Automated Microfluidic Platform for Single Spheroid Culture and Extracellular Vesicle Isolation: Application to Spheroid Transcriptomic Profiling
Marie Hut, Josiane Denis, Frédéric Bottausci, Myriam Cubizolles, Patricia Laurent, Joris Kaal, Mahfod Benessalah, François Boizot, Nadia Cherradi, Yves Fouillet, and Vincent Agache*
Extracellular vesicles (EVs) are key mediators of intercellular communication and carry molecular information that reflects the state of their cell of origin. 3D cell cultures more accurately reflect the in vivo microenvironment and the biogenesis of extracellular vesicles compared to 2D cultures. Despite these advantages, studying EVs in 3D systems such as spheroids remains technically challenging. Conventional EV isolation and characterization methods often require pooling multiple spheroids to obtain sufficient material, which masks the intrinsic heterogeneity between individual spheroids and limits applications in precision medicine. To overcome these challenges, this work develops an automated microfluidic platform capable of single-spheroid culture, continuous secretion collection, and high-efficiency EV isolation. The platform incorporates 200 nm filtration and immunomagnetic capture targeting CD63/CD81-positive EVs, achieving a 60% recovery yield. Using adrenocortical carcinoma spheroids as a model, this work demonstrates that inhibiting ß-catenin signaling selectively reduces the levels of EV-derived miR-139-5p and miR-483-5p, consistent with prior findings from 2D culture studies. This platform represents a groundbreaking approach to EV profiling at the single-spheroid level, unlocking new opportunities for personalized medicine, drug discovery, and targeted therapies by enabling the analysis of cellular heterogeneity and scarce biological samples such as patient-derived organoids.
1. Introduction
Extracellular vesicles (EVs) are nanoscale particles secreted by all cell types into their extracellular environment, carrying a diverse cargo of bioactive molecules, including lipids, proteins, and nucleic acids, that can be transferred intact to recipient cells.[1] EVs facilitate intercellular communication and play pivotal roles in numerous physiological and pathological processes.[2,3] Their ability to reflect the molecular state of their parent cells has made EVs promising biomarkers, partic- ularly in oncology,[4] with malignant cells often secreting elevated levels of EVs.[5] In adrenocortical cancer (ACC)-a rare, aggressive malignancy of the adrenal cortex with poor prognosis and limited therapeu- tic options[6]-tumor and serum-derived miR-139-5p and miR-483-5p have been identified as clinically relevant biomark- ers. miR-483-5p, overexpressed in ACC, serves as a marker of malignancy and tumor aggressiveness, while miR-139- 5p is similarly associated with tumor progression.[7] However, isolating and
M. Hut, F. Bottausci, M. Cubizolles, P. Laurent, J. Kaal, M. Benessalah, F. Boizot, Y. Fouillet, V. Agache Univ. Grenoble Alpes CEA Leti DTIS Grenoble 38000, France
E-mail: vincent.agache@cea.fr
J. Denis, N. Cherradi Univ. Grenoble Alpes CEA Inserm IRIG U1292 IMAC Grenoble 38000, France
İD The ORCID identification number(s) for the author(s) of this article can be found under https://doi.org/10.1002/smll.202508115
@ 2025 The Author(s). Small published by Wiley-VCH GmbH. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.
DOI: 10.1002/smll.202508115
ADVANCED SCIENCE NEWS
NANO · MICRO sma
studying EVs in vivo remains inherently challenging due to their small size and the complexity of biofluid matrices,[8] underscoring the need for robust, physiologically rel- evant in vitro models.
3D models, such as spheroids, organoids and organoid-on- chip, have become essential for replicating the complex 3D ar- chitecture and local microenvironment of in vivo tumors. Unlike traditional 2D cultures, 3D models better mimic tumor hetero- geneity, incorporating crucial features such as cell-cell and cell- matrix interactions, oxygen gradients, and necrotic cores.[9] They also provide a more accurate framework for evaluating tumor re- sponses to therapies, reflecting the molecular and cellular diver- sity observed in clinical cases.[10] These models thus provide a physiologically relevant system for studying EV secretion.[11-13] However, most studies rely on pooling multiple spheroids to obtain enough material for analysis,[14-16] a practice that, while increasing sample quantity, inevitably averages out the varia- tions between individual spheroids. Such heterogeneity is cru- cial to understand, especially in the context of personalized medicine,[17] where individual variability can impact therapeutic outcomes. Addressing this challenge, the focus on EVs secreted by individual spheroids forms the foundation of our work, of- fering the added advantage of reducing tissue usage- particu- larly valuable when working with rare samples, such as Patient- Derived Organoids (PDOs).
Isolating EVs with high purity and yield remains challenging due to their small size and inherent heterogeneity. Traditional methods, such as ultracentrifugation and steric exclusion chro- matography, often fall short in terms of speed, efficiency, and pu- rity, particularly when working with low-concentration or small- volume samples.[8] In contrast, microfluidic technologies offer a promising alternative, enabling precise and scalable handling of small volume samples.[18,19] These systems accelerate the isola- tion process while enhancing purity and cost-effectiveness. Fur- thermore, microfluidic platforms are ideal for spheroid culture, as they can precisely control fluid dynamics to replicate physi- ologically relevant microenvironments.[20,21] However, many ex- isting systems still rely on external components, such as syringe or peristaltic pumps, which complicate integration, increase the risk of contamination, and restrict downstream on-chip analytical applications.[22,23] This underscores the need for more integrated platforms that seamlessly combine EV isolation with organoid culture, providing a streamlined approach to studying EVs in a physiologically relevant context.
To address these challenges, we have developed a fully inte- grated microfluidic platform that combines single-spheroid cul- ture, continuous secretion collection, and high-yield EV isolation in a single device. The platform utilizes a hydrodynamic trapping mechanism to ensure precise spheroid positioning,[24,25] paired with an integrated micro-pump for continuous secretion collec- tion while minimizing environmental disturbances. To optimize EV purification, we implemented an on-chip filtration stage to pre-purify the secretions, efficiently removing cells and cellular debris larger than 200 nm. In conventional immunocapture pro- tocols, an initial purification step - such as centrifugation[26] - is commonly used to ensure that only EV-associated proteins are targeted, therefore minimizing interference from cell-surface proteins. We selected filtration as a microfluidics-compatible alternative, seamlessly integrating it into the workflow while
A
C
Chamber Valve
+
1
+
-P
-P
+P
Fluidic layer
B
Fluidic layer (COC)
Hyperelastic membrane
Pneumatic layer (COC)
+
t
t
+P
-P
-P
Pneumatic layer
Hyperelastic membrane
maintaining high yield. Our two-step EV microfluidic isolation protocol-200 nm filtration followed by immunomagnetic cap- ture of CD63/CD81-positive EVs-achieves a 60% recovery rate. By automating the entire process within a unified device, from spheroid culture to EV isolation, the system reduces manual in- tervention, minimizing risks such as sample loss and contamina- tion, which is particularly critical given the small sample volumes and low EV concentrations secreted by individual spheroids. To validate the platform’s potential, we performed a transcriptomic analysis of EVs derived from an adrenocortical carcinoma model, demonstrating its relevance for advancing personalized cancer therapies. This integrated approach provides an efficient, power- ful tool for exploring tumor molecular heterogeneity and ther- apeutic responses, with profound implications for cancer re- search, drug development, and precision medicine.
2. Results and Discussion
2.1. Operation of the Microfluidic Platform
The aim of this study was to seamlessly integrate three crucial functions-spheroid trapping, secretion collection, and extracel- lular vesicle isolation-into a single, efficient microfluidic plat- form. This approach reduces manual handling, minimizes hu- man error, and lowers the risk of contamination or sample loss. Unlike conventional microfluidic systems that rely on continu- ous flow driven by syringe or peristaltic pumps, our platform employs a discrete microfluidic approach[27,28] (Figure 1A) that divides reagents into independent volumes thanks to individu- ally addressable micromechanical valves and chambers. This en- ables more precise control of the reagents, making it particularly suitable for sensitive biological protocols.
