Migrasome-propagated ITGβ3-TGFβ1 circuit orchestrates breast cancer brain metastasis by reprogramming the microglial niche

Human samples

The following clinical specimens were collected at Liaocheng People’s Hospital (Liaocheng, Shandong, China): (1) six fresh MBC tissue samples, (2) six formalin-fixed, paraffin-embedded BC tissues with matched adjacent tissues, and (3) three formalin-fixed, paraffin-embedded BCBM tissues. All studies involving human samples were approved by the Ethics Committee of Liaocheng People’s Hospital (approval No. 2024193). All participants provided informed consent to participate.

Animal model

All animal procedures were approved by the Animal Care and Use Committee of Liaocheng People’s Hospital (approval No: 2024194) and conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals (8th edition)/ARRIVE guidelines. Female BALB/c nude mice were purchased from Vital River Laboratory Animal Technology (Beijing, China) and housed under a 12 h light/dark cycle at 22 ± 1°C in an environment free of specific pathogenic bacteria.

In animal experiments, the mice were randomly assigned to groups according to a random number generator. To evaluate the effects of migrasomes (Mig-BCBM) on BCBM, we administered tail vein injections of either Mig-BCBM or Mig-BC every 3 days for a total of four administrations over a 12-day preconditioning period. A BCBM model was subsequently established via intracardiac inoculation of 5 × 10⁵ 231-BR cells stably expressing luciferase (231-BR^Luc) on day 0. At 3 weeks post-inoculation, tumor burden was assessed via in vivo bioluminescence imaging (IVIS Spectrum, PerkinElmer, Waltham, MA, USA). Imaging was performed 10 min after intraperitoneal injection of D-luciferin potassium salt (Macklin, cat# D812647) at 150 mg/kg.

To investigate the inhibitory effect of the TGFβR inhibitor galunisertib on tumor growth, we established a subcutaneous xenograft model. Nude mice received two preconditioning intraperitoneal (i.p.) injections of galunisertib (Selleck, cat# S2230; 75 mg/kg) on days -6 and -3. On day 0, 1 × 10⁶ 231-BR^GFP cells pre-treated with galunisertib in vitro and 1 × 10⁶ Mig-BCBM-primed HMC3 microglia in a 1:1 ratio were subcutaneously co-implanted. The mice then received maintenance i.p. injections of galunisertib (75 mg/kg) every 3 days. The maximum tumor size permitted by our ethics committee was 1.5 cm in diameter or 10% of the animal’s body weight. All specimens used in the study remained within these limits. On day 21 postimplantation, the tumor burden was assessed via in vivo GFP fluorescence imaging with an IVIS Spectrum system (PerkinElmer). After imaging, the mice were euthanized via cervical dislocation under anesthesia, and the tumors were harvested for assessment of weight and size.

To investigate the inhibitory effect of galunisertib on BCBM, we established a BM model via intracardiac inoculation of 5 × 10⁵ stably GFP-expressing 231-BR cells (231-BR^GFP) under anesthesia (day 0). The mice subsequently received maintenance i.p. injections of galunisertib (75 mg/kg) every 3 days (q3d). On day 28 post-inoculation, the mice were euthanized via cervical dislocation under anesthesia. The brains were harvested, and the metastatic burden was assessed via ex vivo GFP fluorescence imaging with an IVIS Spectrum system (PerkinElmer). After imaging, the brain tissues were fixed in 10% neutral buffered formalin, paraffin-embedded, sectioned, and stained with hematoxylin and eosin (H&E) for histological analysis of the metastases.

Cell culture

The sources and culture conditions of the 231-BR BC BM cells and parental MDA-MB-231 cells used in the experiment were as previously described^29,32–34. HMC3 human microglia (Sunncell Biotech, Wuhan, China) were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) (BBI, cat# E600003) supplemented with 10% fetal bovine serum (FBS) (BBI, cat# E600001) and 1% penicillin-streptomycin (Beyotime, cat# C0222). HMC3 cells were incubated at 37°C under a humidified atmosphere of 95% air and 5% CO₂.

Purification of migrasomes

Migrasomes were isolated from cells through a previously published protocol²⁹. In brief, the cells were detached with 0.25% trypsin with EDTA and neutralized with DMEM (BBI), and the supernatant was collected. For enrichment of large vesicles including migrasomes, the cell culture supernatant was subjected to sequential centrifugation steps at 4°C: 1,000 × g for 10 min (with supernatant retained and pellet discarded), 4,000 × g for 20 min (with supernatant retained and pellet discarded), and finally 20,000 × g for 30 min (with supernatant discarded and pellet retained)³⁵. The pellet was subsequently resuspended in 400 μL extraction buffer. Migrasome purification was performed via Optiprep (Sigma-Aldrich, cat# LYSISO1)-based density gradient centrifugation. Briefly, the crude extract was combined with 400 μL 1× extraction buffer, 30 μL 2.3 M sucrose, and 505 μL 60% Optiprep solution, and loaded onto a pre-formed Optiprep gradient consisting of the following layers (from bottom to top): 27% (800 μL), 22.5% (1,000 μL), 19% (crude sample layer, 1,585 μL), 16% (1,000 μL), 12% (900 μL), and 8% (300 μL). The gradient was centrifuged at 4°C at 150,000 × g for 4 h. After centrifugation, the fractions were collected from the top downward (500–700 μL per fraction). Each fraction was mixed with an equal volume of PBS and centrifuged at 4°C and 20,000 × g for 30 min. The resulting pellet was designated purified migrasomes.

Negative staining transmission electron microscopy

Migrasomes were fixed in 2.5% glutaraldehyde (diluted from a 25% stock) at room temperature for 2 h. A 20 μL aliquot of the fixed migrasome suspension was applied to a carbon-coated copper grid and incubated for 5 min at room temperature to allow adsorption. Excess liquid was carefully wicked away by touching filter paper to the edge of the grid. The grid was then stained with 2% (w/v) phosphotungstic acid (pH 7.0, adjusted with NaOH). Excess stain was similarly removed by wicking with filter paper, and the grid was fully air-dried at room temperature. Migrasome morphology was examined with a JEM-1200EX transmission electron microscope (Jeol, Tokyo, Japan) operated at 80 kV. Representative images were acquired and analyzed.

Resin-embedded ultrathin section transmission electron microscopy

Purified migrasomes, freshly resected human BC tissues, or BCBM samples derived from model mice were pre-fixed with agarose. The samples were then fixed in 1% osmium tetroxide at room temperature for 2 h in the dark. After osmication, the samples underwent dehydration in a graded ethanol series followed by acetone, infiltration, and resin embedding. The polymerized resin blocks were sectioned with an ultramicrotome to produce ultrathin (60 nm) sections. Sections were collected onto 150-mesh copper grids coated with Formvar film. The grids were subsequently subjected to double staining through incubation in 2% uranyl acetate in ethanol (saturated solution) for 8 min in the dark, followed by incubation in 2.6% lead citrate (aqueous) for 8 min in a CO₂-free environment. After each staining step, the grids were thoroughly rinsed with ultrapure water, and excess liquid was carefully removed by wicking with filter paper. The grids were air-dried at room temperature overnight. Ultrastructural analysis and image acquisition were performed with a JEM-1200EX transmission electron microscope (Jeol, Tokyo, Japan) for observation and image acquisition for analysis.

Wheat germ agglutinin staining for migrasome visualization

Cell culture-treated 14 mm diameter glass coverslips were coated with 2 μg/mL human fibronectin (dissolved in PBS; Sigma, cat# F2006) by incubation at 37°C for 6 h. Cells were seeded onto coated coverslips and cultured until reaching 30–40% confluency. The cells were washed three times with PBS and fixed with 4% paraformaldehyde (PFA) for 15 min. After fixation, the cells were washed and incubated with wheat germ agglutinin (WGA)-Alexa Fluor™ 488 conjugate (WGA-488; Thermo Fisher, cat# W11261) for 30 min in the dark. Unbound lectin was removed by washing three times with PBS. Migrasomes were imaged on a ZEISS Cell Discoverer 7 automated confocal microscope (ZEISS Microscopy, Germany).

