgery.⁴ Corneal nerves have been shown to produce modulators, such as nerve growth factor (NGF) and substance P, that are trophic to the corneal epithelial cells, keratocytes, and corneal fibroblasts and modulate epithelial viability and healing after injury.^5,6 Topical human recombinant NGF (Oxervate; Dompé, Milano, Italy) is approved in the United States by the US Food and Drug Administration and in other countries for the treatment of neurotropic corneas and often stimulates closure of the PED.⁷ 

The authors previously published a study of two spontaneous PEDs that developed after PRK in 168 rabbit eyes.^8–10 In both of those corneas, myofibroblasts developed in the anterior stroma within the PED, and no normal epithelial basement membrane (EBM) was detectable within the epithelial defect, although the EBM had regenerated normally in the other areas of the excimer laser ablation zone where the epithelium healed promptly. In a more recent PRK series in rabbits, in which the status of the epithelium in each eye was monitored carefully with serial slit-lamp evaluations, seven eyes were found to have either defective healing of the epithelium beyond the normal 4 to 5 days after PRK in rabbits or a subsequent breakdown of the corneal epithelium with later development of a PED. In the present study, these eyes were evaluated with immunohistochemistry to extend observations regarding corneal stromal cellularity in the PED, as well as perlecan and collagen type IV, that not only are EBM components but also modulate transforming growth factor–β (TGF-β) entry into the stroma and/or binding of activated TGF-β to the cognate TGF-β receptors. 

Methods

Animals and PRK Surgery

All rabbit surgeries and procedures were approved by the Institutional Animal Care and Use Committee at the Cleveland Clinic Foundation, and animals were treated in accordance with the tenets of the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. Forty-eight female 12- to 15-week-old New Zealand White rabbits weighing 2.5 to 3 kg each were included in a series of corneas from 1 week to 8 weeks after surgery that would be used in studies of growth factors and the cellularity of stromal wound healing. Twenty-four hours prior to PRK, and continuing at least 5 days after PRK, all rabbits received 60 mL children's liquid acetaminophen (Johnson & Johnson, Ft. Washington, PA, USA) per liter of drinking water. One eye of each rabbit was randomly selected to have either –3 diopter (D) or –9 D PRK and received topical 1% proparacaine hydrochloride (Alcon, Fort Worth, TX, USA) prior to the surgery. The rabbit was placed under general anesthesia with 30 mg/kg ketamine hydrochloride and 5 mg/kg xylazine by intramuscular injection. In the surgical eye, manual epithelial debridement was performed using a No. 6400 Beaver blade (MedexSupply, Passaic, NJ, USA), and PRK with a 6.5-mm optical zone was performed with a VISX (Santa Clara, CA, USA) S4 IR excimer laser using previously published method.⁹ No mitomycin C or corticosteroids were administered following surgery, but all eyes that had PRK were treated with one drop of topical ciprofloxacin three times a day until the epithelium had closed (typically by 4 to 5 days after surgery in this series and prior studies of PRK in rabbits).⁹ Unwounded contralateral controls were included in the study since no contralateral effects of PRK have been noted in prior PRK studies.^9,10 

Monitoring Corneas for Closure of Corneal Epithelial Defects

Beginning at 7 days after surgery, all corneas were monitored for epithelial defects with 1% fluorescein drops (Bausch & Lomb, Rochester, NY, USA) and a cobalt blue penlight and then at weekly intervals from 1 week to 8 weeks after surgery, depending on the planned endpoint for each cornea in the PRK study design. If an epithelial defect was still present at 7 days after surgery or development of an epithelial defect was detected from 2 to 8 weeks after PRK, the cornea was considered to have abnormal healing and a PED. In that case, that animal was removed from the normal wound-healing study and replaced, and that eye was entered into the present PED study. Each of these eyes that had a PED identified from 1 week to 3 weeks after PRK had slit-lamp photos, were euthanized, and the corneas removed and cryofixed, as detailed below. 

Each cornea with a PED was dilated with two drops of 1% tropicamide (Akorn Co., Lake Forest, IL, USA) for 30 minutes, and slit-lamp photos, without and with 1% fluorescein drops, were taken at a standardized illumination level, angle of illumination, and imaging angle, with 10× magnification, using a Topcon (Oakland, NJ, USA) SL-D7 slit-lamp photography system with the rabbit under ketamine-xylazine general anesthesia. 

