Urinary extracellular vesicle N-glycomics identifies diagnostic glycosignatures for bladder cancer

Abstract

Bladder cancer (BC) is the most common urologic malignancy, facing enormous diagnostic challenges. Urinary extracellular vesicles (EVs) are promising source for developing diagnostic markers for bladder cancer because of the direct contact between urine and bladder. This study pioneers urinary EV N-glycomics for bladder cancer diagnosis. We have generated a comprehensive N-glycome landscape of urinary EVs through high-throughput N-glycome analysis, identifying a total of 252 N-glycans from 333 individuals. In bladder cancer patients, urinary EVs exhibit decreased fucosylation and increased sialylation level. An Eight N-glycan diagnostic model demonstrates strong performance in both validation cohorts, achieving ROC AUC values of 0.88 and 0.86, respectively. Furthermore, this model successfully differentiates both non-muscle invasive bladder cancer (NMIBC) and muscle-invasive bladder cancer (MIBC) from healthy individuals, underscoring the model's superiority. Moreover, urinary EVs N-glycoproteomic analysis reveals that the glycoproteins carrying cancer-associated N-glycan signatures are closely associated with immune activities. The N-glycome comparative analysis of EVs and their source cells indicate that the glycosylation profiles of EVs do not completely match the glycosylation backgrounds of their source cells. In summary, our study establishes urinary EV N-glycomics as a non-invasive BC screening tool and provide a framework for EV glycan biomarker discovery across cancers.

Fig. 1. Schematic overview of the study.

The urine samples from 333 individuals were collected for EV isolation. Alterations in the urinary EVs N-glycome among urologically related diseases and healthy individuals can unearth potential diagnostic markers for bladder cancer. In the discovery phase, 132 samples were classified into four subgroups: bladder cancer (BC, n = 45), urologic-associated benign patients (UB, n = 31), patients with other urologic malignancies (UM, n = 25) and healthy controls (HC, n = 31). By analyzing the N-glycome data in the discovery cohort, the N-glycome profiles of urinary EVs were constructed. Then, a bladder cancer diagnostic model was constructed using logistic regression algorithms. The diagnostic model was subsequently evaluated and validated in two validation cohorts (Validation cohort 1: n = 83; BC, n = 47; HC, n = 36; Validation cohort 2: n = 118; BC, n = 29; UB, n = 30; UM, n = 29; HC, n = 30;). Blue square: N-acetylglucosamine (GlcNAc); green circle: mannose (Man); yellow circle: galactose (Gal); purple diamond: N-acetylneuraminic acid (Neu5Ac); red triangle: fucose (Fuc). Created with BioRender.com.

Fig. 2. Characterization of EVs and quality control in the discovery phase.

a Characteristics of participants in the discovery cohort, encompassing identified N-glycans number, sex, age, BMI, hematuria, urinary tract infection status, TNM staging, and histologic grading. b NTA results revealed that the diameter range of isolated EVs spanned 40–200 nm across all four sample groups. c WB results demonstrated the presence of characteristic EV markers (CD9, CD81, TSG101) with Calnexin as the negative control in urinary EVs. d–g TEM images depicted urinary EVs in the BC, UB, UM, and HC groups, respectively. h Distributions of pair-wise Pearson’s r within QC samples and non-QC samples. The inner boxplots denote the minimum values, 25^th percentiles, medians, 75^th percentiles, and maximum values, respectively from bottom to top. i Comparative distribution of CVs among QC and biological samples in discovery cohorts. Images shown are representative of three independent experiments. Source data are provided as a Source Data file.

