The gene expression profiles ( raw counts and FPKM-UQ) were downloaded from the TCGA database (https://gdc-portal.nci.nih.gov) by the TCGA biolinks R package (version 2.20.0) (Colaprico et al., 2016). Data from metastatic, male, or formalin-fixed paraffin-embedded tissue samples, as well as repeated data for the same samples, were excluded. After exclusion, 1,191 primary solid tumor (TP) and 112 solid tissue normal (NT) samples were retained for genomic data. The survival and clinical information of breast cancer were obtained from the Xena Functional Genomics Explorer of University of California Santa Cruz (https://xenabrowser.net/).

Monoclonal anti-FLAG antibody and anti-β-actin were obtained from Sigma (St. Louis, United States). Rabbit polyclonal anti-SLC7A11, and rabbit polyclonal anti-ESR1 were from Cell Signaling Technology (CST, United States). Protein G Plus-agarose was from Sangon Biotech (Shanghai, China). RNAiso Plus kits, reverse transcription kits, and real-time PCR kits were from Takara Biomedical Technology (Dalian, China). Ferrostatin-1 (Fer-1), Erastin, and carbobenzoxy-Leu-Leu-leucinal (MG132) were all purchased from MedChemExpress (Shanghai, China). The expression plasmids for MYC-Ub were from Dr. Haishan Huang’s lab (Wenzhou Medical University, China). SLC7A11 and ESR1 genes were obtained by PCR using MCF-7 cDNA as a template and then inserted into pcDNA3.1-FLAG vectors.

The human-source cell lines MCF-7, ZR-75-1, and MDA-MB-231 were obtained from the Cell Bank of Chinese Academy of Sciences and periodically tested for the absence of mycoplasma. The cell lines were grown in high-glucose DMEM medium (Gibco, United States) supplemented with 10% fetal bovine serum and 1% penicillin and streptomycin in a 37 C incubator maintained at 5% CO₂. For transfection, when the cells had grown to 70%–80% confluence in six-well plates, the plasmid (1 μg) or siRNA (100 μM) was transfected using Lipofectamine 2000 (Invitrogen, United States) according to the manufacturer's instructions. The siRNAs were synthesized by GenePharma (Shanghai, China), and the sequences were listed as follows: siRNA for ESR1, CTA​CAG​GCC​AAA​TTC​AGA​TAA; NEDD4L, CGC​CTT​GAC​TTA​CCT​CCA​TAT.

Cell death and lipid peroxidation were assessed by flow cytometry with trypan blue (Gibco, United States) and BODIPY™ 581/591 C11 (Thermo Fisher Scientific, United States) following the manufacturer’s instruction, respectively. In brief, cells were pretreated with 10 µM of Erastin or 10 µM of Fer-1 for 1 h and then exposed to IR. Cells were collected at 48 h for lipid peroxidation or 72 h for cell death testing after X-ray exposure. Cells were resuspended in 400 μl of serum-free medium that contained BODIPY™ 581/591 C11 (5 µM) at 37°C in the dark for 30 min, or in 200 μl of PBS containing 0.04% of trypan blue, respectively to detect lipid peroxidation and cell death. The data were analyzed using the NovoExpress software (ACEA, United States), and a minimum of 10,000 cells were analyzed per condition.

For Western blot, cells were lysed in ice-cold RIPA Lysis Buffer. Lysates were centrifuged at 12,000 rpm for 15 min at 4°C. The supernatant was collected, and the protein content was determined using the BCA Protein Assay Kit (Thermo-Fisher, United States) according to the manufacturer’s instructions. Equal amounts of total protein (20 μg) were separated by 10% or 12% SDS–PAGE and then transferred to an equilibrated PVDF membrane (Millipore, United States). β-actin was used as a loading control. For IP, MCF-7 cells were transfected with expression vectors for FLAG-SLC7A11, and were exposed to IR at 48 h post-transfection. At 24 h after IR, cells were lysed in 1% NP-40 lysis buffer. Whole cell lysates (WCL) were incubated with the specified antibodies overnight at 4°C with constant rotation. Then, protein G Plus-agarose was added for 2 h under rotation at 4°C. After incubation, immunocomplexes were washed five times with IP wash buffer (Sangon Biotech, China) and subsequently resolved by SDS-PAGE followed by Western blot analysis.

