Integrin-Defined Plasma Extracellular Vesicles for Early Breast Cancer Detection and Bone Metastasis Association

Abstract

Sensitive detection of breast cancer and metastatic disease remains critical for improving patient outcomes. Plasma extracellular vesicles (EVs) carrying integrins have shown strong promises for characterizing metastatic progression and organotropism, but their clinical potential remains largely unexplored, particularly for early detection. Here, we applied a gold nanoparticle-based single vesicle imaging method to investigate the feasibility of plasma EV-associated integrins as biomarkers for early breast cancer detection and bone metastasis association. EVs were immunocaptured via the tetraspanin CD81 and labeled with membrane dyes and antibody/AuNPs targeting integrin monomers αV, β3, α6, or β5, enabling quantification of total captured EVs with fluorescence imaging and the target-specific EV subpopulations via dark field imaging. Single vesicle analysis revealed that the fraction of EVs that were positive for integrin αV was significantly elevated in Stage I patients and was further enriched in patients with bone metastasis relative to healthy donors. A similar pattern was observed for integrin β3-positive EVs. In contrast, integrins α6 and β5 showed no significant differences across groups. Receiver operating characteristic analysis revealed strong diagnostic performance for αV and β3, with area under the curve (AUC) values exceeding 0.95 for both markers in distinguishing bone-metastatic disease from healthy controls. Further, integrin αV demonstrated higher sensitivity (AUC = 0.88) for early-stage detection than integrin β3 (AUC = 0.74), while both showed similar performance in distinguishing bone-metastatic disease. These findings suggest that EV-associated integrins αV and β3 are associated with breast cancer presence and progression, with integrin αV showing potential as a biomarker for early detection. This proof-of-concept feasibility study establishes a foundation for mechanistic and translational validation of EV-associated integrins as promising biomarkers in liquid biopsy assays for clinical applications.

Graphical Abstract

1. INTRODUCTION

Breast cancer (BC) remains the most prevalent malignancy and the second leading cause of cancer-related deaths among women in the United States and many other countries worldwide [1]. Approximately one in eight women in the United States will develop BC during their lifetime [2]. Early detection is vital for effective treatment, with the 5-year survival rate of ~99% for patients diagnosed at a localized stage in contrast to ~32% for those with metastatic disease [3]. Mammography is the current standard method for early BC detection. However, it has well-recognized limitations, including reduced sensitivity in women with dense breast tissue, false positives that lead to unnecessary follow-up procedures, and false negatives that delay diagnosis [4–6]. Thus, there is a critical need for new approaches for early BC detection. Importantly, metastasis is the major driver of morbidity and mortality. Metastatic spread is organotrophic rather than random [7]. In BC, bone is one of the most common sites of metastasis, highlighting the need to identify biomarkers associated with bone metastasis.

Extracellular vesicles (EVs) are nano-sized membrane-bound particles actively released by virtually all cell types, consisting of exosomes and microvesicles [8–10]. EVs carry nucleic acids and various proteins that are reflective of their parental cells [11–13]. In addition, EVs are stable under physiological conditions and are present in blood plasma and many other biofluids such as urine and saliva [14–16]. EVs have therefore merged as promising biomarkers for minimally invasive cancer diagnostics, prognostics, and treatment monitoring [17–21]. Tumors release EVs even at early stages of cancer development, positioning circulating EVs as a promising source of biomarkers for early cancer detection [22–24]. Surface proteins on EVs are particularly attractive to biomarkers because they can be analyzed directly without disrupting vesicle structure. Indeed, profiling HER2 or EGFR expression on plasma EVs has been shown to distinguish early-stage HER2-positive BC from healthy individuals [25,26].

Integrins are transmembrane receptors composed of 18 α subunits and 8 β subunits, which pair to form at least 24 functional heterodimers [27]. They physically connect cells to the extracellular matrix (ECM) and regulate adhesion, migration, proliferation, and survival through outside-in and inside-out signaling [28]. Dysregulated integrin expression is a hallmark of cancer progression and contributes to nearly every step of tumor evolution [29]. Several integrin subtypes, including αVβ3, αVβ5, α5β1, and α6β4, are known to promote tumor aggressiveness and metastatic dissemination across multiple solid tumors such as breast, lung, and prostate cancers [30]. Tumor type or disease status can be classified by integrin expressions [31]. Among the various functional dimers, the integrin αVβ3 is especially notable for its role in bone metastasis [32]. Upregulation of αVβ3 enhances tumor cell motility and invasiveness and facilitates interactions with the bone microenvironment, and correspondingly elevated αVβ3 activity is associated with increased bone metastatic potential and poorer clinical outcomes [33–36].

Importantly, integrins are also enriched on the surface of tumor-derived EVs, where their expression is closely associated with disease state [37–40]. For example, integrins α3 and β1 were found to be more abundant in urinary EVs from metastatic prostate cancer patients than in those from non-metastatic disease [37]. A landmark study by Hoshino et al. further demonstrated that specific integrin signatures on tumor EVs dictate organ-specific metastatic behavior [41]. They reported that exosomal integrins α6β4 and α6β1 were enriched in lung-tropic tumors, and αVβ5 was abundant in liver-tropic tumors.

Despite advances in characterizing EV-associated integrins for metastatic progression and organotropism, their clinical potential remains largely unknown. In this study, we applied a single vesicle technology (SVT) to profile several candidate integrin markers on plasma EV samples from healthy donors and BC patients diagnosed at Stage I and Stage IV with bone metastasis. Our results show that CD81-positive plasma EVs carrying integrins αV and β3 are significantly elevated in BC patients of in both Stage I and Stage IV groups relative to healthy controls, whereas integrins α6 and β5 show no such increase. Integrin αV demonstrated higher sensitivity for early-stage detection while both showed similar performance in distinguishing bone-metastatic disease from healthy controls. These findings highlight the potential of EV-associated integrins αV and β3 as candidate minimally invasive biomarkers for BC detection and association with bone-tropic metastasis.

2. EXPERIMENTAL SECTION

2.1. Materials.

All chemicals were purchased from Millipore-Sigma (Burlington, MA) unless otherwise specified otherwise. Custom gold coated (100 nm in thickness) prime grade 4 inch silicon wafer was purchased from Angstrom Engineering Inc (Cambridge, ON, Canada). Purified anti-human CD81 antibody (clone 5A6), anti-human integrin αV antibody (clone NKI-M9), anti-human integrin α6 antibody (clone GoH3), anti-human integrin β3 antibody (clone VI-PL2), anti-human integrin β5 antibody (clone AST-3T), and anti-human CD9 antibody (clone HI9a) were purchased from BioLegend (San Diego, CA). Methoxy polyethylene glycol thiol (mPEG-SH, MW = 5000, Purity: >95%), N-hydroxysuccinimide polyethylene glycol thiol (NHS-PEG-SH, MW = 1000), and cholesterol-PEG-Cy5 (Chol-PEG-Cy5, MW = 2000) were purchased from Nanocs (New York, NY). Fetal bovine serum (FBS), MDA-MB-231, MCF-7, and SK-BR-3 were purchased from ATCC (Manassas, VA). Goat anti-mouse IgG conjugated with horseradish peroxidase (HRP), isotype IgG control, bovine serum albumin (BSA), and sodium pyruvate were purchased from ThermoFisher Scientific (Waltham, MA). Dulbecco’s modified Eagle’s medium (DMEM) with high glucose (4.5 g/L), Rosewell Park Memorial Institute (RPMI) 1640 medium, 0.25% trypsin, penicillin-streptomycin (Pen/Strep), and 10 kDa molecular weight cutoff (MWCO) centrifugal filter were purchased from Avantor (Allentown, PA). 60-nm gold nanoparticles (AuNPs) were purchased from Ted Pella (Redding, CA).

