GSE27262, GSE102287, GSE116959, GSE118370, and GSE19945 were acquired from the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) on the basis of the platform of the Affymetrix Human Genome U133 Plus 2.0 Array, Affymetrix Human Genome U133 Plus 2.0 Array, Agilent-039494 SurePrint G3 Human GE v2 8x60K Microarray 039381, Affymetrix Human Genome U133 Plus 2.0 Array, and Agilent Human 0.6K miRNA Microarray G4471A, respectively.

GEO2R (http://www.ncbi.nlm.nih.gov/geo/geo2r), as a free website toolset, was utilized to compare gene expression data of multiple groups and to identify the differentially expressed genes (DEGs) and differentially expressed miRNAs (DEMs) between the samples of neoplasm and normal control. A p-value <0.05 meant that there was a significant difference.

The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was applied to identify molecular interaction and relation networks through the Database for Annotation Visualization and Integrated Discovery (DAVID; http://david.ncifcrf.gov/). A p-value <0.05 meant that there was a significant statistical difference. The cutoff criteria were considered as a false discovery rate (FDR) <0.01 and gene count >2.

A total of 90 NSCLC tissues and paired adjacent normal ones were taken from patients undergoing primary surgery at the Department of Thoracic Surgery in Huashan Hospital, Fudan University, between February 2016 and July 2020. Referring to the classification tumor node metastasis (TNM) and the standard of the World Health Organization (WHO), all tissue samples were staged and graded by an experienced pathologist. All samples that were immersed in RNA Later stabilization solution (Qiagen, Germany) were maintained in liquid nitrogen and preserved at -80°C. Patients were not subjected to any treatment against carcinoma prior to surgical resection. Our research was approved by the institutional ethics committee of Huashan Hospital of Fudan University, and the participants signed a written informed consent form.

Human NSCLC cell lines were ordered from the National Collection of Authenticated Cell cultures, and Beas-2B cell lines were obtained from MingZhou bio company. NSCLC cell lines were maintained in DMEM (Gibco, USA) with 10% FBS (HyClone, USA) and 1% Penicillin/Streptomycin (Life Technologies, UK) under a 37°C incubator with 5% CO₂. Beas-2B cells were maintained in LHC-9 medium with 0.5 ng/ml EGF, 500 ng/ml hydrocortisone.

A549 cells were transfected with shRNAs (sh-Scrambled and two shRNAs specifics for SLC7A11) and miRNAs (miRNA mimics and controls) by Lipofectamine 3000 (Invitrogen, USA) as manufacturers described. Puromycin (2 µg/ml, Sigma-Aldrich, USA) was added into cells for subsequent screening until stabilization. SLC7A11 cDNA was ligated into lentiviral construct (GenePharma) for subsequent studies.

TRIzol (1 ml, Invitrogen) was employed to harvest the whole RNA, followed by reverse transcription with the use of a First Strand cDNA Synthesis Kit (Thermo Scientific, USA) as manufacturers instructed. qRT-PCR was taken to quantify the RNA level using a SYBR Premix ExTaq Reverse Transcription PCR kit (Takaka, China). GAPDH was utilized as a normalized control. All primers were listed in Supplementary Table 1 .

Cell Counting Kit (CCK-8) (Yesen, China) was employed to detect the cell viability in triplicate for 3 consecutive days at 450 nm by a microplate reader (Thermo Scientific, USA).

Fluorescence activated cell sorting (FACS) was employed to analyze cell proliferation at post-transfection. A total of 1×10⁴/well were incubated with Annexin V-FITC and propidium iodide (Yeasen, China) for 10 min at 4°C in the dark. Subsequently, the cells were rinsed with buffer two times and resuspended by 500 μl of buffer for flow cytometry analysis.

Dihydroethidium (DHE, Merck KGaA) was used to stain ROS, followed by flow cytometry detection as previously described (17). In short, cells were digested and risned and then mixed with 1.25 μM DHE for 30 min at 37°C in the dark. Fluorescence at 610 nm was measured on a FACS Calibur™.

The iron analysis kit (Abcam) was employed to measure the intracellular ferrous (Fe²⁺) level as manual depicted. Taken briefly, we collected samples, washed by precold PBS and homogenized in precold five volumes of iron analysis buffer. We harvested the supernatant and added an iron-reducing agent into every allocation, followed by incubation for 30 min. Each mixture was incubated with iron probes and incubated for 60 min. Subsequently, the OD (optical density) value of 593 nm was measured by a colorimetric microplate reader.

