Ferrostatin‐1‐loaded liposome for treatment of corneal alkali burn via targeting ferroptosis

Abstract Alkali burn is a potentially blinding corneal injury. During the progression of alkali burn‐induced injury, overwhelmed oxidative stress in the cornea triggers cell damage, including oxidative changes in cellular macromolecules and lipid peroxidation in membranes, leading to impaired corneal transparency, decreased vision, or even blindness. In this study, we identified that ferroptosis, a type of lipid peroxidation‐dependent cell death, mediated alkali burn‐induced corneal injury. Ferroptosis‐targeting therapy protected the cornea from cell damage and neovascularization. However, the specific ferroptosis inhibitor ferrostatin‐1 (Fer‐1) is hydrophobic and cannot be directly applied in the clinic. Therefore, we developed Fer‐1‐loaded liposomes (Fer‐1‐NPs) to improve the bioavailability of Fer‐1. Our study demonstrated that Fer‐1‐NPs exerted remarkable curative effects regarding corneal opacity and neovascularization in vivo. The efficacy was comparable to that of dexamethasone, but without appreciable side effects. The significant suppression of ferroptosis (induced by lipid peroxidation and mitochondria disruption), inflammation, and neovascularization might be the mechanisms underlying the therapeutic effect of Fer‐1‐NPs. Moreover, the Fer‐1‐NPs treatment showed no signs of cytotoxicity, hematologic toxicity, or visceral organ damage, which further confirmed the biocompatibility. Overall, Fer‐1‐NPs provide a new prospect for safe and effective therapy for corneal alkali burn.