2.2. Source of Plasma EVs.

Human plasma samples from healthy donors (n = 13) were obtained from BioIVT (Westbury, NY, USA). Plasma samples from Stage I BC patients (n = 19) and bone-metastatic Stage IV BC patients (n = 14) were obtained from Indiana University Melvin and Bren Simon Comprehensive Cancer Center. The samples were not collected specifically for our study, and no identifying information was accessible to us. Under U.S. Department of Health & Human Services regulations, this work is classified as non–human subjects’ research.

2.3. EV Isolation and Characterization.

EVs were purified using 35 nm qEVoriginal columns (Izon Science, Medford, MA) according to the manufacturer’s protocol. Briefly, the column was rinsed with 17 mL of phosphate buffered saline (PBS) and then loaded with a 500 μL plasma. After the sample completely entered the column, 2 mL of PBS was added and allowed to drain into a waste container. Then, samples were collected following a sequential addition of 1.6 mL and 2.0 mL of PBS. Following purification, EVs were concentrated using a VWR centrifugal filter with a 10 kDa molecular weight cutoff at 13,800 × g for 15 minutes (min) at room temperature (RT) (Centrifuge 5415 C, Eppendorf, Hamburg, Germany). The purified samples were stored at −80 °C and used within 2 weeks. Size and concentration of purified EVs were determined via Nanoparticle Tracking Analysis (NTA) using a NanoSight LM10 microscope (Malvern Instruments, Westborough, MA). EV purity and morphology were characterized by scanning electron microscope (SEM) using a Nova NanoSEM650 field emission SEM. For SEM imaging, purified EVs (1 × 10⁹/mL) from a Stage I BC patient were firstly fixed with 1% glutaraldehyde for 30 min at RT. Then, 40 μL of the fixed EV solution was spotted onto a clean silicon chip and incubated for 45 min at RT. After removing the liquid, the adsorbed EVs were sequentially washed with PBS and ultrapure water, three times each. EVs were subsequently dehydrated through a graded ethanol series of 30%, 50%, 70%, 90%, 100%, and 100% ethanol, with each step performed for 8 min. Finally, the sample was air-dried for 2 h before SEM imaging. EVs were sputter-coated with a 2–5 nm gold film and imaged by SEM at an accelerating voltage of 5 or 10 kV at different magnifications.

EVs were further characterized using indirect enzyme-linked immunosorbent assay (ELISA) to determine the expression of tetraspanins CD63 and CD9. In a typical procedure, 100 μL of 2 × 10⁹/mL purified EVs was added to a 96-well polystyrene plate (Corning Incorporated, Corning, NY) in duplicate and incubated overnight at 4 °C to allow EV adsorption. The wells were then washed three times with PBS and blocked with 200 μL of blocking solution (PBS with 5% BSA) for 2 h at RT. After washing three times with PBS, 50 μL of 5 μg/mL anti-human monoclonal CD63 antibody, CD9 antibody, or IgG control (diluted in PBS with 1% BSA) was added to each well and incubated for 2 h. After washing three times with PBS, 50 μL of goat anti-mouse IgG-HRP solution, diluted in 1:1000 in PBS with 1% BSA, was added to each well and incubated for 2 h at RT. After washing as described above, 100 μL of 3,3′,5,5′-tetramethylbenzidine (TMB) substrate was added to each well and incubated for 25 min at 37 °C. The reaction was stopped by adding 100 μL of 1 M sulfuric acid with 5 min incubation. Absorbance was measured at 450 nm using a BioTEK ELx800 microplate reader.

2.4. Cell Culture, Protein Extraction, and Mass Spectrometry (MS)-based Characterization of Cellular Integrin Expression.

MDA-MB-231, MCF-7, and SK-BR3 Cells were cultured in their respective medium (DMEM with high glucose for MDA-MB-231 & MCF-7 and RPMI 1640 for SK-BR-3) supplemented with 10% FBS, 1% Pen/Strep, and 1% sodium pyruvate at 37 °C under 5% CO₂. Integrin expression on the cancer cells was determined using liquid chromatography–tandem mass spectrometry (LC–MS/MS) following standard protocols. First, cellular proteins were extracted using the ThermoFisher Scientific EasyPrep MS Sample Prep Kits according to the manufacturer’s standard protocol. Lysates were centrifuged at 14,000 × g for 10 min at 4 °C. The supernatants were collected and protein concentrations were determined using the Pierce^™ Bicinchoninic Acid (BCA) Protein Assay according to the manufacturer's instructions. The collected cellular proteins were stored at −80 °C before MS measurements.

Proteins were reduced, alkylated, and enzymatically digested according to established protocols at The University of Tennessee Healthy Science Center (UTHSC) Proteomics & Metabolomics Core. Resulting peptides were analyzed by label-free nanoflow LC-MS/MS in data-dependent acquisition (DDA) mode on an Orbitrap Fusion Lumos mass spectrometer operated with Xcalibur 4.3 and coupled to an Ultimate 3000RSLCnano ultra-high pressure liquid chromatography system (ThermoFisher Scientific). Data were also post-acquisition processed at the core facility for peptide/protein identification and relative quantification, generating validated protein abundance datasets ready for downstream biostatistical and bioinformatics analysis. All experiments were performed in five biological replicates. Protein abundance data were analyzed using standard bioinformatics pipelines to generate normalized protein-level quantifications. Proteins with missing values across samples in each batch were removed. Protein abundance was log2-transformed, missing values imputed via Bayesian Principal Component Analysis (PCA) (bpca R package), and batch effects corrected using ComBat (sva R package).

2.5. Preparation and Characterization of Target-Specific Antibody-Conjugated AuNPs (Antibody/AuNPs).

Target-specific anti-integrin antibodies were linked to AuNPs following the procedure described in our recent studies [25]. Briefly, anti-human monoclonal mouse antibodies (20 μL, 0.5 mg/mL) were first buffer-exchanged into carbonate/bicarbonate buffer (pH 9.0) by centrifugal filtration using a 10 kDa MWCO filter (13,800 × g, 4 min). Then, the antibodies were functionalized with free thiol (–SH) groups by reacting with 100-fold molar excess of NHS-PEG-SH (MW = 1000) for 2 h at 37 °C, followed by purification vial centrifugal filtration using a 10 kDa MWCO centrifugal filter. This optimal thiolation condition was determined by measuring the number of –SH groups per protein molecule using Ellman's assay at different reaction temperatures and times, with bovine serum albumin (BSA) serving as the model protein. For the Ellman's assay, 10 μL of Ellman's reagent (DTNB, 4 mg/mL in PBS) and 40 μL of BSA solution (50 μM in PBS) were added to 500 μL of PBS (pH 8.0). The mixture was incubated at 25 °C for 15 min, after which the absorbance spectrum was measured using a 1 cm pathlength cuvette. The concentration of –SH groups was calculated according to the Beer–Lambert law using the molar extinction coefficient of 13,600 M⁻¹ cm⁻¹ for 2-nitro-5-thiobenzoate (TNB) at 412 nm. The number of –SH groups per protein molecule was determined by dividing the measured –SH concentration by the protein concentration. Protein concentration was quantified using the Bio-Rad Protein Assay according to the manufacturer's instructions.

