Identification and validation of oxidative stress signature genes in the trabecular meshwork of glaucoma

Abstract

Oxidative stress (OS) plays an important role in trabecular meshwork (TM) dysfunction in glaucoma, but its molecular mechanism remains unclear. We integrated the GSE27276 dataset and OS-related gene sets from GeneCards to screen 61 differentially expressed OS-related genes (DEOSGs). Functional enrichment analysis revealed that these genes are primarily involved in inflammatory and OS-related signaling pathways, including IL-17, TNF, and NF-κB. A protein–protein interaction (PPI) network constructed via the STRING database identified seven hub genes (CCL3, CXCL1, NFKBIA, VCAM1, LCN2, TNFRSF1A, and HP). Validation using the GSE124114, GSE37474, and GSE65240 datasets showed that the expression of CXCL1 and VCAM1 was downregulated, while NFKBIA was upregulated. Additionally, all three genes exhibited an area under the curve (AUC) greater than 0.7. Immune infiltration analysis demonstrated significant associations between these genes and immune cells, particularly regulatory T cells and neutrophils. Regulatory network analysis suggested that transcription factors (RELA, NFKB1) and microRNAs (hsa-miR-34a-5p) may modulate these core genes. Drug-gene interaction studies identified 35 potential therapeutic agents, including Infliximab and Vitamin B6. This study systematically elucidates the molecular mechanisms of OS in glaucoma, proposing that the identified core genes and their regulatory networks not only serve as novel biomarkers for diagnosis but also provide a theoretical foundation for developing targeted therapeutic strategies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-16534-z.

Introduction

Glaucoma remains one of the leading causes of irreversible blindness worldwide, characterized pathologically by progressive loss of retinal ganglion cells and optic nerve atrophy¹. The pathogenesis of glaucoma involves multifaceted mechanisms extending beyond intraocular pressure (IOP)-induced mechanical stress, encompassing OS, mitochondrial dysfunction, vascular dysregulation, and neuroinflammatory cascades^2,3. Notably, oxidative stress is not limited to primary open-angle glaucoma (POAG), but also plays a crucial role in the pathophysiology of other glaucoma subtypes, including angle-closure glaucoma, normal-tension glaucoma, secondary glaucomas such as pseudoexfoliative and pigmentary glaucoma, and steroid-induced glaucoma. This highlights OS as a universal contributor to glaucoma-related optic neuropathy across clinical variants^4–6. OS driven by excessive reactive oxygen species (ROS) production and compromised antioxidant defenses exacerbates TM degeneration, extracellular matrix (ECM) remodeling^7,8, and axonal damage, positioning it as a critical therapeutic target^5,6,9. Therefore, understanding the molecular basis of OS in glaucoma is imperative for developing early diagnostic biomarkers and precision therapeutic strategies to mitigate disease progression.

Oxidative stress is a critical mediator in the pathogenesis of glaucoma, arising from an imbalance between oxidative and antioxidant defense systems¹⁰, leading to the excessive accumulation of ROS¹¹. Among them, hydroxyl radicals (·OH) are the most reactive species and can cause irreversible damage to DNA¹², particularly mitochondrial DNA¹³. Superoxide anions (O₂⁻), on the other hand, primarily disrupt biological membranes by initiating lipid peroxidation, generating cytotoxic lipid oxidation byproducts. ROS also damage amino acid residues, especially cysteine and methionine, leading to structural disruptions in critical protein regions, which in turn result in misfolding or loss of function¹⁴. Extensive studies have demonstrated that glaucoma patients exhibit reduced expression levels of antioxidant enzymes, such as superoxide dismutase, in both serum and ocular tissues, accompanied by elevated levels of lipid peroxidation markers like malondialdehyde, indicating a disruption in redox homeostasis^15–17. The TM is the most oxidation-sensitive tissue in the anterior chamber, possessing weaker antioxidant defense capacity compared to the cornea and iris¹⁸. Substantial evidence has shown severe DNA oxidative damage in TM tissues of glaucoma patients^4,19. Increased hydrogen peroxide levels and prolonged oxidative stress-induced damage contribute to TM cell structural remodeling, disruption of ECM homeostasis, and subsequent TM fibrosis, leading to impaired aqueous humor outflow and elevated intraocular pressure—a hallmark of glaucoma^6,20. Despite the established association between oxidative stress and glaucoma progression, comprehensive identification of oxidative stress-related core genes and their regulatory networks remains limited.

