Extracellular vesicles expressing CEACAM proteins in the urine of bladder cancer patients

Abstract Early detection and long‐term monitoring are important for urothelial carcinoma of the bladder (UCB). Urine cytology and existing markers have insufficient diagnostic performance. Here, we examined medium‐sized extracellular vesicles (EVs) in urine to identify specific markers for UCB and evaluated their usefulness as diagnostic material. To identify specific markers in urinary EVs derived from UCB, we undertook shotgun proteomics using urine from four UCB patients and four healthy subjects. Next, 29 healthy specimens, 18 noncancer specimens, and 33 UCB specimens, all from men, were analyzed for urinary EVs by flow cytometry to evaluate the diagnostic performance of UCB‐specific EVs. Nanoparticle‐tracking analysis indicated that the size of EVs extracted from urine was mostly <400 nm. By shotgun proteomics, we detected several proteins characteristic of UCB and found that carcinoembryonic antigen‐related adhesion molecule (CEACAM) proteins were increased in patients. Flow cytometric analysis revealed that the degree of expression of CEACAM1, CEACAM5, and CEACAM6 proteins on the surface of EVs varied among patients. Extracellular vesicles expressing CEACAM proteins also expressed mucin 1, suggesting that they were derived from tumorigenic uroepithelial cells. The number of EVs expressing CEACAM1, 5, and 6 proteins was significantly increased in UCB (mean ± SD, 8.6 ± 13%) compared to non‐UCB (0.69 ± 0.46) and healthy (0.46 ± 0.34) by flow cytometry. The results of receiver operating characteristic (ROC) analysis showed a good score of area under the ROC curve of 0.907. We identified EVs that specifically express CEACAM proteins in urine and have potential for diagnostic applications. These EVs are potential targets in a new liquid biopsy test for UCB patients.


| INTRODUC TI ON
Urothelial carcinoma of the bladder, the fourth most common cancer in men, results in significant morbidity and mortality. 1 It is ranked 6th or 7th in estimated cancer incidence in the United States to 2040. 2 At the time of initial diagnosis, approximately 70% of patients have cancer confined to the epithelium or subepithelial connective tissue.
These cancers are usually treated with endoscopic resection and selective intravesical therapy. The recurrence rate of superficial disease exceeds 60%, and <30% of recurrent bladder tumors progress to invasive disease. 3 Thus, UCB requires continuous cystoscopy and cytology because of the high recurrence rate of the disease. 4,5 These tests are invasive and expensive, and there are differences among users and among the facilities that carry them out.
Urine-based biomarkers for the detection of UCB seem to be an attractive option. However, the FDA-approved test will not replace the current diagnostic criteria of urine cytology and cystoscopy, and more sensitive and specific biomarkers are needed. 6 There have been several reports of possible protein UCB markers in urine, including CEACAM1. Tilki et al. reported that urinary CEACAM1 levels discriminate between UCB patients and non-UCB subjects.
Furthermore, urinary CEACAM1 levels increase with advancing stage and grade. 7 Extracellular vesicles play essential roles in cell-cell communication and are diagnostically significant materials. [8][9][10][11] Extracellular vesicles are membrane vesicles that most cells release into the surrounding extracellular environment and can be divided into subgroups such as apoptotic bodies, microvesicles, and exosomes. It is difficult to standardize methods for isolating subgroups of EVs and procedures for purifying mixtures of vesicle types. Therefore, the International Society for Extracellular Vesicles has recently recommended the phrase "extracellular vesicle" as a generic term for isolated and studied vesicles when authors are unable to establish a specific marker. 12 Thus far, we have focused on medium-to largesized (100-1000 nm) EVs in blood and urine to characterize EVs in healthy individuals. 13 Here, we found that urine from healthy individuals contains a large number of EVs expressing MUC1 on their surface, which could be derived from the tubular or uroepithelial surface of normal tissues. 13 We have also characterized EVs that are presumably derived from renal tubules that are positive for CD10, CD13, and CD26 (multipeptidase+ EVs). 13 Although it remains difficult to use EVs as a diagnostic tool for UCB, a large body of evidence has accumulated demonstrating their potential as a biomarker for the noninvasive diagnosis of UCB. [14][15][16] Welton et al. analyzed urinary proteins in EVs and found that several proteins were elevated in UCB patients, including MUC1. 17 In this study, we identified medium-sized uroepithelium-derived EVs in the urine of healthy subjects. We speculated that if cancerous uroepithelial cells were present, EVs distinguishable from those of healthy subjects might also be present in the urine of UCB patients.
To characterize UCB EVs, we undertook a proteomic analysis of EV fractions extracted from UCB patients and searched for proteins that could serve as specific surface antigens.