To conjugate antibodies to AuNPs, thiolated antibodies (20 μL, 0.5 mg/mL) were incubated with AuNPs (200 μL, 43 pM) overnight at 4 °C. Then, the AuNPs were surface-passivated by reacting with 2.6 μL of mPEG-SH (0.1 mM, MW = 5000; corresponding to a 30,000-fold molar excess relative to AuNPs) for 1.5 h at RT with gentle continuous mixing (~600 rpm) using a Deluxe Mixer (American Scientific Products, McGraw Park, IL). The antibody/AuNPs were purified by three rounds of centrifugation (8160 × g, 5 min), with PBS washing between each round. The purified antibody/AuNPs were resuspended in 200 μL of PBS containing 0.05% (v/v) tween-20 (0.05% PBST) and 0.05% (w/v) sodium azide and stored at 4 °C for use within two weeks of preparation. Absorption spectra of the AuNPs before and after antibody conjugation were measured using a Hewlett Packard 8452A diode array UV-Vis spectrometer (HP Inc., Palo Alto, CA) and size by dynamic light scattering (DLS) using a Zetasizer Nano ZS (Malvern Panalytical Inc, Westborough, MA).

The AuNPs before and after antibody conjugation were further characterized by SEM imaging. For bare AuNPs, 5 μL of as-received AuNP solution (43 pM) was spotted onto carbon tape mounted on an aluminum stub and dried under vacuum for 2 h before imaging. For antibody/AuNPs, αV/AuNPs were first buffer-exchanged into water by two rounds of centrifugation (8160 × g, 5 min) and washing with ultrapure water. Then, 10 μL of αV/AuNPs (43 pM in water) was spotted onto a silicon chip and incubated for 60 min at RT. After removal of the liquid, the adsorbed particles were washed twice with ultrapure water and dried in a desiccator overnight before imaging. The samples were sputter-coated with a 2–5 nm gold film and imaged by SEM at an accelerating voltage of 10 kV.

To further evaluate the success of antibody conjugations to AuNPs, 100 μL of DPBS containing 50,000 MDA-MB-231 cells were incubated with target-specific antibody/AuNPs or IgG/AuNP control (final concentration: 20 pM) for 2 h at 37 °C with intermittent mixing. Cells were purified by centrifugation (390 × g, 5 min), fixed with 1% paraformaldehyde (10 min, RT) and purified again via centrifugation. Cellular binding of AuNPs were examined by dark field imaging using an Olympus IX71 inverted microscope.

2.6. Single EV Integrin Profiling Using Dual Imaging Single Vesicle Technology (DISVT).

Targeted surface protein markers on individual EVs were detected using DISVT following the procedure described in our recent studies [25]. Briefly, Au-coated (100 nm in thickness) glass slide was prepared through template stripping from a custom Au-coated silicon wafer. The slide was divided into multiple wells (5 mm in diameter) using black vinyl tape, with each well holding a maximally 10 μL solution. Each well was coated with thiolated anti-CD81 rabbit monoclonal antibodies and then saturated with 11-mercaptoundecyl tetra (ethylene glycol) (MUTEG).

EVs were captured on the Au chamber slide by incubating purified EVs (~10⁹ EVs/mL) in the CD81 functionalized wells for 2 h at 4 °C. Followed by washing with 0.01% PBST, the immobilized EVs were incubated with antibody/AuNPs or IgG/AuNP control at a concentration of 20 pM in 0.05% PBST for 1 h at RT. Finally, EVs were labeled with 20 μM Chol-PEG-Cy5 for 15 min at 37 °C. After additional washing with PBS, EVs were quickly rinsed with water, dried with nitrogen, and imaged instantly.

Dark field and fluorescence images were obtained using a dual imaging system built on a customized Nikon LV150N microscope (Melville, NY, USA). Dark field illumination was provided by a halogen lamp while fluorescence imaging utilized a Melles Griot continuous-wave He laser (model 05-LPH-925, λ = 632.8 nm) at an angle of approximately 45° relative to the sample surface. All images were captured through a high numerical aperture (NA = 0.8), 100× Bright/Dark field objective lens with an extra-long working distance (4.5 mm) using a Photometrics CoolSNAP camera (Teledyne Photometrics, Tucson, AZ, USA). Images were analyze using our recently developed Auto Single EV Dual Imaging Analysis (AutoSEDIA) method, a custom Python image segmentation software that rapidly and nearly automatically analyzes multiple images simultaneously [42]. The image analysis generates a population density of the distribution of all captured EVs according to their scattering intensity. Calibration with IgG control gives the fraction of EVs that are positive to a targeted surface protein marker of interest.

To understand the impact of the Au film on the fluorescence property of Cy5 in the EV membrane, the fluorescence enhancement of Cy5 near a 100 nm thick Au film was calculated under illumination at an incident angle of 45°, consistent with experimental conditions. Simulations were performed both in air and in an environment with refractive index n = 1.45, representing coated molecules on the film surface. At the excitation wavelength (650 nm), the enhancement arises from constructive and destructive interference between the incident and reflected fields. For the emission wavelength (670 nm), the emitting dipole was positioned near an Au film modeled as a 100-nm thick block with lateral dimensions of 500 nm × 500 nm to approximate an extended film. The dielectric function of Au was taken from Palik’s handbook [43] and the discrete dipole approximation (DDA) method [44] was used for the simulation. The enhanced fluorescence signal was calculated using the equation from the previous report [45,46]:

$$f={\mid \frac{E}{E_0}\mid }_{\text{exc}}^2\cdot \frac{k_r^{\prime }/k_r}{(k_t/k_r)\eta +1-\eta }$$ (1)

where the first term represents the local electric field enhancement at the molecule position at the excitation wavelength of 650 nm. The second term accounts for the change in quantum yield of the dye molecule due to the presence of the Au film. Here, $\eta$ is the intrinsic quantum yield of the molecule (taken as 0.3), $k_r$ is the radiative decay rate of the molecule in free space, $k_r^{\prime }$ is the modified radiative decay rate in the presence of the Au film, the ratio $k_r^{\prime }/k_r$ describes the radiative decay rate enhancement in the presence of the Au film relative to free space and $k_t$ represents the non-radiative energy transfer rate from the dye molecule to the Au film at emission wavelength of 670 nm. To understand the impact of the AuNPs on the fluorescence property of Cy5 in the EV membrane, the fluorescence enhancement factor was calculated using the same model as shown in Equation 1 [45,46] and T-matrix method [47]. Calculations are performed for both air and water environments.