Based on this, the present study integrates bioinformatics to analyze datasets from the Gene Expression Omnibus (GEO), including primary open-angle glaucoma (POAG) (GSE27276) and glucocorticoid-induced glaucoma model (GSE124114, GSE37474 and GSE65240). DEOSGs were identified, and a PPI network was constructed to determine hub genes. Furthermore, the biological functions and clinical relevance of these genes were validated through multidimensional analyses, including enrichment analysis, immune cell correlation assessment, and drug prediction. This study aims to elucidate key molecular mechanisms underlying oxidative stress in glaucoma, providing a theoretical foundation for the discovery of diagnostic biomarkers and the development of targeted therapeutic strategies.

Materials and methods

Expression matrix acquisition and differential analysis

By searching the GEO database for “glaucoma,” we identified GSE27276²¹ as the training dataset and GSE124114²², GSE37474²³ and GSE65240²⁴ as the validation datasets for this study. The GSE27276 dataset comprises TM tissue samples from 17 patients with POAG and 19 control samples. The GSE124114 dataset includes TM samples from 8 glucocorticoid-induced glaucoma cases and 8 control samples. The GSE37474 dataset includes TM samples from 5 glucocorticoid-induced glaucoma cases and 5 control samples. The GSE65240 dataset includes TM samples from 3 glucocorticoid-induced glaucoma cases and 3 control samples. We used Sangerbox (http://www.sangerbox.com/)²⁵ to merge the three validation datasets. Briefly, the datasets were first merged using the R package inSilicoMerging²⁴, and then batch effects were removed using the method proposed by Johnson et al. to obtain a combined expression matrix²⁶. Differential expression analysis was performed using the limma package in R (all code has been uploaded to our GitHub repository). During the initial screening phase, to capture more potential differentially expressed genes, we set the threshold at p < 0.05 and |log2FC|> 0.585. In the validation phase, the criteria were adjusted to adj.P.val < 0.05 and |log2FC|> 0.585^27–29. Figure 1 illustrates the overall design of this study.

Fig. 1.

The flow chat of this study.

Identification DEOSGs associated with glaucoma.

We searched the keyword “oxidative stress” in the GeneCards database^30–32 (v5.23, https://www.genecards.org/) and selected genes with a Relevance Score > 7 as the OS gene set. GeneCards integrates information from over 190 curated sources and is widely recognized as a comprehensive gene-centric resource for disease, pathway, and compound associations^33,34. Subsequently, the online venn diagram tool EVenn (https://www.bic.ac.cn/test/venn/#/) was used to identify DEOSGs associated with glaucoma.

Gene ontology (GO) and Kyoto encyclopaedia of genes and genomes (KEGG) enrichment analyses

GO analysis characterizes gene sets through cellular components (CC), molecular functions (MF) and biological processes (BP) to uncover biological significance³⁵. KEGG, an integrated database of genomic and chemical information, provides metabolic pathway maps for understanding gene functions^36–38. Network-based functional enrichment analysis was conducted using ClueGO v2.5.10^39,40, a Cytoscape plugin that integrates GO terms and pathway annotations. The input consisted of 61 DEOSGs. A two-sided hypergeometric test was used for enrichment testing, and p values were adjusted using the Benjamini–Hochberg method. A adjusted p value threshold of < 0.05 was considered statistically significant²⁷. GO terms and pathways from KEGG were included in the analysis. Functional grouping was based on a Kappa Score of 0.4. The resulting network graphs were visualized with node size representing significance and color denoting functional clusters.

PPI network construction and hub gene screening

We input the DEOSGs into the STRING database⁴¹ (https://cn.string-db.org/), selecting Homo sapiens as the species and setting the minimum required interaction score to > 0.4. To identify hub genes within the PPI network, we utilized the CytoHubba plugin in Cytoscape (v3.10.3)⁴², applying two topological analysis algorithms: Degree and Maximal Clique Centrality (MCC). The Degree algorithm reflects the centrality of a node based on its number of direct connections. MCC is a highly sensitive and accurate method for detecting key nodes in complex networks⁴³. The intersection of these two ranking methods was selected as the final hub genes⁴⁴. To further explore the gene interaction network of hub genes, we employed GeneMANIA (https://genemania.org/)⁴⁵ to construct an interaction network encompassing hub genes and their related genes.

