Annexin A2-dependent extracellular vesicle proteome modulates pre-metastatic stromal fibroblast behavior in triple-negative breast cancer

Abstract

Background

Triple-negative breast cancer (TNBC) is an aggressive subclass of breast cancer with limited treatment options and a strong tendency to metastasize to the lung. Tumor-derived extracellular vesicles (EVs) are key mediators of stromal reprogramming in the pre-metastatic niche. Annexin A2 (AnxA2) is overexpressed in TNBC and linked to signal transduction, cytoskeletal remodeling, and poor prognosis, but its role in shaping EV cargo composition and consequent stromal behavior remains unclear.

Methods

EVs were isolated from the conditioned media of MDA-MB-4175 (LM2) TNBC cells with stable AnxA2 knockdown and control cells using differential ultracentrifugation. EVs were characterized and validated using nanoparticle tracking analysis, Western blotting, and cryo-electron microscopy. EV proteomes were profiled using quantitative proteomics, followed by bioinformatic analyses. Functional and pathway enrichment was performed to identify proteome signatures altered by AnxA2 depletion. The internalization of EVs by recipient lung fibroblasts (MRC-5 and WI-38 cells) was assessed using confocal imaging, and EV-mediated effects on fibroblast behavior were evaluated through migration assays and immunoblotting for proteins associated with fibroblast activation.

Results

AnxA2 knockdown reduced pro-metastatic cellular properties in LM2 cells. EVs from AnxA2-deficient cells displayed reduced abundance of proteins involved in cytoskeletal regulation, adhesion, and vesicle trafficking, with enrichment of immune-related proteins. Pathway analyses predicted reduced motility and vesicle trafficking signaling. PKH26 labeled EVs derived from AnxA2 expressing cells exhibited increased internalization by lung fibroblasts as detected by confocal imaging. Functionally, AnxA2-enriched EV-mediated interactions enhanced fibroblast migration, proliferation, and activation marker expression compared with AnxA2-deficient EVs, consistent with the predicted proteomic changes.

Conclusions

AnxA2 emerges as a regulator of the protein composition of TNBC-derived EVs, enhancing EV-mediated interactions with lung fibroblasts and modulating fibroblast motility and activation, linking tumor cell protein expression to stromal remodeling in the lung microenvironment. This work establishes a framework suggestive of an EV cargo-driven mechanism through which EVs derived from AnxA2-expressing cells may influence lung-specific metastatic colonization in TNBC and highlights the importance of evaluating EV-mediated communication in cancer progression.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12964-026-02708-3.

Background

Triple-negative breast cancer (TNBC) is a highly malignant subtype of breast cancer associated with early relapse and poor prognosis due to rapid progression and high metastatic potential. Unlike other breast cancer subtypes, TNBC exhibits a strong propensity to metastasize to visceral organs, contributing significantly to patient mortality [1–3]. Metastatic TNBC has a dismal median overall survival of only 11–13 months and a 5-year survival rate under 12% [4, 5]. TNBC is prone to metastasize to the liver (~ 22%) and the brain (~ 10%), with the lung representing the most frequent initial site (~ 30–40%) [6, 7]. These outcomes emphasize the urgent need to identify the molecular drivers that facilitate systemic spread and condition the lungs for tumor colonization.

Metastasis is a coordinated process that involves the molecular and cellular reprogramming of the future metastatic organ, often occurring before the arrival of circulating tumor cells. Tumor-derived extracellular vesicles (EVs), along with secreted soluble factors, play a central role in this transformation. In TNBC, EVs prime the metastatic microenvironment by modulating recipient stromal cells such as fibroblasts, endothelial cells, and immune cells within the target organ [8–10]. The molecular determinants within TNBC-derived EVs that regulate stromal reprogramming remain poorly defined, with Annexin A2 (AnxA2) emerging as a promising candidate associated with the modulation of EV molecular composition and functional potential.

AnxA2 is a pleiotropic protein involved in cancer progression and metastasis. In TNBC, it is highly upregulated and contributes to poor prognosis by promoting cell migration and invasion, epithelial-to-mesenchymal transition, and angiogenesis [11, 12]. AnxA2 is a well-recognized component of EVs and has been identified both on the vesicle surface and within the lumen, consistent with its membrane-binding and trafficking functions [10, 13, 14]. Notably, tumor-derived EVs are enriched in AnxA2, which has been associated with the recruitment of EV cargo, suggesting a potential role for AnxA2 in modulating EV composition and contributing to tumor-supportive stromal responses at secondary sites [15–18]. Furthermore, elevated levels of AnxA2, both in tumor tissues and circulating EVs, correlate with poor survival in TNBC patients, highlighting its clinical relevance [16, 19].

The lung, as a primary site of TNBC metastasis, presents a pre-metastatic microenvironment characterized by a highly responsive stromal composition, increased vascular permeability, extracellular matrix remodeling, and immune modulation, all of which are influenced by tumor-derived EVs. These features make the lung a particularly relevant site for investigating EV-mediated metastatic priming in TNBC. Within this context, experimental studies have shown that EV-associated AnxA2 promotes angiogenesis and metastatic spread in breast cancer models. Animals primed with AnxA2-enriched EVs exhibit a significantly greater metastatic burden, whereas the reduction of AnxA2 in EVs leads to a 2-fold and 4-fold decrease in lung and brain metastases, respectively [10]. AnxA2 also plays a role in cytoskeletal remodeling and regulates membrane trafficking events such as exocytosis, which support extracellular vesicle biogenesis and secretion [20].

Prior studies suggest that AnxA2 may regulate EV secretion and cargo loading through calcium-dependent membrane trafficking pathways and may influence stromal interactions, positioning it as an important component of EV-mediated tumor-stromal communication. Clarifying the contribution of AnxA2 to stromal remodeling and tumor-stroma communication is essential for understanding EV-driven metastatic progression. Exploring this axis may provide important biological insight into how TNBC cells establish permissive environments for colonization of the lung.

To explore this, we used the LM2 cell line, a lung-selective derivative of the MDA-MB-231 cell line, to model TNBC with preferential metastatic spread to the lungs [21]. EV proteins can exert immediate functional effects on recipient cells by directly engaging surface receptors or modulating intracellular signaling pathways. To evaluate the influence of cellular AnxA2 expression on the global EV protein composition, we employed a quantitative proteomics-based approach to collectively assess proteome-wide alterations in EVs. In parallel, we studied normal lung fibroblasts, the major stromal constituents of the lung microenvironment, as recipient cells of tumor-derived EVs. Fibroblasts are key stromal cells that respond dynamically to tumor-derived cues within the metastatic microenvironment. In response to tumor-derived EVs, fibroblasts can alter cytokine secretion, increase matrix remodeling activity, and adopt contractile phenotypes, collectively contributing to a tumor-supportive stromal environment. These phenotypic changes often include enhanced motility, which may facilitate stromal-tumor interactions during metastatic progression [22, 23]. This experimental framework enables us to evaluate how cellular AnxA2 influences EV cargo and how these EVs collectively modulate fibroblast behavior, reflecting a cell-type dependent context in shaping a permissive microenvironment for metastatic colonization.

