Ferroptosis-dependent small extracellular vesicles ULK1 enhances mitophagy and suppresses breast cancer migration.

Introduction
Breast cancer exhibits a unique epidemiological profile and marked interpatient heterogeneity, and it continues to be a major contributor to cancer-related deaths among women globally [1]. Tumor metastasis is a major contributor to the therapeutic resistance and high mortality associated with breast cancer. Approximately one-third of patients with breast cancer eventually develop distant metastases beyond regional lymph nodes, and once distant metastasis occurs, the five-year survival rate drops dramatically to approximately 23% [2]. Breast cancer often spreads to organs such as the bone, lungs, liver, and brain. Among these, breast cancer liver metastasis is particularly aggressive, with a median survival time of only 4 to 8 months in the absence of timely intervention [3]. Metastasis is a complex, multistep process that depends on dynamic interactions between tumor cells, the tumor microenvironment, and stromal components at distant organs, which together contribute to the formation of the pre-metastatic niche. Emerging evidence indicates that small extracellular vesicles (sEVs) serve as critical mediators of intercellular communication during the establishment of the pre-metastatic niche in breast cancer[4, 5]. sEVs are membrane-bound vesicles typically ranging from 30 to 200 nm in diameter released by diverse cell types. They contain bioactive components such as lipids, proteins, nucleic acids, and cytokines, which can be delivered to recipient cells and influence their biological activity [6]. Cancer cells release substantially more sEVs than their normal counterparts [7], which transmit oncogenic cues to surrounding cells and reprogram the tumor microenvironment, thereby creating a favorable niche for metastatic colonization [8]. However, the specific mechanisms through which breast cancer cell-derived sEVs remodel the tumor microenvironment remain incompletely understood.
Tumor-associated macrophages, as one of the most abundant immune cell populations in the tumor microenvironment, play a pivotal role in promoting tumor growth and metastasis across various cancer types [9]. Accumulating evidence has demonstrated that sEVs derived from tumors such as pancreatic [10] and colorectal cancer [11] can promote metastasis by inducing macrophage polarization toward the M2 phenotype. Notably, our previous research revealed that sEVs derived from ferroptotic breast cancer cells inhibit macrophage M2 polarization, thereby suppressing the migratory and invasive capacities of breast cancer cells [12]. Ferroptosis is a regulated cell death process driven by iron accumulation and marked by lipid peroxidation and membrane oxidative injury [13]. Emerging evidence suggests that signals released from ferroptotic cells can be packaged into sEVs, which subsequently influence macrophage polarization—a process that plays a critical role in the progression of various diseases. Dai et al. were the first to demonstrate that ferroptotic pancreatic tumor cells can modulate macrophage M2 polarization via sEVs protein KRAS crosstalk [14]. Similarly, Sun reported that sEVs released from ferroptotic cardiomyocytes during myocardial infarction enhanced M1 polarization of cardiac macrophages [15]. These findings underscore the pivotal role of sEVs-mediated communication between ferroptotic cells and macrophages in determining macrophage fate and driving disease progression, highlighting the context-dependent and heterogeneous nature of ferroptosis-related intercellular signaling. Nevertheless, the precise molecular mechanism through which ferroptotic breast cancer-derived sEVs modulate macrophage polarization and influence the metastatic behavior of breast cancer cells remain to be elucidated.
Building upon our previous findings [12], the present study aims to further elucidate the molecular mechanisms by which sEVs derived from ferroptotic breast cancer cells inhibit macrophage M2 polarization, thereby suppressing breast cancer invasion and metastasis. To achieve this, we will employ a combination of bioinformatic analyses, molecular docking, RNA pull-down assays, and other experimental approaches to identify and validate key regulatory molecules and pathways involved in this process. This study is expected to provide novel insights into the interplay between ferroptosis, sEVs-mediated intercellular communication, and macrophage polarization in the context of breast cancer progression, offering potential therapeutic targets for the prevention and treatment of metastatic breast cancer.

Materials and methods

Cell culture and transfection
The human breast cancer cell line MDA-MB-231 (Cat# iCell-h133, iCell Bioscience Inc., Shanghai, China) was utilized in this study. Cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM; Corning, 10–013-CVRC) supplemented with 10% fetal bovine serum (FBS; Gibco, 10099-141) and 1% penicillin–streptomycin (P/S; Sangon Biotech, E607011). Human breast cancer MCF-7 cells (iCell-h129, iCell Bioscience Inc., Shanghai, China) were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Corning, 10–013-CVRC) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% P/S (Sangon Biotech, E607011). Cultures were incubated at 37 °C in a humidified atmosphere containing 5% CO₂. For cell passaging, Trypsin-EDTA (Invitrogen, 11668-500) was used.
To induce ferroptosis, MDA-MB-231 and MCF-7 cells were treated with 10 µM erastin (HY-15763, MCE) for 24 h. To inhibit ferroptosis, MDA-MB-231 and MCF-7 cells were treated with 2 µM Ferrostatin-1 (Fer-1; HY-100579, MCE) for 24 h. Erastin and Fer-1 were dissolved into DMSO.
The human monocytic cell line THP-1 and human monocytic U937 cells (iCell Bioscience Inc., Shanghai, China) were cultured in RPMI 1640 medium (CORNING, 10–040-CVR) supplemented with 10% FBS (GIBCO, 10099141 C), 0.05 mM β-mercaptoethanol, and 1% P/S (Sangon Biotech, E607011) at 37 °C in a humidified incubator with 5% CO₂. For macrophage differentiation, THP-1 and U937 cells density was adjusted to 1 × 10⁵ cells/mL, and 2 mL of the cell suspension was seeded into each well of a 6-well plate (Nunc, 140675). Cells were treated with 100 ng/mL of phorbol 12-myristate 13-acetate (PMA; Sigma, P8139) to induce differentiation. The cells were incubated for 48 h to induce differentiation into adherent M0 macrophage-like cells, which were then used for subsequent experiments.
To prepare siULK1 sEVs, MDA-MB-231 and MCF-7 breast cancer cells were transfected with siNC or siULK1 using Lipofectamine™ 2000. Six hours post-transfection, cells were treated with 10 µM erastin for 24 h to induce ferroptosis. sEVs were harvested from serum-free culture supernatants after an additional 48 h and isolated via standard ultracentrifugation. sEVs from untreated MDA-MB-231 and MCF-7 cells were collected as controls.

