Glaucoma has emerged as the leading cause of irreversible blindness worldwide. Among its various forms, primary open-angle glaucoma (POAG) is particularly notable as a complex disease characterized by optic nerve atrophy and progressive visual field loss. Elevated intraocular pressure (IOP), resulting from increased resistance to aqueous humor outflow, has been identified as the primary risk factor contributing to the onset of POAG and its progression toward blindness.1 The dysfunction of trabecular meshwork cells (TMCs), including the imbalance between extracellular matrix (ECM) synthesis and degradation, plays a crucial role in this increased outflow resistance; however, the precise pathophysiological mechanisms remain unclear.2
Oxidative stress has been established as a critical factor in trabecular meshwork dysfunction and damage to retinal ganglion cells (RGCs) and the optic nerve.3 The insufficiency of antioxidant mechanisms within TMCs leads to the accumulation of reactive oxygen species (ROS), fostering chronic inflammatory infiltration, TMC apoptosis, rearrangement, and, ultimately, elevated IOP.4,5 When ROS interacts with membrane-rich polyunsaturated fatty acids, lipid peroxidation (LPO) occurs, primarily affecting plasma and organelle membranes.6 This LPO serves as a cell death signal, capable of inducing regulated cell death pathways, such as ferroptosis. Notably, primary, secondary, and final LPO products have been found to accumulate within trabecular meshwork tissue.7 In patients with glaucoma, increased expression levels of LPO markers, including 4-hydroxy-2-nonenal (4-HNE) and malondialdehyde (MDA), have been identified in both trabecular meshwork tissue and aqueous humor.8 Furthermore, genes associated with LPO, such as cytochrome P450 family member CyP1B1 and thioredoxin reductase 2 (TXNRD2), are implicated in trabecular meshwork injury during glaucoma.9,10 These findings collectively underscore the relationship between oxidative stress, disruptions in lipid metabolism , and the pathogenesis of glaucoma.
Ferroptosis, a novel form of regulated cell death, results from the accumulation of iron-dependent LPO. During this process, the homeostasis of iron metabolism is disrupted, leading to increased levels of intracellular free iron and the excessive production of lipid ROS via iron-dependent oxidase. This accumulation of LPO ultimately results in the destruction of plasma membrane.11 Various studies have established ferroptosis index scoring systems that utilize ferroptosis-related gene expression signatures to predict diagnosis and prognosis across a range of diseases.12–14 However, the specific patterns of ferroptosis-related gene expression in the pathogenesis of POAG remain uncharted.
Prior research has demonstrated that glutamate receptor–mediated ferroptosis of RGCs significantly contributes to glaucoma-induced blindness, with iron-chelating agents markedly mitigating neuronal injury.15 Ajay Ashok et al. have shown that in TGF-β2–induced TMCs, hepcidin orchestrates a self-sustaining feedforward loop through iron-catalyzed ROS. Notably, hepcidin antagonists and antioxidants can partially inhibit this detrimental TGFβ2-hepcidin loop.16 A previous study also revealed that patients with glaucoma exhibit significantly higher serum iron levels compared to the normal population.17 Recently, miR-93 and miR-141 have been implicated in the pathogenesis of POAG by targeting Nrf2, a crucial regulator of ferroptosis.18 By modulating ferroptosis, Nrf2 may contribute to trabecular meshwork injury. Additionally, ferroptosis has been implicated in ECM remodeling in pulmonary and liver fibrosis, where the inhibition of iron accumulation and ferroptosis has been shown to attenuate fibrosis models.19,20 These studies suggest a potential role for ferroptosis in trabecular meshwork injury in glaucoma, indicating that targeting TMC ferroptosis could be an effective strategy for reducing IOP.
Epidemiologic investigations have identified environmental metal exposure (e.g., cadmium) as potential risk modifiers for POAG. However, critical knowledge gaps persist in two domains.21 First, the causal relationship between specific metallic elements and the pathogenesis of POAG remains unsubstantiated by genetic evidence, as current observational studies are unable to disentangle confounding factors from true biological effects. Additionally, emerging mechanistic studies suggest that ferroptosis may constitute a pivotal pathway linking metal homeostasis to retinal ganglion cell degeneration.
Notably, multiple transition metals (e.g., Fe²⁺/Cu⁺) serve as catalytic centers in Fenton reactions, which propagate oxidative damage to RGCs, while selenoproteins such as GPX4 function as metalloenzymes exerting ferroptosis-suppressive roles.22 However, the triadic interaction between environmental metal load, genetic susceptibility loci, and ferroptotic vulnerability in trabecular meshwork cells has yet to be systematically investigated through gene–environment interaction analyses.
This study innovatively integrates multiomics strategies to elucidate the critical role of iron metabolism dysregulation in POAG-associated TM injury. We developed a translational research pipeline encompassing: (1) Mendelian randomization analysis in East Asian populations to prioritize metal–POAG associations, (2) cross-omics integration of TM transcriptomes and ferroptosis signatures to identify regulatory networks, (3) computational drug repositioning via CMap database mining and molecular docking simulations, and (4) hierarchical experimental validation using oxidative stress models and ocular hypertension systems. By bridging environmental exposure analysis with regulated cell death mechanisms, this work pioneers a “metal–ferroptosis–TM injury” paradigm, offering both methodological innovation for pathway-centric ophthalmic research and preclinical foundations for mechanistically driven glaucoma therapeutics.