Immunofluorescence staining

Brain tissue from mice with BCBM was processed through transcardial perfusion with PBS followed by ice-cold 4% PFA. The brains were subsequently fixed in 4% PFA for 24 h at 4°C, cryoprotected in 30% sucrose, and coronally sectioned at a thickness of 20 μm with a cryostat (Leica CM1950). For in vitro studies, HMC3 microglia were treated with migrasomes for 48 h and fixed with 4% PFA for 15 min at room temperature. Both tissue sections and fixed cells were permeabilized with 0.3% Triton X-100 (10 min) and blocked with 5% BSA. The samples were incubated overnight at 4°C with the following primary antibodies diluted in 1% BSA/PBS: anti-Iba1 (Cell Signaling Technology, cat# 17198S, 1:400), anti-arginase-1 (Cell Signaling Technology, cat# 93668, 1:400), anti-CD206 (Proteintech, cat# 60143-1-Ig, 1:400), and anti-iNOS (Proteintech, cat# 18985-1-AP, 1:400). After being washed with PBS, the samples were incubated for 1 h at room temperature with the following secondary antibodies: Alexa Fluor^® 488-conjugated goat anti-rabbit (Abcam, cat# ab150077, 1:1,000), Alexa Fluor^® 555-conjugated goat anti-mouse (Abcam, cat# ab150114, 1:1,000), and Alexa Fluor^® 647-conjugated goat anti-rabbit (Abcam, cat# ab150079, 1:1,000). For HMC3 cells, Phalloidin-iFluor 594 (Abcam, cat# ab176757, 1:200) was co-administered with secondary antibodies to visualize F-actin. All the samples were mounted with DAPI-containing anti-fade medium (MedChemExpress HY-K1048) and imaged with a Vectra 3 Automated Quantitative Pathology Imaging System (PerkinElmer, Waltham, MA, USA).

Western blotting

All cells/migrasomes were lysed with RIPA lysis buffer (Beyotime, cat# P0013B) for protein extraction, and protein concentrations were determined with BCA assays (Beyotime, cat# P0010). The samples were subsequently loaded onto a 10% SDS-PAGE gel for electrophoretic separation. After electrophoresis, the proteins were transferred from the gel to a PVDF membrane (Millipore, cat# IPVH00010). After transfer, the membrane was blocked with 5% skim milk in TBST at room temperature for 1 h and then incubated with primary antibodies overnight at 4°C. The membrane was washed three times with TBST. Subsequently, HRP-conjugated goat anti-rabbit or anti-mouse secondary antibodies were added, and the membrane was incubated at room temperature for 1 h. After three additional 10-min washes with TBST, the membrane was incubated with Immobilon ECL Ultra Western HRP Substrate (Millipore, WBULS0500) for 2 min, and signals were captured with a chemiluminescence imaging system (Tanon, Shanghai, China). Both β-actin (Santa, cat# sc-47778, 1:1,000) or GAPDH (Abcam, cat# ab9485, 1:1,000) served as internal controls, each used to normalize its respective target proteins (cells and migrasomes). The primary antibodies used included the following: anti-PIGK (Abcam, cat# ab201693, 1:1,000), anti-TSPAN4 (Thermo Fisher, cat# PA5-69344, 1:1,000), anti-integrin β3 (Cell Signaling Technology, cat# 13166, 1:1,000), anti-E-cadherin (Bioss, cat# bs-1519R, 1:1,000), anti-vimentin (Bioss, cat# bs-23063R, 1:1,000), anti-Smad2 (Proteintech, cat# 67343-1-Ig, 1:5,000), anti-Smad3 (Proteintech, cat# 66516-1-Ig, 1:5,000), anti-phospho-Smad2 (Ser465/467) (Cell Signaling Technology, cat# 3108, 1:1,000), anti-phospho-Smad3 (Ser423/425) (Cell Signaling Technology, cat# 9520, 1:1,000), anti-AKT (Proteintech, cat# 60203-2-Ig, 1:5,000), anti-phospho-AKT (Ser473) (Proteintech, cat# 66444-1-Ig, 1:5,000), anti-phospho-AKT (Thr308) (Proteintech, cat# 29163-1-AP, 1:5,000), anti-arginase-1 (Cell Signaling Technology, cat# 93668, 1:1,000), anti-CD206 (Proteintech, cat#18704-1-AP, 1:5,000), anti-iNOS (Proteintech, cat# 22226-1-AP, 1:5,000), anti-CD81 (Proteintech, cat# 27855-1-AP, 1:2,000), anti-CD86 (Proteintech, cat# 83523-4-RR, 1:2,000), anti-TMEM119 (Cell Signaling Technology, cat# 98778, 1:1,000), and anti-EGR1 (Bioss, cat# bs-1076R, 1:1,000). The secondary antibodies were HRP-labeled goat anti-rabbit IgG (H + L) (Beyotime, cat# A0208, 1:1,000) and HRP-labeled goat anti-mouse IgG (H + L) (Beyotime, cat# A0216, 1:1,000).

Quantitative reverse transcription polymerase chain reaction analysis

Total RNA was isolated from cell cultures or purified migrasomes with TRIzol™ reagent (Thermo Fisher Scientific, cat# 15596026CN). RNA concentration and purity were quantified with a NanoDrop™ One spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). First-strand cDNA synthesis was performed with a BeyoRT™ II First Strand cDNA Synthesis Kit (Beyotime, cat# D7168L). Quantitative reverse transcription polymerase chain reaction (RT-qPCR) was performed in triplicate in 20 μL volumes containing 1× TB Green Premix Ex Taq II (Takara Bio, cat# RR420A), 0.4 μM gene-specific primers (Table S1), and 10 ng cDNA template. Amplification was conducted on a QuantStudio™ 7,500 Real-Time PCR System (Applied Biosystems).

H&E staining

Brain tissues were fixed in 10% neutral-buffered formalin for 48 h, processed through graded ethanol and xylene, and paraffin-embedded. Consecutive 5-μm coronal sections were cut with a Leica RM2235 microtome. After deparaffinization and rehydration, the sections were stained with hematoxylin (Solarbio cat# G1120; 8 min), differentiated in 0.5% acid ethanol, subjected to bluing in Scott’s solution, counterstained with eosin (3 min), dehydrated, cleared in xylene, and mounted. Metastatic foci were quantified by two blinded pathologists examining whole-slide scans (Olympus VS200, 20× objective, brightfield).

Lentiviral infection

After 231-BR cells were infected with the lentiviral vectors Lv-shITGβ3 and Lv-shEGR1 (Genechem, Shanghai, China), successful knockdown of intracellular ITGβ3 and EGR1 gene expression was confirmed. The Lv-sh-NC lentivirus (Genechem, Shanghai, China) served as a negative control. The infection procedure was conducted in accordance with the viral protocol, by using the specified multiplicity of infection (MOI) for a duration of 12–16 h. After the infection period, puromycin (Beyotime, cat# ST551) was added at a final concentration of 2 μg/mL for selection, and incubation continued for 48 h to eliminate 231-BR cells that were not successfully infected. Through this process, stable 231-BR cell lines with knockdown of ITGβ3 (231-BR-sh-ITGβ3) and EGR1 (231-BR-sh-EGR1) were successfully established. Migrasomes were isolated from the successfully constructed 231-BR-sh-ITGβ3 cells for subsequent experimental investigations.

Flow cytometric analysis

HMC3 microglia were treated with migrasomes for 48 h, detached with 0.25% trypsin-EDTA (Gibco), and quenched with complete DMEM. The cells were washed twice with ice-cold PBS (300 × g, 5 min), resuspended in FACS buffer (PBS + 2% FBS), and adjusted to 1 × 10⁶ cells/mL. Surface staining was performed for 30 min at 4°C in the dark with pre-titrated antibodies to CD163 (Proteintech, cat# 65561-1-MR, 1:200) or CD86 (Proteintech, cat# 83523-4-RR, 1:200). Data acquisition was performed with a flow cytometer (BD, New York, USA), and the collected data were analyzed in FlowJo™ software.

