Introduction

The prevalence of Heart Failure with Preserved Ejection Fraction (HFpEF) is on the rise within clinical settings, a phenomenon attributed to demographic shifts towards an older population and an increase in associated risk factors such as obesity and diabetes.1 HFpEF, which accounts for a substantial proportion of all heart failure incidences, is increasingly acknowledged as a multifaceted clinical entity.

Central to the etiology of HFpEF, as postulated by the Paulus hypothesis, is the role of systemic inflammation. This inflammation, a consequence of various comorbid conditions, incites coronary endothelial inflammation, leading to cardiac malfunctions. The mechanism further implicates a decreased availability of nitric oxide, which contributes to myocardial and vascular rigidity through disruptions in the cyclic guanosine monophosphate (cGMP) pathway, a notion corroborated by tissue analyses from HFpEF patients indicating diminished cGMP concentrations. Proteomic studies underscore a significant correlation between inflammatory markers, HFpEF, and alterations in the extracellular matrix. Moreover, inflammation extends beyond the left ventricle to the left atrium, suggesting that atrial fibrillation may frequently serve as an early indicator of HFpEF, particularly among individuals with obesity or diabetes.2 Histological analysis of atrial tissue samples indicates that the inflammatory milieu linked with HFpEF is pivotal in triggering and sustaining AF.3 In individuals treated with AF ablation, the concentrations of inflammation-related markers, including C-reactive protein, interleukin-6, and matrix metalloprotease-2, showed notable differences when comparing those who maintained sinus rhythm post-procedure to those who relapsed into AF.4 AF contributes to left atrial enlargement, compromised atrial performance, and increased atrial scarring, potentially leading directly to HFpEF.5 Notably, successful cardioversion often results in the recovery of atrial pumping efficiency and enhances ventricular preload, with the atrial share of ventricular filling rising from 30% to 47% one month after sinus rhythm is reestablished.6 Additionally, AF is linked to left ventricular fibrosis, which further exacerbates diastolic dysfunction, a key factor in HFpEF development.7 The process of atrioventricular annular reshaping, which can cause worsening mitral and tricuspid valve leakage, presents yet another pathway through which AF may lead to HFpEF.8 Moreover, the reduction in atrial natriuretic peptide (ANP), commonly seen in persistent AF, could result in increased vasoconstriction and fluid buildup, thereby predisposing individuals to HFpEF.9

Epicardial adipose tissue (EAT) significantly contributes to the development of HFpEF and AF by emitting pro-inflammatory and fibrogenic substances. EAT can directly affect adjacent cardiac structures, including the myocardium and coronary arteries, by releasing adipokines through a paracrine pathway.10 This process is supported by findings that show the potential of these adipokines to contribute to coronary atherosclerosis and cardiomyocyte dysfunction. For instance, exposure to the adipokine interleukin (IL)-1β in EAT has been linked to increased intimal thickening in coronary arteries and prolonged field potential duration in cardiomyocytes, indicating a diffusion mechanism that exacerbates coronary lesions and affects cardiac function.11,12 Furthermore, pro-inflammatory cytokines and profibrotic factors like interleukins, TNF, matrix metalloproteinases (MMPs), and activin A from EAT are implicated in promoting atrial fibrosis and arrhythmias.13 The overexpression of MMPs, heavily produced in EAT, disrupts extracellular matrix homeostasis, leading to fibrosis. EAT-derived factors like transforming growth factor-β (TGFβ) and connective tissue growth factor (cTGF) further contribute to this process, indicating a significant pathogenic role of EAT in atrial fibrosis and thereby in the development of AF.14 Clinical studies correlating high EAT volume with increased intramyocardial triglyceride content and atrial fibrosis reinforce the notion of EAT’s infiltrative and adverse effects on cardiac function.15 This highlights EAT’s pivotal role in the progression of HFpEF and AF through its pro-inflammatory and profibrotic secretions affecting the heart tissue.

In this study, we employed a range of bioinformatics approaches to pinpoint pivotal genes originating from EAT and potential pathways linked to the simultaneous occurrence of HFpEF and AF, using data sourced from GEO database. We also discovered potential therapeutic agents for HFpEF and AF. Additionally, machine learning techniques and receiver operating characteristic curve analysis were applied to identify key genes (ITPKA and WNT9B), whose expression patterns were analyzed in a small patient cohort at our institution.

Transcriptome Data Collection and ProcessingDataset Selection

Datasets detailing the expression profiles of epicardial adipose tissues (EAT), both with and without heart failure with preserved ejection fraction (HFpEF), were acquired from the GEO database (accession numbers: GSE135445, GSE128188 for EAT; GSE75092, GSE31821, GSE79768, GSE115574 for AF). Additional datasets specific to HFpEF (E-MTAB-7454) were sourced from the ArrayExpress database, and cardiac biopsy data were obtained from the European Nucleotide Archive (PRJEB62450). Datasets were selected based on criteria including sample size, relevance to HFpEF and AF, and data quality, prioritizing those with comprehensive metadata and appropriate control groups. Table 1 showed the information of all the datasets used in this study.

Table 1 Information of the Datasets Used in This Study

Preprocessing and Normalization

Raw expression data were preprocessed to remove low-quality reads and non-expressed genes using the R package edgeR (version 3.36.0). Batch effects were corrected using the combat function from the “SVA” package (version 3.48.0), ensuring consistency across merged datasets. Normalization was performed using the TMM (Trimmed Mean of M-values) method to adjust for differences in library size.

Secretory Protein Gene Acquisition

Secretory protein genes were identified using The Human Protein Atlas database, specifically from the “SPOCTOPUS predicted secreted proteins” section under subcellular localization. A total of 3947 genes were compiled for further analysis, ensuring that only those with verified secretion pathways were included.

Differentially Expressed Genes (DEGs) Analysis

Differentially expressed genes were identified using the limma package (version 3.56.2) in R. Criteria for DEG identification included a P-value < 0.05 and an absolute log fold change (|logFC|) exceeding the mean |logFC| plus two standard deviations (SD). To validate the robustness of DEG identification, sensitivity analyses were conducted by varying threshold parameters, confirming consistent results across different settings. Results were visualized using volcano plots and heatmaps generated with the ggplot2 (version 3.4.3) and pheatmap (version 1.0.12) packages.