Methods
Ethical Statement
All experiments involving human participants in this study received approval from the Ethics Committee of the Changsha Aier Eye Hospital (No. KYPJ012), in accordance with the 1964 Declaration of Helsinki and its subsequent amendments, as well as comparable ethical standards.
All animal protocols adhered to the experimental animal welfare ethics guidelines established by the Chinese Association for Laboratory Animal Sciences. Animal care and experimental procedures were approved by the Animal Ethical and Welfare Committee of Beijing Gene Line Bioscience Co., Ltd (No. JL-ZXY-20230101).
Mendelian Randomization
This study employed a two-sample Mendelian randomization (MR) approach to systematically investigate the causal relationship between blood metal levels and POAG. The exposure data incorporated genome-wide association study (GWAS) summary statistics from East Asian populations, encompassing genetic association data for 42 blood metal elements.23 The outcome data for POAG were sourced from BioBank Japan.24,25 The study first identified genetic instruments (single-nucleotide polymorphisms [SNPs]) significantly associated with blood metal levels, adhering to stringent criteria, including genome-wide significance thresholds of P < 1 × 10−5, as employed in the original GWAS by Yang et al.23 to ensure sufficient instrumental variables. Linkage disequilibrium filtering (r2 < 0.001) and instrumental variable strength (F-statistic >10) were also applied.
SNPs associated with potential confounders, such as IOP, retinal nerve fiber layer thickness, and cup-to-disc ratio, were excluded through phenotype scanning. To enhance robustness, the study utilized inverse-variance weighted, MR-Egger regression, and weighted median methods for primary effect estimation, alongside sensitivity analyses such as leave-one-out testing and MR-Egger regression tests, ensuring rigorous control of horizontal pleiotropy and heterogeneity.
ELISA
The levels of MDA (Jianglai Biology, Shanghai; JL-11466 for human, JL13297 for rat) and glutathione peroxidase 4 (GPX4) (Jianglai Biology; JL46163 for human, JL48671 for rat) in aqueous humor were measured using the corresponding species ELISA kits according to the manufacturer's instructions. Glutathione (GSH) (Jianglai Biology, JL-T0906) and iron ion (Jianglai Biology, JL-T1116) levels were assessed using appropriate biochemical kits. The optical density was measured with a microplate reader.
Animal Experiments
Six healthy male Brown Norway rats aged 6 weeks were acclimatized to feeding schedules for 2 weeks. IOP was measured between 10:00 AM and 11:00 AM, using a rebound tonometer (Icare, Helsinki, Finland). Measurements were taken six times daily from 1 week prior to surgery until 4 weeks postoperation, and the average of these readings was recorded as the baseline IOP.
An intravenous injection of hypertonic saline was administered to the episcleral veins of the right eyes, designating them as the chronic ocular hypertension (COH) group, while the left eyes served as the control group. The specific protocol was described in previous studies.26 After maintaining stable IOP for 4 weeks, the rats were sacrificed, and aqueous humor and corneoscleral limbus samples were collected for ELISA and histologic staining analyses.
Histologic Staining
Following sacrifice, the corneoscleral limbus tissues obtained from rats were fixed in 4% paraformaldehyde and subsequently embedded in paraffin wax. The sections were stained using hematoxylin and eosin (H&E). For immunohistochemistry (IHC), after antigen retrieval, the sections were incubated overnight with 4-HNE antibody (BIOSS, Beijing, China, bs-6313R) at 4°C.
The sections were then incubated with polymer–horseradish peroxidase goat anti-rabbit secondary antibody for 40 minutes at room temperature, followed by DBA staining for 10 minutes. Images were captured using a scanning platform. For tissue immunofluorescence (IF), the sections were incubated overnight with 4-HNE, hemoglobin α (HBA1) (Zen-bioscience, Chengdu, China, R25675), SCD (Zen-bioscience, R25675, 1:1000), Glut3 (SLC2A3) (Santa Cruz, Dallas, Texas, USA, sc-74399), α-SMA (Abcam, Cambridge, MA, USA; ab5694), and collagen I (COL-1) (Affinity, Jiangsu Province, China, AF7001) antibodies at 4°C.
The sections were subsequently incubated with goat anti-rabbit IgG (H+L)–Cy3 secondary antibody for 1 hour at room temperature, followed by DAPI staining for 5 minutes.
Transcriptome Data Source and Preparation
We queried the Gene Expression Omnibus (GEO) database to identify TM sequencing datasets from patients with POAG using the following keywords: “POAG,” “trabecular meshwork,” and “Homo sapiens” [organism]. The GSE27276 dataset was selected for subsequent analysis because it represents the largest and highest-quality publicly available sequencing resource of TM tissues from both patients with POAG and matched healthy controls. The GSE27276 dataset compared genome-wide expression profiles of TM in 13 controls and 15 patients with POAG based on the GPL2507 platform.
