Glutamine synthetase shields triple-negative breast cancer cells from ferroptosis in metastasis triggered by glutamine deprivation

Abstract

Background

Epithelial-mesenchymal transition (EMT) in cancer cell metastasis involves complicated metabolic plasticity to survive the highly challenging environment, such as oxidative stress, after subsequent circulation in the bloodstream. Glutamine synthetase (GS) is an enzyme that converts glutamate and ammonia to glutamine (Gln) during Gln deprivation stress. This study revealed for the first time that GS plays an important role in protecting triple-negative breast cancer (TNBC) cells from ferroptosis during Gln deprivation-induced EMT, namely ferroptosis-resistant EMT (FR-EMT).

Methods

To better understand this finding, we focused on the mechanism of GS-mediated FR-EMT in TNBC through transcriptomic analysis and murine metastasis modeling.

Results

This study specifically investigated the effects of GS on lipid peroxidation and iron metabolism, the two major metabolic disorders in ferroptosis. An abnormal increase in monounsaturated fatty acids (MUFAs) mediated by mechanistic target of rapamycin complex 1 (mTORC1) decreased the ferroptosis sensitivity under Gln deprivation. Additionally, aberrant iron metabolism via lipocalin 2 (LCN2) and transferrin receptor (TFRC) affected the sensitivity to ferroptosis. Moreover, this study confirmed that GS protects TNBC cells from ferroptosis and increases their ability to survive during subsequent metastasis through the blood in the lung metastasis mouse model.

Conclusion

This investigation provides insights into the role of ferroptosis in metastasis and demonstrates that GS may be a viable target for preventing metastases in TNBC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13058-025-02115-5.

Background

Triple-negative breast cancer (TNBC), a highly aggressive subtype of breast cancer, has limited treatment options due to the lack of estrogen receptor (ER), progesterone (PR) and human epidermal growth factor receptor 2 (HER2) expression. Thus, conventional chemotherapy is the only treatment option for TNBC, and TNBC metastasis poses a significant challenge in the clinical management of cancer [1]. Alterations in glutamine (Gln) metabolism caused by oncogenic transformation increase the dependency of transformed cells on Gln [2–4]. Various cancer subtypes originating from a particular organ exhibit discernible patterns of Gln metabolism [5]. The requirement for exogenous Gln for TNBC cell survival and proliferation has been confirmed in numerous studies [6, 7]. Nonetheless, Gln is the most abundant amino acid in the serum and is frequently severely depleted in developing cancers [8]. Thus, more research has been conducted to better understand the metabolic plasticity that supports metastasis under Gln deficiency.

Certain cancer cells can synthesize Gln de novo via glutamine synthetase (GS) activity. A gene called glutamate-ammonia ligase (GLUL) encodes GS, an enzyme that converts glutamate and ammonia into Gln, enabling tumor cells to survive under Gln deprivation stress [9]. GS sensitizes cancer cells to autophagy [10] and independently enhances their susceptibility to cytotoxicity mediated by B lymphocytes [11]. The basal breast cancer cells, on the other hand, exhibit high Gln deprivation sensitivity, but not GS expression. This observation has also been demonstrated for melanoma [12]. Importantly, GS plays a crucial role in regulating cellular redox homeostasis by sustaining glutamate levels necessary for the synthesis of glutathione (GSH), which serves as the primary antioxidant defense against lipid peroxidation.

GS gene expression has been linked to epithelial-mesenchymal transition (EMT) and poor prognosis in breast cancer patients [13–15]. The upregulation of GS in certain cancers provides a metabolic advantage in coping with Gln deprivation, which has also been observed at the cellular level and in vivo xenograft tumor rat models [16, 17]. This adaptation presents a paradox: although GS enhances survival under conditions of nutrient stress, it concurrently depletes glutamate reserves essential for GSH synthesis. Consequently, GS gene expression in different cancer cells exhibits considerable variation due to various mechanisms, even within a single tumor [5]. As a result, the GS gene’s inherent ability to respond to Gln deprivation conditions may provide insight into the importance of GS gene expression in cancer cells.

A buildup of harmful lipid peroxidation that is triggered by iron is known as ferroptosis [18]. Specifically, disturbances of metabolism in lipid peroxidation and iron metabolism serve as the principal catalysts for ferroptosis. Recent investigations reveal that therapy-resistant or persistent cancer cells exhibiting a mesenchymal/metastatic phenotype demonstrate heightened susceptibility to ferroptosis-inducing agents [19, 20]. Notably, cells exhibiting mesenchymal phenotypes, which are prevalent in GS-high, therapy-resistant cancers, demonstrate GSH depletion and impaired glutathione peroxidase 4 (GPX4) activity, making them particularly susceptible to ferroptosis. Consequently, we propose the hypothesis that GS modulates redox metabolism, thereby affecting ferroptosis susceptibility in TNBC.

Epithelial cells undergo EMT when junctions and polarity are lost, facilitating morphological changes and enhanced mobility. This crucial phenomenon allows epithelial cells to detach from their intercellular and extracellular matrix adhesion, thereby promoting the invasion of neighboring tissues, disruption of functions, and metastases to distant sites. Additionally, EMT confers chemoresistance to cancer cells, thereby diminishing the efficacy of anti-cancer medications [21]. E-cadherin (E-cad) has been found to activate the intracellular Merlin and Hippo signaling pathway, leading to the suppression of ferroptosis [22]. Furthermore, Gao et al. [23] suggested that the serum components Gln and transferrin play critical roles in ferroptosis regulation in the extracellular environment. Ferroptosis requires the metabolic pathway glutaminolysis and the components involved in intracellular transferrin import [24]. Consequently, this study aimed to investigate the functional connection between GS and ferroptosis sensitivity in TNBC of mesenchymal subtypes.

This study showed that Gln deprivation protects against ferroptosis and induces EMT in TNBC cells. Previous studies show that mesenchymal-state cells are highly sensitive to ferroptosis but resistant to non-ferroptotic oxidative stressors, enhancing their metastatic potential in various cancers [25–27]. Consequently, GS was investigated in this study in the context of lipid peroxidation and iron metabolism to elucidate its role in ferroptosis. Hence, we hypothesized that GS, independent of its catalytic activity, confers resistance to ferroptosis in cells and empowers cells to endure the oxidative stress encountered during metastatic progression. Additionally, we speculated that specifically targeting GS in metastatic TNBC can effectively impede metastatic processes, thereby enhancing the survival rate of individuals with cancer.