The central objective of this study is to determine the molecular role of cellular AnxA2 in regulating the global protein composition of TNBC-derived EVs and to assess their downstream functional impact on lung fibroblasts. By combining proteomic profiling with bioinformatic analyses, we identified key EV proteins that are differentially expressed in an AnxA2-dependent manner. Using pathway and molecular activity prediction tools, we mapped these proteins to signaling networks and biological processes relevant to stromal reprogramming. Functional validation in lung fibroblasts supported the insights gained from our proteomic analysis, revealing that EVs from AnxA2-expressing cells enhanced fibroblast motility and activation-associated features. Taken together, our findings provide novel evidence for AnxA2-dependent modulation of EV cargo composition in shaping early tumor-stromal interactions and lay the groundwork for future studies on the contribution of AnxA2 to metastatic progression.

Materials and methods

Cell culture

The human triple-negative breast cancer cell line MDA-MB-4175 (LM2), a lung-metastatic derivative of the MDA-MB-231 cell line, was kindly provided by Dr. Joan Massagué (Memorial Sloan Kettering Cancer Center, NY, USA). Normal human lung fibroblast lines MRC-5 (#CCL-171) and WI-38 (#CCL-75) were obtained from the American Type Culture Collection (ATCC). MDA-MB-4175 cells were cultured in high-glucose Dulbecco’s modified Eagle Medium (DMEM; Cytiva HyClone), whereas MRC-5 and WI-38 fibroblasts were maintained in DMEM (Cytiva HyClone). All culture media were supplemented with 10% heat-inactivated fetal bovine serum (FBS; Gibco) and a 1% antibiotic-antimycotic solution (ABAM; Gibco). The cells were maintained at 37 °C in a humidified incubator with 5% CO₂. All the cell lines were routinely tested for mycoplasma contamination using a PCR-based detection assay or the MycoAlert™ PLUS Mycoplasma Detection Kit (Lonza), and only mycoplasma-negative cultures were used in the experiments.

Lentiviral transduction and generation of stable knockdown cell lines

Lentiviral particles carrying pGIPZ vectors encoding green fluorescent protein (GFP), puromycin resistance, and either of two AnxA2-targeting shRNAs (KD1, KD2) or a nonsilencing control (NC) were produced in the HEK293T cells using a second-generation packaging system (Δ8.9 and VSV-G helper plasmids) via calcium chloride transfection. Viral supernatants were collected, clarified, filtered, and used to transduce the LM2 cells in the presence of polybrene. Stable knockdown populations were selected by puromycin treatment and validated by GFP fluorescence and immunoblotting.

Fluorescence-activated cell sorting (FACS)

Following puromycin selection, GFP-positive LM2 cells were enriched by fluorescence-activated cell sorting (Sony SH800Z, 100 μm chip), gating the top ~ 30% fluorescent population. Sorted cells were expanded under standard culture conditions, and GFP expression was monitored to confirm stable enrichment prior to downstream applications.

Cell lysate Preparation and Immunoblotting

Cells were lysed in RIPA Buffer (Thermo Scientific) supplemented with protease and phosphatase inhibitors (Millipore Calbiochem), and protein concentration was measured using a Pierce BCA Protein Assay Kit (Thermo Scientific). Equal amounts of protein were separated on 4–12% Bis-Tris gels (Invitrogen) and transferred to nitrocellulose membranes. The membranes were blocked in 5% nonfat dry milk or 5% bovine serum albumin in TBST, incubated with primary antibodies overnight, washed, and then probed with HRP-conjugated secondary antibodies (SouthernBiotech). Signal detection was performed using Pierce ECL Western Blotting Substrate (Thermo Scientific). Details of the primary antibodies used are provided in the Supplementary material (Additional file 1).

Plasmin generation assay

A chromogenic plasmin generation assay was performed to assess cell surface plasmin activity in LM2 NC and AnxA2-knockdown (KD2) cells. The cells were seeded at equal densities in 12-well plates, incubated overnight, and switched to serum- and phenol red-free media. Reactions were initiated by the addition of 100 nM Glu-plasminogen (Diapharma Group, Inc.) and incubated for 24 h. Conditioned media were cleared by centrifugation, and 100 µl of the supernatant was assayed in triplicate by adding an equal volume of 1 mM of the chromogenic substrate S-2403 (Diapharma Group, Inc.). The absorbance was measured at 410 nm over time, and the plasmin activity was expressed as fold change normalized to corresponding untreated controls within each group (NC and KD2). For inhibitor validation, LM2 cells were treated with a competitive AnxA2 hexapeptide inhibitor (LCKLSL) or a control peptide (LGKLSL) at 1 µM concentration for 4 h prior to Glu-plasminogen addition, and plasmin activity was measured as described above.

Wound healing (2D migration) assay

Cell migration was evaluated using a wound healing assay. NC and KD2 cells were seeded in 6-well plates and grown to approximately 80% confluency in complete DMEM. The monolayers were scratched using a sterile pipette tip to create a linear wound, and the wells were gently washed with PBS to remove the detached cells. The medium was then replaced with DMEM containing 2% FBS. Brightfield images (4x magnification) were acquired at 0, 18, and 24 h post-scratch. The wound area was quantified using Fiji (ImageJ, https://imagej.net/ij/) and was expressed as the percentage of wound closure relative to time zero. Quantification was performed for at least three technical replicates per biological replicate, with a minimum of three independent biological replicates.

Transwell invasion assay

Cell invasion was assessed using Transwell inserts with 8 μm pore size membranes (Corning Life Sciences) in a 24-well format. NC and KD2 cells were seeded into the upper chamber in serum-free DMEM. The lower chamber contained complete DMEM to serve as a chemoattractant. After 24 h of incubation at 37 °C with 5% CO₂, non-invading cells on the upper surface of the membrane were removed using a cotton swab. The membranes were sequentially fixed with 4% paraformaldehyde for 15 min and ice-cold methanol for 20 min, followed by staining with 0.5% crystal violet. Invaded cells on the lower surface were imaged under brightfield microscopy (10x magnification), and quantification was performed manually across multiple fields from replicate wells.

Cell proliferation assay

Cell proliferation was assessed using a Cell Counting Kit-8 (CCK-8; Dojindo Molecular Technologies) according to the manufacturer’s protocol. To synchronize the cell cycle, NC and KD2 cells were serum-starved for 18 h prior to seeding. A total of 10,000 cells per well were plated in 96-well plates in complete growth medium and allowed to grow for 48 h. At the endpoint, CCK-8 reagent was added, and the cells were incubated for 2 h. The absorbance was measured at 450 nm using a microplate reader. Proliferation values were normalized to the average absorbance of the NC (control) group, and quantitative analysis of multiple technical replicates for each biological replicate was performed.

Extracellular vesicle isolation

EVs were isolated from the cell culture conditioned medium (CM) of LM2 cells by differential ultracentrifugation and characterized in accordance with the recommendations outlined in the MISEV 2023 guidelines [24]. Cells were cultured in DMEM supplemented with 10% EV-depleted FBS until they reached ~ 80% confluency. After washing with PBS, the cells were incubated in serum-free DMEM for 24 h. The CM was sequentially centrifuged at 500 × g for 10 min and 3,500 × g for 25 min to remove cells and debris. The clarified supernatant was then centrifuged at 38,000 × g for 60 min using a 45Ti rotor (Beckman Coulter Optima XPN) to remove larger vesicles and apoptotic bodies. The resulting supernatant was filtered through a 0.22 μm syringe filter and subjected to ultracentrifugation at 120,000 × g for 100 min. The pellets were subsequently washed in 0.22 μm-filtered PBS and pelleted again under identical conditions. All the centrifugations were performed at 4 °C. The final EV pellets were resuspended in PBS, RIPA buffer, or complete EV-depleted medium, depending on the subsequent analyses. The experimental parameters and all relevant data related to EV associated experiments have been submitted to the EV-TRACK knowledgebase (EV-TRACK ID: EV250095) [25].