A total of 19 control TM samples and 17 POAG TM samples were collected, with 6 controls and 1 POAG case having expression data obtained from both left and right eyes. One technical replicate was performed between two cases. Control TM samples were sourced from donor eyes without a clinical history of glaucoma, while POAG TM samples were collected during trabeculectomy surgery.
The cohort included 5 African American and 10 Caucasian patients with POAG, as well as 3 African American and 10 Caucasian controls, with ages ranging from 40 to 94 years. Detailed information can be accessed from the GEO database and the corresponding article.27
GEO2R was utilized to analyze the differentially expressed genes (DEGs) between control and POAG TM samples, with |logFC| >1 and adjusted P < 0.05 serving as the thresholds for identifying DEGs. LogFC > 0 indicates upregulated genes, while logFC < 0 indicates downregulated genes.
A total of 567 ferroptosis-related genes (FRGs), including drivers, markers, suppressors, and unclassified regulators, were downloaded from FerrDb (http://www.zhounan.org/ferrdb/current/). A Venn diagram was employed to analyze the overlap between FRGs and DEGs. Heatmaps illustrating the expression patterns of differentially expressed FRGs between controls and patients with POAG were generated using the “ggplot2” R package.
The “RCircos” package was used to display the chromosomal locations of FRGs.
Functional Enrichment Analysis
The R packages “clusterProfiler,” “org.Hs.eg.db,” “enrichplot,” “circlize,” “RColorBrewer,” “dplyr,” “ggpubr,” “ComplexHeatmap,” and “ggplot2” were utilized to conduct Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway functional enrichment analyses, as well as visualization based on differentially expressed FRGs. The analysis employed a threshold of adjusted P < 0.05.
Screening Key Genes
The Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) online database (http://stringdb.org/) was employed to construct the protein–protein interaction (PPI) network for differentially expressed FRGs. PPI network facilitates the understanding of the interactions among various proteins, which is crucial for elucidating their functions and biochemical processes. Additionally, it can reveal patterns related to diseases and drug interactions.
Correlation analysis of differentially expressed FRG expression was performed using the “corrplot” R package, while visualization was conducted with the “ggplot2” R package, respectively. The importance score of differentially expressed FRGs was generated using the “randomForest” R package. Nomograms for multivariable models related to differentially expressed FRGs were created using the “rms” and “rmda” R package.
Receiver operating characteristic (ROC) curves were generated to validate the potential predictive effectiveness of differentially expressed FRGs as biomarkers for POAG using the “pROC” R package.
Cell Culture and Treatments
Primary human trabecular meshwork cells (HTMCs) were isolated from the corneal rims of nonglaucomatous donors. Limbal tissues were immersed in Dulbecco's modified Eagle's medium (DMEM)–F12 medium and microdissected along Schwalbe's line to obtain purified pigmented TM strips. Culture flasks (25 cm²) were precoated with FNC coating mix.
The TM strips were then transferred to the precoated flasks containing 1.5 mL of culture medium and maintained undisturbed at 37°C with 5% CO₂ for 10 to 14 days. After observing cell migration, additional medium was supplemented to support clonal growth. HTMCs were cultured in complete DMEM/F12 medium supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin.
Primary HTMCs within seven passages were utilized for subsequent experiments. To validate the identity of the HTMCs, the cells were treated with 500 nM dexamethasone for 7 days, which significantly upregulated myocilin (MYOC) expression, a characteristic response of HTMCs. The induction of MYOC protein was confirmed by Western blot analysis.
Three percent hydrogen peroxide (H2O2) was procured from Sigma-Aldrich (St. Louis, MO, USA; #323381). After screening a concentration gradient (0, 0.1, 0.5, 1, 2, 5, 10, 20 mM) to determine the optimal oxidative stress model, 1.13 mM H2O2 (EC50) was selected for subsequent experiments.
Digoxin (MCE, New Jersey, USA, HY-B1049) was tested across a concentration range (0.1, 1, 5, 10, 100 nM), with 1 nM identified as the optimal intervention concentration. When HTMCs reached approximately 70% confluence, they were pretreated with 1 nM digoxin for 24 hours, followed by exposure to 1.13 mM H2O2 for 1 hour. Finally, all cells were collected for the detection of relevant indicators.
Cell Viability and Cytotoxicity
To assess oxidative stress–induced injury, HTMCs were treated with H2O2. Cytotoxicity was evaluated by measuring the release of lactate dehydrogenase (LDH) into the cell culture media. Briefly, HTMCs were seeded in 96-well plates and subjected to the corresponding interventions.
Following treatment, the cell culture media were collected and analyzed using an LDH Cytotoxicity Assay Kit (Beyotime, Shanghai, China, C0017) according to the manufacturer's protocol. Cell viability was determined using a Cell Counting Kit-8 (CCK-8) assay (APExBIO, Houston, USA, K1018), performed in accordance with the manufacturer's instructions.