2.7. Statistical Analysis.

Statistical analyses were performed to compare the expression levels of the target proteins among human subject groups using analysis of variance (ANOVA) followed by Scheffé’s post hoc test. A p-value < 0.05 was considered of statistical significance. Mean differences between groups were deemed significant when their absolute values exceeded the minimum significant difference derived from the Scheffé procedure. The diagnostic efficiency of the integrin markers in different groups of BC patients was evaluated through receiver operating characteristic (ROC) analysis using function roc in the R package pROC.

3. RESULTS AND DISCUSSION

3.1. Overview of the Study.

Figure 1 shows a schematic overview of the study to examine the feasibility if integrin-defined plasma EVs for BCdetection and bone metastasis association. We investigated two cohorts of BC patients: Stage I patients and Stage IV patients with bone metastasis (Table S1&S2). Plasma samples from healthy donors were used as the control. Each cohort contained at least 10 subjects. To detect and quantify EVs expressing a specific integrin, we applied a dual imaging SVT (DISVT) assay that we recently developed, which captures EVs through CD81-based immunocapture on a MUTEG-modified gold-coated glass chip, enumerates the captured EVs using a customized Chol-PEG-Cy5 membrane dye and fluorescence imaging, and detects targeted proteins using antibody/AuNPs via dark- field imaging [25]. By counting the captured EVs with fluorescence imaging and the subpopulations that are positive to a specific target integrin, the fraction of EVs that are positive for the target integrin can be derived. Owing to its dual imaging modality at single vesicle and single particle level combined with a molecular capture and targeting strategy, this method enables rapid and highly specific molecular profiling of individual EVs, facilitating quantification of EV subtypes of interest. We focus on integrins αV and β3 due to their significant roles in bone-associated metastasis [33–36], with α6 and β5 being used as the controls for comparative studies.

Figure 1.

(A) Study design of breast cancer detection and bone metastasis through quantifying the plasma EVs carrying integrins αV, β3, α6, and β5. (B) Schematic of DISVT for detection of target integrins on individual EVs and quantification of plasma EVs carrying specific integrin markers.

3.2. EV Isolation and Characterization.

To reduce interference from abundant plasma proteins that may compete with EVs for binding to CD81 capture antibodies, we purified plasma EVs using Izon qEV column (35 nm cutoff), a widely used commercial platform for EV purification based on size-exclusion chromatography. qEV column separates particles primarily based on size to enrich EV-containing fractions. Although the primary purpose of qEV isolation is to remove abundant plasma proteins, larger particles, cellular debris, protein aggregates, and large extracellular vesicles are also excluded during the early eluting fractions. Following the manufacturer’s protocol, only the designated EV-enriched fractions were collected for analysis. Figure S1 shows the size distribution of the purified EVs from a patient in each cohort. EVs exhibited the hydrodynamic size (HD) in the range of 100–400 nm with large heterogeneity depending on the subject. However, a major population existed in all subjects with HD around 150–160 nm suggesting dominating contributions by small EVs (≤ 200 nm). Compared with isolation using a 0.2 μm PES filter [48], qEV columns yielded additional EV subpopulations exceeding 200 nm (e.g., ~300 nm and ~460 nm in a Stage I patient sample), which is not surprising as EVs are heterogeneous in size and can range from 30 to 1000 nm.

EVs were also characterized with SEM imaging to evaluate the purity, morphology, and size of qEV-isolated plasma EVs. In this study, plasma from a Stage I BC patient was used as a representative sample for SEM characterization. The qEV isolate was first fixed with 1% glutaraldehyde and then adsorbed onto a silicon chip. After sequential washing with PBS and ultrapure water to remove residual salts, the EVs were dehydrated through a graded ethanol series before SEM imaging. Large-area SEM examination at low magnification (5000×) showed well-dispersed particles across the silicon chip (Figure S2A). No large aggregates or obvious non-EV debris were observed, supporting the effective removal of plasma contaminants during qEV size-exclusion isolation. Closer examination at higher magnification (12,000×) revealed individual particles with round shapes, consistent with the EV morphology (Figure S2B). Sizes of EVs were determined using high-magnification SEM images at 50,000×. Consistent with the NTA results, particles characterized by SEM imaging were dominated by small EVs (≤ 200 nm). Figure S2C shows a representative high--magnification image (50,000×), in which four EVs with similar sizes of ~ 175 nm are observed. It should be noted that the measured EV sizes in SEM images can appear larger than the true size as the EV boundary is blurry or diffuse. Overall, the SEM characterization supports the successful isolation of plasma EVs with expected vesicular morphology and minimal visible contamination.

To further verify the quality of EV preparations at the molecular level, we characterized the presence of tetraspanin proteins commonly enriched on small EVs [49]. Because tetraspanin CD81 was used as the capture marker for small EVs in our platform, CD63 and CD9 were selected as independent small EV markers for ELISA analysis. These experiments were conducted using qEV-isolated plasma EVs from three healthy donors as representative samples. The results showed that the qEV isolates were positive for both CD63 and CD9, although the level of expression of these two markers differed across the three human subjects (Figure S3). The clear presence of tetraspanin proteins, particularly CD9, in the qEV isolates further confirmed that the isolated fractions contained a substantial population of small EVs, consistent with the NTA results.

Importantly, the vesicles analyzed in this study represent a substantially more refined population than the bulk qEV isolate. While NTA, SEM imaging, and ELISA characterize the entire particle population present after qEV purification, our downstream analyses were restricted exclusively to CD81-positive vesicles through immunocapture. CD81 is a tetraspanin commonly enriched on small EVs, including EVs of endosomal origin [48–51]. Because EV biogenesis cannot be definitively assigned based solely on size or surface-marker characterization, we refer to the captured vesicles throughout this study as CD81-positive EVs rather than exosomes. These CD81-positive EVs are expected to originate from multiple cell types, including tumor cells and normal host cells. CD81 expression is relatively low on platelet-derived EVs, a major contributor of normal EVs in plasma. Thus, CD81 immunocapture is expected to reduce the contamination of normal backgrounds, enhancing the sensitivity of detecting tumor-derived EVs. The MUTEG modification effectively blocks nonspecific adsorption of plasma components due to its highly hydrophilic and antifouling EG surface [25]. As shown in Figure S4A, the unmodified Au chip exhibited substantial nonspecific adsorption of plasma components, leading to high background and heterogeneous EV adsorption. In contrast, the MUTEG-modified Au chip remained largely free of nonspecific adsorption, yielding a nearly background-free surface (Figure S4B). By combining qEV column to remove abundant plasma proteins and large non-EV impurities, MUTEG surface modification to minimize nonspecific adsorption of plasma impurities, and CD81 immunocapture for selective enrichment of molecularly defined EV subtypes, we achieved efficient EV capture with minimal background interference (Figure S4C).

