Extracellular vesicles derived by hypoxic preconditioned EPCs attenuated atherosclerosis by suppressing ferroptosis via regulating arginine metabolism Authors: Xing Luo, Xinxin Zhu, Yuwu Chen, Yiyuan Guo, Biyi Xu, Junke Mou, Mengyang Wang, Qishuo Gu, Qianhui Sun, Ying Lv, Xingtao Huang, Ming Zeng, Boling Yi, Xiaolin Liu, Xiaoxuan Bai, Cheng Fang, Ji Li, Bo Yu, Haibo Jia Published: 2026 ABSTRACT Endothelial progenitor cells (EPCs) derived extracellular vesicles (EVs) are involved in maintaining vascular homeostasis, while effects of hypoxic preconditioned EPCs-derived EVs (Hy-EVs) on atherosclerosis (AS) remain unclear. Hy-EVs and normal oxygen EVs (N-EVs) were pretreated for ApoE−/− mice and ox-LDL treated macrophages. N-EVs administration attenuated atherosclerosis progression, Hy-EVs further decreased the plaque area, increased collagen content, reduced lipid deposition and improved inflammatory response. Proteomics and in vitro data suggested that Hy-EVs inhibited macrophages ferroptosis and improved oxidative stress via increasing GPX4 and xCT level. Hy-EVs proteomics revealed that argininosuccinate synthetase1 (ASS1) levels were significantly higher in Hy-EVs than N-EVs, ASS1 knockdown in EPCs blocked the protective effects of Hy-EVs on macrophage ferroptosis and AS progression, whereas ASS1 supplementation restored the effect. Mechanistically, increased ASS1 in Hy-EVs suppressed GCN2 activation by enhancing arginine generation in macrophages. GCN2 agonist or arginine consumption significantly attenuated the protective effect of Hy-EVs on macrophage ferroptosis and AS progression. Furthermore, LC/MS indicated that GCN2 bonded to NRF2, while arginine generated by ASS1 enhanced NRF2 nuclear translocation and increased levels of its downstream target gene GPX4/xCT by reducing GCN2/NRF2 interaction. Our study suggested that Hy-EVs further attenuated atherosclerosis progression by inhibiting macrophage ferroptosis via enhancing arginine generation and inhibiting GCN2/NRF2 interaction. The data provides strong evidence for the translational application of EPCs derived Hy-EVs as atherosclerotic plaques. Abbreviations Atherosclerosis Endothelial progenitor cells Extracellular vesicles glutathione peroxidase 4 hypoxic preconditioned EPCs-derived EVs solute carrier family 7 member 11 Introduction Atherosclerosis (AS), a chronic inflammation of vascular wall, is the pathological basis of numerous cardiovascular diseases [1,2]. Lipid-lowering and anti-inflammatory interventions constitute cornerstone therapeutic interventions for atherosclerotic management [1,3]. However, there are still patients who experience rapid progression of atherosclerotic plaques after receiving standard therapy [4,5]. It is necessary for reducing cardiovascular diseases to explore new treatment strategy. Endothelial progenitor cells (EPCs) are bone marrow-derived stem cells with differentiation potential. EPCs play a role in maintaining vascular homeostasis by repairing endothelial cell function [6]. It has been confirmed that the decreased circulating EPCs is highly correlated with poor clinical prognosis in patients with cerebral atherosclerosis [7]. Similar results have been found in patients with coronary artery disease [[7], [8], [9]]. Furthermore, several studies reported EPCs transplantation limited the atherosclerotic plaque progression in mice and rabbit models [10,11]. Macrophages are the most important immune cells in atherosclerotic plaques, and modulating macrophage inflammatory responses represents a potential therapeutic target for atherosclerosis [12]. Macrophages possess significant phagocytic capabilities. Previous studies have demonstrated that stem cells regulate macrophage function to improve inflammatory responses [13,14]. However, the role and function of EPCs in macrophages within atherosclerotic plaques remain unclear. Extracellular vesicles (EVs) are small lipid membrane vesicles that carried nucleic acids/proteins/metabolites to regulate the function of target cells. Recently, there has been growing evidence that stem cell-derived extracellular vesicles had a therapeutic potential comparable to or even exceeding that of stem cell transplantation [15,16]. In particular, the therapeutic effect of stem cell-derived EVs can be significantly enhanced by pretreatment of stem cells with preconditioned by hypoxia or melatonin [[17], [18], [19], [20]]. Previous studies have also reported that extracellular vesicles derived from stem cells suppressed the progression of atherosclerosis, and the potential mechanism may be through improving macrophage polarization and lipid transport [21,22]. However, it is currently unclear whether hypoxia treatment can enhance the role and potential mechanism of EPCs derived extracellular vesicles in improving the progression of atherosclerosis. Ferroptosis is a novel programmed death characterized primarily by lipid peroxidation of cell membranes [23]. There is growing evidence that ferroptosis is involved in all stages of atherosclerosis progression [24,25]. Meanwhile, our previous study also confirmed that targeted inhibition of ferroptosis may be a potential intervention in atherosclerosis [24,26]. Imbalance of amino acid antioxidant system, disorder of iron metabolism and lipid peroxidation deposition are regarded as the key mechanisms of ferroptosis [27,28]. Cystine-glutamate transporter xCT/solute carrier family 7 member 11 (xCT/SLC7A11) and glutathione peroxidase 4 (GPX4) are responsible for the regulation of cystine transport and lipid peroxidation, respectively, and their decreased expression levels lead to ferroptosis [29]. During atherosclerosis development, the regulatory mechanism of ferroptosis and its feasibility as a target for clinical intervention need to be further explored. In the present study, we investigated the effect of EVs from hypoxic preconditioned EPCs (Hy-EVs) on improving atherosclerosis. We found that hypoxic-pretreated EVs improved the progression of AS by inhibiting macrophages ferroptosis in atherosclerotic plaque, further to confirm the underlying mechanism. Materials and methods Animal model and ethical approval The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the institutional research ethics committee of the Second Affiliated Hospital of Harbin Medical University (YJSKY2022-1140). All animal procedures performed must conform to the guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes or current NIH guidelines. The mice were housed in the animal center of the Second Affiliated Hospital of Harbin Medical University and had free access to food and water (22 ± 2 °C, 55 ± 5% relative humidity with a 12-h light/dark cycle). Euthanasia and statement All animal procedures performed must conform to the guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes or the current NIH guidelines. Mice were anaesthetized by inhalation of an overdose of isoflurane (2%) and euthanized via cervical dislocation. Death was confirmed by the absence of a heartbeat and reflex. Cell culture Bone Marrow-derived macrophages (BMDMs) were isolated from C57BL/6 mice by flushing the femur and tibia with PBS. The bone marrow cells were resuspended in RPMI-1640 (Gibco, 31870) containing 1% l-glutamine, 1% Pen/Strep, and 10% FBS and 20 ng/mL mM-CSF for 6 days. It was reported that EVs can be taken up by cells within 2 h [30]. According to our previous study [24], BMDMs was pretreated with 25 μg/mL EVs and Hy-EVs for 1h, followed by treated with 100 μg/mL of ox-LDL for 48 h [31,32]. According to previous studies [33,34], 200 nM Hal, 5 μM GCN2iB, 100 μM l-arginine, 750 ng/mL ADI-PEG 20 were used to pretreated BMDMs. RNA extraction and PCR According to the manufacturer's instructions. RNA was extracted from cells and mouse aorta using TRIzol® reagent (Invitrogen), purified using chloroform and isopropanol, and reverse transcribed to cDNA. qPCR was performed using a SYBR kit (ROCHE) on 7900HT system (Applied Biosystems; Thermo Fisher Scientific, Inc.) as follows: 60 °C for 1 min, 90 °C for 15 min, followed by 40 cycles of 90 °C for 15 s and 55 °C for 60 s. β-actin was used as an internal control. The standard 2ΔΔCt method was used for data analysis [35]. The primer in the present was shown in Supplemental Table 1. LC/MS‐MS analysis GCN2-Flag‐expressing stable BMDMs were lysed by NP‐40 lysis buffer (Beyotime). The whole cell extracts were incubated with 5 μL anti‐Flag (Cell Signaling Technology) Ab or rabbit IgG control Ab for 4 h in 4 °C followed by ultracentrifugation and 1 h incubation with 10 μL Protein A + G agaroses (Santa Cruz Biotechnology) in 4 °C. The beads were identified on a Fusion Lumos Mass Spectrometer (Thermo Fisher Scientific). Flow cytometry Cells were trypsin-digested (0.25%,37 °C for 1 min) after treatment according to the manufacturer's instructions and incubated for 15 min at 37 °C. utilizing the C11 dye BODIPY (Dojindo) before washing with PBS. Fluorescence intensity was detected using flow cytometry (BD Bioscience) at 581 nm. Isolation and culture of EPCs According to previous study [10], Ficoll density gradient centrifugation was used to isolate EPCs in the tibia and fibula 6-8 weeks male C57BL/6J mice. Briefly, mice were euthanized utilizing 75% ethanol immersion. Bone marrow cells were isolated in a sterile environment, and the bone marrow cavity was flushed with DMEM medium and filtered through a 70 μm cell filter to obtain single cell suspension. Cell suspensions were separated using Ficoll-Paque PREMIUM (Cytiva) according to the instructions and centrifuged at 400 g for 5 min at 4 °C. The middle layer of cells was aspirated, added to the medium and centrifuged again. Immunomagnetic bead sorting was used to further purify EPCs. The cells were then cultured on human fibronectin-coated plates in endothelial cell growth media (EGM-2) supplemented with EGMTM-2 MV Single Quots TM (Lonza, Basel, Switzerland), mouse recombinant vascular epidermal growth factor (20 ng/mL), insulin-like growth factor (4 ng/mL), fibroblast growth factor (4 ng/mL) and 5% fetal bovine serum. Following four days of culturing in a 5% CO2 incubator, nonadherent cells were removed by washing with PBS and adherent cells were incubated in fresh media for further 3 days and were then used for following experiments. Hypoxia preconditioning According to previous study [10], cells were transferred to a three-gas incubator (Thermo Scientific, model 8000) and maintained at 1% O2, 5% CO2, and balanced N2 at 37 °C for 24 h. Normoxic control cells were cultured under standard conditions (21% O2, 5% CO2) for the same duration. Isolation of extracellular