EV-depleted FBS was prepared by ultracentrifugation at 120,000 × g for 18 h at 4 °C, followed by filtration through a 0.22 μm filter and storage at -80 °C. A single batch of FBS and its matched EV-depleted preparation was used across comparative experiments. Complete medium supplemented with EV-depleted FBS was filtered again using a 0.22 μm vacuum filter bottle (Millipore Sigma) prior to use for cell culture.

Nanoparticle tracking analysis (NTA)

Nanoparticle tracking analysis was performed using the NanoSight NS300 system (Malvern Panalytical) to assess the size distribution and concentration of EVs. EVs freshly resuspended in PBS were diluted in 1 mL of filtered PBS, gently mixed, and immediately analyzed. For each sample, five 60-second videos were acquired at a screen gain of 10, camera level of 13–14, and a detection threshold of 5. The data were processed using NTA software (Version 3.4, Build 3.4.4). The size distribution was reported as the mean and mode, and concentration was reported as particles/mL and total particles on the basis of recovered EV volume.

EV protein quantification

EV samples were resuspended in either PBS or RIPA buffer, and the protein concentration was quantified using a Micro BCA Protein Assay Kit (Thermo Scientific), following the manufacturer’s protocol. The absorbance was measured at 562 nm using appropriate standards, and blanks were included to account for background absorbance from the respective dilution buffers. Equal amounts of EV protein were used for Western blot analysis as described above.

EV protease protection assay

EVs derived from NC and KD2 cells were subjected to protease protection analysis. EVs were equally aliquoted at 10–15 µl per condition, corresponding to approximately 5 µg total EV protein, and treated with trypsin at a final concentration of 0.05 mg/mL. Protease digestion was carried out at 37 °C for 20 min. Where indicated, Triton X-100 was added to a final concentration of 0.1% to permeabilize vesicle membranes during digestion. Reactions were terminated by the addition of LDS sample buffer, followed by boiling at 95 °C for 10 min. Samples were resolved by SDS-PAGE and analyzed by immunoblotting as described above.

EV EDTA treatment

EVs derived from NC and KD2 cells were aliquoted at 10–15 µl per condition and treated with Versene solution containing 0.48 mM EDTA. Samples were incubated for 20 min at room temperature. Following treatment, EVs were pelleted as described above. The supernatant was collected, and the EV pellet was resuspended in PBS. Pellet and supernatant fractions were analyzed by immunoblotting as described above.

EV cryo-electron microscopy (cryo-EM)

EV samples from NC and KD2 cells were prepared as 20 µl aliquots in PBS. The samples were applied to Lacey carbon 300 mesh copper grids (Ted Pella) that had been glow-discharged using a PELCO easiGlow system (Ted Pella). The grids were flash-frozen in liquid ethane using the Vitrobot Mark IV system (Thermo Fisher Scientific) with the following settings: a blot time of 3 s, a relative humidity of 95%, and a chamber temperature of 4 °C. Grid screening and imaging were performed on a Talos Glacios transmission electron microscope (Thermo Fisher Scientific) equipped with a K3 direct electron detector (Gatan). Data were acquired in movie mode at a nominal magnification of 45,000× using the SerialEM program [26], yielding a pixel size of 0.877 Å. The dose rate was approximately 14 e⁻/pixel/s, with a total accumulated dose of ~ 50 e⁻/Ų. Each movie stack was dose-fractionated into 40 frames, which were motion corrected and averaged on the fly. The target defocus was approximately − 1.5 μm.

Proteomic sample Preparation and TMT labeling

EV samples were resuspended in RIPA buffer, and the protein concentration was quantified using the Micro BCA Protein Assay Kit (Thermo Scientific) as described above. Following disulfide bond reduction and alkylation, the proteins were digested overnight with trypsin using S-Trap micro columns (Protifi). The resulting peptides were dried and reconstituted in 100 mM triethylammonium bicarbonate (TEAB) buffer. Peptides were labeled using the TMT10plex Isobaric Mass Tagging Kit (Thermo Scientific) according to the manufacturer’s instructions, with the TMT labels 130 C and 131 left unused. The labeling reaction was quenched with 5% hydroxylamine, and equal amounts of peptides from each condition were pooled based on NanoDrop A205 measurements. The labeled peptides were fractionated into eight fractions using a high-pH reverse-phase spin column (Pierce) according to the manufacturer’s protocol. Dried fractions were reconstituted in 2% acetonitrile and 0.1% trifluoroacetic acid (TFA) for LC-MS/MS analysis, and ~ 1 µg of peptides were injected per run.

LC-MS/MS acquisition and data processing

The samples were analyzed on a Thermo Orbitrap Eclipse mass spectrometer coupled to an Ultimate 3000 RSLC-Nano liquid chromatography system. The peptides were separated on a 75 μm x 75 cm EasySpray column (Thermo Scientific) using a 180-minute gradient from 0 to 28% buffer B at a flow rate of 250 nl/min. Buffer A contained 2% acetonitrile and 0.1% formic acid in water, and buffer B contained 80% acetonitrile, 10% trifluoroethanol, and 0.1% formic acid in water. Spectra were acquired in a data-dependent manner across the entire gradient. MS1 scans were performed in the Orbitrap at 120,000 resolution with a standard AGC target. MS2 scans were acquired in the ion trap (turbo scan mode) with an intensity threshold of 5,000, CID collision energy of 35%, isolation width of 0.7 m/z, AGC target set to standard, and a maximum injection time of 35 ms. The charge states of 2–6 were included. Dynamic exclusion was enabled (repeat count 1, exclusion duration 25 s, exclusion mass width ± 10 ppm). A real-time search was employed to select MS2 peaks for synchronous precursor selection (SPS)-MS3 analysis. Peptide matches were searched against the human reviewed UniProt database. MS3 scans were acquired for up to 10 matched MS2 peaks in the Orbitrap at 50,000 resolution using HCD with a collision energy of 65% and a scan range of 100–500 m/z.

Protein identification and quantification were carried out using Proteome Discoverer v3.0 SP1 (Thermo Scientific). Searches were performed against the reviewed human UniProt database (downloaded January 4, 2024; 20,354 entries) using both the Comet and Sequest HT algorithms with INFERYS rescoring. Static modifications included carbamidomethylation (+ 57.0215) of cysteine and TMT labeling (+ 229.1629) at lysine residues and peptide N-termini. Oxidation of methionine (+ 15.9949) was set as a variable modification. Up to one missed tryptic cleavage was allowed. Reporter ion intensities were extracted and normalized using total channel intensity to correct for sample loading variation. Peptide identifications were filtered at a 1% false discovery rate (FDR). The mass spectrometry proteomics data generated during this study have been deposited to the ProteomeXchange Consortium [27] via the PRIDE [28] partner repository with the dataset identifier PXD067600.