3.3. Integrin Expression on BC Cell Lines.

Given that EVs originate from their parent cells, we first examined the expression of candidate integrins on a representative cell line from each BC subtype using LC–MS/MS proteomic analysis. The results revealed distinct expression patterns of integrins αV, β3, α6, and β5 among the three common BC cell lines, MDA-MB-231, MCF-7, and SK-BR-3 that represents triple-negative BC (TNBC), hormone-positive BC, and HER2-positive BC, respectively (Figure 2). All three cell lines exhibit detectable expression of the four integrins. Integrins αV and β5 showed relatively high and comparable abundance across all three cell lines. In contrast, β3 displayed the lowest overall abundance among the four integrins, with slightly higher levels in MCF-7 and SK-BR-3 compared to MDA-MB-231. A clear difference was observed for α6, where MDA-MB-231 showed markedly higher expression than both MCF-7 and SK-BR-3, suggesting enhanced laminin-binding potential in this highly metastatic TNBC cell line. Overall, the data indicate that while αV and β5 are broadly expressed across these BC models, α6 is particularly enriched in MDA-MB-231, whereas β3 remains relatively low in all three lines.

Figure 2.

Expression levels of integrins αV, β3, α6, and β5 on MDA-MB-231, MCF-7 and SK-BR-3 cancer cell lines determined by LC–MS/MS.

3.4. Synthesis and Characterization of Integrin-Targeting AuNPs.

AuNPs were obtained commercially from Ted Pella and used as received. Antibodies against integrins αV, β3, α6, or β5 were first functionalized with thiol groups using NHS-PEG-SH 1000 and then bound to AuNPs via Au-S bonds. Using Ellman’s assay and BSA as the model protein, we found that thiolation with NHS-PEG-SH was more efficient at higher temperatures. The number of free thiol (–SH) groups introduced per protein molecule was approximately 2.5, 2.0, and 1.7 at 37 °C, 21 °C, and 4 °C, respectively (Figure S5A). The number of -SH groups per protein increased rapidly during the first 60 min of reaction and gradually approached a plateau after approximately 120 min, reaching about 1.6 free thiol groups per protein molecule (Figure S5B). Based on these results, a reaction temperature of 37 °C and time of 2 h were selected for antibody thiolation. The antibody/AuNPs were then passivated with mPEG-SH 5000 to improve colloidal stability and minimize non-specific binding of the conjugates to EVs.

Bare AuNPs exhibited an LSPR wavelength at 634 nm. Antibody or IgG conjugation did not induce significant spectral shift on the LSPR wavelength and did not result in any observable color change of the AuNP suspension (Figure 3A). DLS characterization showed the bare AuNPs had a HD of 68.3 ± 0.7 nm. Protein conjugation with IgG, anti-integrin αV, anti-integrin β3, or anti-integrin β5 antibodies increased the HD of AuNPs by 23 nm and with anti-integrin α6 by 35 nm (Figure 3B). It is noted that the anti-integrin α6 antibody is a rat IgG whereas the others are mouse IgG. To determine whether the size differences between α6/AuNPs and other conjugates arose from the host species of the antibody, we performed the same conjugation procedure using anti-integrin α6 antibody derived from mouse. The HD of the AuNPs linked with mouse anti-integrin α6 antibodies was the same as other mouse integrins, with HD of 91.6 ± 0.19 nm (Figure S6). Thus, the rat α6 antibody likely differs from the mouse antibodies in formulation or physicochemical characteristics, which may affect its conjugation behavior and result in the observed size difference.

Figure 3.

Characterization of target-specific antibody/AuNPs. (A) Absorption spectra and photographic images of AuNPs before and after conjugation with anti-integrin antibodies or IgG control. (B) DLS characterization of AuNPs before and after antibody conjugation. (C-H) Dark field images of MDA-MB-231 cells incubated with αV/AuNPs (C), β3/AuNPs (D), α6/AuNPs (E), β5/AuNPs (F), IgG/AuNPs (G), and PEGylated AuNPs (H). (I) Dark field image of untreated MDA-MB-231 cells.

The particles before and after antibody conjugation were further characterized using SEM imaging. Bare AuNPs exhibited a spherical shape with an average diameter of 51.5 ± 4.6 nm, as determined from measurements of over fifty AuNPs in high-magnification SEM image (Figure S7A). After antibody conjugation, the size of the nanoparticles increased to 72.9 nm ± 6.7 nm, as determined using anti-integrin αV antibody as the model protein (Figure S7B). Similar to the DLS characterization, SEM imaging showed a size increase of ~20 nm after antibody conjugation, confirming the success of surface modification. It should be noted that the SEM-based size estimation may not reflect the true sizes of the antibody-conjugated AuNPs because the organic antibody/PEG layer has much lower electron density and weaker contrast than AuNP core under electron microscopy and the apparent diameter can be affected by drying, faint outer boundary, and edge-definition during measurement. In addition, the antibody/AuNPs were buffer-exchanged into water before drying for SEM imaging, which may induce conformational changes in the antibody layer and thus further influence the apparent particle size. It should also be noted that the HD measured by DLS was significantly larger than the particle size measured by SEM, even for bare AuNPs. This difference is not surprising because DLS measures particles in suspension and includes the surface stabilizing layer, electrical double layer, and surrounding solvation shell, whereas SEM measures dried particles under vacuum.

To further evaluate whether the antibodies were successfully conjugated to the AuNPs and retained their ability to specifically recognize and bind the targeted integrins, we assessed their binding to the model cell line MDA-MB-231 via dark field imaging (Figure 3C-F). Controls include IgG/AuNPs, PEGylated AuNPs lacking antibodies, and untreated cells (Figure 3G-I). In consistent with the integrin expression determined by LC-MS/MS, the integrin αV-, α6- or β5- specific antibody/AuNPs targeting showed strong binding to the cancer cells while integrin β3-specific AuNPs showed relatively low binding. In contrast, the IgG/AuNPs and PEGylated AuNPs lacking antibodies showed minimal non-specific binding to the cancer cells. These results demonstrate that antibody/AuNPs can specifically bind to their target integrin receptors, enabling selective detection of integrins of interests. It should be noted that integrin receptor-mediated endocytosis of antibody/AuNPs may occur following initial receptor binding under our experimental condition. However, such internalization is initiated by specific antibody–receptor interactions and therefore does not compromise the assessment of target specificity.