Weighted Gene Co-Expression Network Analysis (WGCNA)

The WGCNA package was used to construct co-expression networks and identify gene modules correlated with clinical traits of HFpEF and AF. The soft threshold power was determined by examining the scale-free topology criterion, with a threshold of 11 selected based on average connectivity and scale independence. Modules containing more than 50 genes were identified using hierarchical clustering and dendrogram analysis. Each module’s eigengene was calculated to represent its overall expression profile, facilitating correlation analysis with clinical traits.

Functional Enrichment Analysis

Functional enrichment of DEGs was performed using the clusterProfiler package (version 4.8.3). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted with significance set at a P-value < 0.05. The biological processes and pathways associated with DEGs were further explored to elucidate their roles in HFpEF and AF pathogenesis.

Connectivity Map (cMAP) Analysis

The Connectivity Map (cMAP) database (https://clue.io/) was utilized to explore potential small-molecule therapeutics for HFpEF and AF. Upregulated genes from protein-protein interaction (PPI) networks were submitted to the cMAP platform. Enrichment scores were calculated to identify compounds that inversely correlate with disease gene signatures, indicating potential therapeutic efficacy. The top 10 compounds for each condition were selected based on their enrichment scores.

Machine Learning Algorithms for Feature Selection

Machine learning techniques, including Lasso, SVM-RFE, and Random Forest, were applied to AF expression profiles. A fixed seed (12345) ensured consistent results across methods. The Lasso algorithm utilized the glmnet package (version 4.1–8) with 5-fold cross-validation, optimizing the lambda parameter for feature selection. SVM-RFE employed the e1071 (version 1.7–13) and MSVM-RFE packages, utilizing 10-fold cross-validation for evaluation. The Random Forest model, built with the randomForest package (version 4.7–1.1), consisted of 500 trees, with the optimal number determined through cross-validation error analysis. Common diagnostic genes were identified across methods and evaluated for discriminative ability using the pROC package (version 1.18.4) with ROC curve analysis.

Immune Infiltration Analysis

The composition of immune cells in HFpEF and AF samples was analyzed using the CIBERSORT algorithm (version 0.1.0). The LM22 signature matrix was employed to quantify 22 immune cell types, with 1000 permutations to ensure accuracy. Immune cell proportions were visualized using ggplot2 (version 3.4.3), with differences between control and disease groups assessed using box plots. Correlations among immune cell types and their associations with diagnostic biomarkers were analyzed using Spearman’s method and visualized with the corrplot package (version 0.92).

Single Gene Set Enrichment Analysis (GSEA)

Single-gene GSEA was conducted using the clusterProfiler package, with gene sets obtained from the Molecular Signatures Database (MSigDB) (c5.go.Hs.symbols.gmt file). The top five enriched pathways were identified for each disease group, with results visualized using the Enrichplot package (version 1.20.3).

Patients’ Sample Collection

Epicardial fat tissues and serum samples were collected from control subjects (n=10), HFpEF patients (n=8), AF patients (n=12), and patients with both HFpEF and AF (n=10) at Changhai Hospital. Ethical approval was obtained from the Institutional Research Ethics Committee at Changhai Hospital (approval number: 2023–01), and all participants provided informed consent in accordance with the Declaration of Helsinki. Samples were processed under sterile conditions to prevent contamination, and RNA quality was assessed using an Agilent 2100 Bioanalyzer.

qRT-PCR

RNA was extracted using Trizol reagent (Thermo Fisher Scientific, Darmstadt, Germany), followed by cDNA synthesis using a reverse transcription kit (Ruizhen Bio, Guangzhou, China). Real-time quantitative PCR was performed using a SYBR Green PCR Kit (Ruizhen Bio) on a Roche LightCycler 480 system. Reactions were conducted in duplicate, and mRNA expression levels were calculated using the 2−ΔΔCt method. Primer sequences were as follows: ITPKA-F 5ʹ-CTTCGACGGACCTTGTGTG-3ʹ; ITPKA-R 5ʹ-CACCGCCAGCATTTTCTTGT-3ʹ; WNT9B-F 5ʹ-TGTGCGGTGACAACCTCAAG-3ʹ; WNT9B-R 5ʹ-ACAGGAGCCTGATACGCCAT-3ʹ; GAPDH-F 5ʹ-GGAGCGAGATCCCTCCAAAAT-3ʹ; GAPDH-R 5ʹ-GGCTGTTGTCATACTTCTCATGG-3ʹ.

Enzyme-Linked Immunosorbent Assay (ELISA)

Serum levels of ITPKA and WNT9B were measured using specific ELISA kits (MyBioSource, USA) according to the manufacturer’s instructions. Calibration curves were generated for each assay, and samples were analyzed in triplicate to ensure accuracy and reproducibility.

Statistical Analysis

Statistical analyses were conducted using GraphPad Prism (version 9.5). Data are presented as mean ± standard deviation (SD). The unpaired Student’s t-test was employed for comparisons between two groups, and one-way ANOVA with Tukey’s post hoc test was used for comparisons among three or more groups. P-values less than 0.05 were considered statistically significant. Multiple testing corrections, such as the Benjamini-Hochberg method, were applied to control for false discovery rates in high-throughput analyses.

ResultsGEO Information and Data Processing

Figure 1 depicts the bioinformatics analysis workflow. Three raw AF datasets were retrieved from the GEO database. Following their amalgamation and correction for batch effects, a consolidated and normalized AF dataset was produced, comprising 23 samples with sinus rhythm and 25 samples with AF.

Figure 1 Flowchart of this research.

As illustrated in Figure 2A and B, the batch effects among the three datasets were effectively mitigated following the batch effect correction process.

Figure 2 The Differential gene expression analysis of heart failure with preserved ejection fraction and atrial fibrillation. (A and B) PCA plot showing before and after batch effect correction in three AF datasets. (C and D) Volcano plot and heatmap showing the DEGs in HFpEF dataset. (E and F) Volcano plot and heatmap showing the DEGs in AF dataset.

