Bioinformatics-based identification of ferroptosis-related biomarkers and immune infiltration in retinoblastoma

Introduction

Retinoblastoma (RB) is the most common primary intraocular malignancy in children (1). Surgical excision is generally performed when tumors present with high-risk clinical features. RB is primarily driven by mutations in the tumor suppressor gene RB1, which encodes the retinoblastoma protein (pRB) (2). Mortality rates remain high in developing countries, where approximately 55% of patients undergo enucleation. In contrast, in developed nations, timely chemotherapy combined with surgery increases survival rates to 95–98% (3,4). Recent advances, such as intra-arterial chemotherapy and intravitreal injection of chemotherapeutic agents, have reduced enucleation rates (5), but these therapies may still induce tumor metastasis and chemotherapy-related toxicities. The molecular mechanisms underlying RB pathogenesis remain incompletely understood.

Ferroptosis is a regulated, iron-dependent form of cell death characterized by phospholipid peroxidation, accumulation of reactive oxygen species (ROS), and polyunsaturated fatty acid-containing phospholipids (PUFA-PLs). Unlike apoptosis or necrosis, ferroptosis is not blocked by conventional inhibitors of these pathways, but can be mitigated by radical-trapping antioxidants (RTAs) such as ferrostatin-1 (Fer-1) (6). RB1 deletion and TP53 mutations are critical in RB pathogenesis, and both are implicated in ferroptosis regulation. Recent studies suggest that drug-tolerant RB cells can be eliminated by inducing autophagy-dependent ferroptosis (7), and that lipid peroxidation-mediated oxidative cell death may represent a promising strategy against drug-resistant RB (8). These findings highlight a potential role for ferroptosis in RB initiation and progression.

The tumor microenvironment (TME) encompasses non-malignant cells and acellular components within the tumor niche. Cytokine secretion induced by oncogenic alterations influences both malignant and stromal cells, including immune cells, fibroblasts, and vasculature. Interactions between RB cells and the TME are critical for tumor initiation, progression, metastasis, and therapeutic response (9).

Bioinformatics approaches provide powerful tools to identify candidate biomarkers in RB. In this study, we analyzed the Gene Expression Omnibus (GEO) dataset to identify differentially expressed genes (DEGs) between RB and healthy tissues. Ferroptosis-related DEGs were obtained by intersecting DEGs with a curated ferroptosis gene set. Key genes were further investigated through single-gene gene set enrichment analysis (GSEA), gene set variation analysis (GSVA), immune infiltration profiling, and drug sensitivity prediction. A schematic overview of the study design is shown in Figure 1. Assessing immune cell subtype distribution in RB tissues may enhance understanding of RB progression and enable the development of more precise diagnostic and therapeutic strategies. Collectively, this study provides new molecular insights into RB progression and identifies potential ferroptosis-related therapeutic targets. We present this article in accordance with the TRIPOD and MDAR reporting checklists (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2568/rc).

Figure 1 Schematic diagram showing the study design. GSEA, gene set enrichment analysis; GSVA, gene set variation analysis; LASSO, least absolute shrinkage and selection operator; RB, retinoblastoma.

Methods

Date collection

Data on RB patients were obtained from the GEO database (http://www.ncbi.nlm.nih.gov/geo), a public repository of gene expression datasets. The raw data of GSE166173 were downloaded for single-cell analysis, and four samples with complete expression profiles were included. For the single-cell dataset GSE166173, this dataset contains RB tissue samples, where the “normal” cells refer to non-malignant cell types identified through unsupervised clustering [such as cone cells, retinal pigment epithelial (RPE) cells, etc.]. These cells were used as intra-tumoral normal references for cell communication and trajectory analysis. For bulk transcriptome analysis, three datasets were used: GSE110811 (training set; 3 normal retinal tissue samples and 28 RB samples), GSE156657 (3 normal retinal tissue samples and 7 RB samples), and GSE97508 (3 normal retinal tissue samples and 6 RB samples). This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Identification of RB cell marker genes by single-cell RNA sequencing (scRNA-seq) analysis

scRNA-seq data were analyzed using the R packages Seurat and SingleR. Data were normalized with the NormalizeData function in Seurat using the “LogNormalize” method, and a Seurat object was created from the normalized matrix. Principal component analysis (PCA) was performed on the top 1,500 variable genes using the RunPCA function. Significant principal components (PCs) were identified by JackStraw analysis, and the first 15 PCs were selected for downstream analyses. Cell clustering was performed with the FindNeighbors and FindClusters functions in Seurat, where Euclidean distance was used to construct a k-nearest neighbor graph. Dimensionality reduction was visualized using t-distributed stochastic neighbor embedding (t-SNE) with the RunTSNE function. DEGs for each cluster were identified with the FindAllMarkers function using the Wilcoxon rank-sum test, applying thresholds of |log₂ [fold change (FC)]| >1 and adjusted P<0.01. Cluster annotation was performed using marker information from the CellMarker and PanglaoDB databases, supplemented by literature review to assign cell identities.