Methods

Study participants

All breast cancer tissue specimens were collected at the Department of Pathology, Air Force Medical Center, with informed consent from all patients according to protocols approved by the institutional ethics committee of Air Force Medical Center.

Cell culture

The MDA-MB-231, 4T-1, HCC1937, BT474 and MCF-7 were procured from the National Infrastructure of Cell Line Resource of China. The cells were cultured in RPMI 1640 medium (BI, Shanghai, China) supplemented with 10% fetal bovine serum (FBS, BI, Shanghai, China). All cell lines were maintained at 37 °C in a humidified atmosphere with 5% CO₂.

Gln deprivation and measurements

During the Gln deprivation experiments, 1640 medium without Gln supplementation was used as the complete medium, as previously described (Gibco, Shanghai, China). The medium was replenished daily during periods of nutrient starvation. According to the manufacturer’s instructions, Gln was quantified using a Glutamine Colorimetric Assay Kit provided by Elabscience.

Small interference RNA transfection (siRNA)

The siRNA used to target human GLUL, ACSL4, VEGFA, and lipocalin 2 (LCN2), as well as the non-targeting siRNA control, were acquired from Sangon Biotech (China). siRNA transfections were performed on 30–50% confluent cells with 30 nM siRNA duplexes using Lipofectamine RNAiMAX (Invitrogen) for 6–8 h before other treatment. The siRNA target sequences are listed in Tables S1 and 2.

Cell death and viability, western blotting (WB), immunofluorescence (IF), immunohistochemistry (IHC), quantitative real-time PCR (qRT-PCR), and transwell cell migration assays

Cell death and viability [28], WB [28], IHC [29], IF [30], qRT-PCR [28] and transwell cell migration [30] were performed according to previous protocols. Table S3 presents a detailed list of primary antibodies, and Table S4 summarizes the primers used in this study.

Reactive oxygen species (ROS) and GSH measurement

The C11-BODIPY (581/591) probe (D3861, Invitrogen, USA) was used to measure lipid ROS [28]. Intracellular GSH/Oxidized glutathione (GSSG) levels were assessed according to the manufacturer’s instructions [28].

Plasmids and cell transfection

The GLUL overexpression plasmid pcDNA3.1-3 × Flag-GLUL, along with its control plasmid pcDNA3.1-vector, were acquired from Addgene. A single colony was selected and cultured in a 10 mL liquid Luria Bertani (LB) medium supplemented with ampicillin (B540112-0100, Sango Biotech). The culture was incubated at 37 °C and 220 rpm overnight and then centrifuged at 8000 x g for 10 min. The supernatant was discarded, and plasmid extraction was performed using the Endo-Free Plasmid Midi Kit (CW21055, CWBIO) according to the manufacturer’s instructions. The plasmid DNA was quantified using Nanodrop One (Thermo Fisher, U.S.A) and stored at − 20 °C for future use. The transient transfection of MDA-MB-231 cells was performed using Lipofectamine 3000 (L3000008, Invitrogen) in accordance with the manufacturer’s guidelines.

Breast cancer patient cohort analysis based on public datasets

To investigate the correlation between GS expression and the mTORC1 signaling pathway, Gene Expression Omnibus (GEO) data (https://www.ncbi.nlm.nih.gov/geo/, accession number GSE103091) was analyzed using Gene Set Enrichment Analysis (GSEA, version 2.2.4). The GSEA analysis incorporated three primary statistical measures: the false discovery rate (FDR), normalized enrichment score (NES), and nominal P-value. Additionally, UALCAN (http://ualcan.path.uab.edu) was used for the clinicopathological analysis of LCN2 and TFRC based on The Cancer Genome Atlas (TCGA) data. To assess the relationship between the presence of target protein and overall survival in patients with BC, the Kaplan-Meier Plotter online tool was used (http://www.kmplot.com). GS expression in breast cancer cell lines was assembled by CCLE (http://www.broadinstitute.org/ccle).

Transcriptomics

RNA was extracted using TRIzol reagent (15,596,026, Invitrogen, USA). Genomic DNA was removed using DNase I (#2270A, TaKara Bio, Shiga, Japan). The RNA samples used for constructing a sequencing library met the following criteria: OD260/280 = 1.8–2.2, OD260/230 ≥ 2.0, RIN ≥ 6.5, 28 S:18 S ≥ 1.0, >1 µg. With a SuperScript double-stranded cDNA synthesis kit (#11917001, Invitrogen) and random hexamer primers (#20022654, Illumina), messenger RNA was isolated using poly(A) selection and oligo (dT) beads. Sequencing the paired-end RNA-sequencing library was conducted on an Illumina HiSeq xten/NovaSeq 6000 sequencer (2 × 150 bp read). Default parameters were used for Sickle and SeqPrep with paired-end reads after trimming and quality control. Each sample’s mapped reads were assembled using StringTie using a reference-based approach using HISAT2 to align them to the reference genome in orientation mode. Statistical significance was determined by considering log2 (fold change) > 1.

Proteomics analysis

Samples were lysed in 8 M urea and centrifuged at 14,100×g for 20 min. Protein concentrations were determined using the Bradford method. Proteins were reduced with 200 mM DTT at 37 °C for 1 h. Proteins were reduced with DTT, alkylated with iodoacetamide, and excess alkylating agent was neutralized with DTT. After dilution, proteins were digested with trypsin overnight. Digested peptides were desalted using C18 columns, eluted with 70% ACN, and lyophilized. Fractionation was performed using HPLC with a gradient elution program, resulting in 6 fractions that were vacuum-dried and reconstituted in 0.1% (formic acid) FA. Peptides were separated using a 25 cm ReproSil-Pur C18-AQ column with a gradient elution and analyzed using a timsTOF_HT mass spectrometer in DDA and DIA modes. Spectronaut software was utilized to generate a hybrid library for protein identification and quantitation against the human UniProt reference proteome. DEPs were identified with a fold change > 1.2 and p-value < 0.05.

