Salidroside sensitizes Triple-negative breast cancer to ferroptosis by SCD1-mediated lipogenesis and NCOA4-mediated ferritinophagy

Graphical abstract

Highlights

• •Salidroside (Sal) inhibits TNBC by inducing ferroptosis, associated with increased intracellular Fe²⁺ and lipid peroxidation.
• •Sal increases the level of intracellular Fe²⁺ by promoting NCOA4-mediated ferritinophagy.
• •Sal induces lipid peroxidation by inhibiting SCD1-mediated lipogenesis of monounsaturated fatty acid.
• •Sal regulates ferritinophagy and lipogenesis via the PI3K/AKT/mTOR pathway.
• •Sal might be a therapeutic drug for treating TNBC.

Abstract

Introduction

Triple-negative breast cancer (TNBC) is the primary cause of breast cancer-induced death in women. Literature has confirmed the benefits of Salidroside (Sal) in treating TNBC. However, the study about potential therapeutic targets and mechanisms of Sal-anchored TNBC remains limited.

Objective

This study was designed to explore the main targets and potential mechanisms of Sal against TNBC.

Methods

Network pharmacology, bioinformatics, and machine learning algorithm strategies were integrated to examine the role, potential targets, and mechanisms of the Sal act in TNBC. MDA-MB-231 cells and tumor-bearing nude mice were chosen for in vitro and in vivo experimentation. Cell viability and cytotoxicity were determined using CCK-8, LDH test, and Calcein-AM/PI staining. Antioxidant defense, lipid peroxidation, and iron metabolism were explored using glutathione, glutathione peroxidase, malondialdehyde (MDA), C11-BODIPY 581/591 probe, and FerroOrange dye. Glutathione peroxidase 4 (GPX4) or stearoyl-CoA desaturase 1 (SCD1) overexpression or nuclear receptor co-activator 4 (NCOA4) deficiency was performed to demonstrate the mechanism of Sal on TNBC.

Results

The prediction results confirmed that 22 ferroptosis-related genes were identified in Sal and TNBC, revealing that the potential mechanism of the Sal act on TNBC was linked with ferroptosis. Besides, these genes were mainly involved in the mTOR, PI3K/AKT, and autophagy signaling pathway by functional enrichment analysis. The in vitro validation results confirmed that Sal inhibited TNBC cell proliferation by modulating ferroptosis via elevation of intracellular Fe²⁺ and lipid peroxidation. Mechanistically, Sal sensitized TNBC cells to ferroptosis by inhibiting the PI3K/AKT/mTOR axis, thereby suppressing SCD1-mediated lipogenesis of monounsaturated fatty acids to induce lipid peroxidation, additionally facilitating NCOA4-mediated ferritinophagy to increase intracellular Fe²⁺ content. The GPX4 or SCD1 overexpression or NCOA4 deficiency results further supported our mechanistic studies. In vivo experimentation confirmed that Sal is vital for slowing down tumor growth by inducing ferroptosis.

Conclusions

Overall, this study elucidates TNBC pathogenesis closely linked to ferroptosis and identifies potential biomarkers in TNBC. Meanwhile, the study elucidates that Sal sensitizes TNBC to ferroptosis by SCD1-mediated lipogenesis and NCOA4-mediated ferritinophagy, regulated by PI3K/AKT/mTOR signaling pathways. Our findings provide a theoretical basis for applying Sal to treat TNBC.

Introduction

Triple-negative breast cancer (TNBC), an aggressive molecular subtype of breast cancer with distinctive biological behavior and clinicopathological features [1], accounts for approximately 15 % of all breast cancers [2]. Due to the absence of targetable hormone receptors and human epidermal growth factor receptor 2 expression, chemotherapy remains the mainstay of treatment [3]. However, chemotherapy benefits only a few early-stage patients with TNBC who are susceptible to it, while those of advanced stage generally suffer from poor response, resulting in increased mortality [3]. Treatment of TNBC remains clinically challenging. As TNBC exhibits a poor prognosis and limited treatment options compared to other breast cancer subtypes, primary prevention, and effective treatments are of particular importance. Consequently, there is an imperative demand to identify the mechanism and innovative treatment approaches of TNBC.

Ferroptosis is a unique form of programmed cell death that is mechanistically and morphologically different from other deaths. Ferroptosis is driven by iron-dependent phospholipid peroxidation and is mainly regulated by lipid metabolism, antioxidant system, iron metabolism, and signaling pathway [4]. According to previous studies, among the various breast cancer subtypes, TNBC is particularly susceptible to ferroptosis [5]. Moreover, emerging evidence indicates that ferroptosis plays an important role in treating TNBC [6]. The progression and metastasis of TNBC could be suppressed by reducing GPX4-driven autophagy-dependent ferroptosis with the treatment of Anomanolide C [7]. Therefore, triggering ferroptosis in tumor cells may emerge as a viable strategy for accelerating the production of anticancer drugs [8].

Active compounds of traditional Chinese medicines have high efficacy and low toxicity, which can promote postoperative recovery, prevent TNBC metastasis and recurrence, and benefit patients with TNBC [9], [10]. Salidroside (Sal), the main active ingredient extracted from Rhodiola rosea (a valuable medicinal plant as a botanical medicine), has the potential efficacy of being anticancer, antioxidant, anti-depressant, anti-hyperlipidemic, anti-inflammatory, and immunomodulatory [11], [12], [13]. Several studies have explored the effect of Sal in treating cardiomyopathy, acute liver injury, and Alzheimer's disease, all through inhibiting ferroptosis [14], [15], [16]. Although early studies have revealed that Sal can regulate EGFR/Jak2/STAT3 signaling through matrix metalloproteinases 2 (MMP2) and inhibit the migration and invasion of MDA-MB 231 cells [17], there is a lack of in vivo animal experiments for further validation. Notably, the exact mechanism of Sal against TNBC concerning ferroptosis remains unclear. Consequently, clarifying TNBC pathogenesis from multiple dimensions is crucial for developing effective therapeutic drugs.

To examine the role and mechanism of Sal in TNBC, the study used strategies such as network pharmacology, bioinformatics, and machine learning algorithms. Moreover, MDA-MB-231 cells and tumor-bearing nude mice were selected to perform in vitro and in vivo experiments to further validate the findings. Our study demonstrated the mechanism of Sal in anti-TNBC, providing fundamental structural evidence for Sal drug development. The workflow chart is presented in Fig. 1.

Fig. 1.

Graphical abstract.

Materials and methods

Data source

The drug-related databases, including HERB (https://herb.ac.cn/), Encyclopedia of Traditional Chines Medicine (ETCM, https://www.tcmip.cn/ETCM/index.php/Home/), Comparative Toxicogenomics Database (CTD, https://ctdbase.org/), and TargetNet (https://targetnet.scbdd.com/), were applied to gather the related targets of Sal. The disease-related databases Online Mendelian Inheritance in Man (OMIM, https://www.ncbi.nlm.nih.gov/omim/), CTD, and GeneCards (https://www.malacards.org/) were utilized to acquire the TNBC-related genes. Besides, TCGA-TNBC datasets were obtained from The Cancer Genome Atlas (TCGA, https://tcga-data.nci.nih.gov/tcga/). The TNBC clinical information was downloaded using the University of California, Santa Cruz (UCSC) Xena (https://xena.ucsc.edu) website.

Additionally, TNBC-related datasets GSE31519, GSE65194, GSE38959, GSE61724, and GSE21653 were downloaded using the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo) database.

Collection of the ferroptosis-related genes (FRGs)

FerrDb, a ferroptosis database, includes ferroptosis-related markers, regulatory factors, and ferroptosis-related diseases that are managed and identified [18]. We gathered FRGs using FerrDb.

Analysis of differentially expressed genes (DEGs)

The “Limma” package of R software was utilized to perform DEG analysis [19]. Besides, the genes that satisfied the criteria of |log (fold change, FC)| > 1 and p < 0.05 were defined as the DEGs.

Weighted gene co-expression network analysis (WGCNA)

The WGCNA is a systematic method for describing gene association patterns between samples [20]. The WGCNA package was used to analyze the TCGA-TNBC and GEO datasets.

Gene ontology (GO), Kyoto encyclopedia of genes and genomes (KEGG), and GeneMANIA functional analyses

The “ClusterProfiler” R package was frequently used for GO enrichment and KEGG functional annotation research. The visualizations of GO and KEGG were created using the “ggplot” package. Moreover, GeneMANIA (https://genemania.org/) was also employed to conduct functional analysis.

Construction of protein–protein interaction (PPI) network

The Search Tool for the Retrieval of Interacting Genes (STRING, https://string-db.org) website tool was applied to determine the protein associations [21]. Here, we imported the shared FRGs into STRING and set the species as Homo sapiens to obtain the PPI network. Subsequently, it was further visualized and analyzed in Cytoscape. Additionally, the Sal-FRGs-TNBC interaction network was generated using Cytoscape.