Quantitative Real-Time PCR Analysis
Total RNA was isolated from HTMCs using Trizol reagent (Transgen, Beijing, China, #ER501), following the manufacturer's protocol. The extracted RNA was then reverse transcribed into complementary DNA (cDNA) using a commercial reverse transcriptase kit (Transgen, #AU341). Quantitative real-time PCR (qPCR) was subsequently carried out with SYBR Green qPCR Super Mix (Transgen, #AQ601) under the following thermal cycling conditions: initial denaturation at 94°C for 30 seconds, followed by 40 cycles of 94°C for 5 seconds and 60°C for 30 seconds.
The relative gene expression levels were calculated using the 2−ΔΔCt method.
Identification of Small Molecular Compounds
Ferroptosis-related DEGs were uploaded into the Connectivity Map (CMap) database (https://clue.io). The top 10 candidate small molecular compounds and their mechanisms of action were identified through CMap mode-of-action (MoA) analysis and visualized using Hiplot (https://hiplot-academic.com/).
Molecular Docking
The three-dimensional (3D) structure of digoxin was retrieved from the PubChem database, while the 3D structure of the receptor protein was obtained from the Protein Data Bank or UniProtKB. Structural files for both the protein receptors and ligand small molecules were subsequently uploaded to the CB-Dock2 web server (https://cadd.labshare.cn/cb-dock2). A structure-based blind docking method, along with the Vina score, was employed to predict and evaluate the binding interactions between the small molecules (ligands) and target proteins (receptors).
The Vina score indicates the predicted ligand–receptor binding free energy (in kcal/mol), where a negative value signifies spontaneous binding, and a more negative value suggests stronger and more stable binding. In this study, a Vina score of less than −10 was established as the screening threshold.
Western Blot
HTMC proteins were extracted using a lysis buffer consisting of 98% radioimmunoprecipitation assay (RIPA), along with 1% protease inhibitor and 1% phosphatase inhibitor. The lysates were then centrifuged at 15,000 rpm for 15 minutes, and the sediment was discarded. Protein expression (50 µg/lane) was detected via SDS-PAGE (12%–20% gel).
Proteins were subsequently transferred onto polyvinylidene fluoride membranes and blocked with PBST containing 5% skim milk for 2 hours at room temperature. Following this, the immunoblots were washed and incubated with primary antibodies at 4°C for 24 hours. The primary antibodies used included GAPDH (HUABIO, Hangzhou, China; EM1901-57), HBA1, SLC2A3, SCD, and α-SMA.
After another wash, the membranes were incubated with the appropriate secondary antibodies (HUABIO; HA1006 and HA1001) at room temperature for 2 hours. The protein bands were visualized using enhanced chemiluminescence reagents and scanned with a Bio-Rad Imager, California, USA.
Gene Set Enrichment Analysis
We divided the samples of patients with POAG enrolled in the GSE27276 dataset into high and low expression groups based on the absolute expression levels of SLC2A3, HBA1, and SCD. The clusterProfiler and ggplot2 package in R (3.6.3) were utilized for the analysis. The reference gene set used was c2.cp.v7.2.symbols.gmt [Curated], sourced from the MSigDB Collections database.
Digoxin Administration in COH Rats
The experiments were conducted in two consecutive phases. Initially, the COH model was established in the right eyes of 10 healthy male Brown Norway rats through intravenous injection of hypertonic saline, as previously described, with the left eyes serving as controls. Four weeks post-COH modeling, rats exhibiting stable elevated IOP were randomly assigned to two groups: a digoxin treatment group and a vehicle control group.
The digoxin intervention was prepared in a vehicle solution containing (in sequential order) 10% DMSO, 40% PEG300, 5% Tween-80, and 45% saline, adhering to previously reported concentrations of 1 mM digoxin.28 Both digoxin and vehicle were administered as eye drops twice daily for a duration of 4 weeks.
Statistical Analysis
Data analysis was performed using the stats [4.2.1], car [3.1-0] packages of R software (4.2.1) and GraphPad Prism (GraphPad Software, La Jolla, CA, USA) software. Data are presented as mean ± SEM. MDA and GPX4 levels in the aqueous humor of the rats were analyzed using paired sample t-tests.
For comparisons between two groups, Student's t-test (when normality and homogeneity of variance were satisfied), Welch's t-test (when normality was satisfied but homogeneity of variance was not), and Wilcoxon rank-sum test (when normality was not satisfied, indicating a nonparametric test) were employed. One-way ANOVA was utilized for comparisons among four groups. A P value of <0.05 was considered statistically significant.
Results
Causal Effects of Blood Metal Elements on POAG
The present study utilized a two-sample MR approach to examine the causal relationships between blood metal levels and POAG. The results revealed a significant positive association between elevated serum iron levels and POAG risk (odds ratio [OR] = 1.192; 95% confidence interval [CI], 1.073–1.325, P = 0.001). This association remained statistically significant after correction for the false discovery rate (FDR = 0.046) (Figs.