and risk of graft rejection. 2 Therefore, more efficient and safe therapy is urgently needed in the treatment of alkali burn-induced injury.
Ocular drug delivery has always been a significant pharmaceutical challenge due to the complex physiology and anatomy of the ocular surface. 3,4 Structural variation of different layers of cornea can pose a barrier following drug administration by any route. 5 Due to the lower toxicity, improved biorecognition, and enhanced drug bioavailability, a variety of nanocarriers have been intensively investigated as possible carriers for delivering therapeutic agents to treat refractory disorders, including cancer, [6][7][8] spinal cord injury, 9 brain diseases, 10 and corneal diseases. [11][12][13] Since the mid-1990s, plenty of liposome formulations combined with drugs or biologics, including liposomal doxorubicin, cytarabine, vincristine, irinotecan, and amphotericin B, have been approved by the Food and Drug Administration for clinical use. 14 Liposomes are unilamellar or multilamellar phospholipid vesicles, where the loaded hydrophobic and hydrophilic agent can be encapsulated within the phospholipid bilayer and the aqueous core, respectively. 15,16 Drugs with high toxicity or low bioavailability benefit from the improved biocompatibility and stabilizing nature of liposomes, which enables the safety and efficacy of drugs. Although the cornea hinders deep ocular drug permeation as a physiological barrier, liposomes have been shown to circumvent this limitation and achieve sustained delivery of the drug into both the anterior and posterior segments. 17,18 Therefore, the liposome could serve as a potential nanocarrier for ocular drug delivery to repair the injured cornea.
Numerous studies have demonstrated that inflammatory response and corneal neovascularization (CNV) play pivotal roles in corneal injury and vision loss. [19][20][21] In the alkali-burned cornea, the wound-healing response could lead to a dysregulated inflammatory response. Various cells, such as inflammatory, stromal, and epithelial cells, are involved in these processes. Meanwhile, the angiogenesis factors are also activated. 22 Alkali burn-induced conjunctival and corneal epithelium injuries may also cause a limbal stem cell deficiency, leading to opacification and neovascularization of the cornea. 23 Further, it has been well documented that alkali burns promote the accumulation of reactive oxygen species (ROS) in the cornea. [24][25][26][27] Following a severe corneal injury, oxidative stress could lead to an antioxidant/prooxidant imbalance in corneal cells. The insufficiently scavenged ROS greatly contribute to excessive intracorneal inflammation, scar formation, and CNV. Moreover, the increased oxidative stress could cause disruption of mitochondrial permeability transition pore, perturbation of electron transfer, and destruction of cell membranes, ultimately leading to cell death. 28 Taken together, oxidative stress, inflammation, and neovascularization collectively aggravate the corneal injury and lead to vision loss. Previous studies indicated that ROS-related oxidases play a role in corneal alkali burn. 24 In addition, the inhibition of alkali-induced oxidative stress appeared to ameliorate corneal inflammation as well as CNV and promote corneal healing. 27,29 These results suggest that oxidative stress may represent a potential therapeutic target for corneal alkali burn. However, the underlying mechanisms are poorly understood.
Ferroptosis is a novel type of programmed cell death characterized by the lethal accumulation of lipid peroxides. It is biochemically, morphologically, and genetically different from other established forms of regulated cell death, such as apoptosis, autophagy, and necroptosis. 30,31 Ferroptosis is associated with numerous pathological processes, including tumorigenesis, neurodegenerative diseases, ischemia/reperfusion (I/R) injury, and liver injury. [32][33][34] Notably, it has been suggested that ferroptosis may contribute to the pathogenesis of some ocular diseases, including retinal pigment epithelium dysfunction, defects in corneal endothelial cells, and photoreceptor degeneration. [35][36][37][38] Nevertheless, the precise role of ferroptosis in the development of corneal injury induced by alkali burn remains unclear.
In this study, we identified that ferroptosis plays a vital role in alkali burn-induced corneal injury. Mechanistically, severe injury results in the accumulation of ROS, up-regulation of the ferroptosis promoter, and down-regulation of the ferroptosis depressor, which facilitates lipid peroxidation and mitochondria disruption, and eventually triggers ferroptosis. Targeting ferroptosis by specific inhibitor ferrostatin-1 (Fer-1) protects corneal cells from damage and angiogenesis, therefore promoting the corneal healing processes. However, as a promising pharmacological molecule, the poor solubility of Fer-1 in aqueous media raises a concern regarding the clinical application of Fer-1 in corneal diseases. A plethora of studies have demonstrated the effectiveness of liposomes for cornea drug delivery applications.
Due to the biodegradable and biocompatible lipid assemblies, liposomes could achieve high solubility and drug-loading, enhanced precorneal retention time, increased transmembrane permeation and absorption, as well as sustained release of the drug along with low toxicity. 14 Therefore, liposomes possess a substantial potential for clinical applications. In light of this, we designed and prepared Fer-1-loaded liposome (Fer-1-NPs) to treat corneal alkali burn via targeting ferroptosis. Interestingly, Fer-1-NPs exerted more curative effects regarding ferroptosis suppression, anti-CNV efficacy as well as corneal injury protection. The Fer-1-NPs were expected to enhance the solubility of Fer-1 and thereby improve its bioavailability. Moreover, the Fer-1-NPs showed low cytotoxicity and no damage to the cornea or visceral organs. Our findings provide solid evidence for the pathogenic mechanisms underlying alkali burn-induced corneal injury.
More importantly, we prepared Fer-1-NPs for safely and effectively inhibiting ferroptosis and CNV, which may provide potential targets for the development of novel therapeutic strategies. anti-glutathione peroxidase 4 (GPX4), anti-acyl-CoA synthetase longchain family member 4 (ACSL4), anti-CD31, and anti-α-smooth muscle actin (α-SMA) antibodies were purchased from Abcam. Rabbit anti-IL-1β and anti-IL-6 antibodies were purchased from Affinity Bioscience. Pri-meScriptTM RT Master Mix was acquired from Takara. ChamQ SYBR qPCR Master Mix was acquired from Vazyme.

| Liposomes preparation and characterization
Liposomes were synthesized using a thin-film hydration method. [39][40][41] Specifically, in a round-bottom flask, 2 mg of Fer-1 (or 3 mg Dex), 8 mg of cholesterol, 10 mg of DSPE-mPEG, and 30 mg of soybean lecithin (molar ratio, 2:5:1:10) were added to 10 ml of methylene chloride and dissolved completely. The mixture was evaporated using a rotary evaporator under reduced pressure until a homogeneous phospholipid film formed. The obtained phospholipid film was then hydrated with 10 ml of ultrapure water for 1 h. Subsequently, the mixture was ultrasonicated with 50 W for 2 min at 4 C (Biosafer). Homogeneous Fer-1-NPs and Dex-NPs were obtained by centrifugation (Thermofisher) at 6000 rpm for 10 min and filtration with a 0.22 μm-filter (Millipore).
It is worth noting that high-speed centrifugation and filtration were used to separate the nonencapsulated insoluble drug and sterilization, respectively, in this study. The hydrophobic fluorescent dye Nile Red loaded liposomes (Nile Red-NPs) were prepared using the same procedure. All the prepared liposomes were stored at 4 C for further use.
The size (hydrodynamic diameter) and polydispersity index (PDI) of the nanoparticles were measured by dynamic light scattering (DLS).
In detail, liposomes were first filtered through a 0.22-μm filter unit (Millipore) and then measured using Zetasizer Nano-ZS (Malvern). The morphology of Fer-1-NPs was characterized with a cryo-transmission electron microscope (cryo-TEM, FEI) operated at an accelerating voltage of 200 kV. The cryo-TEM size of the nanoparticle was quantified using ImageJ software (NIH).
To evaluate the concentration and encapsulation efficiency (EE) of the drugs, the nonencapsulated Fer-1 (Dex) was collected by ultrafiltration centrifugation (6000 rpm, 10 min) and quantified using highperformance liquid chromatography (HPLC) in triplicate. The following HPLC conditions for Fer-1 (Dex) were used: A reverse C18 column
Cells were maintained at 37 C and 5% CO 2 in a humidified incubator.