CCK8 assay
Cells were plated into 96-well plates (Corning, 3599) at a density of 2,000 cells per well in 200 µL of complete culture medium. Each group was set in six technical replicates. Cell viability was evaluated at 0, 24, 48, and 72 h using the Cell Counting Kit-8 (CCK-8; Beyotime, C0039). At each time point, 20 µL of CCK-8 reagent was added directly to each well without removing the medium. Following a 1-hour incubation at 37 °C, the optical density at 450 nm was recorded with a microplate reader.

Lipid peroxidation detection by BODIPY™ 581/591 C11 staining
To prepare the stock solution, 1 mg of BODIPY™ 581/591 C11 (#D3861, INVITROGEN) was dissolved in 200 µL of an appropriate organic solvent to yield a 10 mM storage solution. For staining, cells cultured in 12-well plates were washed once with PBS to remove residual medium, then incubated with 1 mL of fresh complete medium containing 2 µM BODIPY™ 581/591 C11 (prepared by adding 0.2 µL of stock solution per well). The cells were incubated at 37 °C for 60 min under standard conditions. Following incubation, the staining solution was removed, and cells were gently washed once with PBS. Hoechst 33,342 working solution (#C1029, Beyotime) was then added and incubated at room temperature for 15 min to counterstain nuclei. After staining, fluorescence signals were observed under a fluorescence microscope. In its reduced state, BODIPY™ 581/591 C11 exhibits red fluorescence with excitation/emission at 581/591 nm. Upon oxidation of the polyunsaturated butadienyl moiety, the emission spectrum shifts, and green fluorescence becomes detectable at 510 nm following 488 nm excitation, indicating the presence of lipid peroxides.

sEVs isolation
sEVs were isolated from the culture supernatants of treated MDA-MB-231 cells using a series of differential centrifugation steps. After treatment, cells were cultured in fresh serum-free DMEM for 48 h. The conditioned medium was then collected and the supernatant was centrifuged at 300 × g for 10 min at 4 °C (HITACHI himac CPBOWX series) to remove intact cells. The resulting supernatant was transferred and centrifuged at 2,000 × g for 10 min at 4 °C to eliminate dead cells, followed by a centrifugation at 10,000 × g for 30 min at 4 °C to remove cellular debris. The supernatant was then filtered through a 0.22 μm pore-size syringe filter to eliminate remaining large vesicles and particles. The filtered medium was subjected to ultracentrifugation at 120,000 × g for 2 h at 4 °C. After discarding the supernatant, the sEVs-containing pellet was gently resuspended in pre-chilled 1× PBS using a 200 µL pipette. The pellet was mixed thoroughly by pipetting up and down at least 200 times to ensure complete dissolution of the sEVs. The protein concentration of sEV was quantified using a BCA assay kit (Beyotime, P0012).

Transmission electron microscopy (TEM)
The morphology of isolated sEVs was examined by TEM using a JEM-1200EX electron microscope (JEOL Ltd., Japan) operating at an acceleration voltage of 100 kV. Briefly, 5–10 µL of the sEVs suspension was placed onto a carbon-coated copper grid and allowed to adsorb for 1 min. Excess liquid was gently removed from the edge of the grid using filter paper. The grid was then rinsed with PBS and negatively stained with 10 µL of 2% phosphotungstic acid (pH 7.0) for 1 min. After removing excess stain with filter paper, the grid was air-dried at room temperature for 2 min before imaging.

Nanoparticle tracking analysis (NTA)
sEVs particle size and concentration were assessed via NTA using the ZetaView PMX 110 system (Particle Metrix, Meerbusch, Germany) and ZetaView software version 8.04.02 SP2 at VivaCell, Shanghai. Prior to measurement, isolated exosome suspensions were diluted in 1× PBS (Biological Industries, Israel) to achieve optimal particle concentration. Each sample was analyzed across 11 positions, and system calibration was performed using 110 nm polystyrene standards. All measurements were conducted at a controlled temperature at 26.13 °C with pH 7.0.
The sEV dosage of treatment macrophages was quantified based on protein concentration and NTA. The NTA analysis indicated that the isolated Fer- sEVs had a concentration of approximately 3.2 × 10¹¹ particles/mL. The BCA analysis indicated that the isolated Fer- sEVs had a concentration of approximately 2.334 mg/mL. Accordingly, the 100 µg protein dose used for cell treatment corresponds to approximately 1.4 × 1010 particles in total.