1A, 1B). Additionally, serum magnesium (OR = 1.226; 95% CI, 1.064–1.413; P = 0.005) and plasma aluminum (OR = 1.128; 95% CI, 1.021–1.246; P = 0.017) exhibited nominally positive associations; however, these did not achieve significance following FDR adjustment (FDR = 0.103 and 0.191, respectively). The association of serum copper (OR = 1.172; 95% CI, 1.027–1.336, P = 0.018) also necessitated cautious interpretation due to FDR exceeding 0.05.
Sensitivity analyses confirmed the robustness of the findings: heterogeneity tests (Cochran's Q P > 0.05), MR-Egger regression (intercept P > 0.05), and leave-one-out analyses supported the stability of the serum iron effect (Figs. 1C–E). In conclusion, elevated serum iron is identified as an independent risk factor for POAG, whereas the potential roles of other metals warrant further validation in larger cohorts.
Enhanced Ferroptosis Levels in POAG
To investigate the potential role of ferroptosis in POAG, we conducted an analysis of ferroptosis indicators in the aqueous humor of patients with POAG and control individuals. Our findings revealed that, compared to the control group, patients with POAG exhibited elevated levels of iron ions and MDA in their aqueous humor, indicative of increased lipid peroxidation (Figs. 2A, 2B).
Conversely, the expression levels of GPX4 and GSH, key antioxidants that counteract ferroptosis, were significantly lower in the POAG group than in the control group (Figs. 2C, 2D). To further elucidate the role of ferroptosis in glaucoma, we developed a COH rat model by injecting hypertonic saline into the episcleral vein of the right eye, using the left eye as a control.
After 1 week of modeling, the IOP remained stable and elevated for 4 weeks (Fig. 2E). Consistent with our findings in patients with POAG, the MDA content in the aqueous humor of the COH eyes was increased, while the GPX4 content was decreased compared to the control eyes (Figs.
2F, 2G). H&E staining displayed the morphologic structure of the trabecular meshwork in both control and COH eyes (Fig. 2H).
Furthermore, IHC and IF detection revealed that the lipid peroxidation marker 4-HNE was significantly overexpressed in the COH eyes (Supplementary Fig. S1). These results suggest that the imbalance in redox homeostasis, characterized by increased levels of iron ions and MDA, along with decreased levels of GPX4 and GSH, indicates a potential mechanism through which ferroptosis may contribute to the pathogenesis of glaucoma.
Ferroptosis-Related Gene Patterns in POAG
A total of 194 DEGs were identified in TM tissue between POAG and control samples, with 14 of these genes associated with ferroptosis (Fig. 3A). A heatmap was employed to visualize the expression patterns of these 14 FRGs, which comprised five ferroptosis drivers, five ferroptosis suppressors, and four unclassified genes (Fig.
3B). Notably, most of these genes were downregulated, with only FADS2, SCD, and HBA1 exhibiting upregulation. The chromosomal locations of these FRGs are illustrated in Figure 3C.
GO and KEGG enrichment analysis revealed that these 14 DEGs were significantly enriched in processes related to oxygen levels, glucocorticoid response, one-carbon compound transport, gas transport, water transport, and iron ion binding (Figs. 3D, 3E, Supplementary Table S1). To further investigate the interactions among these FRGs, a PPI network was constructed (Fig.
4A). Within this network, SLC2A3 emerged as a hub, participating in the majority of connections. However, HBA1 did not display known PPIs with other genes in the network.
Despite this, correlation analysis demonstrated that HBA1 exhibited a strong expression correlation with the other 13 genes (Fig. 4B). Gene importance scoring was conducted, revealing that HBA1, SLC1A5, and FADS2 (with importance scores of ≥2) may play critical roles in POAG (Fig.
4C). In a nomogram, HBA1 also demonstrated excellent predictive performance for POAG (Fig. 4D).
Additionally, ROC curves were established to assess the potential diagnostic efficiency of the 14 hub genes as biomarkers for POAG (Fig. 4E, Supplementary Table S2). HBA1 showed exceptional ability to distinguish patients with POAG from controls, with an area under the curve of 0.997.
The other 13 hub genes also exhibited favorable accuracy as biomarkers for POAG. These findings collectively reveal a ferroptosis-related gene signature in TM tissues of patients with POAG, highlighting the potential importance of these genes in the pathogenesis of the disease.
Verification of Ferroptosis Gene Expression Trends
Primary HTMCs were identified through myocilin induction following a treatment of 500 nM dexamethasone for 7 days (Supplementary Fig. S2). An oxidative stress model in HTMCs was developed using H2O2, which is a widely recognized cellular model for POAG.
To stimulate the HTMCs, various concentrations of H2O2 (ranging from 0 to 20 mM, including 0.1, 0.5, 1, 2, 5, and 10 mM) were applied for 1 hour to stimulate the HTMCs. Cell viability was evaluated using the CCK-8 kit, which indicated that the median effective concentration (EC50) was 1.13 mM, as illustrated in Figure 5A. Consequently, a concentration of 1 mM H2O2 was chosen for subsequent experimental procedures.