The cartridge consists of three primary components: a flu- idic layer for fluid circulation, a hyper-elastic deformable mem- brane such as silicone,[24,25,27] and a pneumatic layer that regu- lates membrane movement via air pressure (Figure 1B). Apply- ing negative pressure (~200 mbar) causes the membrane to de- flect downward, allowing fluid to flow through the system, while positive pressure (~500 mbar) presses the membrane against the fluidic layer, effectively blocking fluid flow (Figure 1C). Operating
NANO · MICRO sma
A
Culture medium inlet
Organoid inlet
Chamber for secretion storage (20 µL)
Hydrodynamic trap for 300 um spheroid
Secretion outlet
Integrated peristaltic pump made of 3 chambers (~µL)
0
B
C
D
2 mm
2 mm
2 mm
E
m/s 104
1/s
F
2,5
20
2
15
1,5
10
1
0,5
5
G
H
600
500
Y position (um)
400
300
200
100
0
0
100
200
300
400
500
600
700
X position (um)
at relatively low pressures, the system is both energy-efficient and versatile, making it suitable for a wide range of applications. Us- ing this microfluidic architecture, we developed a robust network of valves and chambers that enable precise reagent manipulation.
2.2. Single Spheroid Trapping and Secretion Collection
The first critical step in our platform is the trapping of a single spheroid, achieved through hydrodynamic trapping.[24,29] The
system consists of a serpentine-shaped main channel (400 x 400 um2), along with a bypass channel that connects two facing branches (Figure 2A). This bypass channel features a U-shaped trap leading to a constriction narrower than the spheroid to be trapped (300 um). This configuration offers two possible paths for the spheroid: either attempting to pass through the constriction or bypassing the trap via the main channel if the trap is already occupied. When the trap is empty, the channel design ensures that hydrodynamic forces at the constriction guide the spheroid into the trap with an 80% efficiency (Figure S2, Supporting
ADVANCED SCIENCE NEWS
NANO · MICRO sma
Information), as previously demonstrated with pancreatic islets[24,25] and confirmed with various spheroids including NCI-H295R and HPaSteCs (human pancreatic stellate cells) models (Figure 2B). Representative images of spheroids before and after culture in the microfluidic chip demonstrate that their morphology was preserved across all tested flow rates (Figure S3, Supporting Information), and no deformation capable of blocking the trap was detected. In this study, spheroids were introduced individually via pipetting from well plates to prevent multiple spheroids from being trapped simultaneously. The trap was specifically designed for objects around 300 um in diameter and remained effective for spheroids ranging from 350 to 450 um. Spheroid size was controlled by adjusting the number of cells per spheroid to ensure reproducibility. We assume that this size control may be more challenging when working with patient-derived organoids (PDOs), which often display greater variability in size.
The platform also features an upstream inlet for medium or drug injection, complemented by an integrated micro-pump composed of three spherical caps (each 0.55 uL in volume). This micro-pump facilitates continuous perfusion of culture medium and secretion collection (Figure 2C) at a controlled rate of 0.55 uL min-1. The in situ integration of the micro-pump of- fers two key advantages: secretions are harvested directly from the spheroid, minimizing contamination risks, and are immedi- ately available for on-chip downstream analyses. These secretions are stored in a dedicated collection chamber via pneumatic actu- ation, while reverse actuation of the micro-pump allows for the gentle, intact retrieval of the spheroid at the end of the experi- ment (Figure 2D).
A major challenge associated with peristaltic actuation is the occurrence of periodic changes in flow direction, which we re- fer to as “oscillating flow.” This phenomenon arises in particular during the closure of the first valve of the micro-pump, when a minor fraction of the fluid is transiently redirected toward the trapping site. If uncontrolled, such backflow can generate pres- sure fluctuations that may destabilize or dislodge the spheroid (Video S 4, Supporting Information), thereby compromising its structural integrity. While literature reports that shear stress ex- ceeding a few kPa is detrimental, leading to cellular deformation, disruption of critical cell-cell and cell-matrix interactions, and a loss of physiological relevance of the spheroid model,130-32] few studies provide precise thresholds, and these values are highly dependent on the culture conditions and cell types. To mitigate these risks, we implemented synchronized pressure gradients: as one cap closes, the next opens simultaneously (Video S5, Sup- porting Information). This coordination ensures unidirectional fluid flow, effectively preventing backflow and minimizing shear stress on the spheroid surface.
To define a safe operating point for our system, we performed computational fluid dynamics (CFD) to analyze velocity profiles and shear stresses at the spheroid surface. Steady-state sim- ulations, conducted at a constant flow rate of 0.55 uL min-1 (Figures 2E-G), revealed that shear stress was highest at the constriction zone, where the flow velocity peaked. The supple- mentary information (Video S 7, Supporting Information) also presents time-dependent simulations, offering further insights into the dynamic nature of the flow conditions. By analyzing a full pumping cycle, we estimated the flow rate in the trapping
zone, which can be approximated as constant during the open- ing of V2 and the subsequent filling of the trap. During periods of cross-pressure gradients and fluid transfer from V3 to V4, the flow in the trapping zone is very low and is approximated as zero. A piecewise function was defined to roughly describe the flow evolution over a complete cycle (Figure S6, Supporting Informa- tion). Video S 6, Supporting Information shows the streamlines and the shear at the surface of the spheroid at different time points in the sequence. From this representation, we can con- clude that the shear rate applied to the spheroid surface remains below y = 30 s-1 throughout the entire cycle, corresponding to a shear stress of less than t = 0.03 Pa. In parallel, we conducted multiple CFD simulations at different inlet flow rates to estimate the maximal shear stress on the spheroid surface (Figure S8, Supporting Information). The system’s operating range, between 0.1 µL min-1 and 6 uL min-1, was chosen to prevent backflow, and within this range, the maximal shear stress remains below 0.3 Pa. While these simulations provide an initial approximation, they do not account for key factors such as spheroid deformabil- ity or surface roughness. Nevertheless, they enable us to estimate the maximum shear value on the spheroid surface, which reaches ~0.03 Pa-below the different thresholds for cellular damage re- ported in literature.[30-32] To experimentally validate the velocity profiles, we conducted micro-PIV (Particle Image Velocimetry) measurements during the pumping cycle (Figure 2H).
Velocity profiles were recorded at the peak flow rate, with images recorded at a specific point in the cycle across 560 pumping cycles since the flow is oscillating. From the velocity vector map, we extracted the velocity streamlines (Figure 2I), which closely matched those obtained through numerical sim- ulations (Figure 2G). The strong correlation between the exper- imental and numerical velocity profiles supports the reliabil- ity of the shear rate estimates derived from numerical simula- tions, which remain well within the viability limits reported in the literature.[30-32] Furthermore, we performed viability assays of pancreatic islets from cadaveric donors, a particularly frag- ile model, cultured in our microfluidic system at the identified operating point. The results confirmed that the spheroids re- mained viable after 4 days of culture under these conditions in the microfluidic device (Figure S9, Supporting Information). Alto- gether, these results indicate that the system effectively preserves spheroid integrity. Although we did not experimentally investi- gate the effect of shear stress on EV production in this study, pre- vious reports in 2D systems have suggested that increased shear stress can enhance EV secretion.[33,34] As discussed above, hydro- dynamic trapping in our platform limits the use of higher flow rates, since this could induce backflow and potentially dislodge the spheroid (Supporting Information Video S4). Nonetheless, exploring the influence of shear stress on EV secretion repre- sents an interesting avenue for future work and could be partic- ularly valuable for applications aimed at EV production, where fine-tuning shear conditions may enhance EV yield while main- taining spheroid integrity.