Cell proliferation assays

The supernatant harvested from the Mig-BCBM-treated HMC3 cells after 48 h of treatment is denoted conditioned medium (Co-HMC3). To remove cells and debris, we subjected the harvested supernatant to sequential centrifugation at 1,000 × g for 10 min, followed by 4,000 × g for 20 min. To eliminate residual migrasomes and larger vesicles, we further centrifuged the supernatant at 20,000 × g for 30 min. The resulting supernatant (migrasome-depleted conditioned medium) was used in all functional assays. For subsequent assays, each harvested Co-HMC3 sample was diluted 3:1 (v/v) with fresh complete medium to generate working conditioned medium. The target cells in the logarithmic growth phase were trypsinized, counted, and resuspended in fresh complete medium. The cell suspensions were seeded into 96-well plates at a density of 3,000 cells per well in 100 μL complete medium. After adherence, the seeding medium was aspirated and replaced with 100 μL Co-HMC3. At 24 and 48 h post-treatment, 10 μL Cell Counting Kit-8 (CCK-8) reagent (Beyotime, Shanghai, China; cat# C0038) was directly added to each well. The plates were gently agitated to mix and incubated for 2–4 h. Subsequently, the absorbance (optical density, OD) was measured at 450 nm with a microplate reader (Thermo Fisher Scientific, Waltham, MA, USA).

Cell migration and invasion assays

We pretreated 231-BR cells with freshly prepared conditioned medium for 48 h before the assays. For invasion assays, Matrigel (Sigma-Aldrich, cat# CLS354277) was diluted 1:10 in ice-cold serum-free medium and coated onto the upper chamber of 24-well Transwell inserts (Corning, cat# 3,422; 8-μm pore size) at 100 μL/well, and gelation was subsequently performed at 37°C for 6 h. For migration assays, inserts were used uncoated. Subsequently, 1 × 10⁵ pretreated cells in 100 μL serum-free medium were seeded into the upper chamber. The lower chamber contained 500 μL DMEM supplemented with 10% FBS. After 12 h of incubation at 37°C, non-migrating cells on the upper membrane surface were removed with cotton swabs. The migrating/invading cells on the lower surface were fixed with 4% PFA, washed with PBS, stained with 0.1% crystal violet (Beyotime, cat# C0121) for 15 min, and rinsed thoroughly. The membranes were excised, mounted on slides, and visualized under an inverted microscope (Olympus, Tokyo, Japan).

Chromatin immunoprecipitation assays

Chromatin immunoprecipitation (ChIP) experiments were conducted with a commercial ChIP Assay Kit (Beyotime, cat# P2078). The cells were crosslinked with 1% formaldehyde (diluted from a 37% stock) for 10 min, then quenched with 125 mM glycine. After lysis in SDS lysis buffer containing 1 mM PMSF, chromatin was fragmented by sonication into 200–1,000 bp fragments and centrifuged (14,000 × g, 10 min, 4°C). A 1% aliquot of the supernatant was reserved as the input control. The remaining lysate was precleared with Protein A/G agarose/salmon sperm DNA magnetic beads (4°C, 30 min, rotation). Pre-cleared lysates were then immunoprecipitated overnight at 4°C under rotation with either anti-EGR1 antibody or species-matched normal IgG. Antibody-chromatin complexes were captured via incubation with pretreated magnetic beads (4°C, 60 min, rotation), followed by sequential 5-min washes with low-salt buffer, high-salt buffer, LiCl buffer, and TE buffer. Complexes and input samples were eluted in ChIP elution buffer (1% SDS and 0.1 M NaHCO₃) with rotation, after which crosslinking was reversed by addition of NaCl to a 200 mM final concentration and incubation at 65°C for 4 h. Purified DNA was obtained with a DNA purification kit (Beyotime, cat# D0033), eluted in sterile water, and analyzed through PCR with primers targeting putative EGR1 binding sites in the TSPAN4 promoter.

Dual-luciferase reporter assays

A 2,100-bp fragment of the TSPAN4 promoter region (spanning positions −2,000 to +100 relative to the transcription start site, which was designated +1) was obtained from GenBank. This fragment was subsequently cloned and inserted into the GV534 luciferase reporter vector (GeneChem, Shanghai, China) to generate the TSPAN4 promoter reporter construct (designated GV534-TSPAN4). At 70%–80% confluency, cells were transfected with either GV534-TSPAN4 or empty GV534 control vector with X-tremeGENE™ 9 DNA Transfection Reagent (Sigma-Aldrich, cat# 6365779001). After 48 h, the cells were harvested, and dual-luciferase activity was measured with a Dual-Luciferase^® Reporter Assay System (Promega, cat# E1910) according to the manufacturer’s protocol. Briefly, firefly luciferase activity (promoter-driven) was measured first and was followed by measurement of Renilla luciferase activity (internal control) after quenching of the firefly signal. Luminescence was measured with a GloMax^® 96 Microplate Luminometer (Promega, Madison, WI, USA). Firefly luciferase activity was normalized to Renilla luciferase activity for each sample.

In vivo and ex vivo imaging

For bioluminescence imaging, 231-BR cells stably expressing luciferase were used. After i.p. injection of D-luciferin potassium salt (Macklin), bioluminescence signals from tumor cells were acquired with an IVIS Spectrum imaging system (PerkinElmer, Waltham, MA, USA). This approach enabled the localization, sizing, and quantification of brain metastatic foci according to fluorescence intensity.

For GFP fluorescence imaging, we used 231-BR cells stably expressing green fluorescent protein (GFP). Subcutaneous tumor growth was monitored in live animals with the IVIS Spectrum system (PerkinElmer) and a filter set optimized for GFP detection. Brain tissues were collected at the experimental endpoint and subjected to ex vivo whole-brain imaging with the same IVIS Spectrum system and GFP-optimized filter set. All acquired images were processed and analyzed in Living Image 4.4 software. Specific fluorescence signals were resolved and identified through multispectral unmixing.

Enzyme linked immunosorbent assay

After HMC3 cells were treated with Mig-BCBM, conditioned medium was collected and centrifuged at 1,000 × g for 10 min to remove cellular debris. The concentration of active TGFβ1 in the conditioned medium was quantified with a commercial human enzyme linked immunosorbent assay (ELISA) kit (Boster, cat# EK0513; specific to the bioactive form), according to the manufacturer’s instructions. Levels of IL-1β, IL-2, and TNF-α were also measured with corresponding ELISA kits.

Four-dimensional label-free quantitative proteomics

Proteomic analysis was performed as described previously²⁹, as briefly outlined below. Migrasome proteins were extracted with SDT buffer and quantified with BCA assays. Samples were digested with trypsin via the FASP method, desalted with C18 cartridges, and reconstituted. Data-dependent acquisition (PASEF mode) was performed on a nano-flow Easy nLC system coupled with a timsTOF Pro mass spectrometer. Raw data were identified and quantified with MaxQuant software. Bioinformatic analyses included clustering analysis (ComplexHeatmap), subcellular localization prediction (CELLO), domain annotation (Pfam), and protein-protein interaction network analysis (STRING, visualized with Cytoscape).

Bioinformatic analysis

Transcriptomic data from GSE125989 in the GEO database, comprising 16 BM samples, were used to identify differentially expressed genes (DEGs) between the low and high expression groups, stratified by ITGβ3 expression levels. Subsequent analyses, including immune infiltration assessment, gene set enrichment analysis (GSEA), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, were conducted on the identified DEGs, to explore their biological implications. The ssGSEA algorithm was used for immune infiltration analysis, whereas Spearman correlation was used to assess the correlation between ITGβ3 expression and immune cells. Data processing, analysis, and visualization were performed in R software (v.4.4.2) with packages including limma, clusterProfiler, GseaVis, linkET, corrplot, circlize, and IOBR. The KMplot database (https://kmplot.com/analysis) was used to perform survival analysis for overall survival via the Kaplan–Meier method.