Model construction

Candidate genes were selected and least absolute shrinkage and selection operator (LASSO) regression was used to further build prediction correlation models. After incorporating the expression value of each specific gene, a risk score formula for each patient was constructed and weighted by its estimated regression coefficient in a LASSO regression analysis. According to the risk score formula, with the median risk score value as the cut-off point, patients were divided into low-risk groups and high-risk groups, and the receiver operating characteristic (ROC) curve was used to study the accuracy of model prediction.

Functional enrichment analysis of the ferroptosis-related genes (FRGs) (GSVA + GSEA)

GSEA was performed using the R package clusterProfiler, with significance defined as P<0.05. GSVA was conducted with the GSVA algorithm to systematically evaluate each gene set and identify potential functional alterations across samples.

Immune cell infiltration analysis

Immune cell infiltration in RB tissues was compared with healthy controls. Infiltration levels of 22 immune cell types were estimated using the CIBERSORT algorithm. Correlations among immune cell populations were visualized with the corrplot package. The Wilcoxon rank-sum test was applied to identify statistically significant differences in immune cell infiltration between groups. Visualization of infiltration patterns was conducted using the ggplot2, vioplot packages. Finally, Spearman’s rank correlation analysis was used to examine associations between FRGs and immune cell infiltration levels.

Drug sensitivity analysis

The Genomics of Drug Sensitivity in Cancer (GDSC) database, developed by the Sanger Institute (UK), was used to obtain information on tumor cell drug sensitivity. The R package pRRophetic was applied to predict chemotherapy sensitivity of each tumor sample by estimating the half-maximal inhibitory concentration (IC₅₀) values through regression analysis.

Quantitative real-time polymerase chain reaction (qRT-PCR)

Total RNA was extracted from RB (Y79 line) and RPE (ARPE-19 line) cells using TRIzol reagent (Takara). For the culture and passaging details of the two cells, please see Appendix 1. cDNA was synthesized with the PrimeScript RT Master Mix Kit with gDNA Eraser (Takara, RR047A). qRT-PCR was performed and on a LightCycler 480 system (Roche, Penzberg, Germany) using SYBR Green Master Mix (for more details about qRT-PCR methodology, please see Appendix 1). Relative mRNA expression was calculated using the 2^−ΔΔCT method, with GAPDH as the internal control. Primers for GAPDH, JUN, HSPB1, and NFE2L2 are listed in Table S1 (accurate biology).

Statistical analysis

All statistical analyses were performed using R software (version 4.2.0) and GraphPad Prism 9.0. DEGs were identified using the Wilcoxon rank-sum test with thresholds of |log₂(FC)| >1 and adjusted P<0.01. LASSO regression was conducted using the glmnet package with 10-fold cross-validation to select optimal lambda. ROC curves were generated using the pROC package, and area under the curve (AUC) values were calculated. Correlations between gene expression and immune cell infiltration or metastasis-related genes were assessed using Spearman’s rank correlation coefficient. For qRT-PCR data, comparisons between two groups were analyzed using two‑tailed Student’s t-test. A P value <0.05 was considered statistically significant.

Results

Basic profiling for single cell data

Single cell level analysis in scRNA-Seq data

We processed the expression profiles using the Seurat package and filtered out low-quality cells based on the thresholds nFeature_RNA >500, nCount_RNA >1,000, percent.mt <15, nFeature_RNA <6,000, and nCount_RNA <25,000. After quality control, a total of 10,137 cells were retained. We visualized the filtering results by violin plots and scatterplots (Figure 2A,2B). We then performed data standardization, normalization, PCA, and Harmony analysis, which demonstrated clear clustering and batch correction across samples (Figure 2C-2E).