Machine learning algorithms

Least Absolute Shrinkage and Selection Operator (LASSO) regression is a machine learning algorithm for regression that is based on linear relationships and regularization [22]. Random Forest (RF) is a machine learning algorithm that performs regression using bootstrap resampling and is based on decision trees [23]. Support Vector Machine Recursive Feature Extraction (SVM-RFE) is a machine learning algorithm of ranking features based on a recursive feature deletion process [24]. To identify candidate diagnostic biomarkers for TNBC, this study combined the LASSO, RF, and SVM-RFE algorithms for identifying potential biomarkers [25], [26]. Briefly, we used the top 15 genes obtained from the MCC algorithm in the PPI network as inputs for the expression profiles in the TCGA-TNBC, GSE61724, and GSE21653 datasets. We then use the ‘glmnet’ package in R software for LASSO analysis, the ‘randomForest’ package for RF analysis, and the ‘e1071’ package for SVM-RFE analysis. Finally, the overlapping genes identified by the three machine learning algorithms were defined as candidate biomarkers for the TNBC.

Receiver operating characteristics (ROC) curve

First, we established a ROC curve and calculated the area under the curve (AUC) value. Subsequently, the diagnostic value of the biomarkers was estimated in the TCGA test set, as well as in the GSE61724 and GSE21653 validation sets, according to the AUC value. Biomarkers with an AUC of ≥0.6 were defined as biomarkers with diagnostic value.

Reagents and antibodies

Sal was obtained from Meilunbio (Dalian, China). Erastin, Ferrostatin-1 (Fer-1), and 3-Methyladenine (3-MA) were supplied by Aladdin (Shanghai, China). Deferoxamine (DFO), Glutathione (GSH), VO-Ohpic trihydrate, SC79, Rapamycin, MHY1485, Oleic acid (OA), and Stearic acid (SA) were provided by MedChemExpress (NJ, USA). BODIPY™ 581/591 C11 probe, Lipofectamine 3000, Lipofectamine RNAiMAX, and Opti-MEM were purchased from Thermo Fisher Scientific (MA, USA). FerroOrange probe was acquired from Dojindo (Kumamoto, Japan). CCK-8, LDH assay kit, Calcein-AM/PI staining kit, GSH and GPX assay kit, BCA protein assay kit, and MDA assay kit were sourced from Beyotime (Nantong, China). RNA-easy isolation reagent, HiScript II Q RT SuperMix, and ChamQ SYBR qPCR Master Mix were purchased from Vazyme (Nanjing, China). Plasmids containing the insert SCD1 or GPX4 were procured from GenePharma (Shanghai, China). The primary antibodies against PTGS2, FTH, NCOA4, and SQSTM1/p62 were supplied by Santa Cruz Biotechnology (Santa Cruz, USA), the GPX4, SREBP1, PTEN, PI3K, AKT, mTOR, p-mTOR, 4E-BP1, and LC3 were obtained from Proteintech (Wuhan, China), and the p-PI3K, p-AKT, p-4E-BP1, and SCD1 were supplied by Cell Signaling Technology (MA, USA). All reagents and antibodies used in this study are summarized in sup. Tables S1 and S2.

Cell culture

MDA-MB-231 cells were obtained from the Chinese Academy Cell Resource Center (Shanghai, China) and incubated in Dulbecco's modified Eagle medium (DMEM; KGM12800-500, Keygen Biotech, China), supplemented with 10 % fetal bovine serum (FBS, RY-F22, Royacel, China). Cells were placed at 37 °C with 5 % CO₂, and the culture medium was replaced every 2–3 days.

Cell viability assay

MDA-MB-231 cells were seeded into 96-well plates at a density of 7000 to 9000 cells per well, and cell viability was assessed using the CCK-8 assay according to the relevant user manual (C0043, Beyotime, China). Finally, a microplate reader (Rayto RT-6000, USA) was utilized to determine the OD values of the sample at 450 nm.

Lactate dehydrogenase (LDH) cytotoxicity assay

According to the manufacturer's protocol, the cytotoxicity was evaluated by applying the LDH detection kit (C0017, Beyotime, China). The OD value was measured at 492 nm, employing a microplate reader to determine the relative amount of LDH. The LDH relative release amount was calculated using the following formula: LDH relative release amount (%) = (OD [treated sample] – OD [control sample]) / (OD [maximum enzyme activity of cells] – OD [control sample]) × 100 %.

Calcein-AM/Propidium Iodide (PI) staining

The cell viability and cytotoxicity were conducted as previously described [27]. Briefly, the treated cell samples were collected to be incubated with the Calcein-AM and PI (C2015M, Beyotime, China) at 37 °C for 30 min and photographed using an inverted fluorescence microscope (Nikon Ts2R, Japan).

Evaluated the levels of malondialdehyde (MDA), lipid reactive oxygen species (ROS), glutathione (GSH), and glutathione peroxidase (GPX)

The lipid peroxidation (MDA) assay kit (S0131M, Beyotime, China), C11-BODIPY 581/591 probe (D3861, Thermo Fisher Scientific, USA), glutathione (GSH) detection assay kit (S0052, Beyotime, China), and glutathione peroxidase (GPX) assay kit (S0056, Beyotime, China) were employed to measure the levels of MDA, lipid reactive oxygen species (ROS), GSH, and GPX in cells and tumor tissue, following the instructions provided in the manual.

Detection of Fe²⁺

The intracellular ferrous levels were referred to in the previous study [28], namely, adding the FerroOrange dye (F374, Dojindo, Japan) to the cells and incubating (37 °C, 30 min). Then, pictures were acquired with an inverted fluorescence microscope.

The Fe²⁺ levels in tumor tissue were detected applying the ferrous iron colorimetric assay kit (E-BC-K773-M, Elabscience, China) according to the manufacturer's protocol. And the OD value was measured at 593 nm employing a microplate reader.

Plasmid overexpression and knockdown

The plasmids containing the insert SCD1 (NM_005063.4) or GPX4 (NM_002085.5), and siRNAs targeting the NCOA4 gene used in this study were obtained from GenePharma (Shanghai, China). Additionally, the siRNA sequence is provided in Sup. Table S3. Cell transfection was performed using the Lipofectamine 3000 (L3000150, Thermo Fisher Scientific, USA) or Lipofectamine RNAiMAX reagent (LMRNA015, Thermo Fisher Scientific, USA) following the manufacturer's instructions. Protein levels were analyzed by Western blot (WB) using anti-GPX4, anti-SCD1, and anti-NCOA4, respectively.

Quantitative real-time polymerase chain reaction (RT-PCR) analysis

First, the total RNA was isolated and extracted using an RNA-easy Isolation reagent (R701, Vazyme, China). Second, the HiScript II Q RT Supermix (R223, Vazyme, China) was used in reverse cDNA transcription. The program was as follows: 50 °C, 15 min; 85 °C, 5 s. Third, the cDNA was amplified with the ChamQ SYBR RT-PCR Master Mix (Q311, Vazyme, China) in LightCycler96 RT-PCR detection equipment (Roche, USA). The gene-associated primer sequence is detailed in Sup. Table S4, obtained from Sangon Biotech (Shanghai, China).

Western blotting (WB) analysis

Protein expression levels were detected through WB analysis. The experiment was performed as previously described [29]. The radio immunoprecipitation assay (RIPA) buffer supplemented with phenylmethanesulfonyl fluoride (PMSF), ethylene diamine tetraacetie acid (EDTA), and protease inhibitors was applied to lyse cells and tumor tissue samples. The protein content was detected by the bicinchoninic acid (BCA) protein assay kit (P0010, Beyotime, China). The protein samples were separated by SDS polyacrylamide gel electrophoresis. Subsequently, the polyvinylidene fluoride (PVDF) membranes were transferred and blocked with 5 % skim milk before being incubated with the corresponding primary antibodies, such as PTGS2 (1:500), FTH (1:300), NCOA4 (1:300), SQSTM1/p62 (1:500), GPX4 (1:2000), SREBP1 (1:2000), PTEN (1:6000), PI3K (1:500), AKT (1:6000), mTOR (1:20000), p-mTOR (1:10000), 4E-BP1 (1:2000), and LC3 (1:2000), p-PI3K (1:1000), p-AKT (1:2000), p-4E-BP1 (1:1000) and SCD1 (1:1000). It was incubated with the horseradish peroxidase (HRP)-conjugated secondary antibodies after washing with Tris-buffered saline and Tween-20 buffer. Finally, the enhanced chemiluminescence solution (180–5001, Tanon, China) was added to be imaged by the imaging system (Tanon 5200, China).

In vivo tumor mice model

Healthy female BALB/c nude mice (6 weeks old, 18–20 g) were obtained from the Hangzhou Ziyuan experimental animal cultivation farm (Zhejiang, China). The tumor model was established based on a previous study [30]. Briefly, the MDA-MB-231 cell (4 × 10⁷ cells/mL) suspension was prepared with phosphate-buffered saline (PBS) and injected into the axilla of mice (0.1 mL/mouse).