Validation of hub genes.

Box plots illustrating the differential expression levels of hub genes in the GSE27276 and GSE124114, GSE37474 and GSE65240 datasets were generated using the ggpubr package in R. Subsequently, the diagnostic accuracy of hub genes in these datasets was validated using the glmnet and pROC R packages. Genes with an AUC > 0.7 were considered as potential diagnostic indicators for the disease.

Single-sample gene set enrichment analysis (ssGSEA)

We performed ssGSEA⁴⁶ on glaucoma samples using the GSVA package in R to analyze the distribution of 28 immune cell types within each glaucoma sample. Heatmaps were generated using Sangerbox for visualization. Furthermore, immune cell correlation analysis of hub genes was conducted using the limma, GSVA, and GSEABase packages in R.

Prediction of TFs and microRNAs regulating hub genes

To investigate the transcriptional and post-transcriptional regulation of the identified hub genes, we utilized the miRNet v2.0 platform (https://www.mirnet.ca/)^47,48, an integrated tool that aggregates data from multiple high-quality resources. This platform enables a comprehensive exploration of experimentally validated and predicted interactions involving microRNAs, transcription factors (TFs), and target genes. The regulatory network was constructed by inputting the hub genes into miRNet and selecting the “Human” database. The interaction data were then exported and visualized using Cytoscape (v3.10.3). We accessed the database in March 2025.

Prediction of drugs targeting hub genes

We predicted drugs that regulate the hub genes using the Drug-Gene Interaction Database (DGIdb) (https://dgidb.org/)⁴⁹. The drug-gene interaction network was then visualized using Cytoscape.

Results

Identification of DGEs and DEOSGs

A total of 656 differentially expressed genes (DEGs) were identified from the GSE27276 dataset (Supplementary Table 1), comprising 309 upregulated and 347 downregulated genes (Fig. 2A). The heatmap illustrates the top 50 most significantly upregulated and downregulated genes (Fig. 2B). From the GeneCards database, 1225 oxidative stress-related genes were retrieved (Supplementary Table 2). Venn diagram analysis revealed 61 overlapping genes between glaucoma-related DGEs and oxidative stress-related genes, which were subsequently defined as DEOSGs (Fig. 2C).

Fig. 2.

(A) Volcano plot of DEGs in GSE27276, with green indicating downregulated genes and red indicating upregulated genes. (B) Heatmap of the top 50 differentially expressed genes in the GSE27276 dataset, with red indicating upregulated genes and blue indicating downregulated genes. (C) Venn diagram. The intersecting genes between GSE27276-DEGs and the oxidative stress gene set, referred to as DEOSGs.

Enrichment analyses of DEOSGs

To better understand the biological significance of the 61 DEOSGs, we performed network-based enrichment analysis using the ClueGO plugin in Cytoscape. The analysis identified a total of 117 significant GO terms. Enriched biological processes (BP) included positive regulation of inflammatory response, reactive oxygen species metabolic process, and hydrogen peroxide metabolic process (Fig. 3A). Molecular functions (MF) such as protease binding, oxygen binding, and primaryamine oxidase activity were also significantly enriched (Fig. 3B). Cellular components (CC) significantly enriched included hemoglobin complex, mitochondrial respirasome, and cytoplasmic vesicle lumen (Fig. 3C). These findings suggest that the oxidative stress-related genes are not only involved in inflammation and redox metabolism, but also play crucial roles in mitochondrial function and intracellular vesicle dynamics, thereby contributing to the pathophysiological mechanisms of glaucoma. In the KEGG enrichment analysis, we identified 22 significantly enriched pathways, including those related to inflammatory responses (such as the IL-17 signaling pathway, TNF signaling pathway, and PPAR signaling pathway) and metabolism-related pathways (including lipid and atherosclerosis, mitophagy, and the AGE-RAGE signaling pathway in diabetic complications). These associations are presented in Fig. 3D and detailed in Supplementary Table 3.

Fig. 3.

Enrichment analysis. (A) Gene Ontology biological processes of the DEOSGs. (B) Gene Ontology Molecular functions of the DEOSGs. (C) Gene Ontology Cellular components of the DEOSGs. (D) KEGG enrichment analysis of the DEOSGs. The node size represents the term enrichment significance.

