Peroxisome proliferator-activated receptor alpha agonist suppresses neovascularization by reducing both vascular endothelial growth factor and angiopoietin-2 in corneal alkali burn

We investigated the effect of a peroxisome proliferator-activated receptor alpha (PPARα) agonist ophthalmic solution in wound healing using a rat corneal alkali burn model. After instillation of a selective agonist of PPARα, fenofibrate, onto the burned cornea, PPARα-positive cells were observed in vascular endothelial cells, and there was upregulation of mRNA of PPARα in corneal stroma. Fenofibrate suppressed expression of neutrophils and macrophages during the early phase, and development of neovascularization and myofibroblast generation during the late phase. Fenofibrate reduced not only mRNA expression of vascular endothelial growth factor-A but also angiopoietin-1 and angiopoietin-2. Furthermore, fenofibrate suppressed scar formation by reducing type III collagen expression. These data suggest that a PPARα agonist ophthalmic solution might be a new strategy for treating corneal wounds through not only anti-inflammatory effects but also by preventing neovascularization.

Effect of PPARα agonist on nuclear factor-kappa B (NF-κB). To examine PPARα effects on NF-kB expression, immunohistochemical analysis and western blotting were performed using anti-p65 antibody. In inflammatory cells and vascular endothelial cells after alkali injury, NF-kB expression was localized in the nuclear area in vehicle group, while it was expressed in cytoplasm in PPARα group (Fig. 3a,b,a' and b'). Western blotting revealed there was less expression of NF-kB in the PPAR versus vehicle group ( Fig. 3c and d).
Expression of PPARα in rat cornea. After instilling fenofibrate, PPARα staining was performed in order to investigate changes in expression of PPARα in burned cornea after 6 hours. PPARα-positive cells were observed in epithelial basement cells of normal rat cornea (Fig. 4a). After alkali injury, PPARα-positive cells were observed in inflammatory cells in both groups (Fig. 4b and c). Compared to vehicle group, PPARα group had prominent

Suppression of neovascularization by PPARα agonist.
To investigate neovascularization, nestin and aminopeptidase P (JG12) were immunostained. In both groups, nestin-positive endothelial cells ( Fig. 5a and b) were observed in corneal limbus starting from day 4, with PPARα agonist significantly reducing nestin-positive endothelial cells (Fig. 5c). Subsequently, JG12-stained capillary lumens were noted in the corneal centre on day 14 ( Fig. 5d and e). There was a significantly smaller number of capillary lumens in the PPARα group versus the vehicle group (Fig. 5f). Double immunofluorescence studies demonstrated that PPARα was expressed on vascular endothelial cells, suggesting that expression of PPARα was associated with neovascularization ( Fig. 5g-i). There was lower mRNA expression of VEGF-A (Fig. 5j) in the PPARα versus vehicle group. Furthermore, mRNA expressions of Ang-1 and Ang-2 were significantly lower at 6 hours ( Fig. 5k) and at day 4 (Fig. 5l). These results indicated that PPARα agonist ophthalmic solution prevented development and migration of neovascularization.

Regeneration of corneal stroma and corneal transparency.
To observe fibrotic changes in corneal stroma during wound healing, we focused on alpha-smooth muscle actin (α-SMA) -positive myofibroblasts and type III collagen. Corneal stroma mainly consists of type I collagen ( Fig. 6a and e). On day 14 after injury, we noted accumulation of α-SMA-positive myofibroblasts (Fig. 6b) and deposition of type III collagen (Fig. 6c). However, PPARα group exhibited a lower degree of α-SMA and type III collagen ( Fig. 6f and g). Macroscopic photographs also showed that PPARα group had a higher corneal transparency and a lower central opacity compared to vehicle group ( Fig. 6d and h). Percentages of type III collagen in corneal regions were significantly lower in PPARα group versus vehicle group (Fig. 6i). While there was a gradual increase in type III collagen area in vehicle group, lower percentages of type III collagen area were found in PPARα group. Low-vacuum scanning electron microscopy (LV-SEM) analysis showed the detailed sequence of collagens ( Fig. 6j and k). Although PPARα group exhibited more regulated collagens and less corneal oedema compared to vehicle group, collagens observed in vehicle groups exhibited a wavy-like arrangement.