Bioinformatic analysis and functional annotation

Normalized proteomic data were subjected to comprehensive bioinformatic analyses. An initial comparison of EV cargo with known EV databases was performed using FunRich (http://www.funrich.org) [29] by uploading the dataset alongside the Vesiclepedia [30] and ExoCarta [31] reference databases. A correlation analysis was performed using the built-in Venn diagram function to assess overlap between the identified proteins and known EV-associated proteins. Differentially expressed proteins (DEPs) were identified, and the dataset was analyzed using QIAGEN Ingenuity Pathway Analysis (QIAGEN Inc., https://digitalinsights.qiagen.com/IPA) [32]. For all bioinformatic analyses performed in this study, a statistical significance threshold of p-value ≤ 0.05 and a log₂ fold-change cutoff of ±0.1 were used. Gene Ontology (GO) annotation for cellular component and biological process enrichment analyses was performed using the PANTHER classification system (RRID: SCR_004869; http://pantherdb.org) [33] and g: Profiler tool (https://biit.cs.ut.ee/gprofiler) [34]. Canonical pathway and network analyses were conducted using QIAGEN IPA. A heatmap of normalized abundance values for the DEPs was generated using the Morpheus web tool (RRID: SCR_017386; https://software.broadinstitute.org/morpheus). A chord diagram was created using the SRplot platform (https://www.bioinformatics.com.cn) [35] to visualize the selected GO biological processes and their associated proteins. Complete lists of DEPs, GO terms, canonical pathways, and enrichment statistics are provided in the Supplementary material (Additional file 2).

EV labeling and internalization assay

Equal amounts of EVs from NC and KD2 cells were labeled using the PKH26 Red Fluorescent Cell Linker Kit (Sigma-Aldrich). Briefly, EVs were diluted to 1 mL in Diluent C, and PKH26 dye was added at a final concentration of 4 µM. The mixture was gently mixed for 30 s and incubated at room temperature for 5 min. The labeling reaction was quenched by the addition of 1% bovine serum albumin, and the EVs were pelleted by ultracentrifugation at 120,000 × g for 100 min at 4 °C in 0.2 μm-filtered PBS. The labeled EVs were resuspended in 10% EV-depleted medium and added to fibroblasts plated on glass coverslips. A dye-only control without EVs, processed in parallel, was included for each experiment. After 4 h of incubation, the cells were washed thoroughly with PBS, fixed in 4% paraformaldehyde for 15 min, counterstained with DAPI, and mounted for imaging. Confocal imaging was performed on an inverted Zeiss LSM 880 Airyscan microscope (63x magnification) using Zen Black software. Identical acquisition parameters were maintained across all comparative samples.

Fibroblast functional assays following EV treatment

For the migration assay, fibroblasts were seeded in 6-well plates and cultured until they reached ~ 80% confluence. Equal amounts of EVs derived either from NC or KD2 cells and from PBS controls were added to 10% EV-depleted medium and incubated for 24 h. A scratch was made using a sterile pipette tip, and the medium was replaced with 2% EV-depleted medium containing the respective EVs. The cells were imaged at 0, 18, and 24 h post-wounding. Quantification of wound closure was performed as described above. For proliferation and immunoblotting assays, WI-38 and MRC-5 fibroblasts in log-growth phase were treated with EVs or PBS control for 72 h. Cell proliferation rate was assessed using the CCK-8 assay as described above. For protein expression analysis, fibroblasts were lysed after treatment, and Western blotting was performed as described above. TGF-β1 (5 ng/mL) treatment was used as a positive control for activation.

Statistical analysis

Statistical analyses were performed using GraphPad Prism version 10.0. The data are presented as the mean ± standard error of the mean (SEM) with n representing independent biological replicates. Comparisons between two groups were made using an unpaired two-tailed t-test, whereas comparisons among more than two groups were performed using one-way ANOVA followed by appropriate post hoc tests. A p-value < 0.05 was considered statistically significant. The following notations were used to indicate significance: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****); ns = not significant.

Results

AnxA2 knockdown in LM2 cells reduces Pro-Metastatic cellular behaviors

To determine how AnxA2 influences extracellular vesicle cargo and stromal interactions, we first established a functional baseline in the LM2 TNBC model by evaluating the impact of AnxA2 depletion on intrinsic pro-metastatic traits. Stable knockdown clones were generated using shRNA-mediated transduction followed by GFP-based sorting, yielding two independent AnxA2-targeting clones (KD1 and KD2) and a non-targeting control (NC) (Additional file 3 Fig. S1A-B). Immunoblotting confirmed that AnxA2 protein levels were efficiently reduced, with KD1 cells showing approximately 2-fold reduction and KD2 showing greater than 10-fold reduction relative to NC cells (p < 0.0001; Fig. 1A-B). Given the high knockdown efficiency, KD2 cells were selected for subsequent investigations.

Fig. 1.

Characterization of AnxA2 knockdown in LM2 cells. A Western blot analysis of AnxA2 and β-actin expression in wild-type (WT) LM2 cells, non-targeting control (NC) LM2 cells, and two AnxA2 knockdown clones (KD1 and KD2) of LM2 cells. B Quantification of AnxA2 expression relative to β-actin (n = 3). C Plasmin generation assay quantification for NC and KD2 cells (n = 4). D Cell proliferation assay (n = 6). E Representative images of NC and KD2 cells at 0 h and 18 h after scratching; scale bar, 100 μm. F Quantification of wound closure (n = 3). G Representative images of NC and KD2 cells from the invasion assay; scale bar, 100 μm. H Quantification of invading cells (n = 3). The data represent the mean ± SEM of independent biological replicates. Statistical comparisons were performed using one-way ANOVA with Tukey’s multiple-comparisons test (B) or an unpaired two-tailed t-test with Welch’s correction (C, D, F, H). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant

At the cell surface, AnxA2 forms a heterotetramer with S100A10 that anchors tissue plasminogen activator (tPA) and plasminogen, enabling plasmin generation and the activation of matrix metalloproteinases that degrade extracellular matrix barriers [11, 36, 37]. The interaction between AnxA2 and tPA can be competitively inhibited by the synthetic hexapeptide LCKLSL, which blocks the tPA-binding site of AnxA2 and thereby reduces plasmin formation [38, 39]. To assess this mechanism, LM2 cells were treated with LCKLSL or the inactive control peptide LGKLSL prior to plasminogen addition. LCKLSL treatment resulted in an approximately 2-fold reduction in plasmin activity compared with LM2 cells (p < 0.0001; Additional file 3 Fig. S1C), confirming that AnxA2 is required for efficient plasmin generation at the cell surface. To further validate whether AnxA2 knockdown altered this function, we measured plasmin generation following the addition of exogenous plasminogen. Compared with NC cells, KD2 cells exhibited an approximately 4-fold reduction in plasmin generating activity (p = 0.0002; Fig. 1C), which reflects the reduced cell surface localization of AnxA2 and impaired activity of the plasminogen-plasmin axis.

Given the established role of AnxA2 in cytoskeletal remodeling and membrane dynamics, we evaluated the effects of AnxA2 loss on cell growth and motility. Compared with NC cells, KD2 cells displayed an approximately 20% reduction in proliferation (p = 0.0097; Fig. 1D). Wound healing assays revealed a 2-fold reduction in the migration rate in the KD2 group compared with the NC group (p < 0.0001; Fig. 1E-F), and Matrigel invasion assays demonstrated a greater than 2-fold decrease in invasive capacity (p < 0.0001; Fig. 1G-H).

