Human umbilical vein endothelial cells (HUVECs; CTCC-003-0025) and human corneal epithelial cells (HCECs; cat. no. NCL-10051) were purchased from Mesien CTCC (Zhejiang Meisen Cell Technology Co., Ltd.). HUVECs were cultured in endothelial cell medium (ECM; cat. no. 1001; ScienCell Research Laboratories, Inc.) supplemented with 10% fetal bovine serum (FBS; cat. no. A52556701; Gibco; Thermo Fisher Scientific, Inc.) and 1% penicillin-streptomycin (cat. no. P1400; Beijing Solarbio Science & Technology Co., Ltd.). HCECs were maintained in keratinocyte serum-free medium (Keratinocyte-SFM; Invitrogen; Thermo Fisher Scientific, Inc.) supplemented with 10% FBS, 5 ng/ml epidermal growth factor (cat. no. HY-P7109; MedChemExpress), 0.005 mg/ml insulin (cat. no. 11070-73-8; Merck KGaA), 0.05 mg/ml bovine pituitary extract, 500 ng/ml hydrocortisone and 1% penicillin-streptomycin. All cells were incubated at 37°C in a humidified 5% CO₂ atmosphere. The PI3K inhibitor, LY294002 (cat. no. HY-10108; MedChemExpress), was applied to pretreat cells at a final concentration of 10 µM for 2 h before subsequent NaOH and GA treatment to block PI3K/AKT signaling (18).

Cell viability/proliferation was assessed using a Cell Counting Kit-8 (CCK-8) (cat. no. CK-04; Dojindo Laboratories, Inc.). Briefly, HUVECs were seeded into 96-well plates at 5×10³ cells/well and treated with VEGF-A or VEGF-A plus GA for 48 h, followed by the addition of 10 µl CCK-8 reagent/well and incubation for 1–2 h. Optical density (OD) was measured at 450 nm. HCECs were exposed to 6 mM NaOH for 50 sec, washed with phosphate-buffered saline (PBS; cat. no. BL302A; Biosharp Life Sciences), cultured for 24 h and analyzed with a CCK-8, as aforementioned for HUVECs.

Cells and tissues were lysed using RIPA buffer (cat. no. P0013B; Beyotime Biotechnology) supplemented with protease and phosphatase inhibitors. Protein concentrations were determined using a bicinchoninic acid assay kit (cat. no. 23225; Thermo Fisher Scientific, Inc.). After protein quantification, 10–20 µg total protein/lane was loaded, mixed with 5X loading buffer, heated at 95°C for 5 min and subsequently separated via 10% SDS-PAGE). Proteins were then transferred to polyvinylidene fluoride membranes (cat. no. IPVH00010; MilliporeSigma). After blocking with 5% nonfat milk (1 h, 23–25°C), the membranes were incubated overnight at 4°C with primary antibodies (Table I). After washing with PBS containing 0.1% Tween 20, the membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (Bioss, detailed catalogue number and dilution ratio listed in Table I) for 2 h at room temperature. β-actin was used as the reference protein and loading control for normalization. The signals were visualized using an enhanced chemiluminescence system (cat. no. 34096; Thermo Fisher Scientific, Inc.) and imaged with an Amersham imaging platform (GE Healthcare). Band intensities were quantified using ImageLab software (v.6.1; Bio-Rad Laboratories, Inc.).

Cell migration was assessed using Transwell chambers (8-µm pore size). HUVECs pretreated for 48 h with VEGF-A alone or in combination with GA were resuspended in serum-free DMEM and seeded into the upper chambers at 2×10⁴ cells/well. The lower chambers received 800 µl complete medium (containing 10% FBS) as the chemoattractant. After a 48-h incubation, the migrated cells were fixed with 1% paraformaldehyde (PFA) at room temperature (for 20 min, stained with Giemsa at room temperature for 15 min and quantified via light microscopy.

HCECs seeded on coverslips in 24-well plates at a density of 2×10⁴ cells/well. Cells were incubated with 10 µM EdU at 37°C for 2 h, fixed in 4% paraformaldehyde 15 min at room temperature, permeabilized with 0.3% Triton X-100 in PBS for 10 min at room temperature, then reacted with Click-iT reaction solution for 30 min at room temperature in the dark. Nuclei were counterstained with 1X Hoechst 33342 for 10 min at room temperature. Fluorescence images were acquired using a fluorescence microscope. For quantitative analysis, at least five random fields of view were captured for each group, and EdU-positive cell ratio was quantified with ImageJ software (v1.53; National Institutes of Health). The assay was performed using the BeyoClick EdU-594 Cell Proliferation Detection Kit (C0078S, Beyotime, China) following the manufacturer's instructions.

