The RNA-seq data and clinical information for the LUAD patients were obtained from the Genomic Data Commons (GDC) data portal of the Cancer Genome Atlas (TCGA) database (https://portal.gdc.com). This dataset included 516 LUAD tissues and 59 paracancerous tissues, and the clinical characteristics of the patients in this study can be found in Supplementary Table 1 . The genes related to ferroptosis are sourced from Ze-Xian Liu et al.’s systematic analysis of the aberrations and functions of ferroptosis in cancer (17). To further validate the expression level of target genes, the GSE46539 datasets from the Gene Expression Omnibus database (GEO) were downloaded and utilized (https://www.ncbi.nlm.nih.gov/geo/).

R version 4.0.3 was used to carry out each of the previously listed analytical techniques. A consistency study was performed using a maximum of 6 clusters, 100 repeats, 80% of the total samples randomly picked each time, clusterAlg = “hc”, and innerLinkage = “ward.D2” using the R package ConsensusClusterPlus (v1.54.0). The R program pheatmap (v1.0.12) was used to evaluate the clustering heatmaps, and genes with a variance over 0.1 were retained in the gene expression heatmaps. The top 25% of genes sorted by variance from greatest to lowest were presented if the input target gene count was more than 1000. PCA plots were produced with the R program ggord, and box plots were produced with ggplot2. In the Kaplan-Meier survival study, the differences in survival between the two groups stated above were examined using the log-rank test. To examine the target genes’ prediction accuracy, TimeROC analysis was performed. Log-rank tests and univariate Cox regression were used to get the p-values and hazard ratios (HR) with 95% confidence intervals (CI) for the Kaplan-Meier curve. To examine the effects of smoking, alcohol intake, target genes, and subtypes on LUAD risk, univariate Cox regression analysis was employed. The R package “forestplot” was used to create a forest plot that displayed the p-values, HR, and 95% CI.

A total of 37 pairs of LUAD and adjacent normal tissues and blood were collected from patients who underwent radical nephrectomy at the Second Affiliated Hospital of Xuzhou Medical University between March and December 2015. Written informed consent was obtained from each patient, and the Institutional Ethics Committee at the hospital approved the study. Total RNA was extracted using TRIzol reagent (Takara) and reverse-transcribed into cDNA using the PrimeScript RT Reagent Kit (Takara). The relative expression of genes was determined by RT-qPCR using the SYBR Premix Ex Taq (Takara) and the ABI 7500 fast real-time PCR System (Applied Biosystems), with GAPDH serving as an endogenous normalization reference. The clinical information of all patients and the primers used are provided in Supplementary Tables 2 , 3 .

The A549 human alveolar basal epithelial cancer cell line and BEAS-2B normal human lung bronchial epithelial cells were obtained from the Shanghai Institute of Cell Biology. To overexpress SLC7A11 via plasmid transfection, the plasmid containing the target gene is introduced into the cells using transfection reagent. To knockdown SLC7A11 in A549, a specific small interfering RNA (siRNA) targeting the gene is transfected into the cells using transfection reagents. Cell function assays, such as cell viability, cell migration, and ROS detection, were conducted using established research methods (18, 19). The steps for cell viability CCK-8 assay involve seeding cells, treating them, adding CCK-8 reagent, incubating, reading absorbance, and analyzing data to assess cell viability and proliferation effects. To assess cell migration using the Transwell assay, cells are plated in the upper chamber of a Transwell insert, while a chemoattractant or conditioned media is placed in the lower chamber. After incubation, cells that migrate through the porous membrane to the lower chamber are quantified, providing insights into their migratory capacity. The iron concentration was measured using an iron assay kit (Abcam, ab83366) following the manufacturer’s instructions. The detection of cellular ROS using fluorescence microscopy involves labeling cells with ROS-sensitive fluorescent probes, followed by visualizing the fluorescence signal emitted by the probes to assess the levels of ROS within the cells. The Western blot experiment included sample protein extraction, separation of proteins by gel electrophoresis, antibody detection of PD-L1 (Abcam, ab205921), and analysis of protein expression levels through band intensity quantification.

