A total of 226 consecutive patients were initially screened in this single-center, observational study. All patients with symptomatic, drug-refractory PerAF underwent first-time catheter ablation between July and December 2024 at the Department of Cardiovascular Medicine, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine. Additional inclusion criteria included: age ≥ 18 years, left atrial diameter between 40 and 50 mm, and no documented history of other arrhythmias, including supraventricular tachycardia and atrial flutter. Exclusion criteria included: (1) structural heart disease other than hypertensive or ischemic heart disease; (2) prior pacemaker implantation or known congenital conduction disorder; (3) severe hepatic or renal dysfunction; and (4) active infection or malignancy at enrollment. The patient screening, enrollment, and grouping process is summarized in Figure 1. Written informed consent was obtained from all participants prior to enrollment. The study protocol conformed to the Declaration of Helsinki and was approved by the Institutional Ethics Committee of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine (Approval No. 2021-139).

Following discontinuation of antiarrhythmic drugs for at least five half-lives, patients underwent catheter ablation under conscious sedation. Pre-procedural evaluations, including transthoracic and transesophageal echocardiography, were routinely performed. A decapolar catheter and a bipolar catheter (St Jude Medical, Inc, St. Paul, MN) were placed within the coronary sinus and at the right ventricular apex, respectively. The surface electrocardiogram (ECG) and bipolar endocardial electrograms were continuously monitored and recorded using a computer-based digital amplifier and recording system (Bard Electrophysiology). Intravenous heparin was administered to maintain an activated clotting time of > 250 s during the procedure.

Radiofrequency or cryoballoon ablation was performed according to the operator’s preference. As a standard procedure, pulmonary vein electrical isolation (PVI) was conducted in all patients as the initial ablation strategy. After completion of PVI, a single synchronized biphasic shock (typically 100-200 J) was performed to restore sinus rhythm in all patients. SNRT was defined as the interval from the cardioversion artifact (time 0) to the onset of the first spontaneous sinus P wave observed on the surface ECG and confirmed by intracardiac electrograms. Prolonged SNRT was defined as ≥ 1,500 ms^[12,13]. Patients were classified into the normal-SNRT group and the prolonged-SNRT group according to SNRT duration. Additional ablation procedures were performed subsequently based on the patient’s clinical condition.

To enhance comparability between patients with prolonged SNRT and those with normal SNRT, PSM was performed. The probability of prolonged SNRT was estimated using a multivariable logistic regression model that incorporated age, sex, body mass index, hypertension, diabetes mellitus, coronary artery disease, previous stroke or transient ischemic attack, left atrial diameter, and left ventricular ejection fraction. Patients were paired in a 1:1 ratio using nearest-neighbor matching without replacement, with a caliper set at 0.2 standard deviations (SDs) of the logit of the propensity score. Post-matching balance was assessed using absolute standardized mean differences, and values < 0.10 were considered indicative of satisfactory balance.

For the proteomic discovery analysis, 10 participants were randomly chosen from the propensity score-matched population, including five with prolonged SNRT and five with normal SNRT. Fasting venous blood was obtained before ablation using ethylenediaminetetraacetic acid-containing tubes. Following centrifugation at 3,000 rpm for 10 min at 4 °C, plasma was aliquoted and kept at -80 °C until further analysis.

Briefly, low-abundance plasma proteins were enriched using a proprietary magnetic nanoparticle-based enrichment platform (OE Biotech Co., Ltd., Shanghai, China). After incubation at 37 °C for 1 h, washing steps were performed to remove non-specifically bound plasma proteins. The captured proteins were subsequently denatured, reduced, alkylated, and enzymatically digested at 37 °C for 2 h. Peptides were then purified using C18 spin columns, vacuum-concentrated, and quantified. Following addition of iRT standards at 1:20 (v/v), peptides were separated on an EASY-nLC 1200 system (Thermo Fisher Scientific, Waltham, MA, USA) using a 60-min gradient and analyzed by data-independent acquisition (DIA) on a timsTOF Pro mass spectrometer (Bruker Daltonics, Bremen, Germany). The main acquisition settings were 1.4 kV capillary voltage, 180 °C drying-gas temperature, m/z 100-1,700, and an ion-mobility range of 0.7-1.3 V·s/cm². Raw data were processed in Spectronaut 18.4 (Biognosys, Schlieren, Switzerland). Differentially expressed proteins (DEPs) were defined as those with a fold change > 2.0 or < 0.5 and a nominal P value < 0.05. P values were not adjusted for multiple comparisons.