Discussion
Alkali burn is one of the most severe corneal injuries, as it not only damages corneal epithelium and stroma but also causes acute inflammation that can lead to neovascularization 17,18 . Thus, alkali burn model is a suitable tool for investigating PPARα agonist effects on corneal injury. Our current study demonstrated suppressive effects of PPARα agonist on inflammatory cell infiltration, neovascularization, and scar formation in alkali-injured cornea. We also found that anti-inflammation and anti-neovascularization effects of PPARα were closely related pathologically.
With our present model, we first observed corneal infiltration of cells after injury, with PPARα group leading to a decrease in infiltration of approximately 30% during the early phase. Anti-inflammatory effects of this agonist have been reported to be caused by interference of PPARα on the activity of many proinflammatory transcription factors, such as signal transducer and activator of transcription (STAT), activator protein-1 (AP-1), and NF-kB pathway [19][20][21][22] . Results of immunohistochemistry and western blotting examinations suggested that PPARα agonist effects occurred through NF-kB pathway by reducing its expression. Expression of other transcription factors including c-Jun, c-Fos, STAT5b, and phosphorylated STAT (P-STAT) showed no significant difference between PPARα and vehicle groups ( Supplementary Fig. S1). These agonist effects were confirmed by our findings that showed PPARα suppressed infiltration of inflammatory cells. Our results also indicated that there was a difference in both rate and order of infiltrations in accordance with type of inflammatory cells. We found that neutrophils  infiltrated first followed by macrophages, with infiltrating neutrophil numbers larger than that observed for macrophages. These results were in agreement with previous study findings 12,23 . In addition, fenofibrate suppressed infiltrations of neutrophils and macrophages, regardless of phase.
Subsequently, we observed alterations in localization of PPARα after injury. Although localizations of PPARα have been reported to occur in a variety of organs, such as liver or kidney 24 , localization of PPARα in eye has yet to be definitively clarified. Although we found that PPARα-positive cells were observed within inflammatory cells of both groups after an alkali injury, PPARα group exhibited prominent PPARα-positive cells that showed intense intranuclear staining, particularly in vascular endothelial cells, as compared to controls. A previous cell culture study that examined ratios of PPARα to γ found the ratio was 5.06 in human aortic endothelial cells, while it was 0.043 in monocytes 13 . These differences in cell-specific expression of PPARα and γ may contribute to their different roles in wound healing. In fact, PPARγ has been reported to play a role in M1 and M2 macrophage differentiation during elicitation of immune response [25][26][27] . Abundant localization of PPARα in vascular endothelial cells that was found in our current study suggests that PPARα may play a specific role with regard to neovascularization.