In conclusion, we have designed an architecture that in- corporates an automated micro-pump for the in situ collec- tion of secretions from the hydrodynamically trapped spheroid, along with a network of valves to precisely guide the fluid around the trapped spheroid. Our results demonstrate that this micro-pump operates without detrimental effect on spheroid
NANO · MICRO sma www.small-journal.com
A
4
1
B
C
Porous membrane
Cell, cell debris
Extracellular vesicles
· Biomolecules < 200 nm
1 um
5 mm
D
E
F
Particle concentration (part/mL)
4×10
Particle concentration (part/mL)
Bot (COC)
Before filtration
4×108
ns (p=0.057)
3×106
Ator filtration
Double-sided tape
3.5×108
2×10
Polycarbonate filter
3×10ª
Double-sided tape
1x10
2.5x10ª
Bot (COC)
0
0
200
Size (nm)
400
2×108
Starting solution
After filtration
physiology. The platform’s suitability is further highlighted by its successful applications to pancreatic islet culture-an especially fragile model highly sensitive to mechanical stress.[25] This capa- bility positions our system as a robust and reliable tool for ad- vanced spheroid culture and secretion analysis. Future develop- ments could include integrating a vascularization component to further enhance the system’s physiological relevance, as already explored in our laboratory.[33] Incorporating this element would provide valuable insights for studying its influence on EV secre- tion and transcriptomic cargo.
2.3. Pre-Purification of Spheroid Secretions by 200 nm Filtration
As an initial preparation step for spheroid secretions, we im- plemented a direct pre-purification step using 200 nm filtration to selectively remove larger biological particles, such as cell de- bris and whole cells (Figure 3A). Scanning electron microscopy (SEM) analysis of the filtration membranes revealed an average pore size of 168.5 ± 11.5 nm (Figure 3B). Given this appropri- ately sized pore dimension, we adopted a direct filtration strat- egy, where the fluid flows perpendicularly through the mem- brane pores. To further improve filtration efficiency, the mem- brane surfaces were modified with polyvinylpyrrolidone (PVP), a hydrophilic polymer that enhances fluid interaction and reduces resistance during filtration.
To seamlessly integrate this filtration step into the existing plat- form, we developed a modular, detachable filtration component
(Figure 3C). The microfluidic module was constructed with a cyclic olefin copolymer (COC) layer featuring micromachined in- lets. The filtration membrane is securely enclosed between two layers of double-sided adhesive tape, comprising precisely laser- cut fluidic channels. Fluid is introduced into the module through a 0.2 mm-wide inlet channel and directed toward a central cir- cular filtration region measuring 11 mm in diameter, where it passes through the membrane (Figure 3D). The upper chamber, with a height of 200 um, has a volume of ~19 uL - well suited to the limited secretion volumes typically collected from individual spheroids, minimizing sample dilution.
A common challenge in filtration systems is membrane clog- ging, which requires periodic replacement of the membrane. To prevent the filtration membrane from adhering to the COC sur- faces under pressure-a configuration that would obstruct fluid flow-we incorporated a circular row of micromachined pillars on both sides of the COC. Inspired by standard syringe filters, where micropillars are used as a support grid to hold the mem- brane in place and maintain a minimum gap between the mem- brane and the surrounding surface, we experimentally observed that without the pillars, the membrane collapsed onto the surface under flow, blocking fluid passage and preventing filtration. Con- sequently, we adopted a trial-and-error approach to optimize the pillar placement, ensuring a consistent gap between the mem- brane and the COC surfaces and thereby maintaining optimal fluid dynamics and filtration efficiency.
The performance of our filtration system was evaluated by an- alyzing spheroid secretion solutions before and after filtration
ADVANCED SCIENCE NEWS
using nanoparticle tracking analysis (NTA). The direct filtration process effectively removed particles larger than 200 nm, thus reducing potential contaminants such as cells or cellular debris that may also carry EV-associated proteins (e.g., CD63 or CD81) and could otherwise interfere with subsequent immunocapture (Figure 3E). To investigate potential EV losses during the filtra- tion process, we used a commercial lyophilized EV solution. This pre-purified standard allowed for accurate determination of the initial EV concentration and minimized variability due to non- EV contaminants. By measuring EV concentrations before and after filtration, we estimated potential losses attributable to ad- sorption within the filtration component. Remarkably, the system achieved a 98% EV recovery rate, indicating minimal loss during the process (Figure 3F).
These results demonstrate the filtration system’s ability to effi- ciently pre-purify spheroid secretions without significant loss of EVs. This approach provides a reliable, effective method for the pre-purification of spheroid secretions, offering a robust solution for downstream applications.
2.4. Extracellular Vesicles Isolation Using Immunomagnetic Capture
To isolate extracellular vesicles, we implemented an immuno- magnetic capture method, leveraging beads as the functional support. These beads, a few micrometers in diameter, were coated with antibodies specific for CD63 and CD81, commonly used as EV protein markers.[8] The magnetic beads are controlled by a magnet-equipped arm, integrated into the instrumentation bench and designed to interface directly with the fluidic cartridge (Figure 4A). A recessed groove, 1 mm deep, is positioned within the cartridge at the magnet’s location to optimize bead-magnet interaction. The magnet’s lateral placement ensures that beads remain anchored along the chamber edges, rather than aggre- gating at the center (Figure 4B), thereby maximizing magnetic retention and minimizing the influence of drag forces during chamber closure.
The fluidic design centers around two key chambers: a cal- ibration chamber with a precise 16 uL volume and a reaction chamber (Figure 4C,D). The calibration chamber ensures accu- rate reagent volume measurement, while the reaction chamber accommodates both reagents and beads, creating an optimal mi- croenvironment for incubation. To fully contain both beads and supernatants, the reaction chamber has a slightly larger volume (20 µL) than the calibration chamber. The controlled back-and- forth flow between these chambers promotes efficient mixing, as previously demonstrated in our laboratory.[27,34] In the first study, imaging was used to visualize mixing within the chambers and assess different dilution ranges,[34] while the second study re- ported on the mixing efficiency of magnetic beads for integrating an ELISA assay,[27] highlighting the versatility and effectiveness of this approach.
To prevent bead loss during operations, we implemented a carefully controlled pressure ramping system. Pressure was ap- plied in two stages: an initial increment from -150 to 135 mbar in 15 mbar steps, followed by a finer adjustment from 135 mbar to 150 mbar in 5 mbar increments. This gradual ramping en- sures controlled drainage, while securely retaining the beads.
The “Waste” outlet is connected to a negative pressure source, while the upstream pumping chamber minimizes dead volume between the inlets and the calibration chamber. This configura- tion guarantees precise calibration of reagent volumes at every stage of the process.
The protocol, conducted at 37 ℃ in a CO2 incubator, be- gins with the volume calibration of the magnetic beads, which are then transferred to the reaction chamber for supernatant drainage. Next, the secretions are precisely volume-calibrated and incubated with the beads. Following incubation, the supernatant is removed, and the bead-EV complexes undergo two washing steps with PBS solution. Finally, the purified complexes are re- suspended in PBS (Phosphate Buffer Saline) and collected at the outlet for downstream analysis (Figure 4E).