vesicles EPCs were cultivated in an EVs-free culture medium. After EPCs underwent different treatments, the cell supernatants were collected. Cell supernatants were used to extract N-EVs and Hy-EVs. Firstly, filter cell debris using a 0.22 μm cell filter. The supernatants were centrifuged at 300 g for 10 min, 2000 g for 10 min and 10, 000 g for 30 min the supernatant was then transferred to Ultra-Clear tubes and centrifuged at 100,000 g for 70 min at 4 °C with a SW32Ti rotor (Beckman Coulter, Netherlands). Cold PBS washed and centrifuged at 100,000 g for 70 min. Finally, the pellet was then carefully re-suspended in 100 μL PBS. The samples were used immediately or stored at −80 °C. Nanoparticle tracking analysis (NTA) The particle size and distribution of EVs were analyzed by the ZetaView PMX110 (PMX, Germany) machine. Each sample was analyzed three times. PKH67 labeled N-EVs and Hy-EVs According to previous study [36], A 50 μL N-EVs and Hy-EVs suspension was mixed with 0.5 mL of Diluent C. Separately, 0.5 mL of Diluent C was mixed with 4 μL of PKH67 dye and then added to the EV suspension. After 15-min incubation, 1 mL of 0.5% BSA was added to stop staining. The EVs were re-purified to obtain PKH67-labeled EVs. These labeled EVs were co-cultured with BMDMs for 24 h. The distribution of EVs in cells was observed using confocal microscopy. In vivo imaging: DiR dye was added to PBS to prepare a working solution, which was then mixed with EVs suspension (100 μL working solution per 1000 μL of EV suspension). The mixture was vortexed and incubated for 30 min. After incubation, PBS was added and the EVs were enriched by ultracentrifugation to obtain labeled EVs. Anaesthetized mice were administered 100 μL of the labeled EVs via tail vein injection. After 1h, the distribution of EVs in lung, heart, spleen, liver, and artery was observed using an in vivo imaging system or confocal microscope. Animal model 6-8-week ApoE−/− male mice fed high-fat for 16 weeks were used to construct an atherogenic model. Mice were divided into the following groups. ND group, mice were given a normal diet for 16 weeks. HD group, mice were given a high fat diet for 16 weeks (21% fat, 0.15% cholesterol; MD12015; Medicine Ltd., Jiangsu, China). The grouping and treatment of the animal are presented in Supplemental Fig. 1. Experiment 1: N-EVs groups, ApoE−/− mice were fed with high fat received tail vein injections with 200 μL (100 μg) of EPCs derived N-EVs twice a week for 16 weeks. Hy-EVs groups, the ApoE−/− mice were fed with high fat received tail vein injections with 200 μL (100 μg) of EPCs derived Hy-EVs twice a week for 16 weeks. Experiment 2: Sh–NC–Hy-EVs groups and the ApoE−/− mice were fed with high fat received tail vein injections with 200 μL (100 μg) of Sh-NC EPCs derived Hy-EVs twice a week for 16 weeks. Sh-ASS1-Hy-EVs groups and the ApoE−/− mice were fed with high fat received tail vein injections with 200 μL (100 μg) of Sh-ASS1 EPCs derived Hy-EVs twice a week for 16 weeks. Sh-ASS1-Hy-EVs + AAV-ASS1 groups: ApoE−/− mice were injected with AAV (F4/80) carrying a macrophage-specific promoter to overexpress argininosuccinate synthetase1 (ASS1) in macrophages for two weeks, fed with high fat received tail vein injections with 200 μL (100 μg) of Sh-ASS1 EPCs derived Hy-EVs. Experiment 3: AAC-NC/AAV-GCN2 groups, ApoE−/− mice were injected with AAV (F4/80) carrying a macrophage-specific promoter to overexpress GCN2 in macrophages for two weeks, fed with high fat received tail vein injections with 200 μL (100 μg) of EPCs derived Hy-EVs. After 16 weeks, the mice were euthanized, plasma, aortic plaques and aorta were collected for further analysis. In this study, all animal experiments were conducted in accordance with a rigorously designed protocol to minimize bias. Randomization: Following the establishment of the disease model, animals were randomly assigned to the experimental or control groups using a computer-generated random number sequence. This randomization was stratified by body weight to ensure a comparable baseline distribution across groups. Blinding: A double-blind approach was implemented throughout the subsequent procedures. The investigator responsible for administering treatments and performing daily observations was blinded to the group allocation. Furthermore, a different investigator, who was also blinded to the treatment identities, conducted all pathological and molecular biology experiments analyses. According to the guidelines of the American Heart Association for the study of atherosclerosis, a sample size of 10 mice per group was used [37]. Western blot Proteins from mouse aorta and cells were extracted by RIPA (Beyotime) and lysed on ice for 30 min after combined with ultrasound (30% power) assisted lysis. Centrifuge at 12,000 rpm for 15 min at 4 °C and extract the supernatant. Add Loading buffer (Beyotime) in the ratio of 1:4, heat at 95 °C for 10 min and store at −80 °C. SDS-PAGE gel electrophoresis was used to separate proteins and transferred to PVDF membranes (Beyotime) and incubated overnight in primary antibody, 1 h at room temperature in secondary antibody, and ECL (Beyotime) exposure after TBST wash. Anti-GPX4 antibody (abcam, ab125066, 1:1000), Anti-xCT antibody (abcam, ab307601, 1:1000), Anti-NRF2 antibody (abcam, ab62352, 1:1000), Anti-ASS1 antibody (Cell Signaling Technology, D4O4B, 1:1000), Anti-GCN2 antibody (Cell Signaling Technology, 3302, 1:1000), Anti-Alix antibody (Cell Signaling Technology, 92880, 1:1000), Anti-CD81 antibody (Cell Signaling Technology, D5O2Q, 1:1000), Anti-β-Actin antibody (Cell Signaling Technology, 4967, 1:1000), Anti-GAPDH antibody (Cell Signaling Technology, 14C10, 1:1000), Anti- Histone H3 antibody (Cell Signaling Technology, 9715, 1:1000). Seahorse assays The OCR of BMDMs was measured using XFe96 Extracellular Flux Analyzer (Seahorse Biosciences, USA) according to the manufacturer's protocol. Cells were seeded at 3 × 103 cells per well in an XFe96 cell culture microplate and subjected to the respective treatments. The culture medium was replaced with Seahorse XF Base Medium supplemented with 2 mM glutamine, 10 mM glucose, and 1 mM sodium pyruvate, followed by a 1-h equilibration period. Basal OCR was recorded, followed by sequential injections of oligomycin (1.5 μM) to assess adenosine triphosphate (ATP)-linked respiration, carbonyl cyanide-4- (trifluoromethoxy) phenylhydrazone (FCCP, 0.5 μM) to determine maximal respiratory capacity, and a combination of antimycin A and rotenone (1 μM each) to measure non-mitochondrial respiration. Key parameters, including basal respiration, ATP production, and maximal respiration, were calculated to evaluate mitochondrial function. Proteomics At present, the proteomics part of the experiment is carried out by Shanghai Ouye Biotechnology Co., Ltd. An aliquot of 100 μg of protein was extracted and a corresponding volume of enzyme dilution (protein: enzyme = 50:1 (m/m), 100 μg protein plus 2 μg enzyme) was added to the amount of protein to re-solubilize the protein precipitates, and then the solution was incubated for 12 h at 37 °C. The samples were incubated for 12 h at 37 °C. The samples were frozen or evaporated for digestion. Finally, the samples were lyophilized or evaporated for enzymatic digestion. For TMT labelling, the lyophilized samples were resuspended in 30 μL 100 mM TEAB and Labeling reaction in a 1.5 mL Ep tube. 20 μL acetonitrile were added to TMT reagent vial at room temperature. The centrifuged reagents were dissolved for 5 min and mixed for centrifugation and repeat this step once. Then 10 μL of the TMT label reagent was added to each sample for mixing. The tubes were incubated at room temperature for 1 h. Finally, 5 μL of 5% hydroxylamine were added to each sample and incubated for 15 min to terminate reaction. The labeling peptides solutions were lyophilized and stored at −80 °C. Proteome Discoverer (v.2.4) was used to search all of the raw data thoroughly against the Uniprot Mus Musculus database. Immunofluorescence BMDMs were grown in 24-well plates, fixed for 10 min at room temperature utilizing 4% formaldehyde, and washed with PBS. Triton treatment at 0.1% concentration for 5 min was used to disrupt the cell membranes, after which the cells were closed for 30 min utilizing goat serum, the primary antibody was added overnight. The fluorescent secondary antibody was then utilized to incubate the cells for 1 h at room temperature, and the cells were stained with DAPI and then visualized under a confocal microscope. Paraffin sections of mouse aortic plaques were deparaffinized, cell membranes were disrupted in the same manner, goat serum was closed, and primary and fluorescent secondary antibodies were added. Anti-NRF2 antibody (abcam, ab62352, 1:100), Anti-GCN2 antibody (Cell Signaling Technology, 3302, 1:50). Anti-CD68 antibody (abcam, ab283654, 1:50), Anti-xCT antibody (abcam, ab307601, 1:100) and Anti-GPX4 antibody (abcam, ab125066, 1:100). Proximity ligation assay (PLA) After processing the macrophages in different groups, they were treated using the above cell immunofluorescence method. Primary antibodies for GCN2 (Santa Cruz, sc-374609, 1:100) and NRF2 (abcam, ab325240, 1:50) in DUOLINK Antibody Diluent were then added and incubated at a 4 °C environment overnight. Duolink In Situ PLA Probe Anti-Rabbit PLUS and Anti-Mouse MINUS were added and incubated for 2 h at 37 °C. Finally, the coverslips were mounted onto glass slides with DAPI to stain nuclei. Images were acquired by Zeiss LSM 980 laser-scanning confocal microscope and analyzed using the ImageJ software (version 2.10). Stable isotope tracing In order to track the synthesis of arginine, BMDMs were grown in 60-mm dishes. Cells were rinsed with PBS twice and changed to a medium containing aspartate (labeled with 15N) until plated cells reached approximately 70-80% confluency. To determine isotopologue distributions in soluble metabolites, metabolites were quantitated by Metabo-Profile Biotechnology. Samples were thawed in an ice bath to reduce sample degradation before processing. Add 400 μl of 80% methanol solution to each tube containing the cell samples and sonicate the cells (JY92-IIN, NingBo Scientz Biotechnology Co., Ltd.). The samples were centrifuged at 18,000 g for 15 min at 4 °C (Microfuge 20R, Beckman Coulter, Inc., Indiannapolis, IN, USA), and all the supernatant was collected after centrifugation and concentration (Labconco, Kansas City, MO, USA). For liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) analysis, the samples were analyzed by ultra-performance liquid chromatography