Statistical analysis

Sample sizes for in vitro experiments were based on preliminary data and included at least three independent biological replicates per condition. For animal studies, sample size was calculated a priori in G*Power software (version 3.1) for a two-sample t-test. On the basis of pilot effect size (2-fold difference, SD = 30%), a two-tailed α = 0.05, and 80% power, the minimum required sample size was determined to be three mice per group; to account for potential loss and variability, n = 5 per group was used. All animal experiments adhered to the 3R principles. No samples or animals were excluded. Quantitative data are expressed as mean ± standard deviation (SD). Statistical analyses were performed in GraphPad Prism 9.0 (GraphPad Software, San Diego, CA) or SPSS 26.0 (IBM Corp, Armonk, NY, USA). The normality of the data distribution was assessed with the Shapiro–Wilk test; results for all major datasets are reported in Table S2. For datasets meeting normality assumptions (P ≥ 0.05), parametric tests were applied: unpaired or paired two-tailed Student’s t-test for two-group comparisons, and one-way ANOVA with Tukey’s post hoc test for multi-group comparisons. For datasets violating normality assumptions (P < 0.05), no data transformation was performed; instead, non-parametric tests were consistently applied, comprising Mann–Whitney U test for two-group comparisons, and Kruskal–Wallis test with Dunn’s post hoc test for multi-group comparisons. When ANOVA or Kruskal–Wallis tests were conducted, post hoc corrections for multiple comparisons were applied to control for type I error. Correlation analyses used Pearson (parametric) or Spearman (non-parametric) methods, as applicable. Survival curves were generated through Kaplan–Meier analysis, and statistical significance was determined via log-rank testing. The statistical significance threshold was defined as P < 0.05.

Characterization of migrasomes from BCBM cells and tissues

To determine whether brain-metastatic BC cells (231-BR) and their parental counterparts (MDA-MB-231) produce migrasomes, we isolated these vesicles through optimized density gradient centrifugation. The resulting purified migrasomes, designated Mig-BCBM (from 231-BR cells) and Mig-BC (from MDA-MB-231 cells), were characterized by transmission electron microscopy (TEM) and western blotting. We embedded the purified Mig-BCBM and Mig-BC samples and analyzed them with ultra-thin section TEM. The observed vesicles displayed classic migrasome morphology, characterized by spherical or near-spherical membrane-bound structures, containing pomegranate-like vesicles and retraction fibers (Figure 1A). Negative staining TEM further confirmed the presence of migrasomes, showing vesicles tethered to retraction fibers (Figure 1B). Western blot analysis indicated an enrichment of the migrasome-specific markers PIGK and TSPAN4 in both Mig-BCBM and Mig-BC, whereas the exosomal markers TSG101 and CD81 were absent (Figure 1C), thus confirming successful migrasome isolation.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Mig-BCBM induces M2-like polarization of microglia in vitro and in vivo. (A) Representative TEM images of ultrathin sections showing migrasomes (red arrows) derived from 231-BR cells (Mig-BCBM) or MDA-MB-231 cells (Mig-BC). Scale bars: 500 nm. (B) Representative negative-stain TEM of purified migrasomes. The right panel presents a higher-magnification view of the boxed region. Scale bars: 500 nm. (C) Western blot analysis of migrasome markers (PIGK and TSPAN4) and exosome markers (TSG101 and CD81) in 231-BR cells, MDA-MB-231 cells, and their respective migrasomes (Mig-BCBM and Mig-BC) and exosomes. Band quantification after normalization to total protein loading in ImageJ. (D) Representative ultrathin section TEM of human MBC tissue. Red arrows indicate migrasomes. Sequential magnifications (middle and right) show details of the boxed regions. Scale bars: left panel, 10 μm; middle panel, 2 μm; right panel, 500 nm. (E) Representative ultra-thin section TEM of BCBM tissue derived from model mice. Migrasomes are indicated (red arrows). Sequential magnifications (middle and right) show details of the boxed regions. Scale bars: left panel, 10 μm; middle panel, 2 μm; right panel, 500 nm. (F) Internalization of WGA-labeled Mig-BCBM (green) by human microglial HMC3 cells after 24 h co-culture. Nuclei were counterstained with DAPI (blue). Scale bars: 20 μm. (G–J) Mig-BCBM-driven microglial M2 polarization in vitro. (G, H) HMC3 cells treated with Mig-BCBM or Mig-BC (48 h) were analyzed through immunofluorescence staining showing (G) CD206⁺ (red, M2 marker) and (H) iNOS⁺ (green, M1 marker) subpopulations. Scale bars: 10 μm. (I) Western blot analysis of protein expression of CD206 and ARG1 (M2 markers), and iNOS and CD86 (M1 markers). β-Actin served as a loading control. n = 3 biological replicates. (J) Flow cytometry analysis of CD86⁺ (M1) and CD163⁺ (M2) microglia. (K) Experimental timeline: mice received tail vein injections of Mig-BCBM, Mig-BC, or PBS (control) every 3 days for four total doses over 12 days, followed by intracardiac inoculation of luciferase-expressing 231-BR cells (231-BR^Luc) on day 0 to establish brain metastasis. (L) Representative in vivo bioluminescence images (IVIS Spectrum) showing the brain metastatic burden. (M) Quantification of brain metastasis bioluminescence signal. n = 5 mice/group, biological replicates. (N) In vivo colocalization of intravenously administered WGA-labeled Mig-BCBM (green) with IBA1⁺ microglia (red) in the mouse brain parenchyma (48 h post-administration). Yellow arrows indicate colocalization. Scale bars: 10 μm. (O) Representative IBA1 immunohistochemistry of brain sections indicating activated microglia (amoeboid morphology, Mig-BCBM group) vs. surveillant microglia (ramified morphology, controls). Scale bars: left panels, 200 μm; right panels, 50 μm. (P) Representative immunofluorescence images showing colocalization of ARG1⁺ (red) and the microglia-specific marker TMEM119 (yellow), confirming M2-polarized microglia. Scale bars: 5 μm. Data represent mean ± SD. *P < 0.05, **P < 0.01, ****P < 0.0001.

To extend these observations to clinically relevant contexts, we analyzed ultra-thin TEM sections from human MBC tissues. These analyses revealed migrasomes with ultrastructural features identical to those identified in vitro, including distinct retraction filaments (Figure 1D). Similarly, examination of BCBM tissues derived from model mice confirmed the conservation of migrasome morphology, with retraction filaments present within metastatic lesions (Figure 1E). Collectively, these findings established migrasomes as consistent architectural components of BCBM across experimental, clinical, and in vivo settings.

Mig-BCBM triggers repolarization of brain-resident microglia to a pro-tumorigenic M2-like phenotype

Metastatic carcinoma cells manipulate specific stromal components within the brain microenvironment, thereby creating a tumor-permissive niche that facilitates their colonization and expansion¹². As the primary immune sentinels of the CNS, microglia play crucial roles in regulating neuroimmune surveillance and metastatic clearance. They exhibit context-dependent phenotypes that can either suppress or promote tumor progression³⁶.

After cancer cells infiltrate the brain parenchyma, microglia are recruited to metastatic sites, where they undergo phenotype specification influenced by the local microenvironment. Microglia can differentiate into either the M1 pro-inflammatory phenotype, marked by the upregulation of pro-inflammatory cytokines and anti-tumor responses, or the M2 immunosuppressive phenotype, characterized by the production of anti-inflammatory cytokines that promote tumor progression and stabilize the tumor niche¹². To investigate the effects of Mig-BCBM on microglial behavior, we conducted co-culture experiments with WGA-labeled Mig-BCBM and HMC3 microglia. After 24 h, the observed internalization of WGA-labeled Mig-BCBM (Figure 1F) confirmed successful Mig-BCBM uptake by HMC3 cells.

Subsequent phenotypic analyses of microglia were conducted with immunofluorescence staining, western blotting, and flow cytometry. Because the PBS vehicle alone did not alter microglial phenotype, we evaluated microglial responses specifically after Mig-BCBM or Mig-BC treatment. Migrasome preparations were normalized to total protein concentration (determined with BCA assays) to ensure consistent dosing across experiments. Immunofluorescence revealed significant upregulation of the M2 marker CD206 (Figure 1G), alongside downregulation of the M1 marker iNOS (Figure 1H), in microglia treated with Mig-BCBM, in contrast to those exposed to Mig-BC. Western blot analysis supported this phenotypic shift and indicated increased expression of ARG1 and CD206, and suppression of iNOS and CD86 levels, after Mig-BCBM treatment (Figure 1I). Flow cytometry further confirmed M2 reprogramming, as evidenced by an expanded population of CD163⁺ cells and a diminished subset of CD86⁺ cells in Mig-BCBM-treated microglia (Figure 1J). These findings demonstrated that the internalization of Mig-BCBM drives microglial repolarization toward an M2 immunosuppressive phenotype in vitro.