The efficacy of this isolation platform was validated using lyophilized EVs, with the protocol conducted both in a conven- tional tube and within the microfluidic system. In both config- urations, 500 uL of EV solution were treated with 20 uL anti- CD63 magnetic beads and 6 uL anti-CD81 magnetic beads. In both cases, a consistent capture efficiency of 60% was achieved, demonstrating the robustness of the microfluidic pro- tocol (Figure 4F). These results confirm the negligible adsorp- tion of EVs to both the COC and the hyperelastic membrane, as previously tested. Notably, the microfluidic approach yielded more reproducible results compared to the tube-based method, underscoring the advantages of automation and precision offered by this microfluidic device. Furthermore, cryogenic transmis- sion electron microscopy (cryo-TEM) provided direct visualiza- tion of extracellular vesicle morphology, revealing the character- istic spherical shape and bilayered membrane structure typically associated with EVs (Figure 4F). Complementary flow cytome- try measurements were performed to assess the relative distri- bution of 37 surface epitopes on the isolated extracellular vesi- cles (EVs). Minimal fluorescence was observed for the technical negative controls (REA and mIgG1), indicating low nonspecific binding and confirming the specificity of antibody labeling for EV-associated markers. EVs were positive for canonical EV mark- ers, including the tetraspanin CD9, a well-established exosomal marker.[8] Additionally, NCI-H295R-derived EVs expressed a set of epithelial and tumor-associated markers, including CD126 (IL- 6 receptor a), CD44, integrins CD29 and CD49e, CD146, SSEA-4, CD133/1, and HLA-ABC, consistent with the phenotype of the parental adrenocortical carcinoma cells. These markers reflect both cell adhesion properties and stem-like/tumorigenic charac- teristics. Conversely, immune-related markers (CD3, CD4, CD8, CD14, CD19, CD20, CD56, CD86, HLA-DR) and platelet mark- ers (CD41b, CD42a, CD62p) were absent, as expected. This anal- ysis confirmed the presence of specific EV-associated markers within the samples (Figure S10, Supporting Information), vali- dating both the efficiency of the isolation procedure and the EV phenotype of the captured particles.
We have developed an immunomagnetic capture protocol us- ing anti-CD63 and anti-CD81 antibodies, fully integrated within an automated microfluidic device. A common limitation of such protocols is the binding of EVs to beads, as reported in the litera- ture. Some studies have proposed reversible capture methods to preserve the EV integrity after elution.[35,36] However, since our study focuses on transcriptomic cargo and requires EV lysis, re- versible binding is not a concern in this context. The 60% EV
A
Magnet
B
₣ -P
+ +P
0.5 cm
C
D
Outlet
EXOTRAP VI
Magnet
Reaction chamber
Calibration chamber
Pumping chamber
Inlets
Waste
1 cm
E
1. Magnetic beads calibration
2. Secretions calibration
3. Mixing and incubation
Supernatant removal
4. Washes with PBS (2X)
5. Beads-EVs complexes recovery
Inlet/Outlet
Chamber
Supernatant removal
o
Valve
F
0.701
G
ns (p=0.35)
Capture efficiency
0.65
50 nm
50.0m
0.60
50 nm
0.55
0.50
Microfluidic
In tube
capture yield obtained in our study falls within the range reported in the literature, which varies from 45% to 99%, [37-39] indicating that our method performs comparably to other approaches. For example, Chen et al. reported a 45% yield for plasma-derived EVs using 1 um anti-CD63-coated beads in a microfluidic platform
with a micro-mixer.[38] Guo et al. achieved 75.8% capture effi- ciency using 1-3 um anti-CD63 beads combined with a bubble- driven micromixer, and observed a plateau in efficiency with increasing bead quantity, highlighting the importance of mix- ing time.[37] Yu et al. reported yields between 73% and 99%
ADVANCED SCIENCE NEWS
NANO · MICRO sma
depending on bead size (2, 4, or 7 um, all coated with anti-CD63 antibodies) using horseshoe-shaped micromixers, emphasizing the dependence of yield on bead size.[39] These studies demon- strate that EV yield depends on multiple factors, including bead size, antibody type, incubation time, and mixing efficiency. In our work, we systematically tested different temperatures, bead sizes, antibodies, and incubation times, and identified conditions that provide an optimal balance. Future improvements could involve exploring alternative micromixer geometries to enhance bead-EV interactions and further increase capture efficiency.
2.5. Transcriptomic Profiling of Spheroid-Derived Extracellular Vesicles
To fully showcase the platform’s capabilities, we integrated all previously developed modules into a cohesive and unified system (Figure 5A,B). The hydrodynamic trapping zone and integrated micro-pump were retained as core components, with the filtra- tion module strategically positioned downstream, followed by the EV immunomagnetic capture zone. The microfluidic cartridge was designed to accommodate two fully independent circuits, by duplicating the entire circuit layout within the cartridge. To en- sure secure assembly, micro-machined aluminum frames were used to hold the filtration module in place (Figure 5C), attaching it to the main cartridge via magnets. This modular design not only guarantees uniform clamping forces distribution but also allows for easy removal and replacement of individual modules.
We applied our platform to the study of NCI-H295R spheroids, a 3D model of adrenocortical carcinoma (ACC). ACC is char- acterized by aberrant activation of the Wnt/ß-catenin signaling pathway, particularly in its more aggressive forms.[40] The NCI- H295R cell line was genetically modified to express a doxycycline- inducible shRNA targeting ß-catenin.[41] Previous studies in 2D culture systems showed that silencing ß-catenin inhibits miR- 139-5p expression while maintaining stable levels of miR-483- 5p.[41] Here, we sought to validate these findings using a single 3D spheroid model (Figure 5D).
To ensure compatibility with the platform’s hydrodynamic trapping zone, the spheroid size evolution was closely monitored (Figure 5E). Optimal trapping requires spheroids within a spe- cific size range. This size constraint represents a limitation of the system, particularly for patient-derived samples, which exhibit greater size variability. Spheroids were introduced into the mi- crofluidic cartridge on day 4 of culture (D4). While 3D structures typically secrete higher EV levels compared to 2D cultures,[14] the number of EVs secreted by a single spheroid remains inher- ently low due to the limited number of secreting cells (~4000). To address this limitation, secretions were continuously collected over a 10-h period, during which EVs were progressively isolated. ~500 uL of spheroid secretions were treated using 20 uL anti- CD63 beads and 6 uL anti-CD81 beads. As soon as the secretion collection chamber was filled, the secretions were processed via immunomagnetic capture, allowing EVs to be captured. The su- pernatant was then discarded, while the EV-bead complexes were retained. This process was repeated throughout the 10-h collec- tion period, resulting in the gradual accumulation of EVs on the same set of beads, thereby effectively concentrating the EVs for downstream analysis. To estimate the maximum number of EVs
that can be captured on a magnetic bead, we first calculated the ratio between the bead surface area and the average EV footprint, weighted by the NTA size distribution. Assuming optimal hexag- onal packing (coverage fraction ( ~ 0.84), this analytical calcula- tion provides an upper bound for bead loading. To obtain a more realistic estimate that accounts for both size heterogeneity and steric hindrance, we implemented a random sequential adsorp- tion (RSA) simulation on a spherical surface. In this model, EV diameters were drawn from the experimental NTA distribution and sequentially placed on the bead surface at random positions. A new EV was accepted only if its spherical cap (0 = arcsin(r/Rb), with r the EV radius and Rb the bead radius) did not overlap with previously adsorbed EVs. This iterative process continued until saturation was reached, defined as 20 000 consecutive failed ad- sorption attempts. Simulations were repeated independently 100 times with different random seeds to compute the mean. The analytical estimate predicted a theoretical capacity of ~2337 EVs per bead under dense packing conditions. In contrast, the RSA simulation saturated at ~1937 EVs per bead, corresponding to a surface coverage fraction of ~0.46. This lower value, consistent with the jamming limit of RSA processes, likely provides a more representative measure, reflecting the steric constraints inherent to random adsorption. Using these approximations, the number of EVs that could theoretically be captured under these conditions is on the order of 109. Previous experiments have shown that a single NCI-H295R spheroid secretes ~108 EVs over 10 h (Figure S11, Supporting Information), indicating that the bead capacity exceeds the EV load and is therefore not a limiting factor in the capture process.