HUVECs (2×10³ cells/well) were seeded into 96-well plates precoated with Matrigel (cat. no. 354234; Corning Inc.) and incubated in a medium containing VEGF-A or VEGF-A plus GA at 37°C for 12 h to induce capillary-like structures. Images were captured with a Nikon TS2 inverted microscope (Nikon Corporation). Tube formation was analyzed by measuring length and branch points using ImageJ software (v.1.53; National Institutes of Health).

Apoptosis was evaluated using the FITC Annexin V Apoptosis Detection Kit I (cat. no. 556547; BD Biosciences). Cells were washed twice with PBS, resuspended in 100 µl of 1X binding buffer and stained with 5 µl each of FITC Annexin V and propidium iodide at room temperature for 15 min in the dark, following the manufacturer's instructions. Subsequently, 400 µl binding buffer was added and the samples immediately subjected to flow cytometric analysis using BD FACSCanto II flow cytometer (BD Biosciences). The data were analyzed using FlowJo software (v.10.8; BD Biosciences). The apoptotic rate was calculated as the sum of the percentage of early (Annexin V⁺/PI⁻, Q1-LR) and late apoptotic cells (Annexin V⁺/PI⁺, Q1-UR).

All animal procedures were reviewed and approved by the Animal Ethics Committee of the Wuxi Affiliated Hospital of Nanjing University of Chinese Medicine (Wuxi, China; approval no. 2024-171-01) and complied with Chinese regulations on animal welfare and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. C57BL/6J mice (male, 8 weeks old; weight 20–22 g) were purchased from GemPharmatech Co., Ltd. A total of 120 mice were randomly assigned to six groups (n=20 per group) for this in vivo study. All animals were housed and maintained under standard conditions (20–26°C; 30–70% humidity; 12-h light/dark cycle) with food and water provided ad libitum. Following acclimation, an alkali burn-induced corneal injury model was established as previously described (19,20). The mice were anesthetized by intraperitoneal injections of ketamine (80 mg/kg) and xylazine (16 mg/kg), followed by a topical application of 0.5% proparacaine to the right eye. Subsequently, a 2-mm-diameter circular filter paper soaked in 1 N NaOH (cat. no. 484024; Merck KGaA) was placed on the central cornea of the right eye for 30 sec. The ocular surface was then thoroughly irrigated with PBS for 1 min to remove residual alkali. On day 14 after injury, the mice were anesthetized by intraperitoneal administration of ketamine (80 mg/kg) and xylazine (16 mg/kg) and then sacrificed by gradual-fill CO₂ inhalation at a displacement rate of 20% of the chamber volume per minute. CO₂ exposure was continued for 5 min. Following CO₂ exposure, the mice were continuously observed for at least 3 min, during which the absence of spontaneous respiration and heartbeat was confirmed. The pedal withdrawal reflex and the corneal reflex of the unaffected eye were subsequently assessed. Cervical dislocation was then performed as a secondary physical method to ensure death before eyeball enucleation.

A total of 120 mice were randomly assigned to six groups (n=20/group): normal, PBS, low-dose GA (0.5 mg/ml), medium-dose GA (1 mg/ml), high-dose GA (2 mg/ml) and positive control (1 mg/ml Dex) groups. Mice were divided into two batches. The first batch (n=8/group) was used for in vivo clinical evaluation and immunohistochemistry (IHC) and the second batch (n=12/group) sacrificed for other histological and molecular analyses, with a total of 20 mice/group. The normal group was not subjected to alkali burn-induced corneal injury and only treated with PBS. All other groups were subjected to alkali burn-induced corneal injury and treatments. Each eye received 10 µl eye drops 4 times daily for 14 consecutive days. Corneal lesions were examined under a slit-lamp microscope on days 3, 7 and 14 after injury (day 0=model induction) and scored for corneal opacity, CNV extent and neovascular density, as shown in Table II (7).