Figure 2 Single cell level analysis in scRNA-seq data. (A) Single-cell quality control showing number of cells, number of genes, and sequencing depth for each sample. (B) The left panel shows the relationship between cell sequencing depth and mitochondrial content, and the right panel shows the relationship between sequencing depth and gene number, which are positively correlated. (C) Plot of variance ordering for each. (D,E) Display of PCA and Harmony, dots represent cells and colors represent samples. PC, principal component; PCA, principal component analysis; scRNA-seq, single-cell RNA sequencing.

Next, we performed t-SNE analysis to determine the positional relationships among cells and identified 16 distinct clusters. These clusters were further classified into eight major cell types (Figure 3A,3B). We compared the proportions of these cell types between normal and RB groups (Figure 3C). By integrating differences in cell number and gene expression, we characterized the contribution of specific subpopulations to RB pathogenesis.

Figure 3 Cell typing. (A) We categorized the cells into 16 clusters by tSNE algorithm based on the important components available in the PCA. The 16 clusters were annotated into 8 cell types, i.e., endothelial cell, corneal epithelial cell, microglia cell, NK cell, retinal pigment epithelial cell, fibroblast cell, cycling cone cell and cone cell. (B) Bubble plot of the 8 cell types with dotplot of cell markers. (C) Difference in the percentage of content of the 8 cells in the two groups of samples. NK, natural killer; NKT, natural killer T; PCA, principal component analysis; tSNE, t-distributed stochastic neighbor embedding.

Annotation of cell subpopulations and their contribution to RB

We identified the top 266 DEGs (logFC >1) between RB and control groups to characterize transcriptional changes during this process. Cone cells were found to contribute most prominently to RB progression (Figure 4A,4B). We further reconstructed the developmental trajectories of different cone cell subtypes (Figure 4C-4E).

Figure 4 Annotation of cell subpopulations and their contribution to RB. (A) Cellular interaction networks between the 8 types of cells. (B) Comparison of the total number of interactions in the communication network between the 8 cell types, decreasing from left to right, with the strongest being the cone cell. (C-E) The developmental trajectory of different cone cell subtypes. NKT, natural killer T; RB, retinoblastoma.

Identification of key genes and related analysis

Construction of prediction model and identification of key genes

The marker genes of cone cells (logFC >1) were selected to intersect with FRGs, and a total of 15 intersection genes were obtained (Figure 5A). We downloaded the RB-related dataset GSE110811 from the GEO database as the training set, and used datasets GSE156657 and GSE97508 as validation sets. LASSO regression was performed to screen for characteristic features among the 15 intersecting genes. The results showed that LASSO regression identified three genes as feature genes (Figure 5B-5D), which were designated as key genes for subsequent studies and used to construct a predictive model. The model formula is: RiskScore = NFE2L2 × (−0.0532820821590078) + HSPB1 × (0.0054879682205842) + JUN × (0.0494556842030953). The results demonstrated that the predictive model constructed with the three genes had good diagnostic efficacy, with an area under the AUC curve of 0.95 (Figure 5E). External datasets were used to further validate the diagnostic model, and the results showed that the model possessed strong stability, with AUC values of 0.86 and 0.89 for the two validation sets, respectively (Figure 5F,5G).

Figure 5 Construction of prediction model and identification of key genes. (A) Wayne plots of iron death differential genes with cone cell marker genes. (B) Distribution of LASSO coefficients and gene combinations at the minimum lambda value. (C) Ten-fold cross-validation of tuning parameter choices in the LASSO model to determine the minimum lambda value. (D) LASSO regression identified three genes, JUN, HSPB1 and NFE2L2, as characterized genes. (E) The area of the AUC curve for the three key genes is 0.95. (F,G) The areas of the AUC curves for the two validation sets are 0.86 and 0.89, respectively. AUC, area under the curve; FRG, ferroptosis-related gene; GEO, Gene Expression Omnibus; LASSO, least absolute shrinkage and selection operator.

Analysis of key genes GSEA and GSVA

GSEA revealed that NFE2L2 was enriched in pathways related to DNA replication, necroptosis, and NOD-like receptor signaling (Figure 6A,6B). HSPB1 was mainly associated with cytokine-receptor interaction, neuroactive ligand-receptor interaction, and phototransduction (Figure 6C,6D). JUN was linked to DNA replication, the relaxin signaling pathway, and retinol metabolism (Figure 6E,6F).

Figure 6 Analysis of key genes GSEA. (A-F) KEGG signaling pathways involved in NFE2L2, JUN, and HSPB1, as well as pathway regulation and genes involved. GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes.