For the pan-cancer investigation of SLC7A11, RNA-seq data and accompanying clinical information for 33 cancer types were also gathered from TCGA. The abbreviations for numerous cancer types and ferroptosis regulators in this study can be found in Supplementary Tables 4 , 5 . To undertake a reliable evaluation of immune-related correlations, we deployed immunedeconv, a R software package that includes six state-of-the-art algorithms, including TIMER, xCell, MCP-counter, CIBERSORT, EPIC, and quanTIseq. To examine the expression of these eight immune checkpoint-related genes, transcripts corresponding to SIGLEC15, IDO1, PD-L1, HAVCR2, PDCD1, CTLA4, LAG3, and PDCD1LG2 were extracted. R software version 4.0.3 was used to conduct the statistical analysis. p < 0.05 was deemed statistically significant when comparing the two sets of data using the Wilcoxon rank-sum test.

We used GraphPad Prism 8.0 and R, version 4.0.3, to conduct statistical tests. The Student’s t-test was used to compare two groups of continuous variables. Differences in proportions were compared by the Chi-squared test. For comparing more than two groups, the one-way ANOVA or Kruskal-Wallis test were utilized. Kaplan-Meier analysis were utilized with the log-rank test to compare the overall survival (OS) and progression-free survival between different groups. Additionally, univariate and multivariate Cox regression analyses were to conducted identify independent factors associated with OS. A statistical significance threshold of p < 0.05 was applied to ascertain the results. Mechanism figure is drawn by Figdraw.

A total of 44 ferroptosis regulator genes identified from prior studies were deemed essential ferroptosis regulators because of their critical functions in controlling ferroptosis (8, 17). Using the TCGA dataset, the expression patterns of a few chosen ferroptosis regulators between LUAD and nearby normal pairs were methodically investigated in order to assess the biological roles of these regulators in LUAD. The expression profiles of 59 normal patients and 516 instances of LUAD individuals were retrieved and examined. We screened a total of 1103 differentially expressed genes in LUAD, and 25 out of 44 ferroptosis regulator genes were among them, so we included these 25 genes as key ferroptosis regulator genes in LUAD in the subsequent study ( Supplementary Figure 1 ). In further screening, our analysis revealed distinct expression levels of 25 ferroptosis regulators in LUAD and normal tissues ( Figures 1A, B ). Specifically, we found upregulated ferroptosis regulators HSPA5, SLC7A11, HSPB1, GPX4, FANCD2, CISD1, SLC1A5, RPL8, GLS2, DPP4, CS, CARS1, and ATP5MC3 (p<0.05), as well as downregulated regulators CDKN1A, NFE2L2, FDFT1, SAT1, TFRC, NCOA4, LPCAT3, ALOX15, ACSL4, and ATL1 (p<0.05). However, no significant differences were observed between LUAD and normal tissues in the expression of EMC2 and MT1G (p>0.05). Furthermore, correlation and prognosis analysis demonstrated that the expression of most ferroptosis regulatory genes was positively correlated and played a crucial prognostic role in LUAD ( Figure 1C ). These findings suggest that ferroptosis regulator genes may play a critical role in regulating tumorigenesis and the development of LUAD.

We utilized the Consensus Cluster Plus package of R software for consistency analysis. Based on the expression levels of selected ferroptosis regulators and the proportion of ambiguous clustering measures, we identified k = 2 as the optimal clustering stability from k = 2 to 6 ( Figure 2A ; Supplementary Figure 2 ). Consequently, the 516 LUAD patients were categorized into two subtypes, namely, cluster 1 (n = 358) and cluster 2 (n = 158). To validate the clustering results based on the expression of ferroptosis regulators, we then employed the PCA method to analyze the gene expression profiles between the two subtypes ( Figure 2B ). The gene expression profiles between the two subtypes exhibited distinct differences. Most of the ferroptosis regulators were highly or lowly expressed in cluster 1, while there was no difference in the expression of four regulators (HSPB1, GPX4, GLS2, and SAT1) between the two clusters ( Figure 2C ). Subsequently, differences were observed in the clinicopathological characteristics and prognosis between the two subtypes, as shown in Supplementary Table 1 . Cluster 1 was significantly associated with a lower tumor stage and cancer grade (p < 0.01), while no statistical differences in age, sex, or race were found between the two clusters (p>0.05). Additionally, cluster 1 patients exhibited better overall survival (p< 0.001) and progression-free survival (p<0.001) than cluster 2 ( Figures 2D, E ). These results indicate significant heterogeneity between the two subtypes of LUAD patients.