Corneal Cryofixation and Sectioning

After the PED was diagnosed and imaged, the rabbit was euthanized while under ketamine-xylazine general anesthesia with 100 mg/kg Beuthanasia (Shering-Plough, Kenilworth, NJ, USA) by intravenous injection, followed by bilateral pneumothorax. Both corneoscleral rims were removed with sharp Westcott scissors (Fairfield, CT, USA) and 0.12 forceps (Storz, St. Louis, MO, USA) without touching the cornea surfaces. Each corneoscleral rim was centered in a 24-mm × 24-mm × 5-mm mold (Fisher Scientific, Pittsburgh, PA, USA) that was subsequently filled with optimal cutting temperature (OCT) compound (Sakura Finetek, Torrance, CA, USA). The mold and cornea were then quick frozen on dry ice and stored at –80°C until sectioning was performed. 

Blocks were bisected in the center of the PED in each cornea, and 10-µm-thick transverse sections were cut from the central cornea using a cryostat (HM 505M; Micron GmbH, Walldorf, Germany). Three sections from each cornea were placed on each 25-mm × 75-mm × 1-mm Superfrost Plus microscope slide (Fisher Scientific). Slides with sections were maintained at −20°C prior to immunofluorescence-immunohistochemistry (IF-IHC). 

IF-IHC for α–Smooth Muscle Actin, Keratocan, Vimentin, Perlecan, and Collagen Type IV

Triplex IF-IHC was performed for α–smooth muscle actin (α-SMA; myofibroblast marker), vimentin (mesenchymal cell marker, at a concentration where all corneal fibroblasts and myofibroblasts but only rare keratocytes were positive),^9,10 and keratocan (marker for keratocytes), using previously described methods⁹ and primary antibodies confirmed by Western blotting and IF-IHC to recognize rabbit antigens or isotypic nonspecific control antibodies (ThermoFisher Scientific, Waltham, MA, USA) and previously described secondary fluorescent tagged antibodies (Table 1).⁹ IF-IHC was also performed for perlecan or collagen type IV with 4′,6-diamidino-2-phenylindole staining of nuclei using previously characterized primary and secondary antibodies (Table 1).⁹ The collagen type IV antibody (cat. AB769; Millipore, Temecula, CA, USA) used in this study was generated against purified human and bovine collagen type IV and affinity purified with human and bovine collagen type IV crosslinked to agarose and then cross-absorbed by the manufacturer with human and bovine collagens type I, II, III, V, and VI to eliminate cross-reactivity. This collagen type IV antibody was shown previously to bind rabbit collagen IV in IF-IHC¹⁰ and binds the α-1/α-2 chains but not the α-3 to α-6 chains. Winston Kao, PhD, graciously provided the keratocyte-specific keratocan antibody raised against peptide H2N-LRLDGNEIKPPIPIDLVAC-OH.^9,10 Images were obtained at 100× total magnification on a Leica DM6B upright microscope equipped with an automated stage and Leica 7000 T camera using the LASX software (Leica Microsystems, GmbH, Wetzlar, Germany). 

Table 1.

 

Primary and Secondary Antibodies

 

All images were converted to 600-pixel width × 448-pixel height 300 DPI images with Photoshop 23.5.2 (Adobe, San Jose, CA, USA) to prepare composite image figures and images for α-SMA quantitation in the anterior stroma using ImageJ 1.53a analysis software (National Institutes of Health, Bethesda, MD, USA).¹⁰ The mean pixels of stromal α-SMA intensity were determined in a 600-pixel wide × 150-pixel high rectangle with ImageJ for each PED cornea. This analysis was also performed for four unwounded control corneas, four normal –3 D PRK 2-week postsurgery corneas, and four normal –9 D PRK 2-week postsurgery corneas from the same wound-healing study where these PEDs occurred to allow comparisons to the myofibroblasts generated in PED corneas. 

Statistics

Statistical analyses were performed using the nonparametric Kruskal-Wallis test (https://www.statskingdom.com/kruskal-wallis-calculator.html) followed by a post hoc Dunn's test, and P < 0.05 was considered statistically significant. 