GSVA further showed that high HSPB1 expression was enriched in the ROS pathway and allograft rejection (Figure 7A). Elevated JUN expression was associated with angiogenesis and Wnt/β-catenin signaling (Figure 7B), while NFE2L2 was enriched in PI3K-AKT-mTOR signaling and downregulation of ultraviolet (UV) response pathways (Figure 7C). These findings suggest that the three key genes may contribute to RB progression through distinct signaling mechanisms.

Figure 7 Analysis of key genes GSVA. (A-C) GSVA analysis of NFE2L2, JUN, and HSPB1, with signaling pathways involved in high expression of genes expressed in blue, and signaling pathways involved in low-expressed genes indicated in green, and the background gene set is hallmark. GSVA, gene set variation analysis.

The correlation between key genes and metastasis-related genes

Aberrant activation of the RAS signaling pathway is closely associated with tumorigenesis and metastasis in multiple cancers. To explore its relevance in RB, we obtained RAS pathway-related genes from the GSEA database (https://www.gsea-msigdb.org/gsea/index.jsp) and analyzed their correlations with the key genes. NFE2L2 showed significant positive correlations with CHUK and RHOA, but a negative correlation with HRAS. HSPB1 was positively correlated with RALA and RHOA, while JUN demonstrated a positive correlation with RALGDS (Figure 8).

Figure 8 The correlation between key genes and metastasis-related genes. (A-F) Correlation analysis of NFE2L2, JUN, and HSPB1 with CHUK, HRAS, RALA, RHOA, RALGDS.

Drug sensitivity

Early-stage RB is commonly managed with surgery combined with chemotherapy, and assessing drug sensitivity may help optimize therapeutic efficacy. Using the R package pRRophetic with drug sensitivity data from the GDSC database, we predicted chemotherapy responses in each tumor sample and examined associations between key genes and commonly used chemotherapeutic agents. NFE2L2 expression was significantly correlated with sensitivity to ABT-888, AICAR, and AUY922 (Figure 9A). HSPB1 expression was associated with sensitivity to AG-014699 (Figure 9B), while JUN expression correlated with sensitivity to AKT inhibitor VIII and AMG-706 (Figure 9C). It should be noted that since the drug sensitivity predictions are based on the GDSC database, which mainly contains drug sensitivity data from other tumor types and may not be fully applicable to RB. Therefore, the predicted results should be interpreted with caution and require validation using RB-specific cell lines or organoid models.

Figure 9 Key genes and sensitivity to common chemotherapy drugs. NFE2L2, JUN, and HSPB1 sensitivity to common chemotherapy drugs. (A) NFE2L2 was significantly related to the sensitivity of ABT.888, AICAR, and AUY922. (B) HSPB1 was significantly related to the sensitivity of AG.014699. (C) JUN was significantly related to the sensitivity of AG.014699.

Link between key genes and TME

Immune cell infiltration analysis

The distribution of immune infiltration levels and intercellular relationships was visualized by heatmaps (Figure 10A,10B). Notably, the proportions of CD8⁺ T cells and activated mast cells were significantly increased in RB samples compared with controls (Figure 10C). We further examined the correlations between key genes and immune cell populations. NFE2L2 showed strong positive correlations with naïve B cells, M0 macrophages, M2 macrophages, naïve CD4⁺ T cells, and CD8⁺ T cells. HSPB1 was positively correlated with M1 macrophages, resting NK cells, neutrophils, and resting memory CD4⁺ T cells, while displaying a negative correlation with memory B cells (Figure 10D).

Figure 10 Immune cell infiltration analysis. (A) Relative percentages of the 22 immune cell subpopulations. (B) Pearson’s correlation between the 22 immune cells, with blue indicating a negative correlation and red indicating a positive correlation. (C) Differences in immune cell content between control and disease samples. (D) Correlation of NFE2L2, JUN, and HSPB1 with immune cells. *, P<0.05; **, P<0.01; ***, P<0.001; ns, not significant. NK, natural killer.

We further analyzed the associations between the three key genes and a spectrum of immune-related factors, including immunosuppressive molecules, immunostimulatory molecules, chemokines, and chemokine receptors. These analyses revealed significant correlations, suggesting that the key genes are closely linked to immune cell infiltration and may play critical roles in shaping the immune microenvironment of RB (Figure 11).