After the tumor volume reached approximately 100 mm³, the mice were freely distributed into four groups (15 mice/group) which are as follows: Control group, Fer-1 group (1 mg/kg), Sal group (100 mg/kg), and Sal (100 mg/kg) + Fer-1 (1 mg/kg) combined group. Based on the above, treatment groups were given Sal (intraperitoneal injection, i.p.), Fer-1 (i.p.), or their combination, and the control group was given equivalent saline (i.p.) daily. An administration of once daily for 14 days was conducted. Moreover, the weight and tumor volume of the mice were measured daily, and the tumor volume was calculated using the formula 0.5 × longest diameter × shortest diameter². The nude mice were given unrestricted access to food and water during the experiment. Eventually, the tumors were removed and weighed after the end of treatment.

Hematoxylin and eosin (HE) staining

Xenograft tumor tissue was fixed, dehydrated, embedded, and sectioned. Subsequently, the slides were deparaffinized, rehydrated, and incubated with the HE staining kit (AR1180, BOSTER, China). Meanwhile, the nucleus was stained with hematoxylin, and the cytoplasm was stained with eosin. The slides were then dehydrated, made transparent, and sealed with neutral gum. The image was obtained using a light microscope (Olympus CX31, Japan).

Immunohistochemistry (IHC)

The embedded tumor sections were initially subjected to dewaxed, dehydrated, and rehydrated. Subsequently, the Ki67 (1:200, ab16667, Abcam), 4-HNE (1:200, bs-6313R, Bioss), and PTGS2 (1:200, sc-19999, Santa Cruz) antibodies were added. The polymer anti-rabbit/mouse immunoglobulin G (IgG)-HRP detection kit (SV0004, BOSTER, China) was then added to incubate. Finally, it was stained with the diaminobenzidine (DAB) kit (AR1027, BOSTER, China) and captured by a light microscope.

The analysis of molecular docking and intermolecular interactions

The molecular docking and intermolecular interactions analysis method is vital to evaluate the capability of binding and intermolecular interactions between the protein and ligand [31], [32]. The 3D structure of Sal was downloaded from PubChem Database (https://pubchem.ncbi.nlm.nih.gov/), and the PI3K (PDB ID: 3APF), AKT (PDB ID: 2UZT), and mTOR (PDB ID: 4JT6) were acquired from the PDB database (https://www.rcsb.org/). Furthermore, Autodock vina software was utilized for docking and visualized by PyMOL software.

Furthermore, the Autodock vina molecular docking results were imported to LigPlot + software to explore the intermolecular interactions between ligand (Sal) and proteins (PI3K, AKT, and mTOR).

Cellular thermal shift assay (CETSA) analysis

CETSA was used to assess the protein stability of mTOR, as described in a previous study [33]. Briefly, the MDA-MB-231 cell lysates were first split into two groups and incubated with either PBS or Sal (50 μM) for 1 h at 37 °C. Next, the samples in two groups were split into PCR tubes and heated for 3 min at 37 °C, 41 °C, 45 °C, 49 °C, 53 °C, 57 °C, 61 °C, 65 °C, 69 °C, and 73 °C. Third, the samples were centrifuged at 12000 rpm at 4 °C for 15 min. Finally, the samples were collected and analyzed for mTOR protein expression using WB analysis.

Drug affinity responsive target stability (DARTS) analysis

DARTS is a robust method for detecting small molecule protein targets [34]. MDA-MB-231 cell lysates were incubated with PBS or Sal (50 μM) for 1 h at room temperature, as previously described [35]. The samples were then digested for 30 min by adding pronase. Subsequently, the digestion was terminated by the addition of SDS-PAGE protein loading buffer (5 × ). Finally, mTOR protein expression was analyzed using the WB analysis.

Statistical analysis

The statistical data, if applicable, are represented as mean ± SD. All statistical analyses were performed using GraphPad Prism software (version 9.0). Student's t-test analysis was used to determine differences between only two groups, and the one-way analysis of variance was applied to multiple group comparisons. All experiments were repeated at least three times independently. In all analyses, p < 0.05 was considered statistically significant.

Results

Collection of Sal- and TNBC-related targets

To obtain the targets of Sal, we used “Salidroside” as a keyword to search the drug-related databases. The results found that the Sal-related targets were 37 in HERB, 30 in ETCM, 68 in CTD, and 55 in TargetNet (p > 0). After excluding duplicates, a total of 178 targets remained (Fig. 2A). To determine the TNBC-related targets, OMIM, CTD, and GeneCards databases were employed. The inclusion criteria for the TNBC target screening conditions were set as the Inference score ≥ 15 in the CTD database and the Relevance score ≥ 20 in the GeneCards database. Subsequently, 1374 targets were obtained through conditional screening, including OMIM (100), CTD (473), and GeneCards (1021) (Fig. 2B).

Fig. 2.

Identification of Sal and TNBC-related genes. (A) The drug-related database (HERB, ETCM, CTD, and TargetNet) explores the targets of Sal. (B) TNBC-related targets in various disease databases (OMIM, CTD, and GeneCards). The volcano (C) and heatmap (D) of the differentially expressed genes in the TCGA sample. The PCA analysis results of GEO datasets remove the batch effect before (E) and after rectification (F). The differentially expressed genes were shown in the volcano plot (G) and heatmap (H) in GEO datasets. The WGCNA R package was used to construct co-expression networks in the TCGA sample (I-N). (I) A sample dendrogram and feature heatmap were drawn based on the Euclidean distance by the average clustering method. (J) The scale-free topology fitting index (R²) was 0.85 when the soft threshold was set at 4. (K) The relationship heatmap between control and TNBC module, modules are represented with different colors. The red represents a positive correlation, while the blue represents a negative correlation. (L) Gene dendrogram from Average Linkage Hierarchical Clustering based on Dynamic Tree Cut, with modules merged dynamically at the bottom. (M)Scatterplot of module membership against gene significance. (N) The Venn diagram exhibited the number of intersection genes between WGCNA and DEG analysis in the TCGA sample. The WGCNA analysis results of GEO datasets (O-T). (O) We created a sample dendrogram and heatmap based on Euclidean distances using average clustering. (P) The scale-free topology fitting index (R²) was 0.88 when the soft threshold was set at 3. (Q) An image of the sample dendrogram and module assignment is displayed, with six modules identified based on dynamic tree cutting. (R) The relationship heatmap between the control and TNBC module; modules are represented with different colors. (S) Module membership plotted against gene significance. (T) The TNBC-related genes obtained from the intersection of WGCNA and DEG in GEO datasets. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Identification of DEGs and essential module genes in TNBC

To increase the reliability of the results, we downloaded the TNBC sample from TCGA and GEO datasets to conduct DEG analysis. The results found that 5252 DEGs were recognized in the TCGA sample, with 2666 downregulated DEGs and 2586 upregulated DEGs (Fig. 2C-D). Meanwhile, we performed the DEG analysis in GEO datasets. The first step involved directly removing batch effects from TNBC datasets GSE31519, GSE65194, and GSE38959 (Table 1) using the surrogate variable analysis (sva) algorithm, followed by generating principal component analysis (PCA) plots for all datasets (Fig. 2E-F). Furthermore, 2707 DEGs were identified, which include 1340 downregulated DEGs and 1367 upregulated DEGs (Fig. 2G-H).

Table 1.

The information of TNBC-related datasets.

Dataset	Platform	Total	TNBC	Normal	Note
TCGA-BRCA	/	230	117	113	Test dataset
GSE31519	GPL96	67	67	0	Test dataset
GSE65194	GPL570	66	55	11	Test dataset
GSE38959	GPL4133	43	30	13	Test dataset
GSE61724	GPL6244	20	16	4	Validation dataset
GSE21653	GPL570	104	75	29	Validation dataset

To recognize the critical module connected with TNBC, we performed WGCNA of the TCGA-TNBC sample. With the soft threshold power of 4, the scale-free topological fitting index achieved 0.85 in the co-expression network (Fig. 2I). Additionally, 14 modules were identified, of which the brown module was highly correlated with TNBC (Fig. 2J). Moreover, the scatterplot revealed a significant relationship (cor = 0.83, p < 1e − 200) between the brown module and TNBC (Fig. 2K). Hence, 1442 genes were acquired by intersecting the brown module-related genes of WGCNA with DEGs identified by Limma (Fig. 2L). The GEO dataset revealed that the soft threshold power was 3, and the correspondence unscaled topological fitting index achieved 0.88 (Fig. 2M). Six modules were identified, and the brown module was highly correlated with TNBC (Fig. 2N). Additionally, the scatterplot revealed a positive significant relationship (cor = 0.88, p < 1e − 200) between the blue module and TNBC (Fig. 2O). Hence, 680 genes were acquired through the intersection of the brown module-related genes of WGCNA with DEGs identified by Limma (Fig. 2P).