Functional enrichment analyses of DEPs were performed using clusterProfiler (version 4.8.3) with annotations from the Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases. GO terms were categorized into biological process (BP), cellular component (CC), and molecular function (MF). Adjusted P-values were computed using Benjamini-Hochberg correction for multiple testing. Protein-protein interaction networks were constructed using the STRING database (version 12.0) with an interaction confidence score threshold of 0.7. The resulting networks were visualized in Cytoscape (version 3.10). Hub proteins were identified according to degree centrality, defined as the top 10% of nodes ranked by interaction count.

To identify coordinated protein modules correlated with prolonged SNRT, WGCNA was applied to the proteomic abundance data. Pearson correlations were calculated among the retained proteins and subsequently converted into weighted connection strengths using a soft-thresholding parameter β. The value of β was chosen to obtain an approximate scale-free network topology, with a fit index of 0.90. The adjacency matrix was converted to a topological overlap matrix, and proteins were hierarchically clustered according to topological overlap dissimilarity using the average-linkage method. Protein modules were delineated from the resulting dendrogram with the dynamic tree-cut procedure, using a minimum module size of 40, and were denoted by different colors. The first principal component of each module, referred to as the module eigengene, was calculated and correlated with the binary SNRT phenotype, coded as 1 for prolonged SNRT and 0 for normal SNRT. Module-trait correlations were visualized in a heatmap. Gene significance (GS, representing protein significance in this context) and module membership (MM) were computed for each protein. The relationship between GS and MM was plotted to identify hub proteins strongly associated with both network topology and the prolonged-SNRT phenotype. GO and KEGG enrichment analyses were repeated for key modules to determine biological themes associated with prolonged SNRT and impaired sinus node recovery function.

Based on its relevance to cardiac remodeling and its position within the WGCNA-identified module, Acid phosphatase 5 (ACP5) was selected for validation in the full matched cohort (n = 112). Plasma concentrations were measured using a commercial ELISA kit (CUSABIO, Wuhan, China; catalog no. CSB-E08490h) according to the manufacturer’s protocol. Each sample was assayed in duplicate, and the intra-assay coefficients of variation were < 10%. Absorbance at 450 nm was measured using a BioTek Synergy H1 microplate reader (BioTek Instruments, Winooski, USA).

Continuous data are presented as mean ± SD for normally distributed data or median with interquartile range for non-normally distributed data. Between-group comparisons were performed using Student's t-test or Wilcoxon rank-sum test, as appropriate. Categorical variables are presented as counts and percentages, and are compared using the Pearson chi-square test, Yates' continuity corrected chi-square test, or Fisher's exact test, as appropriate. Normality was assessed using the Shapiro-Wilk test. Correlation analyses were performed using Pearson or Spearman rank correlation coefficients for normally or non-normally distributed data, respectively. Statistical analyses and data visualization were conducted using R software (version 4.3.2) with the ggplot2, cowplot, and ComplexHeatmap packages. All statistical tests were two-sided, with P < 0.05 considered statistically significant.

A total of 226 patients with PerAF were screened according to the predefined inclusion and exclusion criteria [Figure 1], of whom 177 met the eligibility criteria and were included in the study. Based on SNRT measured immediately after PVI and synchronized cardioversion, participants were stratified into the prolonged-SNRT group (SNRT ≥ 1,500 ms) and the normal-SNRT group (SNRT < 1,500 ms) [Figure 2]. Among the 177 eligible patients, 59 patients (33%) had prolonged SNRT, whereas the remaining 118 (67%) patients had normal SNRT.

Baseline demographic characteristics, comorbidities, echocardiographic parameters, and medication use are summarized in Table 1. The sinus rate after cardioversion was significantly lower in patients with prolonged SNRT, whereas all other baseline parameters were comparable between the two groups. To minimize potential confounding, PSM was performed, resulting in 56 matched pairs with excellent covariate balance [Table 2, Supplementary Figure 1A and B], thereby ensuring balanced baseline characteristics for subsequent analyses.