Animal studies

Female BALB/c mice (4–6 weeks old) were obtained from the Department of Animal Science of Peking University Health Science Center for xenograft experiments. To select a highly metastatic lung TNBC cell line, 5 × 105 GFP-tagged 4T-1 cells were injected into the tail veins of mice from the Gln+, Gln-, and GS siRNA + Gln- groups. To investigate the in vivo impact of Artemisinin (ART) on modulating metastatic lung colonization of GS, mice in the Gln- group were subjected to drug treatment for three days starting on day two post-injection. Mice injected with Gln-deficient cells were randomly assigned to three groups and intraperitoneally administered either artesunate (at a dosage of 400 mg/kg of mouse weight) or an equivalent volume of PBS via gavage daily. Mouse lungs were collected and imaged, and the number of lung metastatic nodules was quantified to compare the metastatic ability among the different groups.

In cell line-derived xenograft experiments, 2 × 10⁶ 4T-1 cells in 100 µL PBS were injected into one flank of each mouse for two weeks to measure Gln. Tumor tissues from these mouse xenografts served as primary cell cultures.

Statistical analyses

Each experimental group was replicated thrice for each experiment. GraphPad Prism 7.0 GraphPad Software was used for all statistical analyses. Statistical significance was determined using a two-sided Student’s t-test, an ANOVA followed by Tukey’s post hoc test, or a two-way ANOVA followed by a Bonferroni multiple comparison test was considered statistically significant.

Results

Gln deprivation induces ferroptosis resistance in TNBC cells

BT474 cells indicate triple-positive breast cancer (TPBC), MCF-7 cells represent human ER-positive breast cancer, whereas MDA-MB-231, HCC1937, and 4T-1 cells represent TNBC. We analyzed GLUL mRNA expression in the Cancer Cell Line Encyclopedia (CCLE) [http://www.broadinstitute.org/ccle] database and observed that GLUL expression was progressively higher in breast cancer cell lines BT474, MCF-7, MDA-MB-231, and HCC1937, with no significant difference between MCF-7 and MDA-MB-231 (Fig. 1A). These cell lines have been extensively investigated, and a significant disparity in their sensitivity to ferroptosis has been demonstrated. TNBC cells (MDA-MB-231 and 4T-1) showed significantly increased sensitivity to (1 S, 3R)-RSL3 compared to ER-positive MCF-7 cells (Fig. 1B).

Fig. 1.

GS protects from ferroptosis sensitivity in TNBC cells. (A) GS (GLUL) expression in breast cancer cell lines obtained from CCLE database. (B) The cytotoxic effects of RSL3 on MCF-7, 4T-1, and MDA-MB-231 cells were assessed using the CCK-8 assay after 16 h of treatment, with the cytotoxicity being dose-dependent. (C) Gln levels in the plasma and TIF of mice with 4T-1 tumors on day 14 (n = 3 per group). (D) The cytotoxic effects of Erastin on Xenograft-derived 4T-1 (XD-4T-1) cells and cell lines (4T-1) were analyzed under both normal conditions (Gln+) and Gln-. (E) The dose-dependent cytotoxicity of Erastin and RSL3 was evaluated in HCC1973 and MDA-MB-231 cells under Gln + and Gln- for 36 h. (F) Analysis of GS protein expression through proteomics in Gln + and Gln- using proteomic analyses. (G) Expression of GS was analyzed in MDA-MB-231, 4T-1, and MCF-7 cell lines after 36 h of exposure to either Gln + or Gln- medium. (H) The cytotoxicity of Erastin was assessed for 24 h in Gln-starved cells transfected with GS siRNA for 24 h, and the results were compared with those obtained from normal MDA-MB-231 cells using the CCK-8 assay. (I) In TNBC patients with metastasis and ER-positive patients, GS is detected using immunohistochemistry (brown)

TNBC has a higher demand for Gln and is more sensitive to its depletion, leading to poor prognosis. We analyzed the nutrient composition in the TME by isolating tumour interstitial fluid (TIF) and corresponding plasma from mice injected with 4T-1 cells. Gln levels were lower in TIF than in the matched plasma (Fig. 1C). Specifically, primary cultured mouse cells, like 4T-1 cells, showed resistance to ferroptosis when deprived of Gln (Fig. 1D). TNBC cells cultured under short-term Gln starvation (< 3 days) exhibited resistance to ferroptosis induced by conventional ferroptosis inducers, such as RSL3 or Erastin (Fig. 1E). Subsequently, the protein expression of GS was assessed in breast cancer cell lines subjected to Gln deprivation. As expected, the analysis of proteomic data revealed a notable change in GS expression in 4T-1 cells when deprived of Gln (Fig. 1F). GS expression was increased in the TNBC cell lines 4T-1 and MDA-MB-231 after short-term Gln starvation (Figs. 1G, S1A). Notably, no significant differences were observed in GS expression between Gln deprivation and non-deprivation conditions in MCF-7 cells (Fig. 1G). No significant difference in GS protein expression was detected in MCF-7 and MDA-MB-231 cell lines. This result was consistent with that of CCLE validation (Fig. 1A).

To confirm the specific role of GS, transient transfection was performed to deplete GS expression in human TNBC cell lines. The expression of the GS protein was assessed via WB in Gln-36-h-starved cells (Figs. S1B-C). The results demonstrated that GS siRNA knockout restored sensitivity to ferroptosis induced by Gln deprivation (Figs. 1H, S1D). These findings indicated that GS, as a potent factor, is essential in conferring resistance to ferroptosis under Gln deprivation conditions. In contrast to TNBC cells, epithelial cell lines such as MCF-7 and BT474 cells exhibited no changes in sensitivity to Erastin or RSL3 under Gln deprivation (Figs. S1E-G). Consistently, MCF-7 cells did not exhibit increased Snail expression in response to Gln deprivation (Fig. S1H). In TNBC patients, the expression of GS is positively linked to the risk of metastasis, unlike in ER-positive patients, indicating that GS might play a role in TNBC metastasis (Fig. 1I).