Western blot
sEVs or cell samples were lysed using RIPA buffer supplemented with protease inhibitors. The protein concentration was quantified using a BCA assay kit according to the manufacturer’s instructions. Equal amounts of protein were denatured in loading buffer at 95 °C for 5 min and separated by SDS-PAGE using a 10% polyacrylamide gel, followed by electrophoretic transfer onto PVDF membranes. Membranes were blocked in 5% non-fat milk prepared in TBST and then incubated overnight at 4 °C with the following primary antibodies: anti-CD63 (abcam, ab216130, 1:1000 dilution), anti-ULK1 (Proteintech, 20986-1-AP, 1:20000 dilution), anti-CD81 (Proteintech, 27855-1-AP, 1:5000 dilution), anti-CD9 (Proteintech, 20597-1-AP, 1:5000 dilution), anti-calnexin (Proteintech, 10427-2-AP, 1:10000 dilution), anti-LC3 (Proteintech, 14600-1-AP 1:4000 dilution), anti-TOM20 (Proteintech, 11802-1-AP, 1:20000 dilution), anti-CD86 (Santa cruz, sc-19617, 1:200 dilution), anti-206 (Proteintech, 18704-1-AP, 1:2000 dilution), and anti-GAPDH (Proteintech, 60004-1-Ig, 1:2000 dilution) as a loading control. After washing, membranes were incubated with appropriate HRP-conjugated secondary antibodies Goat Anti-Rabbit IgG H&L (Beyotime, A0208,1:1000 dilution) or Goat Anti-Mouse IgG H&L (Beyotime, A0216, 1:1000 dilution) for 1 h at room temperature. Signals were visualized using an enhanced chemiluminescence (ECL) detection system, and band intensities were quantified using ImageJ software.

RT-qPCR analysis
Total RNA was extracted from cells or sEVs using an RNA isolation kit according to the manufacturer’s protocol. Reverse transcription was carried out using the RevertAid™ First Strand cDNA Synthesis Kit with DNase I (Thermo Scientific, K16225) to synthesize cDNA. Quantitative PCR was performed using 2× PCR Master Mix (Roche) on an ABI QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) in a final volume of 10 µL containing 5 µL of 2× Master Mix, 0.3 µL each of forward and reverse primers (10 µM), 1 µL of cDNA template, and nuclease-free water. The thermal cycling conditions were as follows: initial denaturation at 95 °C for 10 min, followed by 45 cycles of 95 °C for 15 s and 60 °C for 60 s with fluorescence signal collection. Melt-curve analysis was performed at the end of each run by gradually heating from 60 °C to 99 °C at a rate of + 0.05 °C/s to confirm amplification specificity. Primer amplification efficiencies were determined from standard curves and ranged between 90% and 110%. For sEV-derived RNA, no exogenous spike-in RNA was used; instead, equal amounts of total RNA were subjected to reverse transcription, and normalization was performed to GAPDH expression. Relative mRNA expression levels were calculated using the 2^–ΔΔCt method. All primer’s sequences are shown in Supplemental Table 1.

RNase protection assay
Isolated sEVs were resuspended in RNase-free PBS and divided into three treatment groups: PBS control, RNase A alone (100 µg·mL⁻¹, Thermo Fisher) and RNase A (100 µg·mL⁻¹) plus Triton X-100 (0.1% v/v final). Samples were incubated at 37 °C for 30 min. Reactions were terminated by adding an equal volume of phenol: chloroform: isoamyl alcohol followed by RNA extraction. ULK1 mRNA levels were measured by RT-qPCR.

Assessment of mitophagy flux using mCherry-GFP-LC3B reporter in macrophages
To monitor mitophagy flux, THP-1 monocytes were first transduced with the lentiviral construct plenti-cmv-mcherry-GFP-LC3B-LRES-Puro (Heyuan Bio, #H6687) and selected with puromycin to establish stable reporter cells. Following PMA stimulation (100 ng/mL, 48 h), THP-1 cells were differentiated into M0 macrophages. Cells were then divided into five groups: NC, sEVs, Fer-sEVs, siNC-Fer-sEVs, and siULK1-Fer-sEVs. Stable THP-1-derived or U937-derived macrophages were incubated with 100 µg of the respective sEVs for 48 h. The sEV dosage was quantified based on protein concentration and NTA as describes above. Subsequently, mitochondrial staining was performed using a blue fluorescent mitochondrial probe (BestBio, #BB-441146) according to the manufacturer’s instructions. Fluorescence imaging was conducted using an Olympus CKX53 microscope. The tandem mCherry-GFP-LC3B reporter enables visualization of autophagosomes and autolysosomes: GFP fluorescence is quenched in the acidic lysosomal environment, while mCherry remains stable. Thus, yellow (mCherry⁺/GFP⁺) puncta indicate autophagosomes, and red-only (mCherry⁺/GFP⁻) puncta reflect autolysosome formation, allowing quantification of mitophagic flux.