On day 14, mice were humanely sacrificed and their right eyeballs enucleated and fixed in 4% PFA at 4°C overnight. Tissues were washed with PBS, dehydrated through graded ethanol, cleared, paraffin-embedded and sectioned at a thickness of 4 µm. Tissue sections were stained with hematoxylin-eosin (H&E) at room temperature for 5 (hematoxylin) and 2 min (eosin) for histological evaluation and IHC for protein expression analysis. For IHC, the paraffin-embedded sections were deparaffinized, rehydrated and permeabilized with 0.1% Triton X-100 for 15 min. Nonspecific binding was blocked using 10% goat serum (Wuhan Servicebio Technology Co., Ltd.) for 2 h at 25°C. To quench endogenous peroxidase activity, the sections were treated with 3% H₂O₂ in methanol for 15 min at room temperature and subsequently incubated with a primary antibody against VEGF-A (1:200; cat. no. AF5131; Affinity Biosciences) at 4°C for 12 h. After washing with PBS, the sections were incubated with an HRP-conjugated secondary antibody (1:200; cat. no. GB23301; Wuhan Servicebio Technology Co., Ltd.) for 60 min at 37°C. Immunoreactive signals were visualized using DAB. Sections were counterstained with hematoxylin at room temperature for 2 min. Microscopy was performed using a light microscope (Nikon Corporation). The central corneal thickness and average OD of the IHC signals were assessed and quantified using ImageJ (v.1.53; National Institutes of Health).

Mice were sacrificed on day 14 post-alkali burn and their eyes enucleated via blunt dissection. Briefly, the eyelids were retracted using the thumb and index finger, followed by the placement of forceps beneath the globe. The forceps were then closed and used to gently extract the eyeball from the orbit. The enucleated eyes were fixed in 4% PFA at 4°C overnight. After PBS washes, the corneas were transferred to a primary antibody solution and incubated at 4°C for three days. The solution was prepared in 1X PBS containing 1% bovine serum albumin, 0.1% Triton X-100 and the following antibodies: Dylight 550-conjugated anti-PECAM-1/CD31 (1:100, cat. no. AF6191; Affinity Biosciences) and Alexa Fluor 488-conjugated anti-LYVE-1 (1:100, cat. no. AF4202; Affinity Biosciences). Following incubation, the corneas were washed three times with 1× PBS for 20 min each. Whole-mount images were captured using a laser confocal microscope (LSM700; Carl Zeiss AG) and the fluorescence intensities of CD31 and LYVE-1 quantified using ImageJ (v.1.53; National Institutes of Health). CD31 is highly expressed on the vascular endothelial cells of nascent and mature blood vessels, whereas LYVE-1 labels lymphatic endothelial cells and is used to assess lymphangiogenesis (15).

For protein extraction from corneal tissues, the aforementioned western blotting workflow was adopted. Briefly, harvested corneal tissues were fully homogenized in RIPA lysis buffer supplemented with protease and phosphatase inhibitors prior to protein quantification. GAPDH was selected as the loading control for corneal tissue lysates, whereas β-actin served as the internal reference for cell samples. The procedures of SDS-PAGE, membrane transfer, antibody incubation, ECL detection and band intensity quantification were consistent between cells and tissues. Detailed information of all primary antibodies is summarized in Table I.

Following extraction with TRIzol^® reagent (cat. no. 15596018; Thermo Fisher Scientific, Inc.), RNA quality was verified using a NanoDrop spectrophotometer (A260/A280; NanoDrop Technologies; Thermo Fisher Scientific, Inc.). A reverse transcription kit (cat. no. R433; Vazyme Biotech Co., Ltd.) was used to synthesize cDNA from total RNA with incubation at 50°C for 5 min and inactivation at 85°C for 5 sec. For qPCR, reactions containing SYBR Green Master Mix (cat. no. Q711; Vazyme Biotech Co., Ltd.) were run on a CFX96 real-time PCR system (Bio-Rad Laboratories, Inc.) under the following conditions: 95°C for 30 sec, then 40 cycles of 95°C for 10 sec and 60°C for 30 sec, followed by melting curve analysis. Relative expression (2^−ΔΔCq) was calculated using GAPDH as the internal control. The primer sequences used for RT-qPCR analysis were as follows: mouse BAX (NM-007527), F: 5′-AGGATGCGTCCACCAAGAAGCT-3′ and R: 5′-TCCGTGTCCACGTCAGCAATCA-3′; mouse BCL2 (NM-009741), F: 5′-CCTGTGGATGACTGAGTACCTG-3′ and R: 5′-AGCCAGGAGAAATCAAACAGAGG-3′; mouse caspase-3 (NM-009810), F: 5′-GGAGTCTGACTGGAAAGCCGAA-3′ and R: 5′-CTTCTGGCAAGCCATCTCCTCA-3′; mouse-GAPDH (NM-008084), F: 5′-CATCACTGCCACCCAGAAGACTG-3′ and R: 5′-ATGCCAGTGAGCTTCCCGTTCAG-3′.

