Glomerular routing of tumor-derived extracellular vesicles substantiates urinary biopsy

Abstract

Urinary small extracellular vesicles (sEVs), which can reflect systemic conditions, hold great promise for noninvasive cancer diagnostics, yet the mechanism by which tumor-derived sEVs reach urine remains unclear. Here, we demonstrate that the glomerulus actively transcytoses circulating tumor-derived sEVs into urine. Using CRISPR guide RNA–tagged glioma sEVs and bioluminescent/fluorescent green-enhanced nano-lantern (GeNL)–tagged lung and pancreatic cancer sEVs, we tracked their journey from tumors to urine in multiple mouse models. In vivo and in vitro analyses revealed endocytic uptake and transcytotic release by glomerular cells, accompanied by changes in sEV size and surface composition. GeNL-tagged sEVs consistently showed higher signals in urine than plasma, indicating selective excretion. These findings redefine the glomerulus as a dynamic regulator of sEV processing and establish a mechanistic foundation for urinary liquid biopsy.

Tumor-derived extracellular vesicles cross the kidney filter into urine via glomerular transcytosis, enabling urinary diagnostics.

INTRODUCTION

The use of noninvasive liquid biopsy has become increasingly important in personalized medicine, as it enables disease monitoring with minimal patient burden. Among various body fluids, urine offers a uniquely accessible window into systemic pathophysiology because of its ease of collection and potential for longitudinal sampling. Small extracellular vesicles (sEVs), lipid bilayer–enclosed nanoparticles ranging from 30 to 200 nm, are secreted by virtually all cell types and carry cargoes such as small RNAs, mRNAs, and proteins reflective of their cellular origin (1–4). A growing body of research has demonstrated the diagnostic utility of urinary sEVs not only for urological conditions including nephritis (5, 6) and cystitis (7) but also for systemic diseases such as cancer (8–13), diabetes (14, 15), and neurological disorders (16, 17). Notably, even tumors located far from the urinary tract, such as gliomas, have been detected using urine sEV profiling (13, 18). These findings suggest that disease-associated sEVs can enter the circulation, reach the kidneys, and be excreted into urine. However, despite these insights, there remains a critical gap: No study has directly demonstrated the presence of cancer cell–derived sEVs in urine.

Despite their diagnostic potential, the direct demonstration of cancer cell–derived sEVs in urine has remained elusive because of biological and technical limitations. sEVs typically range from 30 to 200 nm in diameter (19–21), while the glomerular filtration barrier, comprising endothelial cells, a dense basement membrane, and podocytes, generally excludes molecules larger than 6 to 8 nm (22–25). This size mismatch has cast doubt on whether intact sEVs can traverse the renal filter under physiological conditions. However, previous studies have reported the presence of synthetic nanoparticles as large as 200 nm (26, 27) and even carbon nanotubes (28) in the urine of experimental animals, and nanoparticle accumulation has been observed beyond the glomerular barrier (29). These findings suggest that particles of sEV size may bypass the filtration barrier, potentially under specific physiological or pathological conditions. Nonetheless, tracing cancer-derived sEVs into urine remains particularly challenging because of two fundamental barriers. The first barrier is the relative scarcity of target sEVs: Most urinary sEVs originate from epithelial cells within the kidney and urinary tract, whereas sEVs from distal tumors represent only a minor fraction. The second barrier relates to their biological behavior: As vehicles of intercellular communication, sEVs are frequently internalized and reprocessed by various tissues, where their molecular contents may be degraded or repackaged before further circulation. These factors collectively obscure the origin and route of tumor-derived sEVs excreted into urine.

To address these limitations, we conducted two independent experiments using complementary reporter systems designed to directly and sensitively trace the excretion of cancer-derived sEVs from tumors to urine. The first system uses CRISPR guide RNA (gRNA) as a molecular tracer selectively packaged into sEVs, enabling detection by quantitative polymerase chain reaction (qPCR). This approach simulates current clinical liquid biopsy strategies on the basis of tumor-derived small RNAs and allows for highly specific tracking of sEVs released from engineered GL261 murine glioma cells in an orthotopic mouse model. These labeled sEVs were found to accumulate in the kidneys and were subsequently detected in urine. The second system uses green-enhanced nanoluciferase (GeNL), a dual luminescent and fluorescent fusion protein composed of NanoLuc and mNeonGreen, which enables high-sensitivity quantification and compatibility with real-time optical imaging. GeNL-tagged sEVs secreted by A549 (human lung) and Panc-1 (human pancreatic) cancer cells were found in notably higher abundance in urine than in plasma in both orthotopic and subcutaneous xenograft models. Together, these orthogonal strategies provide the first direct evidence that cancer cell–derived sEVs originating from distant tumors can be excreted into urine. Furthermore, our findings offer mechanistic insight into the renal trafficking of cancer cell–derived sEVs and establish a robust foundation for urinary sEVs as clinically relevant biomarkers in noninvasive cancer diagnostics.

RESULTS AND DISCUSSION

In vivo tracking of glioma-derived sEVs from tumor secretion to urinary excretion using a gRNA-loaded system in an orthotopic mouse model

To investigate whether sEVs secreted by brain tumors can travel through the bloodstream and ultimately appear in urine, we established an orthotopic glioblastoma model in which GL261 mouse glioma cells expressing CD63 fused to dCas9 (CD63-dCas9) and a synthetic CRISPR gRNA tracer were transplanted into the cerebrum of C57BL/6 mice (Fig. 1A and fig. S1). This system builds upon our recent finding that coexpression of CD63-dCas9 and CRISPR gRNA enables efficient packaging of RNA tracers into sEVs via the RNA binding affinity of dCas9 tethered to an sEV-associated membrane protein (30). The qPCR analysis confirmed that the tracer gRNA was detectable in both the cellular and sEV fractions only when GL261 cells were cotransduced with CD63-dCas9 and gRNA, validating selective incorporation into secreted vesicles (Fig. 1, B and C). Cryo–electron microscopy (cryo-TEM) and nanoflow cytometry (nanoFCM) findings revealed that sEVs derived from GL261/dCas9/gRNA cells were spherical vesicles with diameters ranging from 40 to 100 nm, a size distribution consistent with sEVs secreted by unmodified GL261 cells (Fig. 1, D and E). On the basis of qPCR quantification using a plasmid standard, the gRNA content was estimated at 0.03 copies per sEV, allowing conversion of gRNA copy numbers into particle counts for downstream analyses (Fig. 1F). Furthermore, immunostaining-based single-particle profiling showed elevated levels of CD63-positive sEVs without significant changes in CD9, further supporting the finding that CD63-dCas9 is successfully incorporated into the sEV population (Fig. 1G). The gRNA signal persisted after ribonuclease (RNase) treatment, suggesting that the tracer gRNA was at least partially protected, likely through encapsulation within the lipid bilayer of the sEVs. These results establish a robust system for the selective labeling and quantification of tumor-derived sEVs, enabling in vivo tracking of their biodistribution using qPCR.

Fig. 1. Detection of gRNA encapsulated in GL261/dCas9/gRNA cell–derived EVs.

(A) Schematic overview of the GL261 glioma cell line engineered to secrete gRNA-loaded sEVs and the strategy for tracking their excretion in urine. (B) Intracellular gRNA expression levels in GL261 and GL261/dCas9/gRNA cells. Statistical analysis was performed using the unpaired Mann-Whitney test (*P < 0.05). (C) gRNA expression levels in the culture supernatant of GL261 and GL261/dCas9/gRNA cells. *P < 0.05, Mann-Whitney test. (D) Cryo-TEM images of sEVs derived from GL261 (left) and GL261/dCas9/gRNA (right). Scale bars, 100 nm. (E) Size distribution of sEVs from GL261 (left) and GL261/dCas9/gRNA (right) measured by nanoFCM. Average particle size, 85.98 ± 22.54 nm for GL261-derived EVs and 86.15 ± 29.74 nm for GL261/dCas9/gRNA-derived EVs. No significant difference was observed between the two groups (P = 0.75). (F) Estimated gRNA copy numbers per EV particle calculated from qPCR measurements in GL261/dCas9/gRNA sEVs (n = 3). (G) Surface expression of EV markers CD63 in sEVs from GL261 (left, 4896 events) and GL261/dCas9/gRNA (right, 5136 events). (H) Gadolinium-enhanced T1-weighted MRI image showing brain tumor formation in a representative mouse. (I) Confocal fluorescence image of a brain section from a tumor-bearing mouse showing enhanced green fluorescent protein–positive GL261/dCas9/gRNA cells (green) and nuclear counterstaining (blue). Scale bars, 100 μm (inset) and 1 mm. (J) Concentration of GL261/dCas9/gRNA–derived sEVs in the urine of individual mice (N ≥ 3). (K) Relative gRNA copy numbers detected in kidney samples of tumor-bearing mice as measured by qPCR. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. Statistical significance was determined using the unpaired Mann-Whitney test (**P < 0.01). [(B), (C), (J), and (K)] Data points represent results from independent experimental runs, with error bars indicating the SD for (B) (n = 4), (C) (n = 4), (J) (n ≥ 3), and (K) (n = 5).