Figure 11 Correlation of key genes with different immune factors. The correlation between NFE2L2, JUN, and HSPB1 different immune factors, including immunosuppressive factors, immunostimulatory factors, chemokines, and receptors. MHC, major histocompatibility complex.

The relationship between key genes and immune metabolic pathways

We next examined the expression abundance of the three key genes (NFE2L2, HSPB1, and JUN) across multiple cell types, including cone cells, endothelial cells, corneal epithelial cells, microglia, natural killer T (NKT) cells, RPE cells, fibroblasts, and cycling cone cells (Figure 12A-12C). We further analyzed their roles in immunometabolic regulation at the single-cell level (Figure 12D). The results showed that all three genes positively influenced immunity, metabolism, and signal transduction pathways, while exhibiting negative associations with cell proliferation pathways.

Figure 12 The relationship between key genes and immune pathways. (A-C) Differential expression of NFE2L2, JUN, and HSPB1 in 8 cell types demonstrated by t-SNE, bubble and violin plots. (D) Correlation of NFE2L2, JUN, and HSPB1 with immunometabolic pathways. FC, fold change; FDR, false discovery rate; NKT, natural killer T; t-SNE, t-distributed stochastic neighbor embedding.

qRT-PCR results

Reverse transcription-quantitative real-time polymerase chain reaction (RT-qPCR) analysis confirmed that the mRNA expression levels of NFE2L2, JUN, and HSPB1 were significantly reduced in RB Y79 cells compared with normal RPE cells (Figure 13). Since RB originates from cone precursor cells, the ideal control should be healthy cone cells. However, since primary cone cells are difficult to obtain and cannot be expanded in vitro, we initially selected RPE cells as the control. To more closely approximate the biological context, in subsequent studies we will supplement with another RB cell line, WERI-Rb-1, as well as another normal retinal control cell—human retinal progenitor cells (HRPC) as controls. In future research, we will collect clinical samples to validate the expression of these three genes and their relationship with patient prognosis at the tissue level using techniques such as immunohistochemistry and in situ hybridization.

Figure 13 mRNA expression of three key genes. (A-C) RT-qPCR results showed that the mRNA expression of NFE2L2, JUN and HSPB1 were significantly reduced in RB Y79 cells compared to RPE normal cells (the data are expressed as the mean ± SD. ***, P<0.001). RPE, retinal pigment epithelium; RT-qPCR, reverse transcription-quantitative real-time polymerase chain reaction; SD, standard deviation.

Discussion

RB is a rare pediatric intraocular malignancy with an incidence of approximately 1 in 18,000 live births (10). While early detection and timely treatment can cure the disease, delayed intervention often leads to life-threatening progression. Current therapies—including systemic chemotherapy, enucleation, thermotherapy, and cryotherapy—have improved outcomes in developed countries, yet their efficacy remains limited and patient prognosis is often unsatisfactory (11). According to the Knudson “two-hit” hypothesis, biallelic inactivation of the RB1 gene is the initiating event in RB, and subsequent studies have identified additional molecular pathways and biomarkers that may provide opportunities for targeted intervention (12-14). Thus, further elucidation of RB’s molecular mechanisms is crucial for improving early diagnosis and therapy.

Ferroptosis, an iron-dependent form of regulated cell death characterized by lipid peroxidation, has recently gained attention for its potential to overcome therapy resistance in cancer (15,16) Emerging evidence supports its relevance in ocular diseases (17). For example, ferroptosis induction has been shown to inhibit RB proliferation, enhance chemosensitivity, and overcome multidrug resistance (18-20). Zhang et al. demonstrated that lactate metabolism in the TME promotes RB tumorigenesis and confers resistance to ferroptosis through MCT1-mediated lactate transport, highlighting metabolic vulnerabilities as therapeutic targets (21). Nevertheless, research directly linking ferroptosis to RB remains limited. Our findings contribute to this growing field by identifying three ferroptosis-associated genes—NFE2L2, HSPB1, and JUN—as potential diagnostic biomarkers and regulators of RB progression.