Flow cytometric analysis of CD206⁺ macrophages
Macrophages were harvested for surface marker analysis using flow cytometry. Briefly, cells were gently detached using 12 mM lidocaine (MCE, HY-B0185A) for 3 min followed by 0.25% trypsin digestion for an additional 3 min. The suspension was pipetted gently to facilitate dissociation, then centrifuged at 300 × g for 3 min to collect the cells. The pellet was washed once with PBS and gently resuspended. Cells were counted and adjusted to a density of 2–5 × 10⁵ cells per tube. After centrifugation at 200 × g for 3 min, the supernatant was removed, and the pellet was resuspended in 200 µL PBS. Anti-CD206 primary antibody (Proteintech, 18704-1-AP) was added at a dilution of 1:200, followed by incubation in the dark for 30 min. Cells were then washed and incubated with Alexa Fluor® 488-conjugated goat anti-rabbit IgG secondary antibody (Abcam, ab150077) at a 1:2000 dilution for another 30 min in the dark. Flow cytometric analysis was performed using a Beckman Coulter flow cytometer. Viability dyes to exclude dead cells and doublet discrimination based on SSC-H/FSC-H and FSC-H/FSC-A gating strategies, respectively.

Immunofluorescence staining
Cells cultured in 24-well plates were washed once with PBS and fixed with 4% paraformaldehyde for 30 min at room temperature. After three PBS washes, cells were permeabilized with 0.3% Triton X-100 for 5 min, followed by another set of PBS washes. Blocking was performed with 3% BSA for 30 min to reduce nonspecific binding. Primary antibody against CD206 (Proteintech, 18704-1-AP) was diluted 1:500 in PBS and incubated with cells overnight at 4 °C. After washing, cells were incubated with Alexa Fluor® 488-conjugated secondary antibody (Abcam, ab150077, 1:1000 dilution) at room temperature for 1 h in the dark. Nuclei were counterstained with DAPI for 10 min, followed by final PBS washes. Fluorescent signals were visualized using an Olympus CKX53 fluorescence microscope.

Transwell migration assay
The migratory ability of MDA-MB-231 cells was assessed using a Transwell assay with 8.0 μm pore size inserts (FALCON, 353097) placed in 24-well plates. After being treated with sEVs-conditioned medium from different macrophage groups, MDA-MB-231 cells were harvested, counted, and resuspended in serum-free medium at a concentration of 1 × 10⁵ cells/mL. For each well, 700 µL of complete medium containing 20% FBS was added to the lower chamber as a chemoattractant, and 500 µL of the cell suspension was seeded into the upper chamber. Following incubation, cells remaining on the upper side of the membrane were gently wiped away using a cotton swab. The cells that had migrated to the lower surface were then fixed with 4% paraformaldehyde and stained with crystal violet. The membranes were then rinsed, air-dried, mounted on slides with neutral balsam, and imaged under a light microscope. Cell numbers were quantified by counting three random fields per membrane.

RNA pull-down assay
The full-length sequence of ULK1 was cloned into the pcDNA3.1 expression vector, with EcoRI and XhoI restriction sites introduced at the 5′ and 3′ ends, respectively. The recombinant plasmid pcDNA3.1-ULK1 was linearized using the XhoI restriction enzyme and subsequently purified for in vitro transcription. Biotin-labeled ULK1 RNA probes were synthesized using an in vitro transcription kit according to the manufacturer’s instructions. Target cells were lysed on ice using Pierce™ IP Lysis Buffer (Thermo Fisher Scientific), and the lysates were clarified by centrifugation. The biotinylated ULK1 RNA probes were incubated with streptavidin-coated magnetic beads at 4 °C to form RNA-bead complexes. These complexes were then incubated with pre-cleared cell lysates under gentle rotation to allow for the binding of RNA-associated proteins. Following incubation, magnetic beads were collected and extensively washed to remove non-specifically bound proteins. RNA-bound proteins were then eluted from the beads and subjected to silver staining for visualization. Specific protein bands were excised and analyzed by mass spectrometry to identify ULK1-interacting proteins.

RNA immunoprecipitation (RIP)
RIP was performed using the Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (Millipore) according to the manufacturer’s instructions. Briefly, ~ 5 × 106 donor cells (control or ferroptotic) were lysed in RIP lysis buffer supplemented with RNase inhibitor. Cell lysates were pre-cleared and then incubated with 5 µg of anti-hnRNPA2B1 antibody (Abcam, ab31645) or 1 µg IgG bound to protein A/G magnetic beads at 4 °C overnight with gentle rotation. Beads were washed 6 times with wash buffer, and protein–RNA complexes were digested with proteinase K at 55 °C for 30 min to release RNA. Co-immunoprecipitated RNA was extracted with Trizol method. Enrichment of ULK1 mRNA in IP samples was quantified by qPCR and expressed as fold enrichment relative to IgG or as percentage of input.

Statistical analysis
Statistical analysis was performed by GraphPad Prism 10.0. For all data, normality was assessed using the Shapiro–Wilk test and homogeneity of variance was evaluated using Levene’s test prior to further analysis. When data met the assumptions of normal distribution and equal variance, comparisons between two groups were performed using t tests, and multi-group comparisons were analyzed by one-way ANOVA followed by Tukey’s post hoc test. when data that did not meet assumptions, Kruskal–Wallis test for multiple groups were applied. For all data, data are presented as mean ± SD, with each point in the scatter–bar plots representing one independent biological replicate, and p < 0.05 was considered significant.