To track the urinary excretion of tumor-derived sEVs, we first confirmed the successful transplantation of GL261/dCas9/gRNA cells into the cerebrum of C57BL/6 mice using independent imaging methods: Magnetic resonance imaging (MRI) showed a hyperintense lesion consistent with tumor formation, while confocal microscopy detected green fluorescence from coexpressed enhanced green fluorescent protein in GL261/dCas9/gRNA cells (Fig. 1, H and I, and fig. S1). Urine samples were collected every 3 days starting on day 7 posttransplantation, resulting in nine time points. After RNase treatment to eliminate unprotected RNAs, total RNA was extracted from urinary sEVs, and gRNA levels were quantified by qPCR to estimate the concentration of gRNA-containing sEVs on the basis of Ct (cycle threshold) values (fig. S2). gRNA was consistently detected at high levels in the urine of tumor-bearing mice over an extended period (Fig. 1J).

On the basis of the qPCR quantification of gRNA in urinary sEVs from tumor-bearing mice, the average concentration of tumor-derived sEVs was estimated at 2.5 × 10⁵ particles/ml. Given that the total number of sEVs in urine recovered by ultracentrifugation for the present conditions was about 5.6 × 10⁸ particles/ml, a value consistent with the reported physiological range of urinary sEVs in mice, typically spanning 10⁸ to 10¹¹ particles/ml, tumor-derived sEVs accounted for ~0.044% of the total. In contrast, when the transplanted tumor cells failed to engraft, gRNA levels peaked transiently on day 10 and then declined over time (fig. S3). Although no strict correlation was observed between tumor volume and urinary gRNA abundance, the gradual loss of gRNA in cases of tumor regression suggests that gRNA tracking may serve as a surrogate marker of tumor persistence. These findings are consistent with prior reports on the presence of brain tumor–derived sEVs in urine (13, 18) and support the use of gRNA as a mimic of microRNA-based liquid biopsy for noninvasive tumor monitoring.

To further investigate the in vivo distribution of tumor-derived sEVs before their appearance in urine, we performed in situ hybridization (ISH) and qPCR to detect gRNA signals in kidney and lung tissues from tumor-bearing mice. ISH affirmed that antisense probes yielded stronger signals than negative control probes in both organs, particularly in mice bearing brain tumors (fig. S4). Notably, this signal enhancement was most pronounced in the lungs, which is consistent with a prior report that indicated that glioblastoma-derived vesicles preferentially localize to the lungs because of their anatomical proximity to the brain (31). Supporting the ISH data, qPCR analysis of RNA extracted from kidney and lung tissues showed elevated gRNA levels in tumor-bearing mice relative to sham-operated controls (Fig. 1K and fig. S4). These results suggest that sEVs secreted from brain tumors can traverse the blood-brain barrier in a retrograde manner (32, 33), subsequently accumulate in peripheral organs such as the kidneys and lungs, and ultimately be excreted into the urine.

In vitro analysis of transcytotic excretion of sEVs through glomerular cells

To investigate the mechanism by which sEVs cross the glomerular filtration barrier after accumulating in the kidneys, we examined their uptake and excretion using glomerular endothelial cells (GECs) and podocytes. The glomerular filtration barrier consists of three layers, namely GECs, the basement membrane, and podocytes, with estimated pore sizes of 60 to 80 nm, 300 to 350 nm, and ~12 nm, respectively (22–25). Given that sEVs typically measure around 100 nm in diameter, they are unlikely to pass through these pores via paracellular routes. However, previous studies have reported that lipid nanoparticles as large as 190 nm can traverse the glomerular barrier (27–29), suggesting that transcellular transport via transcytosis may provide a feasible route. Supporting this hypothesis, transcytosis-mediated vesicle transport has been observed across other biological barriers such as the blood-brain barrier (32, 33) and the intestinal epithelium (34), both of which restrict the passage of large macromolecules. On the basis of these considerations, we hypothesized that sEVs released from tumor cells cross the glomerular barrier primarily via transcytosis rather than by passive diffusion through intercellular junctions.

To determine whether glomerular cells internalize sEVs, we examined the uptake of GL261/dCas9/gRNA–derived sEVs by cultured mouse GECs and podocytes differentiated from SVI cells (fig. S5) (35). Cells were incubated with culture medium containing GL261/dCas9/gRNA sEVs, and intracellular gRNA levels were quantified by qPCR at various time points to assess uptake efficiency (Fig. 2A). The intracellular gRNA signal increased with longer incubation, confirming time-dependent sEV uptake by both GECs and podocytes (Fig. 2B). To evaluate whether this uptake was mediated by endocytosis, cells were co-incubated with sEVs at 4°C, a condition known to inhibit endocytic activity. Under this condition, intracellular gRNA levels were markedly reduced compared to those observed at 37°C (Fig. 2C), indicating that endocytosis is the primary mechanism by which glomerular cells internalize sEVs. These findings support the hypothesis that sEVs cross the glomerular barrier via an active, endocytosis-dependent transcellular route.

Fig. 2. Uptake and release of GL261/dCas9/gRNA sEVs by GECs and podocytes.

(A) Schematic representation of the proposed mechanism for uptake and release of GL261/dCas9/gRNA–derived sEVs in GECs and podocytes. h, hours. (B) Relative gRNA levels in GECs after coculturing with GL261/dCas9/gRNA EVs at 37°C. (C) Relative gRNA levels in GECs after coculturing at 4°C to inhibit endocytosis. (D) Relative gRNA levels in the culture supernatant of glomerular cells 20 hours after medium replacement. (E) Relative gRNA levels in glomerular cells 20 hours after medium replacement. [(B) to (E)] The relative gRNA levels were normalized to the gRNA quantity detected at 5 min postcoculture. Data points represent results from independent experimental runs (n = 3), with error bars indicating the SD. (F) Size distribution of sEVs released from human podocytes without (top) or with (bottom) prior exposure to GL261/dCas9/gRNA sEVs, as measured by nanoFCM. (G) Detection of mouse CD63 in sEVs collected from human podocytes before exposure to mouse-derived sEVs. The observed signal represents background reactivity resulting from species cross-reactivity of the anti-mouse CD63 antibody (810 events). (H) Increased signal of mouse CD63 detected in sEVs released from human podocytes after exposure to GL261/dCas9/gRNA sEVs, suggesting the rerelease of internalized sEV components (896 events).

To investigate whether sEVs internalized by glomerular cells are subsequently released, we performed a time-course assay using GL261/dCas9/gRNA sEVs. Following co-incubation with these sEVs, GECs and podocytes were washed and cultured in extracellular vesicle (EV) and serum-free medium for an additional 20 hours (Fig. 2A). As the duration of initial co-incubation increased, higher levels of gRNA were detected in the supernatant after medium replacement, indicating that internalized gRNA was released back into the extracellular environment (Fig. 2D). Conversely, intracellular gRNA levels decreased significantly after 20 hours of incubation in serum-free medium (Fig. 2E). These results suggest that a portion of the sEVs taken up by glomerular cells is excreted, while another portion undergoes degradation within the cells. Together with the evidence of endocytosis-dependent uptake, these findings support the idea that glomerular cells engage in bidirectional transcellular trafficking of sEVs, encompassing both endocytic uptake and subsequent exocytosis.

To determine whether internalized sEVs are rereleased in a form that retains tumor-derived components, we examined whether mouse-specific sEV markers could be detected in EVs secreted by human podocytes following exposure to GL261/dCas9/gRNA sEVs. This cross-species design allows discrimination between exogenous mouse CD63 and endogenous human CD63, which is not possible in mouse podocytes. Human podocytes were incubated with GL261/dCas9/gRNA sEVs, and the sEVs subsequently released into the medium were analyzed. Although the overall size distribution of sEVs remained unchanged regardless of mouse sEV exposure (Fig. 2F), immunodetection analysis identified mouse CD63 in sEVs released by human podocytes (Fig. 2, G and H). While a low-level mouse CD63 signal was also detected in control samples, accounting for ~5.7% of the treated signal and likely reflecting species cross-reactivity of the anti-mouse CD63 antibody, the markedly elevated signal in the treated group suggests that a portion of the internalized sEV components was rereleased. Notably, although gRNA was not quantitatively assessed in this coculture experiment with human podocytes, previous results (Fig. 1, B and C) demonstrated the presence of gRNA both intracellularly and extracellularly at the corresponding stage. Given that gRNA is incorporated into sEVs via binding to the CD63-dCas9 fusion protein and that both gRNA and mouse CD63 were detected extracellularly, it is likely that a portion of the internalized sEVs escaped intracellular degradation and were resecreted. While we did not directly observe hybrid vesicles, these findings raise the possibility that internalized tumor-derived sEVs may acquire host-derived features during intracellular processing in glomerular cells.