A 10-h incubation was selected as a practical duration to achieve sufficient EV concentrations (~108 EVs mL-1) for down- stream analyses, in line with the detection limits of techniques such as NTA. Preliminary experiments with single spheroids cul- tured in individual wells over 24 h confirmed this secretion rate, and the results have been included in Figure S10, Supporting In- formation. Following the 10-h collection phase, transcriptomic analyses were performed on both spheroids and the isolated EVs. In the spheroids, doxycycline-induced ß-catenin silencing led to the inhibition of miR-139-5p while miR-483-5p expression re- mains constant (Figure 5F,G), consistent with untreated con- trols and previous observations in 2D studies.[41] Previous studies have shown that the culture format can strongly influence EV se- cretion. While the comparison of EV secretion under DOX- or DOX+ conditions in 2D versus 3D cultures is currently being in- vestigated as part of ongoing work in our laboratory, several pub- lications have already reported that 3D spheroid cultures enhance EV secretion compared to 2D monolayers.[16,42,43] These observa- tions support the use of spheroid-derived EVs to capture physio- logically relevant secretion profiles, emphasizing the added value of our platform over conventional 2D culture systems. Similarly, transcriptomic analysis of EV-derived miRNA revealed that miR- 139-5p expression was inhibited following doxycycline treatment (Figure 5H) whereas miR-483-5p remains stable (Figure 5I). Fur- thermore, the ratio of miR-139-5p to miR-483-5P copy numbers expressed in isolated EV samples, in untreated and doxycycline- treated conditions, show that miR-483-5p levels remained con- sistently higher than those of miR-139-5p under all conditions (Figure 5J). These findings closely align with data from ACC patients.[7,41]
A
B
IMMUNOMAGNETIC CAPTURE OF EVs
PBS
Wash
Beads
SPHEROID TRAPPING
1 cm
200 nm FILTRATION
C
D
Spheroid formation
Doxycyclin treatment
Load spheroid into microfluidics
DO
D2
D4
D7
RNA extraction & qRT-PCR
D5 Secretion collection and EVs isolation in microfluidics
MICRO-PUMP
E
F
G
450
Not trasstod
400
Daxycyclina
180
120
*
**
Sprkrad sze (n)
Expression af miR-483-5p (%)
160
ns (p=0.46)
Expression of miR-139-50 (%)
100
350
**
140
80
300
120
60
250
100
80
40
200
60
20
150
D3
D4
D5
40
Untrexstod
Daxycydine
0
Untreated
Daxycycline
H
ns (p=0.16)
Expression at mik-483-50 (%)
Expression of miR-139-5p (%)
Copy number rata af miR-139-5g ta miR-483-50
ns (p=0.06)
160
120
0.9
₹140
100
0.8
120
0.7
100
80
0.6
0.5
80
60
0.4
60
40
0.3
40
0.2
20
20
0.1
Unterxx
Doxycycline
Undtrexstod
Daxycydine
Untested
Doxycycline
ADVANCED SCIENCE NEWS
NANO · MICRO sma
Analyzing EVs from a single spheroid allows our platform to capture heterogeneity at the spheroid level, providing insights into the collective behavior of cells within a physiologically rel- evant 3D microenvironment. Compared to single-cell studies, single-spheroid approaches better reproduce the microenviron- ment, including cell-cell interactions and stimuli such as hy- poxia, metabolic gradients, resulting in EVs that more closely resemble those found in native tissues. Previous studies have shown that EV cargo from 3D models is more similar to in vivo sources (e.g., plasma) than EVs from 2D cultures,[44,45] and that 3D spheroid-derived EVs are more effective at modulating target cells. [46-48] A future perspective of this work will leverage this au- tomated platform to perform comprehensive genomic analyses of secreted EVs, comparing profiles from 2D and 3D cultures. Overall, single-cell and single-spheroid approaches are comple- mentary: single-cell studies remain practical for basic mechanis- tic experiments, such as pathway perturbation or gene knock- down/overexpression, whereas single-spheroid-derived EVs pro- vide more physiologically relevant and clinically informative data for applications including biomarker discovery and EV-based therapeutics, particularly using patient-derived organoids.
Importantly, the device already integrates continuous perfu- sion, allowing controlled delivery of compounds directly to the culture. Although this study focused on endpoint analyses, the same setup can readily be used to monitor EV secretion kinetics in response to drug exposure. This capability opens the way to applications in drug screening and prognostic medicine, where monitoring temporal biomarkers dynamics is critical. The next step involves parallelizing the circuits to increase experimental throughput, enabling simultaneous testing of multiple condi- tions or experiments. Achieving this will require system refine- ments, including multiplexed valves control and optimization of key circuit components. This approach holds significant promise for drug screening applications, where high-throughput analysis is essential to accelerate discovery and evaluation processes.
The versatility of the platform opens opportunities for inte- grating additional EV analysis methods directly within the sys- tem, enhancing its analytical power and expanding the scope of possible studies. Building on this modular design, integrating a method for microfluidic PCR could further improve the plat- form by minimizing sample loss and enabling analysis as close to the EV source as possible. Future work will also focus on incor- porating single-EV microfluidic analysis approaches. Single-EV assays remain extremely valuable for characterizing heterogene- ity at the level of individual vesicles, although they are techni- cally demanding, require specialized platforms, and are relatively low-throughput. In our laboratory, we are developing SNR (sus- pended nanochannel resonator)-based sensors that measure the buoyant mass of individual particles as they flow through a mi- crofluidic channel along a microcantilever, providing highly sen-
sitive detection of EVs in solution.[49] Our integrated approach enables the isolation of single spheroids as well as the collection of their secreted EVs, allowing monitoring of secretion dynam- ics over time. These features support translational applications, including the study of EV-mediated mechanisms such as drug resistance, immune escape, and disease modeling using patient- derived organoids. These applications could be further strength- ened by integrating single-EV methods, which would combine the strengths of single-spheroid EV collection-physiological rel- evance and functional context-with the molecular precision of single-EV characterization.
3. Conclusions
We have developed an advanced, automated microfluidic plat- form that revolutionizes the study of spheroid cultures by en- abling the precise, real-time isolation of extracellular vesicles (EVs) directly from spheroid secretions. This state-of-the-art sys- tem integrates a micro-pump for continuous secretion collec- tion, preserving an optimal microenvironment for the spheroids throughout the process. Using a two-step protocol-200 nm pre- filtration followed by immunomagnetic capture targeting CD63 and CD81 proteins-we achieved an impressive 60% EV yield, achieving high-quality, reproducible results.
The platform’s capabilities were demonstrated in the study of adrenocortical carcinoma (ACC)-derived spheroids, demon- strating key transcriptomic insights. Specifically, we observed in- hibition of miR-139-5p expression in EVs following ß-catenin pathway suppression, while EVs-derived miR-483-5p levels re- mained consistently higher than those of EV-derived miR-139-5p. These findings validate the platform’s ability to detect dynamic biomarker shifts in EVs and show a strong correlation with pre- vious observations from 2D studies.
Beyond its current applications, the platform holds signifi- cant potential for scalability. Future developments will focus on increasing throughput, enabling large-scale drug discovery and contributing to targeted therapies development. By isolating EVs at the single-spheroid level, this system offers an unparalleled precision to investigate cellular heterogeneity, offering a substan- tial advantage over traditional bulk analyses. This capability is particularly valuable for rare, patient-derived organoids and the realm of personalized medicine, where studying EVs at the indi- vidual patient level holds the promise of delivering groundbreak- ing insights into treatment responses and therapeutic efficacy.
4. Experimental Section
Experimental Design: This study aimed to develop an advanced mi- crofluidic platform for isolating extracellular vesicles secreted by single spheroids. By integrating spheroid trapping, secretion collection, and EVs
Figure 5. Microfluidic analysis of EVs at the single-spheroid level. A) Photograph of the microfluidic device, B,C) Architecture of the microfluidic platform, top and perspective views, respectively. D) Overview of the protocol, E) Size evolution of NCI-H295R spheroids starting from 4000 cells at DO (Kruskal- Wallis test, n = 8, box = IQR, line = median, whiskers = 1.5 x IQR, p < 0.05 (*), p < 0.01 ( ** )), F,G) Transcriptomic analysis of NCI-H295 spheroids: normalized expression of miR-483-5p (F) and miR-139-5p (G) (Kolmogorov-Smirnov test, n = 8, box = IQR, line = median, whiskers = 1.5 x IQR, statistical significance was defined as p < 0.05, p < 0.001 ( *** )), H,I) Transcriptomic analysis of EVs secreted by a single spheroid normalized expression of miR-483-5p (H) and miR-139-5p (I) (Kolmogorov-Smirnov test, n = 10, box = IQR, line = median, whiskers = 1.5 x IQR, statistical significance was defined as p < 0.05, p < 0.001 ( *** )), J) Copy number ratio of EVs-derived miR-139-5p and miR-483-5p expressions in both conditions (Kolmogorov- Smirnov test, n = 10, box = IQR, line = median, whiskers = 1.5 x IQR, statistical significance was defined as p < 0.05).