To clarify the functional role of tumor-derived migrasomes in the progression of BCBM, we established a preconditioning model in which mice received either Mig-BC or Mig-BCBM via tail vein injection before intracardiac inoculation of 231-BR^Luc cells (Figure 1K). In vivo bioluminescence imaging indicated that Mig-BCBM preconditioning significantly increased the brain metastatic burden, as evidenced by a photon flux 2.56-fold (P < 0.001) and 5.56-fold (P < 0.0001) those observed in the Mig-BC-treated and control groups, respectively (Figure 1L, M).

Further mechanistic investigations of migrasome-microglia interactions in vivo demonstrated rapid brain penetration of intravenously administered WGA-labeled Mig-BCBM. Confocal imaging confirmed colocalization with IBA1⁺ microglia in the brain parenchyma (Figure 1N). Subsequently, IBA1 immunohistochemistry indicated microglial activation in Mig-BCBM-preconditioned mice, characterized by a morphological transition from ramified surveillance states to amoeboid effector phenotypes (Figure 1O). Additionally, immunofluorescence co-staining confirmed significant expansion of ARG1⁺ microglia (TMEM119⁺) in the Mig-BCBM-treated mice (Figure 1P), indicative of M2 polarization. Collectively, these findings mechanistically integrated our in vitro observations by demonstrating that tumor-derived migrasomes accelerate BM by reprogramming resident microglia toward a tumor-promoting M2-like phenotype.

ITGβ3 is a critical regulator of migrasome-driven microglial reprogramming

Our 4D proteomic profiling comparing Mig-BCBM to Mig-BC revealed 121 upregulated proteins and 41 downregulated proteins, among which ITGβ3 and TSPAN4 were significantly elevated migrasome components (Figure 2A). TSPAN4 is a well-known migrasome marker protein, and ITGβ3 was positively correlated with TSPAN4 (Figure S1A). Integrins derived from tumor exosomes have been reported to influence organotropic metastasis patterns³⁷. To assess clinical relevance, we analyzed GEO datasets (GSE125989 and GSE41998), which indicated significantly elevated ITGβ3 expression in primary tumors from patients who developed BM (Figure S1B). Additionally, ITGβ3 expression was positively correlated with M2 phenotypes and negatively associated with M1 phenotypes (Figure S1C, D), thus suggesting its potential role in regulating microglial M2 polarization. These findings collectively provided a strong rationale for investigating ITGβ3 as a key functional candidate. Western blotting validated that TSPAN4 and ITGβ3 levels were significantly higher in Mig-BCBM than Mig-BC (Figure 2B). Additionally, digital PCR analysis demonstrated elevated ITGβ3 transcript levels in Mig-BCBM (Figure 2C).

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

ITGβ3 is selectively enriched in Mig-BCBM and drives microglial M2 polarization. (A) Heatmap of 4D label-free quantitative proteomics comparing Mig-BCBM and Mig-BC. (B) Western blot analysis of ITGβ3, TSPAN4, and PIGK in Mig-BCBM vs. Mig-BC. (C) Digital PCR quantification of ITGβ3 transcripts in Mig-BCBM vs. Mig-BC. n = 3 biological replicates. (D, E) Elevated ITGβ3 expression in 231-BR cells vs. parental MDA-MB-231 cells: (D) RT-qPCR (n = 3 biological replicates) and (E) western blot (n = 3 biological replicates). (F) IHC analysis of ITGβ3 in BCBM lesions vs. primary breast tumors and adjacent non-tumorous tissues. Left: H&E of serial sections; right: matched IHC. Scale bars: 50 μm. (G) Kaplan–Meier overall survival curves for patients with BC, stratified by high (n = 2,073) vs. low (n = 692) ITGβ3 expression levels. Data sourced from the Kaplan–Meier plotter database. (H, I) Validation of ITGβ3 knockdown in 231-BR cells: (H) RT-qPCR (n = 3 biological replicates) and (I) western blot (n = 3 biological replicates). (J) Diminished ITGβ3 protein in migrasomes isolated from ITGβ3-knockdown 231-BR cells. (Mig-sh-ITGβ3; n = 3 biological replicates). (K–L) Mig-sh-ITGβ3 attenuates M2 polarization: (K) RT-qPCR (n = 3 biological replicates) and (L) western blot (n = 3 biological replicates) demonstrating decreased expression of M2 markers (CD206, ARG1). (M) RGD peptide blockade inhibits migrasome-induced microglial M2 polarization. n = 3 biological replicates. Data represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, not significant.

To investigate whether the elevated ITGβ3 expression in Mig-BCBM might be due to its upregulation in tumor cells, we compared ITGβ3 levels in 231-BR cells vs. parental cells (MDA-MB-231). The significantly higher ITGβ3 levels (Figure 2D, E) observed in 231-BR cells suggested that cell-intrinsic upregulation enhances migrasomal loading. To further validate the clinical relevance of this upregulation, we assessed ITGβ3 expression in clinical specimens. IHC analysis revealed significantly greater ITGβ3 expression in BCBM lesions than in primary breast tumors or adjacent non-tumorous tissues (Figure 2F). To assess whether ITGβ3 upregulation in primary tumors might be associated with BM, we performed IHC. Significantly elevated ITGβ3 expression was observed in patients who developed BM (Figure S1E). Kaplan–Meier survival analysis with the Kaplan–Meier Plotter database revealed that patients with BC with high ITGβ3 expression had significantly poorer overall survival than those with low ITGβ3 expression [Figure 2G; hazard ratio (HR) = 1.28, 95% CI: 1.06–1.55, log-rank P = 0.01]. Therefore, elevated ITGβ3 expression is a risk factor for adverse prognosis in patients with BC.

To determine whether ITGβ3 might be necessary for Mig-BCBM-driven M2 polarization of microglia, we generated a lentiviral ITGβ3 knockdown system (Lv-ITGβ3) and stably transfected it into 231-BR cells. Lentiviral knockdown in 231-BR cells (Lv-shITGβ3) significantly decreased ITGβ3 levels (Figure 2H, I). Moreover, migrasomes derived from these cells (Mig-sh-ITGβ3) exhibited depletion of ITGβ3 content (Figure 2J). Importantly, Mig-sh-ITGβ3 did not induce M2 polarization, because treatment led to downregulated CD206 and ARG1 expression in microglia at both the transcript (Figure 2K) and protein levels (Figure 2L), with respect to that observed in control migrasomes (Mig-sh-NC). To investigate whether migrasomal ITGβ3 might function via surface engagement, we performed RGD peptide blocking experiments. Pre-incubation with RGD peptide before Mig-BCBM treatment significantly decreased the expression of M2 markers (CD206 and ARG1) in HMC3 microglial cells (Figure 2M); therefore, ITGβ3 surface engagement is critical for initiating microglial M2 reprogramming. Together, these findings position ITGβ3 as a key mediator of migrasome-mediated immunosuppression in the brain metastatic niche.

The ITGβ3-PI3K/AKT axis reveals a pro-metastatic immunosuppressive niche

To clarify the molecular mechanism through which migrasomal ITGβ3 induces M2 polarization of microglia in BCBM, we conducted KEGG pathway analysis on differentially expressed proteins in Mig-BCBM compared with Mig-BC. The PI3K/AKT signaling pathway was significantly enriched (Figure 3A). Furthermore, to validate these findings in an independent dataset, we analyzed the GSE125989 dataset comprising 16 paired samples of primary BCs and their matched BM. A heatmap of the DEGs is presented in Figure S2A. In agreement with our proteomic results, KEGG pathway analysis of the DEGs indicated significant activation of the PI3K/AKT signaling pathway in BM vs. primary tumors (Figure 3B). Additionally, GSEA demonstrated a positive correlation between ITGβ3 expression and PI3K/AKT activation (Figure 3C), thus supporting the established role of ITGβ3 as an upstream regulator (Figure S2B).