Identification of hub genes

To further explore the interactions among DEOSG-encoded proteins and identify hub genes, we analyzed the PPI network of DEOSGs using STRING, which comprised 61 nodes and 193 edges (Fig. 4A). The PPI network was subsequently imported into Cytoscape for hub gene selection. Using the CytoHubba plugin, we identified the top 10 genes ranked by the MCC and degree algorithms, and the intersection of both methods (CCL3, CXCL1, NFKBIA, VCAM1, LCN2, TNFRSF1A, and HP) was selected as the final hub genes (Fig. 4B).

Fig. 4.

(A) PPI network analysis of DEOSGs. The color intensity and size of the nodes are scaled according to their degree values. (B) The top 10 hub genes identified by the degree and MCC algorithms via the CytoHubba plugin in Cytoscape. (C) Expression of 7 hub genes in GSE27276, *p < 0.05, **p < 0.01, ***p < 0.001. (D) The ROC curve of hub genes in GSE27276. (E) GeneMANIA diagram showing the coexpression interactions between the 7 hub genes and their neighbouring genes.

Expression and interaction network of hub genes

The box plot (Fig. 4C) illustrates the expression levels of hub genes in the GSE27276 dataset, revealing that CCL3 and VCAM1 were upregulated in the glaucoma group, whereas CXCL1, NFKBIA, LCN2, TNFRSF1A, and HP were downregulated. Subsequently, receiver operating characteristic (ROC) curve analysis was performed to assess the diagnostic efficacy of hub genes in the GSE27276 dataset (Fig. 4D). The AUC values for CCL3, CXCL1, NFKBIA, VCAM1, LCN2, TNFRSF1A, and HP were 0.842, 0.777, 0.916, 0.731, 0.895, 0.910, and 0.932, respectively, indicating their high diagnostic value. Furthermore, GeneMANIA network analysis was conducted to elucidate the interaction network of the seven hub genes. In the network visualization (Fig. 4E), inner nodes represent the hub genes, while outer nodes correspond to genes interacting with the hub genes. The different types of interactions are represented by various colored edges, with light purple edges indicating co-expression relationships between genes.

Enrichment analyses of hub genes

The analysis identified a total of 31 significant GO terms. Enriched biological processes (BP) included acute inflammatory response, response to lipopolysaccharide, and cellular response to tumor necrosis factor (Fig. 5A). Molecular functions (MF) such as chemokine receptor binding, and chemokine activity were also significantly enriched. Cellular components (CC) significantly enriched included specific granule and tertiary granule (Supplementary Table S4). In the KEGG enrichment analysis, we found that 41 pathways were significantly enriched (Fig. 5B), such as NF-kappa B signaling pathway, IL-17 signaling pathway, TNF signaling pathway and Lipid and atherosclerosis. These associations are shown in Supplementary Table 4. These results indicate that the hub genes are closely associated with inflammatory and immune responses.

Fig. 5.

Enrichment analysis. (A) Gene Ontology biological processes of the DEOSGs. (B) KEGG enrichment analysis of the DEOSGs. The node size represents the term enrichment significance.

ssGSEA of hub genes

Based on enrichment analysis, we observed that the hub genes were closely related to inflammation and immune modulation, particularly through pathways such as TNF and NF-κB signaling. To further investigate the immunological context, we performed ssGSEA on glaucoma samples from the GSE27276 dataset to assess the enrichment levels of 28 immune cell types. The results indicated that most immune cells were more active in glaucoma samples (Supplementary Fig. 1), including immature dendritic cells and effector memory CD8 T cells, whereas activated B cells and eosinophils exhibited lower activity.

Subsequently, we analyzed the correlation between seven hub genes and immune cell infiltration. The results showed that TNFRSF1A was negatively correlated with Type 1 T helper cells, natural killer cells, gamma delta T cells, and monocytes (Fig. 6A). In contrast, NFKBIA was positively correlated with activated CD8 T cells, regulatory T cells, activated CD4 T cells, effector memory CD8 T cells, and central memory CD8 T cells (Fig. 6B). Additionally, HP was positively correlated with neutrophils, VCAM1 was negatively correlated with natural killer T cells, while LCN2, CXCL1 and CCL3 showed no significant correlation with any immune cell type (Supplementary Fig. 2).

Fig. 6.