Neovascularization after corneal injury can be caused by various factors such as inflammation and hypoxia 28,29 . In our present study, PPARα agonist treatment reduced immature vascular endothelial cell proliferation during early phase, while it suppressed neovascularization during late phase. Double immunofluorescence staining also revealed that PPARα was strongly expressed in vascular endothelial cells in PPARα group, which suggests that PPARα plays a role in angiogenesis. In addition, we also found that there was a significant downregulation of mRNA expression of VEGF-A in PPARα group compared to control. Inflammation is a well-known trigger of angiogenesis, with previous reports documenting effects of anti-inflammatory agents such as NF-κB inhibitors on neovascularization 30,31 . Results of our current study confirm these previous findings. Interestingly, fenofibrate additionally decreased expression of Ang-1 and Ang-2. Furthermore, it has been reported that while blood vessel formation is regulated by VEGF along with Ang-1 and Ang-2, neither Ang-1 nor Ang-2 alone are able to promote neovascularization 32 . Asahara et al. have reported that addition of Ang-1 to VEGF produced an increase in macroscopically evident perfusion of corneal neovasculature, while addition of Ang-2 to VEGF assisted in initiation of neovascularization 33 . At 6 hours after injury in our current study, which is defined as an extremely early phase of angiogenesis, there was a marked increase in mRNA expression of Ang-2. However, instillation of fenofibrate significantly reduced this expression level. Moreover, during early phase neovascularization on day 4, fenofibrate treatment suppressed both Ang-1 and Ang-2. However, on day 14, which is associated with late phase neovascularization, there was no significant difference observed between these groups. Thus, these results suggest that PPARα suppressed angiogenesis by reducing expression of Ang-2 during the early phase. It has been reported that Ang-2 initiates endothelial cell proliferation by dissociating vascular endothelial cells and pericytes 34 . When considering effectiveness of therapies that target VEGF-A and Ang-2 for retinal neovascular diseases, our findings that PPARα suppressed Ang-2 in this study is intriguing from a therapeutic point of view 35 . However, since we could not definitively elucidate a relationship between these effects and the anti-inflammation effects in our current study, further investigations will need to be undertaken.
Normal cornea consists of regulated type I collagens that maintain transparency by helping to properly refract light 36 . Although type III collagens and myofibroblasts have an important role in wound healing, irrelevant deposition of collagens can result in an opaque cornea 37 , which is frequently observed after alkali burn. Our current study showed there was an accumulation of α-SMA-positive myofibroblasts and type III collagen in stromal areas. This was suppressed by fenofibrate, thereby resulting in fewer scar formations. Recent development of LV-SEM has made it possible to observe three-dimensional ultrastructural changes in tissues 38 . Therefore, we used LV-SEM for further detailed observations of these changes. Our findings revealed that arrangement of collagen fiber was smooth and regulated in fenofibrate group, while it was rough with micro-holes and corneal oedema in controls. Thus, these findings pathologically confirm the anti-scar effect of fenofibrate instillation.
In summary, instillation of fenofibrate suppressed inflammation, fibrosis formation, and neovascularization in alkali burned cornea. Results suggest that suppression of NF-kB expression is involved in an anti-inflammation effect, while downregulation of VEGF, Ang-1, and Ang-2 function is associated with an anti-neovascularization effect. Thus, PPARα agonist ophthalmic solution might be an effective treatment for severe corneal wounds associated with inflammation, neovascularization, and scar formation.