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

The ITGβ3-PI3K/AKT axis reveals the immunosuppressive mechanism of migrasome-educated microglia. (A) KEGG pathway analysis of differentially expressed proteins (DEPs) in Mig-BCBM vs. Mig-BC, revealing significant enrichment in PI3K/AKT signaling. (B) KEGG pathway analysis of differentially expressed genes (DEGs) in the GSE125989 dataset (16 paired primary breast cancers and matched brain metastases), indicating significant activation of the PI3K/AKT signaling pathway in brain metastases vs. primary tumors. (C) GSEA of ITGβ3 co-expressed genes. (D) Western blot analysis showing that Mig-BCBM, compared with Mig-BC, strongly promotes AKT phosphorylation at Ser473 (p-AKT Ser473) in HMC3 microglia after 48 h treatment. n = 3 biological replicates. (E) Decreased AKT phosphorylation in HMC3 cells after treatment with ITGβ3-depleted migrasomes (Mig-sh-ITGβ3; n = 3 biological replicates). (F–H) Migrasome-induced microglial M2 polarization is dependent on PI3K/AKT signaling. HMC3 cells were pretreated with the PI3K inhibitor LY294002 or vehicle control, then stimulated with Mig-BCBM. (F) Western blot analysis indicating attenuated Mig-BCBM-induced AKT phosphorylation after LY294002 pretreatment (n = 3 biological replicates). (G) RT-qPCR and (H) western blot analyses demonstrating that LY294002 pretreatment blocks Mig-BCBM-mediated upregulation of M2 markers (CD206 and ARG1) and abrogates the suppression of iNOS expression (n = 3 biological replicates). (I) Time-course analysis of ITGβ3 expression and PI3K/AKT activation in HMC3 cells after Mig-BCBM treatment. n = 3 biological replicates. (J) Correlation between CD163 expression and the abundance of immune cell populations in the GSE43837 dataset. (K) Correlation analysis (GSE43837) between ITGβ3 and the expression of key immunosuppressive checkpoint genes (CTLA4, PDCD1LG2, VTCN1, CD47, SIGLEC9, IDO1, TNFRSF14, PDCD1, and LAG3). Data represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Functional validation indicated that Mig-BCBM enhanced AKT phosphorylation at Ser473 (p-AKT) in HMC3 microglia compared with Mig-BC (Figure 3D). In contrast, ITGβ3-depleted migrasomes (Mig-sh-ITGβ3) significantly decreased the ability to induce AKT phosphorylation (Figure 3E); therefore, migrasomal ITGβ3 is essential for activating the AKT pathway in HMC3 microglia.

To further confirm the role of the PI3K/AKT pathway in mediating Mig-BCBM-induced M2 polarization of microglia, we pretreated HMC3 microglia with the PI3K inhibitor LY294002 before Mig-BCBM stimulation. Mig-BCBM-induced AKT phosphorylation was markedly attenuated by LY294002 pretreatment (Figure 3F). RT-qPCR analysis indicated that LY294002 pretreatment blocked the upregulation of CD206 and the downregulation of iNOS induced by Mig-BCBM (Figure 3G). In agreement with these findings, western blot analysis confirmed that LY294002 pretreatment abrogated Mig-BCBM-induced CD206 and ARG1 upregulation and iNOS suppression (Figure 3H). To further elucidate the expression dynamics of ITGβ3 in microglia after migrasome uptake, we measured ITGβ3 levels in HMC3 cells at various time points. ITGβ3 protein expression was significantly elevated at 6 and 12 h post-treatment vs. baseline, and this sustained presence correlated with enhanced PI3K-AKT pathway phosphorylation during the same time course (Figure 3I). Collectively, these results demonstrated that migrasomal ITGβ3 promotes M2 polarization of microglia through activation of the PI3K/AKT signaling pathway.

To explore the effects of migrasomal ITGβ3-mediated microglial M2 polarization on the immunosuppressive TME, we analyzed transcriptomic data from the GSE43837 dataset. This analysis identified positive correlations between CD163/ITGβ3 expression and the presence of various immunosuppressive cell populations, such as myeloid-derived suppressor cells, mast cells, and neutrophils (Figures 3J and S2C).

Moreover, ITGβ3 expression positively correlated with the expression of key immunosuppressive checkpoint genes, including CTLA4, PDCD1LG2, VTCN1, CD47, SIGLC9, IDO1, TNFRSF14, PDCD1, and LAG3 (Figure 3K). These data collectively indicated that migrasomal ITGβ3 is a key regulator of PI3K/AKT-dependent microglial M2 polarization. This regulation actively contributes to creating an immunosuppressive niche facilitating the progression of BM.

TGFβ1/Smad2/3-dependent EMT promoted by M2-like microglia facilitates BCBM progression

To investigate whether M2-polarized microglia in the brain metastatic microenvironment might influence the proliferation and migration of BCBM cells, we treated 231-BR cells with Co-HMC3. CCK-8 assays demonstrated that Co-HMC3 significantly enhanced the proliferation of 231-BR cells (Figure 4A), whereas Transwell assays demonstrated increased migration and invasion capabilities (Figure 4B). Validation in a microglia-tumor co-culture system (Figure 4C) confirmed that supernatant from Mig-BCBM-treated microglia promoted the migration of 231-BR cells (Figure 4D). Therefore, Mig-BCBM facilitates metastasis through the induction of microglial M2 polarization.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

M2-polarized microglia promote metastatic progression via TGFβ1/Smad2/3-induced EMT. (A) CCK-8 proliferation assays of 231-BR cells treated with conditioned medium from Mig-BCBM-polarized microglia (Co-HMC3; n = 3 biological replicates). (B) Transwell migration and invasion assays of Co-HMC3-treated 231-BR cells. Scale bars: 50 μm. n = 3 biological replicates. (C) Schematic illustration of the microglia-tumor cell co-culture system. (D) Enhanced migration of 231-BR cells exposed to Mig-BCBM-polarized microglial supernatant. Scale bars: 50 μm. n = 3 biological replicates. (E) Representative ELISA results of TGFβ1 levels in supernatants from Mig-BCBM-treated vs. Mig-BC-treated microglia. n = 3 biological replicates. (F) Correlation analysis between TGFβ1 and vimentin/E-cadherin expression in the TCGA-BRCA cohort (n = 1,100). (G) Western blot analysis of vimentin and E-cadherin expression in Co-HMC3-treated 231-BR cells. n = 3 biological replicates. (H) The TGF-β receptor inhibitor galunisertib reverses Co-HMC3-induced vimentin elevation and E-cadherin suppression in 231-BR cells. n = 3 biological replicates. (I) Pre-incubation of Co-HMC3 with TGFβ1 neutralizing antibody (metelimumab) restores E-cadherin and suppresses vimentin, whereas isotype control IgG shows no effect. n = 3 biological replicates. (J) Schematic diagram showing the TGFβ signaling pathway. (K) Western blot analysis of Smad2/3 phosphorylation in Co-HMC3-treated 231-BR cells. n = 3 biological replicates. (L) Nuclear translocation of Smad2/3 in 231-BR cells after treatment with Co-HMC3. Nuclear and cytoplasmic fractions were isolated at the indicated time points. n = 3 biological replicates. (M) Galunisertib abrogates Co-HMC3-induced Smad2/3 phosphorylation in 231-BR cells. n = 3 biological replicates. (N) The Smad2/3 inhibitor NCB-0846 suppresses 231-BR cell migration, whereas TGFβ1 partially rescues this inhibition. n = 3 biological replicates. (O) NCB-0846 downregulates p-Smad2/3 and vimentin while upregulating E-cadherin; TGFβ1 counteracts these effects. n = 3 biological replicates. Data represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, not significant.

To identify the specific factors in M2 microglia-conditioned medium that mediate 231-BR cell proliferation, invasion, and migration, we analyzed cytokine expression in the supernatant of Mig-BCBM-polarized microglia. TGFβ1 levels were significantly greater in the supernatants from Mig-BCBM-treated microglia than Mig-BC-treated microglia (Figure 4E). In contrast, the levels of IL-1β, IL-2, and TNF-α did not show significant differences (Figure S3A–C).