3.5. DISVT for Surface Integrin Profiling.

As mentioned earlier, the workflow of DISVT involved dual fluorescence and dark field imaging of EVs immobilized on an Au-coated glass slide based on EV’s CD81 expression. Integrins of interest were labeled with AuNPs linked with target-specific antibodies (anti-αV, anti-β3, anti-α6, or anti-β5). Fluorescence imaging detected CD81-positive EVs using Cy5 linked with a cholesterol moiety through a PEG linker, providing vesicle localization and enumeration. Simulation studies under illumination at an incident angle of 45° showed a distance-dependent modulation of Cy5 fluorescence by the 100-nm-thick Au film both in air and in the surface coated matrix with MUTEG and capture antibody that had a refractive index of approximately 1.45 (Figure 4A). At very short distances (< 16 nm for n = 1.45 environment and < 38 nm for vacuum), fluorescence is quenched, due to dominant nonradiative energy transfer to the metal at the emission wavelength and destructive interference between the incident and reflected fields at the excitation wavelength. As the distance increases, the enhancement factor rises, reaching a maximum enrichment by 4.7-fold at 100 nm in n = 1.45 environment and 5.1-fold at 150 nm in air, indicating an optimal balance between reduced quenching and plasmon-enhanced emission, as well as interference between the incident and reflected fields. Beyond this optimum, the enhancement gradually decreases, approaching baseline at 236 nm in n = 1.45 environment and 268 nm in air as plasmonic coupling diminishes with distance. Clearly, the higher refractive index environment (n = 1.45) shifts the impact of the Cy5 fluorescence at a shorter distance and slightly reduces the maximum enhancement compared to air, reflecting altered optical interference and plasmonic field distribution. In our methodology, Cy5 molecules were embedded in EV membrane that was separated from the Au film by antibody with its PEG-SH 1000 linker, giving a distance between Cy5 and Au film of 10–200 nm depending on its location in the EV, the size of EV, and the conformation of the PEG (extended, brush or mushroom). These distances were estimated from the known structural dimensions of the assay components, including the Au film, capture antibody, PEG linker, EV membrane, and the typical size range of EVs, and were used to interpret the simulation results rather than representing experimentally measured quantities. Thus, the overall intensity of the Cy5 fluorescence was likely enhanced by the Au film compared to Cy5 alone as majority Cy5 molecules were positioned approximately 15 nm away from the Au surface.

Figure 4.

(A) Calculated fluorescence signal enhancement of a Cy5 molecule near a 100-nm thick Au film as a function of distance in air or in an environment with refractive index n = 1.45. (B) Calculated fluorescence signal enhancement of a Cy5 molecule positioned near a 60-nm AuNP in air or in water.

The target-specific AuNPs bound to EVs are detected by dark field imaging based on the strong LSPR scattering properties of AuNPs. To determine whether the plasmonic properties of AuNPs affect the fluorescence behavior of Cy5, we calculated fluorescence signal enhancement of a Cy5 molecule positioned near a 60-nm diameter AuNP with a comparison between air or water environments. Similar to the Au film, the AuNP exhibited a distance-dependent plasmonic effect on the fluorescence property of Cy5 (Figure 4B). Enhancement increases as the distance increases to 9 nm in air and to 6.5 nm in water and then rapidly decreases, approaching unity (i.e., no effect) by 60–100 nm. The maximal enhancement is 2.1-fold in air and 4.2-fold in water, demonstrating that the dielectric environment significantly amplifies plasmon–fluorophore interactions. In our methodology, Cy5 molecules were embedded in EV membrane that was separated from the surface of the AuNP by the surface modification with the targeting antibody with its PEG-SH linker and mPEG-SH saturation. Based on the DLS characterization studies mentioned above, the thickness of AuNP increased by at least 11 nm. Our images were acquired in air environment. Thus, the fluorescence intensity of Cy5 can be enhanced up to 2-fold depending on the location of Cy5 in the immobilized EVs. Considering the additional separation imposed by the size of EVs (30–200 nm in diameter), the fluorescence enhancement for the majority of Cy5 molecules on the EV membrane is likely attenuated. Consistent with this, we did not observe a significantly stronger Cy5 fluorescence signal in EVs bound to AuNPs compared to EVs without AuNP binding.

To quantify EVs positive and negative for a target integrin, the fluorescence and dark field images were merged. Light scattering signals from individual EVs were extracted using an AutoSEDIA method to give a population density of the distribution of all captured EVs according to their scattering intensity [42]. Figure 5 shows an example of the dual imaging analysis for each integrin marker using a Stage IV BC bone-metastatic patient sample. By counting the total number of captured EVs and AuNP-bound EVs, the fraction of AuNP-bound EVs were calculated. After correction with AuNPs linked with IgG control for nonspecific binding of the antibodies, the fraction of the CD81-positive EVs carrying a given integrin, denoted as F_p where p is the targeted protein marker, was calculated.

Figure 5.

Illustration of image analysis for integrin profiling using a Stage IV BC patient sample. FL: labeled fluorescence image showing captured EVs (red). DF: labeled dark field image showing AuNP-bound EVs (yellow). Overlay image showing labeled particles from FL image and DF image. Expression profile of that targeted integrin was expressed as the population density profile of EVs with blue color for AuNP-bound EVs and orange color for AuNP-free EVs. Scale bar: 30 μm.

To evaluate assay reproducibility, each patient sample was analyzed in duplicate using two replicate chambers available on a single chip. Chips were processed in a randomized order with respect to disease status, and technical reproducibility was assessed by calculating the coefficient of variation (CV) between duplicate measurements for each sample and integrin marker. The median CVs were 9.7% for integrin αV, 10.0% for β3, 11.0% for α6, and 14.7% for β5 (Table S3). The modest differences in CV among markers likely reflect differences in marker abundance, antibody performance, and biological heterogeneity. All median CVs were below 15%, indicating good technical reproducibility for this proof-of-concept study using patient-derived plasma EVs.

3.6. Plasma EVs Carrying Integrin αV for Early BC Detection and Bone Metastasis Association.

Integrin αV is a broadly important adhesion/signaling subunit that pairs with β1, β3, β5, β6, or β8, allowing tumor and stromal cells to sense and respond to RGD-containing extracellular matrix ligands such as vitronectin, fibronectin, osteopontin, and tenascin [52]. Among αV-containing integrins, αVβ3 has been especially linked to aggressive tumor behavior. Abnormal αVβ3 expression has been associated with tumor initiation, persistent growth, metastasis, stemness, and drug resistance across multiple cancer contexts [53]. In BC, elevated αVβ3 activity were found to be associated with increased bone metastatic potential and poorer clinical outcomes [33–36]. Building on prior knowledge, we hypothesize that plasma EVs carrying integrins αV and β3 are associated with the presence of BC and bone metastasis. To test the hypothesis, we first profiled integrin αV among CD81-captured EVs using anti-integrin αV antibody/AuNPs (αV/AuNPs) with IgG/AuNPs as the control in healthy donors, Stage I BC patients, and Stage IV patients with bone metastasis. In total, biobanked plasma samples from 13 healthy donors, 19 Stage I BC patients, and 14 bone-metastatic Stage IV BC patients were analyzed. Figure 6A-C shows an example of the population density profile of each cohort using the αV/AuNPs and the control NPs. The populations for the AuNP-bound EVs were clearly elevated in the BC groups compared to the healthy control.

Figure 6.

Single vesicle integrin αV profiling of plasma EVs from healthy donors and breast cancer patients with different disease statuses. (A−C) Examples of the population density histograms of plasma EVs from different subjects labeled with αV/AuNPs (red) and IgG/AuNPs (gray). (A) Heathy donor. (B) Stage I patient. (C) Bone-metastatic patient. (D) Box plot of F_αV of plasma EVs across different groups of human subjects. (E) F_αV of plasma EVs for each subject.