| Patient specimens
Thirty-three human urine samples from 31 UCB patients and 18 human urine samples from 18 noncancerous patients that were obtained at Kyushu University Hospital between April 2019 and April 2022 were analyzed in this study. The patients' clinical information was obtained from their medical records. All patients provided written informed consent before undergoing the study procedures. Urine samples were also obtained from 29 healthy volunteers. The clinical protocol for this study was approved by the appropriate institutional review boards and ethics committees at Kyushu University Hospital. This study was carried out in accordance with the Declaration of Helsinki.

| Isolation of urinary mEVs
For the isolation of urinary EVs, we modified a urinary exosome extraction protocol. 18 We used a reduction process to degrade THP polymers and a centrifugation process to focus on mediumsized EVs with diameters ranging from 100 to 1000 nm. 13 In a flow cytometric analysis, the volume of urine used for each donor was approximately 10 ml (0.8 ml urine was used per assay, but approximately 10 ml was needed to combine single-stain assays for setting patient-specific compensation in multiple staining). In NTA, the volume of urine used per assay was 10 ml. Collected urine was centrifuged twice at 2330 g for 10 min. Once cells and large debris were removed, the urine supernatant was cryopreserved at −80°C. In the next protocol, the supernatant was thawed in a 37°C water bath and centrifuged at 2330 g to obtain the supernatant. The supernatant was centrifuged at 18,900 g for 30 min in a fixed-angle rotor. The EV pellet obtained from centrifugation was reconstituted by vortex mixing (1-2 min) with 0.2 ml DPBS followed by incubation with DTT (final concentration 10 mg/ml) at 37°C for 10-15 min. The samples were centrifuged again at 18,900 g for 30 min and the supernatant was discarded. Degraded THP monomers were removed from EVs after centrifugation. The DTT-containing DPBS solutions were filtered through 0.1μm filters (Millipore Sigma).

| Nanoparticle tracking analysis
Nanoparticle tracking analysis measurements were carried out using a NanoSight NS300. All samples were diluted in PBS to a final volume of 0.8 ml. Concentrations were determined by pretesting the ideal particle per frame value (20-100 particles/frame). Extracellular vesicles labeled with ExoGlow-NTA Dye (System Biosciences), which binds specifically to intact EV membranes, were also measured by fluorescence NTA. [19][20][21] The settings of the device were in accordance with the manufacturer's software manual (NanoSight NS300 User Manual, MAN0541-01-EN-00, 2017). Particles in the laser beam underwent Brownian motion and videos of these particle movements were recorded. NTA 3.2 software was then used to analyze the video and determine the particle concentration and the size distribution of the particles.
Twenty-five frames per second were recorded for each sample with a "number of frames" setting of 1498. The detection threshold was 5 for both scattered light and fluorescence NTA. The detailed measurement parameters are shown in Supplementary Materials and Methods.