Taken together, these data indicate that AnxA2 depletion in LM2 cells attenuates multiple pro-metastatic cellular behaviors, including proteolytic activity, proliferation, migration, and invasion, providing a functional baseline for examining the role of AnxA2 in EV-mediated communication.

AnxA2 knockdown cells release EVs with reduced AnxA2 levels

To evaluate whether cellular AnxA2 depletion influences the composition of LM2 cell-derived extracellular vesicles, EVs were isolated from the conditioned media of NC and KD2 cells using differential ultracentrifugation in combination with membrane filtration (Fig. 2A). Although the isolated vesicles predominantly fell within the 30–300 nm size range, which is consistent with the established size range for small EVs, we refer to them collectively as EVs throughout this study.

Fig. 2.

Isolation and characterization of EVs from LM2 cells. A Schematic of the differential ultracentrifugation workflow used to isolate EVs from the conditioned media of LM2 cells. B Representative NTA profiles showing the particle size distribution, with the x-axis representing vesicle size in nanometers (range = 0-600 nm) and the y-axis indicating particle concentration in particles per mL (range = 0–9 × 10^6). C Quantification of EV yield expressed as the number of particles per µg of total EV protein (n = 4). D Western blot analysis of EV lysates from NC and KD2 cells showing the presence of EV-associated AnxA2 and canonical EV markers (Alix, Tsg101, CD9 and Hsp90), and the absence of non-vesicular proteins (GM130 and Calnexin). E Relative AnxA2 expression levels in EVs from NC and KD2 cells (n = 4). F Representative cryo-EM images confirming vesicular morphology; scale bar, 50 nm. The data represent the mean ± SEM of independent biological replicates. Statistical comparisons were performed using an unpaired two-tailed t-test with Welch’s correction (E). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant

In accordance with MISEV 2023 guidelines, EV identity and purity were verified using a multimodal approach involving NTA, immunoblotting for EV-enriched and non-vesicular compartment markers, and cryo-EM. NTA revealed comparable size distributions for NC cell- and KD2 cell-derived vesicles, with mean diameters within the expected range. The mean EV concentration was modestly lower in the KD2 group (3.65 × 10⁸ particles/mL) than in the NC group (4.20 × 10⁸ particles/mL), as shown in representative plots (Fig. 2B), with statistical significance observed in the replicate datasets (p = 0.0343; Additional file 3 Fig. S2A). The particle-to-protein ratios in both groups exceeded the widely applied criterion of 3 × 10¹⁰ particles/µg protein for high vesicular purity [40], indicating effective enrichment over soluble protein contaminants (Fig. 2C).

Immunoblotting confirmed the presence of the canonical EV markers TSG101, Alix, HSP90, and CD9, and the absence or negligible levels of the non-vesicular compartment markers GM130 (Golgi) and calnexin (endoplasmic reticulum) (Fig. 2D). AnxA2 levels in KD2 cell-derived EVs were reduced by approximately 2-fold relative to that in NC cell-derived EVs (p = 0.0139; Fig. 2E), demonstrating that AnxA2 knockdown in cells is reflected in the vesicular cargo. In this study, EV-associated AnxA2 refers to the total AnxA2 signal detected in EV samples. Cryo-EM verified that the EVs from both groups were morphologically intact and displayed a characteristic circular shape and double-membrane structure (Fig. 2F).

To further assess the biochemical accessibility of EV-associated AnxA2, protease protection and EDTA treatment assays were performed. Protease treatment revealed that AnxA2 exhibited protease-protected patterns similar to the luminal EV marker TSG101, whereas the EV surface marker integrin beta1 showed greater protease sensitivity (Additional file 3 Fig. S2B). Complete loss of signal for all proteins was observed upon membrane permeabilization. In parallel, EDTA treatment of EVs resulted in partial redistribution of AnxA2 into the supernatant fraction, while the EV markers CD9 and Alix remained pellet-associated (Additional file 3 Fig. S2C-D). Together, these findings support the presence of biochemically distinct pools of EV-associated AnxA2.

These results verified the successful isolation of intact, biochemically pure EVs and demonstrated that AnxA2 depletion in LM2 cells resulted in a corresponding reduction of total AnxA2 levels within their secreted vesicles. The presence of biochemically distinct pools of EV-associated AnxA2 further supports heterogeneity in its EV association. This direct correspondence between cellular and EV AnxA2 levels provides a validated foundation for subsequent proteomic investigations into AnxA2-dependent alterations in vesicular cargo composition.

AnxA2 knockdown alters the proteomic composition of secreted EVs

To investigate whether AnxA2 influences the molecular composition of secreted EVs, we performed quantitative proteomic profiling of EVs isolated from NC and KD2 cells using TMT-based mass spectrometry (Fig. 3A). EV lysates were subjected to tryptic digestion and isotopic labeling, and the resulting peptides were analyzed by LC-MS/MS. Across all the replicates, a total of 3,561 unique proteins were identified at a false discovery rate < 5%.

Fig. 3.

Proteomic profiling of LM2 cell-derived EVs reveals AnxA2-dependent changes in vesicular cargo. A Schematic of the proteomic workflow, including EV protein isolation, TMT labeling, LC-MS/MS, and downstream analysis. B Venn diagram generated in FunRich showing overlap between detected EV proteins and curated Vesiclepedia and ExoCarta datasets. C Heatmap of selected DEPs ranked by statistical significance (p-value) and magnitude of fold change generated in Morpheus; values represent the abundance normalized using z-scores. D Volcano plot generated using IPA depicting DEPs in KD2 cell-derived EVs relative to NC cell-derived EVs. Each data point represents a unique protein, with color indicating upregulated (red), downregulated (blue), or non-significant (gray) abundance. E GO cellular component terms enriched for the DEPs in the EV dataset, plotted as -log10 p-values; color intensities reflect fold enrichment

To confirm the vesicular identity of the identified proteins, we compared our dataset to the Vesiclepedia and ExoCarta EV databases. More than 85% of the proteins overlapped with previously reported EV-associated entries (Fig. 3B). Differential expression analysis using a log₂ fold-change cutoff of ±0.1 and p < 0.05 revealed 93 differentially expressed proteins, including 54 downregulated and 39 upregulated in the KD2 group compared with the NC group. The clustered heatmap displays the top 50 DEPs ranked by significance and effect size to show variation across replicates (Fig. 3C). AnxA2 itself was among the most markedly downregulated proteins, with a log₂ fold-change of approximately -1.6, as depicted in the volcano plot (Fig. 3D).

Gene Ontology cellular component enrichment analysis of the DEPs showed significant overrepresentation of extracellular vesicle, vesicle, and membrane; extracellular region, intracellular vesicle, cytoplasmic vesicle, and endocytic vesicle; and clathrin-coated endocytic vesicle, clathrin-coated vesicle, and endosome. Collectively, these terms suggest a predominant association of the DEPs with membrane-bound compartments and vesicle subtypes central to cargo transport, endocytic trafficking, and intercellular communication (Fig. 3E).