Differential Gene Expression Analysis

The “Limma” package was utilized to identify DEGs between the disease and control groups. In the case of EAT from HFpEF, 575 DEGs were found (P < 0.05, |log2 FC| > 1.02), including 285 genes that were up-regulated and 290 that were down-regulated. For HFpEF heart tissues, 559 DEGs were identified (P < 0.05, |log2 FC| > 0.80), with 97 up-regulated and 462 down-regulated. For EAT in AF, 333 DEGs were discovered (P < 0.05, |log2 FC| > 0.80), comprising 142 up-regulated and 191 down-regulated genes. In AF PBMCs, 579 DEGs were found (P < 0.05, |log2 FC| > 1.46), with 261 up-regulated and 318 down-regulated. Additionally, 613 DEGs were determined for AF heart tissues (P < 0.05, |log2 FC| > 0.29), including 296 up-regulated and 317 down-regulated genes.

The expression patterns of DEGs in the heart tissue dataset of HFpEF and the combined dataset of AF heart tissues were illustrated using volcano plots and heatmaps, depicted in Figure 2C–F, respectively. The visual representations for EAT in HFpEF, EAT in AF, and PBMC in AF were presented in Figure S1A–F, respectively.

Key Modules Identification by WGCNA

To delve deeper into the significant genes associated with HFpEF, WGCNA was employed to pinpoint the gene modules most relevant to HFpEF in heart tissue samples. The optimal soft-thresholding power identified was 11, based on scale-free topology criteria (Figure 3A), resulting in the formation of 20 distinct gene modules. The cluster dendrogram showcasing these modules is depicted in Figure 3B. The correlation heatmap between HFpEF and the gene modules (Figure 3C) revealed that the greenyellow module had the strongest positive correlation with HFpEF (comprising 830 genes, r = 0.61, p = 0.01), while the magenta module was most negatively correlated with HFpEF (comprising 3625 genes, r= −0.85, p = 3e-08). Further analysis demonstrated a robust relationship between module membership and gene significance within both the greenyellow (r = 0.71, p = 3e-128, Figure S2A) and magenta modules (r = 0.84, p = 1e-200, Figure S2B), marking them as pivotal for further exploration. Consequently, a total of 4457 key genes, significantly linked with HFpEF, were pinpointed within the greenyellow and magenta modules.

Figure 3 Screening of key module in the HFpEF dataset via WGCNA and identification of HFpEF key genes through the intersection of key module genes and DEGs. (A) The scale-free topology model was utilized to identify the best β value, and β = 11 was chosen as the soft threshold based on the average connectivity and scale Independence. (B) The map showing the gene dendrogram and module colours. (C) The heatmap revealing the relationship between module eigengenes and status of HFpEF. (D) A total of 88 key genes in HFpEF were identified by taking the intersection between key modules genes, DEGs and secretory proteins via the venn diagram. (E) A total of 112 key genes in EAT from HFpEF were identified by taking the intersection between DEGs and secretory proteins via the venn diagram.

Furthermore, by intersecting genes from DEGs in HFpEF heart tissues, secretory proteins, and pivotal genes from WGCNA, a total of 88 key genes for HFpEF were pinpointed (Figure 3D). Similarly, crossing DEGs in EAT from HFpEF with secretory proteins yielded 112 genes (Figure 3E). Consequently, 200 significant secreted proteins derived from epicardial adipose tissue in HFpEF (DEG_EAT_HFpEF) were identified by amalgamating the sets of 88 and 112 genes.

For the AF group, WGCNA analysis was performed, choosing β = 6 as the ideal soft-thresholding power (Figure 4A). The cluster dendrogram presented in Figure 4B identified 25 modules, with bisque4 demonstrating the highest positive correlation (71 genes, r = 0.42, p = 0.004), and florawhite showing the most significant negative correlation with AF (84 genes, r = −0.57, p = 3e-05) (Figure 4C). A notable link was observed between module membership and gene significance in both bisque4 (r = 0.37, p = 0.0015, Figure S2C) and florawhite modules (r = 0.46, p = 1e-05, Figure S2D), identifying 155 key genes significantly associated with AF in these modules.

Figure 4 Screening of key module in the AF dataset via WGCNA and identification of AF key genes through the intersection of key module genes and DEGs. (A) The scale-free topology model was utilized to identify the best β value, and β = 6 was chosen as the soft threshold based on the average connectivity and scale Independence. (B) The map showing the gene dendrogram and module colours. (C) The heatmap revealing the relationship between module eigengenes and status of AF. (D) A total of 9 key genes in AF were identified by taking the intersection between key modules genes, DEGs and secretory proteins via the venn diagram. (E) A total of 94 key genes in EAT from AF were identified by taking the intersection between DEGs and secretory proteins via the venn diagram. (F) A total of 132 key genes in PBMC from AF were identified by taking the intersection between DEGs and secretory proteins via the venn diagram.

We then further intersected genes from DEGs in heart tissue form AF, secretory proteins and crucial genes from WGCNA to identify the critical genes in AF, obtaining 9 genes (Figure 4D). The DEGs in EAT from AF were intersected with secretory proteins and got 94 genes (Figure 4E). The DEGs in PBMC from AF were intersected with secretory proteins and got 132 genes (Figure 4F). The 232 differentially expressed epicardial adipose tissue-derived secreted proteins in AF (DEG_EAT_AF) were obtained through the combination of 9, 94 and 132 genes.

Candidate Small‑molecular Compounds for HFpEF and AF

To identify small-molecule drugs with therapeutic potential for EAT-associated HFpEF or AF, the upregulated genes from DEG_EAT_HFpEF and DEG_EAT_AF were separately uploaded to the cMAP database. In the case of HFpEF, the top 10 compounds identified, including parachlorophenol, M-3M3FBS, H-8, GBR-12935, ezetimibe, erythrosine, dibenzoylmethane, depudecin, dehydrocholic acid, and bongkrek acid, were suggested as promising pharmacological agents for HFpEF management (Figure 5A). Figure 5B and C provide an overview of the mechanisms targeted and chemical structures of these compounds.