| Cellular uptake
HCECs were seeded at a density of 1.5 Â 10 5 cells/well in a 12-well plate. After culturing for 24 h, the cells were incubated with free Nile     The animal model of alkali burn-induced corneal injury was established as previously reported. 21 In brief, the mice were anesthetized with a sodium pentobarbital (100 mg/kg) intraperitoneal injection and proparacaine (0.5%) topical application to the right eye. After that, a 2 mm-diameter round filter paper pre-soaked into 1 M NaOH was blotted for 5 s and attached to the center of the right cornea for 30 s. Subsequently, the ocular surface was immediately washed using 10 ml saline for 60 s.

| Precorneal retention evaluation
The alkali-burned mice were anesthetized and topically instilled with 10 μl of free Nile Red (50 μM) or Nile Red-NPs (50 μM) before imaging. The head region was imaged using an IVIS Lumina imaging system (PerkinElmer) equipped with filter sets (excitation/emission, 535/600 nm). Imaging was performed at different time points (every 5 min) up to 30 min after administration. The initial fluorescence intensity was set as 100%, and that at following time points were normalized accordingly.

| Clinical assessment
Sixty mice were randomly divided into six groups. In addition to healthy mice without alkali burn, the alkali-burned mice were treated 3. Neovessel size was graded with a score of 0 to 3, where 0 = no detectable neovessels; 1 = neovessels detectable under microscope; 2 = neovessels easily seen under microscope; and 3 = neovessels easily seen without microscope.
If cornea perforation occurred, the total clinical assessment score was rated as 10. CNV was calculated as follows: S = C/12 Â 3.1416 Â [R 2 À (R À L) 2 ]. In this formula, S represented the CNV area, C represented the clock hour points covered by the CNV, R represented the radius of the cornea, and L represented the longest length of the neovessel.

| Histological and immunohistochemical assessment
On day 14, the right eyes were dissected and fixed in 4% paraformaldehyde (Biosharp) overnight. Then, the eyes were washed in PBS and embedded in paraffin. Histological sections (4 μm) were stained with hematoxylin-eosin (H&E) or immunostained with 4-HNE, Acsl4, Gpx4, α-Sma, and CD31, separately. All stained sections were examined and photographed using an Olympus BX-61 microscope. The central corneal thickness and the average optical density (AOD) were quantified by ImageJ software.

| Immunofluorescent staining
After 14-day treatment, the right eyes were dissected, fixed, embedded, and sectioned. Then, paraffin-embedded 4-μm sections were deparaffinized and rehydrated in successive baths of xylene and ethanol, followed by heat-induced epitope retrieval. After blocking with 5% bovine serum albumin/PBS for 1 h, IL-1β and IL-6 antibodies were added into the blocking buffer at 4 C overnight. Secondary antibodies were then incubated in the blocking buffer at room temperature for 1 h. Afterward, nuclei were stained with 4,6-diamidino-2-phenylindole, and slides were mounted with Fluorescent Mounting Medium (Dako). Sections were analyzed using a fluorescence microscope (Olympus BX-63), and the AOD was measured using ImageJ software.