(A) and (B) represent the correlation analysis of TNFRSF1A and NFKBIA with 28 immune cell types, respectively, with p < 0.05 considered statistically significant.

Validation of hub genes

The GSE124114, GSE37374 and GSE65240 dataset was used to validate the expression levels of the identified hub genes (Fig. 7A). The results showed that CXCL1 exhibited expression trends consistent with those observed in the GSE27276 dataset, whereas NFKBIA and VCAM1 displayed an opposite trend. The expression levels of the remaining hub genes showed no significant differences. Furthermore, ROC analysis of hub genes in the GSE124114, GSE37374 and GSE65240 dataset (Fig. 7B) revealed that NFKBIA, VCAM1, and CXCL1 had AUC values > 0.7, suggesting that these genes have strong diagnostic potential for glaucoma.

Fig. 7.

(A) Expression of 7 hub genes in GSE124114, GSE37374 and GSE65240. ns, no statistical significance. (B) The ROC curve of validation dataset GSE124114, GSE37374 and GSE65240 confirmed the diagnostic importance of key genes.

The TF and microRNA network of hub genes

TFs and microRNAs are critical regulatory elements in gene transcription and translation. To further elucidate the regulatory mechanisms governing hub gene expression, we predicted the TFs and microRNAs that modulate hub genes (Supplementary Table 5 and Supplementary Table 6).

In the TF regulatory network of hub genes (Fig. 8A), a total of 24 TFs were identified to regulate six hub genes. Among them, RELA and NFKB1 were the most frequently involved TFs, whereas VCAM1 was the hub gene regulated by the highest number of TFs. Notably, TNFRSF1A had no associated TF information.

Fig. 8.

(A) The transcription factor regulatory network of hub genes. The pink nodes represent hub genes, the green squares represent TFs, and the connections indicate TFs regulating the hub genes. (B) The microRNAs that regulate at least 5 hub genes. The pink nodes represent hub genes, the blue squares represent microRNA, and the connections indicate microRNA regulating the hub genes.

In the microRNA regulatory network, CCL3, CXCL1, NFKBIA, VCAM1, LCN2, TNFRSF1A, and HP were regulated by 17, 68, 153, 116, 48, 100, and 21 microRNAs, respectively. Figure 8B illustrates microRNAs that regulate at least five hub genes, with hsa-miR-34a-5p identified as the most prominent microRNA regulating multiple hub genes. The detailed gene-microRNA regulatory network is provided in Supplementary Fig. 3.

The predicted drugs of hub genes

Using the DGIdb, we analyzed potential drug interactions with hub genes to identify possible therapeutic agents for glaucoma. A total of 35 potential drugs interacting with six hub genes were identified (Fig. 9), while no drug interactions were found for LCN2. Supplementary Table 7 provides detailed information on interaction scores, regulatory approval status, and other relevant data.Among these interactions, CCL3 with Nagrestipen and TNFRSF1A with GSK-1995057 exhibited the highest interaction scores of 17.4. However, these drugs have not yet been approved, indicating their potential for further investigation. Additionally, Infliximab was found to interact with both CCL3 and NFKBIA, while Vitamin B6 demonstrated a moderate interaction with HP, suggesting their possible therapeutic role in glaucoma.

Fig. 9.

The drug prediction network of hub genes. The pink nodes represent hub genes, the yellow squares represent drug, and the connections between them indicate that the drug may act on the corresponding gene.

Discussion

The pathogenesis of glaucoma is complex, and IOP-lowering therapy remains the only proven effective treatment^1,50. As a critical component of the aqueous humor outflow pathway, dysfunction of the TM plays a crucial role in elevated IOP⁵¹. Studies have revealed severe oxidative stress damage in the TM tissues of glaucoma patients, where excessive ROS lead to TM degeneration and remodeling of the ECM, resulting in impaired aqueous humor outflow and elevated IOP^6,52.Therefore, understanding the molecular basis of OS in the glaucomatous TM is crucial for controlling IOP and slowing disease progression. Additionally, glucocorticoid use can induce IOP elevation and even trigger glaucoma⁵³, making glucocorticoid-induced TM damage a widely used model system for studying glaucoma-related TM injury⁵⁴. A Extensive research has shown that the core mechanism of glucocorticoid-induced glaucoma lies in its disruption of TM structure and function^55,56. Based on this, the glucocorticoid-induced TM injury dataset (GSE124114, GSE37374 and GSE65240) was selected in this study as a validation gene set.