Cells were transiently transfected with MYC-tagged ubiquitin (UB), and FLAG-tagged SLC7A11 with or without si-NEDD4L for 48 h, and were exposed to IR for 24 h. At 24 h after IR, cells were pretreated with MG132 (10 μM) for 4 h before harvest. The lysates were subjected to immunoprecipitation with FLAG antibodies overnight at 4°C. Ubiquitinated proteins were determined by immunoblotting with anti-MYC antibodies (Santa Cruz, United States).

Total RNA was isolated from cultured cells using RNAiso Plus kits according to the manufacturer’s instructions. mRNAs were first reverse transcribed into cDNA and quantitative reverse transcription-PCR (qPCR) was performed as the following thermal conditions: 95°C for 30 s followed by 40 cycles of 95°C for 5 s and 60°C for 30 s. qPCR was performed on QuantStudio 3 (Applied Biosystems, United States), and the expression of target genes was normalized to that of GAPDH. The following primers were used: GAPDH forward CCA​TGG​GTG​GAA​TCA​TAT​TGG​A and reverse TCA​ACG​GAT​TTG​GTC​GTA​TTG​G; SLC7A11 forward GCG​TGG​GCA​TGT​CTC​TGA​C and reverse GCT​GGT​AAT​GGA​CCA​AAG​ACT​TC; and ESR1 forward GCT​TAC​TGA​CCA​ACC​TGG​CAG​A and reverse GGA​TCT​CTA​GCC​AGG​CAC​ATT​C.

Double luciferase–reporter gene determination was conducted to determine whether SLC7A11 the downstream target gene of ESR1. Reporter plasmid of SLC7A11 promoter double luciferase was synthesized from Generalbiol (Anhui, China) and inserted into the pGL3-Basic vector. pGL3-Basic SLC7A11 promoter was co-transfected into cells with ESR1 overexpression vector or control vector by Lipofectamine 2000. After transfection for 48 h, luciferase activity was determined on a Centro LB 960 Luminometer (Berthold Technologies, Germany) and the activity of renilla luciferase was used as a standardized control.

Statistical analysis was performed using R version 4.1.1 (http://www.r-project.org/) and GraphPrism 5.0 software (GraphPad Software, United States). The differentially expressed gene (DEG) analyses were done with R package edgeR (version 3.34.0) (Robinson et al., 2010). According to the suggested workflow, raw RNA-Seq read count data were normalized by Trimmed-mean M values (TMM) to balance the counts among the different samples (Maza, 2016). Genes were marked as DEGs if the absolute value of log₂ transformed fold change (log₂FC) ≥ 1.5 and the false discovery rate (FDR) < 10^-10. The FPKM-UQ normalized gene expression data were log-transformed for further analyses. The R package survminer (version 0.4.9) was used to separate patients into higher and lower expression groups by gene expression. The relationship between SLC7A11 expression status and clinicopathologic characteristics of breast cancer was analyzed by Chi-square test. The association with prognosis of breast cancer patients was determined by Kaplan–Meier survival analysis with Log-rank test. A Cox proportional hazard regression model was conducted to perform a univariate and multivariate analysis. All experiments were performed in triplicate, and replicate data were used to calculate means ± standard deviation (SD). ANOVA plus subsequent Dunnett post hoc tests were performed for multiple comparisons. If not specified above, a p-value less than 0.05 was considered to be statistically significant, and p-values were two-tailed.