Correlation analysis revealed a positive correlation between TGFβ1 and vimentin levels, alongside an inverse correlation with E-cadherin (CDH1) expression (Figure 4F). These findings indicated that TGFβ1 induces EMT. Western blot analysis confirmed that Co-HMC3 treatment upregulated vimentin and downregulated E-cadherin in 231-BR cells (Figure 4G), thus further indicating EMT induction. To establish the causal role of TGFβ1, we pretreated 231-BR cells with the TGFβ receptor inhibitor galunisertib before Co-HMC3 exposure. Galunisertib effectively abrogated Co-HMC3-mediated vimentin elevation and E-cadherin suppression (Figure 4H). Therefore, TGFβ1 is a key mediator of microglia-induced EMT. To further determine whether microglia-derived TGFβ1 might be directly responsible, we performed neutralization experiments with a TGFβ1-specific antibody (metelimumab). Pre-incubation of Co-HMC3 conditioned medium with metelimumab, but not isotype control IgG, reversed the EMT phenotype in 231-BR cells (Figure 4I) and significantly decreased their migration and invasion (Figure S4A–C). These findings confirmed that microglia-derived TGFβ1 is necessary for both EMT induction and pro-metastatic functional effects.

TGFβ signaling diagrams indicated that TGFβ1 binds the type II receptor (TGFβRII), which phosphorylates and activates the type I receptor (TGFβRI). The activated type I receptor then phosphorylates Smad2 and Smad3, and consequently regulates cell phenotype (Figure 4J). We hypothesized that TGFβ1 secreted by Mig-BCBM-induced M2 microglia might regulate EMT in 231-BR cells through Smad2/3 signaling. Co-HMC3-treated 231-BR cells exhibited increased phosphorylation of Smad2 and Smad3, thus indicating activation of the Smad2/3 pathway (Figure 4K). To further confirm pathway activation, we performed nuclear-cytoplasmic fractionation and observed time-dependent nuclear translocation of Smad2/3, peaking at 3 h and returning to baseline by 24 h (Figure 4L). Rescue experiments indicating that galunisertib reversed Co-HMC3-induced Smad2/3 phosphorylation (Figure 4M) confirmed that M2 microglia activate Smad2/3 signaling via TGFβ1.

To further establish that TGFβ1 orchestrates migration and EMT through Smad2/3 signaling, we conducted rescue experiments using the Smad2/3 inhibitor NCB-0846. Transwell migration assays demonstrated that NCB-0846 significantly suppressed 231-BR cell migration, whereas co-treatment with TGFβ1 reversed this suppression (Figure 4N). Western blot analysis revealed that TGFβ1 countered NCB-0846-induced E-cadherin upregulation and vimentin downregulation (Figure 4O). Collectively, these data establish that M2-polarized microglia enhance 231-BR cell metastasis through TGFβ1-mediated EMT via activation of the Smad2/3 pathway.

TGFβ1 establishes a self-reinforcing feedback loop by enhancing migrasome biogenesis and ITGβ3 packaging

After having established that 231-BR cells induce M2 microglial polarization through migrasomal ITGβ3, and that these polarized M2 microglia subsequently promote EMT in 231-BR cells via TGFβ1-Smad2/3 signaling, we aimed to explore the regulatory mechanisms involved in migrasome formation in 231-BR cells. Correlation analysis revealed positive associations between TGFB1 expression and key regulators of migrasome biogenesis, including TSPAN4, TSPAN7, and RAB35 (Figures 5A and S5A, B). On the basis of the established roles of these regulators in migrasome formation^19,38,39, we hypothesized that TGFβ1 might regulate migrasome production. Treatment of 231-BR cells with increasing concentrations of TGFβ1 (0–10 ng/mL) resulted in a dose-dependent increase in migrasomal protein content, as quantified with BCA assays (Figure 5B), thus indicating that TGFβ1 promotes migrasome production. At 10 ng/mL TGFβ1, the migrasome markers TSPAN4 and PIGK were upregulated, while TSG101 expression remained unchanged (Figure 5C), thereby confirming the specific induction of migrasomes. Morphological validation with TEM revealed a significant increase in the quantity of migrasomes after TGFβ1 treatment (Figure 5D). In agreement with these findings, WGA staining also demonstrated enhanced migrasome formation (Figure 5E, F). Furthermore, confocal-based super-resolution fluorescence imaging provided enhanced resolution and enabled more reliable visualization and quantification of migrasomes. The findings further confirmed the induction of migrasome biogenesis by TGFβ1 (Figure 5G).

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

TGFβ1 promotes migrasome biogenesis and ITGβ3 packaging. (A) Correlation analysis between TGFB1 expression and the migrasome biogenesis regulator TSPAN4 in the TCGA-BRCA cohort (n = 1,100). (B) Dose-dependent effect of TGFβ1 (0–10 ng/mL) on migrasomal protein levels. Protein concentrations were determined via BCA assays and normalized to equivalent cell numbers. n = 6 biological replicates. (C) Western blot analysis of migrasome markers (TSPAN4 and PIGK) after TGFβ1 treatment. TSG101 served as an exosome marker control. (D) Representative TEM images of ultrathin sections demonstrating TGFβ1-induced migrasome formation in 231-BR cells. Red arrows indicate migrasomes. The right panel shows a magnified view of the boxed region in the left panel. Scale bars: left, 2.0 μm; right, 500 nm. n = 5 biological replicates. (E, F) Confocal microscopy of TGFβ1-induced migrasomes: (E) WGA (green) and phalloidin (red) staining. Red ellipses: representative migrasomes. Scale bar: 20 μm. (F) WGA (green) and TSPAN4 (red) staining. Red ellipses highlight representative migrasomes. Quantification of extracellular migrasomes according to WGA staining intensity within regions of interest (ROIs). n = 3 biological replicates. Scale bars: 20 μm. (G) Confocal-based super-resolution fluorescence imaging of migrasomes (red arrows) in 231-BR cells after TGFβ1 treatment. Enhanced resolution imaging enables reliable visualization and quantification of migrasomes. Scale bars: 10 μm. n = 50 technical replicates. (H) TGFβ1 co-treatment reverses the suppression of migrasomal protein levels induced by the migrasomal inhibitor EHT-1864. Protein concentrations were determined with BCA assays and normalized to equivalent cell numbers. n = 3 biological replicates. (I) Western blot analysis indicating that TGFβ1 co-treatment restores PIGK and TSPAN4 expression suppressed by EHT-1864. (J) Western blot analysis of TGFβ1-induced Smad2/3 phosphorylation and TSPAN4 expression in 231-BR cells. n = 3 biological replicates. (K) TGFβ1 treatment upregulates ITGβ3 protein in BCBM-derived migrasomes. (L) Rescue experiment: TGFβ1 reverses the EHT-1864-induced downregulation of ITGβ3 expression in migrasomes. (M) Decreased TGFβ1 secretion by microglia treated with ITGβ3-knockdown migrasomes (Mig-sh-ITGβ3; ELISA). n = 3 biological replicates. Data represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Rescue experiments demonstrated that the migrasome inhibitor EHT-1864 suppressed migrasomal protein levels, whereas co-treatment with TGFβ1 reversed this inhibition (Figure 5H). Additionally, western blot analysis demonstrated that EHT-1864 decreased TSPAN4 and PIGK expression, whereas TGFβ1 restored their protein levels (Figure 5I).

In agreement with a previous report that TGFβ1/Smad2/3 signaling induces TSPAN4 expression and migrasome production in retinal pigmented epithelium cells²², we observed that TGFβ1 also induced Smad2/3 phosphorylation and TSPAN4 expression in 231-BR cells (Figure 5J). These findings implicate the TGFβ1/Smad2/3 pathway in migrasome induction.

We further assessed the effects of TGFβ1 on the loading of ITGβ3 into migrasomes. Treatment with TGFβ1 significantly increased ITGβ3 expression in Mig-BCBM (Figure 5K). In contrast, EHT-1864 decreased migrasomal ITGβ3 levels; however, this suppression was reversed with TGFβ1 co-administration (Figure 5L). Additionally, knockdown of ITGβ3 in migrasomes (Mig-sh-ITGβ3) resulted in decreased TGFβ1 secretion from microglia after migrasome treatment (Figure 5M). These findings demonstrated that TGFβ1 promotes migrasome biogenesis and ITGβ3 packaging, thus establishing a self-reinforcing feedback loop between BCBM cells and microglia that is mediated by migrasomal ITGβ3 and TGFβ1 and accelerates metastatic progression.

