The trajectory of vesicular proteomic signatures from HBV-HCC by chitosan-magnetic bead-based separation and DIA-proteomic analysis

Abstract

Hepatocellular carcinoma (HCC) is a prevalent primary liver cancer often associated with chronic hepatitis B virus infection (CHB) and liver cirrhosis (LC), underscoring the critical need for biomarker discovery to improve patient outcomes. Emerging as a promising avenue for biomarker development, proteomic technology leveraging liquid biopsy from small extracellular vesicles (sEV) offers new insights. Here, we evaluated various methods for sEV isolation and identified polysaccharide chitosan (CS) as an optimal approach. Subsequently, we employed optimized CS-based magnetic beads (Mag-CS) for sEV separation from serum samples of healthy controls, CHB, LC, and HBV-HCC patients. Leveraging data-independent acquisition mass spectrometry coupled with machine learning, we uncovered potential vesicular protein biomarker signatures (KNG1, F11, KLKB1, CAPNS1, CDH1, CPN2, NME2) capable of distinguishing HBV-HCC from CHB, LC, and non-HCC conditions. Collectively, our findings highlight the utility of Mag-CS-based sEV isolation for identifying early detection biomarkers in HBV-HCC.

Keywords: hepatocellular carcinoma; polysaccharide chitosan; proteomics; small extracellular vesicles.

FIGURE 1

Characterization of sEV separated by distinct methods. (a) Workflow of evaluating sEV separated by IC, CS, UC, SEC, DGC and PEG. (b–d) Evaluation of sEV separated by distinct methods using TEM (b), dynamic light scattering (c), and western blotting (d). In the western blotting analysis, serum (0.2, 0.1 and 0.05 μL) was used as a positive control. sEV from 200 μL of serum separated by IC, CS, UC, SEC, DGC and PEG were loaded (right panel).

FIGURE 2

Proteome analysis of serum sEV separated by distinct methods. (a) Venn diagram of identified proteins in sEV separated by UC, GDC, PEG, IC, SEC, and CS. (b–d) Number (b), CV (c), and PLS‐DA plot (d) of identified proteins in sEV separated by each method. (e) Pearson correlation of identified vesicular proteins. (f&g) Venn diagram of identified vesicular proteins in our study and all proteins (f), or top 100 reported proteins (g) in the Vesiclepedia database.

FIGURE 3

Comparative proteome analysis of sEV separated by distinct methods. (a–c) Levels of 11 conventional sEV protein markers (a), 13 newly defined sEV markers (b), and non‐sEV proteins (c) across sEV separated by distinct methods. (d) The comparison of identified proteins categorized as “blood proteins”, “Vesiclepedia top 100”, and “EV associated GO terms” in sEV separated by distinct methods. (e) HCA and GO enrichment analysis of the clustered vesicular proteins. (f) The comparison of identified proteins categorized for diverse cellular compartments.

FIGURE 4

Comparative proteome analysis of sEV separated by free CS and Mag‐CS. (a) Venn diagram of identified vesicular proteins separated by free CS and Mag‐CS. (b) CV of identified vesicular proteins separated by free‐CS and Mag‐CS. (c–e) Levels of 11 conventional sEV protein markers (c), 13 newly defined sEV markers (d), and non‐sEV proteins (e) across sEV separated by free‐CS and Mag‐CS. (f) The comparison of identified proteins categorized as “blood proteins”, “Vesiclepedia top 100”, and “EV associated GO terms” in sEV separated by free CS and Mag‐CS. (g) HCA and GO enrichment analysis of the clustered vesicular proteins. (h) The comparison of identified proteins categorized for diverse cellular compartments.

FIGURE 5

Proteome analysis of sEV derived from serum sample from HC, CHB, LC and HBV‐HCC. (a) Workflow of the proteome analysis of serum sEV. (b) Venn diagram of identified proteins in serum sEV from HC, CHB, LC and HBV‐HCC. An equal volume of serum (200 μL) was employed for sEV separation and proteomic analysis. (c) Number of identified proteins in sEV from different groups. (d) Rank plot of identified proteins in sEV. Commonly identified vesicular proteins are highlighted in red, and top 5 abundant vesicular proteins are highlighted in blue. (e) CV of identified vesicular proteins in sample cohorts and quality control samples. Pooled serum was used as quality control, that was included within the sample cohorts in proteomic analysis. (f&g) Positivity for conventional (f) and novel (g) sEV markers across different sEV. (h) PLS‐DA plot of identified proteins in serum sEV from HC, CHB, LC and HBV‐HCC. (i) Pearson correlation of identified vesicular proteins.

FIGURE 6

Differences and trajectories of vesicular proteins in the HBV‐HCC progression. (a) Volcano plot of the serum sEV proteomes in comparison of HBV‐HCC versus HC (left), CHB (middle), and LC (right). ‐log10 (p‐value) is plotted against log10 (FC, fold change) using cutoffs of fold change > 1.2, fold change < 0.83, and p‐value < 0.05 (Student's t‐test). Red, upregulation; blue, downregulation. Proteins continuously dysregulated in HBV‐HCC progression were highlighted. (b) Trajectories of serum sEV proteins that differentially altered in the HBV‐HCC progression (p < 0.05, one‐way ANOVA). Trajectories were clustered into four groups according to their expression patterns. (c&d) Hierarchical clustering map (c) and GO enrichment analysis (d) of the clustered vesicular proteins.

FIGURE 7

sEV DAMP molecules and associated potential therapeutic targets according to the Therapeutic Target Database. Fifty‐three sEV DAMPs were differentially altered in the HBV‐HCC progression, of which, 16 sEV DAMPs were identified as drug targets according to the Therapeutic Target Database (TTD). Target type, drug name, and disease were obtained from TTD.

FIGURE 8

Serum vesicular protein signatures distinguishing HBV‐HCC using machine learning. (a) Workflow of feature selection and machine learning modelling. (b&c) ROC curve and confusion matrix performance of the vesicular protein panels in HBV‐HCC versus CHB, HBV‐HCC versus LC, and HBV‐HCC versus nonHCC using random forest model.