Figure 5 Screening of the potential small-molecular compounds for the treatment of HFpEF via cMAP analysis. (A) The heatmap presenting the top 10 compounds with the most significantly negative enrichment scores in 10 cell lines based on cMAP analysis. (B) The description of those top 10 compounds. (C) The chemical structures of those 10 compounds were shown.

Regarding AF, the top 10 compounds identified as potential therapeutic agents included RG-13022, Lypressin, LY-225910, Heraclenol, Gitoxigenin, Dipivefrine, CAY-10585, BX-795, Antimycin-a, and Amperozide (Figure S3A). Information on the pathways targeted by these 10 compounds is presented in Figure S3B.

Discover Common Genes Using Machine Learning Techniques

To investigate the shared pathogenesis between HFpEF and AF, we analyzed the overlap between DEG_EAT_HFpEF and DEG_EAT_AF. As depicted in Figure 6A, 13 genes (HLA-DQB1, HSPG2, ITPKA, NOTCH3, VPS9D1, C4BPA, CCL21, NKG7, NRXN1, PLA2G2D, ST6GAL2, UCHL1, WNT9B) were found in common between the two conditions, suggesting their potential involvement in the pathogenesis of both HFpEF and AF. Through functional annotation and enrichment analysis of these genes (Figures 6B), we aimed to discern potential biological pathways shared by HFpEF and AF. GO analysis indicated a significant representation of these shared genes in pathways linked to the activation of immune cells. Specifically, enriched biological processes such as T cell activation regulation, leukocyte cell-cell adhesion, and G protein-coupled receptor binding were identified, highlighting the role of immune cell activation in the joint pathogenesis of HFpEF and AF.

Figure 6 Identification of common genes for HFpEF and AF, and GO enrichment analysis. (A) The venn diagram showing the 13 overlapping genes of HFpEF and AF. (B) The bar plot showing the GO enrichment analysis results, including biological process, cellular component and molecular function of genes for the coexist of HFpEF and AF.

To discover additional diagnostic gene markers that effectively distinguish between disease and control groups, three separate machine learning methods (LASSO, SVM-RFE, and Random Forest) were applied, focusing on the previously mentioned 13 shared genes. Within the AF cohort, the LASSO regression algorithm pinpointed 10 out of the 13 shared genes as potential diagnostic candidates (Figure 7A and B). Subsequently, these 13 genes were analyzed using the RF classifier, which highlighted the top 5 genes for deeper exploration (Figure 7C and D). Moreover, the SVM-RFE technique, utilizing ten-fold cross-validation, identified 8 genes (Figure 7E). By integrating findings from these three methodologies, we identified 5 shared biomarkers (ITPKA, NKG7, NRXN1, UCHL1, WNT9B) relevant for the AF group (Figure 7F).

Figure 7 Identification of potential diagnostic biomarkers for the coexist of HFpEF and AF by the machine learning methods. (A and B) The minimum (A) and lambda values (B) of diagnostic biomarkers were identified by the LASSO logistic regression algorithm. (C and D) Random forest classifier was used to selected the top 5 genes. (E) The SVM-RFE algorithm identified 8 genes with ten-fold cross-validation. (F) The venn diagram displaying 5 common genes between LASSO and RF algorithms and SVM-RFE.

Assessment and Confirmation of Key Diagnostic Biomarkers

To gain a clearer insight into the link between HFpEF and AF, receiver operating characteristic (ROC) curve analysis was performed on both internal and external datasets. For the HFpEF internal dataset, these five genes showed the results: NKG7 (AUC=0.618), UCHL1 (AUC=0.473), ITPKA (AUC=0.891), WNT9B (AUC=0.800), NRXN1 (AUC=0.527) (Figure 8A). For the AF internal dataset, these five genes showed the results: NKG7 (AUC=0.538), UCHL1 (AUC=0.682), ITPKA (AUC=0.755), WNT9B (AUC=0.720), NRXN1 (AUC=0.583) (Figure 8B). Therefore, ITPKA and WNT9B exhibited the superior specificity and sensitivity for the diagnosis of both HFpEF and AF.

Figure 8 The ROC curve for the diagnostic performance and GSEA for each candidate biomarker. (A) The ROC curve of 5 candidate hub genes in internal HFpEF dataset. (B) The ROC curve of 5 candidate hub genes in internal AF dataset. (C) ITPKA and WNT9B ROC curve in external HFpEF dataset. (D) ITPKA and WNT9B ROC curve in external AF dataset. (E) The GSEA results of ITPKA in HFpEF. (F) The GSEA results of ITPKA in AF. (G) The GSEA results of WNTB in HFpEF. (H) The GSEA results of WNTB in AF.

Additionally, the credibility of ITPKA and WNT9B as fundamental diagnostic markers for HFpEF and AF was established through external validation. Figure 8C demonstrated that both ITPKA (AUC=0.708) and WNT9B (AUC=0.917) exhibited high diagnostic precision in the external cohort for HFpEF. In a similar manner, ITPKA (AUC=0.800) and WNT9B (AUC=0.800) accurately identified AF cases (Figure 8D), thus affirming their significance as crucial discriminative indicators for both HFpEF and AF.

Single Gene GSEA of Diagnostic Genes

Next, we employed single gene GSEA analysis of ITPKA and WNT9B in HFpEF and AF datasets, respectively. The top 5 enriched pathways were visualized by the “GSEA” package. Figure 8E and F showed that ITPKA was involved in cardiac muscle contraction, dilated cardiomyopathy in HFpEF and was involved in oxidative phosphorylation, TGF-β signaling pathway in AF. WNT9B was involved in aging and proline metabolism, calcium signaling pathway in HFpEF and was involved in ECM receptor interaction, focal adhesion in AF (Figure 8G and H).

Immune Infiltration of Diagnostic Genes

The relationship between diagnostic genes and HFpEF and AF were explored. As showed in Figure 9A, ITPKA in HFpEF was positively correlated with T cell CD4 memory activated, while WNT9B in HFpEF exhibited no correlation between all kinds of immune cells (Figure 9B). For AF, ITPKA was positively correlated with Plasma cells and T cell CD8 (Figure 9C) and negatively correlated with NK cells activated and T cell CD4 memory resting (Figure 9D).