| Flow cytometric analysis of urinary EVs
After resuspending EV pellets in 60 μl DPBS, we added saturating concentrations of labeled Abs, annexin V, and normal mouse IgG, and incubated the tubes in the dark without stirring for 15-30 min at room temperature. Various Ab concentrations at the time of staining are listed in Supplemental Materials and Methods. We diluted this Ab-stained solution into 250 μl annexin V binding buffer (10 mM HEPES, 0.14 mM NaCl, 2.5 mM CaCl 2 , pH 7.4; BD Biosciences), which was then subjected to flow cytometry analysis. The DPBS and annexin V binding buffer were filtered through 0.1μm filters (Millipore). Flow cytometry was carried out using a FACSVerse flow cytometer (BD Biosciences). In our previous report, we implemented a method to measure particles smaller than 1 μm in size using polystyrene beads for verification and aggregating them into an observed image using side scatter. 13 The flow cytometer was equipped with 405 nm, 488 nm, and 638 nm lasers to detect up to 13 fluorescent parameters. The flow rate was 12 μl/min. Forward scatter voltage was set to 381, side scatter voltage was set to 340, and each threshold was set to 200. Details of excitation and emission wavelengths as well as voltages are described in the Supplementary Materials and Methods. Flow cytometry was carried out using FACSuite software (BD Biosciences) and data were analyzed using FlowJo software.  were to 1, the better the model; a model is considered good if R2Y is 0.65 or higher and Q2Y is 0.5 or higher. 22,23 Here, the proteins with variable importance in the projection (VIP) value greater than 1.0 were set for differential proteins. The VIP values larger than 1.0 point to the most relevant variables. 24,25 Gene enrichment analysis was carried out using Metascape.

| Statistical analysis
GraphPad Prism 9.2.0 (GraphPad Software) was used for the statistical analysis. Relationships between groups were compared using the Mann-Whitney U-test or Dunn's multiple comparison test (Kruskal-Wallis analyses were carried out). p < 0.05 was considered statistically significant.

| Isolation of urinary mEVs and particle size distribution
To comprehensively recover all types of EVs without depending on the specificity of the membrane surface, we used a classical method of EV extraction by centrifugation (precipitated fraction by centrifugation at 18,900 g). 26,27 The precipitation fraction contains components with similar sedimentation coefficients and densities, except for EV. We treated the precipitated fraction with DTT to degrade and remove THP polymers that interacted with IgG. 13 The extracted EV fractions were subjected to NTA using a Nanosight NS300. 28 Extracellular vesicles labeled with ExoGlow-NTA Dye, which specifically binds to the intact EV membrane, were also analyzed by fluorescence NTA. 21 Histograms of the particle size distributions of fractions extracted from healthy subjects and UCB patients, as measured by scattered light and fluorescence NTA, are shown in Figure 1A-D. The diameters corresponding to 10%, 50%, and 90% of the total number of particles observed by scattered light NTA in eight healthy subjects and eight UCB patients were compared ( Figure 1E). In this enrichment operation, fractions with a diameter of 1 μm or more were rarely included, and the main fraction was distributed in the 200-300 nm diameter range. In addition, some of these fractions contained cell membranes (10%-60%), which were considered to be "mEVs" according to their size. The concentration in urine was calculated from the particle concentration detected in NTA. There were no differences in the particle size distribution or concentration of the extracted EV fractions between healthy subjects and UCB patients.