ADVANCED SCIENCE NEWS
NANO · MICRO sma
isolation into a single automated system, we hypothesized that this ap- proach would enable the identification of spheroid-specific biomarkers, advancing applications in precision medicine.
Chemicals: Magnetic beads conjugated with anti-CD63 (4.5 um in di- ameter) and anti-CD81 (2.7 um in diameter) antibodies (Thermo Fisher Scientific, USA) were employed for immunomagnetic EV capture. Com- mercial lyophilized EVs (HBM-PC3-100/5, HansaBioMed) were reconsti- tuted in phosphate-buffered saline (PBS) for control experiments. These EVs have a mean diameter of ~100 nm. Polycarbonate membranes (WHA104117001, Cytiva) with a pore size of 200 nm were used for filtration.
Cell Culture and Generation of Spheroids: The NCI-H295R human adrenocortical carcinoma cell line was cultured on rat tail collagen-I- coated flasks (Corning, USA) in DMEM/F-12 medium (Thermo Fisher Scientific, USA). The medium was supplemented with 1% (v/v) Insulin- Transferrin-Selenium (BD Biosciences, USA), 2.5% (v/v) Nu-Serum (Corn- ing, USA), 100 UmL-1 penicillin, 50 µg mL-1 gentamicin, and 100 µg mL-1 streptomycin (Thermo Fisher Scientific, USA). To generate spheroids, ~4000 cells were seeded per well in 96-well plates coated with PolyHema (Sigma Aldrich) and centrifuged at 200 g for 5 min to promote spheroids formation.
The Human Pancreatic Stellate Cells HPaSteC were cultured in Stellate Cell Medium (SteCM, Cat. #5301). Spheroids were obtained using ~4000 cells in 200 uL medium per well in 96-wells ultra-low attachment plate (Corning, #4515). For both cell lines, spheroids were cultured in medium containing EV-depleted serum before secretion harvesting.
Design and Fabrication of the Microfluidic Cartridges: The microfluidic chip design followed an ISO/standard[50] credit card format (54 mm x 84 mm) and employed a modular building-block strategy to maximize scal- ability and interoperability. The hybrid chip consisted of three cyclic olefin copolymer (COC) layers for fluidic, pneumatic, and bottom components, separated by a hyperelastic Ecoflex membrane (~150 um thick) as previ- ously described.[24,27]
The fluidic and pneumatic layers were fabricated by micromilling COC sheets (TOPAS, USA) using a Datron M7HP micromilling machine (DA- TRON, Germany). After sequential ultrasonic cleaning in ethanol and deionized water, the layers were thermally bonded. The hyperelastic mem- brane was prepared by spin-coating Ecoflex 00-50 (Smooth-On, USA) on a silicon wafer. Bonding of the COC surfaces and Ecoflex membrane was achieved using O2 plasma activation, followed by thermal sealing at 3 Nm and 90 ℃ for 90 min.
Before use, the microchannels were pre-conditioned with 1% BSA (bovine serum albumin) diluted in PBS 1X to minimize non-specific ad- hesion.
Instrumentation: The custom-built instrument integrated five pres- sure controllers (Fluigent, France), 32 solenoid valves (SMC, Japan), and microcontrollers for high-precision fluidic and pneumatic actuation. A chip holder (Figure 2A) ensured seamless connectivity between pneumatic lines and the microfluidic chip. The system was maintained in a CO2 in- cubator during experiments to provide optimal conditions for cell culture (see S1). Magnetic bead manipulation was performed using a motorized arm with magnets controlled by an electric actuator.
The platform’s operations were managed by proprietary software (UFlu Factory), programmed in C++ and augmented with Python scripts for pro- tocol automation. The software enabled real-time control over valve ac- tuation, pressure modulation, and motorized arm positioning, ensuring reproducibility and streamlined workflows.
Membrane Characterization via Scanning Electron Microscopy: Porous membranes were characterized using SEM to assess pore size and spatial distribution. The membranes were coated with a 10 nm carbon layer on both sides. Imaging was conducted at an accelerating voltage of 2 kV and a current of 150 pA.
Numerical Simulations: The performances of the microfluidic chip were modeled using COMSOL Multiphysics (COMSOL Inc .; Sweden). The CAD geometry, generated in SolidWorks, was directly imported into COM- SOL to simulate fluid dynamics under steady-state and transient condi- tions. The Navier-Stokes equations were solved with boundary conditions
including inlet flowrate, outlet pressure, and no-slip conditions at channel walls.
Micro-Particle Image Velocimetry: Micro-PIV was used to capture ve- locity profiles within the chip. A pulsed Nd:YAG laser (532 nm) illumi- nated the microfluidic channels, and flow fields were captured using two synchronized cameras. Fluorescent tracers (1 um diameter, R020, Fluo- roMax, ThermoFisher Scientific) suspended in 1x TBS (tris buffer saline) were employed.
Images were pre-processed in Image], with adjustments for contrast, brightness, and background subtraction. Velocity profiles were determined using intercorrelation algorithms (Wereley and Meinhart, Purdue Univer- sity; used with permission), with ensemble-averaged data processed for precise flow characterization.[5]]
Extracellular Vesicles Isolation via in Tube Immunomagnetic Capture: A volume of 500 uL of secretion samples or EV-containing solutions was incubated with 20 uL of anti-CD63 magnetic beads (concentration 107 beads mL-1) and 6 uL of anti-CD81 magnetic beads (concentration 108 beads mL-1). The mixture was incubated for 6 h at 37 ℃ under gen- tle agitation using a laboratory rotator (LD79, Labinco) set at 12 rpm to promote efficient binding of EVs to the beads. After incubation, the beads were collected using a magnetic rack, and the supernatant was carefully removed. The bead-EV complexes were washed twice with phosphate- buffered saline to remove unbound material. Finally, the beads were re- suspended in PBS for subsequent analysis.
Extracellular Vesicles Analysis via Nanoparticle Tracking Analysis: EVs size and concentration were measured using a Nanosight NS3000 (Malvern Analytical, UK) equipped with NTA 3.1 software. Camera settings (e.g.,; Shutter: 1259, Gain: 245, FPS: 25) were kept constant across experi- ments. Each sample was analyzed in triplicate, with three 60 s acquisition periods per run.
qRT-PCR Analysis: Total RNA, including miRNA, was extracted using the miRNeasy Micro Kit (Qiagen, Germany). For bead-bound EVs, RNA lysis was performed directly on the magnetic beads using Qiazol reagent. Reverse transcription of 4 uL RNA was conducted with the TaqMan Mi- croRNA Reverse Transcription Kit (ThermoFisher Scientific) and miRNA- specific stem-loop primers.
Real-time qPCR was performed on a CFX96 Real-Time System (Bio- Rad, USA) using TaqMan assays specific to hsa-miR-483-5p (0 02338,) hsa-miR-139-5p (0 02289,) RNU48 (0 01006,) and cel-miR-39 (000200) as previously described.[32] For normalization, exogenous non-human cel- miR-39 was spiked into the EV samples in identical quantities, as no estab- lished circulating small RNA control exists. Normalization was made us- ing RNU4 for spheroid-derived RNA. The expression level of each miRNA is represented by the ratio of its copy number in the sample to the av- erage copy number in the untreated condition for each experiment. This approach allows for the evaluation of variability within the control group.