| Measurement of corneal ROS production
Corneal ROS stress production was measured using CellROX Green reagent (Thermo Fisher) as described in the previous study. 24 NADPH oxidase inhibitor Diphenyleneiodonium Chloride (DPI, Selleck) was added for comparison. In brief, after alkali burn and 14-day treatment, the fresh corneal flat-mounts were dissected, washed with PBS, and permeabilized in 0.5% Triton-X for 10 min. Immediately, 5 μM CellROX reagent was added to the sections followed by incubation for 30 min at 37 C. Following incubation, the flat-mounts were washed three times with PBS. Fluorescence signals of ROS were detected with excitation and emission wavelengths at 485/530 nm under a fluorescence microscope (Olympus BX-63), and the AOD was measured using ImageJ software.

| Mitochondrial membrane potential (ΔΨm) detection
ΔΨm was measured using the sensitive fluorescent probe JC-1, a cationic dye of 5,5 0 ,6,6 0 -tetrachloro1,1 0 ,3,3 0 -tetraethyl benzimidazol carbocyanine iodide. 42 Briefly, the eyes were removed 14 days after treatment, and were immediately frozen in OCT compound (Sakura Finetek). The OCT-embedded slices of the cornea were prepared at 6 μm thickness. Afterward, the cornea slices were stained with 1Â JC-1 (YEASEN) at 37 C for 30 min in a dark place. The stained slides were washed three times with JC-1 buffer and then observed immediately with the fluorescence microscope (Olympus BX-63), and the AOD was measured using ImageJ software.

| Quantitative real-time polymerase chain reaction
On day 14, the mice were sacrificed and the right corneas were enucleated. Total RNA was extracted using TRIzol reagent (Takara). The extracted RNA was quantified by a NanoDrop Spectrophotometer (Thermo Scientific) and reverse-transcribed using the PrimeScript RT reagent Kit (Takara). After that, quantitative real-time polymerase chain reaction was performed using a CFX96 Real-Time System (Bio-Rad) with SYBR Green Supermix (Bio-Rad). Two corneas were combined to get one biological replica, and three biological replicas were used per group. Gene expression was normalized to Gapdh mRNA. The fold difference in gene expression was calculated using the comparative cycle threshold method.

| Mitochondria evaluation
The cornea tissues were immediately fixed in 3% phosphateglutaraldehyde after dissection. The samples were then post-fixed, embedded, cut, and mounted at the Electron Microscopy Core Facility of Zhejiang University. All samples were observed under a transmission electron microscope (FEI) at an accelerating voltage of 100 kV. For pathological conditions (with alkali burn), biocompatibility evaluation was conducted in the four experimental groups (the healthy, saline, Fer-1, and Fer-1-NPs groups) described in Section 2.6 (mouse model of alkali burn-induced corneal injury). After the 14-day treatment, blood was drawn for the white blood cells, red blood cells, hemoglobin, blood platelet, alanine transferase, aspartate transferase, blood urea nitrogen, and creatinine examinations. Five types of visceral organs, including the heart, liver, spleen, lungs, and kidneys, were excised for H&E staining.

| Statistical analysis
All data analyses and plots were generated using GraphPad Prism (version 6.0). The differences among groups were analyzed using a student's t-test or one-way analysis of variance as appropriate. A p < 0.05 was considered statistically significant.