Results

Corneal epithelial defects were present in two corneas at 1 week after –3 D PRK, two corneas at 2 weeks after –3 D PRK (images shown in Fig. 1 and Supplementary Fig. S1), two corneas at 2 weeks after –9 D PRK, and one cornea at 3 weeks after –3 D PRK, in the 24 corneas that had –3 D PRK and the 24 corneas that had –9 D PRK for the normal wound-healing PRK series (seven PEDs total in 48 PRK corneas). Each of the corneas that had an epithelial defect at 2 weeks or 3 weeks after PRK had initially healed by 1 week and, therefore, had subsequent epithelial breakdown. No contralateral control cornea had an epithelial defect. 

Figure 1.

 

PED after PRK in a rabbit. Slit-lamp photograph taken (A) without fluorescein and (B) with topical fluorescein. The arrows delineate the PED. Magnification 10X.

No unwounded cornea (n = 4) had α-SMA–positive myofibroblasts in the stroma (Fig. 2A). As was reported previously,^9,10 at the dilution of the antivimentin antibody used in this study, some anterior stromal cells in unwounded corneas were vimentin positive (Fig. 2A), as were some stromal cells in the far posterior cornea just anterior to Descemet's membrane (not shown). In unwounded corneas, the entire stroma stained positively for keratocan (Fig. 2A). 

Figure 2.

 

Triplex IF-IHC for α-SMA (marker for myofibroblasts), vimentin (marker for mesenchymal cells), and keratocan (marker for keratocytes) in corneas with a PED at different time points after PRK in rabbits. Note in the unwounded cornea (A) there were vimentin-positive cells, most of which were also keratocan positive, in the anterior stroma at the antivimentin antibody concentration used in this study (these vimentin-positive stromal cells were also detected in the posterior stroma anterior to Descemet's membrane [not shown]).^9,10 There were no SMA-positive myofibroblasts detected in unwounded corneas, as has been reported previously.^9,10 Panels A and B show that all seven corneas with a PED had numerous α-SMA–positive myofibroblasts in the anterior stroma devoid of epithelium. Although there was considerable variation between these corneas in the numbers of the myofibroblasts, the numbers of these cells tended to be increased in corneas that had higher correction –9 D PRK. The numbers of keratocan-positive keratocytes in the anterior stroma varied between the PED corneas, but the most anterior stroma had a near absence of keratocytes within the PED. Dashed yellow boxes are the 600-pixel width by 150-pixel height quantitation boxes used for ImageJ quantitation of α-SMA staining intensity in the anterior stroma of each cornea. (B) A triplex isotypic negative control is shown for a section of the –9 D PRK at the 2-week cornea. The α-SMA IF-IHC staining for normal 2-week –3 D and –9 D PRK corneas used for ImageJ analysis comparisons to the PED corneas is shown in Supplementary Figure S2. (C) Graph of ImageJ quantitated α-SMA intensity in 600-pixel width by 150-pixel height quantitation boxes overlying the anterior stroma for unwounded corneas and PED corneas, as well at normally healed –3 D and –9 D PRK corneas from the same series (Supplementary Fig. S2). Note that no α-SMA was detected in the anterior stroma of unwounded corneas or these four –3 D PRK corneas at 2 weeks after surgery. The anterior stromal α-SMA was highly variable in the corneas with a PED after PRK, with the double hatch indicating values beyond the y-axis scale in two corneas where the values are provided for α-SMA intensity. Variability in α-SMA was also noted in the four –9 D PRK 2-week corneas, with two of these corneas having no detectible α-SMA. Statistical analyses for the α-SMA signal intensity in all four of these groups are shown in Table 2.

Figure 2.

 

Continued.