The expression levels of 14 hub genes were subsequently quantified using qPCR. Compared to the control group, three hub genes (HBA1, LCN2, and HILPDA) displayed increased expression, while five hub genes (SLC2A3, DDIT4, SCD, FADS2, and PROM2) showed decreased expression in H2O2-induced HTMCs (Figs. 5B–O).
Notably, the expression trends of HBA1, SLC2A3, DDIT4, and PROM2 were consistent with the sequencing results (Supplementary Table S3). However, no significant differences were observed in the expression levels of six hub genes (ZFP36, DUSP1, AQP3, AQP5, SLC1A5, and GDF15) between the control and H2O2-treated groups. These findings underscore the critical role of HBA1-mediated ferroptosis in the pathogenesis of POAG.
Potential Small-Molecule Compound Screening
To identify potential small-molecule compounds for the treatment of POAG, we uploaded ferroptosis-related DEGs to the Connectivity Map small-molecule drug database for the annotation of chemical compounds and their mechanisms of action. The top 10 potential small-molecule drugs identified include LDN-193189, BX-795, everolimus, ingenol, XMD-892, cycloheximide, digoxin, proscillaridin, calmidazolium, and digitoxin. Additionally, eight mechanisms of action were identified: serine/threonine kinase inhibitor, IκB kinase (IKK) inhibitor, mammalian target of rapamycin inhibitor, protein kinase C activator, mitogen-activated protein kinase inhibitor, protein synthesis inhibitor, calcium channel blocker, and ATPase inhibitor, as shown in Figure 6A.
These findings suggest that digoxin and other ATPase inhibitors may possess therapeutic potential in the treatment of POAG. Furthermore, we conducted molecular docking between digoxin and proteins encoded by eight differentially expressed hub genes in H2O2-induced HTMCs (Fig. 6B).
Based on structure-based blind docking, the Vina score results indicated that digoxin is more likely to regulate ferroptosis in HTMCs by targeting HBA1, SCD, and SLC2A3 (Vina score <−10), as detailed in Supplementary Table S4.
Digoxin Protects HTMCs From Ferroptosis
To validate the potential therapeutic benefits of digoxin in treating HTMC ferroptosis, various concentrations of digoxin were utilized to treat HTMCs exposed to H2O2. Notably, the enhanced the release of LDH into the supernatant, a hallmark of cell death induced by H2O2, was inhibited by 0.1 nM and 1 nM digoxin (Fig. 7A).
Moreover, 1 nM digoxin reversed the H2O2-induced increases in MDA and iron ions, as well as the decrease in GPX4 levels in the cell culture medium (Figs. 7B–D). While statistical analysis did not reveal significant effects of digoxin on GSH content, the trends in GSH-level changes were consistent with expectations (Fig.
7E). Western blot analysis further demonstrated that digoxin reversed the H2O2-induced downregulation of SLC2A3 and SCD, as well as the upregulation of HBA1 (Fig. 7F).
Additionally, digoxin suppressed the H2O2-induced upregulation of α-SMA, which is indicative of extensive ECM remodeling. These findings provide evidence that digoxin may protect HTMCs from ferroptosis by targeting SLC2A3, SCD, and HBA1.
Gene set enrichment analysis (GSEA) was employed to systematically uncover downstream signaling pathway features driven by expression differences of the key genes SLC2A3, SCD, and HBA1 (Fig. 8). These genes functionally converge on pathways critical to TM homeostasis: HBA1 (enriched in metal ion response), SCD (linked to response to metal ions and seleno-amino acid metabolism), and SLC2A3 (associated with glutathione metabolism) regulate redox balance and detoxification, exacerbating iron-induced oxidative stress.
GSEA further highlights ECM remodeling and programmed cell death as key downstream events, suggesting that iron overload disrupts TM integrity via oxidative damage to ECM components and glutathione depletion, impairing aqueous humor outflow.
IOP-Lowering Effect of Digoxin for POAG
To assess the effect of digoxin on IOP reduction in POAG in vivo, a COH rat model was established as previously described, with IOP stabilized for 1 month. Subsequently, COH rats received twice-daily topical administrations of either 1 mM digoxin or a vehicle for an additional month (Fig. 9A).
As anticipated, digoxin significantly decreased IOP in COH eyes compared to the vehicle-treated group (Fig. 9B). Furthermore, digoxin mitigated the elevation of MDA levels in the aqueous humor of COH eyes (Fig.
9C) and enhanced the reduced GPX4 levels in the same (Fig. 9D). Histologic staining of the eyeballs revealed that digoxin inhibited the COH-induced upregulation of 4-HNE in the TM tissue (Fig.
10). Additionally, digoxin reversed the COH-induced downregulation of SLC2A3 and SCD, as well as the upregulation of HBA1, while also inhibiting COH-induced ECM remodeling, as indicated by α-SMA and COL-1 expression (Supplementary Fig. S3).
These findings suggest that digoxin reduces IOP by inhibiting ferroptosis in the trabecular meshwork.
