Atrial fibrillation (AF) is the most common arrhythmia managed in clinical practice with 22%–26% estimated lifetime risk, and is associated with substantial morbidity and increased mortality due to the combination of altered hemodynamics, atrioventricular dyssynchrony, progressive atrial and ventricular mechanical dysfunction, and thromboembolic complications (Hindricks et al., 2020; Brundel et al., 2022). The rising burden of AF, together with the significant increase in coexisting chronic cardiovascular conditions, is evidenced by a global uptrend in AF hospitalization (Kornej et al., 2020).

Atrial remodeling is considered as the most important mechanism involved in the development and progression of AF, including structural and electrical remodeling (Chen et al., 2021a). Recent studies show the metabolic changes of the atrium, especially the mitochondrial metabolism, significantly impact the electrical signaling and structural remodeling of the atrium, and play important roles in the occurrence and development of AF (Bai et al., 2019; Mason et al., 2020; Yuan et al., 2020). In addition, various metabolic disorders (i.e., obesity, diabetes mellitus, and hyperthyroidism) have been proved as risk factors of AF (Lavie et al., 2017; Wang et al., 2019; Lin et al., 2021), through the mechanism of mediating myocyte hypertrophy or fibrosis, etc. (Karam et al., 2017; Sagris et al., 2021). Thus, metabolic disorders play an important role in the occurrence and development of AF (Hajhosseiny et al., 2015; Zheng et al., 2021). However, it is unclear what kind of specific alternation of the metabolism profile is associated with the development of AF, indicating further explorations are required.

The omics technology can generate a large amount of multidimensional data that are amenable to be analyzed by new informatics methods, which have recently been applied to explore the molecular changes and identify the mechanism of AF-related remodeling (Jiang et al., 2019; Liu et al., 2020a; Roselli et al., 2020). As important supplements to genomics and transcriptomics, metabolomics and proteomics are more sensitive to external factors and can better reflect the real physiological state of biological systems (McGarrah et al., 2018; Liu et al., 2020b). In addition, the rapidly-developing mass spectrometry (MS)-based omics techniques provide new tools for numerous metabolite or protein level patterns in biological samples (Challen and Cramer, 2022) and offer more valuable data for biomarker screening and pathological exploration in AF, especially by multi-omics at the myocardial tissue level, for which data are limited.

In this study, integrated multi-omics analysis of metabolomics-proteomics of human atrium tissues were performed and analyzed to explore the metabolism reprogramed in AF patients and describe the key dysregulated metabolites and proteins as well as related metabolic pathways in AF patients relative to patients with sinus rhythm (SR).

In this study, multiple targets and pathways related to mitochondrial function were unveiled and enriched by metabolomics and proteomics profiles of hearts from atrial fibrillation patients including oxidative phosphorylation and ATP biosynthetic process, etc., which have vital functions in maintaining the homeostasis of the myocardium (Covian and Balaban, 2012; Mason et al., 2020), providing a comprehensive analysis of proteomics and metabolomics landscape of human AF.

LAA samples were harvested from 10 persistent non-valvular AF patients during surgical ablation of atrial fibrillation and 10 SR samples matched by age and gender from healthy donors. Metabolomic analyses were then carried out using the 10 AF samples and the 10 SR samples. Differential metabolites were identified with the criteria of |logFC| > 0.263 and p < 0.05. There were 113 significantly upregulated metabolites and 10 significantly downregulated metabolites in AF patients compared to the SR group. Especially, the alterations of Raffinose, Adenine and D-Mannitol with FCs of 0.36, 0.44, and 0.22, respectively, are associated with mitochondrial energy metabolism (Figure 2 and Supplementary Table S1).

Moreover, the altered metabolites in AF were enriched in multiple pathways related to mitochondria and glycolipid metabolism. Notably, among the downregulated metabolites in both positively- and negatively-ionized modes in AF patients, pathways including mitochondrial beta-oxidation, fatty acid metabolism and galactose metabolism were enriched (Figure 3). Moreover, pathways associated with various important amino acid metabolisms were enriched, with the upregulated metabolites, such as beta-alanine metabolism, histidine metabolism, and glycine and serine metabolism (enrichment scores 5.58, 4.425 and 1.923 respectively).

Analysis of pathways sharing both significant proteins and metabolites using MetaboAnalyst was then performed. As a result, multiple glycolipid metabolic pathways, such as glycerophospholipid metabolism and glycerolipid metabolism were mostly enriched using the downregulated proteins and metabolites (Figure 4), and various important amino acid metabolic pathways were significantly enriched using the upregulated metabolomics and proteomics profiles (Supplementary Figure S1), consistent with the results of metabolites analysis.