Breast cancer data were collected from TCGA. Based on the screening principles, 1191 TP and 112 NT samples were selected into this study. DEGs were reflected in volcano plot (Figure 1A). According to the principle of statistics, 4 upregulated and 6 downregulated ferroptosis-related DEGs were associated with breast cancer. Among them, SLC7A11 was a critical regulator of ferroptosis (Koppula et al., 2020). The SLC7A11 mRNA expression was extremely upregulated in tumor tissues compared with normal tissues (Figure 1B). Excluding unavailable SLC7A11 expression value, 1,104 breast cancer samples including 217 high expressions (19.66%) and low 887 expressions (80.34%) were extracted for further study. Chi-Square analysis were applied to study the data and results revealed that there were many clinicopathologic features significantly correlated with SLC7A11 expression, including age, histological type, molecular subtype, ER, PR, and N classification (Supplementary Table S1). Differences in SLC7A11 expression are shown in Figure 1C according to ER subtype. SLC7A11 expression was lower in ER-positive tissues than ER-negative tissues.

In univariate Cox regression for overall survival (OS) model, we included variables that correlated with SLC7A11. The following multivariate analyses confirmed SLC7A11 expression, age, and N classification as independent prognostic factors after adjusting other prognostic indicators (Supplementary Table S2). Analysis demonstrated that patients with higher SLC7A11 were associated with worse OS than those with lower SLC7A11 (Figure 1D). In addition, K–M curves were used to evaluate the prognostic value in differentiation ER subtypes, which revealed that SLC7A11 specifically associated with OS in ER-positive cancers (Figure 3E), but not in ER-negative cancers (Figure 1F).

Then, we further define the relationship between ferroptosis and ER-positive cancers, and eight ferroptosis-related genes were analyzed according to Wu et al. (2020b) and Tang et al. (2021). Results revealed that only low expression of SLC7A11 supported ferroptosis occurrence in ER-positive cancers (Supplementary Figure S1). The above data suggest that SLC7A11 may play an important role in the regulation of ER-positive cell ferroptosis.

The IR-induced ferroptosis in different breast cancer cell lines were evaluated in our previous study (Ma et al., 2021). Fer-1 was able to significantly attenuate IR-induced cell death in ER-positive MCF-7 and ZR-75-1, but not in ER-negative MDA-MB-231 cells. To further verify the result, we analyzed the effects of Erastin and Fer-1 on lipid peroxidation accumulation and cell death in response to IR treatment. Importantly, Fer-1 significantly reduced IR-stimulated lipid peroxidation and cell death. Correspondingly, Erastin significantly increased lipid peroxidation accumulation and cell death, and showed synergistic effect with IR (Figures 2A–D). When we knockdown of SLC7A11 in ZR-75-1, IR-induced lipid peroxidation and cell death significantly increased, which could be rescued by Fer-1 (Figures 2E–J). These results suggested that ferroptosis was one pattern of IR-induced cell death in ER-positive cells, and SLC7A11 may play an important role in this process.

Referring to the above results, we considered that the relationship between SLC7A11 expression and breast cancer survival was probably based on ESR1. Furthermore, a positive correlation of ESR1 with SLC7A11 was found in ER-positive cancers (Figure 3A) and the survival curve of ESR1 expression in ER-positive cancers is shown in Figure 3B. From the K–M curve, higher expression of ESR1 was also related to poor prognosis. From these results, we can conclude that SLC7A11 and ESR1 are both related to survival in ER-positive cancers. However, the precise relationship is unclear.

Then, we designed siRNA targeting ESR1 to verify the function of ESR1. To determine whether ESR1 could interact with SLC7A11, an IP experiment was carried out. Our result suggested that there was no direct protein–protein interaction (Figure 3C). In order to determine the relationship between ESR1 and SLC7A11, MCF-7 and ZR-75-1 cells were transfected with siRNA-NC or siRNA-ESR1; 72 h later, cells were harvested, and the total mRNA and proteins were extracted and subjected to qPCR and Western blotting, respectively. As shown in Figures 3D–G, si-ESR1 significantly reduced ESR1 mRNA and protein levels, following the SLC7A11 levels decreased. Further dual-luciferase reporter gene assay was applied and the results showed that EGR1 was capable of activating the SLC7A11 promoter ESR1 was able to activate the transcription of SLC7A11 (Figure 3H). These findings support that ESR1 suppresses ferroptosis by promoting SLC7A11 expression in IR treatment.