All corneas with epithelial defects (Figs. 2A, 2B) had α-SMA–positive myofibroblasts populating the anterior stroma within the epithelial defect. These α-SMA–positive myofibroblasts extended peripherally up to 0.5 mm into the surrounding stroma that was covered by epithelium in all corneas (not shown). The anterior stroma of each cornea with an epithelial defect had 50% to 100% more vimentin-positive cells in the anterior stroma than unwounded control corneas. In each cornea with an epithelial defect (Figs. 2A, 2B), many of these cells in the anterior stroma were α-SMA negative, indicating they were not myofibroblasts. The latter cells were intermixed with the α-SMA–positive myofibroblasts in the anterior stroma and populated more of the deeper stroma in each cornea than the α-SMA–positive myofibroblasts. Keratocan-positive keratocytes were decreased in the anterior stroma of five of the seven corneas with epithelial defects, but in two PED corneas (–3 D 2-week PRK-2 and –3 D 3-week PRK), keratocytes were present to the anterior stromal surface. The level of α-SMA signal intensity is a quantitative measure proportional to the number of myofibroblasts in an area of analyzed tissue. α-SMA signal intensity was analyzed in uniform-sized quantitation boxes (Figs. 2A, 2B and the Supplementary Fig. S2), with the anterior long side of the box tangent to the anterior stromal surface. These measurements were compared to those in unwounded control corneas, normally healed –3 D PRK 2-week post-PRK corneas, or normally healed –3 D PRK 2-week post-PRK corneas in Fig. 2C. It was noted that all seven PED corneas had high α-SMA signal intensity in the anterior stroma. The unwounded corneas had no α-SMA signal intensity in the stroma. Two of four normally healed –9 D PRK 2-week postsurgery corneas had myofibroblasts in the anterior stroma, and none of the four normally healed –3 D PRK 2-week postsurgery corneas had myofibroblasts in the anterior stroma, as shown in the Supplementary Figure S2. Statistical analyses showed that the PED corneas were significantly different from the unwounded and –3 D PRK 2-week corneas and approached being significantly different from the –9 D PRK 2-week corneas in α-SMA signal intensity in the anterior stroma (Table 2). 

Table 2.

 

Statistical Comparisons Between Groups

 

In the normal unwounded cornea (Fig. 3), most perlecan was localized to the EBM and Descemet's membrane (see –9 D 2-week PRK panel), although small amounts of perlecan were detected deeper in the stroma and likely associated with corneal fibroblasts and keratocytes. All six corneas with a PED had perlecan in the anterior stroma within the epithelial defects, and some of the perlecan was associated with anterior stromal cells that likely included both corneal fibroblasts and myofibroblasts. 

Figure 3.

 

IF-IHC for perlecan in an unwounded cornea and the seven corneas with a PED. In the unwounded cornea, perlecan was detected at high levels in the uniformly thin EBM, as well as in Descemet's membrane (shown for the –9 D 2-week PRK PED corneas [large arrows]). In all PRK corneas with a PED, perlecan was noted at variable levels in the anterior stroma within the PED devoid of epithelium or EBM. An isotypic control performed on a cornea within a PED is also shown.

In normal unwounded corneas, collagen type IV was also localized to the EBM (Fig. 4) and Descemet's membrane (also see de Oliveira et al.^9,10), although small amounts of collagen type IV were detected deeper in the stroma and appeared to be associated with keratocytes. Corneas with a PED after PRK, however, had large amounts of collagen type IV in the anterior stroma within the epithelial defects beginning at 2 weeks after surgery (although the –3 D 2-week PRK-1 cornea had not generated significant collagen type IV at this time point), and this collagen type IV was also associated with anterior stromal α-SMA–negative, vimentin-positive, keratocan-negative cells. 

Figure 4.

 

IF-IHC for collagen type IV in an unwounded cornea and the seven corneas with a PED. In the unwounded cornea, collagen type IV was detected at high levels in the uniformly thin EBM, as well as in Descemet's membrane (shown for one of the –9 D 2-week PRK PED corneas [large arrows]). Small amounts of collagen type IV in the anterior stroma of the unwounded cornea could represent extracellular vesicles. The collagen type IV in corneas with a PED was low at 1 week after surgery and in one cornea at 2 weeks after surgery but was present at high levels within the PED for one of the –3 D 2-week PRK corneas, the –9 D PRK corneas, and the 3-week PRK cornea. An isotypic control performed on a cornea within a PED is also shown.