DIA-based quantitative proteomics was conducted to compare the protein expression profiles of corneal tissues before and after GA treatment. Tissue samples were lysed and digested using an iST sample preparation kit (PreOmics), followed by high-pH reversed-phase fractionation to generate the spectral library. Liquid chromatography-tandem mass spectrometry was performed using an EASY-nLC 1200 system coupled to an Orbitrap Lumos mass spectrometer (Thermo Fisher Scientific, Inc.) equipped with an Acclaim PepMap C18 analytical column (75 µm ×25 cm). Peptides were separated over a 120 min linear gradient of 5–35% solvent B (0.1% formic acid in acetonitrile) at a flow rate of 200 nl/min. Data were acquired in positive ion mode via nano-electrospray ionization (nano-ESI), with an electrospray voltage of 2 kV. The ion transfer capillary temperature was set to 320°C. Raw data were processed against the UniProt database using Spectronaut X (Biognosys AG). Identification thresholds were set at a 1% false discovery rate at both the peptide and protein levels. Differentially expressed proteins (DEPs) were defined by a fold change >1.5 and P-value <0.05. Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses were conducted to identify key pathways. This high-throughput DIA-based proteomic analysis was performed as an exploratory approach to identify DEPs and pathways associated with GA treatment.

Levels of TNF-α (cat. no. BMS607-3TEN), IL-6 (cat. no. BMS603-2) and IL-17 (cat. no. BMS6001) were quantified using commercial ELISA kits (Thermo Fisher Scientific, Inc.), according to the manufacturer's instructions. Samples were diluted 1:5 and added to antibody-precoated 96-well plates (100 µl/well) in duplicate, followed by 2 h incubation at room temperature. After washing, the plates were sequentially incubated with Biotin-conjugated detection antibodies and streptavidin-HRP supplied with the kits, each for 2 and 1 h at room temperature, respectively. Color development was achieved using tetramethylbenzidine substrate for 30 min in the dark; the reaction was terminated with stop solution (1 M phosphoric acid) and absorbance measured at 450 nm. Cytokine concentrations were interpolated by comparison with a standard curve generated using the supplied standards.

The Simplified Molecular Input Line Entry System and Structure Data Format structures of GA were retrieved from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/). Potential targets were screened using SwissTargetPrediction (https://swisstargetprediction.ch/) and disease-related targets for ‘corneal alkali burn’ obtained from the GeneCards database (https://www.genecards.org/). The intersection of GA and disease targets was computed in R (v.4.2.1; https://www.r-project.org; The R Foundation, Vienna, Austria). GO and KEGG enrichment analyses were conducted using Bioconductor packages (http://bioconductor.org/packages/limma) to construct a compound-target-pathway-disease network. PubChem and the Protein Data Bank (rcsb.org/) were used to download three-dimensional structures of the compounds and target proteins. Protein structures were preprocessed (such as removal of solvent molecules) with PyMOL (v.2.6.0; https://www.pymol.org; DeLano Scientific), followed by hydrogenation and charge assignment in AutoDock Tools (v.1.5.7; Scripps Research). Docking was performed using AutoDock Vina and visualization achieved with PyMOL and Discovery Studio (BIOVIA).

All experiments were repeated thrice. Data are presented as mean ± SD. Statistical analyses were conducted using GraphPad Prism 10 (Dotmatics). The Shapiro-Wilk test was used to assess data normality and Levene's test applied to evaluate the homogeneity of variances. If these assumptions were met, one-way analysis of variance followed by Tukey's post hoc test was performed. P<0.05 was considered to indicate a statistically significant difference.

