This study enrolled 10 patients with postoperative PR secondary to CHD who underwent TPVR via a trans-thoracic approach at Guangdong Provincial People’s Hospital, constituting the experimental group. Inclusion criteria were as follows, (1) moderate to severe PR confirmed by echocardiography; (2) evidence of RVOT dysfunction, meeting at least one of the following, right ventricular end-diastolic volume indexed (RVEDVI) ≥ 130 mL/m², right ventricular ejection fraction (RVEF) < 45%, QRS width ≥ 160 ms, sustained atrial or ventricular arrhythmias, or moderate or greater tricuspid regurgitation; (3) mean pulmonary artery pressure ≤ 30 mmHg; and (4) main pulmonary artery length ≥ 20 mm and diameter ≥ 16 mm. Exclusion criteria were as follows, (1) pulmonary hypertension or anatomical contraindications (e.g., abnormal RVOT position, coronary artery anomalies); (2) acute heart failure; (3) active infective endocarditis or a history of prior infective endocarditis; (4) allergy to aspirin, heparin, nickel-titanium alloy, or contrast agents; (5) active rheumatic disease; (6) infectious diseases such as hepatitis B, hepatitis C, HIV, or syphilis. Additionally, 10 healthy individuals were selected as a control group. The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Guangdong Provincial People’s Hospital (approval number: QX2021-011-02). All participants provided written informed consent prior to enrollment.

Venous blood (12 mL) was collected from each of the 10 patients undergoing TPVR at three time points: preoperatively (T1 group), 1 day postoperatively (T2 group), and 7 days postoperatively (T3 group). For the control group (N group), venous blood (12 mL) was collected simultaneously from 10 healthy individuals. Each plasma sample was technically replicated in triplicate for PBA analysis, yielding a total of 120 technical measurements. All samples were drawn into ethylenediaminetetraacetic acid (EDTA) anticoagulant tubes and processed promptly, or temporarily stored at 4 °C and processed within 4 h. The samples were centrifuged at 1,500 × g for 20 min at 4 °C to remove cellular component (CC). The plasma supernatant was carefully transferred to a new 1.5 mL centrifuge tube. Subsequently, the supernatant was centrifuged at 3,000 × g for 15 min at 4 °C to thoroughly remove residual blood cells. The plasma supernatant was then carefully aspirated, aliquoted, and stored at -80 °C.

For EV identification by Nanoparticle Tracking Analysis (NTA), Transmission Electron Microscopy (TEM), and Western Blot (WB), separate plasma aliquots from three healthy controls were pooled and processed by ultracentrifugation independently of the samples used for PBA analysis. NTA, TEM, and WB were performed for EV identification. Frozen plasma was thawed completely at 4 °C and kept on ice. Samples were centrifuged at 500 × g for 10 min at 4 °C, followed by centrifugation at 20,000 × g for 20 min at 4 °C. The supernatant was then centrifuged at 100,000 × g for 70 min at 4 °C. The resulting pellet was resuspended in 40 mL pre-chilled phosphate-buffered saline (PBS) and centrifuged again at 100,000 × g for 70 min at 4 °C.

For PBA analysis, EVs were prepared by size-exclusion chromatography and ultrafiltration (UF) as follows. (1) Plasma Pretreatment: Thaw the plasma and filter it through a 0.8 μm filter to remove large particles. Rinse the filter with 250 μL of PBS, combine the filtrate, and adjust the volume to 1 mL for subsequent steps; (2) Column Equilibration: Retrieve the gel exclusion column (Exosupur® ES9P14e/ES9P11e, Echo Biotech, China) stored at 4 °C, and wash it with 10 mL of 0.1 M PBS to equilibrate the column environment; (3) Sample Loading: After the liquid in the column has completely drained, load the 1 mL pretreated plasma onto the column, allowing it to pass through the resin slowly; (4) Elution & Collection: Once the sample has fully entered the resin, elute with 0.1 M PBS, collecting the eluate in two fractions of 1 mL each (total 2 mL). During this phase, EVs elute with the PBS solution; (5) UF & Lysis: Combine the eluates in a 100 kDa molecular weight cut-off Amicon® Ultra spin filter (Merck, Germany), centrifuge at 4,000 × g for 2 min to concentrate to approximately 200 μL. Add 700 μL of lysis buffer, mix thoroughly, and incubate for 2 min to obtain the lysate for PBA analysis.