Results

Ferroptotic breast cancer cells secrete ULK1-enriched sEVs
In our previous study, we reported that ferroptosis-induced breast cancer cell-derived sEVs (Fer-sEVs) suppressed M2 polarization of macrophages [12]; however, the underlying mechanism remained unclear. Building upon this, the present study aimed to elucidate the molecular mechanisms by which Fer-sEVs regulates M2 macrophage polarization. To this end, we re-analyzed our prior PCR array data comparing macrophages incubated with Fer-sEVs versus those cultured alone. Among the 84 genes, we focused on genes that upregulated in the Fer-sEVs and identified ULK1 as a candidate gene for further investigation based on its high expression and large fold change (Fig. 1A).

To determine whether Fer-sEVs suppresses M2 polarization by delivering ULK1, we first investigated whether sEVs derived from ferroptotic breast cancer cells carry ULK1. MDA-MB-231 cells were treated with the ferroptosis inducer erastin followed by ferroptosis inhibitor Fer-1 treatment. CCK-8 assays showed a significant reduction (at 72 h; p < 0.001, t = 15.49) in cell viability after erastin treatment (Fig. 1B). Lipid peroxidation—a hallmark of ferroptosis—was markedly elevated in the erastin group, as indicated by enhanced green fluorescence (F = 40.12; NC vs. Erastin, p = 0.0005) and diminished red fluorescence (F = 10.61; NC vs. Erastin, p = 0.0148) in oxidized lipid probes (Fig. 1C), but these results was suppressed by Fer-1, confirming successful induction of ferroptosis in MDA-MB-231.
Next, sEVs were isolated from the culture medium of ferroptotic MDA-MB-231 cells. TEM and NTA was subsequently used to examine their morphology and particle size. The results revealed numerous membrane-bound vesicles with a typical cup-shaped or oval morphology, exhibiting a central concavity, and ranging from approximately 50 to 200 nm in diameter as determined by NTA—consistent with the classical size and structure of sEVs (Fig. 1D and E). These morphological features supported the successful enrichment of sEVs from ferroptotic breast cancer cells. To further confirm the identity of the isolated vesicles, Western blot analysis was performed to detect the expression of the classical sEVs surface marker CD63/CD81/CD9. As expected, the protein lysates derived from sEVs exhibited robust expression of CD63, CD9, and CD81 and non-expression of calnexin, whereas calnexin was minimally expressed in the corresponding whole-cell lysates (Fig. 1E). The enrichment of CD63/CD81/CD9 in the sEVs fraction confirmed the purity and identity of the isolated vesicles as bona fide sEVs. Collectively, these results validated the successful isolation and characterization of sEVs derived from ferroptosis-induced breast cancer cells.
Importantly, RT-qPCR analysis showed that ULK1 expression was both significantly upregulated in erastin-treated cells and Fer-sEVs compared to control cells and sEVs derived from untreated MDA-MB-231 cells, respectively (Fig. 1F and G). However, the upregulation of ULK1 in cells and sEVs induced by erastin were restricted by Fer-1. To determine whether ULK1 mRNA was encapsulated within exosomes rather than attached to their surface, RNase protection assays were conducted using intact or detergent-permeabilized vesicles. ULK1 mRNA levels remained stable following RNase treatment alone but were markedly reduced (F = 41.96, p = 0.0004) when sEVs were co-treated with RNase and Triton X-100, demonstrating that ULK1 mRNA is protected within the vesicular lumen (Fig. 1I). These results corroborated our previous PCR array findings and suggested that sEVs ULK1 may be a ferroptosis-responsive factor involved in the regulation of macrophage polarization.

Fer-sEVs enhances mitophagy in macrophages via ULK1 delivery
Previous studies have shown that activated ULK1 forms a complex with ATG13 and FIP200 to recruit downstream ATG proteins and initiate autophagosome formation, thereby regulating autophagy and mitophagy [16]. Importantly, accumulating evidence suggests that ULK1 acts as a metastasis suppressor in breast cancer [17, 18]. Notably, restoration of ULK1-mediated mitophagy has been reported to inhibit NLRP3 inflammasome activation and subsequently reduce breast cancer bone metastasis [19]. Based on these findings, we hypothesized that ULK1-enriched Fer-sEVs may regulate mitophagy in macrophages, thereby influencing breast cancer progression and metastasis.
To test this hypothesis, we generated a ULK1-knockdown MDA-MB-231 cell line using siRNA. These cells were then treated with erastin to induce ferroptosis, and sEVs were isolated as described to obtain ULK1-deficient Fer-sEVs (Fig. 2A; t = 5.089, p = 0.0070). Macrophages were subsequently incubated with either ULK1-intact Fer-sEVs, ULK1-deficient Fer-sEVs, or control sEVs. RT-qPCR and Western blot analyses revealed that ULK1 mRNA and protein expression levels in macrophages were significantly increased following incubation with ULK1-intact Fer-sEVs, compared to those incubated with either control sEVs or untreated controls (Fig. 2B and C). This result was consistent with our previous PCR array findings [12]. In contrast, ULK1 expression was significantly reduced in macrophages incubated with ULK1-deficient Fer-sEVs (Fig. 2B and C).