Discussion
This study revealed a positive causal relationship between elevated serum iron levels and the risk of POAG, consistent with evidence from previous epidemiologic and mechanistic research. As a redox-active metal, iron may contribute to the pathogenesis of POAG by inducing oxidative stress, accelerating mitochondrial dysfunction, and promoting trabecular meshwork fibrosis. Additionally, we conducted a comprehensive, exhaustive analysis of the clinical manifestations associated with ferroptosis patterns in POAG.
Despite the growing body of evidence suggesting that ferroptosis plays a crucial role in glaucoma pathogenesis, the exact mechanisms underlying ferroptosis-induced trabecular meshwork injury remain unclear. Our GO and KEGG analyses suggest that ferroptosis is initiated and executed through pathways involving hypoxia, iron ion binding, and carbon metabolism, consistent with earlier research findings. Notably, we noted significant enrichment in the responses to glucocorticoid and water transport,29 which correspond with the reported pathogenic mechanisms of POAG.
This study unveiled several genes that appear to mediate ferroptosis in POAG, including SLC2A3, HBA1, and SCD. Notably, SLC2A3 has been recognized as a crucial biomarker of ferroptosis across various diseases.30,31 It primarily functions in glucose transmembrane transport and may contribute to ferroptosis by modulating glycolysis and immune infiltration.32 Nag et al.33 highlighted the 12p13.3 CNV locus overlapping the SLC2A3 gene as a vital determinant of IOP control. Furthermore, Xuan et al.34 demonstrated that inhibiting FADS2 disrupts lipid peroxidation, leads to mitochondrial dysfunction, and triggers ferroptosis by directly downregulating GPX4.
Additionally, HBA1 regulates ferroptosis through GSH-related pathways and influences the transport of p-pyruvate and p-serine.35 Intriguingly, HbA1c levels exhibit a positive correlation with IOP in patients with nonproliferative diabetic retinopathy.36 These findings provide valuable insights into the complex interplay between ferroptosis and POAG, suggesting potential therapeutic targets for future interventions.
HBA1 stands out as a distinctive gene in our study, displaying a notable lack of direct interactions while exhibiting a robust correlation in expression with other key genes. This gene demonstrated the highest importance score and significant diagnostic potential for POAG. Remarkably, only SLC2A3 and HBA1 consistently appeared in sequencing results from TM tissues, PCR results from HTMCs, and histologic sections of a COH rat model.
In contrast, SCD expression trends were not consistent between sequencing and qPCR validation. Despite these promising findings, the relationship between these hub genes, ferroptosis, and the underlying pathophysiologic mechanisms of POAG remains unclear due to insufficient evidence. Our data strongly indicate that these hub genes, including HBA1, hold great promise for diagnosing POAG.
However, we observed significant downregulation in most ferroptosis-related genes (including drivers, suppressors, and unclassified genes) in POAG samples. We hypothesize that this downregulation represents a compensatory or adaptive response to early-stage cellular stress within the trabecular meshwork, potentially reflecting the tissue's effort to maintain iron homeostasis and prevent cell death. As POAG progresses, this compensatory mechanism may become inadequate, leading to disrupted iron homeostasis and subsequent ferroptotic cell death.
This perspective offers a more dynamic, stage-specific understanding of the pathogenic process, moving beyond a simplistic linear model of gene expression changes. Further clinical and basic research is imperative to validate their pivotal role in the disease process. By delving deeper into the mechanisms by which these genes influence ferroptosis and POAG pathogenesis, we may uncover novel therapeutic targets and improve our understanding of this blinding condition.
Furthermore, potential small-molecule compounds for the treatment of POAG have been identified, offering a theoretical foundation for the development of therapeutic strategies targeting TM injury. Currently, the effective management of glaucoma encounters numerous challenges, as most antiglaucoma medications primarily focus on lowering IOP by either increasing aqueous humor outflow or reducing its production through the uveoscleral outflow pathway. Traditional antiglaucoma medications face issues such as limited efficacy, poor adherence to long-term use, ocular surface damage, surgical failure, and other adverse reactions.37,38 Prostaglandin analogues, which target extracellular matrix remodeling to increase aqueous humor outflow through the uveoscleral pathway, are currently the most widely used IOP-lowering agents.39 Another relatively novel class of antiglaucoma medications, Rho kinase inhibitors, reduces IOP by directly targeting the conventional outflow pathway and simultaneously relaxing trabecular meshwork cells.40 Since the drainage of aqueous humor through the TM channel accounts for 50% to 75% of total outflow, the development of medications that specifically target TM injury to promote increased aqueous humor outflow represents the most promising avenue for glaucoma treatment.
In this study, we screened eight pharmacologic and genetic inhibitors of ferroptosis as potential antiglaucoma medications. Notably, three of these inhibitors are Na+, K+-ATPase inhibitors: digoxin, proscillaridin, and digitoxin. Previous studies have demonstrated that iron overload increases Na,K-ATPase activity in erythrocyte plasma membranes.41 Na,K-ATPase interacts with Parkin, preventing Parkin from translocating to the mitochondria and inactivating mitophagy, which leads to ferroptosis.42 On the other hand, Na,K-ATPase also interacts with SLC7A11, inhibiting ferroptosis in Parkinson disease.