EV samples were diluted with PBS to a detection concentration of 1 × 10⁸/mL for analysis. First, the instrument flow rate was calibrated using fluorescent silica microspheres with a known concentration and a diameter of 250 nm by correlating detection time with particle count. Under the same pressure conditions, the actual concentration of the EV sample was calculated based on this flow rate and particle count. Meanwhile, standard silica microspheres with diameters of 68, 91, 113, and 155 nm were used to establish a standard curve of scattered light intensity versus particle size. The scattered signals from the EV samples were converted to size data using this curve, thereby obtaining their final size distribution.

10 µL of EV suspension was deposited onto a copper mesh and incubated at room temperature for 10 min to facilitate adsorption. The mesh was then rinsed with sterile distilled water, and residual liquid was removed with absorbent paper. Negative staining was performed by applying 10 µL of 2% uranyl acetate for 1 min; excess stain was removed, and the grid was dried for 2 min under an incandescent lamp. The prepared grid was then examined under a transmission electron microscope (H-7650, Hitachi Ltd., Tokyo, Japan) at an accelerating voltage of 80 kV. This protocol enhances contrast through negative staining and reduces sample damage via low-voltage imaging, thereby clearly revealing the ultrastructural details of the EVs.

The protein concentration of EVs was determined using the Bicinchoninic Acid assay. Based on the calculated concentration, a loading amount of 10-30 µg was mixed with 5× SDS loading buffer, vortexed, and denatured by heating at 95 °C for 5 min. Denatured samples were then subjected to SDS-PAGE electrophoresis. Target gel regions containing exosomal markers Alix, CD9, CD81, and the negative control Calnexin were excised and transferred onto PVDF membranes via electroblotting. After blocking the membranes with 3% BSA, they were sequentially incubated with primary and secondary antibodies. Finally, the target proteins were visualized and analyzed through development, fixation, and exposure.

The surface proteins on EV were profiled using the PBA (ExoSeek® panel260, Secretech, Shenzhen, China)^[13]. The PBA procedure was described in the PBA method paper^[13]. Briefly, we follow the steps below. The PBA detection leverages a panel of 260 oligonucleotide-labeled antibodies to detect EV proteins simultaneously. Each antibody was conjugated with oligonucleotides consisting of a unique protein tag to mark protein type and a molecular tag to label duplicate detection. The EVs were captured on a surface with cholera toxin subunit B (CTB) coating. The antibody-labeled oligonucleotides located on the same EV were assigned the same unique EV tag. The DNA sequences comprising barcoding information of EV tag - protein tag - molecule tag were prepared into a sequencing library for each sample. The libraries were prepared for PE150 sequencing with the DNBSEQ-T7 platform (MGI, Shenzhen, China) or NovaSeq S4 (Illumina, USA). The sequencing raw data in bcl files were transferred into fastq files with DNA sequences using the bioinformatic tools MegaBolt for the MGI instrument or bcl2fastq for the Illumina platform.

This approach utilizes PBA technology to achieve combinatorial detection of membrane protein assemblies on individual EVs. It facilitates the exploration of co-expression and co-localization patterns among multiple membrane proteins at the single-EV level, thereby elucidating their potential collaborative involvement in intricate biological processes. In this study, we enumerated the number of vesicles containing each possible protein combination within each sample. Based on the single EV protein expression matrix for each sample, we initially categorized EVs according to the number of protein types they carried into four groups. EVs carrying only one protein were designated as combination_1; those carrying two proteins as combination_2; those carrying three proteins as combination_3; and EVs carrying four or more proteins as combination_4More. For each EV category, we profiled the expressed protein combinations and their respective abundances. Subsequently, we performed count per million (CPM) normalization on the EV protein combination data for each category. CPM normalization is calculated by dividing the protein expression level by the total expression level of a sample and then multiplying by 10^e6.

Plasma EVs were isolated as described above. For surface protein staining, purified EVs were resuspended in PBS and incubated with the following fluorochrome-conjugated antibodies^[14], PE anti-human LGR5, APC/Cyanine7 anti-human CD8a, ABflo® 647 anti-human CD111 (targeting NECTIN1), PE/Cyanine7 anti-human CD223 (targeting LAG-3), and ABflo® 488 anti-human CD166 (targeting ALCAM). All antibodies were used at the dilutions recommended by the manufacturer. The staining reaction was carried out at 4 °C for 30 min in the dark. After incubation, the stained EV samples were diluted with PBS to a final volume of 200 µL and analyzed immediately on a flow cytometer (FACSCanto^TM II, BD Biosciences). To discriminate EV events from background noise, a threshold was established based on the side scatter (SSC) and forward scatter (FSC) properties of size-calibrated reference beads (100 and 200 nm). At least 10,000 EV events were acquired for each sample. Data analysis was performed using FlowJo software (version 10). The percentage of EVs positive for each target protein was determined by gating against the corresponding isotype controls and fluorescence-minus-one (FMO) controls.