In this study, a total of 656 DEGs (including 309 upregulated and 347 downregulated genes) were identified from the POAG dataset (GSE27276). By intersecting these DEGs with the OS gene set, 61 DEOSGs were obtained. GO and KEGG enrichment analyses revealed that these DEOSGs were primarily enriched in oxidative stress response, antioxidant activity, and immune regulation-related biological functions and molecular pathways. Further PPI analysis identified seven hub genes: TNFRSF1A, CXCL1, CCL3, NFKBIA, VCAM1, LCN2, and HP. These hub genes were significantly enriched in inflammatory and immune regulatory pathways, suggesting a synergistic role of oxidative stress and immune microenvironment dysregulation in glaucoma pathogenesis. Notably, CXCL1 was consistently downregulated in both the training dataset (GSE27276) and validation dataset (GSE124114, GSE37374 and GSE65240). Additionally, ROC analysis demonstrated the high diagnostic performance, indicating its potential as biomarker for oxidative stress-induced damage in glaucoma.

The results of ssGSEA and immune cell correlation analysis indicated that the expression of TNFRSF1A was significantly negatively correlated with multiple pro-inflammatory immune cell types (Fig. 6A). In contrast, NFKBIA showed positive correlations with several subsets of adaptive immune cells (Fig. 6B). These findings are consistent with previous in vitro studies demonstrating that NF-κB is an essential mediator of TGFβ2-induced ECM production and ocular hypertension⁵⁷. In our study, although derived from transcriptome-level bioinformatics analysis, the consistent correlation patterns between TNFRSF1A/NFKBIA and immune infiltration mirror the in vitro findings, suggesting that TM inflammation in glaucoma may be partially mediated through TNF-NFκB-driven transcriptional reprogramming. Collectively, our results propose a potential regulatory axis of oxidative stress, TNFRSF1A, NFKBIA signaling, and immune cell infiltration in the glaucomatous TM microenvironment.

TNFRSF1A (Tumor Necrosis Factor Receptor Superfamily Member 1A, also known as TNFR1) is a primary receptor for tumor necrosis factor-alpha (TNF-α)⁵⁸. TNF-α exerts its opposing biological effects through two receptors: TNF receptor I (TNFR1) and TNF receptor II (TNFR2). TNFR1 predominantly mediates pro-inflammatory and pro-apoptotic signaling pathways, whereas TNFR2 is associated with neuroprotection and anti-apoptotic signaling⁵⁹. Notably, TNFR1 blockade has been shown to promote the pro-survival effects of TNF in acute inflammation, but its deficiency in chronic inflammation may exacerbate inflammatory damage⁶⁰. While TNF-TNFR signaling has been extensively studied in inflammation-related research, most glaucoma studies have focused on its role in retinal layers and the optic nerve head⁶¹, with limited attention to the TM. In this study, we observed that TNFRSF1A expression was significantly reduced in the TM tissues of POAG patients, which may be associated with chronic inflammation in POAG. This down-regulation was further validated in a glucocorticoid-induced glaucoma TM model, demonstrating a high diagnostic value (AUC = 0.844). Additionally, the down-regulation of TNFRSF1A in glucocorticoid-induced glaucoma damage may be related to the anti-inflammatory properties of glucocorticoids⁶². However, the precise role of TNFRSF1A in OS-induced TM damage in glaucoma requires further experimental validation.

CXCL1 is a crucial chemokine primarily involved in recruiting immune cells and participating in various inflammatory diseases⁶³. Studies indicate that CXCL1 is essential for retinal leukocyte recruitment and ischemia/reperfusion-induced retinopathy⁶⁴. Shuai Wang et al. found that selective blockade of CXCL1-CXCR2 activation ameliorated angiotensin II-induced retinopathy, and this beneficial effect may correlate with reduced infiltration of CXCR2 + immune cells (particularly macrophages) and decreased expression of pro-inflammatory cytokines (IL-1β, IL-6, TNF-α, and MCP-1)⁶³. Previous research on CXCL1 in glaucoma is limited. Jeffrey Freedman et al. reported elevated CXCL1 levels in the aqueous humor of glaucoma patients, though not statistically significant⁶⁵. In this study, reduced CXCL1 expression was observed in the TM of POAG patients. We hypothesize that this down-regulation may be linked to compensatory mechanisms or severe TM cell loss in advanced glaucoma stages⁶⁶, leading to overall decreased CXCL1 expression.