Discussion

A PED occurs infrequently in a human cornea that has PRK³ and also occurs in corneas associated with disorders such as primary or recurrent herpes simplex keratitis, herpes zoster ophthalmicus, bacterial keratitis, trigeminal nerve disorders, or diabetes mellitus, where there is also damage to the corneal sensory nerves.^1,2 A PED often leads to severe corneal scarring and vision loss, even if the epithelial defect eventually heals.^1,2 

The current study included seven eyes that developed a spontaneous PED in 48 rabbit eyes that underwent –3 D PRK or –9 D PRK, without mitomycin C or corticosteroids, to produce a series of corneas at different time points for future wound-healing studies. The relatively high frequency of a PED in this series compared to a previous series^8–10 likely relates to the diligence of the investigators in monitoring the status of epithelial healing in these corneas using topical fluorescein at the first week and at weekly intervals after surgery. The two corneas did not heal by 1 week after PRK and were deemed to have a PED based on the presence of greater than 3-mm diameter epithelial defects at 1 week after surgery compared to complete healing of the epithelial defect by 4 to 5 days after normal PRK in rabbits.^9,10 The other five corneas with a PED were initially healed at 1 week after surgery but subsequently developed a PED at 2 to 3 weeks after the PRK surgery. These PED corneas were removed from the normal PRK series and replaced so that the future wound-healing studies would not include corneas with a PED. 

The most important finding in this series was that all corneas with a PED, even as early as 1 week after PRK, had α-SMA–positive myofibroblasts populating the anterior stroma within the epithelial defect (Figs. 2A, 2B) and approximately 0.5 mm beyond the peripheral edge of the epithelial defect. This result is consistent with the findings in two previous corneas with a spontaneous PED after PRK.⁸ Two of the normally healed –9 D PRK corneas in the current series, but none of the normally healed –3 D PRK corneas, also had α-SMA–positive myofibroblasts in the subepithelial stroma at 2 weeks after PRK (Supplementary Fig. S2). However, the myofibroblasts in those corneas are destined to undergo apoptosis by 6 to 8 weeks after PRK once the EBM is normally regenerated and these cells are deprived of adequate and ongoing TGF-β1 and/or TGF-β2 that enter the stroma primarily from the tears and epithelium after PRK.^9,10 In corneas with a PED, the area within the epithelial defect cannot begin the process of regenerating the EBM because epithelial cells that produce self- and copolymerizing laminins, like laminin 511 and laminin 512, which act as a scaffold for the recruitment of the other components in EBM regeneration, are not present.^9–13 Subsequently, in normally healing corneas, other components such as perlecan, collagen type IV, and nidogens are incorporated into the nascent EBM, and many of these later basement membrane components appear to be inserted through the actions of corneal fibroblasts and keratocytes.^9–16 While myofibroblasts can themselves produce some of these EBM components, including perlecan,^9,10,17 these cells do not appear competent to insert these components into the nascent EBM during the regeneration process.^9,10,14 

Each of the PED corneas had numerous vimentin-positive, α-SMA–negative, keratocan-negative cells that likely included corneal fibroblasts^9,10 and possibly bone marrow–derived fibrocytes,^18,19 intermingled with the myofibroblasts in the anterior stroma of the PED but also populating the more posterior stroma beneath the myofibroblasts (Figs. 2A, 2B). In PED corneas, there also tended to be few, but variable, keratocan-positive keratocytes in the anterior stroma within the PED (Figs. 2A, 2B). 

High levels of perlecan (Fig. 3) and collagen type IV (Fig. 4) tended to be present in the anterior stroma within the epithelial defect of PED corneas, although collagen type IV was not as prominent in corneas with a PED at 1 week to 2 weeks after PRK. Non-EBM-associated perlecan is produced by corneal fibroblasts, and possibly myofibroblasts,¹⁷ in the anterior stroma underlying the PED.^9,10 Similarly, collagen type IV is produced by corneal fibroblasts in the anterior stroma of PED corneas, and that production is upregulated in corneal fibroblasts by TGF-β that enters into the stroma and binds TGF-β receptors on the corneal fibroblasts.^20,21 High levels of perlecan in the anterior stroma within an epithelial defect produce a high negative charge due to the perlecan molecule's three heparan sulfate side chains and, therefore, provides a nonspecific barrier to TGF-β penetration into the deeper corneal stroma.^22–24 Collagen type IV directly binds TGF-β1 and TGF-β2 and prevents them binding to the cognate TGF-β receptors and thereby also downregulates TGF-β signaling.^25,26 It is likely that both of these components that are not associated with an EBM in these PED corneas are produced by corneal fibroblasts to modulate activated TGF-β effects on stromal cells. 