Whole protein was harvested in samples by lysis buffer (Beyotime, China). Proteins (25 µg) with the corresponding volume of loading buffer were boiled in water bath for 10 min. SDS-PAGE gel was used to resolve the proteins and then transferred onto PVDF membranes (Millipore, USA). Primary and secondary antibodies were incubated as indicated. xCT/SLC7A11 (CST,12691), GPX4 (CST, 52455), and GAPDH (CST, 5174) were used in this study.

The rehydrated sections were retrieved by citrate buffer for 3 min at 100°C and incubated with primary antibodies at 4°C overnight. On the following day, second antibody (goat anti-rabbit IgG) was incubated with them at RT for 30 min. The washed sections were stained with diaminobenzidine and captured under a microscope. The images were analyzed by Image Pro Plus software (Media Cybernetics, USA).

We utilized TargetSan (http://www.targetscan.org/vert72), miRanda (http://www.microrna.org/microrna/home.do), microT (http://www.microrna.gr/microT), PITA (http://genie.weizmann.ac.il/pubs/mir07/mir07_data.html), miRmap (http://mirmap.ezlab.org), and PicTar (http://www.pictar.org/) databases to predict the upstream gene of SLC7A11. miR-27a-3p was chosen. The 3’UTR region with the miR-27a-3p binding site was cloned into pGL3-basic luciferase construct (Promega, USA). Cotransfection of luc-SLC7A11-wt/luc-SLC7A11-mut with miR-27a-3p mimic/NC mimic into A549 cells was performed by Lipofectamine 3000 (Invitrogen). At 48 h post-transfection, luciferase activity was measured by a Dual-Luciferase Reporter Assay System (Promega) as manufacturers instructed. Renilla luciferase was regarded as normalized control.

The represented data were shown as the mean ± SEM from three separate experiments in triplicate. Statistical significance was determined by the unpaired Student’s t test, *p < 0.05.

First, we performed bioinformatic interrogation of GEO datasets. In total, 876, 1,292, 846, and 1,438 DEGs were obtained from GSE27262, GSE102287, GSE116959, and GSE118370, respectively. A total of 180 genes were screened out in all four datasets ( Figure 1A ). Furthermore, KEGG pathway analysis showed seven identified pathways including ferroptosis with four DEGs, SLC7A11, CP, GCLM, and STEAP3 ( Figure 1B ). To verify the expression trends of SLC7A11, CP, GCLM, and STEAP3 in NSCLC, TCGA-LAUD data were obtained and reanalyzed. We found that all the four DEGs were indeed upregulated in NSCLC ( Figure 1C ). Recent studies revealed that SLC7A11 overexpression promotes tumor growth partly through suppressing ferroptosis. As the upstream and key gene in the ferroptosis pathway, SLC7A11 was selected for further study.

A total of 90 NSCLC patients were classified into two groups in the light of the median value to assess in-depth the association of SLC7A11 expression with clinicopathological traits. First, IHC array was conducted to determine the SLC7A11 level in patients with NSCLC, and the result showed that its levels significantly increased in NSCLC patients’ tissues ( Figure 1D ). Our data revealed that SLC7A11 expression had an association with tumor size (chi-square test, P = 0.009), smoking history (chi-square test, P = 0.010), lymph node metastasis (chi-square test, P = 0.009), and TNM stage (chi-square test, P = 0.013) ( Table 1 ). Nonetheless, no obvious correlation was observed between the SLC7A11 level and age, gender, differentiation, and primary location (chi-square test, P > 0.05; Table 1 ). Collectively, our results indicated that SLC7A11 may function essentially in the progression of NSCLC. In addition, the Kaplan–Meier method analysis revealed that the overall survival (OS) rate in the highly expressed SLC7A11 group was largely lower than that in the lowly expressed SLC7A11 group ( Figure 1E ).

Next, we identified SLC7A11 mRNA and protein expression in NSCLC cells. SLC7A11 was higher in NSCLC cells than in normal control ( Figures 2A, B ). Next, we analyzed erastin activity in A549 and Calu-3 cells. Both erastin (IC50 (A549) = 5.56 μM; IC50 (Calu-3) =2.72 μM) could induce cell death in A549 and Calu-3 cells, which could be reversed after treatment of ferrostatin-1 (ferroptosis inhibitor) ( Figure 2C and Supplementary Figure S1 ). We used A549 cells for further analysis. Images showed erastin induced A549 cells around, swelled, and rescued by ferrostatin-1 ( Figure 2D ). GPX-4 is a ferroptosis maker, which protects cells against membrane lipid peroxidation. Notably, SLC7A11 expression increased and GPX-4 significantly attenuated during ferroptosis process ( Figure 2E ). Considering oxidative stress, lipid peroxidation, and iron accumulation were major signaling processes in generating ferroptosis, we focused on these three indicators. First, we found that erastin led to ROS accumulation in both A549 and Calu-3 cells and rescued by ferrostatin-1. Similarly, the H₂O₂ induced ROS could not be inhibited by ferrostatin-1, which confirmed that the oxidative stress happened during ferroptosis ( Supplememntary Figure S2 ). Considering malondialdehyde (MDA) is one of primary lipid peroxidation end products, we investigated whether MDA was gathered in NSCLC cells. Figure 2F illustrated that the MDA level was raised by erastin with dosage increasing. Besides, ferrous iron (Fe²⁺) is another important factor leading to ferroptosis ( Figure 2G ). We also found the intracellular iron concentrations were increased depending on the ferroptosis progress.