To evaluate the in vivo effects of GA on corneal inflammation and neovascularization, a mouse corneal alkali-burn model was established and administered different concentrations of GA, Dex, or PBS. Slit-lamp microscopy was performed on days 3, 7 and 14 after injury to monitor corneal morphological changes (Fig. 1A). In the PBS group, extensive CNV was observed by day 14, covering nearly the entire cornea, accompanied by severe opacity and vascular congestion in the central pupillary area (Fig. 1B). By contrast, both Dex and GA treatments markedly improved corneal pathology (Fig. 1C). Notably, the high-dose GA group (2 mg/ml) showed a therapeutic efficacy comparable to that of Dex on day 14 (scores, 4.125 vs. 3.125). Gross observations further confirmed these findings: corneal transparency was substantially restored in GA-treated mice, with background details clearly discernible (Fig. 1D). Histological examination supported these observations. H&E staining revealed a significant reduction in inflammatory-cell infiltration within the corneal stroma and restoration of epithelial thickness in GA-treated mice compared with those in the model group, suggesting GA exhibits both anti-inflammatory and wound-healing properties.

To further elucidate the anti-inflammatory mechanism of GA, inflammatory cytokine levels in corneal tissues were analyzed. ELISA demonstrated that GA markedly suppressed IL-6, IL-17 and TNF-α expression compared with that in the PBS-treated group (Fig. 1E), indicating that GA alleviated corneal inflammation by inhibiting pro-inflammatory cytokine production. Alkali injury induces pathological corneal angiogenesis and lymphangiogenesis, which are closely associated with elevated inflammatory cytokines (21). Immunofluorescence staining showed that vascular (18.05%) and lymphatic vessel (17.20%) densities in GA-treated corneas were markedly reduced compared with those in the PBS group (31.27 and 33.89%, respectively; Fig. 1F), further confirming that GA effectively inhibits pathological-blood and lymphatic-vessel formation. The present findings highlighted that GA ameliorates alkali burn-induced corneal injury by suppressing inflammatory cytokine expression and pathological angiogenesis, thereby alleviating corneal inflammation and promoting tissue recovery.

To elucidate the role of GA in angiogenesis, the present study systematically evaluated its effects on VEGF-A-induced proliferation, migration and tube formation in HUVECs. Initially, GA cytotoxicity toward HUVECs was assessed using the CCK-8 assay (Fig. 2B). GA showed no significant effect on cell viability within the concentration range of 0–40 µg/ml. However, a slight inhibitory effect was observed at 80 µg/ml and marked cytotoxicity occurred at 160 µg/ml (P<0.05). Accordingly, 80 µg/ml was selected as the working concentration for subsequent experiments. Further results demonstrated that 80 µg/ml GA markedly suppressed VEGF-A-induced HUVEC proliferation, exhibiting an inhibitory effect comparable to that of Dex (Fig. 2C).

Endothelial cell migration is a critical step in angiogenesis (22). Transwell migration assays were employed to evaluate the effect of GA on VEGF-A-stimulated HUVEC migration (Fig. 2D-E). VEGF-A treatment markedly enhanced endothelial cell migration, whereas GA treatment markedly reduced the number of migrated cells. Quantitative analysis confirmed that GA intervention substantially attenuated VEGF-A-induced migratory capacity. The formation of capillary-like tubular networks by endothelial cells is another key phase of angiogenesis (23,24). To further examine the anti-angiogenic potential of GA, Matrigel tube-formation assays were performed (Fig. 2F-H). VEGF-A markedly promoted the formation of complex, highly branched tubular structures, whereas GA treatment markedly reduced tube length, branch points and total lumen numbers, indicating a potent inhibition of angiogenesis. These results indicate that GA effectively suppresses VEGF-A-mediated HUVEC proliferation, migration and tube formation, suggesting that it exerts considerable anti-angiogenic activity in vitro.

To investigate the regulatory effect of GA on VEGF-A expression, experiments were performed in HUVECs and a mouse corneal alkali-burn model. Western blotting revealed that GA treatment markedly decreased VEGF-A protein levels in HUVECs compared with that in the control group (Fig. 3A), indicating that GA directly suppresses VEGF-A upregulation in endothelial cells.