Microphysiological system for analyzing the selective transcytosis of sEVs across glomerular barriers

As described above, our experiments showed that GL261/dCas9/gRNA sEVs are internalized by GECs and podocytes via endocytosis and subsequently rereleased. Therefore, we developed a microphysiological glomerulus system to directly observe sEV passage across the glomerular filtration barrier. Two complementary devices were constructed: an insert-well model in which mouse GECs and podocytes were cultured on opposite sides of a polycarbonate membrane (pore size, 1 μm; density, 1.6 × 10⁶ pores/cm²; thickness, 10 μm) and a microfluidic glomerulus-on-a-chip device that simulates perfusion-driven flow. In the insert-well system, GECs were cultured on the upper side of the membrane and podocytes on the lower side, forming a glomerular bilayer structure (Fig. 3A). The formation of confluent cell layers on both surfaces of the membrane was confirmed by live/dead staining and confocal microscopy (fig. S6). To more closely mimic physiological conditions, a microfluidic chip was fabricated by sandwiching the same polycarbonate membrane between two microchannels to apply shear stress via controlled flow. Given that sEV size and surface properties likely influence barrier permeability, both systems were used to analyze the dynamics of sEV transcytosis through the glomerular barrier.

Fig. 3. Analysis of particles passing through a microphysiological glomerulus system.

(A) Schematic illustrations of the insert-well glomerular device (left) and the microfluidic glomerular device (right) designed to mimic the glomerular filtration barrier. (B) gRNA copy numbers in sEVs recovered after passage through the insert-well glomerular device. (C) Principal components analysis (PCA) of single-particle surface SERS spectra from GL261/dCas9/gRNA sEVs (blue), sEVs derived from the glomerular device alone (gray), and sEVs collected after passage through the device (pink). Postdevice EVs refer to the sEVs collected after passage through the glomerular device, which include both GL261/dCas9/gRNA sEVs and sEVs secreted by glomerular cells. Two distinct clusters were observed, indicating compositional differences in surface molecular features. Original spectra are shown in fig. S7. (D) gRNA copy numbers in sEV fractions after passage through the insert-well glomerular device. Blue bars indicate fractions with a density of 1.08 to 1.21 g/ml, consistent with the expected range for sEVs. (E) Size distribution of GL261/dCas9/gRNA sEVs (top) and sEVs collected after passage through the insert-well glomerular device (bottom), both isolated from fractions with a density of 1.08 to 1.21 g/ml. The postpassage sEVs consist of a mixture of GL261-derived sEVs and sEVs secreted by glomerular cells. (F) Permeability coefficients of calcein, Alexa Fluor 555–labeled albumin, carboxylated polystyrene beads (50-, 100-, and 200-nm diameters), and GL261/dCas9/gRNA sEVs across the insert-well glomerular device. (G) Relative permeability coefficients of the same set of molecules and particles in the insert-well and microfluidic glomerular devices. [(B), (F), and (G)] Data points represent results from independent experimental runs, with error bars indicating the SD (n = 3).

Using the insert-well glomerular device, we evaluated the behavior and structural characteristics of GL261/dCas9/gRNA sEVs during their passage through the glomerular filtration barrier. The sEVs were introduced into the upper chamber containing GECs, and sEVs collected from the lower chamber, beneath the podocytes, were analyzed to confirm successful traversal through the device (Fig. 3A). gRNA was detected in the collected sEVs, indicating that a portion of the GL261/dCas9/gRNA sEVs had passed through the GECs, membrane, and podocytes (Fig. 3B). To assess potential changes in surface molecular composition during passage, we measured surface-enhanced Raman scattering (SERS) spectra using a plasmonic nanopore sensor, which enables label-free, single-particle molecular analysis. This method revealed that the postdevice EVs, defined as sEVs collected after passage through the device, exhibited both conserved and altered surface features compared to the original GL261/dCas9/gRNA sEVs and glomerular device sEVs (i.e., sEVs secreted by glomerular cells alone). Among the 800 particles analyzed, which included both transcytosed sEVs and glomerular device sEVs, 495 particles exhibited surface molecular profiles distinct from both the original GL261/dCas9/gRNA sEVs and glomerular device sEVs (Fig. 3C and fig. S7).

To further interpret the SERS results, we examined the particle yield and gRNA content of sEV fractions obtained after passage through the glomerular device. Density gradient centrifugation showed that gRNA was retained in fractions with a density range of 1.08 to 1.21 g/ml, indicating that these fractions contained transcytosed GL261/dCas9/gRNA sEVs on the basis of established criteria (Fig. 3D) (36). These sEV fractions, collected after passage through the device and corresponding to the density range of 1.08 to 1.21 g/ml, exhibited a shift toward larger particle sizes and a broader size distribution (Fig. 3E). Their concentration (3.6 × 10⁸ particles/ml) was reduced compared to that of the original GL261/dCas9/gRNA sEVs (6.2 × 10⁸ particles/ml) recovered from the same density range.

We next considered whether the particle size shift observed in the postdevice sEV fractions (Fig. 3E, bottom) could be attributed solely to the presence of glomerular device sEVs. However, this interpretation was unlikely given that a substantial proportion of particles in this fraction also exhibited surface molecular profiles distinct from both the original GL261/dCas9/gRNA sEVs and glomerular device sEVs, as shown in the SERS analysis (Fig. 3C and fig. S7). These findings support the idea that transcytosed GL261/dCas9/gRNA sEVs underwent compositional remodeling during passage, potentially involving vesicle-vesicle interactions, changes in the composition of the protein corona on the sEV membrane, or partial membrane fusion with glomerular device sEVs. Such hybridization events could account for the concurrent increase in particle size and surface heterogeneity. This interpretation is further supported by our earlier observation of the mouse CD63 signal in human podocyte–derived sEVs, indicating that internalized vesicular components can be rereleased following molecular exchange. Together, these findings support the hypothesis that the glomerular transcytosis of sEVs involves not only selective transport but also partial structural and molecular remodeling.

To evaluate the permeation selectivity of the glomerular barrier, we quantified the permeability of various fluorescent molecules and particles using the insert-well glomerular device. Calcein (hydrodynamic diameter, ~1 nm), albumin-Alexa Fluor 555 (hydrodynamic diameter, ~10 nm), 50- or 100-nm-diameter carboxylated polystyrene beads, and GL261/dCas9/gRNA sEVs (mean size, 78 nm by nanoFCM) were added to the upper chamber. After 20 hours of incubation at 37°C, fluorescence intensity and qPCR were used to determine the permeability coefficients of each species in the lower chamber, as described previously (Fig. 3F and fig. S8) (37). Across all particle types, permeability was significantly reduced in the presence of glomerular cells, indicating formation of a size-restrictive barrier. The reduction was more pronounced for particles larger than albumin, which is typically retained in the bloodstream and only appears in urine under pathological conditions such as nephrotic syndrome. sEVs exhibited higher permeability than 50- and 100-nm beads, despite being similar in size, suggesting that sEVs may cross the barrier through an active, selective transcytosis pathway involving glomerular endothelial and epithelial cells.

To assess the influence of shear stress on sEV transcytosis, we developed a microfluidic glomerular device that simulates blood flow through the glomerular capillary interface. In vivo, physiological shear stress in renal capillaries has been reported in the range of ~0.3 to 1.2 dyne/cm² in healthy kidneys (38), with higher values potentially reflecting pathological states. This range is far lower than the maximum values cited in some literature, and it is not reproducible in our static insert-well system. To replicate this physiological condition, we fabricated a glomerulus-on-a-chip by sandwiching the polycarbonate membrane used in the insert-well device between two cyclo-olefin polymer microfluidic chips (figs. S9 and S10). When RPMI 1640 medium with a viscosity of 0.733 mPa·s (39) was perfused through the upper microchannel at a rate of 1.0 μl/min, the resulting shear stress at the membrane interface was calculated to be 0.69 dyne/cm². This shear stress closely approximates physiological conditions and has previously been reported to be effective for mimicking in vivo environments (40), thereby providing a controlled platform for evaluating transcytosis under dynamic flow.