Flow Cytometry: To determine the proteomic profile of our EVs sam- ples, a bead-based flow cytometry technique was employed using the Mac- sPlexExo kit (Miltenyi Biotec), according to the manufacturer’s instruc- tions. This kit is capable of detecting 37 surface epitopes. The kit provides two technical negative controls (REA and mIgG1) to monitor nonspecific binding. Measurements were made using the Attune NxT (Thermofisher) cytometer.
Cryo Transmission Electron Microscopy: For these analyses, EVs were specifically eluted from the magnetic beads to allow downstream charac- terization thanks to ExoFlow buffer (System Biosciences). Samples were deposited on a plasma cleaned (Solarus, Gatan) EM grid with lacey carbon (S166-3, Ted Pella) through two successive deposits of 1 µL of sample and a 10 s waiting time between both deposits. Automated vitrification was ensured by a Vitrobot instrument (FEI Company) to perform cryo-fixation at a controlled 20.5 ℃ temperature, 100% relative humidity, blotting con- ditions and freezing velocity. Excess sample was removed with filter paper with the following blotting parameters: blot time: 3 s, relative blot force: - 5, wait time: 1 s, drain time: 0, blot total: 1. The grids were then plunged into liquid ethane in equilibrium with solid ethane at about - 185 ℃. The grids were stored in liquid nitrogen before mounting in a Gatan 626 cryo-holder for imaging with an FEI Technai Osiris transmission electron microscope
ADVANCED SCIENCE NEWS
operated at 200 kV. Images were recorded on a US 1000XP camera (Gatan) with 2048 x 2048 pixels.
Statistical Analysis: All statistical analyses were performed using GraphPad Prism version 9.0 (GraphPad Software, USA). Data normality was assessed using the Shapiro-Wilk test. Data are presented as box-and- whisker plots, where the box represents the interquartile range (IQR), the line indicates the median and whiskers extend to 1.5xIQR. For each ex- periment, the sample size (n) and the statistical test used are provided in the corresponding figure legend. Statistical significance was defined as p < 0.05 and exact p values are reported in the figures.
Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.
Acknowledgements
The authors gratefully acknowledge L. Golanski for her contribution to the acquisition of SEM images. We also thank C. E. Goujon and J. Porcherot for their involvement in programming and electronic board development. F. Kermarrec is sincerely acknowledged for her valuable assistance with the RNA extraction protocol. Finally, we extend our thanks to F. Navarro for his support throughout this work. This work used the facility of the Grenoble Instruct-ERIC Center (ISBG:UMS 3518 CNRS-CEA-UGA-EMBL) with sup- port from FRISBI (ANR-10 INSB-05-02) and GRAL (ANR-10-LABX49-01) within the Grenoble Partnership for Structural Biology (PSB). The electron microscope facility is supported by the Rhone-Alpes Region, the Fonda- tion Recherche Medicale (FRM), the fonds FEDER, the Centre National de la Recherche Scientifique (CNRS), the CEA, the University of Grenoble, EMBL, and the GIS-Infrastrutures en Biologie Sante et Agronomie (IBISA). This work was supported by the FOCUS OSP program of CEA. Scanning Electron Microscopy was carried out at the NanoCharacterization Plat- Form (PFNC) and was supported by the “Recherches Technologiques de Base” program of the French National Research Agency (ANR).
Conflict of Interest
The authors declare no conflict of interest.
Author Contributions
Conceptualization: M.H., N.C., Y.F., V.A .; Microfluidic design: M.H., Y.F., J.K .; Microfabrication: F.Boizot, M.B .; microPIV experiments and numer- ical simulations: M.H., F. Bottausci, Y.F .; Filtration module development: M.H .; Immunomagnetic capture protocol development: M.H., M.C., P.L .; PCR experiments and cell culture: J.D .; Extracellular vesicles characteri- zation: M.H .; Funding acquisition: V.A., Y.F .; Data curation: M.H .; Formal analysis: M.H., N.C .; Writing-original draft: M.H. Writing-editing: M.H., N.C., J.D., F. Bottausci, M.C., Y.F., V.A.
Data Availability Statement
The data that support the findings of this study are available within the article and its Supporting Information. Additional data are available from the corresponding author upon reasonable request.
Keywords
extracellular vesicles, microfluidics, microphysiological systems, precision medicine, single-spheroid
Received: July 7, 2025
Revised: October 3, 2025
Published online: October 14, 2025
[1] J. Lotvall, H. Valadi, Cell Adhes. Migr. 2007, 1, 156.
[2] J. Liu, Y. Zhang, Y. Tian, W. Huang, N. Tong, X. Fu, Theranostics 2022, 12, 1342.
[3] R. E. Sayed, S., J. Cristante, L. Guyon, J. Denis, O. Chabre, N. Cherradi, Cancers 2021, 13, 2680.
[4] A. Rodríguez-Martínez, D. De Miguel-Pérez, F. G. Ortega, J. L. García- Puche, I. Robles-Fernández, J. Exposito, J. Martorell-Marugan, P. Carmona-Sáez, M. D. C. Garrido-Navas, C. Rolfo, H. Ilyine, J. A. Lorente, M. Legueren, M. J. Serrano, Breast Cancer Res. 2019, 21, 21.
[5] S. A. Melo, L. B. Luecke, C. Kahlert, A. F. Fernandez, S. T. Gammon, J. Kaye, V. S. LeBleu, E. A. Mittendorf, J. Weitz, N. Rahbari, C. Reissfelder, C. Pilarsky, M. F. Fraga, D. Piwnica-Worms, R. Kalluri, Nature 2015, 523, 177.
[6] M. Fassnacht, O. M. Dekkers, T. Else, E. Baudin, A. Berruti, R. R. De Krijger, H. R. Haak, R. Mihai, G. Assie, M. Terzolo, Eur. J. Endocrinol. 2018, 179, G1.
[7] O. Chabre, R. Libé, G. Assie, O. Barreau, J. Bertherat, X. Bertagna, J .- J. Feige, N. Cherradi, Endocrine-Related Cancer 2013, 20, 579.
[8] C. Théry, K. W. Witwer, E. Aikawa, M. J. Alcaraz, J. D. Anderson, R. Andriantsitohaina, A. Antoniou, T. Arab, F. Archer, G. K. Atkin- Smith, D. C. Ayre, J .- M. Bach, D. Bachurski, H. Baharvand, L. Balaj, S. Baldacchino, N. N. Bauer, A. A. Baxter, M. Bebawy, C. Beckham, A. Bedina Zavec, A. Benmoussa, A. C. Berardi, P. Bergese, E. Bielska, C. Blenkiron, S. Bobis-Wozowicz, E. Boilard, W. Boireau, A. Bongiovanni, et al., J. Extracell. Vesicles 2018, 7, 1535750.
[9] H. Xu, D. Jiao, A. Liu, K. Wu, J. Hematol. Oncol. 2022, 15, 58.
[10] E. Verjans, J. Doijen, W. Luyten, B. Landuyt, L. Schoofs, J. Cell. Physiol. 2018, 233, 2993.
[11] G. Bordanaba-Florit, I. Madarieta, B. Olalde, J. M. Falcón-Pérez, F. Royo, Cancers 2021, 13, 307.
[12] M. Schuster, F. K. Braun, D. M .- L. Chiang, C. Ludwig, C. Meng, C. Grätz, B. Kirchner, M. Proescholdt, P. Hau, O. K. Steinlein, M. W. Pfaffl, M. J. Riemenschneider, M. Reithmair, Front. Immunol. 2024, 15.
[13] S. Kwak, C. L. Song, J. Lee, S. Kim, S. Nam, Y .- J. Park, J. Lee, Biomate- rials 2024, 307, 122522.
[14] S. Abdollahi, Biotechnol. Bioeng. 2021, 118, 1029.
[15] B. B. Silver, R. Fannin, K. Gerrish, E. Tokar, Curr. Res. Toxicol. 2025, 8, 100229.
[16] J. Tu, Biochem. Biophys. Rep. 2021, 26, 101026.