| Ferrostatin-1 ameliorates alkali burn-induced corneal injury in mice
A schematic representation of the workflow of our in vivo modeling and the assessment is provided in Figure 1a. As two main injured manifestations of alkali burn, corneal opacity, and neovascularization (CNV) were continuously examined by slit-lamp microphotography.
On day 1 before administration, the corneas of both groups were comparable with respect to opacity and neovessels, indicating a good homogeneity of modeling. In both groups, the corneal opacity aggravated over time. Similarly, the neovessels grew gradually from the limbal vascular plexus toward the corneal center by days 3, 7, and 14 (Figure 1b,c). On day 14, the Fer-1 group showed more effective inhibition of CNV than the saline group, which was in accordance with the quantitative analyses of CNV length and area ( Figure 1d). Accordingly, the clinical corneal assessment score based on CNV size, CNV area, and opacity significantly favored the Fer-1 treatment (Figure 1e).
The gross view demonstrated that the saline-treated-burned cornea group exhibited obvious opacification and slight hemorrhaging, with the background pattern visually indistinguishable (Figure 1f). In contrast, the cornea in the Fer-1 group showed a moderate opacity, and the background pattern was easier to distinguish. As for the H&E staining results (Figure 1f), the saline group exhibited prominent corneal edema, as the thickness of the epithelium and stroma increased significantly compared with the healthy group, which was also shown in the quantitative analysis of the central corneal thickness (Figure 1g). In addition, remarkable increases in inflammatory cell infiltration and CNV formation were observed in the corneal stroma.
These findings may be due to the lack of structural integrity and the increased vascular permeability of the neovessels, which could result in corneal edema, inflammation, and fibrosis. 21 In the Fer-1 group, the edema was partially ameliorated, reflected by a statistically significant decrease in central corneal thickness compared with that in the saline group ( Figure 1g). This is clinically important since, even without CNV, edema in the cornea could independently lead to a loss of vision. 43 Although there was still evident infiltration of inflammatory cells and CNV, it was improved significantly compared to the saline group. Nonselective inhibitors of NADPH oxidase could ameliorate the inflammation and CNV caused by alkali burn, suggesting that anti-ROS can be an effective strategy for alkali burn treatment. 24 We then detected ROSrelated genes in the corneas of the three groups. As shown in Figure 2a Results were presented as the mean ± SEM. *p < 0.05 and **p < 0.01. Significance was calculated by one-way analysis of variance. AOD, average optical density; Fer-1, ferrostatin-1 could attack the mitochondria membrane, resulting in the destruction of mitochondrial membrane structure, reduced cristae, swelling, and rupture of mitochondria, which further aggravates intracellular lipid peroxidation. 48 Previous studies have suggested that in the rat model of corneal alkali burn, ROS were elevated and mitochondrial membrane potential decreased, leading to increased cytochrome C release and dysregulation of intracellular homeostasis. 49 To further explore the pathological mechanism of ferroptosis-induced injury during corneal alkali burn, we examined the morphology of mitochondria in corneal epithelial cells with a transmission electron microscope. Our result showed that the corneal mitochondria in the saline group were distinctly enlarged and distorted. Moreover, these effects induced by alkali burn were partially rescued (Figure 2e). These results suggest that mitochondrial damage may be a crucial pathological mechanism in corneal alkali burn-induced ferroptosis.

| Preparation and characterization of the drug-loaded liposomes
Although Fer-1 treatment exhibited improvement effects for corneal injury to some extent, its therapeutic efficacy is limited. In terms of F I G U R E 3 Self-assembly and characterization of Fer-1-loaded liposomes and their predicted effects on corneal cells after alkali burn injury. (a) Liposomes were prepared using a thin-film hydration method with Fer-1 encapsulated into the hydrophobic layer. When administrated to the alkali-burned cornea by eye drops, nanocarriers-encapsulated Fer-1 exhibit enhanced drug bioavailability than free Fer-1. After the internalization of Fer-1-NPs, the drugs are released inside the cell, where they exert their inhibitory effects of ferroptosis, inflammation, and neovascularization.   (n = 3). # The Dex-NPs-treated group is not included in the CNV measurements, H&E staining images, and central corneal thickness analyses since most corneas in this group presented perforation and thus were not applicable for analyses. Results were presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, and n.s. stands for not statistically significant. Significance was calculated by one-way analysis of variance. CNV, corneal neovascularization; Fer-1-NPs, ferrostatin-1-loaded liposomes This behavior of fast internalization with improved cellular uptake is extremely beneficial for the administration of eye drops which have a short period of retention time.
To further validate that liposome formulation improves drug efficacy, we studied their precorneal retention by in vivo imaging technology. Eyes were imaged after topical administration of free Nile Red or Nile Red-NPs. As shown in Figure 3f, the fluorescence intensity of the free Nile Red group rapidly decreased, with more than 50% reduction within 30 min (Figure 3g), verifying rapid and time-dependent precorneal elimination. Conversely, Nile Red-NPs possessed enhanced retention performance on the ocular surface. The fluorescence signal of the Nile Red-NPs group remained stronger than that of the free Nile Red group. About 78% of fluorescence was still retained even after 30 min. These results suggested that liposome was a promising drug delivery vehicle to prolong precorneal retention time, which was beneficial for drugs with more time to permeate the cornea.