We conducted subpopulation and subpopulation marker identification for EVs using large-scale vesicle datasets. The analysis of EV subpopulations was carried out in the following steps: (1) Data Cleaning: We performed rigorous cleaning on massive vesicle datasets, filtering out low-abundance vesicles. The criterion retained vesicles that contained at least two distinct proteins, each with a count ≥ 2. To reduce computational burden, we randomly sampled 1/9 of vesicles from each sample for subpopulation identification. Proportional downsampling was applied uniformly across all samples to reduce computational bias caused by uneven EV counts while preserving relative subpopulation structure; (2) Data Normalization: We mitigated excessive dispersion in EV data to achieve a more uniform vesicle distribution. For the EV expression vector, the specific vesicle normalization formula is as follows: ev (normalization) = log₁₀ (cpm + 1); (3) Scale: We applied scaling to the EV data features (i.e., protein vectors) to eliminate the influence caused by unit and scale differences between features. The method employed was the z-scale method; (4) Feature Extraction: We utilized Principal Component Analysis (PCA) to further reduce data dimensionality, obtaining the principal dimensions that contribute to variation and diminishing the impact of dimensions with minor contributions; (5) EV Subpopulation Identification: We employed the self-organizing map (SOM) algorithm provided by FlowSOM^[15] to cluster single EV data, identifying EV subpopulations. Clustering robustness was evaluated through repeated subsampling and independent FlowSOM runs to ensure consistent identification of major EV subpopulations. This includes calculating the proportion of each subpopulation and comparing protein expression levels across subpopulations; (6) EV Subpopulation Visualization: We utilized t-distributed stochastic neighbor embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction and visualization algorithms to graphically represent single EV protein expression data.

The sequencing reads were analyzed with the software EVisualizer® decoding package (version 1.0, Secretech, Shenzhen, China) to generate EV ID-protein expression datasets as PBA raw data and then to analyze protein expression, combinations, and EV subpopulations. Briefly, the algorithm used in EVisualizer® is as follows. Based on the raw data, the total number of detected EVs and proteins was summarized as mean ± standard error (SE). The total expression of each protein was summarized as a protein expression dataset, which was subsequently normalized via the trimmed mean of M-values (TMM) algorithm to obtain a normalized total protein expression dataset. The co-expression of two or more proteins on the same single EVs was defined as a protein combination dataset. CPM normalization was performed for the protein combination dataset. For analysis of differentially expressed proteins (DEPs) and differentially expressed protein combinations (DEPC), we performed statistical analysis as follows. We examine the normality of data with the Shapiro-Wilk test. We examine the homogeneity of data with the F-test or Bartlett’s test. When comparing data of two groups, we use Student’s t-test if the data show normal distribution and homogeneous variance. The Mann-Whitney U test is used if the data is not normally distributed. The Welch’s t-test is used for data with normal distribution but not homogeneous variance. When comparing more than two groups (> 2 variables or categories), we choose the Analysis of Variance (ANOVA) test for a dataset with a normal distribution and homogeneous variance. The significant differential expression is analyzed with Duncan’s test. If the input data doesn’t have a normal distribution or homogeneous variance, the non-parametric Kruskal-Wallis test was used instead. The significant differential expression is analyzed with the Pairwise Wilcoxon Rank Sum Test. We leveraged the Benjamini-Hochberg (BH) method to adjust the P-values. Adjusted P-values (false discovery rate, FDR) < 0.05 were considered statistically significant. We also constructed the receiver operating characteristic (ROC) curves and calculated the area under the curve (AUC) values using the pROC package (v. 1.18.0)^[16]. Spearman correlation analysis was used to assess the associations between EV protein levels and clinical parameters in TPVR patients. To generate EV subpopulations, the unsupervised FlowSOM algorithm was used^[15]. The t-SNE and UMAP methods could be employed for plotting EV subpopulations^[17,18]. All analyses were conducted using R software and the EVisualizer® platform (version 1.0, Secretech, Shenzhen, China).