Figure 9 The immune filtration analysis of hub genes in HFpEF and AF. (A) Lollipop of ITPKA and immune cells in HFpEF. (B) Lollipop of WNT9B and immune cells in HFpEF. (C) Lollipop of ITPKA and immune cells in AF. (D) Lollipop of WNT9B and immune cells in AF.

Validation of Diagnostic Genes in Human Samples

The mRNA expression of ITPKA and WNT9B in human samples were also validated. As showed in Figure 10A and B, upregulated expression was found for ITPKA and WNT9B in the epicardial fat tissues from HFpEF and AF patients, which were consistent with the results of serum samples from HFpEF and AF patients (Figure 10C and D).

Figure 10 Validation of the expression patterns of two hub genes in HFpEF and AF. (A) RT-qPCR showing increased mRNA levels of ITPKA in HFpEF and AF samples. (B) RT-qPCR showing increased mRNA levels of WNT9B in HFpEF and AF samples. (C) ELISA analysis displaying elevated serum ITPKA levels in HFpEF and AF samples. (D) ELISA analysis displaying elevated serum WNT9B levels in HFpEF and AF samples.

Discussion

HFpEF and AF are increasingly prevalent conditions, especially among the elderly, and often coexist due to shared risk factors and pathophysiology. The interaction between HFpEF and AF is complex and multifaceted, impacting diagnosis, treatment, and outcomes. HFpEF and AF share several clinical features and have a mutually reinforcing relationship. AF can exacerbate HFpEF symptoms by impairing diastolic filling and reducing cardiac output, while HFpEF creates a substrate for AF through atrial remodeling and increased filling pressures. Systemic inflammation has been recently recognized as a potential link between HFpEF and AF, adding to the complexity of their interrelationship16.

The coexistence of HFpEF and AF poses diagnostic challenges due to overlapping symptoms. The use of diuretic therapy for symptom relief can support the presence of HFpEF in patients with concomitant AF. Non-vitamin K oral anticoagulants, left atrial appendage occlusion devices, and catheter ablation have emerged as alternative treatment options for these patients.17 Moreover, novel pharmacological agents, including neprilysin inhibitors, are being explored for their potential benefits in this patient population, though more research is needed. Senile amyloidosis, characterized by the deposition of transthyretin amyloid in the heart, has also been identified as a factor contributing to HFpEF and AF, highlighting the role of atrial structural changes in the pathophysiology of these conditions.

To summarize, the intricate interplay between HFpEF and AF underscores the need for a comprehensive and multidisciplinary approach to diagnosis and management, aiming to address the complex pathophysiology and improve patient outcomes. Further research is necessary to better understand these interactions and develop more effective treatments for patients affected by both conditions.

EAT plays a significant role in the pathophysiology of HFpEF and AF, acting through various mechanisms including inflammation, effects on cardiac structure and function, and the secretion of bioactive substances.18 EAT may contribute to the pathophysiology of HFpEF by affecting coronary flow reserve, cardiac structure and function, and quality of life.19 Moreover, EAT’s unique properties distinguish it from other fat depots, potentially influencing heart health by secreting proinflammatory adipokines that can cause atrial and ventricular fibrosis, contributing to HFpEF.20

Obesity and diabetes, conditions associated with an increase in EAT, are significant risk factors for AF. EAT produces proinflammatory adipocytokines, leading to microvascular dysfunction and fibrosis of the underlying myocardium, resulting in an atrial myopathy manifest as AF.21 EAT thickness is associated with the incidence and severity of AF, suggesting its role in AF pathogenesis through adipocyte infiltration, pro-fibrotic, and pro-inflammatory paracrine effects.22 EAT is a metabolically active organ that not only stores lipids but also secretes a variety of bioactive substances, including inflammatory mediators and adipokines, affecting the heart and vascular system. Its close anatomical relationship with the myocardium and coronary arteries allows EAT to influence cardiac function directly and may play a role in the development of cardiovascular diseases including HFpEF and AF.23 Understanding the role of EAT in HFpEF and AF has potential therapeutic implications. Treatments that reduce EAT volume and ameliorate its inflammatory characteristics may offer new approaches to managing these conditions.24

However, few studies focused on the EAT-derived secreted proteins who might play their roles in the coexist of HFpEF and AF. Hence, we performed one comprehensive bioinformatics analysis aiming to bridge the HFpEF and AF based on EAT-derived secretory proteins. The entire research could be separated into three parts. First, we emphasized on HFpEF and filtered out the common DEGs in EAT and heart tissues, which were intersected with secretory proteins. Next, PBMC dataset, EAT dataset and integrated AF dataset were downloaded to filter the secretory protein related DEGs. Finally, two acquired gene sets (DEG_EAT_HFpEF and DEG_EAT_AF) were taken intersected, the common genes were considered as the EAT-derived proteins which could have an active role in both HFpEF and AF. However, in interpreting the origins of differentially secreted protein genes in both EAT and the myocardium, caution is essential due to the complex interplay between these tissues. Shared protein expression pathways, extensive paracrine signaling, and technical challenges in proteomics can obscure the precise source of proteins. To accurately determine protein origins, it is vital to utilize advanced proteomic techniques, integrative omics approaches, and functional validation studies. This careful interpretation is crucial for understanding the role of these proteins in HFpEF pathogenesis and for guiding the development of targeted therapies.