To compare the effects of dynamic and static conditions on sEV permeability, we evaluated particle translocation in both microfluidic and insert-well glomerular devices. After perfusing 800 μl of medium into both upper and lower microchannels of the microfluidic chip, fluorescence intensity and qPCR were used to calculate the permeability coefficients of each particle type (Fig. 3G). Across all particle sizes, permeability was higher in the microfluidic device than in the insert-well system. While the presence of glomerular cells decreased the permeability of all particles in both devices, GL261/dCas9/gRNA sEVs showed a more pronounced increase in permeability under flow conditions compared to calcein, albumin, and both sizes of synthetic beads (fig. S11). Although shear stress is generally known to enhance barrier function via tight junction reinforcement (41), our results showed increased sEV permeability under flow conditions. This paradoxical result suggests that shear stress does not compromise paracellular integrity but rather selectively promotes transcellular transport of sEVs via transcytosis. Together, our findings demonstrate that combining microphysiological models and molecular tracers enables quantitative analysis of sEV excretion and highlights the role of glomerular transcytosis in determining sEV passage.

In vivo dual-reporter tracking of urinary excretion of cancer-derived sEVs

To complement our previous findings using GL261/dCas9/gRNA sEVs, which model the behavior of small RNA in liquid biopsy, we developed an independent reporter system to evaluate the urinary excretion of cancer-derived sEVs across different cancer types. We constructed a dual-reporter probe by fusing human CD9, an sEV-enriched membrane protein, with GeNL, a hybrid of the fluorescent protein mNeonGreen and the highly sensitive bioluminescent protein NanoLuc (Fig. 4A). Stable cell lines expressing CD9-GeNL were generated from A549 (lung cancer) and Panc-1 (pancreatic cancer) cells. Luminescence assays as well as Western blot analysis (fig. S12) confirmed that sEVs released from both cell lines were successfully labeled with GeNL. On the basis of our findings that proteinase K pretreatment minimizes background signals from free proteins and enhances signal specificity for intact sEVs (fig. S13), which is consistent with other previous studies as well (42–44), we applied this step consistently in all luminescence measurements throughout the present investigation. Further characterization by nanoFCM and cryo-TEM demonstrated that a high proportion of sEVs was GeNL-positive (A549, 71%; Panc-1, 85%) and that the labeled vesicles retained their normal size and morphology comparable to sEVs released from wild-type cells (fig. S14).

Fig. 4. Excretion of cancer-derived sEVs into urine tracked with the CD9-GeNL system.

(A) Schematic illustration of the CD9-GeNL tracking system. Cancer cells expressing CD9-GeNL, a fusion protein of mNeonGreen and NanoLuc, are transplanted into mice. sEVs released from these cells can be monitored via luminescence [using furimazine as a substrate; it should be noted that no normalization (e.g., by sample volume or internal protein biomarker) was done before the measurement] and fluorescence (excitation at 488 nm). (B) Luminescence-based estimation of sEVs in plasma and urine of mice transplanted with A549 CD9-GeNL cells, either orthotopically (lung) or subcutaneously. (C) Luminescence-based estimation of sEVs in plasma and urine of mice transplanted with Panc-1 CD9-GeNL cells, either orthotopically (pancreas) or subcutaneously. [(B) and (C)] Two independent mouse experiments were performed on different days using distinct sets of mice (Exp. #1: N = 3; Exp. #2: N = 5), and luminescence signals were recorded at 2 and 4 weeks posttransplantation. Raw luminescence values are provided in fig. S15. (D) Ratio of luminescence intensity in urine to plasma across all mice and time points. wks, weeks. Statistical significance was determined by a two-tailed one-sample t test against a null hypothesis ratio of 1. **P < 0.005 and ***P < 0.0005. Data points represent samples from individual mice in two independent experiments (Exp. #1 and Exp. #2), each including two time points (2 and 4 weeks), with a total of n = 14 to 16 after excluding mice that died before the final measurement. Error bars indicate SD. (E) NanoFCM analysis of sEVs purified from the urine of a control mouse (top, 4695 events) and a mouse orthotopically transplanted with A549 CD9-GeNL cells (bottom, 10291 events). Fluorescent sEVs were detected at 4 weeks posttransplantation using a sample with the highest luminescence signal.

To monitor the in vivo excretion of cancer-derived sEVs, we established mouse tumor models by injecting CD9-GeNL–expressing A549 and Panc-1 cells either subcutaneously or orthotopically (lung and pancreas, respectively). Tumor growth was assessed by both size and firefly luciferase activity, which was coexpressed with CD9-GeNL to enable noninvasive monitoring (fig. S15). Urine and plasma samples were collected periodically, and the luminescence intensity of GeNL was measured to quantify the sEV levels in each body fluid. In both A549 and Panc-1 models, GeNL signals increased over time in urine and plasma in parallel with tumor progression (Fig. 4, B and C, and fig. S16). These results confirm that cancer-derived sEVs circulate systemically and are excreted into urine, consistent with our findings using GL261/dCas9/gRNA sEVs. Moreover, the time-dependent increase in the GeNL signal demonstrates that tumor progression can be tracked by monitoring cancer-derived sEVs in both urine and plasma.

Notably, GeNL luminescence was consistently higher in urine than in plasma, suggesting that cancer-derived sEVs are more concentrated in urine (Fig. 4D and fig. S17). Although the urine concentration can vary considerably among individuals and even within the same individual depending on physiological conditions, this trend was observed quite consistently, indicating that the phenomenon is rather robust. (It should be noted that we did not normalize particle numbers in urine by any internal control or sample volume.) To verify that this luminescence originates from intact sEVs, we purified urinary sEVs using Tim4 (T cell membrane protein 4) beads, which bind phosphatidylserine on the sEV membrane, and analyzed them by nanoFCM. In a representative A549 orthotopic model sample with high luminescence, estimated to contain ~5.0 × 10⁷ particles/ml of cancer-derived sEVs on the basis of luminescence values, about 0.2% of the purified sEVs were found to be GeNL-positive (Fig. 4E). On the basis of this observation, we estimated the total urinary sEV concentration to be ~2.5 × 10¹⁰ particles/ml, which again falls within the previously reported physiological range of sEVs in mice (10⁸ to 10¹¹ particles/ml), supporting the reliability of our assay. Although we observed a roughly 40-fold discrepancy between the estimated urinary sEV particle numbers on the basis of GeNL luminescence and those observed in our gRNA-based in vivo experiment, we consider this reasonable given the differences in detection methods and the known inter- and intraindividual variabilities in urinary sEV concentrations, which can fluctuate dynamically depending on physiological conditions. In addition to these methodological and physiological sources of variability, other biological factors may also contribute to the discrepancy. These include differences in tumor location, labeling stoichiometry or packaging efficiency between reporters, differential stability or degradation of GeNL and gRNA in circulation (especially when the membrane integrity is compromised), or a difference in renal filtration dynamics depending on the sEV source (i.e., sEV characteristics would be different depending on the originating cells). We consider that a summary of these effects is observed as the discrepancy rather than the effect of any single factor. (A summary of these biological and methodological differences is provided in table S1.)

Together, our results suggest that urine may represent a richer reservoir for cancer-derived sEVs in the context of liquid biopsy because urine can be harvested in larger volumes than plasma. Moreover, considering that overall sEV concentrations are higher in plasma, but cancer-derived sEVs appear more enriched in urine, transcytosis via specific interactions with membrane molecules may underlie their selective urinary excretion. This enrichment tendency was consistently observed, supporting the notion that active urinary excretion occurs through systemic mechanisms. The urine-to-plasma ratio of the GeNL signal was generally higher in subcutaneous models than in orthotopic models (Fig. 4D and fig. S17), implying that the biodistribution and clearance kinetics of cancer-derived sEVs depend on tumor location.

Mechanistic insights and diagnostic implications of urinary excretion of tumor-derived sEVs

The present study provides direct experimental evidence that tumor-derived sEVs are excreted into urine following systemic circulation, a phenomenon that has long been hypothesized but not previously demonstrated with molecular specificity. By engineering two distinct reporter systems, CD63-dCas9/gRNA and CD9-GeNL, we tracked the biodistribution and urinary excretion of cancer-derived sEVs in mouse models representing brain, lung, and pancreatic tumors. Both reporter systems consistently detected tumor-derived sEVs in urine, confirming that these vesicles traverse the circulatory system and pass into the urinary tract. Notably, luminescence- and fluorescence-based measurements, in combination with nucleic acid quantification, enabled orthogonal validation of sEV origin and transport. These findings establish a mechanistic foundation for the presence of tumor-derived sEVs in urine and reinforce the utility of urine as a minimally invasive and robust source of cancer biomarkers.