Statistical comparisons of the fraction of integrin αV-positive EVs among CD81-captured EVs across groups are presented in Figure 6D, while the corresponding subject-level data are shown in Figure 6E. The results showed that BC patients had a markedly higher proportion of αV-positive plasma EVs compared to healthy controls. The healthy controls had a mean F_αV of ~9.9% (range 2.2–29.4%), whereas BC patients exhibited much higher levels, with 27.6% on average in the Stage I group (range 1.0–52.3%), and 39.4% in the bone-metastatic group (13.2–66.2%). Statistical analysis (one-way ANOVA with post-hoc comparisons) confirmed that both Stage I and Stage IV patient groups had significantly greater F_αV than healthy controls (p = 3.4 × 10⁻⁴ for healthy donors vs. Stage I patients and p = 4.8 × 10⁻⁶ for healthy vs. Stage IV patients). These results demonstrate that even patients at early-stage had more αV-bearing EVs in circulation than disease-free individuals, suggesting the potential of this marker for early cancer detection. We also observed that the F_αV for Stage IV group significantly exceeded that of the Stage I group (39.4% vs 27.6%, p = 0.03). Together, these data suggest that the prevalence of αV-positive EVs is associated with tumor presence and bone metastasis association.

From a clinical perspective, the elevation of integrin αV on circulating EVs could be a useful liquid biopsy indicator. The fact that F_αV is already higher in Stage I patients compared to healthy donors is particularly noteworthy. It implies that early-stage tumors may shed EVs enriched in αV integrins, which could be probed for early-stage disease detection. This finding is in line with our prior report that single-EV analysis can sensitively detect early BC through surface protein markers (e.g. HER2) while traditional bulk assays fail [25]. Moreover, the significantly higher F_αV in patients with bone metastases suggests that αV-positive EVs might also serve as a marker associated with metastatic disease.

3.7. Plasma EVs Carrying Integrin β3 for Early BC Detection and Bone Metastasis Association.

Integrin β3 is a key β subunit that forms heterodimers primarily with αV and αIIb, functioning as a transmembrane receptor that mediates cell–extracellular matrix (ECM) interactions and intracellular signaling [27]. Integrin β3 is widely recognized as a multifunctional regulator of tumor progression, influencing adhesion, migration, survival, and communication with the tumor microenvironment, serving as a valuable biomarker for early detection and prognostic assessment [54,55]. As noted above, elevated αVβ3 activity has been linked to enhanced bone metastatic potential and poorer clinical outcomes in BC. Thus, we next profiled integrin β3 among the CD81-catpured EVs across the same human subjects as in the αV studies. Figure 7A–C shows an example of the population density profile of each cohort using the anti-integrin β3 antibody/AuNPs (β3/AuNPs) and the control IgG/AuNPs. Similar to integrin αV, the populations for the AuNP-bound EVs were clearly elevated in the BC groups compared to the healthy control.

Figure 7.

Single vesicle integrin β3 profiling of plasma EVs from healthy donors and breast cancer patients with different disease statuses. (A−C) Examples of the population density histograms of plasma EVs from different subjects labeled with β3/AuNPs (red) and IgG/AuNPs (gray). (A) Heathy donor. (B) Stage I patient. (C) Bone-metastatic patient. (D) Box plot of F_β3 values of plasma EVs across different groups of human subjects. (E) F_β3 of plasma EVs for each subject.

Statistical comparisons of the fraction of integrin β3-positive EVs among CD81-captured EVs across groups are presented in Figure 7D, while the corresponding subject-level data are shown in Figure 7E. The distribution of F_β3 paralleled that of F_αV in many respects. Healthy donor plasma displayed a low baseline of β3-bearing EVs (mean F_β3 ~9.6%, range 0.5–22.1%), whereas BC patient plasmas contained elevated fractions of β3-positive EVs. On average, F_β3 was 21.6% in Stage I patients (range 0–54.8%) and 37.3% in bone-metastatic patients (13.7–58.7%). The level of β3-bearing plasma EVs was approximately doubled in Stage I patients and tripled in Stage IV patients with bone metastasis compared to healthy donors. Statistical comparisons showed that Stage I patients had significantly greater F_β3 than healthy controls (p = 5.9 × 10⁻³), as did the bone-metastatic patients (p = 8.4 × 10⁻⁸). The bone-metastatic patients exhibited a markedly higher F_β3 compared to the Stage I group (37.3% vs 21.6% mean; p = 2.2 × 10⁻³). These results suggests that prevalence of integrin β3-positive EVs is associated with tumor presence and bone metastasis association.

Overall, the integrin β3 on EVs mirrored the trends observed for integrin αV. Both integrin αV-positive EVs and β3-positive EVs were enriched in BC at early-stage and further enriched in advanced stage with bone metastasis. The concordance between F_αV and F_β3 results also hints that many of the EVs in patients may carry the complete αVβ3 integrin heterodimer. As a matter of fact, pervious studies reported that prostate cancer cells released αVβ3 integrin through EVs in vivo [56]. Here, we show the diagnostic relevance of EVs carrying αV or β3 integrin subunits by independent assessment of αV and β3 expressions on plasma EVs in pilot cohorts of BC patients. It is worth noting that our results do not imply that αV and or β3 EVs is exclusively associated with bone metastasis, as these integrins have been implicated in multiple metastatic sites and cancer types.

3.8. Profiling and Quantification of Plasma EVs Carrying Integrin α6 or β5.

As a comparative study for integrin αV in the bone metastasis association, we examined integrin α6, a marker with implications in lung metastasis [41]. Figure 8 shows an example of the EV population density profiles in each cohort using anti-α6 antibody/AuNPs (α6/AuNPs) and the control IgG/AuNPs as well as the comparison of F_α6 across all subjects. The median fraction of integrin α6-positive EVs among CD81-positive EVs was 18.0% (range 2.9–38.6%), 24.8% (5.1–43.3%), and 15.9% (13.0–39.2%) for healthy control, Stage I patients, and bone-metastatic patients, respectively. No statistically significant difference was observed across the three groups (e.g., healthy vs Stage I patient p = 0.10; healthy vs bone-metastatic patient p = 0.63; and Stage I vs bone-metastatic patients p = 0.5). The lack of statistically significant variation suggests that EV- associated integrin α6 is not influenced by disease status in our cohorts, indicating that it is unlikely to serve as a marker for BC detection or bone-tropic disease.

Figure 8.

Single vesicle integrin α6 profiling of plasma EVs from healthy donors and breast cancer patients with different disease statuses. (A−C) Examples of the population density histograms of plasma EVs from different subjects labeled with α6/AuNPs (red) and IgG/AuNPs (gray). (A) Heathy donor. (B) Stage I patient. (C) Bone-metastatic patient. (D) Box plot of F_α6 of plasma EVs across different groups of human subjects. (E) F_α6 of plasma EVs for each subject.