Each of these genes plays diverse roles in cancer biology and ferroptosis. NFE2L2 (Nrf2) is a master regulator of antioxidant responses and detoxification pathways. Persistent activation of NFE2L2 promotes RB1 protein degradation and contributes to tumor progression and drug resistance (22-24). While NFE2L2-mediated ferroptosis regulation has been studied in cataract, diabetic retinopathy, and age-related macular degeneration (25-30), its role in RB remains unexplored. JUN, a transcription factor involved in cell proliferation and stress responses, has been implicated in ferroptosis sensitivity across several cancers (31,32). Although evidence in ocular diseases is scarce, one study suggested JUN as a biomarker in retinal neurodegeneration via ferroptosis (33). HSPB1 (HSP27), expressed in multiple ocular tissues, has protective roles in RB, glaucoma, and cataract (34,35), and functions as a negative regulator of ferroptosis (36,37). Together, these results suggest that NFE2L2, JUN, and HSPB1 may influence RB development by modulating ferroptosis-related signaling networks.

Beyond ferroptosis, our study highlights the interplay between these genes and the RB immune microenvironment. The RB TME consists of diverse immune cells, including dendritic cells, macrophages, and T lymphocytes, which influence tumor growth and therapeutic response (38,39). Recent single-cell transcriptomic analyses by Zhang et al. have systematically characterized TME heterogeneity in RB, revealing distinct subclusters of tumor-associated macrophages (TAMs) and astrocyte-like cells that correlate with invasive phenotypes (40). This study demonstrated that cell-cell communications within the TME, particularly between TAMs and tumor cells, play pivotal roles in RB progression and metastasis. Our results demonstrate that NFE2L2, JUN, and HSPB1 are significantly correlated with immune infiltration patterns and immune-related molecules, suggesting that they may shape the immune landscape of RB. Given that immunotherapy remains underutilized in RB, dissecting these immune interactions could help identify new strategies to enhance anti-tumor immunity. Emerging approaches such as circRNA-based immunomodulation and TAM-directed therapies represent promising frontiers for RB treatment (41,42).

This study provides new insights into ferroptosis and immune regulation in RB; however, several limitations must be acknowledged. First, the analyses relied on publicly available datasets with relatively small sample sizes and variable clinical annotations, which may introduce bias. Second, while our integrative bioinformatics approach identified robust candidate genes, functional validation is still required. It is worth noting that the core findings of this study are based on correlation analysis, which can only reveal trends of association between variables but cannot establish causality. The three key genes (NFE2L2, HSPB1, JUN) did not show significant commonalities in their associations with immune regulation and metastasis-related genes, suggesting that they may function through independent signaling pathways—a speculation that requires validation through subsequent in vitro and in vivo functional experiments. Moreover, larger patient cohorts are needed to validate their diagnostic and therapeutic potential.

Conclusions

The dysregulated expression of NFE2L2, JUN, and HSPB1, together with distinct patterns of immune cell infiltration, appears to contribute to RB progression. These genes may serve as ferroptosis-related biomarkers and potential therapeutic targets in RB. Collectively, our findings provide novel insights that could aid in the prevention, monitoring, and precision treatment of RB.

Acknowledgments

The authors appreciate the GEO database. The authors thank Zhihui Gene Technology Co., Ltd. for their technical assistance and consultation on the bioinformatics analysis.

Footnote

Reporting Checklist: The authors have completed the TRIPOD and MDAR reporting checklists. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2568/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2568/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2568/prf

