Screening and identification of genes related to ferroptosis in keratoconus

Abstract

Corneal keratoconus (KC) is a dilated (ectatic) corneal disease characterized by a central thinning of the cornea, which causes protrusion into a conical shape that seriously affects vision. However, due to the complex etiology of keratoconus, its entire mechanism remains unclear and there is no mechanism-directed treatment method. Ferroptosis is a novel programmed cell death mechanism related to lipid peroxidation, stress, and amino acid metabolism, which plays a crucial role in various diseases. This study aimed to explore the relationship between keratoconus and ferroptosis, to provide new insights into the mechanism of keratoconus development, and potential treatment options based on further elucidation of this mechanism. The corresponding mRNA microarray expression matrix data of KC patients were obtained from GEO database (GSE204791). Weighted co-expression network analysis (WGCNA) and support vector machine recursive feature elimination (SVM-RFE) were selected to screen hub genes, which were overlapped with ferroptosis genes (FRGs) from FerrDb. GO and GSEA were performed to analyze differential pathways, ssGSEA was used to determine immune status, and then, feasible drugs were predicted by gene-drug network. Additionally, we predicted the miRNA and IncRNA of hub genes to identify the underlying mechanism of disease so as to predict treatment for the disease. The epithelial transcriptome from keratoconus tissue mRNA microarray data (GSE204791) was extracted for the main analysis, including eight epithelial cells and eight epithelial control cells. The differential genes that were overlapped by WGCAN, SVM-RFE and FRGs were mainly related to oxidative stress, immune regulation, cellular inflammation, and metal ion transport. Through further analysis, aldo-keto reductase family 1 member C3 (AKR1C3) was selected, and negatively correlated with mature CD56 natural killer (NK) cells and macrophages. Then, gene-drug interaction network analysis and miRNA prediction were performed through the website. It was concluded that four immune-related drugs (INDOMETHACIN, DAUNORUBICIN, DOXORUBICIN, DOCETAXEL) and a miRNA (has-miR-184) were screened to predict potential drugs and targets for disease treatment. To our knowledge, this was the first report of KC being associated with ferroptosis and prompted search for differential genes to predict drug targets of gene immunotherapy. Our findings provided insight and a solid basis for the analysis and treatment of KC.

Figure 1

The workflow chart in this study.

Figure 2

Representative results of GSEA analysis in the expression data of EN and EKC. (A) A significant gene set negatively correlated with EKC group was Cytokine-Cytokine Receptor Interaction (NES = − 1.839, P.adj < 0.001). (B) Overview of proinflammatory and profibrotic Mediators was negatively correlated with EKC (NES = − 1.567, P.adj < 0.001). (C) Ferroptosis-related genes are negatively correlated with keratoconus (NES = − 1.566, P.adj = 0.007). NES, normalized enrichment score.

Figure 3

Ferroptosis related genes in the central gene module (p < 0.05, cor ≥ 0.5). (A-C) Gene significance versus Module Membership (MM) in black, brown, and magenta modules. (D) Venn diagram screened ferroptosis genes in 3 gene modules.

Figure 4

SVM—RFE filtered characteristic genes. (A, B) When n = 14, 10xCV Accuracy was 0.95, and the classifier had the minimum error (0.05).

Figure 5

GO analysis of ferroptosis genes. (A) GO analysis of differentially expressed genes screened by WGCNA. (B) GO analysis of differentially expressed genes screened by SVM—RFE. (C) Venn diagram shows intersection genes.

Figure 6

The differential gene expression and Receiver operating characteristic analysis (ROC). (A) The mRNA expression of AKR1C3. (B) The mRNA expression of SLC7A11. (C) The AUC analysis of AKR1C3. Receiver operating characteristic, ROC; AUC, area under the ROC curve.

Figure 7

ssGSEA analysis. (A, B) Differential identification of immunophenotypic cells. (C) Immune subtype cells affected by AKR1C3. *p < 0.05.

Figure 8

Gene-drug therapy and mRNA target prediction. (A) Prediction of gene-targeted drugs, using the website (https://dgidb.genome.wustl.edu/). (B) miRNA and IncRNA combined with characteristic genes.