| In vivo treatment effects of different drug formulations
We treated alkali-burned corneas with different drug formulations and investigated the therapeutic effects. As shown in Figure 4a These findings indicate that liposomes might provide a valid way to deliver Fer-1 into the alkali-burned cornea, thus alleviating inflammation and neovascularization more effectively.
Regarding the improved therapeutic effect of Fer-1-NPs, when encapsulated in liposomes, the hydrophobic Fer-1 was suspected to increase solubility and disperse in tears more adequately instead of being cleared quickly. The Dex and Dex-NPs groups served as positive controls because steroids were well-established clinically for the treatment of alkali burn. In our results, the Dex group showed a remarkable therapeutic effect on CNV length (0.67 mm) and CNV area (3.02 mm 2 ), with a mean clinical assessment score of 7.7, which was not significantly different from that in the Fer-1-NPs group. Similar results were found regarding the gross view and H&E-stained sections. However, unexpectedly, six of 10 mice in the Dex-NPs group presented corneal perforation on day 14. The mean clinical assessment score of the group was 9.8, indicating that the injuries were at least as severe as those in the saline group. The gross view also showed atrophy and synechia of the cornea (Figure 4e). We speculated that liposomal entrapment enhanced the bioavailability of Dex, thereby mimicking the long-term use of corticosteroids. Clinically, it could compromise the corneal integrity and increase the risk of corneal melting if the topical administration of corticosteroids lasts for 2 weeks after an alkali burn, 51 which is consistent with our results.
Moreover, the administration of Dex could also have unsatisfactory side effects, including secondary glaucoma, cataracts, and infectious keratitis. 52 Taken together, Dex could ameliorate alkali burn-induced corneal injury to an extent equivalent to that of Fer-1-NPs, but it also increases the risk of ocular complications. Hence, we excluded Dex and Dex-NPs therapy in our further experiments.