The present study identified 20 small molecules through cMAP analysis that exhibit potential therapeutic effects on EAT in the context of AF and HFpEF. These molecules were categorized based on their pharmacological mechanisms, offering diverse approaches to modulating EAT-related pathways implicated in these conditions. Anti-Inflammatory Agents: Compounds such as Parachlorophenol,25 Erythrosine,26 Heraclenol,27 Amperozide,28 CAY-10585,29 and BX-79530 demonstrate significant anti-inflammatory properties. By inhibiting the production of pro-inflammatory cytokines and modulating inflammatory signaling pathways within EAT, these agents can mitigate fibrosis and improve cardiac remodeling. This reduction in inflammation is critical for addressing the pathological processes underlying AF and HFpEF, where inflammation contributes to disease progression. Calcium Signaling Modulators: M-3M3FBS31 and Gitoxigenin32 are identified as key modulators of calcium signaling pathways. Calcium signaling is essential for maintaining cardiac contractility and rhythm. By stabilizing these pathways within EAT, these compounds enhance cardiac function and reduce the risk of arrhythmias, offering a promising therapeutic strategy for AF patients. Metabolic Modulators: Ezetimibe,33 Dehydrocholic Acid,34 and Bongkrek Acid35 primarily impact lipid and energy metabolism in EAT. By reducing lipid accumulation and improving metabolic function, these molecules decrease the pro-fibrotic potential of EAT, supporting healthier cardiac tissue. Enhanced metabolic efficiency is particularly beneficial in managing HFpEF, where metabolic dysregulation plays a significant role. Antioxidants: Dibenzoylmethane36 and Antimycin-a37 exert antioxidant effects that reduce oxidative stress within EAT. By minimizing oxidative damage, these compounds prevent excessive fibrosis and maintain cardiac tissue integrity. Antioxidant therapy is crucial in protecting the heart from the deleterious effects of oxidative stress associated with both AF and HFpEF. Kinase Inhibitors: H-838 and BX-79539 target kinases involved in inflammatory and fibrotic signaling pathways. By modulating these pathways, they reduce fibrosis and enhance myocardial compliance. This mechanism is particularly important in HFpEF management, where reducing myocardial stiffness can significantly improve diastolic function. Neuromodulators: Compounds such as GBR-12935,40 Dipivefrine,41,42 LY-22591043 act as neuromodulators affecting neurotransmitter systems, particularly the sympathetic nervous system. These compounds modulate neural activity, thereby decreasing inflammation and improving cardiac function. Neuromodulation offers a novel approach to managing EAT inflammation and metabolism, crucial for treating AF and HFpEF. Histone Deacetylase Inhibitors: Depudecin44 represents the category of histone deacetylase inhibitors, which influence gene expression by altering chromatin structure. This compound promotes anti-inflammatory pathways and inhibits fibrotic gene activation, reducing fibrosis and improving cardiac outcomes. Modulating gene expression provides a targeted approach to addressing the complex gene regulatory networks involved in EAT pathology. Vasopressin Analogs: Lypressin,45 a vasopressin analog, modulates water balance and vascular tone, impacting EAT function and inflammation. This mechanism can lead to improved cardiac performance and reduced inflammatory responses, offering therapeutic benefits in both AF and HFpEF.

By categorizing the identified small molecules based on their pharmacological mechanisms, our study highlights the diverse approaches available for targeting EAT in AF and HFpEF. These compounds represent promising therapeutic strategies that address specific pathological processes, emphasizing the importance of personalized treatment approaches. Future research should focus on validating these findings through experimental and clinical studies to establish the efficacy and safety of these compounds in treating AF and HFpEF. The integration of bioinformatics analysis and pharmacological research offers a powerful framework for discovering novel therapeutic agents and advancing the treatment of complex cardiovascular diseases.

Owing to the limited sample size in HFpEF, machine learning algorithms were only applied in AF datasets. However, as the two hub genes, both ITPKA and WNT9B passed the internal and external validation and showed superior discriminability for HFpEF and AF, which reminded us the possible common pathogenic genes for these two diseases.

The role of the WNT/β-catenin signaling pathway in cardiac development and disease has been increasingly recognized. Among its constituents, WNT9B is emerging as a significant factor that may influence the development and progression of HFpEF and AF. Both conditions are characterized by structural and functional cardiac alterations, suggesting a possible common pathway involving WNT9B. The study by Wolke et al46 highlights the activation of WNT signaling, including WNT9B, in the context of AF. This activation is associated with atrial remodeling, a hallmark of AF, suggesting that WNT9B may contribute to the fibrotic and structural changes observed in the atria during AF progression. The elevated levels of WNT9B mRNA observed in AF patients underline its potential as a therapeutic target or biomarker for atrial remodeling processes. Paul et al47 explored the role of Wnt signaling in HFpEF, suggesting that WNT9B could play a role in the myocardial stiffness and fibrosis typical of HFpEF through its regulatory effects on cardiac fibroblasts and extracellular matrix composition. Given the complexity of HFpEF pathophysiology, the involvement of WNT9B in these processes provides a novel perspective on disease mechanisms and potential therapeutic interventions. WNT9B may contribute to the fibrotic process by influencing fibroblast proliferation and differentiation, leading to increased extracellular matrix deposition and atrial stiffness. By affecting the extracellular matrix and myocardial stiffness, WNT9B could directly impact diastolic function, a key feature of HFpEF. The regulatory effects of WNT9B on cardiomyocyte proliferation and growth may contribute to the ventricular hypertrophy observed in HFpEF.

ITPKA (Inositol-Trisphosphate 3-Kinase A) plays a crucial role within the inositol phospholipid signaling pathway by modulating the levels of inositol phosphates, which are key secondary messengers in cellular signaling processes.48 ITPKA phosphorylates inositol 1,4,5-trisphosphate (IP3), a product of the phospholipase C (PLC)-mediated hydrolysis of PIP2, converting it into inositol 1,3,4,5-tetrakisphosphate (IP4).49 This conversion plays a critical role in the regulation of intracellular calcium levels and further modulates various cellular processes triggered by calcium signaling. By converting IP3 into IP4, ITPKA indirectly influences calcium signaling pathways.50,51 IP3 is a key molecule that binds to IP3 receptors on the endoplasmic reticulum, prompting the release of calcium ions into the cytosol. By influencing IP3 and calcium signaling, ITPKA indirectly affects the broader network of signaling pathways that interact with PI3K/AKT, potentially impacting cellular responses to growth factors and other extracellular signals.