Furthermore, IHC and western blotting were performed to examine VEGF-A expression in alkali-burned corneal tissues. IHC results showed that corneas in the PBS group exhibited marked stromal edema and thickening, accompanied by strong VEGF-A immunoreactivity. By contrast, treatment with 2 mg/ml GA markedly reduced VEGF-A expression, with the corneal structure largely preserved (Fig. 3B). Consistently, western blotting analysis confirmed that GA effectively reversed the alkali burn-induced upregulation of VEGF-A expression in corneal tissues (Fig. 3C).

Collectively, these findings demonstrated that GA exerted anti-angiogenic effects in vivo and in vitro by downregulating VEGF-A expression, thereby inhibiting CNV.

To further elucidate the potential molecular mechanisms by which GA exerts its therapeutic effects in alkali-burned corneal tissues, DIA-based quantitative proteomics was performed on corneal samples collected on day 7 after injury. Differential protein expression profiles were compared between the PBS and high-dose GA groups (2 mg/ml). In total, 779 DEPs were identified, including 638 upregulated and 141 downregulated proteins (Fig. 4A). Distribution and expression patterns of the top 20 DEPs are presented in Fig. 4B. GO functional enrichment analysis (Fig. 4C) indicated that these DEPs were mainly enriched in 26 biological process, 19 molecular function and 2 cellular component categories. KEGG pathway enrichment analysis (Fig. 4D) revealed that the top five markedly enriched pathways included ‘metabolic pathways’, ‘complement and coagulation cascades’, ‘pathways in cancer’, the ‘PI3K/AKT signaling pathway’ and ‘human papillomavirus infection’.

Given that corneal alkali burn is typically accompanied by severe inflammation, pathological neovascularization, corneal ulceration and loss of transparency and based on previous studies (25–28), the PI3K/AKT pathway serves a key role in regulating inflammatory responses and tissue repair. Proteomic analysis showed increased abundance of FAK-and PI3K-related proteins in GA-treated corneas, indicating that the FAK/PI3K/AKT pathway may contribute to GA-mediated corneal repair.

Therefore, both the corneal alkali-burn model and NaOH-induced HCECs were employed in subsequent experiments to further validate the regulatory effects of GA on the PI3K/AKT signaling pathway in vivo and in vitro, aiming to clarify the molecular mechanisms underlying corneal repair and anti-angiogenic activity.

Network pharmacology has emerged as a powerful approach for identifying potential targets and elucidating the signaling pathways of bioactive compounds in traditional Chinese medicine (29,30). To further investigate the molecular mechanisms by which GA regulates CNV following alkali burn injury, a network pharmacology-based strategy was applied to screen and analyze potential targets and signaling pathways associated with GA.

Potential targets of GA were predicted using multiple databases, including BATMAN (probability ≥50%), PharmMapper (score ≥0.5), TargetNet (score ≥0.5) and SwissTargetPrediction (probability >0). After removing duplicate entries, 389 candidate targets were identified (Fig. 5A). Disease-related targets associated with ‘corneal alkali burn’ were retrieved from the GeneCards database (score >1), yielding 244 disease targets. The intersection of the two datasets identified 43 potential overlapping targets that may mediate the therapeutic effects of GA on corneal alkali burn (Fig. 5C). KEGG pathway enrichment analysis of these 43 targets revealed that the top 30 markedly enriched pathways included the PI3K/AKT signaling pathway, consistent with previous proteomic findings (Fig. 5B). Based on these results, a compound-target-pathway-disease network was constructed using the core pathways, key targets and bioactive compound (Fig. 5D). The network comprised 75 nodes and 241 edges, representing 43 targets, 30 signaling pathways and one compound. The color intensity of the nodes corresponded to their degree values, where higher degree values indicate greater importance within the network.

Finally, molecular docking analysis was performed to evaluate the binding affinity between GA and the core targets associated with corneal alkali burn (Fig. 5E). The binding energy was used to estimate the stability and strength of the ligand-protein interactions. Generally, a binding energy ≤-5.0 kcal/mol indicates good binding activity, whereas a binding energy ≤-7.5 kcal/mol reflects strong binding affinity. The visualized docking interactions are presented in Fig. 5F.