Proteomics samples and metabolomic samples were then sent to correlation analysis using interaction networks of proteins with p < 0.05 and metabolites with VIP >1 and p < 0.05. Correlations with the top 100 r values and p < 0.05 were next constructed, which demonstrated obvious clustering patterns of proteins and metabolites that represented the metabolism changes in AF. Network diagram was drawn on Cytoscape by sorting the absolute values of correlation coefficient (Figure 5B, Supplementary Figure S2 and Supplementary Tables S2, S3). As a result, several hub proteins with metabolites were identified, among which glycerol-3-phosphate dehydrogenase 2 (GPD2) was found at the top of the network string interactions ranked by maximal clique centrality (MCC) method in positively-ionized mode (MCC score = 27; Figures 5A, C), which is located at the inner membrane of mitochondria and closely related to mitochondria energy metabolism (Mráček et al., 2013; Ishihama et al., 2021; Liu et al., 2021). Together with GPD2, SYNM ranked at the top three (MCC score = 17,16, respectively) as well, which have also been reported to be associated with mitochondrial function (Winter et al., 2015; Kuznetsov et al., 2020; Paulin et al., 2020; Baek et al., 2021), indicating its potential participation of mitochondria metabolism in AF.

Proteomics samples were tested by univariate analysis using the criteria for differentially expressed proteins (DEPs): p < 0.05 and logFC >0.263 for upregulation, and p < 0.05 and logFC < −0.263 for downregulation. All 330 DEPs including 225 upregulated DEPs and 105 downregulated DEPs were analyzed, which identified the downregulation of GPD2 (p = 0.02 and FC = 0.77), PLEC (p < 0.001 and FC = 0.71) and SYNM (p = 0.04 and FC = 0.76) in AF patients in AF patients (Figure 6 and Supplementary Table S4).

Pathway enrichment analysis was then performed. For biological processes (BP) by GO analysis, DEPs were enriched in heart development, respiratory electron transport chain and actomyosin structure organization (Figures 7A–C). For cellular components (CC), DEPs were enriched in mitochondrial protein containing complex, organelle inner membrane, and mitochondrial proton transporting ATP synthase complex coupling factor. For molecular functions (MF), DEPs were enriched in proton transmembrane transporter activity, active transmembrane transporter activity and inorganic molecular entity transmembrane transporter activity.

Moreover, oxidative phosphorylation and cardiac muscle contraction were enriched by KEGG (Figure 7D). Ion channel transport, and organelle biogenesis and maintenance were enriched by REACTOME (Figure 7E). Oxidative phosphorylation was enriched with the largest score (NES = −2.47) by WIKI (Figure 7F). GSEA for function enrichment showed several pathways related to mitochondrial energy metabolisms were significantly downregulated in AF samples, including oxidative phosphorylation (p < 0.01), ATP biosynthetic process (p < 0.01) and NADH dehydrogenase complex (p < 0.01) (Figure 8), which were also reportedly associated with the function of GPD2 (Mráček et al., 2013; Ishihama et al., 2021; Liu et al., 2021).

Gene Set Variation Analysis (GSVA) was then carried out to analyze the correlation of enriched gene sets and the expression of GPD2, PLEC, and SYMN (Figure 9). The GPD2 expression level was associated with multiple pathways related to cardiac structural remodeling [e.g., contractile fiber (r = 0.33), cardiac muscle contraction (r = 0.2)], and various pathways relevant to myocardial electrical remodeling, including calcium signaling (r = 0.27) and ion homeostasis (r = 0.7), and metabolic pathways such as lipid metabolism (r = 0.28) and AMP activated protein kinase signaling (r = 0.46). The expression levels of SYNM showed association with pathways related to cardiac remodeling and energy metabolism [e.g., cardiac muscle contraction (r = 0.38), ion homeostasis (r = 0.71), lipid metabolism (r = 0.28)], and PLEC demonstrated association with pathways related to mitochondria function [e.g., oxidative phosphorylation (r = 0.43), ATP biosynthetic process (r = 0.48)].

This study provides new insights into the mechanisms of AF progression, especially the reprogramming mitochondrial metabolism, and identifies multiple genes related to mitochondrial function as novel targets for non-valvular AF, which may participate in and maintain AF.