To further assess whether Fer-sEVs-transferred ULK1 expression influenced mitophagy in macrophages, we employed mCherry-GFP-LC3B tandem fluorescence reporters combined with Mitochondrial Blue staining. As shown in Fig. 2D, incubation with ULK1-intact Fer-sEVs significantly enhanced mitophagy levels in macrophages, evidenced by increased autophagosome/autolysosome formation (strong red and weak green fluorescence) and colocalization with mitochondria (Blue fluorescence). However, this effect was abolished when ULK1-deficient Fer-sEVs was used, indicating that ULK1 was required for Fer-sEVs-mediated induction of mitophagy. To further confirm that this observation reflected genuine mitophagic flux rather than static accumulation of LC3 puncta, bafilomycin A1 (BafA1), a late-stage autophagy inhibitor that prevents lysosomal acidification, was introduced. Western blot analysis (Fig. 2E) showed that Baf A1 treatment led to a pronounced accumulation of LC3 in siNC cells (F = 27, siNC + DMSO vs. siNC + BafA1, p = 0.0027), consistent with normal autophagic flux, whereas LC3 levels remained low in siULK1-transfected cells regardless of BafA1 treatment, suggesting impaired autophagosome formation. Meanwhile, TOM20, a mitochondrial outer membrane marker, exhibited progressive accumulation upon ULK1 knockdown (F = 18.35, siNC + DMSO vs. siNC + BafA1, p = 0.0221) and further increased after BafA1 treatment (F = 18.35, siULK1 + DMSO vs. siULK1 + BafA1, p = 0.0050), indicating defective mitochondrial clearance. Collectively, these data demonstrate that ULK1 deficiency suppresses mitophagic flux, and that Fer-sEVs-induced mitophagy in macrophages is ULK1-dependent.
To overcome the potential limitation of using a single cell line, we performed parallel experiments using another breast cancer cell line, MCF-7, and the human macrophage cell line U937, to validate that sEVs derived from ferroptotic MCF-7 cells induce mitophagy in U937 macrophages through the transfer of ULK1. Western blotting confirmed the purity of MCF-7–derived sEVs by detecting the presence of sEVs markers (CD63, CD9, and CD81) and the absence of the negative marker calnexin (Supplemental Fig. 1 A). RT-qPCR analysis revealed that ULK1 expression was markedly upregulated in MCF-7 cells following erastin-induced ferroptosis, whereas this upregulation was effectively reversed by the ferroptosis inhibitor Fer-1 (Supplemental Fig. 1B). A similar trend was observed in sEVs derived from MCF-7 cells under the same treatments (Supplemental Fig. 1 C), suggesting that ULK1 can be encapsulated and transmitted via sEVs released from ferroptotic MCF-7 cells. Subsequently, sEVs were isolated from ULK1-silenced MCF-7 cells (Supplemental Fig. 1D) and used to treat U937 macrophages. After incubation, U937 cells were treated with the autophagy inhibitor BafA1, and the expression of mitophagy-related proteins LC3B and TOM20 was examined by Western blotting. As expected, ULK1 knockdown markedly reduced LC3B expression while increasing TOM20 levels, indicating impaired mitophagy (Supplemental Fig. 1E). Treatment with BafA1 further enhanced TOM20 accumulation and partially restored LC3B expression. These results were consistent with the findings obtained from MDA-MB-231 and THP-1 cell models, showing a similar trend and further confirming that ferroptotic breast cancer cell–derived sEVs promote mitophagy in macrophages through ULK1 transfer.

Fer-sEVs suppresses macrophage M2 polarization and breast cancer migration via ULK1-dependent mitophagy
ULK1 delivered by Fer-sEVs mediates mitophagy and thereby promotes macrophage M2 polarization, we conducted loss-of-function experiments in recipient macrophages. Macrophages were transfected with siULK1 or siNC and subsequently treated with Fer-sEVs. As shown in Fig. 3A, mCherry-GFP-LC3B tandem fluorescence analysis revealed that Fer-sEVs treatment markedly increased autophagosome and autolysosome formation in siNC macrophages, whereas ULK1 knockdown significantly reduced mitophagic activity. Furthermore, to determine whether ULK1-induced mitophagy contributes to macrophage polarization, macrophages were treated with ULK1-deficient Fer-sEVs in the presence or absence of the autophagy inducer rapamycin. Western blot analysis showed that ULK1-deficient Fer-sEVs decreased CD86 expression (F = 144.3, siNC-Fer-sEVs vs. siULK1-Fer-sEVs, p < 0.0001) and increased CD206 expression (F = 7.519, siNC-Fer-sEVs vs. siULK1-Fer-sEVs, p = 0.0478), indicative of enhanced M2 polarization, while rapamycin partially reversed these effects (Fig. 3B). Consistently, immunofluorescence staining showed that compared to the siNC-Fer-sEVs group, CD206 expression was significantly upregulated in macrophages treated with siULK1- Fer-sEVs, while this effect was markedly reversed by rapamycin treatment (Fig. 3C). Flow cytometric analysis (Supplemental Fig. 2) also revealed a significantly increased proportion of CD206⁺ macrophages in the ULK1-deficient Fer-sEVs group compared to the siNC-Fer-sEVs group (F = 1840, siNC-Fer-sEVs vs. siULK1-Fer-sEVs, p < 0.0001), indicating enhanced M2 polarization upon ULK1 depletion. This increase was significantly attenuated by co-treatment with rapamycin (Fig. 3D; F = 1840, siULK1-Fer-sEVs vs. siULK1-Fer-sEVs + Rap, p < 0.0001), suggesting that ULK1 suppresses M2 polarization by promoting mitophagy.