Predictive and experiment analysis results elucidating Sal anti-TNBC by promoting ferroptosis

Furthermore, 3296 TNBC-relevant genes were acquired by intersecting the disease, TCGA, and GEO databases (Fig. 3A). This intersection of Sal genes and TNBC genes acquired 75 shared genes (Fig. 3B), which represented the therapeutic genes of Sal inhibiting TNBC. To explore the potential mechanism of Sal on TNBC, the GeneMANIA tool was utilized to perform a functional analysis of 75 shared genes. The result revealed that the shared genes are mainly involved in processes related to oxidative stress, response to oxidative stress, cell death in response to oxidative stress, regulation of lipid biosynthetic process, and regulation of lipid metabolic process (Fig. 3C). Ferroptosis is an iron-dependent form of programmed cell death driven by oxidative abnormalities of lipid peroxidation and redox imbalance [36]. Among the factors that cause oxidative stress in cells, oxidative modification of lipids in membrane bilayers, particularly lipid peroxidation, has emerged as an important regulator of cell fate. Extensive lipid peroxidation commits cells to death via a distinct cell death mechanism termed ferroptosis [37]. Accordingly, we speculated that the underlying mechanism of Sal against TNBC could be linked to ferroptosis.

Fig. 3.

Sal can promote ferroptosis based on both predictive and experimental evidence. The predictive evidence supports that Sal can induce ferroptosis (A-C). (A) The TNBC-related genes from disease, TCGA, and GEO database. (B) The common genes in the Venn plot of Sal therapeutic TNBC. (C) The functional enrichment analysis results of Sal and TNBC-related genes via GeneMANIA. The in vitro experimental evidence supports the idea that Sal can induce ferroptosis (D-G). (D) Cell viability was measured after being treated with different concentration of Erastin (0, 1, 2.5, 5, 10 μM) for 24 h (n = 6). (E) Cell viability was measured after being pretreated with different concentration of ferroptosis inhibitors Fer-1 (0, 0.5, 1, 2.5, 5 μM) for 1 h, and then being treated with or without Erastin (5 μM) for 24 h (n = 6). Cell viability (F) and LDH release (G) were measured after being pretreated with Fer-1 (2 μM), GSH (1 mM), or DFO (50 μM) for 1 h, and then being treated with or without Sal (50 μM) for 24 h (n = 6). The levels of cell viability/cytotoxicity (H) measured with Calcein-AM/PI cell viability/cytotoxicity assay kit (n = 3), GSH (I) measured with GSH detection assay kit (n = 6), GPX (J) measured by GPX assay kit with NADPH (n = 6), MDA (K) measured with lipid peroxidation (MDA) assay kit (n = 6), ferrous (L) were detected with FerroOrange dye (n = 3), and lipid ROS (M) measured with C11-BODIPY 581/591 probe (n = 3), in MDA-MB-231 cells treated with Fer-1 (2 μM), Sal (50 μM) or their combination for 24 h. The in vivo experimental evidence supports the idea that Sal can induce ferroptosis (N-X). (N) MDA-MB-231 cells suspended in PBS were injected subcutaneously into the left flank of each mouse to build model. Seventeen days after inoculation (when tumors reached a volume of ∼ 100 mm³), mice were randomly divided into four groups, and then received daily intraperitoneal injection Fer-1 (1 mg/kg), Sal (100 mg/kg) or the drug combination for 14 days (n = 15). (O) Photograph of tumors in four groups (n = 6). (P) Tumor volume was monitored every day (n = 6). (Q) Tumor weight (n = 6). (R) Body weight of nude mice was measured every day during treatment (n = 6). (S) H&E staining and Ki67 immunohistochemical staining in different groups (n = 3). The levels of intracellular GSH (T), GPX (U), and MDA (V) and Fe²⁺ (W) were measured in different treatment groups (n = 6). (X) The expressions of 4-HNE and PTGS2 were determined by immunohistochemical staining (n = 3). Scale bar, 50 μm. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

Initially, we validated the effect of Sal on TNBC in MDA-MB-231 cells. It was observed that Sal suppressed cell activity in correlation with dose and time for treatment (Supplementary Figure S1). Based on this, the IC₅₀ of Sal was chosen for further investigation. Furthermore, Erastin, an inducer of ferroptosis, could suppress cell activity in correlation with the dose for treatment (Fig. 3D). Conversely, the effect could be reversed by Fer-1 (Fig. 3E), demonstrating the occurrence of ferroptosis in MDA-MB-231 cells. We further validated whether Sal improves TNBC via triggering ferroptosis. The ferroptosis inhibitors Fer-1, DFO, and GSH were added to MDA-MB-231 cells. The results revealed that the suppression of cell activity by Sal was partially alleviated through the treatment of ferroptosis inhibitors concerning Fer-1, DFO, and GSH (Fig. 3F-G). Fer-1 was selected for further research in the following experiments, which inhibited ferroptosis mainly by regulating the intracellular antioxidant system and iron metabolism. The Calcein-AM/PI staining revealed that the cell activity suppression was partially alleviated through treatment of Fer-1 administration (Fig. 3H). Subsequently, Fer-1 addition significantly enhanced the GSH level. In contrast, GPX activity was reduced by Sal (Fig. 3I-J), and MDA, Fe²⁺, and lipid ROS levels all increased significantly in the Sal group (Fig. 3K-M). However, the results mentioned above in the Sal group were partially reversed by Sal + Fer-1, indicating that Sal induces ferroptosis in MDA-MB-231 cells.

Meanwhile, for the potential anticancer activity of Sal in vivo, a TNBC animal model was obtained through inoculation with MDA-MB-231 cells in female nude mice (Fig. 3N). As depicted, the tumor growth in the Sal group was markedly slower compared with the Fer-1 alone and co-administration groups (Fig. 3O-P). Furthermore, the tumor weight in the Sal group was significantly lower than in other groups (Fig. 3Q). The body weight also exhibited small changes during the treatments, indicating that the treatments had no apparent toxicity and were well-tolerated (Fig. 3R). Furthermore, HE staining results of tumor tissues revealed that the area of necrosis in the Sal treatment group was much higher than in other groups (Fig. 3S).

Additionally, following Sal and Fer-1 administration, the Ki67 positive region upregulated more distinctly than that received Sal alone (Fig. 3S). All data uncovered that Sal could fight TNBC without overt toxic effects. Afterward, GSH and GPX activity levels were significantly reduced, but MDA and Fe²⁺ levels increased in the Sal group compared with other groups (Fig. 3T-W). However, this result was contrary to the results of the combination group. Moreover, IHC results found that 4-HNE was enhanced in the Sal group, demonstrating that Sal promotes the accumulation of lipid peroxides (Fig. 3X). Meanwhile, the positive expression of PTGS2 was elevated in the Sal group in the IHC analysis. This supported the idea that Sal is crucial in inhibiting TNBC tumor growth by inducing ferroptosis (Fig. 3X).

Functional enrichment analysis results of shared FRGs

We identified 22 genes in Sal and TNBC that were associated with ferroptosis, which were defined as shared FRGs (Fig. 4A). To understand the biological functions and signaling pathways of the common FRGs, we performed GO, KEGG, and GeneMANIA functional analyses. Initially, the GO analysis was significantly enriched. The results revealed that the primary biological processes (BP) involved cellular response to biotic stimulus (GO:0071216), external stimulus (GO:0071496), and chemical stress (GO:0062197). Moreover, cellular component analysis disclosed that the genes were primarily associated with the caveola (GO:0005901), endoplasmic reticulum lumen (GO:0005788), and plasma membrane raft (GO:0044853). Additionally, the kinase activity concerning MAP (GO:0004707), protein serine (GO:0106310), and protein serine/threonine (GO:0004674) was the primary molecular function (Fig. 4B-C).

Fig. 4.

Functional enrichment analysis and PPI network analysis of FRGs. (A) The shared genes of Sal, TNBC, and ferroptosis. (B) A column diagram showing the significance of the results of GO enrichment. (C) GO enrichment diagrams, including BP, CC, and MF. (D) KEGG enrichment bubble diagram. (E) The functional association of targets was analyzed using GeneMANIA. (F) A PPI network based on the STRING database. (G) Cytoscape visualizes the PPI network. (H) The Cytohubba plug-in screened the hub genes in the PPI network. (I) The Sal-FRGs-TNBC network. (J) LASSO regression of the 15 hub FRGs calculated by Cytohubba. (K) LASSO regression parameter selection using cross-validation. (L) In RF analysis, TNBC and total group diagnostic errors were visualized. (M) Nine biomarkers with a score greater than 4.5 were selected based on the RF algorithm presenting the MeanDecreaseGini of the 15 genes in TNBC. (N) Using the SVM-RFE algorithm, the FRGs with the lowest error and highest accuracy were considered the most suitable candidate biomarkers. (O) Triangle Venn diagram exhibiting the genes shared/unique among the LASSO, RF, and SVM-RFE algorithms. The expression comparison of 9 candidate biomarkers (HIF1A, IL6, mTOR, SIRT1, HMOX1, HSPA5, MAPK3, GSK3B, and XBP1) between TNBC and normal groups in the TCGA test dataset (P), GSE61724 validation dataset (Q), and GSE21653 validation dataset (R). (S) The IHC staining results were obtained from the HPA database. The ROC curves of each candidate biomarker (HIF1A, IL6, mTOR, SIRT1, HMOX1, HSPA5, MAPK3, GSK3B, and XBP1) in TCGA test set (T), GSE61724 validation set (U), and GSE21653 validation dataset (V).