GS regulates the metastasis of TNBC cells

Initially, this study revealed that Gln depletion augmented the EMT mechanism in TNBC cells (Fig. 2A). Furthermore, knocking down GS expression impeded the EMT-like alterations observed during Gln deprivation (Fig. 2A). Subsequently, an in vivo metastasis assay was conducted. The growth rate of body weight was notably decelerated in the Gln-deprived group, which was reversed by GS knockdown (Fig. 2B). Metastatic nodules in the lungs of BALB/c mice after intravenous injection of GFP-labelled 4T-1 cells were visually monitored using GFP-based fluorescence imaging. As revealed in Fig. 2C injecting 4T-1 cells with Gln depletion resulted in a higher GFP-fluorescence density in mouse lungs than injecting control cells. Interestingly, GS siRNA treatment resulted in a significant decrease in the fluorescence density of metastatic tumors. This observation was further confirmed by the size of these nodules in the photomicrographs of hematoxylin and eosin (HE)-stained specimens (Fig. 2D) and the number of metastatic lung cancer nodules (Fig. 2E). Significantly elevated GS and Snail expression was observed in breast cancer lung metastasis subjected to Gln deprivation compared with the control and GS knockdown groups (Fig. 2F). Based on these findings, we hypothesized that GS may play a pivotal role in determining the propensity for poor distant metastasis or ferroptosis resistance.

Fig. 2.

The inhibition of GS effectively mitigates tumor metastasis and enhances prognosis. (A) Immunoblotting was performed to evaluate the protein expression of E-cad and Snail in MDA-MB-231 cells transfected with GS siRNA and cultured in Gln- medium for 36 h compared with the control cells. (B) The body weight of each mouse was monitored every other day. (C) Live imaging was conducted on 4T-1 cells expressing GFP, with a sample size of 5 mice in each of the scrambled control (Gln+), Gln-, and Gln- medium with GS knockdown groups. (D) HE staining was used to analyze sections of lung tissues from each group of mice, and tumor size was measured using ImageJ software. (E) Measurement of metastatic lung tumor nodules was performed using a fluorescence-based imaging system. The number of metastasis nodules was recorded manually. (F) Representative IF of three groups of mouse tissues stained with anti-GS, anti-Snail, and DAPI for nuclei stain. (G) Kaplan-Meier plots were generated to assess the OS of patients with breast cancer, ER + breast cancer, and TNBC, with hazard ratio (HR) calculated. **P < 0.01, ***P < 0.001, ****P < 0.0001

Subsequently, the Kaplan-Meier meta-analysis was conducted using the Kaplan-Meier plotter for breast cancer survival analysis database (http://www.kmplot.com). OS analysis revealed that high GS expression (represented by the red line) was associated with an increased risk of adverse outcomes in patients with TNBC (Fig. 2G). Interestingly, high GS expression in TNBC tissues predicted an unfavorable survival outcome compared with low GS expression. However, low GS expression (represented by the black line) was associated with an increased risk of adverse outcomes in patients with all subtypes of breast cancer. These findings indicated that elevated GS expression exclusively influences the progression of TNBC, as opposed to exerting an impact on other breast cancer subtypes. Collectively, these findings provided compelling evidence for the highly specific targeting of GS in inhibiting metastasis development in TNBC cells.

The non-catalytic involvement of GS is necessary for the induction of FR-EMT

GS expression and activity are upregulated in response to Gln deprivation. The GS enzyme converts glutamate to Gln [9]. Both GS inhibitors, MSO treatment and GS knockdown, resulted in a significant decrease in the catalytic activity of GS (Fig. S2A). However, the protein expression of GS was significantly upregulated following treatment with MSO, an inhibitor that impedes glutamine synthesis within cells and inhibits GS catalytic activity. This effect was observed at MSO concentrations ranging from 1 to 4 mM (Figs. S2B-C). Notably, MSO exhibited no toxicity up to a concentration of 5 mM (Fig. S2B). The accumulation of ammonia is known to trigger a compensatory mechanism. Upon the inhibition of GS activity by MSO, intracellular NH₃ levels rise due to the impaired conversion of ammonia to glutamine by GS. As a toxic metabolite, ammonia activates transcription factors such as c-Myc and NF-κB, which in turn directly enhance the transcription of the GS gene. A critical function of c-Myc is its ability to bind to the promoter region of the GS gene, thereby promoting its expression [31]. Under conditions of glutamine deficiency, c-Myc upregulates GS expression to maintain metabolic homeostasis by responding to adenosine levels [31].

Previous research has indicated that GS activity is a key metabolic feature associated with the pro-metastatic, immunosuppressive, and proangiogenic functions of M2-like macrophages [32]. However, the specific mechanism through which GS regulates metastasis, irrespective of its catalytic activity, has not been thoroughly explored. We hypothesized that GS protein accumulation may have a noticeable effect on cell progression. A study was conducted to investigate this hypothesis by inducing GS protein overexpression using MSO. EMT was significantly enhanced in the MSO-treated TNBC cells (Fig. S2D) and exhibited resistance to the classic ferroptosis inducer RSL3 (Figs. S2E-F). However, no significant differences were observed in MCF-7 cells treated with MSO (Figs. S2G-H). Collectively, these findings suggested that GS, in a non-catalytic manner, induces EMT-like changes and promotes resistance to classic ferroptosis inducers in TNBC cells.

GS activates LCN2 and restrains TFRC to lower the iron level

To investigate the impact of the Gln starvation period on the sensitivity to ferroptosis and the occurrence of EMT-like changes, this study analyzed both short-term (Days 2 to 3) and long-term (30% Gln, Days 10 to 30) starvation. The findings indicated that resistance to ferroptosis and activation of the EMT program, which are mediated by long-term starvation, were significantly reduced compared with those induced by short-term starvation (Figs. S3A-D). These results suggested that GS primarily affects the early stages of Gln shortage.