To assess the functional consequences of macrophage polarization on tumor cell behavior, MDA-MB-231 breast cancer cells cultured with conditioned medium derived from sEVs-treated macrophage and then performed Transwell migration assays (Supplemental Fig. 3). The results showed that macrophages pretreated with Fer-sEVs suppressed the migration of breast cancer cells (F = 24.10, Mac vs. siNC-Fer-sEVs-Mac, p = 0.0002), whereas macrophages incubated with ULK1-deficient Fer-sEVs promoted tumor cell migration (Fig. 3E; F = 24.10, siNC-Fer-sEVs-Mac vs. siULK1-Fer-sEVs-Mac, p = 0.0037). Collectively, these findings indicate that ULK1 transported by Fer-sEVs promotes mitophagy in macrophages, thereby suppressing M2 polarization and ultimately inhibiting breast cancer cell migration.

hnRNPA2B1 mediates the selective loading of ULK1 mRNA into Fer-sEVs upon ferroptosis induction
To further elucidate the mechanism by which ferroptosis induces the selective loading of ULK1 into sEVs, we performed an RNA pull-down assay using a biotin-labeled probe targeting ULK1 mRNA. Silver staining of the pull-down complex revealed multiple protein bands (Fig. 4A), and subsequent mass spectrometry analysis identified a total of 145 ULK1-specific RNA-binding proteins. GO and KEGG pathway enrichment analyses were performed on these 145 ULK1-binding proteins. GO analysis revealed that their molecular functions were predominantly associated with binding activity, with 11 out of the top 20 enriched GO terms related to various types of molecular binding (Fig. 4B). KEGG analysis further indicated that these proteins are involved in pathways such as ferroptosis and glycolysis (Fig. 4C). These findings suggest a functional link between ULK1 and ferroptotic signaling, thereby supporting our experimental observations. Protein–protein interaction (PPI) analysis revealed extensive interactions among the 145 ULK1-binding proteins (Fig. 4D), suggesting that they may function as part of coordinated regulatory networks involved in ferroptosis or related cellular processes.

To narrow down potential candidates responsible for ULK1 mRNA sorting into sEVs, we compiled a list of 13 RNA-binding proteins known to mediate selective packaging of mRNAs or noncoding RNAs into sEVs, based on prior literature [20, 21]. These included NSUN2, MEX3C, MVP, La protein, MTR4, Annexin-2, SYNCRIP, hnRNPA2B1, hnRNPK, YBX1, Ago2, and FMR1. Cross-referencing these known RNA-sorting proteins with the 145 proteins identified in our ULK1 RNA pull-down dataset revealed a single overlapping candidate: hnRNPA2B1 (Fig. 4E), suggesting that hnRNPA2B1 may bind to ULK1 mRNA and facilitate its selective packaging into sEVs. Figure 4F presents the MS2 spectrum of the peptide sequence IDTIEIITDR, confirming the identification of hnRNPA2B1 by mass spectrometry.
To confirm the direct interaction between hnRNPA2B1 and ULK1 mRNA, RIP assays were performed in control and ferroptotic cancer cells using an anti-hnRNPA2B1 antibody. Based on predictions from the catRAPID omics web server (http://service.tartaglialab.com/page/catrapid_omics2_group), hnRNPA2B1 was found to possess three potential binding sites on ULK1 mRNA, and specific primers targeting these regions were designed for RIP-qPCR analysis. The qPCR results (Fig. 5A) showed that in normal MDA-MB-231 cells, ULK1 mRNA was significantly enriched in hnRNPA2B1-immunoprecipitated complexes compared with the IgG control (approximately a sixfold increase), whereas this enrichment was further elevated in erastin-induced ferroptotic cells (approximately a thirteenfold increase), indicating enhanced hnRNPA2B1–ULK1 mRNA binding during ferroptosis. To further directly demonstrate that the selective loading of ULK1 mRNA into exosomes depends on hnRNPA2B1, we mutated the putative EXO-motif regions within the ULK1 3’UTR and constructed both wild-type (WT) and mutant (MUT) plasmids. These constructs were transfected into ULK1-knockout cancer cells, followed by hnRNPA2B1 knockdown or overexpression. sEVs were then isolated, and ULK1 mRNA levels were quantified by qPCR. Efficient knockdown of hnRNPA2B1 in MDA-MB-231 cells was confirmed by RT-qPCR and Western blot (Fig. 5C and D). As shown in Fig. 5B, hnRNPA2B1 silencing markedly reduced ULK1 mRNA enrichment in sEVs derived from cells expressing ULK1 3’UTR-WT, whereas hnRNPA2B1 overexpression restored ULK1 levels to near baseline. In contrast, sEVs from cells carrying the ULK1 3’UTR-MUT construct exhibited consistently low ULK1 expression regardless of hnRNPA2B1 manipulation, indicating that hnRNPA2B1-dependent recognition of EXO-motifs within the ULK1 3’UTR is required for its selective loading into exosomes.

Furthermore, after silencing hnRNPA2B1 in MDA-MB-231 cells, we examined ULK1 levels in both MDA-MB-231-derived sEVs after erastin treatment and in macrophages exposed to these vesicles. As expected, erastin stimulation significantly upregulated ULK1 expression in MDA-MB-231-derived sEVs; however, this upregulation was significantly attenuated when hnRNPA2B1 was silenced (Fig. 5E). Similarly, ULK1 expression in Fer-sEVs-incubated macrophages was significantly reduced in the hnRNPA2B1 knockdown group compared to controls (Fig. 5F). Collectively, these results demonstrate that the incorporation of ULK1 into Fer-sEVs is dependent on hnRNPA2B1-mediated RNA transport.