AnxA2 depletion was associated with reduced levels of proteins involved in vesicle trafficking and sorting, including syntaxin-5, REPS1, multivesicular body subunit MVB12B, and inositol polyphosphate-5-phosphatase (OCRL). The abundance of proteins linked to membrane organization and adhesion, such as claudin-3, EPCAM, and vinculin was also decreased. Lower levels of cytoskeletal regulators, including the actin-related proteins ARPC5 and ARPC5L and septin-9 were observed. Additional downregulated cargo included SERPINB5, PRC1, EPHB3, and EPS8. A smaller subset of upregulated proteins included the MHC class II molecules HLA-DRA, HLA-DRB3, and HLA-DPB1, along with the surface proteins CD82, CD83, and IL13RA2, which are associated with immune signaling and antigen presentation. The expression of PTGS2, an enzyme involved in inflammatory lipid mediator synthesis, showed the highest fold increase. Additionally, increases in the levels of WNT5A, GPR68, and RHOJ, which have intercellular signaling and cytoskeletal regulatory functions, were observed. Proteins linked to vesicle trafficking and recycling, including EHD1, DNAJC13, and CPNE8, also showed higher abundance.

Proteomic profiling revealed that cellular AnxA2 depletion selectively altered secreted EV cargo associated with vesicle biogenesis, cytoskeletal regulation, and membrane organization while modifying immune-related cargo composition. Together, these shifts highlight EV cargo alterations with putative links to stromal remodeling, vesicle trafficking, and immune-related processes within the lung microenvironment.

AnxA2 knockdown is associated with altered EV cargo signatures linked to motility and vesicle trafficking

To investigate the functional implications of alterations in the EV proteome following AnxA2 knockdown, we performed canonical pathway and biological function enrichment analyses using Ingenuity Pathway Analysis (Fig. 4A-B). Among the pathways with the most significant downregulation in KD2 cell-derived EVs were clathrin-mediated endocytosis, cargo recognition for clathrin-mediated endocytosis, actin cytoskeleton signaling, regulation of actin-based motility by Rho, actin nucleation via the ARP-WASP complex, CDC42 signaling, and RhoA signaling. These pathways are tightly linked to vesicle trafficking, regulation of cytoskeletal dynamics, and directional cell migration, indicating that EVs secreted from AnxA2-deficient cells may have a reduced capacity to modulate recipient cell interactions and cytoskeletal remodeling. In parallel, several immune-associated pathways, including T-cell receptor (TCR) signaling, Robo-Slit signaling, and MHC class II antigen presentation, were upregulated, which aligns with the increased levels of immune-related cargo observed in the proteomic dataset.

Fig. 4.

Enrichment analysis identifies pathways and biological processes altered in EVs from AnxA2-deficient cells. A Canonical pathways with downregulated activity in KD2 cell-derived EVs, as identified by IPA. The bars represent -log10 p-values; color intensities reflect inhibition z-scores. B Canonical pathways with upregulated activity in KD2 cell-derived EVs, as identified by IPA. The bars represent -log10 p-values; color intensities reflect activation z-scores. C IPA-generated network showing downregulated DEPs in KD2 cell-derived EVs and their association with functional processes. D GO biological processes enriched for the DEPs in the EV dataset, plotted using -log10 p-values. E Chord diagram created with SRplot showing associations between selected EV proteins and functional categories

The functional outcomes were predicted using the Molecule Activity Predictor (MAP) tool in IPA. MAP analysis indicated decreased activity of biological processes associated with cell motility, cytoskeleton organization, microtubule dynamics, and morphogenesis of cellular protrusions (Fig. 4C). The corresponding IPA pathway schematics highlighting these processes are provided in additional file 3 (Fig. S3).

The results of GO biological process enrichment analysis supported these predictions, revealing significant enrichment of terms including cell migration, cell adhesion, cell-matrix adhesion, and cell-substrate adhesion, as well as regulation of locomotion, regulation of cell motility, and wound healing (Fig. 4D). To visualize the overlap and distribution of EV proteins across these biological processes, we generated a chord diagram of DEPs mapped to several motility-associated functions, including the regulation of migration, adhesion, and morphogenesis (Fig. 4E). These proteins have established roles in actin branching, membrane dynamics, and junctional stability, underscoring their potential relevance to stromal cell reprogramming.

Pathway and functional enrichment analyses suggested that cellular AnxA2 depletion shifts the global proteome of secreted EVs away from proteins associated with actin-cytoskeletal dynamics, motility, adhesion, and EV-mediated interactions, which is suggestive of altered vesicular cargo composition with the potential to reprogram recipient stromal cells.

EVs derived from AnxA2-Knockdown cells differentially modulate functional and molecular responses in lung fibroblasts

To assess whether alterations in global EV proteomic cargo following cellular AnxA2 knockdown affect stromal cell interactions, we examined EV internalization as well as EV-mediated functional responses in healthy lung fibroblasts. The MRC-5 and WI-38 cell lines, two well-characterized non-transformed human lung fibroblast lines, were used as representative stromal components of the lung microenvironment.

Considering the proteomic enrichment analyses indicated alterations in vesicle trafficking and cytoskeletal organization in KD2 EVs, we first evaluated EV internalization using PKH26-labeled EVs. Fibroblasts were incubated with equal amounts of PKH26-labeled EVs for 4 h, visualized by confocal fluorescence microscopy (Figs. 5A and 6A). Both cell types displayed a lower mean fluorescence intensity (MFI) normalized by cell number per region and corrected for background using dye control. MRC-5 cells showed about 3-fold decrease in MFI following KD2 cell-derived EV treatment than those following NC cell-derived EV treatment (p = 0.0051; Fig. 5B), whereas WI-38 cells showed about 2-fold reduction in MFI following KD2 cell-derived EV treatment when compared with NC cell-derived EV treatment (p = 0.0485; Fig. 6B). Background-corrected fluorescence measurements and differential interference contrast imaging verified the intracellular localization of PKH26 signal (Additional file 3 Fig. S4 A-E).

Fig. 5.

EVs derived from AnxA2-expressing cells elicit distinct functional and activation-associated responses in MRC-5 fibroblasts. A Representative confocal images of MRC-5 fibroblasts treated for 4 h with PKH26-labeled EVs from NC or KD2 LM2 cells. DAPI (blue), PKH26 (red), and merged channels are shown; scale bar, 20 μm. B Quantification of normalized PKH26 fluorescence intensity (n = 3). C Representative scratch wound images at 0 h and 24 h; scale bar, 100 μm. D Quantification of wound closure at 24 h (n = 3). E Western blot analysis of α-SMA and FAP expression in MRC-5 cells treated with PBS, NC-EVs or KD2-EVs. F Quantification of protein expression relative to GAPDH and normalized to the TGF-β1 treated group (n = 3). The data represent the mean ± SEM of independent biological replicates. Statistical comparisons were performed using an unpaired two-tailed t-test with Welch’s correction (B) or one-way ANOVA with Tukey’s multiple-comparisons test (D). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant

Fig. 6.