A statistically significant inverse association was observed between sinus rate after cardioversion and SNRT [Figure 3]. This finding demonstrated that slower intrinsic sinus rates are associated with longer recovery time and supported the value of prolonged SNRT as an indicator of SND.

To identify potential molecular biomarkers of SND, label-free plasma proteomic profiling was conducted in 10 discovery samples comprising five randomly selected patients from each group in the matched cohort. Quality-control analyses, including distributional assessment and missing-value filtering, revealed no apparent outliers or batch effects [Supplementary Figure 2A and B]. Principal component analysis (PCA) showed separation between the prolonged-SNRT and normal-SNRT samples along the main components [Figure 4A]. Using predefined thresholds, we identified 435 DEPs, including 138 upregulated and 297 downregulated proteins in the prolonged-SNRT group [Figure 4B]. Unsupervised clustering of these DEPs revealed clear group-coherent clusters [Figure 4C], supporting a biological divergence between the two groups. Functional enrichment analysis showed that upregulated proteins were mainly associated with extracellular matrix (ECM) organization and collagen assembly, whereas downregulated proteins were enriched in oxidative phosphorylation and mitochondrial metabolic pathways [Figure 4D and E]. A protein-protein interaction network constructed from the DEPs [Figure 4F] showed high-degree hubs in inflammation and matrix remodeling clusters, consistent with matrix remodeling observed in enrichment analyses.

To characterize the modular organization of the proteome, WGCNA was performed using all quantified proteins. Sample clustering identified no outliers [Figure 5A]. Using a soft-thresholding power of β = 5 based on the scale-free topology criterion [Supplementary Figure 3A], the network was partitioned into multiple color-coded modules of varying sizes [Figure 5B and C]. Module-trait correlation analysis identified two modules associated with the prolonged-SNRT phenotype [Figure 5D]. The brown module was positively correlated with prolonged SNRT (r = 0.64, P = 0.04), whereas the green module was negatively correlated with the prolonged-SNRT phenotype (r = -0.86, P = 0.001). Eigengene dendrogram and adjacency heatmap analysis further showed that these modules formed distinct eigengene clusters, supporting coordinated proteomic programs associated with the prolonged-SNRT phenotype [Figure 5E].

Within each SND-associated module, MM was strongly correlated with GS [Figure 5F], indicating that topologically central proteins tend to be the most trait-informative. Functional dot plots [Figure 5G] summarized the GO and KEGG themes for these modules, aligning closely with the DEP-level enrichment shown in Figure 4. To further visualize network interconnectivity, a TOM plot was constructed based on protein dissimilarity. In the resulting heatmap, lighter colors indicate stronger topological overlap and higher levels of co-expression among proteins within identified modules, whereas progressively darker red colors indicate weaker overlap [Supplementary Figure 3B]. Modules other than brown/green are shown in Supplementary Figure 3C and D to complete the network survey.

ECM and collagen-assembly features differed consistently between the two groups across both analytical approaches, suggesting that ECM remodeling may be one of the biological processes associated with prolonged SNRT in patients with PerAF. ACP5, a protein previously implicated in cardiac fibrosis, was identified as a consistently robust hub overlapping both DEP and WGCNA analysis [Figure 4B; Figure 5F]. ACP5 was therefore selected for orthogonal clinical validation.

ELISA was performed for ACP5 in the full matched cohort (56 matched pairs; n = 112). ACP5 levels were significantly elevated in patients with prolonged SNRT compared with those with normal SNRT [Figure 6A]. However, plasma ACP5 levels were not significantly correlated with SNRT duration when SNRT was analyzed as a continuous variable (Pearson r = 0.088, R² = 0.008, P = 0.528; Figure 6B).

In conclusion, this study identifies a distinct plasma proteomic signature associated with prolonged SNRT in patients with PerAF. Using DIA-based proteomics and WGCNA, ACP5 was identified as a candidate plasma biomarker associated with prolonged SNRT. Collectively, these findings provide exploratory evidence that plasma ACP5 may serve as a candidate biomarker for assessing sinoatrial node function in patients with PerAF before rhythm control therapy.