| Possible mechanisms of action of Fer-1-NPs on alkali burn
To examine the role that Fer-1-NPs play in the alkali burn-induced ferroptosis, corneal ROS stress production was directly detected by CellROX Green staining according to the previous study. 24 NADPH oxidase inhibitor DPI was added for comparison. We found that alkali burn treatment significantly increased corneal ROS stress; by contrast, DPI, Fer-1, and Fer-1-NPs all rescued the alkali burn-induced ROS stress in corneas, while with the best effect in the Fer-1-NPs group ( Figure S3). These results suggest that Fer-1-NPs may function as a more potent radical trap in the treatment of alkali-burned corneas.
To further clarify the mechanisms of Fer-1-NPs in the treatment of alkali-burned corneas via targeting ferroptosis, regulatory genes on the ferroptosis pathway were determined after Fer-1-NPs administration. As shown in Figure 5a, compared with Fer-1, the effect of Fer-1-NPs was more potent with respect to rescuing the mRNA expression of Ptgs2 and Acsl4. Consistent with the property of potent radical trap for Fer-1-NPs, the lipid peroxide 4-HNE was most significantly reduced in the Fer-1-NPs group (Figure 5b,c). Notably, in addition to the ungoverned protein levels of Gpx4, alkali burn caused a significant accumulation of Acsl4, both of which were almost fully recovered by Fer-1-NPs treatment (Figure 5b,c). Acsl4 is a crucial regulatory protein in lipid metabolism. It changes the membrane lipid composition by promoting the oxidation of polyunsaturated fatty acids, thereby sensitizing cells to ferroptosis. 53 Our results suggest that Acsl4 is involved in the process of corneal injury caused by ferroptosis, and Fer-1-NPs may improve corneal injury by inhibiting this pathway. Based on the results above, in alkali burn-induced corneal injury, the induction pathway of ferroptosis is activated, while the inhibition pathway is attenuated, resulting in ferroptotic cell death in the cornea. Moreover, we found that Fer-1-NPs have significantly greater therapeutic effects than free Fer-1 on restoring the expression of ferroptosis regulators and improving corneal injury caused by ferroptosis.
Taken together, these results demonstrate that Fer-1-NPs play an important role in the treatment of alkali burn-induced ferroptosis in corneas by effectively preventing corneal lipid peroxides production.
As for mitochondria morphology, although Fer-1 treated corneas were somewhat recovered as the shape was more regular and cristae were more abundant than those in the saline group, nevertheless, Fer-1-NPs treatment showed a further improvement in revising the mitochondrial damage. These results suggest that Fer-1-NPs may play a therapeutic role by remodeling mitochondrial morphology (Figure 5d  Results were presented as the mean ± SEM. *p < 0.05, **p < 0.01, and n.s. stands for not statistically significant. Significance was calculated by one-way analysis of variance. AOD, average optical density; Fer-1-NPs, ferrostatin-1-loaded liposomes; 4-HNE, 4-hydroxynonenal Inflammation is crucial in the pathological progress of corneal alkali burn. Once stimulated by alkali burn, inflammatory cells (e.g., white blood cells) and mesenchymal cells (e.g., macrophages, myofibroblasts) can be activated. In addition, the expression of a large number of cytokines, such as IL-1β, IL-6, and IL-10 in the cornea will be significantly increased, promoting inflammation and the formation of CNV. 54 Previous studies have suggested that anti-ROS therapy could effectively alleviate corneal inflammation and inhibit CNV. [24][25][26][27] F I G U R E 6 Fer-1-NPs exhibited more potent anti-inflammatory and anti-angiogenic effects than free Fer-1. Expression of the indicated inflammatory-related (a) and angiogenic-related (b) genes was measured using real-time polymerase chain reaction in corneal tissue (n = 3). Representative immunofluorescence images (c) and quantitative summary (d) of corneal sections stained with IL-1β (top) or IL-6 (bottom); scale bars, 50 μm; n = 3. Representative immunohistochemistry images (e) and quantitative summary (f) of corneal sections stained with α-Sma (top) or CD31 (bottom); scale bars, 50 μm; n = 3. Results were presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, and n.s. stands for not statistically significant. Significance was calculated by one-way analysis of variance. AOD, average optical density; Fer-1-NPs, ferrostatin-1-loaded liposomes To determine whether Fer-1-NPs can exert a therapeutic effect by inhibiting inflammation, we tested the mRNA expression levels of IL-1β and IL-6. As shown in Figure 6a, corneas treated with Fer-1-NPs significantly rescued the high expression of IL-1β (p < 0.001 vs. saline, p < 0.05 vs. Fer-1) and IL-6 (p < 0.01 vs. saline, p > 0.05 vs. Fer-1) in response to alkali-burned injury. Similar results were found for the IL-1β and IL-6 protein levels by immunofluorescence (Figure 6c,d). These results indicate that Fer-1-NPs could improve corneal injury by reducing the inflammatory response caused by alkali burn.
Neovascularization is another important pathological feature of corneal alkali burn. As an important factor affecting the visual quality of patients and the prognosis of corneal transplantation, CNV has become a topic of interest in the field of alkali burn. We observed that Fer-1-NPs could significantly inhibit the growth of CNV after alkali burn in mice. To explore the anti-angiogenesis mechanism of Fer-1-NPs, we examined the mRNA expression levels of neovascularization-related genes. As shown in Figure 6b, the Fer-1-NPs group had reduced expression of Vegfa (p < 0.01 vs. saline, p < 0.01 vs. Fer-1) compared to the alkali-burned groups. A significant decrease in Vegfc, Vegfr3, and Mmp2 mRNA levels was also found in corneas treated with Fer-1-NPs compared with those in the saline group. VEGFA could facilitate the migration of vascular endothelial cells and increase vascular permeability, which is vital to the growth of pathological neovascularization. 55,56 Moreover, VEGFC could promote the regeneration of lymphatic vessels by activating VEGFR3. 57 MMP-2 is known to provide growth space for CNV by degrading the extracellular matrix. 58 Meanwhile, μM Fer-1-NPs-treated groups were evaluated by slit-lamp examination, fluorescein sodium staining, and hematoxylin-eosin (H&E) staining on day 14; scale bars represent 1 mm for slit-lamp images (white) and 100 μm for H&E staining images (black). For pathological conditions (with alkali burn), in vivo biocompatibility of the indicated groups was evaluated by blood routine examination analyses and blood biochemistry (c) and H&E staining of main visceral organs (d); scale bars, 100 μm; n = 3. Results were presented as the mean ± SEM. ALT, alanine transferase; AST, aspartate transferase; BUN, blood urea nitrogen; CREA, creatinine; Fer-1-NPs, ferrostatin-1-loaded liposomes; HGB, hemoglobin; PLT, blood platelet; RBC, red blood cells; WBC, white blood cells pathological neovascularization is also closely related to corneal stroma fibrosis. Therefore, α-Sma (myofibroblasts marker) and CD31 (vascular endothelial cells marker) are generally considered tissue indicators of CNV. 21 Our immunohistochemical analyses show that, in the Fer-1-NPs group, the expressions of both α-Sma and CD31 were markedly downregulated compared to the other two groups (Figure 6e,f), suggesting the potent anti-CNV effect of Fer-1-NPs in vivo.
To further demonstrate the anti-angiogenesis effect of Fer-1-NPs, we carried out a VEGF-induced migration assay and a tube formation assay in vitro. Figures S5 and S6 show that HUVEC-GFP migration was induced by VEGF, with the maximum healing rate among all treatments.
However, a significantly decreased healing rate was induced by Fer-1 and Fer-1-NPs. Remarkably, the healing rate in the cells treated with Fer-1-NPs was notably less than that in the cells treated with Fer-1. Regarding the Matrigel tube formation assay, we observed a similar trend. The tube formation was strikingly decreased when cells were exposed to Fer-1 or Fer-1-NPs, as quantified by the number of junctions. In addition, Fer-1-NPs treatment led to a more significant reduction of tube junctions than that of Fer-1 treatment. The stronger inhibitory effect of Fer-1-NPs might be attributed to the enhanced intracellular uptake of liposomes, 59 which was also demonstrated in our cellular uptake experiment ( Figure 3e). Therefore, these results demonstrate that both Fer-1 and Fer-1-NPs could efficiently inhibit VEGF-induced angiogenesis, offering the important potential for the development of ferroptosistargeted anti-CNV treatment in the clinic.
The anti-angiogenesis effect by targeting ferroptosis warrants consideration. ROS was reported to act as second messengers in triggering inflammation through the activation of nuclear factor-κB (NF-κB), and subsequently stimulate the release of inflammatory cytokines, 60 the expression of VEGF, 61 and MMPs, 62 suggesting that the anti-angiogenesis effect of Fer-1 may be due to the reduced inflammation signaling from the relived ROS stress by this potent ferroptosis inhibitor. Moreover, increasing evidence supported that ferroptosis plays an important role as the trigger for initiating inflammation in the injury of liver, 63 intestine, 64 brain, 65 heart, 66 and renal, 67 which implicated in several signaling pathways including NF-κB, TGF-beta, Nrf2/HO1 and TLR. More pathway analyses are needed in further study to characterize the underlying mechanism of the anti-angiogenesis effect in alkali-burned injury by targeting ferroptosis.