Next, SLC7A11 was ablated n by SLC7A11 shRNAs in two NSCLC cell lines to investigate SLC7A11’s role ( Supplementary Figure S3 ). Silencing SLC7A11 facilitated A549 cell viability compared with the control group ( Figure 3A ). Consistently, MDA accumulation assay demonstrated that the reduction of SLC7A11 resulted in the decrease in ROS and MDA accumulation under the stimulation of erastin in A549 cells ( Figures 3B, C ). Thus, we further uncovered the impacts of ablated SLC7A11 on Fe²⁺ levels. The intracellular Fe²⁺levels were reduced in A549 cells upon erastin treatment ( Figure 3D ). The same results were also illustrated in Calu-3 cells ( Figures 3E–H ). Collectively, our results suggested that SLC7A11 exhibited a key role in ferroptosis.

We explored in-depth the mechanism of SLC7A11 in ferroptosis. The miRNAs targeting SLC7A11 were predicted using TargetSan, miRanda, microT, PITA, miRmap, and PicTar ( Figure 4A ). A total of five overlapping miRNAs were screened out including miR-1297, miR-26a-5p, miR-26b-5p, miR-27a-3p, and miR-27b-3p. We further detected these miRNAs’ expression in the GSE19945 dataset, and the results showed that miR-26a-5p, miR-26b-5p, and miR-27a-3p were downregulated in NSCLC ( Figures 4B, C ). Among the three miRNAs, miR-27a-3p expression was remarkably decreased in cancer tissues and A549 cells ( Figure 4D ). The association of SLC7A11 mRNA expression with miR-27a-3p was further investigated. Pearson correlation analysis showed that SLC7A11 mRNA expression displayed a greatly negative correlation with miR-27a-3p (r = 0.797, P < 0.001) ( Figure 4E ). SLC7A11 mRNA expression was significantly upregulated, and ablating miR-27a-3p gave rise to increasing SLC7A11 mRNA level in miR-27a-3p mimic-transfected NSCLC cells ( Figure 4F ). Luciferase reporter assay was carried out to verify the relation of SLC7A11 with miR-27a-3p, and the data revealed that decreased luciferase activity was observed in A549 cells with cotransfection of miR-27a-3p and WT-SLC7A11-3’UTR rather than the control group, suggesting that SLC7A11 was a target of miR-27a-3p ( Figures 4G, H ).

CCK-8 cell viability assay was employed to analyze cell death after erastin treatment. Cell viability was reduced after downregulating miR-27a-3p in comparison with control ( Figure 5A ), suggesting that miR-27a-3p negatively modulated ferroptosis. To unearth miR-27a-3p’s role in ferroptosis, miR-27a-3p mimic was transfected into cells, and the data illustrated that upregulating miR-27a-3p contributed to the promotion of cell viability ( Figure 5B ). Similarly, we also assessed miR-27a-3p’s role in regulating MDA accumulation in NSCLC cells. The data indicated that the elevated MDA after erastin treatment was largely reversed upon overexpressing miR-27a-3p ( Figure 5C ). In addition, after erastin treatment, the overexpressed miR-27a-3p decreased Fe²⁺ levels in A549 and Calu-3 cells ( Figure 5E ). And the inhibition in endogenous miR-27a-3p caused by the antagonist led to the increase in MDA accumulation and intracellular Fe²⁺ levels in A549 and Calu-3 cells ( Figures 5D, F ). Of interest, miR-27a-3p exerted more significant effect on lipid peroxidation relative to iron accumulation (**p<0.01 vs *p<0.05), indicating that miR-27a-3p mediated the inhibition of ferroptosis majorly by regulating lipid peroxidation in NSCLC cells. Schematic of the working details of miR-27a-3p/SLC7A11 in NSCLC was shown in Figure 6 .