coupled to a tandem mass spectrometry (UPLC-MS/MS) system (ACQUITY UPLC-Xevo TQ-S, Waters Corp., Milford, MA, USA). The optimized instrument settings are briefly described as follows. For HPLC, column: UPLCHILIC (4.6-mm column, Amide X Bridge, Waters), column temp. 40 °C, sample manager temp.: 10 °C, mobile phases: A = 5% acetonitrile in water (with.20 mM ammonium hydroxide and ammonium acetate); and B = acetonitrile, gradient conditions: 0-3.5 min (85-32% B), 3.5-12 min (32-2% B), 12-16.5 min (2% B),16.5-17 min (2-85% B), 17-25.5 min (85% B), flow rate: 0.40 mL min−1, and injection vol.: 5.0 μL. For the mass spectrometer, capillary 2.5 (ESI-) Kv, source temp.: 150 °C, desolvation temp.: 550 °C, and desolvation gas flow: 1000 L h−1. The raw data files generated by UPLC-MS/MS were processed using MassLynx software (v4.1, Waters, Milford, MA, USA) to perform peak integration, calibration, and quantitation for each metabolite. CCK8 assay Cells were grown in 96-well plates according to the manufacturer's instructions. After configuring the CCK8 working solution (Dojindo), 10 μL of working solution and 90 μL of medium were added to each well and incubated for 1 h at 37 °C. Absorbance was detected at 481 nm. JC-1 staining Follow the manufacturer's instructions (Beyotime). Cells were grown in 48-well plates. The working solution was configured according to the ratio of 1 mL of buffer to 5 μL volume of JC-1 storage solution, and 200 μL of working solution was added to each well and incubated for 30 min at 37 °C. The cells were observed using confocal microscope at 488 nm and 594 nm, respectively. Targeted metabolomic Targeted metabolomic analysis of intracellular arginine and ornithine concentrations was performed using LC-MS/MS. Cultured cells were rapidly washed twice with ice-cold PBS to remove extracellular metabolites, followed by immediate quenching of metabolism using cold 80% methanol (pre-chilled to −80 °C) directly on the culture dish/plate. Cells were then scraped into the quenching solvent, transferred to pre-cooled microcentrifuge tubes, and snap-frozen in liquid nitrogen. After thawing on ice, samples were vortexed vigorously and subjected to probe sonication on ice (3 cycles of 10 s pulse, 20 s rest) to ensure complete cell lysis and metabolite extraction. The resulting lysates were centrifuged at 16,000 g for 15 min at 4 °C to pellet protein and cellular debris. The supernatant, containing the extracted metabolites, was carefully transferred to new autosampler vials and dried completely under a gentle stream of nitrogen gas. Dried extracts were reconstituted in 50 μL of initial LC mobile phase (0.1% formic acid in water) immediately prior to analysis. Chromatographic separation was achieved using hydrophilic interaction liquid chromatography (HILIC) on a ZIC-pHILIC column (150 mm × 2.1 mm, 5 μm) maintained at 40 °C, with a binary gradient elution (mobile phase A: 20 mM ammonium carbonate, 0.1% ammonium hydroxide in water; B: acetonitrile) at a flow rate of 0.25 mL/min. Detection and quantification utilized electrospray ionization in positive mode (ESI+) on a triple quadrupole mass spectrometer operating in scheduled multiple reaction monitoring (MRM). Optimized precursor-to-product ion transitions were monitored for arginine (m/z 175.1 → 70.1, 175.1 → 116.1), ornithine (m/z 133.1 → 70.1, 133.1 → 116.1), and their corresponding stable isotope-labeled internal standards (e.g., l-Arginine-13C6, l-Ornithine-13C5,15N2). Quantitation was based on peak area ratios of analytes to their respective internal standards, referenced against calibration curves prepared in matrix-matched solvent. Data acquisition and processing were performed using instrument-specific software (e.g., Skyline or Analyst). Intracellular concentrations were normalized to total cellular protein content determined by BCA assay on the precipitated pellet. Statistical analysis All the results were presented as mean ± SD. The exact sample size in each group was listed in the corresponding figure legends. Kolmogorov-Smirnov test and Shapiro-Wilk test were employed to test normality of the data distribution. For comparisons, normally distributed data were analyzed by nonpaired 2-tailed Student's t-test (2-group analysis), and analysis of one-way ANOVA followed by Bonferroni post hoc test for multiple pairwise comparisons. For animal studies, the sample size for experiment was based on power calculations, published reports and our experience. Mice in the study were randomly allocated to experimental groups using simple randomization. Whenever possible, the investigators were blinded to group allocation during experiments and outcome assessment. Statistical analysis were performed using IBM SPSS Statistics (version 21.0), and P < 0.05 was considered statistically significant. Results Hypoxia precondition enhanced the atherosclerotic protective ability of EVs derived from EPCs EPCs were isolated from mice as previous study described [38,39], EPCs exhibited a ‘cobblestone-like’ morphology typical of mature endothelial cells at confluence (Supplemental Fig. 2A). Flow cytometry was used to analyze the markers of EPCs, the results showed that EPCs expressed surface markers CD31, KDR, VE-Cadherin and e-NOS, while not express CD45 (Supplemental Fig. 2B). After treating EPCs with normoxia and hypoxia respectively, the N-EVs and Hy-EVs in the supernatant were collected. The morphology and diameter of EVs were analyzed by electron microscopy and NTA, confirming a diameter range of 50-150 nm (Supplemental Fig. 2C and D). Western blot was used to detect the markers of extracellular vesicles, data suggested that hypoxia promoted the secretion of EVs by EPCs. We also measured the expressions of Histone H3 and calnexin to confirm the purity of N-EVs and Hy-EVs. (Supplemental Fig. 2E). All of these results indicated that the methods for isolating EVs and their purity meet the requirements of MISV [40]. In order to compare the protective effects of N-EVs and Hy-EVs on atherosclerosis, equal amounts of N-EVs and Hy-EVs were administered to high-fat fed ApoE−/− mice through the tail vein of mice. HE staining showed that N-EVs reduced the atherosclerotic plaques area in the aortic root caused by high-fat feeding, while Hy-EVs further reduced the plaque area and necrotic core (Fig. 1A, C, F). Masson staining showed that N-EVs administration increased the collagen content in aortic root, while Hy-EVs further increased the collagen level (Fig. 1B, D). Oil red staining of the mouse aorta revealed that N-EVs treatment reduced lipid deposition in the aorta induced by high fat. Compared with the N-EVs group, the lipid deposition area in the Hy-EVs group was significantly lower ( Fig. 1E, G). Moreover, mice serum was collected to detect the inflammatory factors, the levels of pro-inflammatory factors (IL-1β, IL-18 and TNF-α) were increased in HD group significantly, N-EVs groups showed a decrease of pro-inflammatory factors, while Hy-EVs groups decreased the pro-inflammatory factors level (Fig. 1H-J). Moreover, compared with N-EVs, the expression level of serum IL-4 and IL-10 in mice of the Hy-EVs group was higher (Fig. 1K and L). These results demonstrated that EPCs derived N-EVs attenuated atherosclerosis development, hypoxic pretreatment further enhanced the atherosclerotic protective effect of EVs derived from EPCs.Fig. 1Hy-EVs further attenuated atherosclerosis development. (A) HE staining of aortic root plaque. n = 10. Scale bars:200 μm. (B) Masson staining of aortic root plaque. n = 10. Scale bars:200 μm. (C) Quantification of the area of HE-stained plaques. n = 10. (D) Quantification of the plaque content of Masson-stained plaques. n = 10. (E) Area of oil Red Staining. n = 10. Scale bars:1 cm. (F) Area of necrotic core in the plaque. n = 10. (G) Quantification of the area of the aortic oil red positive region. n = 10. Inflammatory factors in the plasma of different mice were detected. (H) IL-1β level. n = 10. (I) IL-18 levels. n = 10. (J) TNF-α level. n = 10. (K) IL-4 levels. n = 10. (L) IL-10 levels. n = 10. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05.Fig. 1 Hy-EVs further attenuated atherosclerosis development. (A) HE staining of aortic root plaque. n = 10. Scale bars:200 μm. (B) Masson staining of aortic root plaque. n = 10. Scale bars:200 μm. (C) Quantification of the area of HE-stained plaques. n = 10. (D) Quantification of the plaque content of Masson-stained plaques. n = 10. (E) Area of oil Red Staining. n = 10. Scale bars:1 cm. (F) Area of necrotic core in the plaque. n = 10. (G) Quantification of the area of the aortic oil red positive region. n = 10. Inflammatory factors in the plasma of different mice were detected. (H) IL-1β level. n = 10. (I) IL-18 levels. n = 10. (J) TNF-α level. n = 10. (K) IL-4 levels. n = 10. (L) IL-10 levels. n = 10. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. Hy-EVs inhibited macrophage ferroptosis via increasing GPX4/xCT expression Macrophages are potent phagocytes with a high efficiency for EV uptake, making them key targets for circulating EVs [[41], [42], [43]]. To determine whether N-EVs and Hy-EVs reached the sites of atherosclerotic plaque, N-EVs and Hy-EVs were stained with DiR and observed 12 h after tail vein injection. The results showed that DiR-labeled N-EVs and Hy-EVs were distributed in major organs including liver, lung, artery, heart, kidneys, and spleen (Supplemental Fig. 3A). Further localization of EVs in mouse AS plaques revealed that EVs were predominantly co-localized by CD68-positive macrophages in AS plaques (Supplemental Fig. 3B). Therefore, we focused on the effects of N-EVs and Hy-EVs on macrophages. In the in vitro model, BMDMs were isolated and treated with N-EVs and Hy-EVs respectively under the stimulation of ox-LDL. Immunofluorescence confirmed that both N-EVs and Hy-EVs labeled with PKH67 could be phagocytosed by macrophages (Supplemental Fig. 2E). Furthermore, we observed that some Hy-EVs were engulfed by endothelial cells and smooth muscle cells. Therefore, we evaluated the effects of Hy-EVs on endothelial dysfunction and smooth muscle cell proliferation in atherosclerotic plaques. We found that Hy-EVs significantly improved endothelial dysfunction and reduced smooth muscle cell proliferation (Supplemental Fig. 3C–G). Further to study the effects of Hy-EVs on macrophages, we subjected the macrophages treated with ox-LDL to pretreat with Hy-EVs and conducted proteomic analysis (Fig. 2A). Ferroptosis is a type of programmed cell death caused by cell membrane oxidation [44,45]. GO enrichment analysis revealed that the ferroptosis pathway