EV identification was conducted using newly processed healthy-control plasma samples that were independent of the PBA samples. NTA, TEM, WB were performed for EV identification. NTA showed a predominant particle population in the small-EV size range, with a modal diameter of 84.4 nm and a mean diameter of 131.7 ± 0.7 nm [Figure 1A]. The particle concentration was 4.33 × 10¹¹ ± 5.30 × 10⁹ particles/mL [Figure 1A]. TEM revealed membrane-bound vesicular structures with a round to cup-shaped morphology [Figure 1B]. WB detected the EV-associated proteins CD9, CD81, and ALIX in the EV preparation, whereas these markers were not detected in the plasma supernatant [Figure 1C]. In contrast, the endoplasmic reticulum marker Calnexin was not detected in the EV but showed weak expression in the plasma supernatant. NTA, TEM, WB were performed solely for EV identification.

We performed a differential expression analysis of proteins between the N and T1 groups, and generated a heatmap to visualize the DEPs between EVs from the T1 group and those from the N group [Figure 2A]. Specifically, the proteins DSCAML1, ALCAM, and LAG3 were markedly downregulated in the T1 group (P < 0.001), while CD8A, NECTIN1, and LGR5 were significantly upregulated (P < 0.001) [Figure 2B-G]. Subsequent ROC curve validation showed AUC values of 0.94, 0.96, and 0.96 for DSCAML1, ALCAM, and LAG3, respectively [Figure 2H-J]. Conversely, the AUC values for CD8A, NECTIN1, and LGR5 were 0.96, 1.00, and 1.00, respectively [Figure 2K-M]. We then performed functional enrichment analysis on these DEPs. Biological process (BP) analysis indicated that these proteins are primarily involved in cellular processes [Figure 2N]. Molecular function (MF) analysis showed that these proteins are mainly associated with binding [Figure 2O]. CC analysis revealed that the proteins are predominantly located in integral components of the plasma membrane, intrinsic components of the membrane, and integral components of the membrane [Figure 2P]. Furthermore, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis demonstrated that these proteins are significantly enriched in pathways such as the intestinal immune network for IgA production, tight junction, and natural killer cell-mediated cytotoxicity [Figure 2Q].

We compared the protein expression profiles between the preoperative and postoperative groups, and generated a heatmap of DEPs [Supplementary Figure 1A]. The expression levels of ALCAM and LAG3 proteins on postoperative day 1 were significantly higher than preoperative levels (P < 0.05) [Supplementary Figure 1B and C]. Conversely, CTLA4 protein expression on postoperative day 1 was significantly lower than preoperative (P < 0.01) [Supplementary Figure 1D]. Subsequent ROC curve analysis demonstrated that the AUC values for ALCAM, LAG3, and CTLA4 proteins were 0.80, 0.79, and 0.87, respectively [Supplementary Figure 1E-G]. We further performed functional enrichment analysis on these DEPs. Gene Ontology (GO) analysis revealed that, in terms of BP, these proteins are primarily involved in immune system processes and the regulation of T cell differentiation; in MF, they predominantly participate in transmembrane signaling receptor activity, cytokine receptor activity, and immune receptor activity; and in CC, they are mainly associated with integral components of the plasma membrane, intrinsic components of membranes, and integral membrane components [Supplementary Figure 1H-J]. KEGG pathway enrichment analysis indicated that these proteins are significantly involved in the side of the membrane and the external side of the plasma membrane [Supplementary Figure 1K]. Additionally, comparative analysis was performed between the T1 and T3 groups [Supplementary Figure 2A-K].

Subsequently, multiprotein combination analysis revealed that combinations such as ALCAM&CD36&DSCAML1&PECAM1, ALCAM&CD8A&DSC1&DSCAML1, ALCAM&DSCAML1 &ITGA1&ITGB1, and ALCAM&CD36&DSCAML1&ITGA1 were expressed at significantly higher levels in the N group compared to the T1 group (P < 0.001) [Figure 3A-E], The expression levels of the combination of CD36&CD8A&DSCAML1&TACSTD2 and CD8A&DSCAML1&KIT&SIGLEC6 were significantly lower in the N group compared to the T1 group (P < 0.05) [Figure 3F and G]. ROC curve analysis showed that the AUC values for these combinations all exceeded 0.94 [Figure 3H-K]. BP analysis indicated that these proteins are primarily involved in defense response, response to stress, and immune system processes [Figure 3L]. MF analysis showed that these proteins mainly participate in protein binding and cytokine binding [Figure 3M]. CC analysis revealed that the proteins are predominantly located on the cell surface, on the side of the membrane, and on the external side of the plasma membrane [Figure 3N]. KEGG enrichment analysis demonstrated that these proteins are significantly enriched in pathways such as the PI3K-Akt signaling pathway, neutrophil extracellular trap formation, and hematopoietic cell lineage [Figure 3O].