The present study intersected the DEPs identified through quantitative proteomics with candidate targets predicted via network pharmacology and the two shared core positive proteins, albumin (ALB) and plasminogen (PLG), were screened for cellular validation using western blotting assays. The western blotting results confirmed that protein levels of ALB and PLG in HCECs were markedly upregulated in the GA-treated group compared with those in the alkali-injured model group, which validated the reliability of the DEPs screened through proteomics and the target prediction results of network pharmacology. Mechanistically, ALB neutralizes toxic oxidized lipids, eliminates reactive oxygen species and stabilizes epithelial junctions to suppress inflammatory infiltration. PLG participates in extracellular matrix remodeling and facilitates epithelial cell migration, thereby accelerating the healing of alkali-induced corneal wounds. Moderate upregulation of PLG expression by GA maintains physiological fibrinolytic balance and avoids excessive degradation of the corneal stroma. Collectively, ALB and PLG synergistically restore the activity of the protective FAK/PI3K/AKT signaling cascade, thereby alleviating alkali-triggered corneal injury and restraining pathological CNV. The western blotting verification data for ALB and PLG are presented in Fig. S1.

To clarify the mechanism underlying how GA attenuates corneal epithelial-cell apoptosis, we assessed the expression of key apoptosis regulators (Bax, Bcl-2 and caspase-3). Alkali burn markedly increased pro-apoptotic Bax and caspase-3 expression levels while decreasing those of anti-apoptotic Bcl-2 compared with the expression levels measured in the normal control (Fig. 6A and B). GA treatment effectively reversed these changes, with the most potent anti-apoptotic effect observed at 2 mg/ml.

Collectively, these results suggested that GA exerts a protective effect against alkali burn-induced corneal epithelial injury by modulating apoptosis-related signaling pathways and inhibiting excessive cell apoptosis.

To evaluate the effects of GA on HCEC viability, its cytotoxicity at different concentrations was initially determined using the CCK-8 assay. The results showed that GA had no significant effect on cell viability within the concentration range of 0–200 µg/ml; however, a notable decrease in cell viability was observed at 400 µg/ml (P<0.05; Fig. 7A). Subsequently, HCECs were exposed to 6 mM NaOH for 50 sec to establish a corneal injury model and then co-incubated with GA (0–300 µg/ml) to determine the optimal treatment concentration. The results indicate that 250 µg/ml GA provided the most pronounced protective and proliferative effects on NaOH-damaged cells; hence, this concentration was chosen for subsequent experiments (Fig. 7B). Transwell migration and EdU proliferation assays revealed that GA markedly enhanced the migratory and proliferative ability of HCECs, whereas these effects were markedly attenuated by co-treatment with a PI3K inhibitor (Fig. 7C and D). In addition, flow cytometric analysis demonstrated that GA effectively reduced NaOH-induced apoptosis in HCECs, whereas the inhibition of PI3K reversed this effect and increased the apoptotic rate (Fig. 7E).

To clarify whether the anti-neovascularization effect of GA relies on the FAK/PI3K/AKT signaling pathway and identify the upstream-downstream order of VEGF-A and PI3K, the present study performed western blotting-based PI3K inhibitor (LY294002) rescue assays in alkali-stimulated HCECs. Consistent with previous data, GA intervention elevated phosphorylated FAK, PI3K and AKT levels, whereas co-treatment with LY294002 markedly abolished the activation of PI3K/AKT without altering VEGF-A expression. These results indicated that LY294002 blocks downstream PI3K/AKT signaling but does not reverse GA-mediated VEGF-A suppression, suggesting that the inhibitory effect of GA on VEGF-A is largely independent of PI3K/AKT activation.

Combined with the functional data showing that GA inhibits VEGF-A-induced angiogenesis in HUVECs, the present study confirmed that GA exerts anti-CNV effects by primarily repressing VEGF-A production and that the FAK/PI3K/AKT pathway acts as a downstream effector mediating epithelial-protective functions rather than as an upstream regulator of VEGF-A (Fig. 8A-B). Moreover, GA treatment markedly increased FAK, PI3K and AKT phosphorylation, with a concomitant upregulation of the downstream targets, Bcl-2 and cyclin D1. By contrast, co-treatment with the PI3K inhibitor markedly attenuated these effects. (Fig. 8C-I).

Overall, GA promotes the proliferation and migration of NaOH-injured HCECs and inhibits apoptosis by activating the FAK/PI3K/AKT pathway, thereby exerting cytoprotective effects during corneal injury repair.