Fig. 3. Comprehensive landscape of the N-Glycome profile in urinary EVs.

a The boxplot illustrates N-glycan identifications in BC (n = 45), UB (n = 31), UM (n = 25), and HC (n = 31) samples. P-values from two-sided Wilcoxon rank-sum test (Benjamini-Hochberg adjusted) are annotated. b The venn diagrams illustrates the overlap of N-glycan identification among the four groups. 194 N-glycans were identified in all 4 groups. c The sankey diagram demonstrates the number of different types of N-glycans identified in urinary EVs N-glycome. Distinct N-glycan characteristics include N-glycan branching numbers, the presence or absence of fucose, sialic acid, and the N-glycosylation types. The thickness of the lines represents the number of N-glycans in each category. d The heatmap demonstrates the differential expression of various types of glycans across the four groups. Source data are provided as a Source Data file.

Fig. 4. Eight N-glycans were identified and screened as potential biomarkers.

a The Nightingale rose diagram illustrates the quantity of differential N-glycans among BC, UB, UM, and HC groups. b The heatmap depicts the intensity distribution of 21 differential N-glycans across urology-related diseases and healthy individuals. c The volcano plot represents differentially expressed N-glycans in three pairwise comparisons, including BC versus HC, BC versus UB, and BC versus UM (log₂(FC) > 1, p < 0.05). P-values were calculated from Tucky’s HSD post-hoc test (two-sided), adjusted by the Benjamini-Hochberg method. d The strategy for screening candidate N-glycans biomarkers. e Intensity distribution of candidate N-glycan biomarkers in the N-glycome of urinary EVs. f Performance benchmark of 10 machine learning classifiers based on three evaluation metrics: accuracy, F1 score and ROC AUC.Source data are provided as a Source Data file.

Fig. 5. Development and assessment of diagnostic model for bladder cancer.

a Characteristics of participants in the validation cohort 1, encompassing identified N-glycans number, sex, age, BMI, hematuria, urinary tract infection status, TNM staging, and histologic grading. b Characteristics of participants in the validation cohort 2, encompassing identified N-glycans number, sex, age, BMI, hematuria, urinary tract infection status, TNM staging, and histologic grading. c The receiver operating characteristic (ROC) curve evaluates the model’s capacity to distinguish between bladder cancer patients and healthy individuals. d EVGScores for each sample in BC (n = 47) and HC (n = 36) groups within the validation cohort 1, showcasing the model’s discriminatory capacity. P-value was derived from Wilcoxon rank-sum test (two-sided). The boxplots denote the minimum values, 25th percentiles, medians, 75th percentiles, and maximum values, respectively from bottom to top. e Same as (d) but for validation cohort 2 (BC: n = 29, HC: n = 30). Source data are provided as a Source Data file.

Fig. 6. A comprehensive characterization of EVcans proteins and pairwise comparative analysis between urinary EVs and tissue N-glycome.

a Pie charts show distribution of EVcans sites and EVcans proteins. b The co-occurrence probability of EVcans on the same protein or the same glycosylation site. c The heatmap demonstrates the differential expression of various glycoforms across the four groups. d Most dysregulated EVcans glycoforms were up-regulated in BC groups. e Hematoxylin and eosin (H&E) staining of the formalin-fixed, paraffin-embedded (FFPE) section used for MALDI-MSI. Scale bar: 2.5 mm. f Spatial clustering of N-glycans in formalin-fixed, paraffin-embedded (FFPE) sections based on MALDI-MSI outcomes. g The intensity distribution of differentially expressed N-glycans in tissues and EVs (Benjamini-Hochberg-adjusted ANOVA p < 0.05, log₂ | FC | > 1). H&E staining data shown are from a single experimental replicate. Source data are provided as a Source Data file.

Fig. 7. Analysis of N-glycome differences between bladder cancer cell lines and their derived EVs.

a The average numbers of identified N-glycans in the five cell lines and their derived EVs. Error bars denote standard errors (SDs) among three biological replicates. b The Venn diagrams illustrate the overlap of N-glycan identification in the five cell lines and their derived EVs. c The abundance distribution of top ten N-glycans identified in five cell lines and their EVs. d The abundance characteristics of N-glycome in five cell lines and their EVs. e The heatmap demonstrates the differential expression of 27 N-glycans in the five cell-derived EVs. Source data are provided as a Source Data file.