Therefore, the relationships between Na,K-ATPase and ferroptosis are complicated and vary across different cell types and diseases. Our results suggest a therapeutic role for digoxin in POAG. Consistent with our findings, Katz et al.43 reported that digoxin and its derivatives significantly lowered IOP in animal models of ocular hypertension.
Initial hypotheses proposed that this effect might involve digoxin's inhibition of aqueous humor production in the ciliary body.28 However, our cellular studies suggest that digoxin has a protective effect in the TM. Critically, a recent study by Barathi et al.44 employing an ocular hypertensive nonhuman primate model (OHT-NHP; Macaca fascicularis) reassessed the efficacy and mechanism of the cyclobutyl digoxin derivative (DcB). This investigation demonstrated that DcB potently lowered IOP in OHT-NHP under both acute (24 hours) and extended (7–10 days) conditions.
Notably, this reduction was accompanied by an increased aqueous flow rate, whereas normotensive animals (ONT-NHP) exhibited a minimal response. Mechanistic studies utilizing isolated porcine ciliary epithelium (aqueous humor (AH) inflow) and perfused enucleated eyes (AH outflow) revealed two key findings: (1) The application of DcB to the aqueous compartment in contact with the nonpigmented epithelial cell layer resulted in a significant increase in fluid secretion. The subsequent addition of 1 mM ouabain to the aqueous surface markedly suppressed (∼70%) this DcB-induced secretion.
(2) The pressure–flow relationship slope (mm Hg versus µL/min) in DcB-treated eyes was significantly lower than that in controls, indicating an enhanced outflow facility contributing to IOP reduction. Collectively, these data necessitate a revision of the original hypothesis: DcB lowers IOP primarily by enhancing AH outflow rather than suppressing inflow in OHT-NHP and likely other species. The study by Barathi et al. provides evidence that informs the site of digoxin's action.
While their digoxin derivative DcB increased fluid secretion from the ciliary epithelium in vitro, its in vivo effects point to a different dominant mechanism. First, DcB produced an additive IOP-lowering effect when combined with latanoprost (a uveoscleral outflow enhancer), suggesting distinct primary sites of action. Second, and more critically, DcB was effective in an ocular hypertensive NHP model but not in a normotensive model with altered outflow (ONT-NHP).
This contrasts with prostaglandins, which lower IOP in normal eyes, and implies that digoxin's IOP-lowering efficacy is condition-dependent, requiring a specific state of the conventional outflow pathway (e.g., sensitized by high IOP or oxidative stress). This aligns with our in vitro findings that digoxin protects TM cells under stress. We acknowledge that our study design cannot definitively rule out complex effects on aqueous dynamics; however, the collective evidence strongly suggests that facilitation of conventional outflow is the principal mechanism underlying the observed IOP reduction.
In addition, IKK inhibitor BX-795 has shown potential in protecting RGCs in glaucoma, while no research has established its correlation with IOP.45 Protein synthesis inhibitor cycloheximide was reported to reverse glucocorticoid-induced responsiveness in HTMCs, such as upregulated MYOC.46,47 Several calcium channel blockers, such as verapamil and diltiazem, have demonstrated significant reductions in IOP and improvements in ophthalmic blood flow.48,49 These novel small-molecule compounds are promising candidates for innovative therapeutic strategies in glaucoma. Nevertheless, further basic experiments and clinical cohorts are necessary to validate the therapeutic effect of these medications.
Nevertheless, this study has several limitations. First, significant scarring in the hypertonic saline model limits its utility for evaluating TM-specific mechanisms. Future studies will employ POAG-relevant models (e.g., dexamethasone-induced or AAV-TGFβ2) to address this constraint.
Second, while our data strongly suggest digoxin's action on the TM, direct measurement of aqueous humor outflow facility was precluded by equipment limitations. This measurement is crucial to definitively confirm that the IOP reduction is mediated specifically through enhanced conventional outflow, as opposed to alterations in aqueous production or uveoscleral outflow. Future studies employing perfusion systems or tracer studies will be essential to directly quantify the effect of digoxin on outflow facility and precisely delineate its site of action within the aqueous dynamics pathway.
Third, while ferroptosis inhibition was identified as one mechanism for digoxin's IOP reduction, its potential effects on TM contractility—mediated through Na⁺/K⁺-ATPase inhibition and subsequent calcium elevation—remains insufficiently explored.50
This study establishes a positive causal relationship between elevated serum iron levels and the risk of POAG, elucidating the pivotal role of ferroptosis-related gene signatures in the underlying mechanisms of POAG. It offers a novel perspective on the clinical manifestations and ramifications of ferroptosis in glaucoma, thereby establishing a robust theoretical framework for the prevention and early identification of this condition. Furthermore, the potential for developing targeted small-molecule drugs that modulate ferroptosis-related pathways presents promising avenues for innovative therapeutic strategies in the treatment of glaucoma.
Supplementary Material