EVs derived from AnxA2-expressing cells elicit distinct functional and activation-associated responses in WI-38 fibroblasts. A Representative confocal images of WI-38 fibroblasts treated for 4 h with PKH26-labeled EVs from NC or KD2 LM2 cells. DAPI (blue), PKH26 (red), and merged channels are shown; scale bar, 20 μm. B Quantification of the normalized PKH26 fluorescence intensity (n = 3). C Representative scratch wound images at 0 h and 24 h; scale bar, 100 μm. D Quantification of wound closure after 24 h (n = 3). E Western blot analysis of α-SMA and FAP expression in WI-38 cells treated with PBS, NC-EVs or KD2-EVs. F Quantification of protein expression relative to GAPDH and normalized to the TGF-β1 treated group (n = 3). The data represent the mean ± SEM of independent biological replicates. Statistical comparisons were performed using an unpaired two-tailed t-test with Welch’s correction (B) or one-way ANOVA with Tukey’s multiple-comparisons test (D). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant

To independently assess the functional consequences of EV treatment, we next evaluated fibroblast migration and proliferation. Compared with those treated with KD2 cell-derived EVs, fibroblasts exposed to NC cell-derived EVs exhibited greater wound closure (Figs. 5C and 6C). Quantitative analysis revealed that, compared with control treatment, NC cell-derived EV treatment increased the migration of MRC-5 cells by approximately 20% (p = 0.0001; Fig. 5D), whereas KD2 cell-derived EV treatment resulted in reduced migration by about 13% relative to the NC group (p = 0.0023; Fig. 5D). A similar trend was observed in WI-38 cells, NC cell-derived EV treatment increased migration by approximately 21% compared with the control (p < 0.0001; Fig. 6D), while KD2 cell-derived EV treatment led to decreased migration by about 11% relative to the NC group (p = 0.0090; Fig. 6D). These functional differences are consistent with proteomic predictions of diminished actin remodeling and motility-associated signaling. Consistent with these findings, moderate differences in proliferation rates were also observed, with NC EV-treated fibroblasts exhibiting increased proliferation compared with KD2 EV-treated cells in both MRC-5 and WI-38 lines (Additional file 3 Fig. S5A-B).

To determine whether these EV-mediated effects were associated with fibroblast activation, we next examined the expression of activation-associated proteins commonly used to assess fibroblast contractile and remodeling states, including α-smooth muscle actin (α-SMA) and fibroblast activation protein (FAP), using immunoblotting to evaluate potential molecular changes indicative of fibroblast reprogramming. Fibroblasts treated with NC EVs showed higher expression of α-SMA (~ 6-fold; p < 0.01) and FAP (~ 2-fold; p < 0.0001) relative to PBS control, as compared with that of KD2 EV treatment (Figs. 5E-F and 6E-F). Similar expression trends were observed for collagen 1 (COL1A1) and fibronectin (FN1) with moderate fold changes (Additional file 3 Fig. S5C-F).

Collectively, these findings demonstrate that EVs derived from AnxA2-expressing cells elicit stronger functional and molecular responses indicative of fibroblast remodeling. In conjunction with the proteomic data, these results indicate that differences in cellular AnxA2 status are associated with distinct EV profiles and corresponding fibroblast responses.

Discussion

TNBC is marked by early relapse, limited treatment options, and high rates of visceral metastasis [41–43]. The lung is a frequent site of colonization, and communication between tumor cells and the stromal microenvironment plays a pivotal role in the development of this niche [44–49]. AnxA2 is linked to poor prognosis, extracellular matrix degradation, and cytoskeletal remodeling in breast cancer [19, 50]. Although the role of AnxA2 in tumor cell migration and invasion is well established, its influence on EV composition and subsequent stromal reprogramming is less understood.

This study provides new insight into the role of cellular AnxA2 in regulating the proteomic composition of EVs derived from metastatic TNBC cells and in directing their downstream effects on recipient fibroblasts. We show that AnxA2 knockdown in TNBC cells reduces cellular migration and invasion and significantly alters the secreted EV proteome. These vesicles, in turn, exhibit distinct EV-mediated interactions with lung fibroblasts, processes that are key in pre-metastatic niche formation in the lung microenvironment. Accordingly, these findings should be interpreted within the functional context of early stromal remodeling in the lung microenvironment, rather than as evidence of fibroblast-exclusive or fully delineated EV signaling mechanisms.

Consistent with the findings of earlier studies in breast and other cancers, AnxA2 depletion in LM2 cells reduced their migratory, invasive, and proliferative potential, indicating a selective effect on metastatic behavior [36, 51–59]. These phenotypic changes were reflected in the secreted vesicular proteome, suggesting that the status of AnxA2 in donor cells shapes the vesicle cargo landscape and its influence on stromal cells. In this context, we further assessed the biochemical accessibility of EV-associated AnxA2 using protease protection and EDTA-based dissociation approaches. These analyses support the presence of biochemically distinct pools of EV-associated AnxA2, consistent with differential membrane association and protected vesicular localization. Together with prior reports describing both surface-associated and luminal AnxA2 in EVs [10, 13, 14], these findings refine the interpretation of EV-associated AnxA2 in the present study. Although our experiments were conducted using a single high-efficiency knockdown clone, the magnitude of the changes and consistency across multiple functional assays strengthen the confidence in the model.

Proteomic analysis revealed differences in the levels of proteins involved in actin organization, focal adhesion, vesicle trafficking, and metabolic regulation. Control EVs contained proteins previously linked to fibroblast motility, contractility, or intracellular transport, whereas EVs derived from AnxA2-depleted cells presented a reduced abundance of several cytoskeletal and trafficking regulators. These patterns align with prior reports that tumor-derived EVs can reprogram fibroblasts into a pro-migratory state through the delivery of cytoskeletal and matrix-modulating cargo [60–64]. Although EVs were collected from serum-free conditioned media, which avoids contamination from serum-derived vesicles but may alter vesicle release profiles, the vesicles met established purity criteria and were validated by multiple orthogonal approaches, adding strength to the interpretation of downstream functional data.

EVs derived from AnxA2-expressing cells elicited stronger responses in lung fibroblasts, including enhanced migration, proliferation, and expression of activation-associated markers. These functional outcomes indicate that EVs generated from cells differing in AnxA2 status exert distinct effects on fibroblast behavior. Notably, prior work from our laboratory using a comparable non-targeting versus AnxA2-knockdown cells-derived EV framework demonstrated differential responses in additional stromal populations, including endothelial cells and macrophages, where AnxA2-dependent EVs promoted angiogenic and activation-associated phenotypes [10]. Together, these findings support the concept that cellular AnxA2 status shapes EV states with the capacity to differentially modulate multiple stromal cell types within the tumor microenvironment, while the nature of the response remains context dependent.

PKH26-labeled EV internalization assays revealed greater internalization of EVs derived from AnxA2-expressing cells compared with EVs derived from AnxA2-knockdown cells in fibroblasts. This observation indicates that cellular AnxA2 status is associated with differences in EV internalization efficiency under the experimental conditions used. However, the specific uptake routes engaged by these EVs were not defined in the present study. EV internalization is known to occur through multiple, partially overlapping mechanisms, including clathrin-mediated endocytosis, caveolin-dependent pathways, macropinocytosis, and direct membrane fusion [65–67]. Given that AnxA2 depletion alters the global EV proteome and may influence vesicle surface composition, the observed differences in internalization likely reflect changes in overall EV properties including potential effects on multiple uptake mechanisms. Delineating the molecular determinants and pathways governing EV internalization will require targeted inhibition, receptor-blocking, or surface-protein mapping approaches and represents an important direction for future investigation.