In order to explore the relationship between PD-L1 and ferroptosis in LUAD, this study examined the divergent expression in two clusters and the association of PD-L1 with the ferroptosis regulators in LUAD. Unlike normal adjacent tissues and cluster 1 patients, the expression levels of PD-L1 in LUAD tissues and cluster 2 patients were notably low (p< 0.01; Figures 3A, B ). An analysis involving 516 LUAD individuals revealed that PD-L1 exhibited an association with the expression levels of most ferroptosis regulators ( Figure 3C ). We further investigated the impact of ferroptosis regulators on the tumor immune microenvironment of LUAD. The two subtypes classified based on the expression levels of selected ferroptosis regulators showed significant differences in immune cell infiltration ( Figure 3D ). Cluster 2 exhibited higher levels of infiltrated activated Common lymphoid progenitor, T cell CD4+ Th2, T cell CD4+ Th1, while cluster 1 was more correlated with the B cell plasma, T cell CD4+ naive, B cell, B cell memory ( Figure 3D ; Supplementary Figure 3 ). The GSEA method was utilized to elucidate the underlying regulatory mechanisms causing the difference between the two clusters. Gene set enrichment analysis indicated that pentose phosphate, ERBB signaling, and alpha linolenic acid metabolism pathways are significantly enriched in cluster 1 ( Supplementary Figure 4A ).

As previously mentioned, the two subtypes of LUAD, distinguished by the levels of ferroptosis regulators, exhibited distinct PD-L1 expression and tumor immune environments. These findings indicate the crucial role of ferroptosis regulators in tumor development and immune infiltration. By identifying the highly expressed ferroptosis regulators in LUAD associated with poor prognosis and negatively correlated with PD-L1 expression, we identified SLC7A11, DPP4, and GLS2 as potential key poor prognostic ferroptosis regulators differently expressed in LUAD and correlated with PD-L1 expression ( Figure 4A ). To further illustrate the expression of SLC7A11, DPP4, and GLS2 in LUAD, we analyzed expression data from the GEO database. The results from the GSE46539 datasets showed significantly higher expression of SLC7A11, DPP4, and GLS2 in LUAD compared to normal tissues, while PD-L1 was expressed at lower levels (p < 0.05, Figures 4B, C ; Supplementary Figures 5A, D ). Additionally, the prognostic value of highly expressed SLC7A11, DPP4, and GLS2 was analyzed in LUAD, revealing that only the upregulated SLC7A11 group exhibited worse overall survival compared to the lowly expressed group (p<0.05, Figure 4D ; Supplementary Figures 5B, E ). This result was further confirmed in an independent LUAD cohort from the Kaplan–Meier plotter (p<0.05, Figure 4E ; Supplementary Figures 5C, F ). Similarly, LUAD tissue had higher levels of SLC7A11 protein expression than normal tissue did in the human protein atlas ( Supplementary Figure 6 ). Further correlation analysis demonstrated a significant negative correlation between SLC7A11 and PD-L1 expression in LUAD (p<0.001, Spearman =-0.15) ( Figure 4F ). Therefore, SLC7A11 emerged as the sole highly expressed ferroptosis regulator in LUAD and cluster 2, exhibiting a negative correlation with PD-L1 expression and worse overall survival.