In addition, the mRNA levels of SLC7A11 and ESR1 were evaluated by qRT-PCR based on time course. After exposure to X-ray, there was an increase in the mRNA levels of SLC7A11 and ESR1 (Figures 4A,B). The protein levels of SLC7A11 and ESR1 based on IR time course were evaluated by western blot (Figure 4C). The results suggested that IR-induced ferroptosis occurred at 48 h, but at 4–12 h SLC7A11 had a rising period to against cell death.

Lipid peroxidation and cell death were assayed by flow cytometry using the fluorescent probes C11-BODIPY and trypan blue, respectively. ESR1 knockdown rendered ER-positive cells susceptible for ferroptosis and lipid peroxidation accumulation induced by IR. When ESR1 was knockdown, the increase of SLC7A11 induced by IR was shown to significantly attenuate in MCF7 (Figure 5A) and ZR-75-1 (Figure 5F). The suppression of ESR1 by siRNA increased IR-induced cell death as well as rise of lipid peroxidation in MCF-7 (Figures 5B–E). Moreover, IR had similar effects on cell death and lipid peroxidation with ESR1 knockdown in ZR-75-1 (Figures 5G–J). Indeed, ESR1 knockdown showed a significant effect on ferroptosis induced by IR. The increased lipid peroxidation and cell death can be significantly rescued by Fer-1.

On the basis of these findings, a rescue assay was conducted to confirm the role of ESR1 in SLC7A11-mediated IR-induced ferroptosis. As shown in Figures 6A–D, SLC7A11 can also rescue siESR1-increased lipid peroxidation and cell death. Over-expression of SLC7A11 rescue IR-induced ferroptosis in ESR1 knockdown cells.

For further verification of the ESR1 effect, we overexpressed ESR1 in ER-negative cells. The data showed that the upregulation of ESR1 could increase SLC7A11 (Figure 6E) and rescue IR-induced lipid peroxidation (Figures 6F,G) and cell death (Figures 6H,I). Our results indicate that ESR1 promotes SLC7A11 transcription to resistance ferroptosis and cytotoxic stress, thereby promoting tumor malignancy.

We next detected whether the reduction of SLC7A11 was regulated at the transcriptional level or posttranscriptional level. The mRNA and protein levels of SLC7A11 inconsistently changed in IR treated at 24 h (Figure 4B), indicating that IR regulated the expression of SLC7A11 post-transcriptionally. To test this assumption, MG132, which acts as a blocker in the ubiquitin–proteasome pathway, was used and we found that SLC7A11 expression was accumulated by MG132 (Figures 7A,B). These observations suggest that SLC7A11 is regulated by the ubiquitin–proteasome pathway. By using the UbiBrowser database, NEDD4L is the one with the highest confidence to interact with SLC7A11 (Supplementary Table S3) and serves as a potential E3 ligase (http://ubibrowser.ncpsb.org.cn/ubibrowser/) (Li et al., 2017). Our data showed that NEDD4L knockdown rescued IR-induced degradation of SLC7A11 (Figure 7C). We then confirmed the interaction between SLC7A11 and NEDD4L by IP experiments, and a stronger binding was detected after IR treatment (Figure 7D), suggesting that NEDD4L associates with SLC7A11 in ER-positive cells. Consistent with these observations, we found that knockdown of NEDD4L markedly reduced the ubiquitination of SLC7A11 in ER-positive cells (Figure 7E). Taken together, IR could increase the interaction between NEDD4L and SLC7A11.