Discussion
Ferroptosis, a regulated form of iron-dependent cell death characterized by lipid peroxidation, has been increasingly recognized for its role in modulating tumor progression and immune responses. Recent studies suggest that ferroptotic tumor cells can influence the tumor microenvironment through the release of damage-associated molecular patterns and cytokines [22, 23]. However, the role of ferroptosis-induced sEVs as mediators of intercellular communication between tumor cells and immune cells, particularly macrophages, remains largely unexplored. To date, no studies have reported that sEVs from ferroptotic breast cancer cells modulate macrophage polarization via specific RNA or protein cargoes—highlighting a significant knowledge gap. In this study, we report that ULK1-enriched sEVs derived from ferroptotic breast cancer cells suppress macrophage M2 polarization by promoting mitophagy, a process dependent on the RNA-binding protein hnRNPA2B1 for ULK1 mRNA packaging (Fig. 5G). These findings not only uncover a novel ferroptosis–sEVs–mitophagy axis and offers potential therapeutic insights for harnessing sEVs ULK1 to modulate macrophage function and suppress breast cancer metastasis.
ULK1, a 112 kDa serine/threonine kinase, comprises an N‑terminal kinase domain, a serine-rich linker region, and a C-terminal scaffold domain containing MIT/EAT motifs that mediate assembly with ATG13, FIP200, and ATG101 to initiate autophagy [24]. Cryo‑EM studies have revealed that ULK1 forms a supercomplex with PI3KC3‑C1 via dynamic stoichiometry changes, underpinning its central role in macroautophagy [25]. Beyond general autophagy, ULK1 has been increasingly recognized for its direct involvement in mitophagy [26]. Poole et al. demonstrated that ULK1 phosphorylates the mitochondrial receptor BNIP3 at Ser17, enhancing LC3 interaction and receptor stability—thereby promoting mitophagy [27]. Additionally, AMPK-mediated activation of ULK1 stimulates FUNDC1-dependent mitophagy [28]. Despite these insights, no studies have yet examined how ULK1 transported via sEVs can regulate mitophagy in recipient macrophages. Moreover, although ULK1 is well-established as a master regulator of autophagy/mitophagy, in recent years, several studies have suggested that ULK1 plays a regulatory role in ferroptosis execution, mainly through autophagy-dependent mechanisms. For instance, AMPK/ULK1 has been shown to facilitate ferritinophagy, a selective form of autophagy that degrades ferritin to release free iron and thereby promotes lipid peroxidation, a hallmark of ferroptosis, leading to ventilator-induced lung injury [29]. A study by Li et al. reported that the activation of AMPK-ULK1 pathway by M2-type microglia-derived sEVs enhances ferroptosis sensitivity in neurons [30]. However, these studies have primarily focused on ULK1 acting upstream to promote ferroptosis. In contrast, whether ferroptosis itself can regulate ULK1 expression or influence its sorting into sEVs has not been previously reported. Our findings provide the evidence that induction of ferroptosis in breast cancer cells by erastin enhances ULK1 expression and promotes its selective packaging into sEVs via hnRNPA2B1. This suggests a novel upstream regulation of ULK1 by ferroptosis, expanding our understanding of ferroptosis-autophagy crosstalk beyond existing models.
hnRNPA2B1 is characterized by two RNA recognition motifs (RRM1 and RRM2) and an RGG-rich domain, enabling it to bind diverse RNA species and shuttle between the nucleus and cytoplasm [31]. It plays key roles in mRNA processing, transport, stabilization, and translation regulation [32]. Critically, hnRNPA2B1 has been well-documented to mediate the sorting of specific RNAs into extracellular vesicles. Villarroya-Beltri et al. demonstrated that sumoylated hnRNPA2B1 recognizes conserved EXOmotifs (e.g., GGAG/UGCA) in miRNAs, directly binding them and facilitating their loading into sEVs [33]. This is supported by structural studies showing that its RRMs specifically bind AGG and UAG RNA motifs, and that hnRNPA2B1 sumoylation is essential for this function [31, 34]. Additional reports confirmed that depletion of hnRNPA2B1 impairs miRNA packaging into sEVs [35]. Liu et al. proved that hnRNPA2B1-mediated sorting of miR-6881-3p into sEVs controlled by LSD1 to promote bone metastasis of breast cancer [36]. Our findings extend this established mechanism to mRNA cargo. Consistent with its known function, we identified hnRNPA2B1 as the sole overlapping binder between the ULK1 mRNA pull-down dataset and sEVs-sorting proteins, and demonstrated that hnRNPA2B1 knockdown significantly reduces ULK1 loading into Fer-sEVs. These results strongly suggest that hnRNPA2B1 mediates selective ULK1 mRNA sorting into sEVs, highlighting a broader role in sEVs RNA cargo selection beyond miRNAs.

Conclusion
In summary, our study identifies a novel ferroptosis-driven sEVs signaling axis in which ULK1 is selectively packaged into breast cancer cell-derived sEVs via hnRNPA2B1 and subsequently delivered to macrophages. This sEVs ULK1 promotes mitophagy and suppresses M2 polarization, ultimately inhibiting breast cancer cell migration. These findings provide new insights into the interplay between ferroptosis, sEVs RNA transport, and immune modulation, and may offer potential therapeutic targets for preventing tumor metastasis through reprogramming of the tumor immune microenvironment.

Supplementary Information