Corneal fibroblasts (and bone marrow–derived fibrocytes) within a PED serve as precursor cells to develop into additional myofibroblasts when their differentiation is driven by high levels of TGF-β1 and TGF-β2 entering the stroma.^9,10,18,19 However, the corneal fibroblasts also serve a dual function to cooperate in the regeneration of the EBM and also to produce non-EBM perlecan and collagen type IV to modulate TGF-β activity in the stroma,^9,10,20–26 and they may directly modulate myofibroblast apoptosis via the production of interleukin-1α (IL-1α)^27–29 and hepatocyte growth factor (HGF).^30–33 It remains enigmatic what factors control whether a particular corneal fibroblast in the stroma differentiates into a myofibroblast or, conversely, remains a corneal fibroblast that produces perlecan and collagen type IV to decrease TGF-β availability to myofibroblasts, regulates myofibroblast viability, and is prepared to cooperate with epithelium that heals over the PED in regenerating the EBM to further modulate TGF-β passage into the stroma.^9,10 This may be related to the localized levels of TGF-β, IL-1, HGF, and the other modulators that coordinate the functions and survival versus apoptosis of the myofibroblasts and other stromal cells in the vicinity of the epithelial defect. 

Each PED in this study was a relatively short-lived epithelial defect since in each case, they had existed less than 1 week. It is likely that the myofibroblast density in the anterior stroma of each of these corneas would continue to increase over time, with worsening of stromal fibrosis and corneal opacity, since there could be no EBM regeneration unless the epithelium eventually healed. Myofibroblasts themselves inhibit corneal nerve regeneration in injured corneas.³⁴ Therefore, the development of myofibroblasts in a PED due to a neurotropic cornea likely negatively impacts corneal sensory nerve recovery. 

Each of the corneas in this study, and similarly all corneas with a PED, regardless of etiology, becomes trapped in a pathologic state characterized by a corneal epithelial defect and stromal fibrosis, despite the efforts of the lacrimal glands,^35,36 residual peripheral epithelium, and corneal fibroblasts to repair the injury.^9,10 In a cornea with a PED, it is imperative to facilitate closure of the epithelial defect as quickly as possible to decrease myofibroblast generation and corneal scarring, as well as the risk of secondary microbial keratitis or corneal perforation. Clinical measures that can expedite closure of the epithelial defect include prescribing artificial tears, gels, and ointments; autologous serum solutions; bandage contact lenses; amniotic membranes; surgical freshening of the edges of the epithelial defect; tarsorrhaphy; and other approaches.^1,2 Importantly, occult herpes simplex keratitis notoriously produces persistent epithelial defects after PRK, and testing and potentially treatment with antivirals should be considered for a PED that occurs after PRK and other corneal surgical procedures.³⁷ 

Topical nerve growth factor (Oxervate; Dompé) is now available for clinical use and has shown promise in facilitating healing of a corneal PED in patients.⁷ Cost can be prohibitive for some patients since an 8-week course of topical Oxervate can exceed $80,000. Even when this treatment is effective in closing a PED, however, the affected cornea frequently has moderate to severe scarring stromal fibrosis that limits the vision in the eye. Several recent rabbit studies,^38–40 as well as a clinical report in humans,^41,42 have demonstrated the efficacy of topical losartan in decreasing stromal myofibroblasts and fibrosis after a variety of corneal injuries. It may be that combined topical Oxervate and topical losartan could provide the best treatment for neurotropic corneas with a PED and stromal scarring fibrosis. It is even possible that topical losartan alone could facilitate closure of persistent epithelial defects by triggering myofibroblast apoptosis and repopulation of the anterior stroma with corneal fibroblasts and keratocytes that produce HGF and keratinocyte growth factor, which modulate the proliferation, migration, and differentiation of corneal epithelial cells.^33,43 Clinical trials in human patients will be needed to answer these questions since a spontaneous PED occurs so infrequently in normal rabbits after PRK or other corneal surgeries. 

Acknowledgments

The authors thank Winston Kao, PhD, Cincinnati, OH for providing the keratocan antibody utilized in this study. 

Supported in part by Department of Defense grant VR210001 (SEW) and P30-EY025585 from the National Eye Institute, National Institutes of Health, Bethesda, MD; Research to Prevent Blindness, New York, NY; and The Cleveland Eye Bank Foundation, Cleveland, OH. 

Disclosure: L.P. Sampaio, None; V.V. Martinez, None; T.M. Shiju, None; G.S.L. Hilgert, None; M.R. Santhiago, None; S.E. Wilson, Cleveland Clinic (P)