| Shotgun proteomic analysis of urinary EVs
Proteomic analysis using LC-MS/MS was undertaken to identify proteins specifically present in EVs in the urine of UCB patients. 29 Urine from four UCB patients (patient 1: Tis, urine cytology class III; patient 2: T2, urine cytology class V; patient 3: T2, urine cytology class V; patient 4: Ta, urine cytology class II) and four healthy subjects were used to enrich the EV fraction in urine. Two fractions were prepared: one was a mixed pool of samples from four patients and four healthy individuals, with the equivalent amount per patient after protein quantification for each group, and the other was an EV fraction extracted from each patient and one healthy individual. To gain an overview of the proteins contained in the urinary EVs of UCB patients, we first undertook a shotgun analysis of the pooled samples. The proteins detected in each of the two groups were selected from the Proteome Discoverer search results with a score of 1.0 or higher. These proteins are shown in a Venn diagram in Figure 2A and are listed in the Data S1. Enrichment analysis using Metascape was carried out on the proteins detected in UCB patients only (585 proteins) ( Table 1).
Although there were some functional categories related to cell adhesion, which is a characteristic of cancer, the most frequently categorized functions of proteins included eukaryotic translation elongation and ribonucleoprotein complex assembly RNA.
Ribosome-related functional categories (protein complex, translation) were calculated. This could be one feature of the proteins contained in EVs derived from cancer patients.
The analysis was undertaken using individually extracted EVs.
The obtained MS/MS data were subjected to LFQ analysis using the MaxQuant platform, and the quantitative results for each patient were calculated. 30,31 In the analysis of extracted EV fractions from four healthy subjects and four UCB patients, 1957 different proteins were detected (Data S1). MetaboAnalyst 5.0 was used for the statistical analysis. Principal component analysis was undertaken from the results of each of the four UCB patients and four healthy subjects ( Figure 2B). Although Patient 2 had a slightly different profile to the other three patients, there were some differences in the detected protein profiles between the groups of four UCB patients and four healthy individuals. We carried out OPLS-DA analysis, which allows discriminant analysis between groups. In the score plot shown in Figure S1, R2X = 0.253, R2Y = 0.847, and Q2Y = 0.608. Because R2Y and Q2Y were good, the patient group and the healthy group were significantly discriminated. Next, screening was carried out using and VIP values and S-plot to identify key markers that differentiate UCB from healthy subjects. Those with a large contribution to the discrimination of the two groups and an increase in UCB were extracted from the s-plot, and those with a VIP value exceeding 1.0 were selected (Data S1). A heat map of the 64 proteins extracted by this analysis was used to separate the healthy group from the patient group by clustering ( Figure 2C). Of the 64 selected proteins, 31 proteins, or approximately half, were categorized as both "plasma membrane" and "extracellular exosome" in Gene Ontology (AGRN,   ANXA2, APOA1, APOB, C3, C9, CEACAM5, CFB, CP, F2, FAM129B,   FGB, FGG, FN1, IGHA1, IGHG1, IGHG2, IGHM, IGKC, MARCKS,   PLG, PROS1, PSCA, RAB27B, SDCBP2, SERPINC1, SLC2A1, SRC, TACSTD2, TF, and UPK3A). Proteins that have been reported to be upregulated in UCB (UPK3A, PSCA, ANXA2, and ANXA9) were included, [32][33][34] indicating the validity of the analysis.
We found several proteins that are increased in UCB, but we needed to select among these for diagnostic applications, that is, membrane proteins that can detect EVs from the outside. Proteins that contribute to the grouping of patients and healthy subjects include CEACAM5 and CEACAM7, and we wondered whether these  Comparison of the particle size distributions at the 10th, 50th, and 90th tiles of the total particle size distributions of eight UCB patients and eight healthy subjects, using scattered light measurement. (F) Comparison of detected particle concentrations per urine between UCB patients and healthy subjects. Significance was determined by Mann-Whitney test. N.S., not significant  CD66a, CD66c, and CD66e alone, or that had two or three of these antigens present simultaneously, was suggested. Patient 2 showed CEACAM8 in shotgun analysis; therefore, we attempted to detect it by flow cytometry and were able to observe CEACAM8-positive EVs ( Figure S2). However, in flow cytometry analysis, the EVs in which CEACAM8 expression could be observed were infrequent in patient urine. We also observed a population of EVs that merged with CD66a/CD66c/CD66e, suggesting that the EVs were simultaneously expressing one of these antigens.
Because many EVs with major CD66a, CD66c, and CD66e (CEACAM1, CEACAM6, and CEACAM5) were observed in the patient's urine, we decided that a measurement system using an Ab that recognizes these antigens without distinction would be appropriate, and conducted a flow cytometric analysis using this Ab (Clone ASL-32). Figure 3C shows a typical example. This is in order to achieve a more sensitive diagnostic system. Extracellular vesicles that were CD66a-/CD66c-/CD66e-positive by Clone ASL-32 (CEACAM+ EVs) were selected after gating to remove multiple aggregated particles, to remove particles reactive to IgG (e.g., THP polymer), and to distinguish them from multipeptidase+ EVs derived from renal tubules ( Figures 3D and S3).
The CEACAM+ EVs were also MUC1-positive EVs ( Figures 3E   and S3), although many MUC1+ EVs were detected in the urine of healthy individuals. It is possible that CEACAM+ EVs were originally derived from uroepithelial cells and secreted with CEACAM family proteins on their surface as antigenic.
We also observed the nature of the lipid bilayer in CEACAM+ EVs with annexin V (exposure of PS). Phosphatidylserine appeared to be facing out of EV membranes to varying degrees, depending on the patient ( Figure S4).