CCL3 (MIP-1α, macrophage inflammatory protein-1-alpha), another key member of the chemokine family, primarily mediates immune cell migration, inflammatory regulation, and tissue repair by binding to chemokine receptors such as CCR1 and CCR5⁶⁷. Garweg JG et al. found significantly elevated levels of CCL3 and other chemokines in the aqueous humor of patients with pseudoexfoliative glaucoma⁶⁸. Similarly, Burgos-Blasco et al.⁶⁹ reported increased levels of inflammatory factors, including CCL3, in the aqueous humor of POAG patients. However, a recent meta-analysis integrating five studies revealed no significant difference in CCL3 levels between POAG patients and controls⁷⁰. The relationship between chemokines and glaucoma pathogenesis, as well as their underlying mechanisms, still requires extensive experimental and clinical investigation to elucidate.

Haptoglobin (HP) alleviates iron toxicity and OS caused by free hemoglobin through irreversible binding⁷¹. Down-regulation of HP leads to the accumulation of free hemoglobin and iron ions⁷². Free iron catalyzes the generation of ·OH via the Fenton reaction, activating the ferroptosis and exacerbating mitochondrial dysfunction and lipid peroxidation⁷³. HP inhibits ECM deposition and regulates fibroblast migration, both essential for wound healing and repair^71,74. Therefore, Hp down-regulation may serve as a critical contributor to ECM deposition and tissue remodeling in the TM.

Lipocalin-2 (LCN2) is an acute-phase protein⁷⁵ with iron-chelating, iron-transporting, and antioxidant properties^76–79. Studies have shown that LCN2 expression is up-regulated in various neurodegenerative diseases^80,81 and tumors^82,83, and it is positively correlated with disease severity and poor prognosis. We propose that the increased expression of LCN2 in neurodegenerative diseases, which correlates with disease severity, may represent a defensive response of the body to injury. Currently, there are limited reports on LCN2 in glaucoma. Azusa Yoneshige et al. found that LCN2 expression is up-regulated in the retina of glaucoma patients⁸⁴. However, Yutao Liu et al. conducted transcriptomic sequencing on 35 human TM tissues and observed significant down-regulation of LCN2 expression in TM tissues⁸⁵. This discrepancy may be related to tissue specificity or disease progression stages. In conclusion, this study suggests that down-regulation of LCN2 in glaucoma TM tissues leads to functional dysregulation, which exacerbates oxidative stress and inflammatory responses by disrupting iron metabolism.

Vascular Cell Adhesion Molecule 1 (VCAM1), a member of the immunoglobulin superfamily, is primarily expressed in activated vascular endothelial cells, macrophages, and certain tumor cells. Its core functions involve mediating intercellular adhesion and regulating immune cell migration, inflammatory responses, and disease progression⁸⁶. The up-regulation of VCAM1 is closely associated with oxidative stress-induced endothelial cell activation⁸⁷. A genome-wide meta-analysis identified VCAM1 as a high-risk gene for primary open-angle glaucoma⁸⁸. In aqueous humor, VCAM1 may serve as a biomarker for predicting the progression of diabetic retinopathy, while it is over-expressed in the iris of uveitis patients^89,90. Its negative correlation with natural killer T cells may suggest an immune regulatory imbalance, leading to insufficient anti-inflammatory responses.

Subsequently, we predicted TFs, microRNAs, and potential therapeutic drugs regulating the hub genes, providing researchers with a multidimensional perspective to understand the mechanisms of hub genes in glaucoma. In the TF regulatory network, RELA (NF-κB p65 subunit) and NFKB1 (NF-κB p50 subunit) were identified as the TFs regulating the most hub genes, further validating the central role of the NF-κB signaling pathway in oxidative stress and inflammation in glaucoma. RELA/NFKB1 drives inflammatory cell infiltration and ECM deposition⁹¹ in the TM by activating the transcription of pro-inflammatory genes such as CCL3 and VCAM1^92,93, thereby exacerbating aqueous humor outflow resistance. Additionally, VCAM1 is the hub gene regulated by the largest number of TFs, suggesting its expression is co-regulated by multiple signaling pathways and may act as a pivotal player in the oxidative stress-inflammation vicious cycle of the TM.