Funding: This work was supported by Kuanren Talents Program of The Second Affiliated Hospital of Chongqing Medical University (No. 202417-32), China Postdoctoral Science Foundation (No. 2024M753888), Chongqing Postdoctoral Research Project Special Support (No. 2023CQBSHTB3106), and Hami Science and Technology Program Project (No. hmkj2025026).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2568/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Stenfelt S, Blixt MKE, All-Ericsson C, et al. Heterogeneity in retinoblastoma: a tale of molecules and models. Clin Transl Med 2017;6:42. [Crossref] [PubMed]
2. Alonso J, Frayle H, Menéndez I, et al. Identification of 26 new constitutional RB1 gene mutations in Spanish, Colombian, and Cuban retinoblastoma patients. Hum Mutat 2005;25:99. [Crossref] [PubMed]
3. Cruz-Gálvez CC, Ordaz-Favila JC, Villar-Calvo VM, et al. Retinoblastoma: Review and new insights. Front Oncol 2022;12:963780. [Crossref] [PubMed]
4. Martínez-Sánchez M, Hernandez-Monge J, Rangel M, et al. Retinoblastoma: from discovery to clinical management. FEBS J 2022;289:4371-82. [Crossref] [PubMed]
5. Ancona-Lezama D, Dalvin LA, Shields CL. Modern treatment of retinoblastoma: A 2020 review. Indian J Ophthalmol 2020;68:2356-65. [Crossref] [PubMed]
6. Liu K, Li H, Wang F, et al. Ferroptosis: mechanisms and advances in ocular diseases. Mol Cell Biochem 2023;478:2081-95. [Crossref] [PubMed]
7. Kuganesan N, Dlamini S, Tillekeratne LMV, et al. Tumor suppressor p53 promotes ferroptosis in oxidative stress conditions independent of modulation of ferroptosis by p21, CDKs, RB, and E2F. J Biol Chem 2021;297:101365. [Crossref] [PubMed]
8. Liu K, Huang J, Liu J, et al. Induction of autophagy-dependent ferroptosis to eliminate drug-tolerant human retinoblastoma cells. Cell Death Dis 2022;13:521. [Crossref] [PubMed]
9. Kitajima S, Li F, Takahashi C. Tumor Milieu Controlled by RB Tumor Suppressor. Int J Mol Sci 2020;21:2450. [Crossref] [PubMed]
10. Lin FY, Chintagumpala MM. Neonatal Retinoblastoma. Clin Perinatol 2021;48:53-70. [Crossref] [PubMed]
11. Byroju VV, Nadukkandy AS, Cordani M, et al. Retinoblastoma: present scenario and future challenges. Cell Commun Signal 2023;21:226. [Crossref] [PubMed]
12. Kaewkhaw R, Rojanaporn D. Retinoblastoma: Etiology, Modeling, and Treatment. Cancers (Basel) 2020;12:2304. [Crossref] [PubMed]
13. Sun J, Xi HY, Shao Q, et al. Biomarkers in retinoblastoma. Int J Ophthalmol 2020;13:325-41. [Crossref] [PubMed]
14. Bouchoucha Y, Matet A, Berger A, et al. Retinoblastoma: From genes to patient care. Eur J Med Genet 2023;66:104674. [Crossref] [PubMed]
15. Lei G, Zhuang L, Gan B. Targeting ferroptosis as a vulnerability in cancer. Nat Rev Cancer 2022;22:381-96. [Crossref] [PubMed]
16. Zeng F, Nijiati S, Tang L, et al. Ferroptosis Detection: From Approaches to Applications. Angew Chem Int Ed Engl 2023;62:e202300379. [Crossref] [PubMed]
17. Yang Y, Lin Y, Han Z, et al. Ferroptosis: a novel mechanism of cell death in ophthalmic conditions. Front Immunol 2024;15:1440309. [Crossref] [PubMed]
18. Zhong L, Meng X, Huang J, et al. Expression of YAP suppresses cell proliferation and elevates the sensitivity of chemotherapy in retinoblastoma cells through lipid-peroxidation induced ferroptosis. Chin Clin Oncol 2023;12:52. [Crossref] [PubMed]
19. Liu H, Gan Q, Lai Y, et al. USP14 increases the sensitivity of retinoblastoma to cisplatin by mediating the ferroptosis. Naunyn Schmiedebergs Arch Pharmacol 2024;397:8671-80. [Crossref] [PubMed]
20. Li C, Sun S, Zhuang Y, et al. CTSB Nuclear Translocation Facilitates DNA Damage and Lysosomal Stress to Promote Retinoblastoma Cell Death. Mol Biotechnol 2024;66:2583-94. [Crossref] [PubMed]
21. Zhang Z, Tang J, Liu Y, et al. The role of lactate metabolism in retinoblastoma tumorigenesis and ferroptosis resistance. Tissue Cell 2025;95:102893. [Crossref] [PubMed]
22. Kobayashi A. Roles of NRF3 in the Hallmarks of Cancer: Proteasomal Inactivation of Tumor Suppressors. Cancers (Basel) 2020;12:2681. [Crossref] [PubMed]