Subsequently, we obtained tumor samples from 37 patients with LUAD who underwent monthly follow-up visits. We assessed the expression of SLC7A11 and PD-L1 in the lung tissues of LUAD patients and normal paraneoplastic tissues (control) ( Figures 5A, B ). Our findings indicated high expression of SLC7A11 in LUAD and low expression of PD-L1, consistent with the previous section. Furthermore, there were no statistically significant differences in SLC7A11 expression across smoking, gender, or clinical stage ( Figures 5C-E ). Subsequently, we categorized the 37 LUAD patients into high and low SLC7A11 expression groups based on SCL7A11 expression in tumor tissue and plotted Kaplan–Meier curves ( Figure 5F ). Similarly, the group with upregulated SLC7A11 exhibited a trend toward worse overall survival compared to the lowly expressed group (logrank p=0.08). We also employed the Cox analysis method to investigate the relationship between SLC7A11 expression and overall survival in LUAD. Univariate analysis revealed a significant correlation between pM-stage (p<0.01) and SLC7A11 expression (p<0.01) with overall survival ( Figure 5G ). Furthermore, multivariate analysis demonstrated that SLC7A11 expression (p< 0.01) is an independent prognostic factor in LUAD patients ( Figure 5H ).

Human LUAD cell lines A549 and normal lung epithelial cells BEAS-2B were used to detect the expression levels of SLC7A11 and PD-L1. Compared with normal cells, lung cancer cells A549 had higher levels of SLC7A11 expression and lower levels of PD-L1 expression, and the difference was statistically significant ( Figures 6A, B ). Meanwhile, we overexpressed SLC7A11 in A549, which resulted in reduced PD-L1 expression ( Figure 6C ). In contrast, knockdown of SLC7A11 showed a increase in PD-L1 expression in A549 ( Supplementary Figure 7A ). The expression level of PD-L1 protein after SLC overexpression or knockdown was explored using Western Blot, and the results were consistent with qRT-PCR ( Supplementary Figure 7B ). Overexpression of SLC7A11 also leads to an accelerated proliferation rate and migration of A549 cells ( Figures 6D, E ). We next explored the effect of overexpression of SLC7A11 on ferroptosis levels, overexpression of SLC7A11 resulted in lower iron ion concentration and reduced ROS accumulation in A549 cells ( Figures 6F, G ). All of these results suggest that high expression of SLC7A11 inhibits the ferroptosis levels of lung cancer cells, leading to accelerated proliferation and migration, which may be associated with poor prognosis.

Given the widespread involvement of ferroptosis and immune infiltration in the development of various cancer types, we aimed to investigate whether SLC7A11, the key ferroptosis regulator in our study, also predicts poor prognosis and influences the tumor immune microenvironment in other cancer types. To begin, we analyzed the relative expression of SLC7A11 in over 10,000 tumor and normal tissue samples in the TCGA and GTEx databases. The results revealed that, compared to normal tissues, SLC7A11 was significantly upregulated in 11 cancer types (BRCA, CESC, ESCA, HNSC, LUAD, LUSC, PRAD, READ, SARC, STAD, and UCEC) and downregulated in 9 cancer types (CHOL, COAD, KICH, KIRC, KIRP, LIHC, PAAD, PCPG, and THCA) (p< 0.05; Figure 7A ). Furthermore, we investigated the prognostic significance of SLC7A11 across multiple cancer types. Univariate Cox regression analyses in pan-cancer indicated that upregulation of SLC7A11 was associated with poor overall survival in 10 cancer types (ACC, KIRP, LAML, LIHC, LUAD, MESO, OV, SARC, UCEC, UVM) (p< 0.05; Figure 7B ). Additionally, we conducted an expression correlation analysis between SLC7A11 and the eight most common immune checkpoint-related genes (including PD-L1, CTLA4, HAVCR2, LAG3, PDCD1, PDCD1LG2, SIGLEC15, and TIGIT). The results indicated a close relationship between SLC7A11 and the expression of checkpoint-related genes in most cancer types, particularly PD-L1 expression ( Figure 7C ). In summary, high expression of SLC7A11 was associated with the expression of immune checkpoint-related genes and poor prognosis in various prevalent cancers.