Based on the role of ITPKA in inositol phospholipid signaling pathway, we speculated two possible mechanisms of ITPKA in the coexist of HFpEF and AF. One is that Calcium signaling, regulated by the inositol phospholipid pathway, is crucial for cardiac muscle contraction and electrophysiological stability. ITPKA, by modulating the levels of inositol phosphates, could influence calcium dynamics, potentially impacting heart rhythm and contractility. This could have implications for both HFpEF, characterized by diastolic dysfunction, and AF, associated with electrical instability in the atria. The other is that the PI3K/AKT pathway. Aberrant activation of this pathway can contribute to pathological cardiac remodeling and fibrosis, key features of HFpEF. Similarly, altered signaling could affect the atrial substrate in AF, promoting the development and maintenance of the arrhythmia.

The positive correlation between ITPKA and CD4 Memory T cell in HFpEF, as well as the negative correlation between ITPKA and CD4 Memory T cell in AF, highlights the distinct biological functions of this molecule in both HFpEF and AF. Although direct studies specifically addressing ITPKA’s role in CD4 T cells are limited, several related findings highlight the importance of inositol kinases in T cell biology. Unutmaz, Derya et at.52 discusses the role of ITPKB in the positive selection of T lymphocytes, with implications for Erk activity modulation, which reminds us the similar functions for ITPKA in regulating T cell receptor signaling and development. The potential mechanism by which ITPKA influences HFpEF and AF through CD4 T cells may involve several interconnected pathways related to inflammation, immune response modulation, and cardiovascular function. Chronic inflammation is a key feature in the development and progression of HFpEF and AF.

CD4+ T cells, through the production of various cytokines, can contribute to the inflammatory milieu associated with these conditions. By affecting CD4+ T cell function, ITPKA might indirectly influence the extent and nature of inflammation in the heart, promoting fibrotic and hypertrophic processes in HFpEF or the atrial remodeling seen in AF.

Calcium signaling is essential for cardiac muscle contraction and electrical conduction. Aberrant calcium signaling in immune cells, influenced by ITPKA, could have systemic effects that alter cardiac function. For instance, dysregulated calcium signaling in CD4+ T cells might affect cytokine production and release, exacerbating inflammatory processes that compromise cardiac function, contributing to HFpEF and promoting the arrhythmogenic substrate for AF.

HFpEF and AF are associated with cardiac remodeling, characterized by changes in myocardial structure and function. Immune cells, including CD4+ T cells, can infiltrate cardiac tissue, driving fibrosis and altering myocardial stiffness — key features of HFpEF. Similarly, in AF, immune cell-driven inflammation can contribute to atrial remodeling, affecting electrical properties and promoting arrhythmogenesis. ITPKA’s role in regulating CD4+ T cell activity could therefore impact the degree of immune cell infiltration and subsequent remodeling in these conditions.

However, there are several limitations in the research. First, all the analysis were based on transcriptomic level, which can not completely represent the protein expression. Second, larger sample size should be applied to validate the hub gene expression in HFpEF and AF. Secondly, wet experiment should be applied to explore the role of ITPKA in HFpEF and AF by regulating CD4+ T cell.

Conclusion

For the first time, this research comprehensively analyzed the EAT-derived secretory proteins bridging HFpEF and AF. It was found that ITPKA and WNT9B could be the two hub genes for the coexist of HFpEF and AF.

Data Sharing Statement

The data supporting this study’s conclusions are available in both the GEO database (https://www.ncbi.nlm.nih.gov/geo/) and the referenced publication. The “Materials and methods” section details all analytical approaches and software tools used.

Ethics Approval and Consent to Participate

Human samples protocols obtained approval from the Institutional Research Ethics Committee at the Changhai Hospital.

Consent for Publication

Written informed consent was obtained from all individuals.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Funding

This study was supported by National Natural Science Foundation of China (grant numbers 82270379 and 81770383) and 2022 Shanghai “Science and Technology Innovation Plan” Biomedical Science and Technology Project (grant number 22S31904300).

Disclosure

The authors declare that they have no competing interests.

References

1. Campbell P, Rutten FH, Lee MM, Hawkins NM, Petrie MC. Heart failure with preserved ejection fraction: everything the clinician needs to know. Lancet. 2024;403(10431):1083–1092. doi:10.1016/S0140-6736(23)02756-3

2. Kotecha D, Lam CSP, Van Veldhuisen DJ, Van Gelder IC, Voors AA, Rienstra M. Heart failure with preserved ejection fraction and atrial fibrillation. J Am Coll Cardiol. 2016;68(20):2217–2228. doi:10.1016/j.jacc.2016.08.048

3. Frustaci A, Chimenti C, Bellocci F, Morgante E, Russo MA, Maseri A. Histological substrate of atrial biopsies in patients with lone atrial fibrillation. Circulation. 1997;96(4):1180–1184. doi:10.1161/01.CIR.96.4.1180

4. Sasaki N, Okumura Y, Watanabe I, et al. Increased levels of inflammatory and extracellular matrix turnover biomarkers persist despite reverse atrial structural remodeling during the first year after atrial fibrillation ablation. J Interv Card Electrophysiol. 2014;39(3):241–249. doi:10.1007/s10840-013-9867-6

5. Shantsila E, Shantsila A, Blann AD, Lip GY. Left ventricular fibrosis in atrial fibrillation. Am j Cardiol. 2013;111(7):996–1001. doi:10.1016/j.amjcard.2012.12.005

6. Manning WJ, Silverman DI, Katz SE, et al. Impaired left atrial mechanical function after cardioversion: relation to the duration of atrial fibrillation. J Am Coll Cardiol. 1994;23(7):1535–1540. doi:10.1016/0735-1097(94)90652-1

7. Zile MR, Baicu CF, Ikonomidis JS, et al. Myocardial stiffness in patients with heart failure and a preserved ejection fraction: contributions of collagen and titin. Circulation. 2015;131(14):1247–1259. doi:10.1161/CIRCULATIONAHA.114.013215

8. Pai RG, Varadarajan P, Tanimoto M. Effect of atrial fibrillation on the dynamics of mitral annular area. J Heart Valve Dis. 2003;12(1):31–37.