Furthermore, the efficacy of Sal treatment of TNBC through triggering ferroptosis was tightly associated with the KEGG enrichment pathway, namely, lipid and atherosclerosis (hsa05417), mTOR signaling pathway (hsa04150), PI3K/AKT signaling pathway (hsa04151), and Autophagy-animal (hsa04140) signaling pathway (Fig. 4D).

Afterward, the GeneMANIA tool conducted a functional analysis to reveal the molecular-functional relationship. The results suggested that these genes were closely connected with oxidative stress, lipid and fatty acid metabolic processes, lipid biosynthetic processes, and the regulation of autophagy processes (Fig. 4E).

PPI network construction and identification of the key FRGs

To study protein interactions at common FRGs, we constructed a PPI network employing the STRING online tool (Fig. 4F). Subsequently, the CytoNCA plug-in was used to calculate topological parameters and extract the core PPI network. The degree of targets was indicated by the node size, from large to small (Fig. 4G). Concurrently, a pharmacological network was constructed to represent the relationship between Sal-FRGs-TNBC, which provided a general understanding of the association between Sal, FRGs, and TNBC (Fig. 4H). Subsequently, we calculated the top 15 hub genes using the Cytohubba plug-in, which contains HIF1A, IL6, mTOR, IL1B, PARP16, SIRT1, HMOX1, HSPA5, TGFB1, MAPK8, MAPK3, RELA, GSK3B, XBP1, and MAPK1 (Fig. 4I).

Screening hub signature genes in shared FRGs

This study combined three machine learning algorithms to recognize shared FRGs (Sal and TNBC)-associated potential diagnostic biomarkers from 15 hub genes screened by the PPI network. First, 11 biomarkers were obtained from the LASSO algorithm, including HIF1A, IL6, mTOR, SIRT1, HMOX1, HSPA5, TGFB1, MAPK3, RELA, GSK3B, and XBP1 (Fig. 4J-K). Meanwhile, the RF algorithm identified nine potential biomarkers based on their importance (HSPA5, SIRT1, XBP1, MAPK3, HIF1A, HMOX1, GSK3B, IL6, and mTOR; Fig. 4L-M). Furthermore, the SVM-RFE analysis exhibited that the model involving ten genes (HSPA5, SIRT1, XBP1, MAPK3, GSK3B, HMOX1, HIF1A, IL6, MAPK1, and mTOR) was modeled to achieve the highest accuracy (Fig. 4N). Finally, three machine learning algorithms were intersected that identified genes to obtain the candidate biomarkers, namely HIF1A, IL6, mTOR, SIRT1, HMOX1, HSPA5, MAPK3, GSK3B, and XBP1 (Fig. 4O).

To further examine the importance of HIF1A, IL6, mTOR, SIRT1, HMOX1, HSPA5, MAPK3, GSK3B, and XBP1 in patients with TNBC, we analyzed the expression of the above genes in patients with TNBC and normal sample. In the TCGA-TNBC test set, the expression levels of HIF1A, mTOR, HMOX1, HSPA5, and GSK3B, were significantly elevated in the TNBC samples, while IL6, SIRT1, MAPK3, and XBP1 expression was reduced. The differences were all statistically significant (Fig. 4P). Meanwhile, the validation dataset GSE61724 was consistent with the above results, which exhibited a consistent trend for all genes except MAPK3 (Fig. 4Q). Besides, the results from validation dataset GSE21653 found that except for the decreased expression of SIRT1, MAPK3, and XBP1, the remaining genes were expressed in patients with TNBC (Fig. 4R). Furthermore, we verified the expression of the above genes using IHC results in the human protein atlas (HPA) database. The IHC results were in line with the above results. Except for XBP1, there were no data in the HPA database. HIF1A, IL6, mTOR, SIRT1, HMOX1, HSPA5, MAPK3, and GSK3B exhibited the same trend of increasing in the BC tissue (Fig. 4S).

Following this, we established ROC curves and calculated AUC values to assess the diagnostic value of the above genes. In the TCGA-TNBC dataset, the AUC of all genes was ≥ 0.6, suggesting an excellent diagnostic value (Fig. 4T). Meanwhile, the results obtained in GSE61724 and GSE21653 validation sets were almost consistent with those mentioned above. The average AUC values of the listed genes were ≥ 0.6, except for IL6 and HSPA5, in validation sets (Fig. 4U-V). These results provide strong evidence for HIF1A, mTOR, SIRT1, HMOX1, GSK3B, MAPK3, and XBP1 as diagnostic biomarkers for TNBC.

Sal-induced ferroptosis by regulating the PI3K/AKT/mTOR signaling pathway in MDA-MB-231 cells and TNBC animal model

The above bioinformatics and network pharmacology predicted results exhibited that Sal could improve TNBC pathogenesis mainly by targeting the PI3K/AKT and mTOR pathways. Besides, mTOR is a hub target in treating TNBC by Sal. In this context, we selected mTOR as a critical target to detect the PI3K/AKT/mTOR pathway-related protein (PTEN, p-PI3K, PI3K, p-AKT, AKT, p-mTOR, and mTOR) expression to confirm the predictive result.

In the in vitro experiment, the results revealed that the protein expression levels of phosphatase and tensin homolog (PTEN) increased. However, p-PI3K/PI3K and p-AKT/AKT decreased in the Sal group (Fig. 5A), indicating that Sal-triggered ferroptosis may inhibit the PI3K/AKT pathway. To further invalidate the mechanism of Sal-induced ferroptosis, the PTEN inhibitor VO-Ohpic trihydrate and AKT activator SC79 were applied to explore the vital role of the PTEN and AKT in PI3K/AKT pathway-regulated ferroptosis. It was demonstrated that VO-Ohpic trihydrate and SC79 conferred MDA-MB-231 cells resistance to ferroptosis induced by Sal (Fig. 5B-C). Apart from this, VO-Ohpic trihydrate and SC79 also significantly decreased the Sal-induced MDA and elevated ferrous levels (Fig. 5D-E). PI3K is only one of many mechanisms that regulate AKT activation. Therefore, we used the PTEN inhibitor to elucidate the role of the PI3K/AKT pathway in Sal-mediated ferroptosis. Furthermore, the VO-Ohpic trihydrate could improve the Sal-triggered p-PI3K/PI3K and p-AKT/AKT (Fig. 5F). All these data explored that, on the one hand, VO-Ohpic trihydrate can suppress Sal-induced ferroptosis. On the other side, the mechanism of Sal-induced ferroptosis is mainly through the inhibited p-AKT/AKT pathway.

Fig. 5.

Sal-induced ferroptosis is regulated by the PI3K/AKT/mTOR signaling pathway in-vitro and in-vivo experiment. (A) The protein levels of PTEN, PI3K, p-PI3K, AKT, and p-AKT were detected in MDA-MB-231 cells treated with Sal (50 μM) for 24 h (n = 3). (B-D) Cells were pretreated with VO-Ohpic (5 μM) or SC79 (10 μM) for 1 h, and then treated with or without Sal (50 μM) for 24 h. Cell viability (B) and LDH release (C) were detected (n = 6). The levels of MDA (D) detected by lipid peroxidation (MDA) assay kit (n = 6), and Fe²⁺ (E) detected by FerroOrange dye (n = 3). (F) The protein levels of PI3K, p-PI3K, AKT, and p-AKT were detected after cells were pretreated with VO-Ohpic (5 μM) for 1 h, and then Sal (50 μM) for 24 h (n = 3). (G) The protein levels of mTOR, p-mTOR, 4E-BP1, and p-4E-BP1 were detected in MDA-MB-231 cells treated with Sal (50 μM) for 24 h (n = 3). (H-J) Cells were pretreated with MHY1485 (2 μM) or Rapamycin (50 nM) for 1 h, and then treated with or without Sal (50 μM) for 24 h. Cell viability (H), LDH release (I) and MDA levels(J) were detected (n = 6). (K) The protein level of PTGS2 was detected after cells were pretreated with MHY1485 (2 μM) for 1 h and then Sal (50 μM) for 24 h (n = 3). (L) The animal protein levels of PTEN, PI3K, p-PI3K, AKT, p-AKT, mTOR and p-mTOR were measured in the Control and the Sal group (n = 3). Scale bar, 50 μm. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

To further verify the predictive results, we detected the hub biomarker mTOR's expression level by WB. The p-mTOR/mTOR protein was downregulated significantly after Sal administration (Fig. 5G). Furthermore, as the substrate of mTORC1, the p-4EBP1/4EBP1 was downregulated after Sal administration at the protein level (Fig. 5G). Afterward, the mTORC1 activator MHY1485 and inhibitor rapamycin were applied to further explore the role of mTORC1 on Sal-regulated ferroptosis. The findings exhibited that pre-treatment with MHY1485 conferred cell resistance to ferroptosis caused by Sal, whereas rapamycin enhanced Sal-induced ferroptosis (Fig. 5H-I). Additionally, MHY1485 greatly downregulated Sal-regulated MDA generation, while Rapamycin instead increased (Fig. 5J). The increased PTGS2 protein level induced by Sal was significantly inhibited by MHY1485 (Fig. 5K). The above findings were consistent with predictive results. The activity of PI3K/AKT/mTOR might be associated with the increasing sensitivity of MDA-MB-231 cells to Sal-induced ferroptosis.