Furthermore, comparative analysis between the T1 and T2 groups revealed that combinations such as CD36&CD8A&DSCAML1&FCGR1A, CD36&CD8A&DSCAML1&Ly6G, ALCAM&CD36& CD8A&DSCAML1, and CD36&DSCAML1&ITGA1&ITGA2 were expressed at significantly higher levels in the postoperative group than in the preoperative group (P < 0.001) [Supplementary Figure 3A-E], and ROC curve analysis showed that the AUC values for these combinations all exceeded 0.85 [Supplementary Figure 3F-I]. BP analysis revealed that these proteins are primarily involved in complement activation, the alternative pathway, and the regulation of phagocytosis [Supplementary Figure 3J]. MF analysis indicated that these proteins are mainly associated with C-X3-C chemokine binding and protease binding [Supplementary Figure 3K]. CC analysis showed that the proteins are predominantly located in phagocytic vesicles, the cell surface, and intrinsic membrane components [Supplementary Figure 3L]. KEGG enrichment analysis demonstrated significant enrichment in pathways such as the Hematopoietic cell lineage and Efferocytosis [Supplementary Figure 3M]. Additionally, we performed a comparative analysis between the T1 and T3 groups [Supplementary Figure 4A-K]. Based on the co-expression data of proteins detected at the single EV level, we performed PCA and Multidimensional Scaling (MDS) on the co-expression patterns of various protein combinations, including single proteins [Supplementary Figure 5A and B], pairs of proteins [Supplementary Figure 5C and D], triples of proteins [Supplementary Figure 5E and F], and groups of four or more proteins [Supplementary Figure 5G and H].

We performed cluster analysis on EVs based on proteomic characteristics to identify EV subpopulations. We calculated the proportion of each subpopulation and compared the protein expression levels across them. In total, we classified the EVs into 14 subpopulations and visualized the characteristic proteins of each subpopulation using a heatmap [Figure 4A]. Subsequently, two dimensionality reduction and visualization algorithms, t-SNE and UMAP, were employed to visualize the single-EV protein expression data. The t-SNE and UMAP scatter plots revealed that EVs across all samples primarily clustered in Cluster 1 (18.79%), Cluster 2 (27.79%), Cluster 4 (20.5%), and Cluster 10 (10.23%) [Supplementary Figure 8A and B]. Individual sample t-SNE and UMAP visualizations further showed the distribution of EV subpopulations across the N, T1, T2, and T3 groups [Supplementary Figure 8C and D]. Subsequently, we performed clustering analysis on EVs from the N, T1, T2, and T3 groups [Figure 4B-E].

We then performed differential expression analysis of each EV subpopulation’s proportion and protein expression levels across groups. Cluster 1 shows a significantly lower proportion in the N group (12.81%) compared to T1 (22.02%), T2 (19.58%), and T3 (19.68%) (P < 0.05) [Figure 4F]. Conversely, Cluster 2 has a significantly higher proportion in the N group (37.31%) compared to T1 (23.88%) and T3 (20.6%) (P < 0.05) [Figure 4G]. Additionally, the proportion of Cluster 4 is significantly higher in the T3 group (24.34%) compared to the N group (19.46%) (P < 0.05) [Figure 4H]. Subsequently, we performed a detailed analysis of each subpopulation and listed the top ten most abundantly expressed proteins. Notably, the top three proteins in Cluster 1, Cluster 2, and Cluster 4 were CD8A, DSCAML1, and CD36; DSCAML1, ALCAM, and CD36; and DSCAML1, CD8A, and CD36, respectively [Figure 4I-K]. Moreover, Clusters 3 and 5 were significantly less abundant in the control group compared to the disease group (P < 0.05) [Supplementary Figure 9A and B]. In contrast, Clusters 7, 8, and 11 were markedly more abundant in the T2 and T3 groups than in the control group (P < 0.05) [Supplementary Figure 9C-E]. Conversely, Cluster 12 was significantly more abundant in the control group than in the disease group (P < 0.05) [Supplementary Figure 9F]. The top ten proteins for Clusters 3, 5, 7, 8, 11, and 12 are shown in Supplementary Figure 9G-L.