| Biocompatibility evaluation of Fer-1-NPs
The cytotoxicity of the Fer-1-NPs was evaluated using calcein-AM/PI and CCK-8 assays (Figure 7a). Based on the live/dead cell staining, the majority of HCECs treated with Fer-1-NPs displayed green fluorescence. Accordingly, the CCK-8 assay showed that the cell viability still exceeded 90%, even when the concentration of Fer-1-NPs reached 20 μM, which indicated little inhibitory effect on cell proliferation.
The in vivo biocompatibility of Fer-1-NPs was evaluated by corneal stimulation assessment in normal mouse eyes (without alkali burn) treated with different formulations, followed by corneal examination using a slit-lamp microscope (Figure 7b). After administration of Fer-1-NPs for 14 consecutive days, there was no evidence of corneal opacity, CNV, inflammation, or hyperemia in the treated corneas. In addition, the integrity of the corneal epithelium was evaluated by fluorescein staining, which revealed no visible green staining, indicating that the corneal epithelium was intact without observable impairment during the Fer-1-NPs therapy.
Furthermore, H&E staining, which was used to examine corneal anatomical morphology, showed that the corneas in each group maintained a regular appearance, with a tight and orderly arrangement, but without inflammatory cells or CNV.
In addition to physiological conditions, we also investigated the toxicity of Fer-1-NPs in pathological conditions (with alkali burn). As shown in Figure 7c, the routine blood parameters as well as liver and renal function were within the normal ranges, without intergroup differences. Moreover, H&E staining of the main visceral organs (heart, liver, spleen, lung, and kidney) showed that the Fer-1-NPs treatment did not exhibit noticeable histological changes (Figure 7d). Taken together, no obvious toxic effects and excellent biocompatibility of Fer-1-NPs were verified, laying the foundation for clinical application.

| CONCLUSIONS
In this study, Fer-1-loaded liposomes, Fer-1-NPs were prepared via the film hydration method. The solubility, as well as the precorneal retention time of encapsulated Fer-1, could be improved, resulting in enhanced bioavailability. An improved therapeutic effect of alkali burn was achieved by Fer-1-NPs in the mouse model due to the inhibition of ferroptosis, inflammation, and neovascularization. Moreover, the suppression of angiogenesis by Fer-1-NPs was verified by a VEGF-induced endothelial cell migration assay and a tube formation assay. The Fer-1-NPs exhibited no evident cytotoxicity, ocular surface damage, or systemic toxicity.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.