Notably, the ferroptosis PI3K/AKT/mTOR signaling pathway was validated in vivo. The protein level of PTEN was upregulated, and p-PI3K/PI3K, p-AKT/AKT, and p-mTOR/mTOR were downregulated after Sal administration (Fig. 5L). Additionally, Sal could be stable-bound with mTOR (–8.3 kcal/mol) to form dense complexes (Supplementary Fig. S2C; Sup. Table S5). Additionally, intermolecular interactions revealed that the Sal docked with mTOR, mainly through hydrogen bonding and hydrophobic interactions (Supplementary Fig. S2F). The CETSA results indicated that after incubation with Sal, the thermal stability of mTOR was improved (Supplementary Fig. S3A). Meanwhile, DARTS data revealed that coincubation with Sal inhibits the degradation of mTOR by pronase (Supplementary Fig. S3B).

This further indicates that Sal induces ferroptosis primarily by targeting mTOR. In summary, these results revealed that Sal might improve TNBC by regulating the mTOR-related PI3K/AKT/mTOR pathway to induce ferroptosis.

Sal-induced ferroptosis by regulating mTOR-mediated GPX4 antioxidant defense in MDA-MB-231 cells and TNBC animal model

Functional enrichment results predicted that Sal-induced ferroptosis was primarily associated with lipid peroxidation. As known, GPX4 can prevent ferroptosis by removing lipid peroxides, while PTGS2 has a positive effect on regulating ferroptosis. Surprisingly, mTOR could suppress ferroptosis by promoting GPX4 protein synthesis [38]. The GPX4 and PTGS2 were measured to test the predicted results through RT-PCR and WB methods. The results discovered that Sal could remarkably reduce the GPX4 levels, while the PTGS2 was increased in the mRNA and protein levels (Fig. 6A-B). Consistently, the mTOR activator, MHY1485 significantly improved Sal-induced repression of GPX4 expression as well (Fig. 6C). To further investigate whether GPX4 was connected with Sal-regulated ferroptosis, GPX4 overexpression was obtained in MDA-MB-231 cells (Fig. 6D). Consistent with Fer-1 administration, GPX4-overexpressed MDA-MB-231 cell activity suppression was partially alleviated through Sal treatment (Fig. 6E, H). Commonly, MDA and PTGS2 levels significantly downregulated (Fig. 6G, I), while the GPX activity increased because of GPX4 overexpression (Fig. 6F). Consequently, the Sal-induced ferroptosis might be attributed to decreased GPX4. Besides, the data confirms the occurrence of ferroptosis in Sal-treated MDA-MB-231 cells, and the mechanism of Sal-triggered ferroptosis may be by regulating mTOR-mediated lipid peroxidation through the GPX4 signaling pathway.

Fig. 6.

Sal-induced ferroptosis in MDA-MB-231 cells and TNBC animal model via regulating GPX4-medidated lipid peroxidation. (A) RT-PCR measured the mRNA levels of GPX4 and PTGS2 in MDA-MB-231 cells treated with Sal (50 μM) for 12 h or 24 h (n = 3). (B) The protein levels of GPX4 and PTGS2 were measured in MDA-MB-231 cells treated with Sal (50 μM) for 24 h (n = 3). (C) The protein level of GPX4 was measured after cells were pretreated with MHY1485 (2 μM) for 1 h and then Sal (50 μM) for 24 h (n = 3). (D) The protein levels of GPX4 were measured in MDA-MB-231 cells transfected with plasmids expressing GPX4 (n = 3). (E) Cell viability was measured by CCK-8 assay (n = 6) in Sal (50 μM)-treated cells with GPX4 overexpression. The levels of GPX (F) measured by GPX assay kit with NADPH (n = 6), MDA (G) measured by lipid peroxidation (MDA) assay kit (n = 6), and cell viability/cytotoxicity (H) was measured by Calcein-AM/PI cell viability/cytotoxicity assay kit (n = 3), in Sal (50 μM)-treated cells with GPX4 overexpression. (I) The protein level of PTGS2 was measured in cells with GPX4 overexpression for Sal (50 μM) treatment (n = 3). (J) The animal mRNA levels of GPX4 and PTGS2 were measured with RT-PCR in the Control and the Sal group (n = 3). (K) The animal protein levels of GPX4 and PTGS2 were measured in the Control and the Sal group (n = 3). Scale bar, 50 μm. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

Moreover, the GPX4 signaling pathways in the processes of antioxidant defense were tested in the TNBC animal model. The result was consistent with the in-vitro experiment. The mRNA and protein levels of GPX4 were decreased. In contrast, the PTGS2 level was increased after Sal treatment (Fig. 6J-K), suggesting that Sal can therapeutically treat TNBC. The mechanism may be to induce ferroptosis by suppressing the mTOR-mediated GPX4 signaling pathway in lipid peroxidation.

Sal-induced ferroptosis in MDA-MB-231 cells and TNBC animal model by suppressing the mTOR-mediated SCD1 pathway in fatty acid synthesis

Our previously predictive analysis suggested that fatty acid synthesis is one of the most important pathways by which Sal improves TNBC. The PI3K/AKT/mTOR signaling can also inhibit ferroptosis by promoting SCD1 transcription-mediated lipogenesis [39]. Therefore, we investigated the SCD1-associated signaling pathway in fatty acid synthesis. The result was consistent with our observations. The mRNA and protein levels of sterol regulatory element binding protein 1 (SREBP1) were all substantially reduced by Sal treatment (Fig. 7A-B). Additionally, as SREBP1 influences genetic transcription connected with lipid synthesis, the mRNA of downstream associated genes was detected, containing acetyl-CoA carboxylase (ACACA), ATP citrate lyase (ACLY), fatty acid synthase (FASN), and stearoyl-CoA desaturase (SCD). Comparing, the SCD downregulated significantly (Fig. 7A). Meanwhile, the protein expression of SCD1 was considerably decreased when treated with Sal (Fig. 7B). Furthermore, the mTOR activator MHY1485 increased the Sal-induced protein expression of SREBP1 and SCD1 (Fig. 7C). SCD1 can modulate monounsaturated fatty acids (MUFAs) generation, connected with the transformation of saturated fatty acids (SFAs). Consequently, MDA-MB-231 cells were pre-treated with MUFA OA (18:1) as well as SFA SA (18:0). Finally, owing to OA rather than SA, MDA-MB-231 cells appeared to be resistant to Sal-induced ferroptosis to a greater extent (Fig. 7D). The inhibitory effects were also confirmed by the reduction of Sal-induced MDA levels (Fig. 7E). Overall, the findings revealed that Sal inhibited the synthesis of MUFA OA produced by SCD1, which was the primary cause of SCD1-induced ferroptosis.

Fig. 7.

Sal-induced ferroptosis by suppressing the SREBP1/SCD1 signaling pathway in fatty acid synthesis. (A) The mRNA levels of SREBF1, SCD, FASN, ACLY and ACACA were detected with RT-PCR in MDA-MB-231 cells treated with Sal (50 μM) for 12 h or 24 h (n = 3). (B) The protein levels of SREBP1, mSREBP1 and SCD1 were detected in MDA-MB-231 cells treated with Sal (50 μM) for 24 h (n = 3). (C) The protein levels of SREBP1, mSREBP1 and SCD1 were detected after cells were pretreated with MHY1485 (2 μM) for 1 h and then Sal (50 μM) for 24 h (n = 3). (D, E) Cells were treated with OA (100 μM) or SA (100 μM), with simultaneous treatments of Sal (50 μM) for 24 h. Cell viability (D) was detected with CCK-8 assay (n = 6), and the levels of MDA (E) were detected with lipid peroxidation (MDA) assay kit (n = 6). (F) The protein levels of SCD1 were measured in MDA-MB-231 cells transfected with plasmids expressing SCD1 (n = 3). (G-K) Cells were transfected with plasmids containing the insert SCD1 and then treated with Sal (50 μM) for 24 h. Cell viability (G) and LDH release (H) were detected (n = 6). The levels of MDA (I) were detected with a lipid peroxidation (MDA) assay kit (n = 6), and lipid ROS (J) were detected with a C11-BODIPY 581/591 probe (n = 3). GPX4 and PTGS2 (K) protein levels were also detected via WB (n = 3). (L) The animal mRNA levels of SREBF1 and SCD were measured with RT-PCR in the Control and the Sal group (n = 3). (M) The animal protein levels of SREBP1 and SCD1 were measured in the Control and the Sal group (n = 3). Scale bar, 50 μm. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

To demonstrate Sal-induced ferroptosis attributed to SCD1, overexpression plasmids insert SCD1 was constructed in MDA-MB-231 cells (Fig. 7F). We found that the overexpressed SCD1 could increase the Sal-induced cell viability (Fig. 7G-H). Meanwhile, MDA and lipid ROS levels decreased significantly upon SCD1 overexpression (Fig. 7I-J). Significantly, the Sal-induced protein expression of GPX4 and PTGS2 was also inhibited by overexpressed SCD1 (Fig. 7K). These findings combined revealed that Sal might facilitate ferroptosis in MDA-MB-231 cells by suppressing the mTOR-mediated SCD1 signaling pathway in fatty acid synthesis.