[17] Y. Man, Y. Liu, Q. Chen, Z. Zhang, M. Li, L. Xu, Y. Tan, Z. Liu, Adv. Healthcare Mater. 2024, 13, 2401843.
[18] E. M. Clarissa, S. Kumar, J. Park, M. Karmacharya, I .- J. Oh, M .- H. Kim, J .- S. Ryu, Y .- K. Cho, ACS Nano 2025, 19, 5526.
[19] X. Qin, Y. Xiang, L. Mao, Y. Yang, B. Wei, H. Lu, X. Li, Y. Zhang, F. Yang, ACS Nano 2024, 18, 14569.
[20] K .- C. Lin, H .- Y. Lin, C .- Y. Yang, Y .- L. Chu, R .- H. Xie, C .- M. Wang, Y .- L. Tseng, H. Chen, J. H. Y. Chung, J .- W. Yang, G .- Y. Chen, ACS Nano 2025, 19, 2228.
[21] L. Xu, H. Ding, S. Wu, N. Xiong, Y. Hong, W. Zhu, X. Chen, X. Han, M. Tao, Y. Wang, D. Wang, M. Xu, D. Huo, Z. Gu, Y. Liu, ACS Nano 2024, 38, 26201.
[22] P. M. Misun, B. Yesildag, F. Forschler, A. Neelakandhan, N. Rousset, A. Biernath, A. Hierlemann, O. Frey, Adv. Biosyst. 2020, 4, 1900291.
[23] H .- R. Seo, H. E. Jeong, H. J. Joo, S .- C. Choi, C .- Y. Park, J .- H. Kim, J .- H. Choi, L .- H. Cui, S. J. Hong, S. Chung, D .- S. Lim, Sci. Rep. 2016, 6, 28832.
[24] C. Quintard, E. Tubbs, J .- L. Achard, F. Navarro, X. Gidrol, Y. Fouillet, Biosens. Bioelectron. 2022, 202, 113967.
[25] M. Hut, Microfluidic Devices for Single Organoid Culture and Isola- tion of Secreted Extracellular Vesicles. phdthesis, Université Greno- ble Alpes [2020 -… ], 2024. https://theses.hal.science/tel-05110611 (accessed: August 2025).
ADVANCED SCIENCE NEWS
NANO · MICRO sma
[26] S. Chen, S.-C. Shiesh, G.-B. Lee, C. Chen, PLoS One 2020, 15, 0229610.
[27] C. Parent, P. Laurent, C .- E. Goujon, X. Mermet, A. Keiser, F. Boizot, R. Charles, L. Audebert, Y. Fouillet, M. Cubizolles, Lab Chip 2022, 22, 3147.
[28] M. A. Unger, H .- P. Chou, T. Thorsen, A. Scherer, S. R. Quake, Science 2000, 288, 113.
[29] W .- H. Tan, S. Takeuchi, Proc. Natl. Acad. Sci. USA 2007, 104, 1146.
[30] M. Klak, P. Kowalska, T. Dobrzański, G. Tymicki, P. Cywoniuk, M. Gomółka, K. Kosowska, T. Bryniarski, A. Berman, A. Dobrzyń, W. Sadowski, B. Górecki, M. Wszoła, Micromachines (Basel) 2021, 12, 304.
[3]] F. Montel, M. Delarue, J. Elgeti, D. Vignjevic, G. Cappello, J. Prost, New J. Phys. 2012, 14, 055008.
[32] T. Masiello, A. Dhall, L. P. M. Hemachandra, N. Tokranova, J. A. Melendez, J. Castracane, Cells 2018, 7, 277.
[33] C. Quintard, E. Tubbs, G. Jonsson, J. Jiao, J. Wang, N. Werschler, C. Laporte, A. Pitaval, T .- S. Bah, G. Pomeranz, C. Bissardon, J. Kaal, A. Leopoldi, D. A. Long, P. Blandin, J .- L. Achard, C. Battail, A. Hagelkruys, F. Navarro, Y. Fouillet, J. M. Penninger, X. Gidrol, Nat. Commun. 2024, 15, 1452.
[34] C. Parent, F. Boizot, M. Cubizolles, N. Verplanck, J .- L. Achard, Y. Fouillet, Sens. Actuators, B 2018, 261, 106.
[35] N. Zhu, Y. Zhang, J. Cheng, Y. Mao, K. Kang, G. Li, Q. Yi, Y. Wu, J. Colloid Interface Sci. 2022, 611, 462.
[36] D. Brambilla, L. Sola, A. M. Ferretti, E. Chiodi, N. Zarovni, D. Fortunato, M. Criscuoli, V. Dolo, I. Giusti, V. Murdica, K. Kluszczyńska, L. Czernek, M. Düchler, R. Vago, M. Chiari, Anal. Chem. 2021, 93, 5476.
[37] X. Guo, F. Hu, Z. Yong, S. Zhao, Y. Wan, B. Wang, N. Peng, Anal. Chem. 2024, 96, 7212.
[38] Y .- S. Chen, Y .- D. Ma, C. Chen, S .- C. Shiesh, G .- B. Lee, Lab Chip 2019, 19, 3305.
[39] H. Yu, J. Kim, J. Yu, K .- A. Hyun, J .- Y. Lim, Y .- J. Yoon, S. Park, H .- I. Jung, BioChip J. 2023, 17, 468.
[40] G. Assié, E. Letouzé, M. Fassnacht, A. Jouinot, W. Luscap, O. Barreau, H. Omeiri, S. Rodriguez, K. Perlemoine, F. René-Corail, N. Elarouci, S. Sbiera, M. Kroiss, B. Allolio, J. Waldmann, M. Quinkler, M. Mannelli, F. Mantero, T. Papathomas, R. De Krijger, A. Tabarin, V. Kerlan, E. Baudin, F. Tissier, B. Dousset, L. Groussin, L. Amar, E. Clauser, X. Bertagna, B. Ragazzon, et al., Nat. Genet. 2014, 46, 607.
[41] J. Cristante, S. R. El Sayed, J. Denis, B. Ragazzon, C. Hantel, O. Chabre, L. Guyon, N. Cherradi, bioRxiv 2023, https://doi.org/10. 1101/2023.02.10.527992.
[42] M. Kim, H .- W. Yun, D. Y. Park, B. H. Choi, B .- H. Min, Tissue Eng. Re- gener. Med. 2018, 15, 427.
[43] K. Cook, H. Li, Nanoscale 2025, 17, 15110.
[44] S. Thippabhotla, C. Zhong, M. He, Sci. Rep. 2019, 9, 13012.
[45] P. Ramamoorthy, S. M. Thomas, G. Kaushik, D. Subramaniam, K. M. Chastain, A. Dhar, O. Tawfik, A. Kasi, W. Sun, S. Ramalingam, S. Gunewardena, S. Umar, J. M. Mammen, S. B. Padhye, S. J. Weir, R. A. Jensen, G. S. Sittampalam, S. Anant, Cancer Res. 2019, 79, 1681.
[46] L. Yan, X. Wu, Cell Biol. Toxicol. 2020, 36, 165.
[47] S. Hu, Z. Li, H. Lutz, K. Huang, T. Su, J. Cores, P .- U. C. Dinh, K. Cheng, Sci. Adv. 2020, 6, aba 1685.
[48] W. Chen, P. Wu, C. Jin, Y. Chen, C. Li, H. Qian, J Nanobiotechnology 2024, 22, 215.
[49] S. Olcum, N. Cermak, S. C. Wasserman, K. S. Christine, H. Atsumi, K. R. Payer, W. Shen, J. Lee, A. M. Belcher, S. N. Bhatia, S. R. Manalis, Proc. Natl. Acad. Sci. USA 2014, 111, 13 10.
[50] ISO 22916:2022. ISO. https://www.iso.org/fr/standard/74157.html (accessed : May 2025).
[5]] S. T. Wereley, C. D. Meinhart, Exp. Fluids 2001, 31, 258.