was significantly enriched. The key regulators of ferroptosis, GPX4/xCT, also significantly increased in the Hy-EVs group (Fig. 2A and B). GPX4/xCT maintained the cellular oxidative stress homeostasis by inhibiting lipid peroxidation of the cell membrane and regulating cystine transport [23]. It was observed that N-EVs increased the GPX4/xCT level, while Hy-EVs further enhanced their expression (Fig. 2C and D). We also detected the level of GPX4/xCT in aortic root plaque macrophage. Consistent with the in vitro experiments, immunofluorescence revealed that the co-localization levels of CD68 and GPX4/xCT in aortic plaques in the HD group decreased significantly, while those of Hy-EVs and N-EVs increased after treatment, and the expression of GPX4/xCT in macrophages of the Hy-EVs group was higher than that of N-EVs (Fig. 2 E-H). The mRNA of the mouse aorta was used to assess the mRNA level of GPX4/xCT, and the experimental results were consistent with those of immunofluorescence (Fig. 2I and J).Fig. 2Proteomics revealed that Hy-EVs improve macrophage ferroptosis caused by ox-LDL and high-fat feeding. Bone marrow-derived macrophages (BMDMs) were pretreated with N-EVs and Hy-EVs for 1 h, then stimulated with 100 μg/ml ox-LDL for 48 h. (A, B) Proteomics was employed to analyze the differential proteins in macrophages after treatment with ox-LDL and Hy-EVs. n = 3. (C, D) WB was used to analysis GPX4 (n = 3) and xCT (n = 3) levels. (E-H) The CD68 and GPX4/xCT immunofluorescence co-staining was used to analyze the expression of GPX4 and xCT in macrophages within the aortic root plaques. n = 10. Scale bars:100 μm. (I, J) GPX4 and xCT mRNA level in aorta tissue. n = 10. (K) MDA level in macrophages. (n = 4). (L) 4HNE content in macrophages. (n = 3). (M) Iron content in macrophage (n = 4). (N) GSH level in macrophages (n = 4). (O, P) The lipid ROS level was evaluated by flow cytometry using C11-Bodipy fluorescent probe. (n = 3). (Q) Representative transmission electron microscopy pictures of macrophages. (n = 3). Scale bars:0.5 μm. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05.Fig. 2 Proteomics revealed that Hy-EVs improve macrophage ferroptosis caused by ox-LDL and high-fat feeding. Bone marrow-derived macrophages (BMDMs) were pretreated with N-EVs and Hy-EVs for 1 h, then stimulated with 100 μg/ml ox-LDL for 48 h. (A, B) Proteomics was employed to analyze the differential proteins in macrophages after treatment with ox-LDL and Hy-EVs. n = 3. (C, D) WB was used to analysis GPX4 (n = 3) and xCT (n = 3) levels. (E-H) The CD68 and GPX4/xCT immunofluorescence co-staining was used to analyze the expression of GPX4 and xCT in macrophages within the aortic root plaques. n = 10. Scale bars:100 μm. (I, J) GPX4 and xCT mRNA level in aorta tissue. n = 10. (K) MDA level in macrophages. (n = 4). (L) 4HNE content in macrophages. (n = 3). (M) Iron content in macrophage (n = 4). (N) GSH level in macrophages (n = 4). (O, P) The lipid ROS level was evaluated by flow cytometry using C11-Bodipy fluorescent probe. (n = 3). (Q) Representative transmission electron microscopy pictures of macrophages. (n = 3). Scale bars:0.5 μm. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. As previously reported in the literature, GPX4 and xCT regulate ferroptosis by modulating cellular oxidative stress [46,47]. Therefore, we speculated whether EVs derived from EPCs after hypoxia treatment inhibited the progression of atherosclerosis by improving the macrophage ferroptosis. CCK8 and LDH assay showed that N-EVs mitigated the decrease of cell viability induced by ox-LDL, and the degree to which Hy-EVs restores cell viability was higher than that of N-EVs (Supplemental Fig. 4A and B). MDA and 4HNE are the product of lipid peroxidation in cell membranes. We observed that the MDA and 4HNE levels in the ox-LDL-treated macrophages and atherosclerotic plaque were increased obviously. Both N-EVs and Hy-EVs decreased the serum iron ions, MDA and 4HNE levels, and the decreased level in the Hy-EVs group was higher (Fig. 2L–N, Supplemental Fig. 4H–J). In addition, compared with N-EVs, Hy-EVs further increased the intracellular GSH level and SOD activity, suggesting that Hy-EVs increased the antioxidant level of macrophages in vivo and in vitro (Fig. 2O, Supplemental Figs. 4C, K, L). We further investigated the effect of EPCs-derived large EVs (larger than 200 nm) after hypoxia treatment on ferroptosis. After filtering with a 0.45 nm filter, we extracted large Hy-EVs (lHy-EVs) using the same method as for extracting small Hy-EVs (sHy-EVs), and then treated the macrophages with same concentration. Our results show that lHy-EVs have a significant effect in improving ferroptosis, but it is slightly lower than that of sHy-EVs (Supplemental Fig. 4M–P). Furthermore, we observed that Hy-EVs inhibited a higher level of lipid ROS than N-EVs (Fig. 2P). Mitochondrial damage is a key feature of ferroptosis, and we observed from scanning electron microscopy that mitochondrial ridges disappeared after ox-LDL treatment, whereas N-EVs and Hy-EVs treatments significantly reversed this alteration (Fig. 2Q). JC-1 staining demonstrated that compared with N-EVs group, Hy-EVs further improved the decrease of mitochondrial membrane localization (Supplemental Fig. 5A). Seahorse assay was performed to analysis mitochondrial respiratory function, the results demonstrated that Hy-EVs treatments have increased basal respiration, maximal respiration, ATP production and spare respiration to a much higher level than N-EVs (Supplemental Fig. 5B–F). DCFH staining showed that Hy-EVs further improved oxidative stress level in macrophage induced by ox-LDL (Supplemental Fig. 5G). Notably, it was found that ox-LDL increase the inflammatory factors level, we observed lower iron ions and inflammatory factors expression in the Hy-EVs treated group, which is consistent with previous results (Supplemental Fig. 4D–G). These data demonstrated that hypoxia treatment enhanced the protective effects of EVs derived from EPCs on macrophage ferroptosis induced by ox-LDL. Hy-EVs attenuated atherosclerosis progression by carrying ASS1 In order to explore the mechanism by which Hy-EVs improves AS, proteomics was used to analyze the differentially expressed proteins in EVs and Hy-EVs. It can be observed from the volcano map that ASS1 is most significantly upregulated in Hy-EVs (Fig. 3A). The Western blot experiments for detecting ASS1 level in N-EVs and Hy-EVs also exhibited the same results (Fig. 3B and C). After treating macrophages with Hy-EVs and EVs, ASS1 in macrophages was increased significantly in Hy-EVs group, while ox-LDL had no obvious effect on the expression of ASS1 in macrophages (Fig. 3D–F). Si-RNA was used to knockdown the ASS1 in EPCs (Fig. 3E–G), it was found that after knockdown of ASS1, the expression of ASS1 was obviously decreased in macrophages treated with Hy-EVs derived from EPCs (Fig. 3D–F). However, neither Hy-EVs nor N-EVs had a significant effect on the mRNA level of ASS1 in macrophages (Fig. 3H, I). Previous study has reported that hypoxia-inducible factor-1α (HIF-1α) plays a crucial role in enhancing the therapeutic effect of stem cell therapy under hypoxic conditions. After knocking down HIF-α in EPCs, the levels of ASS1 in Hy-EVs and EPCs significantly decreased (Supplemental Fig. 6A and B). Furthermore, compared with the levels of ASS1 in Hy-EVs, mouse circulating EVs and EVs derived from aortic plaques. ASS1 was expressed at extremely low levels in both the circulating and aortic plaque EVs, While the level of ASS1 in Hy-EVs was 50 times higher than that in serum EVs and aortic EVs (Supplemental Fig. 6C). CHIP-PCR further confirmed that hypoxic preconditioning treatment increased the binding of HIF to the promoter region of ASS1(Supplemental Fig. 6D). These results suggested that hypoxia treatment increases the expression of ASS1 in EVs of EPCs.Fig. 3Proteomics revealed that ASS1 expression was significantly increased in Hy-EVs. (A) Proteomics was used to analyze the differential proteins in N-EVs and Hy-EVs. n = 3. (B, C) Western blot was used to detected ASS1 level in N-EVs and Hy-EVs. n = 3. (D, F) N-EVs and Hy-EVs were used to treat macrophages, and the expression level of ASS1 was detected by Western blot. n = 5. (E, G) Si-RNA was used to knock down ASS1 in EPCs, and N-EVs and Hy-EVs derived from these EPCs were used to treat macrophages. n = 5. (H, I) PCR was used to analysis ASS1 mRNA level in macrophage. n = 5. Results were expressed as mean ± SD. All the data were in accordance with the normal distribution. Comparisons of parameters were performed with Student's t-test (C). Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test (F, H, G, I). ∗, p < 0.05.Fig. 3 Proteomics revealed that ASS1 expression was significantly increased in Hy-EVs. (A) Proteomics was used to analyze the differential proteins in N-EVs and Hy-EVs. n = 3. (B, C) Western blot was used to detected ASS1 level in N-EVs and Hy-EVs. n = 3. (D, F) N-EVs and Hy-EVs were used to treat macrophages, and the expression level of ASS1 was detected by Western blot. n = 5. (E, G) Si-RNA was used to knock down ASS1 in EPCs, and N-EVs and Hy-EVs derived from these EPCs were used to treat macrophages. n = 5. (H, I) PCR was used to analysis ASS1 mRNA level in macrophage. n = 5. Results were expressed as mean ± SD. All the data were in accordance with the normal distribution. Comparisons of parameters were performed with Student's t-test (C). Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test (F, H, G, I). ∗, p < 0.05. In order to further investigate whether Hy-EVs exerts a protective effect on atherosclerosis through ASS1 in vivo, EPCs with ASS1 (sh-ASS1) knocked down was constructed. Results showed that the plaque area of the artery root in the Sh-ASS1-Hy-EVs group was larger compared with Hy-EVs group. Further to confirms the role of ASS1 in the improvement of plaque progression by Hy-EVs, AAV carrying a macrophage-specific promoter (F4/80) was injected into ApoE−/− mice to overexpress ASS1 (AAV-ASS1) in macrophages (Supplemental Fig. 6E). The results demonstrated that the combination of EVs with AAV-ASS1 and Hy-EVs could further reduce the plaque area of the aortic root (Fig. 4A–C). Moreover, AAV-ASS1 significantly reversed the increase in plaque area caused by ASS1 knockout EPCs derived Hy-EVs (Sh-ASS1-Hy-EVs) (Fig. 4A–C). Masson staining revealed that Sh-ASS1-Hy-EVs treatment leaded to less collagen content than Hy-EVs groups, while AAV-ASS1 increased the content of collagen deposition in the N-EVs and Sh-ASS1-Hy-EVs groups, and there was no significant difference in the level compared with the Hy-EVs group (Fig. 4B–E). It was also observed that AAV-ASS1 reduced the aorta lipid deposition levels in the Sh-ASS1-Hy-EVs and N-EVs groups (Fig. 4D–F). The serum inflammatory factors in mice was further evaluated, compared with the mice in the Hy-EVs group, Sh-ASS1-Hy-EVs and N-EVs groups showed higher levels of IL-18 and IL-1β, lower levels of IL-4 and IL-10, while AAV-ASS1 reduced the levels of IL-18 and IL-1β in the Sh-ASS1-Hy-EVs and N-EVs groups, and the levels of IL-4 and IL-10 were also increased (Fig. 4G–J).Fig. 4Hy-EVs suppressed atherosclerosis progression via ASS1. The ApoE−/− mice were divided into 6 groups. Adeno-associated virus vectors carrying macrophage-specific promoters(F4/80) used to overexpress ASS1. (A) HE staining of aortic root plaque. n = 10. Scale bars:200 μm. (B) Masson staining of aortic root plaque. n = 10. Scale bars:200 μm. (C) Quantification of the area of HE-stained plaques. n = 10. (D) Aortic Oil Red Staining. n = 10. Scale bars:1 cm. (E) Quantification of the plaque content of Masson-stained plaques. n = 10. (F) Quantification of the area of the aortic oil red positive region. n = 10. Inflammatory factors in the plasma of different mice were detected. (G) IL-1β level. n = 10. (H) IL-18 levels. n = 10. (I) IL-4 levels. n = 10. (J) IL-10 level. n = 10. Results are expressed as mean ± SD. All the data were in accordance with the normal distribution. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. AAV-ASS1: Adeno-associated virus vectors overexpress ASS1. Sh-ASS1-Hy-EVs: EVs derived from EPCs after ASS1 knockdown and hypoxia treatment. HD: high fat diet.Fig. 4 Hy-EVs suppressed atherosclerosis progression via ASS1. The ApoE−/− mice were divided into 6 groups. Adeno-associated virus vectors carrying macrophage-specific promoters(F4/80) used to overexpress ASS1. (A) HE staining of aortic root plaque. n = 10. Scale bars:200 μm. (B) Masson staining of aortic root plaque. n = 10. Scale bars:200 μm. (C) Quantification of the area of HE-stained plaques. n = 10. (D) Aortic Oil Red Staining. n = 10. Scale bars:1 cm. (E) Quantification of the plaque content of Masson-stained plaques. n = 10. (F) Quantification of the area of the aortic oil red positive region. n = 10. Inflammatory factors in the plasma of different mice were detected. (G) IL-1β level. n = 10. (H) IL-18 levels. n = 10. (I) IL-4 levels. n = 10. (J) IL-10 level. n = 10. Results are expressed as mean ± SD. All the data were in accordance with the normal distribution. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. AAV-ASS1: Adeno-associated virus vectors overexpress ASS1. Sh-ASS1-Hy-EVs: EVs derived from EPCs after ASS1 knockdown and hypoxia treatment. HD: high fat diet. Increased ASS1 in Hy-EVs meditated the protective effect of Hy-EVs on ferroptosis In order to confirm whether ASS1 in Hy-EVs meditated the protective effect of Hy-EVs on ferroptosis, Sh-ASS1-Hy-EVs were used to treat macrophages. It was found that the protective effect on macrophage viability decreased significantly compared with Hy-EVs, while the protective effect of Sh-ASS1-Hy-EVs on macrophages viability was restored again after ASS1 overexpression (OE-ASS1) (Supplemental Fig. 7A and B). We further detected the levels of MDA and 4HNE in macrophages and artery, the results showed that the level of MDA and 4HNE significantly increased in the Sh-ASS1-Hy-EVs group, and OE-ASS1 diminished the increased levels of lipid MDA and 4HNE caused by ASS1 knockout (Fig. 5A, and Supplemental Figs. 7C, G, H). Flow cytometry and immunofluorescence shown that compared with N-EVs, Hy-EVs reduced the levels of lipid ROS and ferrous iron, while this effect was attenuated by ASS1 knocking down, and OE-ASS1 in macrophages compensated for the effect again (Fig. 5E–H). In vivo data also demonstrated that AAV-ASS1 decreased the levels of serum iron ions and artery ROS level, while also enhancing the expression and activity of GSH and SOD in EVs or Sh-ASS1-Hy-EVs treated ApoE−/− mice (Supplemental Fig. 7I–L). Moreover, it was also found that after knockdown of ASS1 in EPCs, the effect of Hy-EVs on increasing GPX4/xCT level were significantly inhibited, while overexpression of ASS1 in macrophages reversed this effect of Hy-EVs (Fig. 5B–C). Similar results were observed in ApoE−/− mice, immunofluorescence and PCR showed that the effect of hypoxic treated EVs on increasing GPX4/xCT was weakened after knocking out ASS1 in EPCs. While specific overexpression ASS1 in macrophages not only increased GPX4/xCT level in Sh-ASS1-Hy-EVs groups, but also significantly increased the levels of GPX4/xCT in the EVs treated ApoE−/− mice (Fig. 5I–L and Supplemental Fig. 7E and F). Furthermore, JC-1 staining showed that EVs and Hy-EVs from ASS1 knockdown EPCs had limited protective effect on improving mitochondrial membrane potential, ASS1 overexpression further strengthened the effect of Hy-EVs from ASS1 knockdown EPCs and EVs in improving mitochondrial membrane potential (Supplemental Fig. 8A). The respiratory function of mitochondria was also evaluated, seahorse assay suggested that the mitochondrial respiratory function of macrophages was further enhanced in Sh-ASS1-Hy-EVs group and N-EVs with ASS1 overexpression (Supplemental Fig. 8B–F). These data suggested that N-EVs derived from EPCs treated with hypoxia inhibited the macrophage ferroptosis through ASS1.Fig. 5ASS1 mediated the protective effect of Hy-EVs on preventing ferroptosis in macrophages. (A). The levels of 4HNE level were measured in macrophages by commercial kits. n = 5. (B-D) Western blot was used to analysis GPX4 and xCT protein levels. n = 5. (E, F) The lipid ROS level was evaluated by flow cytometry using C11-Bodipy fluorescent probe. n = 5. (G, H) Immunofluorescence was used to detect the intracellular level of ferrous ions. n = 5. Scale bars:20 μm. (I-L) Adeno-associated virus vectors carrying macrophage-specific promoters(F4/80) used to overexpress ASS1. The CD68 and GPX4/xCT immunofluorescence co-staining was used to analyze the expression of GPX4 and xCT in macrophages within the aortic root plaques. n = 10. Scale bars:100 μm. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. AAV-ASS1: Adeno-associated virus vectors overexpress ASS1. Sh-ASS1-Hy-EVs: EVs derived from EPCs after ASS1 knockdown and hypoxia treatment. HD: high fat diet.Fig. 5 ASS1 mediated the protective effect of Hy-EVs on preventing ferroptosis in macrophages. (A). The levels of 4HNE level were measured in macrophages by commercial kits. n = 5. (B-D) Western blot was used to analysis GPX4 and xCT protein levels. n = 5. (E, F) The lipid ROS level was evaluated by flow cytometry using C11-Bodipy fluorescent probe. n = 5. (G, H) Immunofluorescence was used to detect the intracellular level of ferrous ions. n = 5. Scale bars:20 μm. (I-L) Adeno-associated virus vectors carrying macrophage-specific promoters(F4/80) used to overexpress ASS1. The CD68 and GPX4/xCT immunofluorescence co-staining was used to analyze the expression of GPX4 and xCT in macrophages within the aortic root plaques. n = 10. Scale bars:100 μm. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. AAV-ASS1: Adeno-associated virus vectors overexpress ASS1. Sh-ASS1-Hy-EVs: EVs derived from EPCs after ASS1 knockdown and hypoxia treatment. HD: high fat diet. Increased ASS1 in Hy-EV suppressed macrophage ferroptosis via generating arginine ASS1 is the rate-limiting enzyme in the urea cycle that converts l-citrulline to l-arginine, which is considered a semiessential amino acid as it can be synthesized within a cell as well as obtained from external environment [48]. Depletion of circulating l-arginine has been shown to halt the growth of cell, while exogenous arginine has been proven to exert the effect of improving oxidative stress [33,49]. We hypothesized that ASS1 in hypoxic preconditioned EPCs derived EVs inhibits ferroptosis by improving oxidative stress via increasing the synthesis of arginine in macrophages. Previous study has demonstrated that ornithine and arginine are the main products of ASS1 in the uric acid cycle [35]. Therefore, targeted metabolomics was employed to detect the levels of ornithine and arginine in macrophages. Results exhibited that EVs treatment did not increase the levels of ornithine and arginine, but the Hy-EVs group showed a significant increase of ornithine and arginine, while the effect was attenuated in Hy-EVs from ASS1 knockdown EPCs, the level of ornithine and arginine in the macrophages was further increased after EVs and Sh-ASS1-Hy-EVs group with ASS1 addition administration (Supplemental Fig. 9A and B). According to previous study, aspartate and citrulline are synthesized into arginine under the action of ASS1 [35]. Stable isotope tracing was used to confirm that Hy-EVs synthesize arginine through ASS1. Following the addition of N15-labeled aspartate to the culture medium, the detection of N15-positive arginine indicated that Hy-EVs significantly enhanced the incorporation of the label into arginine (Supplemental Fig. 9C and D). To further confirm the effect of arginine in the protective effect of Hy-EVs on macrophage ferroptosis, exogenous arginine was administered to macrophages. CCK8 assay demonstrated that arginine supplementation strengthened the effect of N-EVs in improving the decline of macrophage viability induced by ox-LDL, however, exogenous arginine had no significant effect on the viability of macrophages in the Hy-EVs group (Supplemental Fig. 9E). MDA level in lysates and GPX4/xCT mRNA levels were also evaluated, the results showed that arginine supplementation decreased the MDA level, increased the GPX4/xCT mRNA levels in N-EVs groups, while exhibited no obvious effects in Hy-EVs group (Supplemental Figs. 9F, G, H). ADI-PEG20 is a kind of PEGylated arginine deiminase consuming arginine by converting it into citrulline [50,51]. Conversely, it was found that ADI-PEG20 reversed the improvement effects of arginine supplementation and Hy-EVs on cell viability (Supplemental Fig. 9I). Furthermore, we also observed that ADI-PEG20 increased the MDA levels and decreased the mRNA levels of GPX4/xCT in the arginine supplementation and Hy-EVs groups (Supplemental Figs. 9J, K, L). Similarly, ADI-PEG20 significantly inhibited the effect of Hy-EVs in improving ferroptosis in macrophages (Supplemental Figs. 9M, N, O). These data illustrated that ASS1 in hypoxic preconditioned EPCs derived EVs protect macrophage from ferroptosis via enhancing arginine generation. Arginine synthesized by ASS1 inhibited ferroptosis via inhibiting GCN2, but not mTOR pathway The level