As a comparative study for integrin β3 in the bone metastasis association, we studied integrin β5, a biomarker that pairs with αV on EVs to home to live metastasis [41]. As shown in Figure 9, the healthy donors had a median of 15.8% (range 2.0–30.9%) integrin β5-positive EVs among the CD81-captured EVs. Stage I and bone-metastatic patients had similar levels of β5-positive EVs, with the mean of 17.8% (1.8–44.6%) for Stage I patients and 17.5% (5.1–30.8%) for bone- metastatic patients. Like integrin α6, β5 showed no significant differences in the abundance of integrin β5-positive EVs among the studied groups, with p-value of 0.60 between Stage 1 patients and healthy donors, 0.64 between bone-metastatic patients, and 0.90 between Stage I and bone-metastatic patients. In other words, the level of integrin β5-positive EVs did not change in association with cancer presence or bone metastasis. This uniformity reinforces that integrin β5 is not a discriminatory marker in our patient population.

Figure 9.

Single vesicle integrin β5 profiling of plasma EVs from healthy donors and breast cancer patients with different disease statuses. (A−C) Examples of the population density histograms of plasma EVs from different subjects labeled with β5/AuNPs (red) and IgG/AuNPs (gray). (A) Heathy donor. (B) Stage I patient. (C) Bone-metastatic patient. (D) Box plot of F_β5 values of plasma EVs across different groups of human subjects. (E) F_β5 of plasma EVs for each subject.

In our companion analysis, the levels of integrin αV-positive EVs and integrin β3-positive EVs among CD81-capured EVs were found to be significantly elevated in Stage I patients and bone-metastatic patients. The fact that α6 and β5 did not show such increase indicates that the changes in αV and β3 are not due to a general upregulation of all EV integrins in cancer patients. Instead, the elevation of αV and β3 is likely tied to their roles in cancer initiation and bone metastasis biology. Comparing integrins αV, α6, β3, and β5 using the same human subjects, we found that integrin markers can be selectively altered in disease and conveyed through plasma EVs, suggesting that probing integrin-carrying plasma EVs may provide a promising approach for non-invasive cancer detection and metastasis assessment.

3.9. Diagnostic Performance of Plasma EVs with Distinct Integrins.

To assess the diagnostic performance of each integrin marker, we performed ROC analysis to differentiate the defined patient groups. Each marker’s ability to distinguish Stage I and bone-metastatic patients from healthy donors, as well as to differentiate between these two patient groups, was quantified by the area under the curve (AUC). The results showed that integrin αV achieved high AUC values when distinguishing cancer patients from healthy donors (Figure 10A). Notably, αV yielded an AUC of 0.97 for metastatic versus healthy groups, indicating excellent discrimination between bone-metastatic patients and healthy individuals. It also showed strong performance for Stage I patients versus healthy donors (AUC = 0.88), underscoring αV’s robust ability to discriminate disease presence. Similar to αV, integrin β3 showed strong diagnostic performance in distinguishing bone-metastatic group and healthy control, with AUC = 0.98 (Figure 10B). For early detection, β3 exhibited much lower sensitivity than αV, with AUC = 0.74.

Figure 10.

Examination of the diagnostic potential of each integrin marker by ROC analysis for integrins αV (A), β3 (B), α6 (C), and β5 (D). H: Healthy donors. SI: Stage I BC patients. BM: bone-metastatic BC patients.

In contrast to αV and β3, the integrin α6 and β5 monomers demonstrated low diagnostic utility (Figure 10C&D). For α6, the AUC values fell in the poor (0.5–0.6) to weak range (0.6–0.7). For β5, AUC values hovered around 0.5, which indicates no discriminative power. These results suggest that integrins α6 and β5 expression does not differ meaningfully between healthy individuals and Stage I patients or bone-metastatic patients. In particular, β5 provided essentially no diagnostic value for this disease context.

Overall, the ROC analysis suggests integrins αV and β3 were the best-performing markers among the four candidate integrins for distinguishing BC patient categories, especially for identifying patients with bone-metastatic disease. Both αV and β3 achieved AUC values in the good-to-excellent range for differentiating BC patients from healthy donors. They showed excellent discriminatory performance in differentiating bone-metastatic patients from healthy donors, supporting an association of these markers with bone-metastatic disease in this pilot cohort. In addition, integrin αV showed good performance in differentiating Stage I patients from healthy donors, suggesting its potential as a candidate biomarker for early BC detection.

4. CONCLUSION

We performed the first quantitative single EV analysis of integrin expressions on plasma EVs. By leveraging a DISVT, we overcame the signal dilution inherent to bulk EV assays and captured diagnostically meaningful differences in EV subpopulations. Using this ultrasensitive technique, we characterized individual plasma EVs in pilot cohorts of healthy donors, Stage I BC patients and bone-metastatic BC patients based on the expression of four integrin subunits and demonstrated that the fraction of CD81-captured EVs carrying integrin αV or β3 is significantly elevated in BC patients compared to healthy controls. This elevation was consistently observed across both groupwise statistical comparisons and ROC curve analyses. Specifically, both integrins αV and β3 achieved excellent performance in distinguishing bone-metastatic BC patients from healthy controls, with AUC exceeding 0.95. Integrin αV demonstrated good diagnostic strength in differentiating Stage I patients from health donors (AUC = 0.88), showing strong potential for early BC detection. However, integrin β3 exhibited moderate diagnostic performance (AUC = 0.74), indicating more limited but still meaningful discriminatory ability. In contrast, α6 exhibited very limited discriminatory ability (AUCs = 0.59–0.71) and β5 no meaningful discriminatory ability (AUCs = 0.48–0.53) across the three cohorts. This contrast highlights the association of integrins αV and β3 in BC presence and bone-metastatic progression, rather than a generalized upregulation in integrin expression in cancer.

While these findings are promising, the current study is limited by its inability to directly assess αVβ3 heterodimer formation on individual EVs. Accordingly, the observed associations are interpreted as potential rather than definitive evidence of αVβ3 involvement. Future studies employing approaches capable of directly assessing α and β co-localization or heterodimer formation at the single-EV level will be needed to confirm the presence and biological significance of specific integrin heterodimers.

Due to the limited cohort size, the findings from this proof-of-concept feasibility study need further validation in larger, independent cohorts and more diverse populations with comprehensive clinical annotation to evaluate the potential impact of baseline clinical factors and the robustness of EV-associated integrin markers for early BC detection and bone-metastasis association. In addition, future validation studies should include patients with benign breast disease as another control to better define the specificity of the identified signatures from this feasibility study. Further, the broader involvement of integrins αV and β3 in metastasis needs to be evaluated in patients with metastasis or recurrence at non-bone sites, although such cases are relatively rare in breast cancer. At present, the DISVT platform is limited to the detection of a single biomarker on the individual EVs and relies on manual operation, which constrains throughput and scalability. Future development is needed to enhance multiplexing capability to enable simultaneous detection of co-expressed biomarkers on individual EVs, as well as improving throughput through integration with high-throughput imaging modalities and automated workflows.

Supplementary Material

ASSOCIATED CONTENT

Supporting Information. Additional information on the patients’ clinical data and size distribution of EVs in exemplary healthy donors, Stage I patients, and bone-metastatic patients, as well as size distribution of AuNPs conjugated with mouse anti-integrin α6 antibodies is available in the supporting information. This information is available free of charge via the Internet at http://pubs.acs.org/.