The diminished migration of fibroblasts treated with AnxA2-deficient EVs reinforces the functional differences in EV-mediated effects on stromal cells. Fibroblast motility is a defining feature of the activated stromal phenotype that supports pre-metastatic niche development through extracellular matrix remodeling and recruitment of immune and endothelial cells. Our results are consistent with studies in pancreatic and colorectal cancers, where tumor EVs have been shown to condition fibroblasts through integrin and matrix-binding protein delivery [23, 68]. In addition to migration, EVs from AnxA2-expressing cells also enhanced fibroblast expression of canonical activation-associated markers, αSMA and FAP, confirming that these effects are consistent with fibroblast activation and indicative of early stromal remodeling. Extending these observations, the cytoskeletal and transport regulators detected in AnxA2-enriched EV proteome likely contribute to downstream effects on fibroblast behavior. However, the present study does not distinguish whether these functional responses arise predominantly from EV surface-associated interactions or from internalized EV cargo. Resolving this distinction will further refine our understanding of how EV-mediated signaling is executed at the stromal interface and is an important focus of future investigations.

Not all cargo changes were linked to enhanced motility. Some proteins that showed elevated levels in EVs from AnxA2-deficient cells are associated with tumor progression but not typically with stromal activation. These findings indicate that not all shifts in EV composition favor pro-migratory programming and that the recipient cell context, cargo functionality, and potential compensatory changes in donor cells shape EV outcomes [69, 70].

Our results indicate that in addition to affecting motility-related cargo, cellular AnxA2 influences the levels of EV-associated metabolic and redox regulatory proteins. The absence of AnxA2 was linked to reduced abundance of proteins involved in glutathione metabolism, the stress response, and vesicle-mediated transport. These shifts may limit fibroblast adaptability beyond cytoskeletal dynamics, echoing the findings of recent studies in which tumor EVs were found to modulate stromal metabolism and oxidative balance to create a niche supportive of tumor colonization [71, 72].

From a translational perspective, these findings raise the possibility that disrupting cellular AnxA2 in tumors [73–75] could affect tumor growth while also impairing early stromal conditioning mediated by tumor-derived EVs [10, 18], potentially slowing metastatic progression. Given that EV-associated AnxA2 is significantly reduced in vesicles derived from AnxA2-knockdown cells, defining the potential functional contributions of EV-associated AnxA2 would provide additional biological insight into how cellular AnxA2 status shapes EV behavior. While such studies would advance biological understanding, the present findings underscore cellular AnxA2 as the upstream determinant of EV composition and function, supporting cell-level targeting as the more relevant translational strategy. The overall EV proteomic signatures identified here may also inform biomarker strategies for aggressive diseases. Nevertheless, variability is well recognized across TNBC cell lines and patient-derived models, which differ in metastatic tropism, stromal interactions, and EV secretion profiles. Extending these observations to additional TNBC systems will be essential to establish the translational relevance of the AnxA2-EV axis and to strengthen its clinical significance.

Looking ahead, the findings presented here establish a foundation for further dissection of cellular AnxA2-dependent EV cargo remodeling and its role in stromal modulation during metastatic progression. While this study demonstrates that EVs derived from AnxA2-expressing and AnxA2-deficient cells elicit distinct stromal responses, these functional differences are observed alongside broad, coordinated alterations in EV protein composition identified by quantitative proteomics. As such, the proteomic dataset generated here provides a molecular framework for prioritizing candidate pathways and cargo classes for future mechanistic interrogation. Future studies will be required to delineate the specific molecular mechanisms underlying these differences. In particular, defining the uptake pathways engaged by EVs with distinct cargo profiles will be important for understanding how cellular AnxA2-dependent alterations in EV composition influence recipient cell responses. Beyond fibroblasts, extending these analyses to additional stromal and immune cell populations will be necessary to fully capture the breadth of EV-driven niche conditioning, particularly in light of prior work demonstrating differential EV effects on endothelial cells and macrophages. Finally, in vivo studies tracking EV biodistribution, persistence, and temporal activity within metastatic target organs will be essential for defining the physiological relevance of AnxA2-dependent EV signaling in organ-specific niche formation. Together, these approaches will build upon the current findings to advance a more comprehensive and mechanistic understanding of EV-mediated tumor-stroma communication in metastatic TNBC.

In summary, this study identifies cellular AnxA2 as a determinant of EV cargo composition and stromal engagement in TNBC. By demonstrating that differences in cellular AnxA2 status are associated with distinct EV profiles and corresponding fibroblast responses, including altered motility and early activation-associated features, this work establishes a conceptual framework linking tumor cell protein expression to stromal reprogramming potential. These insights underscore the broader principle that extracellular vesicles act as precision tools of intercellular communication, shaping stromal behavior in ways that influence metastatic trajectories.

Conclusion

By demonstrating that cellular AnxA2 status is associated with distinct extracellular vesicle cargo profiles and corresponding fibroblast responses, this study links tumor cell protein expression to stromal remodeling in metastatic TNBC. These findings clarify the contribution of cellular AnxA2 to EV-mediated tumor-stroma communication while highlighting a broader paradigm in intercellular signaling, in which vesicle composition encodes multifaceted cues capable of reprogramming the host microenvironment. This concept extends beyond a single protein or tumor type, providing a framework to investigate comparable communication pathways across cancers. Cargo-driven signaling networks in different tumor contexts and their integration with other niche-modulating factors remain important areas for future exploration to fully elucidate the role of EVs in cancer progression.

Abbreviations

TNBC

Triple-negative breast cancer

EV

Extracellular vesicles

AnxA2

Annexin A2

LM2

MDA-MB-4175 TNBC cell line

DMEM

Dulbecco’s Modified Eagle Medium

FBS

Fetal bovine serum

GFP

Green fluorescent protein

KD

Annexin A2 knockdown LM2 cells

NC

Non-targeting control LM2 cells

FACS

Fluorescence-activated cell sorting

PBS

Phosphate buffer saline

CM

Cell culture conditioned media

NTA

Nanoparticle tracking analysis

EM

Electron microscopy

TMT

Tandem mass tag

FDR

False discovery rate

DEP

Differentially expressed proteins

IPA

Ingenuity Pathway Analysis

GO

Gene ontology

tPA

Tissue plasminogen activator

MAP

Molecule Activity Predictor

MFI

Mean fluorescence intensity

α-SMA

Alpha-smooth muscle actin

FAP

Fibroblast activation protein

COL1A1

Collagen type I alpha 1

FN1

Fibronectin

Authors’ contributions

R.T. and J.K.V. conceived the idea and planned the study. R.T. developed the experimental design and conducted all the experiments. P.R. assisted with immunoblotting, migration and invasion assays and confocal staining. R.T. collected and analyzed the data. R.T. and J.K.V. interpreted the data. R.T. organized the data and wrote the manuscript. J.K.V. provided supervision and ongoing guidance. All the authors have reviewed and approved the final version of the manuscript.

Funding

This work was funded by the National Institutes of Health under the grant R01CA220273 (NCI) and S21MD012472 (NIMHD). The study was supported in part by the Cancer Prevention and Research Institute of Texas (CPRIT) under the grant RP210046. The funding bodies had no role in study design, data collection, analysis, interpretation, or manuscript writing.

Data availability

The experimental parameters and all relevant data related to EV associated experiments have been submitted to the EV-TRACK knowledgebase (EV-TRACK ID: EV250095) (25). The mass spectrometry proteomics data generated during this study have been deposited to the ProteomeXchange Consortium (27) via the PRIDE (28) partner repository with the dataset identifier PXD067600.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.