TGFβ1/Smad2/3 signaling upregulates TSPAN4 via transcriptional activation by EGR1

To assess whether TGFβ1 might transcriptionally upregulate TSPAN4, we screened six transcription factor databases (Harmonizome 3.0, JASPAR, hTFtarget, GTRD, ChIP-Atlas, and TFDB) and identified EGR1 and YY1 as potential regulators of TSPAN4 (Figure 6A). Correlation analysis revealed positive associations between EGR1 and both TGFB1 and TSPAN4 levels (Figure 6B), whereas YY1 exhibited a negative correlation (Figure S6A, B). On the basis of these findings, we focused on the regulatory role of EGR1. Compared with the sh-NC control, lentivirus-mediated EGR1 knockdown (sh-EGR1) significantly decreased both EGR1 mRNA (Figure 6C) and protein levels (Figure 6D) in 231-BR cells. This knockdown also led to a concomitant downregulation of TSPAN4 expression (Figure 6E, F), thereby establishing EGR1 as a regulator of TSPAN4. Furthermore, because TGFβ1 treatment partially mitigated the sh-EGR1-induced suppression of TSPAN4 expression (Figure 6G), TGFβ1 was found to exert its effects at least partly through EGR1.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

TGFβ1 promotes TSPAN4 expression through EGR1-dependent transcriptional activation. (A) Prediction of potential TSPAN4 transcriptional regulators with six databases (Harmonizome 3.0, JASPAR, hTFtarget, GTRD, ChIP-Atlas, and TFDB). (B) Positive correlations between EGR1 and TGFB1/TSPAN4 expression in the TCGA-BRCA cohort (n = 1,100). (C, D) EGR1 knockdown (sh-EGR1) in 231-BR cells decreases (C) EGR1 mRNA (RT-qPCR) and (D) EGR1 protein (western blot) levels. n = 3 biological replicates. (E, F) sh-EGR1 downregulates (E) TSPAN4 mRNA (RT-qPCR) and (F) TSPAN4 protein (western blot) expression. n = 3 biological replicates. (G) TGFβ1 partially reverses sh-EGR1-induced TSPAN4 suppression. n = 3 biological replicates. (H) ChIP assays confirming EGR1 binding at the TSPAN4 promoter region. IgG served as a negative control. (I) sh-EGR1 decreases TSPAN4 promoter activity (dual-luciferase assay; n = 5 biological replicates). (J) Galunisertib suppresses Smad2/3 phosphorylation and EGR1 expression; TGFβ1 co-treatment restores these effects (western blot; n = 3 biological replicates). (K) Super-resolution confocal microscopy showing less migrasome formation in sh-EGR1-treated cells than sh-NC control cells. Scale bars: 10 μm. n = 50 technical replicates. (L) Western blot analysis of ITGβ3 protein levels in migrasomes isolated from sh-NC and sh-EGR1 231-BR cells. (M) Western blot analysis showing lower expression of M2 markers (CD206 and ARG1) in the Mig-sh-EGR1 group than the control group. n = 3 biological replicates. Data represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Mechanistically, a ChIP assay revealed direct binding of EGR1 to the TSPAN4 promoter, with the strongest enrichment observed in the −700 to −350 bp (P5) region (Figure 6H). Dual-luciferase reporter assays confirmed that EGR1 knockdown significantly decreased TSPAN4 promoter activity (Figure 6I). Furthermore, rescue experiments demonstrated that TGFβ1 treatment counteracted the galunisertib-induced suppression of Smad2/3 phosphorylation and EGR1 expression (Figure 6J). Collectively, these findings indicated that the TGFβ1/Smad2/3 signaling axis promotes TSPAN4 expression through EGR1-dependent transcriptional activation in 231-BR cells.

To functionally validate the role of EGR1 in the proposed feedforward loop, we examined migrasome production and ITGβ3 loading after EGR1 knockdown. Super-resolution confocal microscopy revealed that EGR1 knockdown significantly decreased migrasome formation (Figure 6K). Western blot analysis further showed that migrasomes derived from EGR1-knockdown cells (Mig-sh-EGR1) contained markedly lower levels of ITGβ3 protein than control migrasomes (Figure 6L). Functionally, Mig-sh-EGR1 also exhibited a diminished ability to induce M2 polarization in microglia, as reflected by attenuated expression of M2 markers (CD206 and ARG1) (Figure 6M). Together, these results demonstrated the essential roles of EGR1 in migrasome biogenesis, ITGβ3 loading, and subsequent microglial reprogramming.

Therapeutic validation of the migrasomal ITGβ3-TGFβ1 loop by galunisertib in preclinical models

Galunisertib, a selective TGF-β/Smad pathway inhibitor, was assessed in BC models with subcutaneous xenograft and intracardiac metastasis assays. Nude mice received two i.p. pre-treatment injections of galunisertib, followed by subcutaneous co-implantation of 231-BR^GFP cells pre-treated with galunisertib and Mig-BCBM-primed HMC3 microglia. These mice were then maintained on i.p. injections every 3 days (Figure 7A). By day 21, in vivo imaging showed a significantly lower total photon flux in galunisertib-treated tumors than control tumors (Figure 7B). This finding was accompanied by decreases in tumor volume and weight, thus confirming growth inhibition (Figure 7C, D). Importantly, an absence of significant changes in body weight or organ indices indicated favorable biosafety (Figure 7E, F). Mechanistically, the downregulation of TSPAN4 expression in galunisertib-treated tumors suggested that TGFβ1 blockade disrupted migrasome regulation (Figure 7G).

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Galunisertib inhibits tumor growth and brain metastasis without causing systemic toxicity. (A) Experimental timeline for the tumor growth study: nude mice received two pre-treatment i.p. injections of galunisertib followed by subcutaneous co-implantation of 231-BR^GFP cells pre-treated with galunisertib and Mig-BCBM-primed HMC3 microglia, with maintenance i.p. injections every 3 days. (B) Representative in vivo fluorescence images and quantitative analysis of total photon flux in the galunisertib vs. vehicle control groups. n = 5 mice/group, biological replicates. (C) Ex vivo tumors from treatment cohorts. (D) Decreased tumor weight after galunisertib treatment. n = 5 mice/group, biological replicates. (E) Comparable body weights, indicating no systemic toxicity. n = 5 mice/group, biological replicates. (F) Organ-to-body weight ratios, showing no treatment-related abnormalities. n = 5 mice/group, biological replicates. (G) IHC results showing decreased expression of TSPAN4 in tumor tissues. Scale bars: 50 μm. (H) Experimental timeline for the brain metastasis study: intracardiac injection of 231-BR^GFP cells with q3d i.p. galunisertib administration. (I) Ex vivo brain imaging (left) and metastatic burden quantification (right), demonstrating galunisertib efficacy. n = 5 mice/group, biological replicates. (J) H&E-stained brain sections, showing decreased metastatic nodules (red arrows) in the galunisertib group vs. the control group. Scale bars: 1,000 μm (overview), 50 μm (insets). Nodule quantification from ten sections/group. (K) Normal histology of major organs (heart, liver, spleen, lung, and kidneys), according to H&E staining. Scale bars: 50 μm. Data represent mean ± SD. *P < 0.05, **P < 0.01; ns, not significant.

To assess BM, we established an intracardiac 231-BR^GFP injection model, with galunisertib i.p. injections every 3 days (Figure 7H). Ex vivo brain imaging on day 28 revealed a lower metastatic burden in the galunisertib cohort than the control cohort (Figure 7I). H&E-stained brain sections confirmed a decrease number of metastatic foci in the Galunisent-treated group compared with the control group (Figure 7J), whereas a thorough organ histopathology analysis showed no drug-induced damage (Figure 7K). Collectively, these findings established galunisertib as an effective therapeutic agent that inhibits the progression of BCBM without causing systemic toxicity.