23. Homma S, Ishii Y, Morishima Y, et al. Nrf2 enhances cell proliferation and resistance to anticancer drugs in human lung cancer. Clin Cancer Res 2009;15:3423-32. [Crossref] [PubMed]
24. Yang S, Xie Z, Pei T, et al. Salidroside attenuates neuronal ferroptosis by activating the Nrf2/HO1 signaling pathway in Aβ1-42-induced Alzheimer's disease mice and glutamate-injured HT22 cells. Chin Med 2022;17:82. [Crossref] [PubMed]
25. Zhang H, Yang X, Xu Y, et al. KCNQ1OT1 regulates the retinoblastoma cell proliferation, migration and SIRT1/JNK signaling pathway by targeting miR-124/SP1 axis. Biosci Rep 2021;41:BSR20201626. [Crossref] [PubMed]
26. Mi Y, Wei C, Sun L, et al. Melatonin inhibits ferroptosis and delays age-related cataract by regulating SIRT6/p-Nrf2/GPX4 and SIRT6/NCOA4/FTH1 pathways. Biomed Pharmacother 2023;157:114048. [Crossref] [PubMed]
27. Zuo X, Zeng H, Wang B, et al. AKR1C1 Protects Corneal Epithelial Cells Against Oxidative Stress-Mediated Ferroptosis in Dry Eye. Invest Ophthalmol Vis Sci 2022;63:3. [Crossref] [PubMed]
28. Li S, Lu S, Wang L, et al. Effects of amygdalin on ferroptosis and oxidative stress in diabetic retinopathy progression via the NRF2/ARE signaling pathway. Exp Eye Res 2023;234:109569. [Crossref] [PubMed]
29. Shi W, Dong Y, Liu S, et al. Corilagin alleviates ferroptosis in diabetic retinopathy by activating the Nrf2 signaling pathway. Biomed Pharmacother 2024;179:117409. [Crossref] [PubMed]
30. Zhi X, Lu H, Ma D, et al. Melatonin protects photoreceptor cells against ferroptosis in dry AMD disorder by inhibiting GSK-3B/Fyn-dependent Nrf2 nuclear translocation. Biochim Biophys Acta Mol Basis Dis 2024;1870:166969. [Crossref] [PubMed]
31. Dai C, Chen X, Li J, et al. Transcription factors in ferroptotic cell death. Cancer Gene Ther 2020;27:645-56. [Crossref] [PubMed]
32. Yang Y, Lin Y, Wang M, et al. Targeting ferroptosis suppresses osteocyte glucolipotoxicity and alleviates diabetic osteoporosis. Bone Res 2022;10:26. [Crossref] [PubMed]
33. Lei XL, Yang QL, Wei YZ, et al. Identification of a novel ferroptosis-related gene signature associated with retinal degeneration induced by light damage in mice. Heliyon 2023;9:e23002. [Crossref] [PubMed]
34. Park AM, Tsunoda I, Yoshie O. Heat shock protein 27 promotes cell cycle progression by down-regulating E2F transcription factor 4 and retinoblastoma family protein p130. J Biol Chem 2018;293:15815-26. [Crossref] [PubMed]
35. Sooraj K, Shukla S, Kaur R, et al. The protective role of HSP27 in ocular diseases. Mol Biol Rep 2022;49:5107-15. [Crossref] [PubMed]
36. Sun X, Ou Z, Xie M, et al. HSPB1 as a novel regulator of ferroptotic cancer cell death. Oncogene 2015;34:5617-25. [Crossref] [PubMed]
37. Liang Y, Wang Y, Zhang Y, et al. HSPB1 facilitates chemoresistance through inhibiting ferroptotic cancer cell death and regulating NF-κB signaling pathway in breast cancer. Cell Death Dis 2023;14:434. [Crossref] [PubMed]
38. Esfahani K, Roudaia L, Buhlaiga N, et al. A review of cancer immunotherapy: from the past, to the present, to the future. Curr Oncol 2020;27:S87-S97. [Crossref] [PubMed]
39. Sarver AL, Xie C, Riddle MJ, et al. Retinoblastoma tumor cell proliferation is negatively associated with an immune gene expression signature and increased immune cells. Lab Invest 2021;101:701-18. [Crossref] [PubMed]
40. Zhang X, Liu H, Lan Z, et al. Single-cell transcriptomic analyses provide insights into the tumor microenvironment heterogeneity and invasion phenotype in retinoblastoma. Pathol Res Pract 2025;271:156009. [Crossref] [PubMed]
41. Wang X, Ma J, Dang Y, et al. Circular RNAs and immunotherapy in retinoblastoma: emerging biomarkers and precision therapeutic strategies. Front Immunol 2025;16:1666606. [Crossref] [PubMed]
42. Chen M, He H, Cheng H, et al. EIF4A3-induced hsa_circ_0078136 inhibits the tumorigenesis of retinoblastoma via IL-17 signaling pathway. Int Ophthalmol 2024;44:352. [Crossref] [PubMed]