Furthermore, the SREBP1/SCD1 signaling pathways in fatty acid synthesis were verified in the TNBC animal model. The result was consistent with the in vitro experiment. The mRNA and protein levels of SREBP1 and SCD1 were decreased (Fig. 7L-M), suggesting Sal can therapeutically treat TNBC. The mechanism may induce ferroptosis by suppressing the mTOR-mediated SCD1 signaling pathway in fatty acid synthesis.

Sal promotes NCOA4-mediated ferritinophagy regulated by the mTOR in MDA-MB-231 cells and TNBC animal model

Based on the predicted functional enrichment results, Sal-induced ferroptosis to upgrade TNBC might be through regulating autophagy-related pathways. Importantly, mTOR also inhibits ferroptosis by suppressing ferritinophagy [40]. To further explore the association between autophagy-related pathways and ferroptosis in TNBC, we detected alterations in iron regulatory proteins at mRNA levels, such as transferrin receptor (TFRC), iron-responsive element binding protein 2 (IREB2), ferritin heavy chain (FTH), and NCOA4 after Sal treatment. It was observed that FTH mRNA downregulated and NCOA4 mRNA upregulated significantly, while others did not change (Fig. 8A), demonstrating that Sal-induced ferroptosis might be correlated with the ferritinophagy. Subsequently, we analyzed the changes in ferritinophagy-related protein levels, including FTH, NCOA4, P62 and LC3 II/I after Sal treatment. As depicted, NCOA4 and LC3 II/I levels were upregulated, while FTH and P62 were significantly downregulated in the Sal-treated MDA-MB-231 cells (Fig. 8B). Surprisingly, the mTOR activator MHY1485 significantly inhibited LC3 II/I levels increased by Sal and improved P62 and FTH protein levels decreased by Sal (Fig. 8C), suggesting that Sal-activated ferritinophagy may by suppressing the mTOR pathway.

Fig. 8.

Sal promotes NCOA4-mediated ferritinophagy and ferroptosis of MDA-MB-231 cells and TNBC animal model. (A) The mRNA levels of TFRC, IREB2, FTH and NCOA4 were measured with RT-PCR in MDA-MB-231 cells treated with Sal (50 μM) for 12 h or 24 h (n = 3). (B) The protein levels of FTH, NCOA4, P62, LC3 I and LC3 II were measured in MDA-MB-231 cells treated with Sal (50 μM) for 24 h (n = 3). (C) The protein levels of P62, LC3 I, LC3 II and FTH were measured after cells were pretreated with MHY1485 (2 μM) for 1 h and then Sal (50 μM) for 24 h (n = 3). (D-G) Cells were pretreated with 3-MA (1 mM), and then Sal (50 μM) for 24 h. Cell viability (D) and LDH release (E) were measured (n = 6). The levels of MDA (F) were measured with a lipid peroxidation (MDA) assay kit (n = 6). NCOA4 and FTH (G) protein levels were measured (n = 3). (H) The protein levels of NCOA4 were measured in MDA-MB-231 cells that transfected with siNCOA4#1 or siNCOA4#2 (n = 3). (I-M) Cells were transfected with siNCOA4#1, and then treated with Sal (50 μM). Cell viability (I), LDH release (J), and MDA levels (K) were measured (n = 6). The changes in Fe²⁺ (L) were detected with FerroOrange dye (n = 3). The protein levels of FTH and PTGS2 (M) were also detected (n = 3). (N) The animal mRNA levels of FTH and NCOA4 were measured with RT-PCR in the Control and the Sal group (n = 3). (O) The animal protein levels of FTH, NCOA4, P62, LC3 I and LC3 II were measured in the Control and the Sal group (n = 3). Scale bar, 50 μm. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

Furthermore, to investigate whether ferritinophagy is engaged in ferroptosis caused by Sal, the cells were pre-treated with 3-MA to suppress autophagosome assembly. The result revealed that 3-MA pre-treatment significantly improved cell viability and inhibited MDA expression induced by Sal (Fig. 8D–F). Additionally, the FTH protein levels were downregulated in the Sal group but upregulated in the Sal + 3-MA group (Fig. 8G). Next, siRNAs targeting NCOA4 were used to investigate whether NCOA4 mediates ferritinophagy triggered by Sal, and the transfection efficacy was confirmed by WB (Fig. 8H). Depending on this, siNCOA4#1 was chosen for further research. Knockdown of NCOA4 made MDA-MB-231 cells anti-ferroptosis (Fig. 8I-J), which was proven by the downregulated MDA and Fe²⁺ levels (Fig. 8K-L). Consistent with this, the knockdown of NCOA4 also inhibited PTGS2 protein levels while improving the FTH protein levels induced by Sal (Fig. 8M). All results demonstrated that autophagy-related passage in ferroptosis is NCOA4-mediated ferritinophagy, and Sal-induced iron overload may be attributed to NCOA4-mediated ferritinophagy regulated by the mTOR pathway.

Subsequently, an in vivo experiment confirmed the effect of NCOA4-mediated ferritinophagy and ferroptosis in Sal-treated TNBC. The levels of LC3 II/I and NCOA4 were increased, while P62 and FTH were decreased significantly in the Sal group compared with the control group (Fig. 8N-O). This was consistent with the in vitro results, indicating that Sal might promote ferroptosis occurrence through mTOR-mediated NCOA4-regulated ferritinophagy.

Discussion

TNBC is a highly aggressive form of breast cancer with few available treatment choices, accounting for 5 % of yearly cancer-related fatalities [41], requiring effective therapeutic interventions [42]. Consequently, there is an urgent need to identify TNBC biomarkers and confirm whether interventions on these biomarkers can achieve therapeutic efficacy. Sal has been widely proven to be non-toxic or with fewer side effects in numerous animal experiments and clinical trials. In addition, it is considered a safe, natural ingredient [43]. Studies have revealed that Sal has anti-migration and invasion effects on many types of malignant tumors, including melanoma, gastric cancer, liver cancer, and TNBC [17], [44], [45], [46]. Notably, Sal was used without clinical adverse events in treating epirubicin-induced early left ventricular regional systolic dysfunction in patients with breast cancer [47]. Consequently, we speculate that Sal's clinical safety and efficacy conferred natural advantages in tumor treatment, making it possible to develop it as a clinical therapeutic agent for TNBC.

Previous studies proposed that toosendanin and isotoosendanin could inhibit TNBC growth by inducing necrosis, apoptosis and autophagy [48]. Besides, glycyrrhetinic acid could induces ferroptosis in MDA-MB-231 cells to suppress TNBC [49]. All of these studies demonstrated the pathogenesis of TNBC was tightly associated with cell death, including necrosis, apoptosis, autophagy, and ferroptosis. This further elucidated that programmed cell death could be a new target for anticancer drug development. Unfortunately, there are still a lack of studies to determine which form of cell death dominates in TNBC. Notably, there is currently only one study on Sal in relation to TNBC to date. It was demonstrated only at the cellular level that Sal could inhibit tumor cell proliferation and migration via regulated apoptosis pathways [17]. Considering that ferroptosis is a newly identified form of programmed cell death and that studies have elucidated that TNBC is sensitive to ferroptosis [50]. Meanwhile, based on the research gap of how Sal regulates ferroptosis against TNBC, our study takes ferroptosis as a clue to explore the potential molecular mechanism of Sal-induced ferroptosis against TNBC by multiple dimensions.

Our study presents several novel findings. First, we demonstrated that Sal can exert anti-TNBC effects, potentially inducing ferroptosis. Besides, in TNBC treatment, Sal mainly modulates lipid metabolism, antioxidant defense, fatty acid synthesis, and ferritinophagy. It also participates in lipid and atherosclerosis, mTOR, PI3K/AKT, and autophagy signaling pathways. Moreover, it was proven that seven biomarkers, including HIF1A, mTOR, SIRT1, HMOX1, GSK3B, MAPK3, and XBP1, have diagnostic value in TNBC and are critical regulators of Sal-induced ferroptosis, delaying TNBC progression. Hence, the biomarker mTOR and related pathways (mTOR and PI3K/AKT signaling pathways) were selected to be further investigated by careful consideration.