Our data indicate that urinary excretion of tumor-derived sEVs is not driven by passive filtration through the glomerular barrier but rather involves active, selective, and possibly remodeling-mediated transcytosis. The in vitro study with GECs and podocytes demonstrated the endocytic uptake of tumor-derived sEVs and subsequent release of vesicles containing tumor-specific components. These observations suggest that a portion of the internalized sEVs escapes degradation and is recycled via exocytic pathways. Notably, transcytosed sEVs showed increased particle sizes and altered surface molecular profiles, consistent with surface/structural modification or fusion with glomerular cell–derived vesicles during intracellular processing. The emergence of such hybrid-like vesicles is further supported by single-particle SERS analysis, which revealed a substantial subset of sEVs that were molecularly distinct from both input and background vesicles. These findings imply that the glomerular filtration interface is not a static molecular sieve but a dynamic processing system capable of sorting and reshaping EVs before excretion. This revised view of glomerular function raises the possibility that transcytosis through glomerular cells may selectively enrich certain subpopulations of tumor-derived sEVs in urine.

Conventional approaches for detecting tumor-derived sEVs in urine have relied largely on bulk RNA profiling or immunoaffinity capture using surface markers, both of which are indirect and susceptible to background signals from urogenital tissues. In contrast, our dual-reporter strategy allows for direct and orthogonal tracking of tumor-origin vesicles using either nucleic acid (gRNA) or protein (GeNL) reporters, enabling quantification with high specificity. Both systems yielded consistent results across multiple tumor types and anatomical sites, including the brain, lung, and pancreas, underscoring the generalizability of the observed urinary excretion phenomenon. Our experimental design also permitted temporal monitoring of sEV dynamics, linking signal intensities in urine and plasma to tumor progression, which adds functional relevance beyond endpoint measurements. Together, these features establish our approach as a robust platform for dissecting sEV trafficking in vivo and for validating urinary sEVs as dynamic, disease-linked biomarkers.

Despite the strength of our dual-reporter systems and multimodel validation, several limitations should be acknowledged. First, the use of engineered sEVs bearing synthetic reporters such as gRNA or GeNL may not fully recapitulate the native behavior of unmodified tumor-derived sEVs in human patients. Future work should evaluate whether endogenous tumor markers, such as oncogenic microRNAs or membrane proteins, exhibit similar trafficking and urinary enrichment. Second, while our data suggest partial remodeling and possible hybridization of sEVs during glomerular transcytosis, direct visualization of fusion events or colocalization of tumor and host-derived markers within individual vesicles remains to be achieved. Advanced imaging or single-vesicle multiomics approaches may help resolve these questions. Third, although murine models offer valuable mechanistic insights, translation to human physiology will require further validation in clinical specimens or humanized models. In addition, while our glomerular-on-chip model successfully recapitulates the bidirectional sEV trafficking observed in vivo, it uses immortalized GECs and podocytes, which may differ from primary cells or organoids in glycocalyx composition, endocytic routing, and transcytosis kinetics—differences that could affect the selectivity, dynamics, and generalizability of sEV transport. Last, given the heterogeneity of sEV populations, identifying the molecular determinants that govern selective transcytosis into urine, such as lipid composition or surface glycosylation, represents an important direction for future research.

Here, we demonstrated that tumor-derived sEVs could be excreted into urine through systemic circulation and active transcytosis across the glomerular barrier. By using complementary reporter systems based on gRNA and GeNL, we directly tracked the dynamics of sEV trafficking in vivo and established that this excretion mechanism was consistent across multiple tumor types and anatomical sites. Mechanistic investigations revealed that glomerular cells not only internalized circulating sEVs via endocytosis but they also released modified or hybrid vesicles, suggesting a dynamic molecular remodeling process during transcytosis. These findings redefine the role of the glomerulus from a passive filtration interface to an active regulatory checkpoint in EV biology. Furthermore, the consistent enrichment of tumor-derived sEVs in urine, relative to plasma, highlights the diagnostic potential of urine as a minimally invasive and information-rich source for liquid biopsy. This work provides a foundation for future efforts to exploit urinary sEVs as real-time biomarkers for cancer detection and monitoring.

MATERIALS AND METHODS

Cell culture and stable cell line generation

GL261 cells (provided by A. Natsume) were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Thermo Fisher Scientific). A549 cells (RCB0098; provided by the RIKEN BRC through the National BioResource Project of the MEXT, Japan) were cultured in high-glucose DMEM (FUJIFILM Wako). Panc-1 cells (RCB2095; RIKEN BRC) were cultured in RPMI 1640 medium (FUJIFILM Wako). All media were supplemented with 10% fetal bovine serum (FBS; Thermo Fisher Scientific or Biosera) and 1% penicillin-streptomycin or penicillin-streptomycin-amphotericin B solution (Life Technologies or FUJIFILM Wako). Cells were maintained at 37°C in a humidified 5% CO₂ incubator.

Stable expression of CD63-dCas9 or CD9-GeNL was established using the Sleeping Beauty transposon system. Cells were seeded at 2.5 × 10⁵ cells/ml in 12-well plates to reach 70 to 80% confluency on the day of transfection. A total of 1 μg of plasmid DNA, including 50 ng of transposase-encoding plasmid (Addgene, no. 34879, pCMV(CAT)T7-SB100), was transfected using Polyethyleneimine “Max” (Polysciences, no. 24765) for GL261 cells or Lipofectamine 3000 (Invitrogen) for A549 and Panc-1 cells. Sixteen hours posttransfection, the culture medium was replaced, and antibiotic selection was initiated 48 hours later: GL261 CD63-dCas9 cells were selected with hygromycin (300 μg/ml); A549 and Panc-1 CD9-GeNL cells were selected with puromycin (0.5 and 1.0 μg/ml, respectively). The transfection scale was adjusted on the basis of the culture area.

To express the model tracer gRNA, lentivirus was produced by transfecting human embryonic kidney 293T cells (RCB2202; RIKEN BRC) with pMD2.G, psPAX2, and pRK389 (encoding tracer gRNA and blue fluorescent protein) at a ratio of 2:3:4 (ng). Transfection was performed in DMEM supplemented with 10% FBS. After 16 hours, the medium was refreshed, and the virus-containing supernatant was collected at 46 hours posttransfection, filtered (0.45 μm), and used to infect GL261 CD63-dCas9 cells. Selection with puromycin (0.5 μg/ml) was started 48 hours postinfection.

Fluorescence-activated cell sorting was done with the Aria IIIu cell sorter (BD) to get the enrichment of stably expressing cells. GL261/dCas9/gRNA cells were sorted on the basis of strong green fluorescent protein (CD63-dCas9) and blue fluorescent protein (gRNA) signals (top 20%). A549 and Panc-1 CD9-GeNL cells were sorted on the basis of high green fluorescence (GeNL, top 5%). Stable lines were maintained in selection medium: GL261/dCas9/gRNA in hygromycin (300 μg/ml) and puromycin (0.5 μg/ml); A549 and Panc-1 CD9-GeNL in puromycin (0.5 and 1.0 μg/ml, respectively).

Immortalized mouse GECs (Cell Biologics) were cultured in endothelial cell–specific medium supplemented with growth factors (Cell Biologics) and maintained up to passage 6 at 37°C in 5% CO₂. Immortalization was achieved by SV40 lentivirus (Applied Biological Materials) via polybrene-assisted transduction using a Lentiviral High Titer Packaging Mix (Takara Bio).

Mouse podocytes (SVI cell line, Cell Lines Service) were cultured in RPMI 1640 supplemented with 10% FBS and GlutaMAX-I (Thermo Fisher Scientific). Proliferation was maintained at 33°C, and differentiation was induced at 37°C for more than 2 weeks. Before sEV collection, the medium was replaced with advanced RPMI (Thermo Fisher Scientific). Human podocytes (provided by Y. Kobayashi) were cultured similarly in RPMI 1640 supplemented with 10% FBS and insulin-transferrin-selenium (ITS-G; FUJIFILM Wako), with temperature shifts and medium changes for differentiation and sEV collection as described above.