of arginine is precisely regulated during cell development, and its homeostasis is crucial for maintaining cellular oxidative stress and survival. General control nonderepressible 2 (GCN2) and mTOR are sensitive elements that control the concentration of arginine. Cellular arginine deficiency activates GCN2 via the accumulation of uncharged tRNAs. Conversely, under conditions of arginine sufficiency, mTOR signaling is activated [33,52,53]. We hypothesized that arginine synthesized by ASS1 in Hy-EVs exerts ferroptosis protection by regulating the activity of GCN2 or mTOR. Rapamycin, a highly selective mTOR inhibitor, was used to pretreated macrophages. It was found that rapamycin has no obvious effect on the cell viability in Hy-EVs group (Supplemental Fig. 10A). The MDA and 4HNE levels in macrophage lysate were also evaluated, the results showed that rapamycin did not influence the MDA and 4HNE levels in Hy-EVs groups (Supplemental Fig. 10B and E). Furthermore, compared with Hy-EVs group, the GPX4/xCT level in rapamycin treated groups exhibited no obvious changes (Supplemental Figs. 10C, D, F, G, H). On the contrary, halofuginone (Hal), a highly selective GCN2 agonist, significantly inhibited the effect of Hy-EVs in improving the decline of macrophage viability induced by ox-LDL (Supplemental Fig. 11A). It was found that Hal increased the MDA and 4HNE levels in Hy-EVs group (Supplemental Fig. 11B and E). Similarly, compared with Hy-EVs groups, the mRNA and protein levels of GPX4/xCT were also decreased in Hal treated group (Supplemental Figs. 11C, D, F, G, H). Moreover, we also found that GCN2iB, an inhibitor of GCN2, could directly improve the decreased cell viability, increased 4HNE/MDA levels and decreased expression of GPX4/xCT caused by ox-LDL (Supplemental Fig. 11A–H). Previous studies have confirmed that the activation of GCN2 lead to the phosphorylation of the Thr899 site [33,52]. Therefore, we detected the phosphorylation of GCN2Thr899 in different groups. The immunoblotting results suggested that GCN2iB and Hy-EVs significantly reduced the phosphorylation level of GCN2Thr899, while Hal increased the phosphorylation level of GCN2Thr899 (Supplemental Fig. 12A and B). The above results suggested that arginine synthesized by ASS1 suppressed macrophage ferroptosis caused by ox-LDL by inhibiting the activity of GCN2. Hy-EVs inhibited atherosclerosis progression and macrophage ferroptosis via inhibiting GCN2 In order to further confirm in vivo that Hy-EVs inhibit plaque progression by suppressing GCN2, AAV carrying macrophage-specific promoters (F4/80) were used to overexpress GCN2 in the macrophages of ApoE−/− mice (Supplemental Fig. 6F). HE staining showed that AAV-GCN2 increased the plaque area in aortic root compared with Hy-EVs groups (Fig. 6A–C). Masson and oil red staining demonstrated that AAV-GCN2 decreased the collagen content and increased lipid deposition (Fig. 6B–D, E, F). Moreover, AAV-GCN2 reversed the effect of anti-inflammatory response of Hy-EVs, it is manifested as AAV-GCN2 increasing the expression of IL-18 and IL-1β and reducing the levels of IL-4 and IL-10 (Fig. 6G–J). Furthermore, immunofluorescence revealed that the expression of GPX4/xCT in macrophages in the AAV-GCN2 group was significantly lower than that in the Hy-EVs group (Fig. 6K–N). The same results were also observed for the detection of mRNA levels of GPX4 and xCT in the aorta (Fig. 6O and P). In order to further confirm that AAV-GCN2 weakens the protective effect of Hy-EVs on ferroptosis, the contents of serum iron ions, GSH, MDA, 4HNE and the activity of SOD in the aortic lysate were evaluated. The results showed that AAV-GCN2 increased serum iron ions/MDA/4HNE/ROS level and decreased the GSH content and SOD activity again in Hy-EVs groups (Fig. 6Q–V). However, AAV-GCN2 did not influence the arginine level (Fig. 6W). These data suggested that Hy-EVs inhibited atherosclerosis progression and macrophage ferroptosis via inhibiting GCN2 activation.Fig. 6Hy-EVs suppressed atherosclerosis progression via inhibiting GCN2. Adeno-associated virus vectors carrying macrophage-specific promoters(F4/80) used to overexpress GCN2. (A) HE staining of aortic root plaque. n = 10. Scale bars:200 μm. (B) Masson staining of aortic root plaque. n = 10. Scale bars:200 μm. (C) Aortic Oil Red Staining. n = 10. Scale bars:1 cm. (D) Quantification of the area of HE-stained plaques. n = 10. (E) Quantification of the plaque content of Masson-stained plaques. n = 10. (F) Quantification of the area of the aortic oil red positive region. n = 10. Inflammatory factors in the plasma of different mice were detected. (G) IL-1β level. n = 10. (H) IL-18 levels. n = 10. (I) IL-4 levels. n = 10. (J) IL-10 level. n = 10. (K, L, M, N) The CD68 and GPX4/xCT immunofluorescence co-staining was used to analyze the expression of GPX4 and xCT in macrophages within the aortic root plaques. n = 10. Scale bars:100 μm. (O, P) GPX4 and xCT mRNA level in aorta tissue. n = 10. (Q). The level of MDA in aorta lysates. n = 10. (R) The level of 4HNE in aorta lysates. n = 10. (S) The levels of serum iron. n = 10. (T, U) The SOD activity and GSH level in aorta lysates. n = 10. (V) The ROS level in aorta lysates. n = 10. (W) The level of arginine in aorta lysates. n = 10. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. ns: no significant statistical significance.Fig. 6 Hy-EVs suppressed atherosclerosis progression via inhibiting GCN2. Adeno-associated virus vectors carrying macrophage-specific promoters(F4/80) used to overexpress GCN2. (A) HE staining of aortic root plaque. n = 10. Scale bars:200 μm. (B) Masson staining of aortic root plaque. n = 10. Scale bars:200 μm. (C) Aortic Oil Red Staining. n = 10. Scale bars:1 cm. (D) Quantification of the area of HE-stained plaques. n = 10. (E) Quantification of the plaque content of Masson-stained plaques. n = 10. (F) Quantification of the area of the aortic oil red positive region. n = 10. Inflammatory factors in the plasma of different mice were detected. (G) IL-1β level. n = 10. (H) IL-18 levels. n = 10. (I) IL-4 levels. n = 10. (J) IL-10 level. n = 10. (K, L, M, N) The CD68 and GPX4/xCT immunofluorescence co-staining was used to analyze the expression of GPX4 and xCT in macrophages within the aortic root plaques. n = 10. Scale bars:100 μm. (O, P) GPX4 and xCT mRNA level in aorta tissue. n = 10. (Q). The level of MDA in aorta lysates. n = 10. (R) The level of 4HNE in aorta lysates. n = 10. (S) The levels of serum iron. n = 10. (T, U) The SOD activity and GSH level in aorta lysates. n = 10. (V) The ROS level in aorta lysates. n = 10. (W) The level of arginine in aorta lysates. n = 10. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. ns: no significant statistical significance. Arginine synthesized by ASS1 inhibited ferroptosis by activating NRF2 via suppressing GCN2/NRF2 interaction To further explore the mechanism by which GCN2 regulates ferroptosis in macrophages, we analyzed the proteins bound to GCN2 using LC/MS in GCN2 overexpressed macrophages. It was found that nuclear factor erythroid 2-related factor 2 (NRF2) was the binding protein with the highest score (Fig. 7A). Immunofluorescence and co-immunoprecipitation confirmed the directly binding of GCN2 and NRF2 in macrophages (Fig. 7B and C). NRF2 is a transcription factor that is widely present in the cytoplasm and cell nucleus. NRF2 regulates cellular oxidative stress balance by transcribing antioxidant proteins such as GPX4. When intracellular oxidative stress occurs, NRF2 undergoes nuclear translocation and initiates the transcription of antioxidant proteins [24,54]. Previous studies have found that regulating the nuclear translocation by binding to NRF2 can control intracellular redox reactions [[55], [56], [57]]. We further analyzed whether the activity of GCN2 affected GCN2/NRF2 interaction. The PLA results revealed that inhibiting the activity of GCN2 by Hy-EVs or GCN2iB reduced the effect of GCN2/NRF2 interaction, while activating GCN2 increased the binding of GCN2/NRF2 interaction (Fig. 7D–F. Supplemental Fig. 12C and D). Therefore, we speculated that inhibition of GCN2 activity reduced the binding of GCN2/NRF2, thereby promoting the nuclear translocation of NRF2. Immunofluorescence showed that inhibiting the activity of GCN2 increased the expression of NRF2 in the nucleus, while the level of NRF2 in the nucleus of the GCN2 activation group decreased significantly (Fig. 7E–G). Furthermore, CHIP-PCR analysis also found that the transcriptional activity of GPX4/xCT in the GCN2 activity inhibition group increased, while the transcriptional activity of GPX4/xCT in the GCN2 activation group decreased significantly (Fig. 7H and I). These results suggested that Arg reduced GCN2 and NRF2 interaction and promoted NRF2 nuclear translocation.Fig. 7Hy-EVs enhanced NRF2 activation by suppressing GCN2/NRF2 interaction via generating arginine. (A) GCN2 was purified from macrophages using GCN2 antibody, IgG was used as a negative control, and protein profiling was performed to analyze proteins in lysates. (B) Immunofluorescence co-localization of GCN2/NRF2 in macrophages. Scale bars:20 μm. n = 3. (C) GCN2/NRF2 immunoprecipitation in macrophages. n = 3. (D, F) GCN2 inhibitor (GCN2iB) and GCN2 agonist (Hal) was used to pre-treated macrophages, respectively. Immunofluorescence co-localization of GCN2/NRF2 in macrophages. n = 4. Scale bars:20 μm. (E, G) GCN2 inhibitor (GCN2iB) and GCN2 agonist (Hal) was used to pre-treated macrophages, respectively. Immunofluorescence was used to analyze the nuclear translocation of NRF2. n = 5. Scale bars:20 μm. (H, I) The enrichment ability of NRF2 on GPX4 and xCT promoter using ChIP-qPCR. n = 5. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05.Fig. 7 Hy-EVs enhanced NRF2 activation by suppressing GCN2/NRF2 interaction via generating arginine. (A) GCN2 was purified from macrophages using GCN2 antibody, IgG was used as a negative control, and protein profiling was performed to analyze proteins in lysates. (B) Immunofluorescence co-localization of GCN2/NRF2 in macrophages. Scale bars:20 μm. n = 3. (C) GCN2/NRF2 immunoprecipitation in macrophages. n = 3. (D, F) GCN2 inhibitor (GCN2iB) and GCN2 agonist (Hal) was used to pre-treated macrophages, respectively. Immunofluorescence co-localization of GCN2/NRF2 in macrophages. n = 4. Scale bars:20 μm. (E, G) GCN2 inhibitor (GCN2iB) and GCN2 agonist (Hal) was used to pre-treated macrophages, respectively. Immunofluorescence was used to analyze the nuclear translocation of NRF2. n = 5. Scale bars:20 μm. (H, I) The enrichment ability of NRF2 on GPX4 and xCT promoter using ChIP-qPCR. n = 5. All the data were in accordance with the normal distribution. Results were expressed as mean ± SD. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. In order to further analyze the mechanism by which arginine regulated the GCN2/NRF2 interaction, molecular dynamics were employed to further analyze the stability of GCN2/NRF2 interaction in the presence of arginine. It can be observed that NRF2 bind to GCN2 (Fig. 8A and B), consistent with the results of in vitro experiments. However, the binding of GCN2 and NRF2 altered in the presence of arginine, it was also observed a large number of binding sites for Arg and GCN2 (Fig. 8C and D). Furthermore, by analyzing the root-mean-square deviation (RMSD), it can be found that when NRF2 exist alone in the system, the RMSD is approximately 18 Å, while when the protein NRF2 and GCN2 complex exist in the system, the RMSD is approximately 19 Å, no significant difference was observed. However, when protein NRF2, GCN2 and Arg exist in the system, the RMSD is approximately 25 Å (Fig. 8E). These data suggested that adding Arg to the system increased the instability of GCN2/NRF2 and decreased the binding ability. This phenomenon was also supported in the protein solubility accessibility aspect. After adding Arg to the system, the protein solvent accessible surface area (SASA) value significantly increased, and the stability of GCN2/NRF2 decreased (Fig. 8F). The root means square fluctuation (RMSF) analysis further revealed that the regions that significantly affected by NRF2 were amino acid segments of 100-200, 250-300, 350-500 (Fig. 8G).Fig. 8Molecular dynamics revealed that Arg inhibits the binding of GCN2 and NRF2. (A) Molecular docking of GCN2 and NRF2. (B) The residues detail of the interaction between GCN2 and NRF2. (C) Molecular docking of Arg, GCN2 and NRF2. (D) The residues detail of the interaction between Arg, GCN2 and NRF2. (E) Root-mean-square deviation analysis of NRF2, NRF2/GCN2 and Arg/NRF2/GCN2. (F) solvent accessible surface area analysis of NRF2, NRF2/GCN2 and Arg/NRF2/GCN2. (G) root mean square fluctuation analysis of NRF2, NRF2/GCN2 and Arg/NRF2/GCN2.Fig. 8 Molecular dynamics revealed that Arg inhibits the binding of GCN2 and NRF2. (A) Molecular docking of GCN2 and NRF2. (B) The residues detail of the interaction between GCN2 and NRF2. (C) Molecular docking of Arg, GCN2 and NRF2. (D) The residues detail of the interaction between Arg, GCN2 and NRF2. (E) Root-mean-square deviation analysis of NRF2, NRF2/GCN2 and Arg/NRF2/GCN2. (F) solvent accessible surface area analysis of NRF2, NRF2/GCN2 and Arg/NRF2/GCN2. (G) root mean square fluctuation analysis of NRF2, NRF2/GCN2 and Arg/NRF2/GCN2. To further confirm that Hy-EV inhibited ferroptosis by inhibiting the activity of GCN2 via activating NRF2, ML385, a specific NRF2 inhibitor, was used to pretreat macrophages. It was observed that ML385 suppressed the NRF2 nuclear translocation in GCN2 inhibition induced by Hy-EVs and GCN2iB (Fig. 9A–D). PCR and Western blot demonstrated that ML385 diminished the increased level of GPX4 and xCT levels in Hy-EVs and GCN2iB groups (Fig. 9C–F). Moreover, ML385 diminished the effect of Hy-EVs and GCN2iB on improving the cell viability, decreasing MDA/4HNE content and increasing GSH level (Fig. 9G–J). Lipid ROS and ferrous iron levels were also evaluated, the results demonstrated that the lipid ROS and ferrous iron levels were increased again in ML385 pretreated in Hy-EVs and GCN2iB groups (Fig. 9K–O). These results illustrated arginine synthesized by ASS1 inhibited ferroptosis by activating NRF2 via suppressing GCN2/NRF2 interaction.Fig. 9NRF2 inhibition by ML385 abolished the protective effect of Hy-EVs on macrophage ferroptosis. ML385, a specific NRF2 inhibitor. GCN2iB, a specific GCN2 inhibitor. (A, B) Immunofluorescence was used to analyze the nuclear translocation of NRF2. n = 5. Scale bars:20 μm. (C, D) Western blot was used to detected the NRF2 level in nuclear and GPX4/xCT level in cytoplasm. n = 3. (E, F) GPX4 and xCT mRNA level in macrophages. n = 5. (G) CCK-8 assay was used to analysis cell viability. n = 5. (H). The levels of MDA in macrophages detected by commercial kits. n = 5. (I) The level of GSH in macrophages. n = 5. (J) The level of 4HNE in macrophages. n = 5. (K, M) The lipid ROS level was evaluated by flow cytometry using C11-Bodipy fluorescent probe. n = 5. Scale bars:100 μm. (L, N) Ferrous irons staining. n = 5. (O) The iron ion level in macrophages. n = 5. Results were expressed as mean ± SD. All the data were in accordance with the normal distribution. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05.Fig. 9 NRF2 inhibition by ML385 abolished the protective effect of Hy-EVs on macrophage ferroptosis. ML385, a specific NRF2 inhibitor. GCN2iB, a specific GCN2 inhibitor. (A, B) Immunofluorescence was used to analyze the nuclear translocation of NRF2. n = 5. Scale bars:20 μm. (C, D) Western blot was used to detected the NRF2 level in nuclear and GPX4/xCT level in cytoplasm. n = 3. (E, F) GPX4 and xCT mRNA level in macrophages. n = 5. (G) CCK-8 assay was used to analysis cell viability. n = 5. (H). The levels of MDA in macrophages detected by commercial kits. n = 5. (I) The level of GSH in macrophages. n = 5. (J) The level of 4HNE in macrophages. n = 5. (K, M) The lipid ROS level was evaluated by flow cytometry using C11-Bodipy fluorescent probe. n = 5. Scale bars:100 μm. (L, N) Ferrous irons staining. n = 5. (O) The iron ion level in macrophages. n = 5. Results were expressed as mean ± SD. All the data were in accordance with the normal distribution. Comparisons of parameters were performed with one-way ANOVA with Bonferroni post hoc test. ∗, p < 0.05. Discussion Stem cell therapy has been regarded as a highly promising treatment strategy for cardiovascular diseases. However, the role and potential mechanism of EVs derived from stem cells in atherosclerosis remain unclear. In the present study, we have confirmed that hypoxia preconditioning enhances the atherosclerotic protection effect of EVs derived from EPCs. Furthermore, we found that Hy-EVs attenuated atherosclerosis progression by inhibiting macrophage ferroptosis via carrying ASS1. Mechanistically, ASS1 in Hy-EVs enhanced NRF2 nuclear translocation and improved oxidative stress by increasing arginine synthesis. The increased arginine in Hy-EVs treated macrophage activated NRF2 by suppressing GCN2 via reducing the interaction between GCN2 and NRF2. The current results emphasized that hypoxia preconditioning further enhanced the protective effect of EVs derived from EPCs on atherosclerosis. Our research has proposed a new potential therapeutic approach for atherosclerosis, and provided new evidence for the clinical translation of hypoxia-preconditioned EPCs-derived EVs in the treatment of atherosclerosis. Increasing evidence suggested that stem cells have protective effect on cardiovascular diseases, including myocardial infarction and ischemia-reperfusion injury, however, there are limited studies on the protective effects of EVs derived from stem cells on cardiovascular diseases. It was reported that miRNA-96 derived from bone marrow mesenchymal stem cells attenuated doxorubicin-induced myocardial toxicity [58]. Several studies also found that BMSCs attenuated atherosclerosis progression via secreting EVs [59,60]. Moreover, previous studies have shown that hypoxic preconditioning further enhanced the biological activity of EVs derived from stem cells [31,61,62]. As a type of stem cell, studies have shown that compared with BMSCs, EPCs have stronger targeting ability and more efficient repair effects on blood vessels [63]. However, it is not clear whether hypoxia treatment can enhance the protective effect of EVs derived from EPCs on the progression of atherosclerosis. In the current study, we compared the effects of Hy-EVs and N-EVs on the progression of atherosclerosis in high-fat-fed ApoE−/− mice. The pathological results showed that after hypoxia treatment, compared with N-EVs, Hy-EVs further reduced the plaque area, increased collagen fiber deposition, decreased lipid area and improved inflammatory factors. In vivo and in vitro studies further confirmed that Hy-EVs suppressed atherosclerosis development via inhibiting macrophage ferroptosis. These results strongly confirmed that compared with N-EVs, Hy-EVs not only further inhibited the progression of plaques but also enhanced plaques stability. In the present study, we only analyzed the differentially expressed proteins between the Hy-EVs and ox-LDL groups using proteomics, without further examining differences between Hy-EVs and N-EVs. This limitation may hinder our understanding of the protective mechanism of Hy-EVs in atherosclerosis. However, in the current manuscript, we find that N-EVs enhance macrophage ferroptosis, while hypoxia further amplifies the protective effect of EPC-derived EVs against ferroptosis. Recent studies have clarified the benefits of arginine for cardiovascular diseases. As a donor of NO, arginine enhances arginine production and exerts vascular protective effects by improving endothelial dilation function [64,65]. In ApoE−/− mice, Sirt3 overexpression inhibits plaque progression by increasing the synthesis of arginine and improving endothelial cell dysfunction and inflammatory responses [66]. However, it has been found that arginine is not only dependent on endothelial cells for its role in improving vascular function. Katrin et al. [65], found that exogenous arginine supplementation reduced arterial plaque area by reducing CD4-positive T cells and inhibiting the inflammatory response by the specific mechanism of inhibiting the action of muscle protein chain 9 to reduce T-cell motility. We observed that the majority of PKH67-labeled EVs were co-localized with CD68-positive macrophages. Endothelial cells, smooth muscle cells and macrophages are the main cells in AS plaques. Phagocytosis and internalization of foreign tissues are the main functions of macrophages, which may be the potential reason for macrophages to engulf more Hy-EVs. Macrophages can efficiently absorb EVs through various pathways such as phagocytosis and receptor-mediated endocytosis [67]. However, we also observed that endothelial cells and smooth muscle cells engulfed a portion of EVs. Our in vitro experiments further confirmed that Hy-EVs ameliorated endothelial dysfunction and smooth muscle cell proliferation induced by ox-LDL. This suggests that Hy-EVs also inhibit plaque progression by improving endothelial and smooth muscle cell function. We believe that EVs