Materials and Methods
Animals. Eight-week-old male Wistar rats (Sankyo Laboratory Service, Tokyo, Japan) were used for all experiments in our present study (n = 10 per time point). All animal experiments were conducted in compliance with the Experimental Animal Ethics Review Committee of Nippon Medical School, Tokyo, Japan, and all procedures conformed to the Association for Research in Vision and Ophthalmic and Visual Research.
Alkali burn model. To create a corneal alkali burn, a 3.2 mm diameter circular piece of filter paper that had been soaked in 1 N NaOH was placed on central cornea of each rat for 1 min while under general isoflurane anaesthesia. After 1 min of alkali exposure, corneas were then rinsed with 40 mL physiologic saline. All procedures were performed unilaterally in right eyes of each rat.

Treatment with PPARα ophthalmic solution.
This study used two kinds of ophthalmic solutions: a vehicle solution and a 0.05% fenofibrate solution. Ophthalmic vehicle solution was prepared using 0.1 mL polyoxyethylene sorbitan monooleate (Wako Pure Chemical Industries, Osaka, Japan) and 100 mL NaCl-based PBS (0.01 M; pH 7.4), which was prepared with disodium hydrogen phosphate dodecahydrate (232 g), sodium dihydrogen phosphate dihydrate (23.7 g), and distilled water (4000 mL). Ophthalmic solutions, including 0.05% fenofibrate, were prepared in 20 mL of vehicle solution with 10 mg fenofibrate (Wako Pure Chemical Industries). Ophthalmic solutions were kept at 4 °C. During solution administration, either 0.05% fenofibrate ophthalmic solution (PPARα group) or vehicle (vehicle/control group) was topically instilled onto ocular surfaces of each rats' eye. Topical administration was continued in each group twice a day until reaching appropriate endpoints (at 6 hours and on days 1, 2, 4, 7, and 14 after alkali burn). At each endpoint, macroscopic photographs of each group were obtained and reviewed. Rats were euthanized at each endpoint by exsanguination under general isoflurane anaesthesia. Enucleated eyes were used for histological and immunohistochemical analyses, with real-time RT-PCR and LV-SEM performed after macroscopic examinations. For RT-PCR analyses, all corneal tissues were immediately transferred into RNAlater solution (Life Technologies, Carlsbad, CA, USA) and stored at −80 °C. Histological and immunohistochemical analyses. Enucleated eyes were fixed in 10% buffered formalin and embedded in paraffin for light microscopic analysis. Deparaffinized tissues were stained with HE for histopathological examination. EST staining was performed to detect infiltrating neutrophils 39 .
For the ED1, α-SMA, type I and type III collagens, PPARα, nestin, and JG12 immunohistochemical analyses, deparaffinized tissues were stained using a standard avidin-biotin-peroxidase complex technique. Percentage of positive pixel intensity of type III collagens in 200X corneal regions was quantitatively analysed on days 4, 7, and 14 using a computer-assisted image analysis system in conjunction with colour image-analysing software (WinROOF; Mitani, Tokyo, Japan).
PPARα and JG12 were detected by examining frozen tissue sections by double immunofluorescence staining for PPARα (mouse; Texas Red) or JG12 (goat; fluorescein isothiocyanate).

LV-SEM.
After periodic acid-methenamine silver staining without a mounting cover glass, all sections were then examined using LV-SEM (Hitachi Tabletop Microscope TM3030, Hitachi High-Technologies Corp., Tokyo, Japan) 44 . LV-SEM observations in our present study compared the same corneal central area in each group. Ultrastructural alterations of corneal stroma were assessed by LV-SEM using acceleration voltages at 15 kV with 30 Pa for the backscattered electron detector.

Real-time RT-PCR.
Real-time RT-PCR was performed to investigate changes in mRNA expression of PPARα after instillation of fenofibrate. We also analysed VEGF-A, Ang-1, and Ang-2 mRNA expressions. Total RNA was extracted from corneas using Qiagen RNeasy Mini kit (Qiagen, Hilden, Germany) in accordance with manufacturer's protocol. To ensure RNA concentration and purity (A 260 /A 280 ), a NanoDrop ND-1000 V3. Western blot analysis. For western blotting analyses, whole cell lysates containing equal amounts of protein (25 μg) from corneal tissue were separated on 7.5% acrylamide gel by SDS-PAGE. After electrophoresis, separated proteins were transferred to polyvinylidene difluoride membranes (Invitrogen, Carlsbad, CA, USA) and incubated with anti-NF-kB (Thermo Scientific) and anti-β-actin (Sigma, St Louis, MO, USA) in order to confirm equal loading of each protein. Bound antibody was detected using appropriate HRP-conjugated second antibodies (Promega, Madison, WI, USA) for more than 60 min. Membranes were then washed and developed with SuperSignal West Femto Luminol/Enhancer solution (Thermo Scientific). Immunoreactivity on blots was detected using a LAS-4000 Luminescent Image Analyzer with CCD Camera (Fujifilm, Tokyo, Japan) and quantified, with all results expressed as ratios to β-actin protein amounts.

Statistical analyses.
All results are expressed as mean ± standard error. All values were determined by an unpaired Student's t-test using an analytical software program (Excel, Microsoft, Redmond, WA). A value of P < 0.05 was considered statistically significant.

Study approval. This experiment was approved by the Animal Care and Use Committee of Nippon Medical
School .