For the immunophenotyping of glomerular cells, cultured cells were harvested and stained for flow cytometry analysis. After discarding the culture medium, cells were washed with phosphate-buffered saline (PBS) and dissociated using trypsin (Cell Lines Service) for 1 to 2 min. The reaction was neutralized with culture medium, and gentle shaking facilitated cell detachment. The cells were pelleted by centrifugation at 300g for 5 min, resuspended in PBS, and washed once more under the same conditions. The final pellet was resuspended in ice-cold 1% bovine serum albumin (BSA) in PBS to achieve a concentration of 10⁶ to 10⁷ cells/ml. Aliquots of 100 μl were transferred into 1.5-ml tubes and incubated with 1 μl of primary antibody [anti-podocin, anti-nephrin, anti-CD2AP (all from Bioss Inc. and GeneTex), or CD31 (Thermo Fisher Scientific)] in the dark at 4°C for 90 min. After centrifugation at 300g for 5 min at 4°C, the supernatant was discarded, and the cells were washed twice with 500 μl of ice-cold 1% BSA in PBS. Cells were then incubated with 1 μl of goat anti-rabbit IgG (H + L) cross-adsorbed secondary antibody (Thermo Fisher Scientific) in 500 μl of 1% BSA in PBS for 90 min at 4°C in the dark. Following two additional wash steps, cells were resuspended in 200 μl of ice-cold 1% BSA in PBS and analyzed using a flow cytometer (Beckman Coulter). Fluorescence data were collected for downstream analysis.

Isolation of EVs from cultured cells

Cells were cultured in T75 plastic flasks (IWAKI) until they reached ~80% confluency, as confirmed by inverted phase-contrast microscopy. After aspirating the culture supernatant and washing the cells with PBS (pH 7.2, Thermo Fisher Scientific) when necessary, cells were incubated in serum-free medium for 48 hours at 37°C in a 5% CO₂ atmosphere. The culture supernatants were then collected. For EV collection, GL261/dCas9/gRNA cells were maintained in Advanced DMEM (Thermo Fisher Scientific) supplemented with 1% exosome-depleted FBS (Thermo Fisher Scientific) and 1% penicillin-streptomycin. A549 CD9-GeNL and Panc-1 CD9-GeNL cells were cultured in Opti-MEM (Gibco) supplemented with 1% penicillin-streptomycin.

Collected supernatants were centrifuged at 300g for 5 to 10 min at 4°C, followed by 2000g for 10 min at 4°C to remove dead cells and debris. The supernatant was then filtered through a 0.22-μm membrane (Merck Millipore) and subjected to ultracentrifugation. For GL261-derived sEVs, samples were ultracentrifuged at 110,000g for 80 min at 4°C using a CS150FNX micro ultracentrifuge with an S50A-2181 rotor (Hitachi). For A549 and Panc-1 cells, ultracentrifugation was performed at 120,000g for 120 min using an Optima XE-90 ultracentrifuge with an SW32Ti rotor (Beckman Coulter). The pellet was washed in PBS and reultracentrifuged under the same conditions. Final sEV pellets were resuspended in filtered PBS and stored at 4°C.

For density gradient separation, 4 ml of each sEV preparation was concentrated using Amicon Ultra filters (100-kDa molecular weight cutoff, 4-ml format; Merck). The filters were prewashed with 4 ml of 0.22-μm-filtered PBS and centrifuged at 10,000 rpm for 5 min at 4°C. Samples were then loaded and centrifuged under the same conditions, yielding ~2 ml of concentrate. OptiPrep gradient solutions (60% stock diluted to 2.5, 5, 10, and 20%) were layered sequentially (2.5 ml per layer), and 2 ml of the sample was carefully loaded on top. Gradients were centrifuged at 39,300 rpm for 160 min at 4°C. Twelve 1-ml fractions were collected from the top and stored for downstream analysis.

The concentrations of sEVs were assessed using an NTA (nanoparticle tracking analysis) instrument (Malvern Panalytical). Samples were diluted to ~10⁸ particles/ml. Five 60-s videos were acquired with a camera level of 15 and a detection threshold of 5. Data were analyzed using NanoSight NTA 3.2 software.

Cryo–transmission electron microscopy (cryo-TEM)

Cryo-TEM was performed to assess vesicle morphology. A 3-μl aliquot of sEV suspension was applied to a glow-discharged Quantifoil R1.2/1.3 200-mesh Cu grid (Quantifoil Micro Tools), blotted for 3 s, and plunge-frozen into liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). Imaging was performed under cryogenic conditions using a CRYO ARM 300 II transmission electron microscope (JEOL) operated at 300 kV, and images were recorded with a K3 direct electron detector (Gatan) at ×60,000 magnification using SerialEM software.

Single-particle characterization of sEVs used in in vitro experiments

Single-particle characterization of sEVs was performed using nanoFCM (Flow NanoAnalyzer, NanoFCM Co., Ltd.). For the immunophenotyping of sEVs from GL261 or GL261/dCas9/gRNA cells, sEVs collected from 50 ml of culture supernatant were resuspended in 50 μl of filtered PBS and incubated overnight at 4°C with FITC (fluorescein isothiocyanate)–conjugated anti-mouse CD63 (BioLegend, at 0.05 μg/μl). Following incubation, 450 μl of PBS was added, and sEVs were washed twice by ultracentrifugation at 110,000g for 80 min at 4°C. The final pellet was resuspended in 50 μl of PBS and subjected to nanoFCM analysis. For GeNL-labeled sEVs, purified sEVs were directly analyzed by nanoFCM after appropriate dilution. During sample acquisition, the laser power was set to 10 mW for both 488- and 638-nm lasers. Unless otherwise specified, gating was performed using the autothreshold function of the instrument software.

Western blotting of CD9-GeNL–expressing cells and EVs

Whole-cell lysates were prepared with radioimmunoprecipitation assay buffer (cat. no. 182-02451, FUJIFILM Wako). The lysate and sEV (isolated by ultracentrifugation from culture supernatants) were denatured with 4× Laemmli sample buffer (cat. no. 1610747, Bio-Rad) without reduction. Ten micrograms of whole-cell lysate or 0.5 to 1 μg of sEV was separated by SDS–polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes using the iBlot 3 Western blot Transfer System (cat. no. 34002, Invitrogen with iBlot-3, Thermo Fisher Scientific). The membranes were blocked with Blocking One (cat. no. 03953, Nacalai Tesque) for 30 min at room temperature and incubated with each primary antibody (anti-CD9, 1:1000, cat. no. MAI-80307, Thermo Fisher Scientific/anti-CD63, 1:1000, cat. no. SHI-EXO-M02, CosomoBio) at 4°C overnight, followed by incubation with a horseradish peroxidase–conjugated secondary antibody (cat. no. 7074 or no. 7076, Cell Signaling Technology) diluted at 1:2000. Protein signals were detected using an enhanced chemiluminescence substrate (cat. nos. RPN2232V1 and RPN2232V2, Cytiva) and iBright FL1500 Imaging Systems (Thermo Fisher Scientific). Antibodies were removed from the membrane using stripping buffer (cat. no. 7135A, Takara) for subsequent reprobing with another antibody.

RNA extraction and qPCR analysis

For RNA extraction from cells, culture supernatants were removed from cells grown in 24-well plates, and the cell surface was washed with PBS. A total of 200 μl of TRIzol reagent (Thermo Fisher Scientific) was added to each well, and the cells were incubated for 5 min at room temperature to allow complete lysis. The lysate was transferred to 1.5-ml tubes, and chloroform (FUJIFILM Wako) was added at one-fifth the volume of TRIzol. After gentle inversion and a 2-min incubation, the samples were centrifuged at 12,000g for 15 min at 4°C to separate phases. The aqueous phase was transferred to a fresh tube, mixed with an equal volume of isopropanol, and centrifuged at 12,000g for 15 min at 4°C to pelletize the RNA. The pellet was washed with 70% ice-cold ethanol and centrifuged at 7500g for 5 min at 4°C. After air-drying for 10 min, the RNA was resuspended in 10 μl of nuclease-free water (Thermo Fisher Scientific).

For RNA extraction from cell culture supernatants, the medium was first filtered through a 0.22-μm membrane (Merck Millipore) to remove cellular debris. A volume of 330 μl was collected, and RNase (Nippon Gene) was added to get a final concentration of 10 μg/ml. After incubation at 37°C for 10 min to degrade free RNA, TRIzol (4× volume) was added, and the mixture was incubated for 5 min at room temperature. Chloroform (one-fifth volume) was then added, followed by gentle inversion and centrifugation at 12,000g for 15 min at 4°C. The aqueous phase was recovered, mixed with 2 μl of GlycoBlue (Thermo Fisher Scientific) and an equal volume of isopropanol, and centrifuged at 12,000g for 15 min at 4°C. The RNA pellet was washed with 70% ice-cold ethanol, centrifuged at 7500g for 5 min at 4°C, air dried, and resuspended in 10 μl of nuclease-free water.

For urine samples, 330 μl of supernatant was obtained after centrifugation at 12,000g for 15 min at 4°C to remove debris such as fecal matter. RNase treatment and RNA extraction were performed using the same procedure as described for cell culture supernatants. RNA was resuspended in nuclease-free water.