| Verification of possible diagnostic applications of CEACAM+ EVs using urine from UCB patients
Erythrocyte-derived EVs were present in the urine of hematuric patients ( Figure S3). Multipeptidase+ EVs derived from renal tubules were characterized in the same way as in previous studies ( Figure S3). 13 In the flow cytometric analysis, there were four types of EVs in patients' urine that could be characterized on the basis of our previous studies 13 : (i) MUC1+ uroepithelial EVs (CD66a-/ CD66c-/CD66e-negative); (ii) CEACAM+ EVs; (iii) multipeptidase+ EVs derived from renal tubules; and (iv) erythrocyte-derived EVs ( Figure 4A).
We confirmed that CEACAM+ EVs were not abundant in the urine of healthy subjects and non-UCB patients ( Figure 4B). Therefore, 33 UCB urine specimens, 18 non-UCB (urological disease) specimens, and 29 healthy specimens, all from men, were used to analyze the four types of EVs by the characterization method described above using a flow cytometer ( Table 2, Figure S5). The proportion (%) of these four types of EVs observed in the urine in the entire image is shown in Figures 4C and S6. The number of CEACAM+ EVs was significantly increased in UCB (mean ± SD, 8.6% ± 13%) compared to non-UCB (0.69 ± 0.46) and healthy (0.46 ± 0.34) ( Figure 4D). The number of MUC1+ EVs was slightly decreased in UCB (10 ± 5.8) compared to non-UCB (16 ± 8.5) and healthy (17 ± 9.0) ( Figure 4D).
The diagnostic performance was evaluated by ROC curve in UCB and others (non-UCB and healthy specimens) ( Figure 4E). The area under the ROC curve was 0.907 (95% confidence interval, 0.833-0.981; p < 0.0001). On the basis of these results, the provisional cut-off point was also calculated, taking into account both sensitivity (81.82%) and specificity (97.87%).
The provisional CEACAM+ EVs percentage cut-off value was 1.34%. Twenty-nine urine samples from patients with UCB were compared for the results of urine cytology (class [2][3][4][5]  Abbreviations: VEGFA, vascular endothelial growth factor A; VEGFR2, VEGF receptor 2. a Number of genes in the user-provided lists with membership in the given ontology term. b Percentage of all the user-provided genes that were found in the given ontology term (only input genes with at least one ontology term annotation are included in the calculation).  Table 2). Using this method, seven of the nine cases with urine cytology results in class 2 (negative) were judged as positive ( Figure 4F). This suggests that the method covers the oversights of existing urine cytology and could be implemented as a valid test.