In the microRNA regulatory network, hsa-miR-34a-5p was identified as the microRNA regulating the largest number of hub genes. miR-34a-5p exacerbates oxidative stress damage and inflammatory responses by targeting SIRT1^94,95, and its upregulation may aggravate TM cell injury in glaucoma^96,97. Additionally, NFKBIA is regulated by 153 microRNAs, far exceeding other genes, indicating that its expression is subject to highly refined post-transcriptional regulation.

Through screening via the DGIdb database, 35 potential drugs were identified. Some approved drugs have shown moderate interactions with hub genes, demonstrating direct translation potential. Infliximab, a TNF-α monoclonal antibody, is currently primarily used in the clinical treatment of inflammatory bowel disease⁹⁸. In ophthalmic diseases, it is mainly employed to control ocular neovascularization and intraocular inflammation⁹⁹. Glaucoma, as an inflammation-related disease², holds significant research and clinical value for the future application of Infliximab as a therapeutic agent. Vitamin B complex, an important antioxidant, may be linked to symptoms of optic nerve dysfunction due to its deficiency¹⁰⁰. A recent study suggests that Vitamin B6 supplementation is associated with reduced glaucoma prevalence in males¹⁰¹, possibly due to its antioxidant properties in maintaining iron metabolism balance and inhibiting lipid peroxidation¹⁰². Additionally, the interaction between glucocorticoids and VCAM1 requires cautious interpretation, as prolonged use may induce elevated intraocular pressure, underscoring the complexity of targeted therapeutic strategies.

Our study revealed that the hub genes associated with oxidative stress in the glaucomatous TM are also closely linked to inflammatory responses. The pathological progression of glaucoma centers on the dynamic interplay between oxidative stress and inflammation, which synergistically exacerbate TM dysfunction and optic neurodegeneration¹⁰³. OS directly impairs mitochondrial function in TM cells through excessive accumulation of ROS, while activating key inflammatory signaling pathways to promote the release of pro-inflammatory factors¹⁰⁴. ROS induces oxidative modifications of lipids, proteins, and DNA, leading to abnormal deposition of ECM in the TM and increased resistance to aqueous humor outflow⁹¹. Concurrently, inflammatory responses recruit immune cells to the TM, releasing proteases and cytokines that further disrupt ECM homeostasis and amplify oxidative stress signaling^105,106.

Conclusion

This study systematically identified hub genes associated with oxidative stress in the glaucomatous TM (TNFRSF1A, CXCL1, CCL3, NFKBIA, VCAM1, LCN2, HP) through integrated bioinformatics analysis and validation. Enrichment analysis of these hub genes revealed a critical mechanism by which OS and inflammation synergistically impair TM function. Furthermore, validation using external datasets combined with ROC curve analysis demonstrated the diagnostic accuracy of these hub genes, with TNFRSF1A and CXCL1 potentially playing significant roles in regulating oxidative stress in the glaucomatous TM. Subsequently, we constructed transcriptional factor (e.g., RELA/NFKB1) and microRNA (e.g., hsa-miR-34a-5p) regulatory networks for the hub genes and screened potential therapeutic agents such as Infliximab and Vitamin B6, providing novel strategies for targeted interventions. These findings may offer early diagnostic biomarkers for oxidative stress damage in the glaucomatous TM and lay the foundation for potential therapeutic approaches in glaucoma treatment.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, W.Y.X. and F.S.Z.; Methodology, W.Y.X.; Formal analysis, F.S.Z.; resources, W.Y.X.; investigation, W.Y.X.; visualization, F.S.Z.; software, W.Y.X.; validation, F.S.Z; funding acquisition, J.F.H.; supervision, J.F.H.; project administration, J.F.H.; writing—original draft, W.Y.X. and F.S.Z.; writing—review and editing, W.Y.X. and J.F.H.; All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Key Research and Development Program of Hunan Province (2022SK2079).

Data availability

The datasets used in this study are publicly available in the GEO database (GSE27276, GSE124114, GSE37374 and GSE65240). The R scripts used for data analysis are available at the following GitHub repository: https://github.com/xuweiye1996/glaucoma-oxidative-stress-analysis.

Declarations

Competing interests

The authors declare no competing interests.