For tissue samples, frozen lung and kidney specimens stored at −80°C were sectioned into ~100-mg pieces and homogenized in 2-ml tubes using a TissueRuptor system (QIAGEN). RNA was extracted using a commercial RNA purification kit (Thermo Fisher Scientific) following the manufacturer’s protocol and eluted in 30 μl of nuclease-free water.

qPCR was performed using the Luna Universal One-Step RT-qPCR Kit (New England Biolabs, Japan) and specific primers and probes listed in table S2. Amplification and detection were carried out using a QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific) with the following thermal cycling protocol: reverse transcription at 55°C for 10 min, initial denaturation at 95°C for 1 min, followed by 45 cycles of denaturation at 95°C for 10 s, annealing at 58°C for 10 s, and extension at 60°C for 30 s with real-time fluorescence acquisition.

Mouse models and sample collection

All animal experiments were approved by the University of Tokyo and Nagoya University and performed in accordance with institutional and national guidelines. To establish an orthotopic brain tumor model, C57BL/6 mice were anesthetized, and a burr hole was drilled into the skull. A total of 2 × 10⁵ GL261/dCas9/gRNA cells were stereotactically injected into the striatum of the cerebral cortex using a Hamilton syringe. Sham-operated control mice received an equal volume of PBS. Beginning on day 7 postimplantation, mice were transferred to metabolic cages for urine collection, which was performed every 3 days over nine time points until day 34. On the final day, whole brains were harvested and fixed in 4% paraformaldehyde overnight. The brains were sectioned through the injection site, and each sample was placed with the cut surface facing down on a coverslip. Fluorescence images of engrafted tumors, visualized via green fluorescent protein expression in GL261/dCas9/gRNA cells, were acquired using a confocal laser scanning microscope (TiE-A1R, Nikon Intech). Image processing was performed using Adobe Photoshop CS6 (Adobe Inc.).

For MRI imaging, a separate cohort of mice was used from those subjected to fluorescence observation. On the final day, intracranial tumor imaging was performed using a 3-T MRI system (MRS 3000; MR Solutions). Mice were anesthetized with 1 to 2% isoflurane delivered in air at a flow rate of 1.5 liters/min and positioned on an animal holder. The respiratory rate was monitored using a respiratory sensor connected to a gating system and maintained at 80 to 100 breaths per minute. For contrast-enhanced imaging, each mouse received an intraperitoneal injection of 0.05 mmol/kg body weight of gadolinium-diethylenetriamine penta-acetic acid (Magnevist, Bayer).

On the same day, kidneys and lungs were perfusion fixed with 4% paraformaldehyde and stored in PBS or cryopreserved in liquid nitrogen depending on the downstream analysis. Fixed tissues were embedded in paraffin using the CT-Pro20 system (Genostaff) with G-Nox (Genostaff) as a less toxic organic solvent substitute for xylene and sectioned at a 6-μm thickness. ISH was performed using the ISH Reagent Kit (Genostaff) according to the manufacturer’s instructions. Deparaffinized sections were fixed in 10% neutral buffered formalin for 30 min at 37°C, rinsed with distilled water, immersed in 0.2% HCl for 10 min at 37°C, and washed with PBS. Sections were then treated with proteinase K (FUJIFILM) in PBS for 10 min at 37°C at concentrations of 4 μg/ml for lung sections and 10 μg/ml for kidney sections, followed by PBS washing. Subsequently, the sections were heat treated in PBS for 10 min at 95°C, cooled immediately to room temperature in PBS, and transferred to a Coplin jar containing 1× G-Wash (Genostaff), equivalent to 1× SSC. The following antisense RNA probe was synthesized to detect the model gRNA sequence: gtttaagagctaagctggaaacagcatagcaagtttaaataaggctagtccgttatcaacttgaaaaagtggcaccgagtcggtgc.

Hybridization was carried out with the probe (250 ng/ml) in G-Hybo-L (Genostaff) for 16 hours at 40°C for lung sections and at 25°C for kidney sections. After hybridization, sections were washed three times with 50% formamide in 2× G-Wash for 30 min at 40°C (lung) or 25°C (kidney), followed by five washes in TBST [0.1% Tween 20 in tris-buffered saline (TBS)] at room temperature. Sections were then incubated in 1× G-Block (Genostaff) for 15 min at room temperature and subsequently with an anti-DIG-AP conjugate (Roche), diluted 1:2000 in G-Block (1:50 dilution in TBST), for 1 hour at room temperature. After two TBST washes, sections were incubated in a solution containing 100 mM NaCl, 50 mM MgCl₂, 0.1% Tween 20, and 100 mM tris-HCl (pH 9.5). Color development was performed using an NBT/BCIP (nitro blue tetrazolium–bromochloroindolyl phosphate) solution (Sigma-Aldrich), followed by PBS washing. Sections were counterstained with a Kernechtrot stain solution (Muto) and mounted sequentially with G-Mount (Genostaff) and Malinol (Muto). Images were acquired using the NanoZoomer S210 digital slide scanner (C13239-01; Hamamatsu Photonics) and visualized using NDP.view2 Plus software (U12388-02; Hamamatsu Photonics).

Analysis of glomerular uptake, release, and molecular remodeling of tumor-derived sEVs

Mouse GECs and SVI podocytes were seeded separately in 24-well plates. SVI cells were cultured at 33°C for 3 days to promote proliferation, followed by incubation at 37°C for at least 2 weeks to induce differentiation. GECs were cultured for more than 1 week to ensure maturation. After washing the cells with PBS, 500 μl of advanced RPMI 1640 medium containing GL261/dCas9/gRNA sEVs (1 μg/ml) was added to each well, and the cells were incubated at 37°C. The protein concentration of the sEVs was quantified using the Qubit Protein Assay Kit (cat. no. Q33211, Thermo Fisher Scientific, Inc.). On the basis of nanoFCM results, this concentration corresponded to about 1 × 10⁸ particles/ml (fig. S18). Following incubation, the sEV-containing medium was aspirated. In a subset of wells, cells were washed three times with PBS and subjected to RNA extraction to quantify intracellular gRNA by qPCR. In the remaining wells, cells were washed three times with PBS and incubated with fresh advanced RPMI 1640 medium for an additional 20 hours at 37°C. After this incubation, RNA was extracted separately from both the medium and cells, and gRNA levels were quantified by qPCR to evaluate vesicle release.

SERS was performed to analyze structural changes in individual sEVs following interaction with glomerular cells. SERS measurements were performed using a plasmonic nanopore sensor with a pyramidal structure fabricated by chemical wet etching of silicon substrates in potassium hydroxide (45). Gold was deposited on the nanopore surface via sputtering to generate a plasmonic enhancement zone at the apex. Before measurement, the nanopore was filled with a buffer solution, and sEVs were directed to the apex via electrophoresis, allowing single-vesicle capture. Raman spectra were recorded using a 785-nm laser at 10 mW with a 10-s integration time. After each measurement, the laser was turned off to release the single vesicle. This process was repeated to obtain spectra from 2400 individual sEVs.

Principal components analysis (PCA) was conducted on the obtained SERS spectra to assess molecular heterogeneity. Three types of samples were analyzed: (i) gRNA-containing sEVs, (ii) sEVs derived from glomerular cells, and (iii) a mixed population consisting of sEVs derived from glomerular cells and sEVs that had passed through the glomerular barrier. The variance-covariance matrix was calculated, and the first two principal components (PC1 and PC2) were defined on the basis of the largest eigenvalues. The mixed sEV sample exhibited two distinct clusters in the PC1-PC2 plot, corresponding to separate molecular profiles. One cluster overlapped with glomerulus-derived sEVs, while the other likely represented gRNA-containing sEVs that underwent membrane fusion or surface remodeling during transcytosis. The SERS spectra corresponding to the latter group suggested significant compositional alterations. These molecular signatures, while not yet fully understood in biological function, provide evidence of vesicle remodeling during glomerular processing.

Permeability analysis using glomerular filtration devices

To construct the insert-type glomerular device, a 24-well cell culture insert (Corning) was inverted, and 100 μl of SVI podocyte suspension (2 × 10⁴ cells/ml) was carefully applied to the underside of the membrane to prevent the suspension from dripping. The insert was incubated at 33°C in 5% CO₂ for 1 hour to allow cell attachment. It was then placed upright into a 24-well plate, with 750 μl of medium added to the lower compartment and 250 μl to the upper compartment. Cells were cultured at 33°C for 1 day and then at 37°C for 1 week to induce differentiation. After aspirating the upper medium and washing with PBS, 250 μl of GEC suspension (5 × 10⁴ cells/ml) was seeded into the upper compartment. The device cells were cultured for at least one additional week at 37°C before use. For sEV permeability assays, 330 μl of GL261/dCas9/gRNA sEVs in a suspension (1 μg/ml) was introduced into the upper chamber, and after 20 hours of incubation at 37°C, 990 μl of medium was collected from the lower chamber for analysis.

To construct the microfluidic glomerular device, a six-well cell culture insert was inverted, and 1000 μl of SVI cell suspension (2 × 10⁴ cells/ml) was applied to the underside of the membrane. After 2 hours of incubation at 33°C in 5% CO₂, the insert was placed into a six-well plate with 2 ml of medium in the lower chamber and 1.5 ml in the upper chamber. Cells were cultured at 33°C for 2 days and then at 37°C for 1 week. Following PBS washing, 1.5 ml of GEC suspension (5 × 10⁴ cells/ml) was seeded into the upper compartment and cultured for at least 1 week. The membrane was then cut into rectangles of ~5 × 20 mm and sandwiched between the microfluidic channels to complete the device assembly. The microchannel dimensions were 12 mm in length, 3 mm in width, and 188 μm in height. PEEK tubing was connected to the inlets and outlets of the microchannels to allow perfusion.

To evaluate the permeability of fluorescent tracers through the insert-type device, both compartments were filled with serum-free medium. The upper chamber received a solution containing calcein (1 μM, Thermo Fisher Scientific), Alexa Fluor 555–conjugated albumin (1 μM, Thermo Fisher Scientific), carboxylated fluorescent polystyrene beads (50-, 100-, or 200-nm diameters; 2.5 × 10⁻⁴% w/v; Techo Chemical Corp.), or gRNA-containing sEVs (1 μg/ml). The medium-filled device was incubated at 37°C for 20 hours in 5% CO₂. After incubation, the medium from the lower chamber was collected, and the concentrations of permeated tracers were measured using a fluorescence plate reader and qPCR.

The permeability coefficient P was calculated using the following equation (37)

$$P=\frac{∆C_{\mathrm{L}}{V}_{\mathrm{L}}}{C_{\mathrm{U}}A∆t}$$ (1)

where C_U is the initial concentration in the upper compartment, ΔC_L is the concentration change in the lower compartment, V_L is the volume of the lower compartment, A is the membrane surface area, and Δt is the incubation time.

For the microfluidic glomerular device, PEEK tubing was connected to the inlets and outlets of both upper and lower channels. A syringe pump delivered the same test solution used in the insert-type assay into the upper channel at a flow rate of 1 μl/min. The lower channel was perfused with the advanced RPMI. After perfusing 800 μl through both upper and lower channels, effluents were collected and analyzed by fluorescence and qPCR. Shear stress within the microfluidic device was calculated using the following equation (46)

$$\mathrm{\tau }=\frac{6uQ}{w{h}^2}$$ (2)

where τ is the shear stress (dyne/cm²), μ is the dynamic viscosity (g/cm·s), Q is the volumetric flow rate (cm³/s), h is the channel height (cm), and w is the channel width (cm).

Bioluminescence analysis of GeNL-tagged sEVs

To evaluate the GeNL signals in the intact sEVs, purified CD9-GeNL–tagged sEVs in PBS were treated with proteinase K (0.1 mg/ml; Thermo Fisher Scientific, EO0491) at 37°C for up to 2 hours. As a control for membrane-compromised sEVs, a subset of samples was lysed with 0.1% Triton X-100 before proteinase K treatment. After incubation, 4 μl of the treated sEV solution was transferred into a U-bottom white 384-well plate (Corning, 4513), and an equal volume (4 μl) of Nano-Glo luciferase assay reagent (Promega) was added. Luminescence was measured immediately after reagent addition using a plate reader.

For in vivo monitoring, BALB/cAJcl-nu/nu mice (female, 7 weeks old) were implanted with A549 CD9-GeNL or Panc-1 CD9-GeNL cells either subcutaneously (5 × 10⁶ cells per mouse) or orthotopically (1 × 10⁶ cells per mouse; into the lung or pancreas). Tumor growth in the subcutaneous model was assessed by a caliper-based volume calculation using the following formula: volume = (long diameter) × (short diameter)²/2. In the orthotopic model, tumor engraftment was confirmed by bioluminescence imaging using the IVIS Lumina II system at 10 min after an intravenous injection of 100 μl of d-luciferin potassium (30 mg/ml) in PBS.

To collect body fluids, blood (~100 μl) was obtained from the mouse tail vein using a heparinized capillary and transferred into a heparin-pretreated tube. Plasma was isolated via standard centrifugation. For urine collection, mice were placed individually in metabolic cages (KN-645, Natsume Seisakusho Co., Ltd.) overnight under fasting conditions with access to water only. Plasma and urine samples were centrifuged at 3000g for 15 min, and the supernatants were diluted fivefold with PBS. [This dilution factor was predetermined through spike-in experiments to ensure that the luminescence signal would not be affected by the physicochemical properties of urine (e.g., pH).] These diluted samples were treated with proteinase K (0.1 mg/ml) and subjected to luminescence measurement following the same procedure as described above. The luminescence intensity was converted into particle concentration by referencing a standard curve generated from purified sEVs with known concentrations determined by nanoparticle tracking analysis (NanoSight). We note that the luminescence value/particle concentration was not normalized by sample volume or internal biomarkers.

Fluorescence analysis of GeNL-tagged sEVs in urine

Urine was collected as described in the “Bioluminescence analysis of GeNL-tagged sEVs” section. The urinary sEVs were purified using the MagCapture Exosome Isolation Kit PS (FUJIFILM Wako), which uses Tim4 protein–coated magnetic beads to capture phosphatidylserine-expressing vesicles, according to the manufacturer’s instructions. This kit-based method was selected for urine samples because of their limited volume and complexity, and it allowed for the efficient recovery of vesicles while minimizing sample loss. Details of the protocol are as follows, and a schematic of this procedure is presented in fig. S19.

Sample preparation

Urine was brought to a final volume of 500 μl with TBS (Nippon Gene). Exosome binding buffer (×500) was added, followed by centrifuging at 1200g for 20 min at 4°C and then at 10,000g for 30 min at 4°C; the final supernatant was collected for capture

Bead preparation

Biotin capture magnetic beads (kit-supplied) were thoroughly resuspended, and a 60-μl aliquot was dispensed into a 1.5-ml tube. Beads were equilibrated with 500 μl of exosome capture immobilizing/washing buffer (I/W; 1×), vortexed, briefly centrifuged, and placed on a magnetic stand for 1 min before the supernatant was removed. The I/W buffer (1×; 500 μl) and biotin-labeled exosome capture solution (10 μl) were then added, mixed by vortexing, and incubated on a rotator per the bead manufacturer’s instructions. After a brief spin, beads were placed on the magnetic stand again for 1 min before the supernatant was discarded. Beads were washed twice with 500 μl of I/W buffer (1×) and kept at 4°C until use.

sEV purification

Functionalized beads were combined with the precleared supernatant containing a binding enhancer (×500), incubated on the rotator (1 hour at room temperature), and then placed on the magnetic stand for 1 min to remove the supernatant. Beads were washed twice with I/W buffer supplemented with a binding enhancer (×500), each time magnetizing for 1 min before supernatant removal. sEVs were eluted with exosome elution buffer in two rounds of 50 μl each: 1-min hold at room temperature, magnetization for 1 min, and collection of the supernatant. The combined eluate was applied to nanoFCM after appropriate dilution to achieve the manufacturer-recommended event rate. During sample acquisition, the laser power was set to 10/50 mW for the 488-nm laser. Gating was performed using the autothreshold function of the instrument software, and it was confirmed that more than 99.99% of the sEVs of the urine of a negative control sample (sEVs purified from a mouse without tumor implantation) were in the nonfluorescent fraction.

Statistical analysis

Statistical analyses were performed using GraphPad Prism [version 10.4.2 (633); GraphPad Software] and Microsoft Excel. Data are presented as the means ± standard deviation (SD), unless otherwise indicated. For comparisons between two groups, a two-tailed unpaired Student’s t test was used. For multiple group comparisons, the one-way ANOVA followed by Tukey’s post hoc test was applied. In all tests, a P value <0.05 was considered statistically significant. The number of biological replicates (n) and specific statistical tests applied are described in the corresponding figure legends.

Animal experiment

The animal experiments were carried out according to protocols approved by the Experimental Animal Ethics Committees of National Center for Geriatrics and Gerontology and the University of Tokyo, Japan (approval numbers: Dou-5-41 and A2023M053-01, respectively).

Supplementary Materials

This PDF file includes:

Figs. S1 to S19

Tables S1 to S3