Transcriptomic analysis was conducted to elucidate the mechanism by which GS induced ferroptosis resistance. A filter was applied to identify gene clusters that exhibited differential expression in short-term starvation than that in long-term starvation in the presence of Gln and GS knockdown conditions. The analysis identified 23 genes as primary targets of GS transcriptional regulation (Fig. 3A), including the transcriptional regulator nuclear protein 1 (NUPR1). Moreover, a volcano plot was generated to illustrate the disparities in gene expression between vehicle-treated and Gln-deprived cells. This study revealed that the ATF4-NUPR1-LCN2 pathway was significantly activated in Gln-deprived cells (Fig. 3B). This pathway has been reported to inhibit ferroptosis by diminishing iron-dependent oxidative damages [33]. ATF4, a pivotal transcription factor implicated in endoplasmic reticulum (ER) stress and possessing anti-ferroptosis properties [34, 35], has been identified as a suppressor of ferroptosis in pancreatic ductal adenocarcinoma cells, exerting regulatory control over NUPR1 expression in response to diverse stressors, including ER stress [36]. NUPR1 transactivates the gene encoding LCN2, which downregulates the cellular iron level by promoting its secretion. Similarly, the qPCR results confirmed that ATF4, NUPR1, and LCN2 expression were significantly upregulated in Gln-deprived cells (Fig. 3C). The involvement of GS in the activation of the ATF4-NUPR1-LCN2 pathway was further investigated. The suppression of GS by siRNA resulted in decreased expression of ATF4, NUPR1, and LCN2 (Fig. 3C).

Fig. 3.

NUPR1/LCN2 genes participate in GS-mediated FR-EMT. (A) Transcriptomic analysis was conducted on cells subjected to Gln deprivation for varying durations (30, 3, 0, and 3 days transfected with GS siRNA), and the resulting differential expression heatmap of genes is presented. (B) A volcano plot was generated to visualize transcriptome expression profiles in MDA-MB-231 cells under Gln deprivation conditions compared to non-deprived cells. (C) The mRNA expression levels of ATF4, NUPR1, and LCN2 were assessed using qPCR in MDA-MB-231 cells subjected to Gln deprivation and Gln deprivation cells with GS knockdown. (D) Western blotting was used to analyze the protein expression levels of GS, LCN2, and TRFC in MDA-MB-231 cells that underwent lipid peroxidation induced by Erastin (Gln+) or Gln deprivation. (E) The impact of GS knockdown on LCN2 protein expression was investigated. (F) MDA-MB-231 cells were transfected with vectors harboring a plasmid for GS overexpression and subsequently verified through immunoblotting. (G) The protein expression of LCN2 was assessed in MDA-MB-231 cells with either GS overexpression plasmid or Gln deprivation. (H) The mRNA expression of TFRC in MDA-MB-231 cells with different treatments was analyzed using qPCR. (I) Western blot analysis of cellular lysates of GS-overexpressing and LCN2-knockdown MDA-MB-231 cells under non-starvation and Gln starvation conditions. (J) Representative IF images of LCN2 and IHC staining images of TFRC expression in metastatic lung cancer. Scale bars = 100 mm. Original magnification: 40х. **P < 0.01, ***P < 0.001

This study also focused on changes in LCN2 expression at the protein level, which directly modulates iron metabolism. WB analysis showed that LCN2 expression increased under Gln deprivation, and a similar increase was identified following Erastin treatment (Fig. 3D). GS knockdown decreased LCN2 expression (Fig. 3E), while GS overexpression increased its expression in MDA-MB-231 cells (Figs. 3F-G).

Furthermore, the expression of other genes implicated in iron metabolism was studied. The TFRC gene encodes transferrin receptor 1 (TfR1), which serves as the primary iron transporter on the cellular membrane [37] and plays an essential role in ferroptosis. Both transcriptomic and qPCR analyses revealed that TFRC was downregulated in Gln-deprived cells (Figs. 3B and H). In comparison, changes in TFRC level were negligible following Erastin treatment. GS, like the ATF4-NUPR1-LCN2 pathway, regulates TFRC, as evidenced by the fact that GS knockdown increased the TFRC expression (Fig. 3H).

Additionally, this study explored the role of LCN2 in EMT. GS overexpression or Gln deprivation increased the expression of the EMT biomarker Snail. In both cases, LCN2 knockdown significantly abolished the increase in Snail expression (Fig. 3I).

The changes in LCN2 and TFRC expression were further examined in the in vivo model (Fig. 3J). The mouse lung tissues treated with Gln-deprived cells showed an increase in LCN2 protein levels as measured by IF and a decrease in TFRC as measured by IHC. Consistently, when GS was simultaneously knocked down, changes in both LCN2 and TFRC were restored in different treatment groups. These results, as well as the in vitro findings, indicated that LCN2 and TFRC play crucial roles in regulating ferroptosis and EMT by GS under Gln deprivation.

Kaplan-Meier curves demonstrated that NUPR1, LCN2, and TFRC served as robust prognostic indicators associated with unfavorable survival outcomes (Figs. S4A-C). Compared with other pathological classifications and normal breast tissue, TNBC exhibited higher LCN2 expression according to TCGA data analysis (Fig. S4D). TFRC exhibited substantial upregulation in tumor tissues compared to the adjacent normal controls (Fig. S4E). TFRC expression was markedly elevated in major breast cancer subclasses, including luminal, HER2+, and TNBC breast tissue compared with normal breast tissue (Fig. S4E).

mTORC1 pathway in response to GS governs FR-EMT

Enrichment analysis of gene ontology (GO) categories was conducted using DAVID software (https://david.ncifcrf.gov/) in Gln-deprived cells. The most prominent of the top 10 enriched pathways were the cell adhesion pathways involved in EMT (Fig. 4A). Interestingly, lipid metabolism was also significantly enriched in Gln-deprived cells (Fig. 4A). Except for its canonic role in EMT, mTORC1 plays an important role in ferroptosis. It maintains cellular redox homeostasis by activating NRF2-mediated signaling and regulates lipid metabolism via SREBP1/SCD1-mediated synthesis of monounsaturated fatty acids (MUFAs) [38].

Fig. 4.

Gln starvation triggers mTORC1 activation, facilitating FR-EMT. (A) Transcriptome data of MDA-MB-231 cells was subjected to GO analysis to investigate the effects of Gln deprivation. (B) GSEA was conducted on breast cancer datasets obtained from the GEO database using gene sets regulated by mTORC1 based on the expression levels of GS. (C) The co-localization of GS and P-mTOR (Ser2448) was examined using fluorescence imaging, with quantitative analysis of fluorescence intensity in Gln deprivation cells versus WT controls. (D-E) The expression of EMT-related genes (E-cad and Snail) and MMPs (MMP-2 and MMP-7) in MDA-MB-231 cells under Gln-free conditions treated with 100 nM rapamycin (Rapa) and/or 3 mM MSO for 24 h was assessed using immunoblotting. (F) Migration assays were conducted in MDA-MB-231 cells cultured in Gln-free conditions treated with 100 nM Rapa and/or 3mM MSO for 24 h. The data are presented as mean ± standard error of the mean (SEM), with n = 3–5 random field measurements taken from a total of three transwell chambers. (G) The effects of Rapa on the inhibition of LCN2 and TFRC induced by Gln deprivation and GS overexpression were investigated using western blot analysis. **P < 0.01. (H) In the context of TNBC cells, a conceptual framework was developed to elucidate the various roles and mechanisms of GS in safeguarding against ferroptosis. Specifically, when subjected to Gln deprivation-induced stress, GS plays a protective role by mitigating lipid peroxidation and iron metabolism dysregulation, which are recognized as the principal metabolic aberrations underlying ferroptosis. Lipid peroxidation encompasses the aberrant elevation of MUFAs regulated by the mechanistic target of mTORC1. In contrast, iron metabolism perturbations involved dysregulated iron homeostasis facilitated by LCN2 and TFRC, ultimately conferring enhanced survival and metastatic potential to TNBC cells

The correlation between GS and mTORC1 expression in breast cancer samples was assessed using GSEA software. GSEA revealed that mTORC1 exhibited the highest enrichment as a gene expression signature in samples with high GS expression (Fig. 4B). The downstream targets S6K1 and 4E-BP1 were stimulated by mTORC1 phosphorylation at Ser2448 (p-mTORC1) [39]. Under normal Gln conditions, the degree of colocalization between phosphorylated mammalian target of pmTOR and GS is low. However, following Gln deprivation, colocalization was significantly enhanced (P < 0.001), indicating the formation of functional complexes between pmTOR and GS in response to metabolic stress (Fig. 4C).

Rapamycin, an allosteric inhibitor of mTORC1, along with its analogs, was used to inhibit the kinase activity of mTORC1. Rapamycin disrupted the EMT-like changes induced by MSO under Gln deprivation (Fig. 4D), as well as the expression of MMPs (MMP2 and MMP7) (Fig. 4E) and the migration ability (Fig. 4F). The transcriptomic analysis and qPCR revealed that VEGFA might act as downstream of the mTORC1 signaling pathway (Figs. S6A, B). VEGFA knockdown significantly impaired Snail expression under Gln deprivation (Figs. S6C-D). Therefore, it can be concluded that the mTORC1 pathway is necessary for GS to induce metastasis.

Furthermore, this study explored the involvement of the mTORC1 signaling pathway in GS-mediated ferroptosis resistance. GS expression was induced by MSO, and sensitivity to ferroptosis was examined after rapamycin treatment. Rapamycin effectively reversed the ferroptosis resistance induced by MSO (Fig. S6E), and this effect was correlated with downstream SREBP1/SCD1-induced lipogenesis (Figs. S6F-G), consistent with the previous study [38]. Moreover, the potential involvement of mTORC1 in iron metabolism was investigated. Rapamycin effectively mitigated the elevation of LCN2 induced by Gln deprivation or GS overexpression but exerted a negligible effect on TFRC (Fig. 4G). These findings indicated that mTORC1 plays an important role in lipid peroxidation and plays a vital role in regulating iron metabolism, which requires further investigation (Fig. 4H).

GS-mediated FR-EMT is independent of elevated lipid ROS

Previous studies have demonstrated that prolonged deprivation of growth factors, amino acids, or glucose leads to gradual cell death [40]. To examine potential cell death resulting from Gln deprivation, the human TNBC cell line MDA-MB-231 and mouse TNBC cell line 4T-1 were cultured in a basal media deficient in Gln for 0–3 days. CCK-8 detection indicated a significant decrease in cell viability (Fig. 5A). The addition of Gln to short-term-starved cells resulted in a rapid elevation in lipid peroxidation levels (Fig. 5B).

Fig. 5.

Lipid ROS induces EMT through Gln deprivation, whereas GS-mediated ferroptosis does not. (A) The viability of MDA-MB-231 and 4T-1 cells was assessed using the CCK-8 assay during different periods of Gln starvation. (B) The oxidation of C11-BODIPY was analyzed by flow cytometry to determine the time dependence in MDA- 231 and 4T-1 cells under both Gln starvation and non-starvation conditions. (C) GPX4 protein expression in MDA-MB-231 cells. (D) GSH content in MDA-MB-231 cells. (E) Immunoblotting was performed to examine the expression of E-cad and Snail in MDA-MB-231 cells cultured in either Gln-replete or Gln-free medium for 36 h, followed by treatment with Lip-1 for 24 h. β-actin was used as the loading control. (F) Representative images from a transwell migration assay illustrate the comparison between cells cultured in Gln-replete or Gln-free medium for 36 h, followed by treatment with Lip-1 for 24 h. The migrating cells were stained using HE stain. (G) The viability of MDA-MB-231 cells cultured in Gln-replete or Gln-free medium for 36 h and subsequently treated with Lip-1 for 24 h. Statistical analysis was performed using one-way ANOVA to compare three or more groups at different time points. (H) Lipid ROS was quantified with BODIPY-C11 using flow cytometry. **P < 0.01, ***P < 0.001

This study also examined the reasons leading to the elevation of lipid ROS. The absence of GPX4 leads to the buildup of peroxides in the membrane, causing the aggregation of harmful lipid ROS. In the Gln-deprived group, a significant decrease in GPX4 expression was observed (Fig. 5C). Furthermore, Gln deprivation resulted in a notable GSH depletion (Fig. 5D), of which Gln is an indispensable component.

Elevated lipid peroxidation played a crucial role in facilitating EMT (Fig. 5E) and migration (Fig. 5F) of TNBC cells under Gln deprivation, as evidenced by the significant blocking effect of liproxstatin-1 (Lip-1), an efficient antioxidant to lipid ROS. However, the addition of Lip-1 failed to rescue the cell death induced by Gln deprivation (Fig. 5G). These results indicated that the lipid ROS elevated levels under Gln deprivation contributed to EMT rather than triggering ferroptotic cell death in the absence of ferroptosis inducers.

Additionally, the correlation between lipid ROS and GS was investigated. The results showed no changes in GPX4 protein expression in MDA-MB-231 cells after GS knockdown (Fig. 5C). MSO did not further decrease the GSH level (Fig. 5D) or increase lipid ROS levels (Fig. 5H). In contrast, inhibiting lipid peroxidation did not affect the expression of GS or TFRC (Fig. 3D), with only a partial decrease in LCN2 expression. These results suggested that GS-mediated FR-EMT is independent of lipid ROS.

Artemisinin can be used as a potential therapeutic agent for GS-mediated FR-EMT

Endoperoxides, specifically endoperoxide-containing 1,2-dioxolane (FINO₂)and artemisinins, trigger ferroptosis through mechanisms distinct from the dysfunction of the GSH-GPX4 axis, as demonstrated by Erastin and RSL3 [41]. Interestingly, TNBC cells exposed to short-term Gln starvation remained sensitive to endoperoxides-induced ferroptosis (Figs. 6A-B). To elucidate the mechanism of artemisinins, this study initially examined the expression of GS after the addition of artemisinin derivatives dihydroartemisinin (DHA). As expected, the addition of DHA significantly reduced the GS levels (Fig. 6C). Remarkably, DHA also clearly disrupted the metastasis triggered by Gln deprivation and its combination with MSO (Figs. 6D-E).

Fig. 6.

ART inhibits TNBC cell lung-metastatic potential with Gln deprivation. (A, B) The CCK-8 assay demonstrated that FINO₂ and ATS exhibited a dose-dependent reduction in cell viability in MDA-MB-231 cells subjected to Gln deprivation. (C) The protein expression of GS in MDA-MB-231 cells was assessed using immunoblotting after treatment with either vehicle (H₂O) or 100 µM DHA in a Gln-replete or Gln-free medium for a duration of 24 h. (D) Snail expression in MDA-MB-231 cells was analyzed through western blotting after treatment with 100 µM DHA and/or 3 mM MSO under Gln starvation conditions. (E) Migration assays were conducted in MDA-MB-231 cells cultured in Gln-free conditions and treated with 100 µM DHA and/or 3 mM MSO for a duration of 24 h. The data are presented as mean ± SEM, with measurements taken from three to five random fields in a total of three transwell chambers. (F) The body weight of the mice was monitored every other day. (G) Approximately 5 × 10⁵ GFP-tagged 4T-1 cells were intravenously administered into the tail veins of nude mice. After 11 days of administration, the mice underwent in vivo bioluminescence imaging. (H) The number of lung metastasis nodules was quantified. Representative images of lung tissue are presented in the upper inset. The data are graphed as mean ± SEM, and P-values were determined using two-way ANOVA. *P < 0.05, **P < 0.01, ****P < 0.0001

Moreover, ACSL4 is essential for maintaining PUFA-containing phospholipids (PUFA-PLs) and ultimately influencing ferroptosis sensitivity [42]. We assessed the expression levels of ACSL4 in TNBC cells under Gln deprivation, and the results revealed that ACSL4 expression significantly decreased in 4T-1 cells (Fig. S6A). The indispensable role of ACSL4 in the induction of ferroptosis by RSL3 in MDA-MB-231 cells was also verified. Consistent with previous studies [43], ACSL4 knockdown significantly reduced the sensitivity of MDA-MB-231 cells to RSL3 under normal or Gln-deprived conditions (Figs. S6B-C). In contrast, FINO₂ and DHA exhibited comparable sensitivities to wild-type and ACSL4 knocked-down cells under normal and Gln-deprived conditions (Figs. S6D–F).

Inspired by the in vitro results, the effect of artesunate (ATS) (an artemisinin derivative with improved pharmaceutical kinetics) was further investigated in a pulmonary metastasis model using GFP-labelled 4T-1 cells. ATS can mitigate the metastasis of 4T-1 cells treated with Gln deprivation (Fig. 6G), as monitored by live imaging of its GFP label. Furthermore, the number of metastatic nodules in the lungs of five-week-old mice was also examined. Consistently, ATS significantly reduced the number of metastatic lung nodules resulting from Gln-deprived cells (Fig. 6H). Although a reduction in body weight was observed in the mice treated with ATS, this effect was gradually relieved after one week of ATS treatment (Fig. 6F). Therefore, these findings indicated that artemisinin may have the potential as a treatment option for hematogenous metastasis of human breast cancer cells mediated by GS.

Discussion

Tumors can remodel their metabolic systems to maintain malignant growth and survival under challenging conditions. GS is required for cell proliferation regardless of its catalytic activity. During mitosis, the interaction between GS and nuclear pore protein NUP88 ensures the proper functioning of the anaphase-promoting complex or cyclosome [44]. This study revealed additional functions of GS in TNBC under Gln deprivation. In the study by Kung et al. [16], Gln deprivation was maintained for 12 to 24 h, during which a high expression of GLS was observed, potentially serving as a stress response in the tumor microenvironment and aiding in sustaining the metabolic demands of cancer cells. In contrast, in this study, TNBC cells were subjected to Gln deprivation for up to 2 to 3 days, and it was found that the high expression of GS triggered metabolic reprogramming. Its activation induces EMT-like changes and also promotes resistance to classic ferroptosis inducers. Cancer cells in a therapy-resistant state, including those undergoing EMT, have been observed to become sensitive to ferroptosis, specifically the GPX4 pathway. The main reason for their enhanced sensitivity was ascribed to the remodeling of lipid metabolism to favor PUFAs. In contrast, the EMT observed in TNBC cells under Gln deprivation exhibited resistance to ferroptosis. Such resistance to ferroptosis has been found to facilitate metastasis, given the oxidative stress in blood [45].

GS-mediated EMT and ferroptosis are related to non-catalytic function of GS. The GS activity inhibitor MSO increased GS expression and promoted both EMT and ferroptosis. These findings raise concerns that the pharmacological blockade of GS activity with small molecules can exacerbate ferroptosis resistance and contribute to tumor metastasis.

Under Gln deprivation, GS activation lowers iron levels through at least two pathways. First, the NUPR1/LCN2 pathway was upregulated by GS during the initial phase of Gln deprivation. The activation of this pathway has been previously identified as a repressor of ferroptosis using canonic ferroptosis inducers [33] by mitigating iron-induced oxidative damage. Second, the iron uptake receptor TFRC, which is essential for ferroptosis, was downregulated by GS. Collectively, GS-mediated adaptations, characterized by the upregulation of NUPR1/LCN2 and the downregulation of TFRC, work synergistically to decrease intracellular iron availability, thereby inhibiting ferroptosis in Gln-deprived TNBC cells.

Under conditions of Gln deficiency, cells sustain survival and adaptive stress responses through the activation of genes associated with lipid metabolism [46]. This mechanism appears to be intricately linked to the induction of EMT, as the reprogramming of lipid metabolism facilitates the enhancement of cell membrane remodeling and migratory capabilities [47]. For instance, the upregulation of fatty acid synthase and genes involved in sphingolipid synthesis may promote EMT and subsequently increase cell invasiveness by influencing the Wnt/β-catenin signaling pathway [47]. These findings underscore a complex interplay between lipid metabolism and EMT, offering novel insights for the development of anti-tumor strategies that target metabolic reprogramming. Moreover, it was found that GS mediated the mTORC1 pathway. In addition to its classic role in EMT, mTORC1 activation can inhibit ferroptosis in cancer cells via downstream modulation of SREBP1/SCD1-mediated MUFA [38]. This study elucidated that rapamycin, an allosteric inhibitor of mTORC1, disrupted the regulation of GS, thereby conferring resistance to RSL3-induced ferroptosis and metastasis.

The strategy of targeting GS can have short-term anti-tumor effects during glutamine deprivation by inhibiting tumors’ internal Gln synthesis [48], enhancing chemotherapy sensitivity [49]. However, long-term use is ineffective due to metabolic adaptation [50]. Clinically, Gln status requires dynamic monitoring: short-term deprivation (≤ 72 h) shows plasma Gln < 420 µmol/L and a high glutamate/glutamine ratio, often seen in acute interventions like preoperative fasting [51]. Long-term deprivation (> 7 days) leads to metabolic remodeling and risks organ failure [52]. Combining GS inhibition with Gln deprivation poses risks like tissue toxicity [4], immune suppression [53], and metabolic imbalance, potentially causing hyperammonemia due to impaired GS function in the CNS [32]. The clinical translation strategy should focus on three key areas: (1) Implement time-limited interventions with strict control of combination therapies within 72 h and use intermittent administration (e.g., a 48-hour cycle) to prevent compensation mechanisms; (2) Accurately define patient status by dynamically stratifying based on plasma Gln levels, ¹⁸F-FGln PET imaging, and metabolomics; (3) Mitigate risks by avoiding use in patients with liver/kidney dysfunction and developing tumor-targeted prodrugs like JHU083 to minimize systemic toxicity. Ultimately, the clinical potential of GS targeting hinges on precisely defining Gln status and regulating combination regimens spatiotemporally. Future efforts should focus on personalized experimental validation using biomarkers like GAC expression.

Additionally, we found that organic peroxides, including ART, ATS, and FINO₂, which, independent of ACSL4, trigger ferroptosis, may still provide a strategy to inhibit metastasis even when cancer cells become resistant to classical ferroptosis inducers. Therefore, this study sheds light on the mechanism by which organic peroxides trigger ferroptosis and potentially intercept metastasis in TNBC cells under Gln deprivation. However, the specific mechanism by which organic peroxides regulate GS expression remains unknown. Although ATS has demonstrated certain efficacy on the metastasis mediated by Gln deprivation, additional specifically targeted molecules are definitely worth investigating to understand the value of GS inhibition completely.

Conclusions

In summary, Gln deprivation in TNBC cells inhibited cell viability and conferred resistance to ferroptosis, initiated and promoted the EMT process. GS plays a crucial role in modulating the cellular response to Gln deprivation by decreasing the sensitivity of TNBC cells to ferroptosis. Additionally, GS facilitates the EMT process, mitigates oxidative stress within the bloodstream, and exhibits a positive correlation with metastatic potential in vivo. This study contributes to understanding the functions of ferroptosis in metastasis, along with providing evidence suggesting that GS can serve as a potential target for the prevention of metastasis in TNBC.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

TNBC

Triple negative breast cancer

EMT

Epithelial mesenchymal transition

GS

Glutamine synthetase

Gln

Glutamine

FR-EMT

Ferroptosis resistant EMT

MUFAs

Monounsaturated fatty acids

LCN2

Lipocalin 2

TFRC

Transferrin receptor

ER

Estrogen receptor

PR

Progesterone

HER2

Human epidermal growth factor receptor 2

GLUL

Glutamate ammonia ligase

CCLE

Cancer cell line encyclopedia

TIF

Tumour interstitial fluid

NUPR1

Transcriptional regulator nuclear protein 1

Author contributions

Conceptualization and design: Liu Q, Yang Z. Manuscript drafting: Liu Q, Yang Z. Manuscript revision: Liu Q, Yang Z. Experiment and statistical analysis: Yang Z, Lian X, Guo G. Data analysis: Yang Z, Lian X, Luo Y, Ye Q. Figure design: Yang Z, Lian X。.

Funding

This work was supported by grants from the National Natural Science Foundation of China (82203788, 91643112).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval and consent to participate

This study involving human participants was approved by the institutional ethics committee of Air Force Medical Center, Air Force Medical University (approval no. 2024-69-PJ01). Written informed consent was obtained from each participant enrolled in the study. Peking University Health Science Center’s Department of Animal Science approved the project license and all animal studies were conducted in compliance with institutional welfare guidelines (protocol n. LA2022256).

Consent for publication

All authors have consented to the publication of this article.

Competing interests

The authors declare no competing interests.