| DISCUSS ION
The release of microvesicles or microparticles (in other words, mEVs) by tumor cells is a very common event in the tumor microenvironment. 39 As a result, tumor-derived mEVs not only influence tumor cell biology but also profoundly advance tumor immunology. 40 The described methods for mEV isolation include step-wise centrifugation, which removes large cellular debris, followed by ultracentrifugation (14,000 g) to pellet the mEVs. 42 A method to observe tumor-derived mEVs in blood by flow cytometry was also introduced. 43 We extracted EV fractions from urine by centrifugation at 18,900 g. From the results of NTA, particle size fractions with a diameter of approximately 300 nm were mainly extracted from both healthy subjects and patients.
From the fluorescence NTA results, we also observed that the extracted fractions contained particles with lipid-bilayer membranes. In addition, there are reports that blood cell-derived mEVs express PS and are related to cancer malignancy, 44,45 but there are also reports that tumor-derived mEVs express PS. 46 In Figure S4, we observed that some CEACAM+ EVs were positive for annexin V. These might be involved in some cancer characteristics, as described above. 46 In the shotgun proteomic analysis of pooled samples of patients tissue. 49 In the future, nucleic acid content could become a new cancer biomarker by analyzing RNA in UCB EVs.
In an individual analysis of four healthy subjects and four UCB patients, we were able to select proteins that contributed to each grouping and were particularly increased in the patient groups.
These selected proteins could be potential biomarkers specific to UCB because their dynamics can be divided into two groups in healthy subjects and patients by clustering ( Figure 2C), and some proteins have been reported to be increased in UCB. Many of the extracted proteins are adjacent to the plasma membrane, and annexins are an example of such proteins. Recently, the expression of annexin family members in UCB tissues was reported in detail. 34 In this report, ANXA2, ANXA3, ANXA4, ANXA8, and ANXA9 were signifi- introduced the sensitivity and specificity of various FDA-approved urine-based tests for UCB, but these tests do not replace urine cytology or cystoscopy because of their poor diagnostic performance and problems with false positives. 6 These kits have a mean sensitivity not exceeding 80%, and false positives resulting from hematuria have been a problem. In our system, erythrocyte-derived EVs originating from hematuria can be measured separately, and thus any resulting nonspecific reactions and false positives are not a problem.
The clinical sensitivity must be verified by large-scale clinical studies, but we believe that our system also requires improvement. Tilki et al. reported clinical sensitivity of 74% (69/93) and specificity of 95% (40/42) for measurement using CEACAM1 in urine, 7 and our results were close to this. In addition to CEACAM proteins, shotgun proteomics analysis revealed several proteins that could be potential markers for UCB in EVs, and we would like to further examine the possibility of improving sensitivity by combining these proteins.
Urothelial carcinoma of the bladder has a high recurrence rate, and prognostic monitoring tests in the postoperative or treatment course are very important. The urine samples examined in this study, including those from patients who presented with recurrence after transurethral resection of bladder tumor and during treatment, had a high number of CEACAM+ EVs (>10%) ( Table 2). It is necessary to verify these findings in the future, along with the elucidation of the generation mechanism of CEACAM+ EVs and their relationship with cancer progression.
In this study, we used flow cytometry for mEV measurement to detect multiple antigens and to confirm a certain particle size.
Considering the actual specimen processing and testing, the application of a simple automated assay system is desirable. An assay system that can directly measure exosomes in body fluids without EV extraction was recently reported. One is ExoScreen (Theoria Science, Inc.), which targets smaller EVs (e.g., exosomes) and could be implemented for biomarker screening in a variety of diseases. 52 Another system, ExoCounter (JVCKENWOOD Corporation), can determine the exact number of exosomes in the serum of cancer patients. 53 These have potential applications in automated systems that can process large numbers of specimens, but at present, both systems are limited to EVs with a size of 200 nm or less, and thus they need to be improved before their application to CEACAM+ EVs. Furthermore, EV detection technology has advanced rapidly in recent years, and their application to better measurement systems is expected. 54,55 Extracellular vesicles expressing CEACAM proteins in the urine of UCB patients could form a new liquid biopsy test. In the future, we aim to improve this protocol to make it easier and more practical, as well as to deepen the analysis of EV contents to expand the possibilities of liquid biopsies.

ACK N OWLED G M ENTS
We thank the laboratory members of Clinical Chemistry and Laboratory Medicine, Kyushu University, for reagents, discussions, and carefully reviewing the manuscript. We would like to express our gratitude to the Department of Urology and Laboratory Medicine, Kyushu University, for collecting valuable clinical samples.
We are grateful to LSI Medience Corporation for its generous support during this study. This research was supported by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (grant numbers 17H01550, 18 K15421, and 20H00530) and AMED (grant number JP21lm0203009). Finally, we thank H. Nikki March, PhD, from Edanz for editing a draft of this manuscript.

CO N FLI C T O F I NTE R S E T
K. Igami is a full-time employee of LSI Medience Corporation. K.
Igami is also seconded to and belongs to Kyushu Pro Search Limited Liability Partnership, a subsidiary of LSI Medience Corporation. M.
Eto is an Editorial Board Member of Cancer Science. No disclosures were reported by the other authors.