derived from EPCs exert a greater influence on macrophages than on endothelial and smooth muscle cells, since macrophages phagocytose the majority of EVs. We found that the level of arginine significantly increased in macrophages treated with Hy-EVs. Proteomics further revealed that Hy-EVs increased the expression of ferroptosis regulatory proteins such as GPX4 and xCT, and inhibited ferroptosis in macrophages, these effects were weakened after ASS1 knockout or ADI-PEG20 induced arginine consumption, while exogenous arginine or supplementation with ASS1 once again restored the protective effect against ferroptosis. These results suggested that ASS1-mediated arginine synthesis in Hy-EVs slows plaque progression in ApoE−/− mice by inhibiting macrophage ferroptosis. However, in the study, we did not directly overexpress ASS1 in the mice fed with high-fat diet, which prevented us from establishing a direct link between ASS1 and atherosclerosis. Furthermore, we confirmed that depletion of arginine via ADI-PEG20 suppresses the protective effects of Hy-EVs, while supplementation with arginine enhances the ability of N-EVs to inhibit ferroptosis. These data highlight arginine as a key factor in Hy-EVs' suppression of ox-LDL-induced macrophage ferroptosis. However, we lack in vivo experiments with arginine supplementation following ASS1 knockdown, which could further confirm arginine's pivotal role in Hy-EVs' improvement of plaque progression. We also investigated the origin of ASS1 after hypoxia. HIF-1α is a well-known transcription factor that regulates gene transcription to resist the hypoxic environment. Previous studies have confirmed that HIF-1α binds to the ASS1 promoter to regulate the expression of ASS1 [68]. CHIP experiment further confirmed that hypoxia in EPCs increased the binding of HIF-1α to the ASS1 promoter region in the present study. Future research should focus on the specific mechanism by which HIF-1α regulates the expression of ASS1. Previous study reports that the secretion of stem cell EVs induced by hypoxia is regulated in a manner dependent on nSmase2 [69]. Current data shown Hypoxic treatment not only increases the secretion of EVs by EPCs, but also raises the levels of ASS1 in both EPCs and Hy-EVs. Furthermore, under hypoxic conditions, EPCs improve the local inflammatory response by secreting ASS1 to synthesize arginine, thereby resisting the hypoxic stress. These data indicate that the ASS1 secretion by EPCs is selective rather than passive. Our research has discovered a new mechanism by which arginine protected vascular function, further enhancing our understanding of the cardiovascular protective effect of arginine. GCN2 and mTOR are the main sensors of intracellular arginine, and the mTOR pathway is activated when arginine is in excess, whereas GCN2 activity is inhibited in arginine deficiency [49,70,71]. Previous studies reported arginine metabolism disorder influenced cell fate through regulating GCN2 and mTOR. Arginine stripping triggered cell cycle arrest in cancer cell by activating GCN2, which was further exacerbated by inhibition of GCN2 [33,72]. In the present study, we observed that the activity of GCN2 was inhibited in the Hy-EVs group, potentially due to increased macrophage arginine synthesis by ASS1 after Hy-EVs treatment. It was further found that activation of GCN2 by AAV-GCN2 significantly reversed the protective effect of Hy-EVs on macrophage ferroptosis, while direct inhibition of GCN2 by GCN2iB exerted the same protective effect on macrophage ferroptosis. Our in vivo data also showed that AAV-GCN2 repressed the protective effect of Hy-EVs on plaque progression and macrophage ferroptosis in ApoE−/− mice. However, an inhibitor of the mTOR pathway, rapamycin, had no significant effect on the improvement of macrophage ferroptosis in Hy-EVs. These results suggested that increased arginine synthesis caused by ASS1 in Hy-EVs inhibited macrophage ferroptosis through inhibition of GCN2 rather than through the mTOR pathway. These findings further supported the importance of de novo arginine biosynthesis for inflammatory response and vascular homeostasis, and the critical role of GCN2 in mediating plaque progression and macrophage ferroptosis in Hy-EVs. We also observed the increase of ornithine in Hy-EVs treated macrophages. The ASS1-mediated synthesis of arginine is followed by its metabolism via arginase within the urea cycle, leading to the production of ornithine. In line with this metabolic pathway, ornithine levels tend to mirror those of arginine [73]. Previous studies have shown that as an endogenous amino acid, ornithine also has the effect of reducing reactive oxygen species. Therefore, we speculate that the ornithine produced after arginine metabolism may play a role in protecting ferroptosis by reducing reactive oxygen species [74,75]. There was evidence that GCN2 regulated NRF2 activation. In high glucose-induced retinal pigment epithelial cells injury, knockdown of GCN2 improved oxidative stress and apoptosis by activating NRF2 activity [76]. Furthermore, a study based on the high-throughput chemical library screening found that Hal, an agonist of GCN2, which improved the drug resistance of tumor cells by inhibiting the activity of NRF2. Even so, the mechanism by which GCN2 regulates the NRF2 pathway remains unclear. In our study, LC/MS analysis of purified GCN2 lysates from macrophages revealed that GCN2 binds to NRF2. Further mechanistic studies suggested that GCN2 dissociated from NRF2 in the presence of sufficient arginine. Studies on the subcellular localization of GCN2 confirmed that GCN2 is localizes in the cytoplasm [77]. We hypothesized whether GCN2 inhibits the translocation of NRF2 to the cytosol by binding it in the cytosol. In contrast, arginine inhibited GCN2 activity and reduced the interaction of GCN2 and NRF2, which in turn promoted NRF2 entry into the nucleus and initiated antioxidant gene transcription. Subsequent experiments confirmed our speculation that in Hy-EVs-treated macrophages, GCN2/NRF2 binding was decreased, NRF2 level was increased in the nucleus and NRF2 binding to the GPX4/xCT promoter was increased. These effects were suppressed by ASS1 knockdown and GCN2 agonists and exogenous ASS1 or arginine administration restored the above effects again. Meanwhile, molecular dynamics simulation validation further verified our speculations. Similarly, Yuan et al. [78]., found that knockdown of GCN2 ameliorated oxidative stress and inflammatory responses in hepatocytes by increasing NRF2 activity in a mouse model of fatty liver. This study also identified increased binding of GCN2 and NRF2 in high-fat-induced hepatocytes, and disruption of GCN2 and NRF2 binding had a significant ameliorative effect on hepatic lipid deposition in mice. Thus, our results clearly described the specific mechanism by which arginine ameliorated oxidative stress and ferroptosis. Study limitations: There were some limitations in our study. Firstly, we found a protective effect against atherosclerosis in ApoE−/− mice by intravenous administration, but Hy-EVs lacked vascular targeting. Therefore, future studies should focus on the preparation of engineering Hy-EVs with vascular targeting capability. Secondly, regarding the dynamic distribution of Hy-EVs in vivo, we have only observed a single time point. Future research should focus on analyzing the distribution of Hy-EVs at different time points. Moreover, we only focused on the protective effects of small EVs derived from EPCs on AS and macrophage ferroptosis, and discovered the protective effect of large EVs on macrophage ferroptosis induced by ox-LDL. However, it remains unclear which of the small EVs and the larger ones offers a stronger protective effect against the progression of AS. Thirdly, there is a lack of further confirmation of the clinical translational value of Hy-EVs in improving AS at the large animal level. Conclusion Taken together, our study found that Hy-EVs further attenuated atherosclerosis progression by carrying ASS1 to enhance arginine generation, the mechanistic data clearly showed the increased arginine derived by ASS1 inhibited macrophage ferroptosis and atherosclerosis by promoting NRF2 nuclear translocation via reducing GCN2/NRF2 interaction. Therefore, this study elucidates the protective effects and the underlying mechanism of Hy-EVs against atherosclerosis, providing compelling evidence that pave the way for the translational application of EPCs-derived EVs in managing atherosclerotic plaques. Source of funding This work was supported by National Natural Science Foundation of China (82425030, U24A20648, 82370464, 820720031 to H.J. 82500554 to L.X. and 82421214 to G.Y.), the National Key R&D Program of China (2023YFC3043504 to H.J.), Key Research and Development Program of Heilongjiang Province (22022ZX06C07 to H.J.), Heilongjiang Graduate Excellence Funding Program (LJYXL2024-018), Postdoctoral Research Funding of Heilongjiang Province (LBH-Z24219 to X.L.), Postdoctoral Innovation Talent Support Program (BX20250260 to X.L.), Heilongjiang Province Chunyan Young Talents Support Program (Grant No. CYQN24004), Fund of Key Laboratory of Myocardial Ischemia, Ministry of Education (KF202514, KF202410), China Postdoctoral Foundation General Project (2025M772156 to X.L.), Natural Science Foundation of Heilongjiang Province (JJ2025BS0050 to Y.C.), The State Key Laboratory of Precision Welding &Joining of Materials and Structures (MSWJ-24M15), Harbin City Science and Technology Innovation Talent Project (2023HBRCCG001) and Heilongjiang Province Natural Science Foundation (LH2022H041). CRediT authorship contribution statement Xing Luo: Investigation, Methodology. Xinxin Zhu: Data curation, Formal analysis, Resources. Yuwu Chen: Formal analysis, Funding acquisition, Investigation, Resources. Yiyuan Guo: Validation, Visualization. Biyi Xu: Methodology, Project administration, Resources. Junke Mou: Investigation, Methodology. Mengyang Wang: Resources, Software. Qishuo Gu: Data curation, Validation, Visualization. Qianhui Sun: Data curation, Investigation, Software. Ying Lv: Formal analysis, Investigation. Xingtao Huang: Project administration, Resources. Ming Zeng: Formal analysis, Methodology. Boling Yi: Data curation, Formal analysis, Investigation. Xiaolin Liu: Validation, Visualization. Xiaoxuan Bai: Data curation, Formal analysis, Methodology. Cheng Fang: Investigation, Validation. Ji Li: Methodology, Visualization. Bo Yu: Methodology, Resources, Validation. Haibo Jia: Funding acquisition, Methodology, Visualization. Declaration of competing interest No.