PI3K/AKT is among the essential pathways for regulating signal transduction in various BP, closely associated with tumor development. In the TNBC, the PI3K/AKT pathway is hyperactive due to loss-of-function alterations of tumor suppressor PTEN [51]. We found that Sal could activate PTEN and inhibit PI3K/AKT in TNBC, which could be reversed by PTEN inhibitor VO-Ohpic trihydrate and AKT activator SC79. The oncogenic activation of the PI3K/AKT pathway enhances the downstream effector mTOR, concerning tumor growth, and mainly inhibits apoptosis and autophagy in cancer cells [52]. The pharmacological inhibition of mTOR has been validated to efficiently suppress TNBC tumorigenesis [53]. The mTORC1, one complex of mTOR, negatively mediates ferroptosis owing to iron accumulation [54]. Consistent with this, we found that Sal inhibited mTOR in TNBC and downregulated the mTORC1 substrate 4E-BP1. Furthermore, the mTOR agonist MHY1485 could significantly reduce Sal-induced ferroptosis in TNBC, whereas rapamycin significantly enhanced, implicating the critical role of the mTOR pathway. The above experimental data again supported the predictive findings that Sal-induced ferroptosis anchored TNBC through the inhibited PI3K/AKT/mTOR pathway.

Studies have found that GPX4 protein synthesis is closely linked to mTORC1-4EBP signaling axis. Oncogenic activation of mTORC1 signaling suppresses ferroptosis by promoting GPX4 synthesis [55], [56]. As a selenoprotein type, GPX4 is the primary enzyme that catalyzes the reduction of phospholipid hydroperoxides in TNBC cells, thereby reducing toxicity. GPX4 senses and translates oxidative stress into the cell-death pathway associated with PTGS2 and 12/15-lipoxygenase [57]. PTGS2 upregulation is simply a downstream marker of ferroptosis. Cells with downregulated GPX4 expression are more vulnerable to ferroptosis [58]. It has been reported that GPX4 ubiquitination to reduce the abundance of GPX4 can lead to ferroptosis in TNBC [59]. Post-transcriptionally, guanine sequence binding factor 1 binds to the 5′ untranslated region of the mitochondrial form of the GPX4 mRNA, leading to an increase in the translation of the mitochondrial GPX4 [60]. This is in line with our work, and we found that Sal-induced suppression of mTOR further inhibits GPX4 antioxidant defense in TNBC, which could be improved by mTOR agonist and GPX4 overexpression, further indicating that mTOR mediates cystine-induced GPX4 synthesis. Accordingly, the PI3K/AKT/mTOR pathway may regulate the antioxidant defense function of GPX4, and its inactivation allows lipid peroxidation accumulation and Sal-induced ferroptosis.

Hyperactive mutation of the PI3K/AKT/mTOR pathway must confer ferroptosis resistance by upregulating SREBP1/SCD1-mediated lipogenesis [39]. Lipogenic regulator-transcription factor SREBPs have been reported to be essential transcription factors that regulate most lipid metabolism genes [61]. Among these, lipogenic regulator-transcription factor SREBP1 was the downstream protein of the mTORC1 signal pathway [62], and it is closely associated with cancer progression and poor outcomes. Gabitova-Cornell et al. [63] have reported that disruption of cholesterol biosynthesis by statins treatment or NAD(P)-dependent steroid dehydrogenase-like (NSDHL) loss activates SREBP1, which induces Tgfb1 expression, enabling epithelial-mesenchymal transition and basal differentiation in pancreatic cancer. Liu et al. [64] have revealed that regulatory T (Treg) cells break immune homeostasis and contribute to tumor evasion by promoting SREBP1-mediated lipid metabolism and repressing CD8 + T cell-derived secretion of interferon-γ (IFN-γ) in M2-like tumor-associated macrophages. In this study, we revealed that Sal could inhibit SREBP1 and its mature form (mSREBP1) in TNBC for the first time. The mSREBP1 can translocate into the nucleus and activate genes connected with lipid homeostasis. This transcription factor controls fatty acid biosynthesis by regulating SCD1 expression [39]. SCD1 in the lipogenesis pathway is required for the body to produce MUFA from SFAs [65]. It has been demonstrated that SCD1 could protect cancers from ferroptosis. SCD1 was upregulated in recurrent human breast cancer samples and correlated with a poor prognosis [66]. Magtanong et al. [67] also reported that exogenous MUFA could inhibit ferroptosis in an acyl-CoA synthetase long-chain family member 3-dependent manner. Our research supports the above idea and confirms that Sal inhibits SCD1, which could be significantly improved by SCD1 overexpression and OA. It suggested that SCD1 and MUFA could inhibit ferroptosis in TNBC and were correlated with a poor prognosis. Additionally, we found that mTOR agonists could reverse SREBP1/SCD1-mediated lipogenesis of MUFA in Sal-induced ferroptosis. Consequently, we assume that the PI3K/AKT/mTOR pathway may participate in SREBP1/SCD1-mediated lipogenesis of MUFA, and its inhibition confers TNBC sensitive to Sal-induced ferroptosis, which is correlated with a pleasant treatment effect.

Furthermore, autophagy signaling pathways have also been found to be associated with TNBC. Further exploration is required to confirm any definitive association with either autophagy or ferroptosis. Ferritin—a widespread iron storage protein—is highly expressed in breast cancer. Accordingly, ferritin may have considerable reference value in the clinical diagnosis of patients with breast cancer and in monitoring their response to therapy [68]. In this study, we found that the FTH expression was increased in mRNA and protein levels. In contrast, the Sal treatment reversed the trend, indicating that Sal could somewhat improve TNBC. NCOA4 can selectively recognize and bind to FTH to form a complex [69], enter the autophagosome, finally degrade ferritin, and release free iron [70]. Under normal physiological conditions, ferritinophagy is tightly regulated by an iron-dependent protein network to maintain the intracellular Fe²⁺ balance and related physiological functions. However, excessive ferritinophagy can lead to ferroptosis [69]. Our study suggested that Sal could promote NCOA4-mediated FTH degradation in TNBC cells, and NCOA4 knockdown or autophagic inhibitors could significantly suppress Sal-induced ferroptosis. This was consistent with previous studies that inhibiting NCOA4-mediated ferritinophagy increases iron storage and limits ferroptosis in cancer cells [71], [72]. Therefore, it is reasonable to speculate that the association between autophagy-related pathways and ferroptosis in TNBC may be connected with ferritinophagy mediated by NCOA4.

Aside from this, the active mTORC1 inhibits autophagy by binding to both ULK1 and Vps34 complexes, inhibiting autophagosome and lysosome biogenesis by inhibiting transcription factor EB (TFEB) [73], [74]. Furthermore, the thiazide-sensitive cotransporter (TSC) protein (the complex formed by TSC2 and TSC1), as a critical regulator, can inhibit the activity of mTORC1 by converting ras homolog enriched in brain (RHEB) from activated form to inactivated form. Activated PI3K/AKT phosphorylates TSC2 and prevents the formation of the TSC complex, further leading to the activation of mTORC1 inhibiting autophagy [74], [75]. These studies reveal that the PI3K/AKT/mTOR is possible upstream of NCOA4-mediated ferritinophagy. This observation was further confirmed in the present study that Sal could induce autophagic death by inhibiting mTOR in TNBC, as proved by the elevation of the LC3 II/LC3 I ratio and the decrease in P62. Importantly, we found that the mTOR agonists could significantly increase FTH expression and decrease free iron induced by Sal. It is suggested that Sal may activate NCOA4-mediated ferritinophagy by inhibiting mTOR.

Although both predictive and experimental evidence demonstrated that Sal could target the mTOR-related pathway to facilitate ferroptosis and arrest TNBC progression, the current study still has some limitations. For instance, seven targets with diagnostic value were screened in this study, but only mTOR was further validated by considering the comprehensive role of Sal. Besides, as our study was primarily designed to investigate how Sal regulates ferroptosis to exert anti-TNBC effects, we use ferroptosis inhibitors rather than TNBC-associated positive drugs for comparison with Sal in the animal study. Additionally, the research about the therapeutic role of Sal on TNBC remains in the pre-clinical stage. Given this, our following studies will focus on the roles of the remaining targets, excluding mTOR. Meanwhile, the clinical efficacy of Sal in treating TNBC will be explored to bring the study closer to the clinic. In conclusion, this study lays the theoretical foundation for clinically treating TNBC with Sal.

Conclusion

Our research indicated that Sal sensitized TNBC to ferroptosis by suppressing the PI3K/AKT/mTOR axis, thereby suppressing SCD1-mediated lipogenesis of MUFA and GPX4-mediated antioxidant defense. Additionally, Sal facilitates NCOA4-mediated ferritinophagy to increase intracellular Fe²⁺ content. These findings comprehensively explore the prospective therapeutic targets and related mechanisms connected with Sal against TNBC and provide a theoretical basis for applying Sal in the clinical treatment of TNBC.

Compliance with ethics requirements

All experimental procedures involving animals in our studies were approved by the Animal Ethics Committee of China Pharmaceutical University (Permit number SYXK-2021–0010) as well as strictly in accordance with the Guide for Care and Use of Laboratory Animals (8th Edition, 2011). Extensive efforts have been made to minimize suffering of animals during experiments.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Fig. 1 and Fig. 3N are drawn using Figdraw (https://www.figdraw.com).

Appendix A. Supplementary data

The following are the Supplementary data to this article: