Utility of Urinary Extracellular Vesicles for Colorectal Cancer: Comparison of DNA Quality and MRD Detection

ABSTRACT

This study evaluated the feasibility of urinary extracellular vesicles (EVs) as a liquid biopsy source for colorectal cancer (CRC) using two predefined cohorts: 97 patients for analytical validation and 63 patients for postoperative survival analysis. We compared urinary EVs, urinary cell‐free tumor DNA (ctDNA), plasma EVs, and plasma ctDNA across multiple molecular dimensions, and assessed the utility of urinary EV‐DNA for minimal residual disease (MRD) detection. Urinary EVs exhibited larger particle size (median 175.6 nm) and markedly higher purity, demonstrated by a substantially broader proteomic profile (4674 vs. 476 proteins) and significantly higher expression of EV‐specific markers (CD63, CD81, CD9). DNA extracted from urinary EVs showed the highest concentration and best preservation of nucleic acid integrity, with the greatest Long/Short fragment ratio, outperforming all plasma‐derived samples. Mutation detection sensitivity was comparable across sample types but was highest in plasma ctDNA (39.2%) and urinary EV‐DNA (34.0%), exceeding urinary ctDNA (21.6%). In the survival cohort, MRD positivity—defined as detectable RAS/BRAF mutations in urinary EV‐DNA 1 month after curative surgery—was strongly associated with inferior outcomes (HR for recurrence 5.25, 95% CI 1.74–15.80), and overall survival was significantly worse in MRD‐positive patients. These findings indicate that urinary EVs provide highly pure vesicles with superior DNA quality, making them a robust and completely non‐invasive source for molecular profiling. Urinary EV‐DNA–based MRD assessment shows significant prognostic value and may serve as a practical tool for postoperative monitoring and risk stratification in colorectal cancer.

Urinary EVs provide highly pure vesicles with superior DNA quality, making them a robust and completely non‐invasive source for molecular profiling. Urinary EV‐DNA–based MRD assessment shows significant prognostic value and may serve as a practical tool for postoperative monitoring and risk stratification in colorectal cancer.

Abbreviations

CRC

colorectal cancer

ctDNA

cell‐free tumor DNA

EVs

extracellular vesicles

IRB

Institutional Review Board

MRD

minimal residual disease

NTA

nanoparticle tracking analysis

PBS

phosphate‐buffered saline

1. Introduction

Colorectal cancer (CRC) is one of the most common malignancies worldwide, and reliable biomarkers for early detection and postoperative monitoring remain essential for improving clinical outcomes. Liquid biopsy using plasma cell‐free tumor DNA (ctDNA) has been well established for assessing tumor burden, treatment response, and minimal residual disease (MRD) [1, 2]. However, several alternative biofluids and analytes—including urine [3], extracellular vesicles (EVs) [4], and circulating tumor cells [5]—have recently gained attention as complementary sources of tumor‐derived molecular information.

EVs, particularly small EVs, encapsulate DNA, RNA, and proteins within a lipid bilayer that protects their cargo from enzymatic degradation, offering a more stable analyte than free‐circulating nucleic acids [6, 7, 8, 9]. Importantly, EVs provide a more stable platform for biomarker preservation compared to free‐circulating nucleic acids, as their membranous structure protects the internal contents from enzymatic degradation [10]. This stability makes EVs particularly attractive for clinical applications, including cancer diagnostics and prognostics [11, 12, 13, 14, 15]. Studies have shown that tumor‐derived EVs carry cancer‐specific molecular signatures, making them promising biomarkers for various malignancies, including CRC. However, most EV‐based liquid biopsy studies have focused on blood‐derived samples, with limited investigation of urinary EVs. Urine represents an attractive alternative biofluid for liquid biopsy due to its completely non‐invasive collection, large available volumes, and ease of serial sampling. While urinary sEVs have shown promise in bladder and urothelial cancer [16, 17], their utility in CRC remains largely unexplored, and direct comparisons with plasma‐based approaches are lacking.

The potential of urinary EVs for MRD detection is of particular interest, as postoperative monitoring is crucial for identifying patients at high risk of recurrence who may benefit from adjuvant therapy [18, 19]. However, it remains unclear whether urinary EV–based MRD assessment is useful for predicting prognosis in colorectal cancer. This study aimed to characterize urinary EVs in comparison with plasma‐derived vesicles and ctDNA, and to determine the clinical utility of urinary EV‐DNA for MRD detection following curative surgery in colorectal cancer.

2. Material and Methods

2.1. Patients and Outcome

This retrospective study enrolled colorectal cancer patients who underwent surgical resection between April 2023 and December 2024. Inclusion criteria were: histologically confirmed colorectal cancer; age 20–80 years; preserved major organ function; presence of RAS or BRAF mutation in tumor tissue; and availability of paired blood and urine samples stored in our biobank. Cases with missing data were excluded.

For the quality‐validation cohort, we included patients with tissue‐confirmed RAS/BRAF mutations and compared DNA obtained from urinary EVs, urinary ctDNA, plasma EVs, and plasma ctDNA to assess the analytical performance of urinary EV‐DNA. For the survival‐analysis cohort, we included patients with RAS/BRAF mutations detected in urinary EV‐DNA before surgery and evaluated its prognostic utility for survival risk stratification.

2.2. Ethics and Sample Collection

This study utilized tissue, blood, and urine samples from our institutional biobank. The biobank prospectively collects and stores resected tumor specimens and blood samples from patients treated at our facility following acquisition of written informed consent. Blood and urine samples were collected preoperatively, at 1 month after surgery, and every 3 months thereafter. The biobank was approved by the Institutional Review Board (IRB; approval no. 28‐03‐738). The present study using biobank specimens received separate IRB approval (approval no. M‐2024‐239), and an opt‐out consent process was employed. This study was conducted in accordance with the Declaration of Helsinki.

Samples collected within 1 month immediately prior to surgery were used. The survival‐analysis cohort included patients who underwent preoperative treatment; in these patients, samples collected after completion of preoperative treatment and within 1 month before surgery were analyzed. In addition, for the survival‐analysis cohort, postoperative urine samples collected approximately 1 month after surgery (between 2 and 6 weeks postoperatively) were also analyzed. These samples were picked up within 2 h of collection. Blood samples (10 mL in BD Vacutainer EDTA tube: Becton Dickinson) were transferred in BD Vacutainer EDTA tubes (Becton Dickinson, Oxford, UK) and processed within 2 h. Plasma was separated by centrifuging the blood at 1900 g for 10 min at 4°C and then stored at −80°C until DNA extraction or EV isolation. Urine samples were also stored at −80°C until analysis. Urine samples were centrifuged at 1900 g for 10 min at 4°C and aspirated 4 mL of supernatant. The supernatant was mixed with 400 μL of preservation solution (Streck Urine Preserve; Streck, La Vista, NE, USA) and stored at −80°C.

2.3. EV Isolation

Plasma EVs were isolated from the plasma using the Total Exosome Isolation Kit (from plasma) (Invitrogen, California, USA), according to the manufacturer's instructions. Briefly, the plasma sample was centrifuged at 2000 × g for 20 min at room temperature to remove cells and debris. The supernatant was transferred to the new tube and centrifuged at 10,000 × g for 20 min at room temperature to remove debris. The sample was mixed with phosphate‐buffered saline (PBS) and the isolation reagent. And then the mixture incubated at room temperature for 10 min. After the mixture was centrifuged at 10,000 × g for 5 min, the pellet was dissolved in 200 PBS to obtain the EVs.

Urinary EVs were isolated from the urine using the Total Exosome Isolation Kit (from Urine) (Invitrogen, California, USA), according to the manufacturer's instructions. Briefly, the urine sample was mixed with the isolation reagent. And then the mixture incubated at room temperature for 60 min. After the mixture was centrifuged at 10,000 × g for 60 min, the pellet was dissolved in 200 μL PBS to obtain the EVs.

2.4. EV Analysis

First, the isolated EVs were examined using transmission electron microscopy (TEM). Samples were directly adsorbed onto glow‐discharged (60 seg low discharging using PELCO easy‐glow device) carbon‐coated copper grid (300 mesh). Afterwards, grids were fixed with 2% paraformaldehyde (PFA) in phosphate buffer (PB 0.2 M pH 7.4) for 20 min and washed with distilled water. Then, the contrast staining was made by incubating the grids with 4% uranyl acetate (UA) at 4°C for 15 min. TEM images were obtained by using TECNAI G2 20 C‐TWIN high‐resolution transmission electron microscope, at an acceleration voltage of 200 kV.

Membrane proteins of isolated sEVs (CD63, CD81 and CD9) were analyzed by Western blotting. Samples (40 μL) were suspended in RIPA buffer containing protease inhibitor cocktail at −4°C for 15 min. After adding loading buffer and heat denaturation at 75°C for 10 min, equal samples (10 μL) were separated on 4%–20% SDS‐PAGE gradient gels (Bio‐Rad Laboratories Inc.). After transfer, membranes were blocked with blocking buffer (Bio‐Rad Laboratories Inc.) and probed overnight at 4°C with primary antibodies [anti‐CD63 (sc‐5275, Santa Cruz), anti‐CD81 (sc‐166,029, Santa Cruz), and anti‐CD9 (MCA469GT, Bio‐Rad)] at 1:200 dilution. Following three washes with TBS‐0.1% Tween (5 min each), membranes were incubated with HRP‐conjugated secondary antibody (Goat Anti‐Mouse IgG, Bio‐Rad) at 1:5000 dilution for 1 h at room temperature. Protein bands were detected using ECL substrate (Bio‐Rad) with ImageQuant LAS 4000 (Cytiva, Tokyo, Japan).

Size distribution and concentration of the isolated EVs were evaluated by nanoparticle tracking analysis (NTA) using a NanoSight NS300 System (Malvern, UK) further equipped with fast video capture and particle‐tracking software. NTA post‐acquisition settings were kept constants for all samples. Each video was analyzed for obtaining the mean and mode vesicle size as well as particle concentration.

2.5. DNA Extraction

EV‐DNA was extracted using QIAamp DNA Mini Kit (Qiagen, Limburg, Netherlands) according to the manufacturer's protocol. Briefly, 200 μL PBS, 40 μL QIAGEN Protease, and 4 μL RNase (Qiagen, 19,101) were mixed and transferred to the column membrane. After adding 400 μL AL buffer and incubating at 56°C for 10 min, the column was centrifuged at 3200 g for 10 min to collect the lysate. Following the addition of 400 μL ethanol, the lysate was applied to the QIAmp Mini spin column and centrifuged at 6000 g for 1 min. The column was washed with 500 μL AW1 and AW2 buffers sequentially.

ctDNA was extracted from 800 μL plasma and urine using a QIAamp Circulating Nucleic Acid kit (Qiagen, Limburg, Netherlands), in accordance with the manufacturer's protocol, with minor modifications, including washing the column twice with 30 mL elution buffer and waiting 5 min before processing the cfDNA. ctDNA concentrations were measured using the Qubit quantification assay (Thermo Fisher Scientific, Tokyo, Japan).

2.6. DNA Analysis

We compared the fragment lengths of EV‐DNA and ctDNA in both plasma and urine samples using the Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA). Since a higher proportion of long DNA fragments indicates less fragmentation and better DNA quality, we calculated the Long/Short ratio (the amount of long DNA [> 300, < 600 bp] divided by the amount of short DNA [> 50, < 180 bp]) as an indicator of DNA quality, which is widely used as a measure of DNA integrity [20, 21].

Tissue RAS/BRAF status was evaluated using the RASKET assay (LSI Medience Corporation, Tokyo, Japan). Genetic alterations in ctDNA and EV‐DNA were examined by ddPCR using the QX200 Droplet Digital PCR system (Bio‐Rad, Hercules, CA, USA). For RAS mutations, we used ddPCR KRAS Screening Multiplex Kit (detecting G12 and G13D mutations), ddPCR KRAS Q61 Screening Kit (detecting Q61 mutations), ddPCR NRAS G12 Screening Kit, and ddPCR Mutation Assay for KRAS p.A146T (Bio‐Rad). For BRAF mutations, ddPCR BRAF V600 Screening Kit (detecting V600E/K/R) was used (Bio‐Rad). Variant allele frequencies of 0.1% or higher were considered positive.

2.7. Survival Analysis

We further assessed MRD by analyzing urinary EV‐DNA 1 month after surgery in patients who had undergone R0 resection and had mutations detected in preoperative urine. Recurrence‐free survival (RFS) and overall survival (OS) were compared between the MRD‐positive and MRD‐negative groups.

2.8. EV Proteomic Analysis

EV proteins were lysed in SDS/TEAB buffer, and total protein was quantified using the Micro BCA assay. Up to 50 μg of protein was processed using S‐Trap micro columns with standard reduction, alkylation, washing, and on‐column digestion using Trypsin/Lys‐C. Eluted peptides were dried and stored at −80°C.

Proteomic profiling was performed by LC–MS/MS using an EASY‐nLC 1200 system coupled to an Orbitrap Exploris 480. A 0.1‐μg peptide input was analyzed using a 70‐min gradient. DIA was acquired with FAIMS, and data were processed with DIA‐NN (v1.8) using an in silico spectral library. Protein abundances were quantified with MaxLFQ, applying a 1% FDR at both peptide and protein levels.

2.9. Statistical Analysis

All statistical analyses were performed with R Commander and EZR, which is a modified version of R Commander. Fisher's exact tests were performed to compare frequencies between groups. DNA concentrations and Long/Short ratio among the four groups (urine‐derived EVs, urinary ctDNA, plasma‐derived ctDNA, and plasma‐derived EVs) were compared using the Friedman test, a non‐parametric test for repeated measures or matched samples. For post hoc pairwise comparisons, the Wilcoxon signed‐rank test was employed with Bonferroni correction to adjust for multiple comparisons. Survival outcomes were compared between MRD positive and negative patients based on urinary EV analysis using the Kaplan–Meier method. Differences in survival distributions were assessed using both the log‐rank test and the Wilcoxon (Breslow) test. All statistical tests were two‐sided, and a p‐value < 0.05 was considered statistically significant.

3. Results

3.1. Patients

During the study period, 504 patients underwent surgery for colorectal cancer. After excluding those with RAS wild‐type (N = 195) or without RAS testing (N = 90), 219 patients remained eligible for screening. Among these, cases without adequately preserved paired blood and urine samples (N = 117) and those with minor RAS mutations (N = 5) were further excluded. Ultimately, 97 patients were included in the cohort for evaluating the quality of urinary extracellular vesicles (Quality‐validation cohort).

Of the 504 patients, 380 underwent R0 resection. Of those, both preoperative and one‐month postoperative urine samples were available in 204 patients. Among them, 80 had both preoperative and one‐month postoperative urine samples available and RAS or BRAF mutation was identified in urinal EV‐DNA before surgery. After excluding 17 patients with missing data, 63 were finally included in the survival‐analysis cohort. Notably, 24 patients (38.1%) exhibited molecular heterogeneity, defined as the absence of mutations in tumor tissue despite their presence in preoperative urinary EV‐DNA. The patients flow diagram is shown in Figure 1. The baseline characteristics of the quality‐validation and survival‐analysis cohorts are summarized in Tables 1 and 2, respectively. The survival‐analysis cohort included a total of five patients with stage IV or recurrent disease, comprising four MRD‐positive cases and one MRD‐negative case. The surgical procedures performed in these five patients were as follows: sigmoid colectomy with peritoneal metastasectomy, peritoneal metastasectomy with para‐aortic lymph node dissection, total pelvic exenteration, low anterior resection with peritoneal metastasectomy, and right hemicolectomy with partial hepatectomy.

FIGURE 1.

Patient inclusion flowchart for quality‐validation and survival‐analysis cohorts. From 504 colorectal cancer patients who underwent surgery, 380 patients with R0 resection were identified. After excluding patients with RAS wild‐type (N = 195), insufficient RAS testing (N = 90), RAS minor mutations (N = 5), and insufficient samples (N = 117), 97 patients were included in the quality‐validation cohort. For survival analysis, 204 patients with both preoperative and 1‐month postoperative urine samples were evaluated. Among these, 80 patients showed RAS or BRAF mutations in preoperative urinary EV‐DNA, and after excluding 17 patients with insufficient data, 63 patients comprised the final survival‐analysis cohort. A total of 21 patients overlapped between the quality validation cohort and the survival analysis cohort. These 21 cases specifically refer to those with R0 resection and confirmed RAS/BRAF mutations.

TABLE 1.

Patients' characteristics of quality‐validation cohort.

	N = 97
Age	67.4 ± 12.6
Sex (M/F)	49/48
Tumor location (right/left/rectum)	30/19/48
Pathological stage (0/I/II/III/IV and rec)	1/11/38/29/18
Preoperative chemotherapy (%)	29 (30.0)
Mutation in the tissue (KRAS/NRAS/BRAF)	81/4/12

Note: Values are presented as mean ± standard deviation, number of patients, or number (percentage).

Abbreviations: F, female; Left, left sided colon; M, male; rec, recurrence; Right, right sided colon.

TABLE 2.

Patients' background of survival‐analysis cohort.

	Positive group (N = 20)	Negative group (N = 43)	p
Age	69.0 ± 14.3	67.9 ± 13.7	0.79
Sex (male/female)	10/10	14/30	0.18
Tumor location (right/left/rectum)	6/8/6	10/14/19	0.66
Pathological stage (I/II/III/IV and rec)	1/8/7/4	16/10/16/1	0.04
Preoperative chemotherapy (%)	6 (30.0)	10 (23.3)	0.55
Preoperative heterogeneity in urinary EV (%)	10 (50.0)	14 (32.6)	0.18
Preoperative urine EV‐DNA Concentration (ng/mL)	232.6 ± 342.6	235.8 ± 418.7	0.97
Postoperative urinary EV‐DNA Concentration (ng/mL)	345.0 ± 650.6	215.1 ± 571.9	0.45
Adjuvant chemotherapy yes	5 (25.0)	12 (27.9)	1.00

Note: Values are presented as mean ± standard deviation, number of patients, or number (percentage).

Abbreviations: Left, left sided colon; rec, recurrence; Right, right sided colon.

3.2. Quality of Extracted EVs

The successful isolation of urinary EVs was confirmed using electron microscopy (Figure 2A) and Western blot analysis detecting EV‐enriched tetraspanins (Figure 2B). According to nanoparticle tracking analysis, urinary EVs had a significantly larger particle size than plasma EVs, with median diameters of 175.6 nm and 98.9 nm, respectively (p < 0.001, Figure 3A). The concentration of particles was significantly lower in urinary EV than in plasma EV (7.3 particles/pL vs. 98.9 particles/pL, p < 0.001, Figure 3B), suggesting that serum EVs were more frequently contaminated with smaller particles, such as proteins.

FIGURE 2.

Characterization of extracellular vesicles from plasma and urine. (A) Representative transmission electron microscopy images showing the morphology of plasma‐derived and urinary EVs. Both samples display characteristic cup‐shaped or spherical vesicular structures. Scale bars = 200 nm. (B) Western blot analysis confirming the presence of EV‐enriched tetraspanin markers (CD63, CD9, CD81) in both urinary and plasma EVs.

FIGURE 3.

Nanoparticle tracking analysis of urinary and plasma extracellular vesicles. (A) Particle size distribution comparison between urinary and plasma EVs. Urinary EVs demonstrated significantly larger median particle sizes (175.6 nm, IQR: 160.5–193.4 nm) compared to plasma EVs (98.9 nm, IQR: 46.7–251.0 nm) (p < 0.001). (B) Particle concentration analysis showing significantly lower EV concentrations in urine (median: 7.3 particles/pL, IQR: 5.1–6.9) compared to plasma (median: 98.9 particles/pL, IQR: 84.3–120.2) (p < 0.001). Box plots display median, interquartile range, and outliers (×).

A total of 4674 proteins were identified on average in urinary EVs, compared to only 476 in plasma EVs (Figure S1). Major proteins such as albumin and immunoglobulin showed no difference between the two groups (albumin; p = 0.32, immunoglobulin; p = 0.80, Figure 4A,B). On the other hand, EV‐specific membrane proteins such as CD63, CD9, and CD81 were detected at significantly higher levels in urinary EVs than in plasma EVs (CD63; p < 0.01, CD81; p < 0.01, CD9; p = 0.02, Figure 4C–E).

FIGURE 4.

Comparative proteomic analysis of urinary and plasma extracellular vesicles. Paired analysis of protein abundance in matched urinary (red) and plasma (blue) EV samples from the same patients. MaxLFQ values represent label‐free quantification of protein abundance. (A) Albumin showed no significant difference between urinary and plasma EVs (ns). (B) IgG levels were comparable between the two EV sources (ns). (C) CD63, a classical EV marker, was significantly more abundant in urinary EVs compared to plasma EVs (p < 0.01). (D) CD9, another tetraspanin marker, showed significantly higher abundance in urinary EVs (p < 0.01). (E) CD81 expression was significantly elevated in urinary EVs (p < 0.05). Connected lines represent paired samples from individual patients, demonstrating consistent patterns of EV marker enrichment in urine‐derived samples.

3.3. Quality of DNA

The Friedman test revealed significant differences in DNA concentrations among the four groups (χ ² = 76.98, p < 0.001, Figure 5A). Post hoc analysis with Bonferroni correction revealed that plasma‐derived EVs had significantly lower DNA concentrations than all other sample types (all p < 0.001).

FIGURE 5.

Comparative analysis of DNA quality and quantity across different liquid biopsy sources. (A) DNA concentration comparison among urinary EV‐DNA (median: 70.9 ng/mL, IQR: 39.3–192.0), urinary cfDNA (median: 97.3 ng/mL, IQR: 46.7–251.0), plasma EV‐DNA (median: 41.9 ng/mL, IQR: 30.0–61.2), and plasma cfDNA (median: 153.0 ng/mL, IQR: 101.0–239.0). Urinary EV‐DNA and cfDNA showed significantly higher concentrations than plasma EV‐DNA (both p < 0.001). Plasma cfDNA showed significantly higher concentrations than plasma EV‐DNA (p < 0.001). (B) Long/short fragment ratio analysis reflecting DNA fragmentation patterns. Urinary EV‐DNA exhibited the highest median ratio (0.97, IQR: 0.31–2.35), significantly higher than plasma EV‐DNA (0.31, IQR: 0.10–0.46, p = 0.002) and plasma cfDNA (0.35, IQR: 0.21–0.42, p = 0.001). Urinary cfDNA (median: 0.59, IQR: 0.24–0.97) also showed significantly higher ratios than both plasma sources (p = 0.04 for each comparison). Higher ratios indicate better preservation of longer DNA fragments within EV membranes. Box plots show median, interquartile range, and outliers (×). Statistical significance was determined using Friedman test with Bonferroni correction.

The Long/Short ratios showed significant differences among the four groups (Friedman test, χ ² = 16.63, p = 0.0008, Figure 5B). Post hoc analysis with Bonferroni correction revealed that the Long/Short ratios in urinary EVs were significantly higher than those in plasma EVs (p = 0.001) and plasma ctDNA (p = 0.002). Similarly, urinary ctDNA showed significantly higher Long/Short ratios compared to plasma EVs (p = 0.04) and plasma ctDNA (p = 0.04). No significant differences were observed between urinary EVs and urinary cfDNA (p = 0.27) or between plasma EVs and plasma ctDNA (p > 0.99).

3.4. Detection of KRAS, NRAS and BRAF Mutations

RAS or BRAF mutations were detected in 33 (34.0%) patients using urinary EVs, 21 (21.6%) using urinary ctDNA, 29 (29.9%) using plasma EVs, and 38 (39.2%) using plasma ctDNA in the Quality‐validation cohort. No significant differences were observed among the four groups (p = 0.06).

3.5. MRD Analysis

The median follow‐up period was 544 days (IQR: 363–692 days, range: 157–1010 days), calculated using the reverse Kaplan–Meier method. Among the 63 patients in the survival‐analysis cohort, RAS/BRAF mutations were identified in urinary EVs at 1 month postoperatively in 20 patients, who were classified as MRD‐positive. MRD‐positive group included a higher proportion of patients with more advanced disease (Table 2). Among the 20 patients who were MRD‐positive, recurrence was observed in 9 cases. The sites of recurrence included the liver in 6 patients, the lung in 2 patients, peritoneal dissemination in 3 patients, and lymph nodes in 4 patients, with overlapping among sites. Both PFS (Hazard ratio 5.25, 95% confidence interval: 1.74–15.80, Log rank: p = 0.001, Wilcoxon: p < 0.001 Figure 6A) and OS (Log rank: p < 0.001, Wilcoxon: p < 0.001, Figure 6B) were significantly worse in the MRD‐positive group. The median overall survival was 728 days in the MRD‐positive group, while it was not reached in the MRD‐negative group, with no death events observed during the follow‐up period. The hazard ratio was not estimable due to complete separation.

FIGURE 6.

Kaplan–Meier survival analysis based on minimal residual disease status in postoperative urinary EV‐DNA. MRD status was determined by detection of RAS/BRAF mutations in urinary EV‐DNA collected 1 month after curative resection. Survival is shown in days from surgery. Number of patients at risk and censored events are indicated by tick marks on the curves. (A) Progression‐free survival for all patients. Kaplan–Meier curves comparing MRD‐positive (black line) and MRD‐negative (gray line) patients. MRD‐positive patients showed significantly worse progression‐free survival (hazard ratio = 5.25, 95% CI: 1.74–15.80; log‐rank test: p = 0.001; Wilcoxon test: p < 0.001). (B) Overall survival for all patients. MRD‐positive patients demonstrated significantly worse overall survival compared to MRD‐negative patients (log‐rank test: p < 0.001; Wilcoxon test: p < 0.001). No deaths occurred in the MRD‐negative group; therefore, the hazard ratio for overall survival could not be estimated due to complete separation. (C) Progression‐free survival for Stage I‐III patients. Among Stage I‐III patients, MRD‐positive status was associated with significantly worse progression‐free survival (hazard ratio = 5.47, 95% CI: 1.51–19.76; log‐rank test: p = 0.004; Wilcoxon test: p = 0.003). (D) Overall survival for Stage I‐III patients. Among Stage I‐III patients, MRD‐positive status was associated with significantly worse overall survival (log‐rank test: p = 0.01; Wilcoxon test: p = 0.01). No deaths occurred in the MRD‐negative group; therefore, the hazard ratio for overall survival could not be estimated due to complete separation.

We also analyzed the data after excluding patients with stage IV disease and those with recurrent disease. Both PFS (Hazard ratio 5.47, 95% confidence interval: 1.51–19.76, Log rank: p = 0.004, Wilcoxon: p = 0.003, Figure 6C) and OS (Log rank: p = 0.01, Wilcoxon: p = 0.01, Figure 6D) remained significantly worse in the MRD‐positive group.

4. Discussion

To date, evidence for urinary EVs in colorectal cancer has been limited, with most prior studies focusing on bladder and urothelial malignancies [16, 17]. In this context, our study addresses this gap through systematic comparison of four liquid biopsy modalities—urinary EV‐DNA, urinary cfDNA, plasma EV‐DNA, and plasma cfDNA—within the same patient cohort. We report three principal findings. First, urinary EVs exhibited superior purity compared to plasma‐derived EVs, as evidenced by enrichment of EV‐specific markers and markedly broader proteomic profiles. Second, DNA extracted from urinary EVs demonstrated the highest quality among all sample types, with optimal concentration and fragment integrity. Third, postoperative detection of RAS/BRAF mutations in urinary EV‐DNA proved to be a robust prognostic biomarker, with MRD‐positive patients showing significantly inferior survival outcomes.

Our nanoparticle tracking analysis revealed fundamental differences between urinary and plasma EVs. Urinary EVs exhibited significantly larger particle size and lower particle concentration compared to plasma EVs. These findings suggest that plasma samples contain abundant smaller particles, likely representing non‐EV contaminants such as protein aggregates and lipoproteins. In contrast, the larger size distribution and lower concentration in urine indicate reduced contamination with such particles. This interpretation was strongly supported by our proteomic analysis, which identified approximately 4674 proteins in urinary EVs compared to only 476 in plasma EVs. Notably, while contaminating proteins such as albumin and immunoglobulin showed no significant difference between the two sources, EV‐specific membrane proteins (CD63, CD81, CD9) were significantly enriched in urinary EVs. These results collectively demonstrate that urine provides EVs of higher purity than blood, which is consistent with previous reports suggesting that the lower protein content of urine facilitates cleaner EV isolation [7, 22, 23].

The second major finding concerns DNA quality. Bioanalyzer analysis showed that urinary EVs consistently yielded the highest Long/Short ratio among all sample types, indicating superior preservation of longer DNA fragments that are more suitable for molecular analyses (20). Both urinary EVs and urinary ctDNA exhibited significantly higher ratios than their plasma counterparts, suggesting that urine provides a more favorable environment for DNA preservation, likely due to reduced nuclease activity and the protective EV membrane [10, 24]. Urinary EVs also provided the highest DNA concentrations, whereas plasma EVs yielded the lowest. Although plasma contains more EV‐like particles overall, their lower purity likely results in fewer genuine vesicles carrying DNA cargo. From a practical standpoint, these features—higher DNA concentration and better fragment integrity—make urinary EVs an advantageous analyte for liquid biopsy, enabling improved analytical sensitivity and more comprehensive molecular profiling, including mutation detection, copy‐number assessment, and potentially epigenomic analysis.

The third and most clinically meaningful finding is the prognostic value of MRD detection using urinary EV‐DNA. In our cohort, patients with detectable RAS/BRAF mutations in urinary EVs 1 month after surgery had significantly worse RFS (HR 5.25, 95% CI 1.74–15.80, p = 0.001) and OS (p < 0.001). These results parallel prior evidence showing that postoperative ctDNA detection predicts recurrence in colorectal cancer, as demonstrated by Tie et al. [18] and further validated in the prospective study by Nakamura et al. [19]. Our results extend these observations to urinary EVs, suggesting that this non‐invasive sample source can serve as an effective alternative to plasma for MRD assessment. From a clinical standpoint, urine offers multiple practical advantages: complete non‐invasiveness, elimination of clinical visits for sampling, feasibility of frequent home‐based collection, and high patient acceptability. These characteristics may improve monitoring compliance and enable earlier detection of recurrence, particularly when frequent serial sampling is required. Notably, MRD‐positive cases may be associated with a shorter time to recurrence. Nakamura et al. [19] reported a median time to recurrence of 5.34 months in MRD‐positive patients, and in the present study, the median time to recurrence was also short, at 4.46 months.

It should be noted that the rates of adjuvant chemotherapy were comparable between the MRD‐positive and MRD‐negative groups because postoperative treatment decisions were made without knowledge of MRD status. Therefore, the observed differences in outcomes are unlikely to be explained by differences in adjuvant treatment intensity. Although the MRD‐positive group included a higher proportion of patients with advanced pathological stage and residual confounding by stage cannot be completely excluded, the persistence of significantly worse outcomes in MRD‐positive patients despite similar use of adjuvant chemotherapy suggests that conventional postoperative therapy may be insufficient to eradicate residual molecular disease and that MRD status may provide prognostic information beyond conventional staging systems. In contrast, previous studies have suggested that among MRD‐negative patients, postoperative outcomes may not differ substantially according to the administration of adjuvant chemotherapy [19, 25], implying that MRD negativity identifies a biologically low‐risk population with limited absolute benefit from standard adjuvant treatment. Together, these findings support the clinical utility of urinary EV‐DNA–based MRD assessment for tailoring postoperative management.

Emerging evidence indicates that EV‐DNA originates not only from apoptotic processes but also from viable tumor cells through active secretion [11, 26]. Tumor cells frequently exhibit chromosomal instability and micronuclei formation, and rupture of these micronuclei releases nuclear DNA fragments into the cytoplasm, which can subsequently be packaged into EVs [13, 27]. As a result, EV‐DNA often contains longer double‐stranded fragments that better preserve genomic architecture, including copy‐number changes and methylation patterns [27, 28]. The lipid bilayer further protects these fragments from nuclease‐mediated degradation, providing higher DNA integrity than plasma ctDNA [27]. These biological characteristics suggest that EV‐DNA may reflect ongoing activity of residual viable tumor cells rather than passive release from dying cells [29]. Consequently, urinary EV‐DNA could offer enhanced sensitivity for detecting minimal residual disease, particularly in the early postoperative period when tumor burden is extremely low.

Beyond the three principal findings, this study provides additional insights into the expanding field of liquid biopsy in colorectal cancer. Although plasma ctDNA has been the primary focus of previous work, our comparative analysis demonstrates that different biofluids and molecular carriers can offer complementary information. Mutation detection sensitivities were similar across sample types, indicating that each modality may be appropriate depending on clinical context and practical considerations. Notably, molecular heterogeneity—defined as mutations detected in urinary EV‐DNA but absent in tumor tissue—was observed in 38% of patients. This finding suggests that liquid biopsy can capture tumor‐derived alterations not represented in single‐site tissue sampling, emphasizing the biological complexity underlying MRD assessment.

Our study has several limitations that warrant consideration when interpreting the findings. First, we used a precipitation‐based EV isolation method, which is feasible for clinical implementation but may co‐isolate non‐vesicular materials. This could influence the absolute DNA yield and may partially account for the observed purity differences between urinary and plasma EVs. Second, the molecular analysis focused solely on RAS and BRAF mutations. While clinically relevant, this limited panel does not capture the full genomic landscape of colorectal cancer, and therefore the prognostic performance of urinary EV‐DNA–based MRD may be underestimated. Third, urine is inherently variable in concentration and composition depending on hydration status, renal function, and timing of collection. Such variability may affect DNA yield and could limit the reproducibility of urinary EV‐DNA measurements across different clinical settings. Fourth, although the detection of oncogenic RAS or BRAF mutations in urinary EV‐DNA strongly suggests a tumor‐derived origin, the cellular origin of urinary EV‐DNA could not be directly demonstrated. Fifth, the retrospective single‐center design and modest sample size restrict generalizability. The magnitude of the hazard ratios observed for MRD may therefore not fully translate to larger, more diverse populations. Finally, the MRD‐positive group tended to include patients with more advanced disease than the MRD‐negative group; thus, the observed prognostic differences between the two groups may, at least in part, be attributable to differences in disease stage. Prospective multicenter studies with standardized EV isolation protocols are needed to confirm the robustness of these findings.

In summary, this study demonstrates that urinary EVs represent a feasible, informative, and clinically relevant liquid biopsy platform for colorectal cancer. The superior purity of urinary EVs compared to plasma EVs, combined with their higher DNA concentration and better fragment integrity, establishes urine as an attractive alternative biofluid source. Most importantly, the strong prognostic value of MRD detection using urinary EV‐DNA suggests potential clinical utility for postoperative monitoring and risk stratification. These findings support further investigation of urinary EVs through prospective, multicenter studies with comprehensive molecular profiling to fully realize their potential in precision oncology and postoperative surveillance.

Author Contributions

Shintaro Kanaka: conceptualization, methodology, investigation, data curation, writing – original draft, writing – review and editing, formal analysis, project administration. Takeshi Yamada: conceptualization, writing – review and editing, investigation, methodology, supervision, project administration. Fumi Asai: data curation, investigation, formal analysis, writing – review and editing. Yuta Shimizu: conceptualization, supervision, data curation, investigation, writing – review and editing, visualization. Kay Uehara: conceptualization, data curation. Akihisa Matsuda: formal analysis, data curation. Seiichi Shinji: data curation. Yasuyuki Yokoyama: data curation. Goro Takahashi: data curation. Takuma Iwai: data curation. Kohki Takeda: data curation. Sho Kuriyama: data curation. Toshimitsu Miyasaka: data curation, investigation, formal analysis. Koki Hayashi: investigation, formal analysis, data curation. Hiroshi Yoshida: conceptualization, writing – review and editing, supervision.

Funding

The authors have nothing to report.

Ethics Statement

This retrospective study was conducted in accordance with the Declaration of Helsinki. We included patients who participated in a bio‐bank project at our institution. This project was approved by local ethics review boards (28‐03‐738 and M‐2024‐239) and written informed consent was obtained from all patients who participated in this project.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Supplied as separate files. It shows the types of proteins detected from EVs isolated from urine and plasma in a bar graph.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.