9. van den Berg MP, van Gelde IC, van Veldhuisen DJ. Depletion of atrial natriuretic peptide during longstanding atrial fibrillation. Europace. 2004;6(5):433–437. doi:10.1016/j.eupc.2004.04.006

10. Sacks HS, Fain JN. Human epicardial adipose tissue: a review. Am Heart J. 2007;153(6):907–917. doi:10.1016/j.ahj.2007.03.019

11. Nalliah CJ, Bell JR, Raaijmakers AJA, et al. Epicardial adipose tissue accumulation confers atrial conduction abnormality. J Am Coll Cardiol. 2020;76(10):1197–1211. doi:10.1016/j.jacc.2020.07.017

12. Shimokawa H, Ito A, Fukumoto Y, et al. Chronic treatment with interleukin-1 beta induces coronary intimal lesions and vasospastic responses in pigs in vivo. The role of platelet-derived growth factor. J Clin Invest. 1996;97(3):769–776. doi:10.1172/JCI118476

13. Gaborit B, Venteclef N, Ancel P, et al. Human epicardial adipose tissue has a specific transcriptomic signature depending on its anatomical peri-atrial, peri-ventricular, or peri-coronary location. Cardiovasc Res. 2015;108(1):62–73. doi:10.1093/cvr/cvv208

14. Goldberger JJ, Arora R, Green D, et al. Evaluating the atrial myopathy underlying atrial fibrillation: identifying the arrhythmogenic and thrombogenic substrate. Circulation. 2015;132(4):278–291. doi:10.1161/CIRCULATIONAHA.115.016795

15. Kankaanpää M, Lehto HR, Pärkkä JP, et al. Myocardial triglyceride content and epicardial fat mass in human obesity: relationship to left ventricular function and serum free fatty acid levels. J Clin Endocrinol Metab. 2006;91(11):4689–4695. doi:10.1210/jc.2006-0584

16. Naser JA, Lee E, Scott CG, et al. Prevalence and incidence of diastolic dysfunction in atrial fibrillation: clinical implications. Eur Heart J. 2023;44(48):5049–5060. doi:10.1093/eurheartj/ehad592

17. Adamson C, Kondo T, Jhund PS, et al. Dapagliflozin for heart failure according to body mass index: the DELIVER trial. Eur Heart J. 2022;43(41):4406–4417. doi:10.1093/eurheartj/ehac481

18. Koepp KE, Obokata M, Reddy YNV, Olson TP, Borlaug BA. Hemodynamic and functional impact of epicardial adipose tissue in heart failure with preserved ejection fraction. JACC Heart Fail. 2020;8(8):657–666. doi:10.1016/j.jchf.2020.04.016

19. Lund LH, Savarese G, Venkateshvaran A, et al. Eligibility of patients with heart failure with preserved ejection fraction for sacubitril/valsartan according to the PARAGON-HF trial. ESC Heart Fail. 2022;9(1):164–177. doi:10.1002/ehf2.13705

20. Packer M. Derangements in adrenergic-adipokine signalling establish a neurohormonal basis for obesity-related heart failure with a preserved ejection fraction. Eur J Heart Fail. 2018;20(5):873–878. doi:10.1002/ejhf.1167

21. Packer M. Effect of catheter ablation on pre-existing abnormalities of left atrial systolic, diastolic, and neurohormonal functions in patients with chronic heart failure and atrial fibrillation. Eur Heart J. 2019;40(23):1873–1879. doi:10.1093/eurheartj/ehz284

22. Hatem SN, Sanders P. Epicardial adipose tissue and atrial fibrillation. Cardiovasc Res. 2014;102(2):205–213. doi:10.1093/cvr/cvu045

23. Mazurek T, Zhang L, Zalewski A, et al. Human epicardial adipose tissue is a source of inflammatory mediators. Circulation. 2003;108(20):2460–2466. doi:10.1161/01.CIR.0000099542.57313.C5

24. Gaborit B, Dutour A. Looking beyond ectopic fat amount: a SMART method to quantify epicardial adipose tissue density. Eur J Prev Cardiol. 2017;24(6):657–659. doi:10.1177/2047487317689976

25. Barid I, Nawawi S, Sosroseno W. Effects of parachlorophenol and camphorated parachlorophenol on the phagocytic activities of a murine macrophage cell line (RAW264.7). J Endod. 2002;28(12):806–810. doi:10.1097/00004770-200212000-00004

26. Zabucchi G, Menegazzi R, Cramer R, Nardon E, Patriarca P. Mutual influence between eosinophil peroxidase (EPO) and neutrophils: neutrophils reversibly inhibit EPO enzymatic activity and EPO increases neutrophil adhesiveness. Immunology. 1990;69(4):580–587.

27. García-Argáez AN, Ramírez Apan TO, Parra Delgado H, Velázquez G, Martínez-Vázquez M. Anti-inflammatory activity of coumarins from decatropis bicolor on TPA ear mice model. Planta med. 2000;66(03):279–281. doi:10.1055/s-2000-14894

28. Fajardo O, Galeno J, Urbina M, Carreira I, Lima L. Serotonin, serotonin 5-HT1A receptors and dopamine in blood peripheral lymphocytes of major depression patients. Int Immunopharmacol. 2003;3(9):1345–1352. doi:10.1016/S1567-5769(03)00116-4

29. Gupta N, Sahu A, Prabhakar A, et al. Activation of NLRP3 inflammasome complex potentiates venous thrombosis in response to hypoxia. Proc Natl Acad Sci. 2017;114(18):4763–4768. doi:10.1073/pnas.1620458114

30. Lu R, Qu Y, Wang Z, et al. TBK1 pharmacological inhibition mitigates osteoarthritis through attenuating inflammation and cellular senescence in chondrocytes. J Orthop Transl. 2024;47:207–222. doi:10.1016/j.jot.2024.06.001

31. Krjukova J, Holmqvist T, Danis AS, Akerman KE, Kukkonen JP. Phospholipase C activator m −3M3FBS affects Ca 2+ homeostasis independently of phospholipase C activation. Br J Pharmacol. 2004;143(1):3–7. doi:10.1038/sj.bjp.0705911

32. De Pover A, Godfraind T. Influence of 16 beta formylation on Na, K-ATPase inhibition by cardiac glycosides. Naunyn-Schmiedeberg’s Arch Pharmacol