DNA methylation is well-characterized as a heritable regulation of gene expression. A methyl group is catalyzed by DNA methyltransferases (DNMTs) to shift from the S-adenosyl-L-methionine to the 5' carbon of cytosine which mostly located in cytosine-phosphate-guanine (CpG) islands (9). The gene promoter hypermethylation correlates with transcriptional silencing, whereas hypomethylation leads to increased expression of the gene (10). DNA methylation regulation may serve an important role in AF pathogenesis.

It has been identified the methylation of CpG sites in prevalent, permanent, and paroxysmal AF. Differential methylation of CpG sites were significantly related to prevalent AF (two CpG sites) and incident AF (five other CpGs) by GWAS of the participants' peripheral blood in the Framingham Heart Study (11). The majority of 417 differentially methylated CpG sites discovered in the fibrillating atrium were located in intergenic regions outside of CpG islands (12).

DNMTs dysregulations likely plays an important role in the pathogenesis of AF. Significantly higher level of the whole DNA methylation was found in the AF group than SR group (13). DNMT3b likely contributes to the DNA methylation dysregulations in valvular AF (13). DNMT3A was shown to be involved in Ras association domain family 1 isoform A (RASSF1A)-mediated upregulation of extracellular signal regulated kinases 1/2 (ERK1/2) in cardiac fibrosis (14).

DNA hypermethylation also participated in the association of transcription factor, fibrosis and potassium ion transport with pathophysiology of AF (15, 16). Hypermethylation of paired-like homeodomain 2 (Pitx2) promoter was shown to be associated with AF in humans and aging spontaneously hypertensive rats (SHR) (15). DNA methylation inhibitor 5-Aza-2'-deoxycitidine treatment reduced the left ventricular fibrosis in SHR (13). The CpG hypermethylation of SUR2, a subunit of the ATP-sensitive potassium channel, leads to its silencing expression in the HL-1 atrial cardiomyocyte cell line (16).

Heart failure (HF), a risk factor of AF, may induce Pitx2c promoter hypermethylation in vivo and in vitro (17) Pitx2c promoter methylation increased and Pitx2c protein level decreased with the increase of DNMT 1 in isoproterenol-induced HF atria compared with normal atria (17). The same trend was shown in Angiotensin II (AngII)-treated HL-1 cells compared with control cells (17). The methylation inhibitor 5-AZa-2 '-deoxycytitine and the AngII receptor blocker Losartan attenuated these effects (17). However, isoproterenol did not alter the expression of Pitx2c and DNMT1 (17).

Chromatin is the state in which DNA is wrapped in the cell. The nucleosome, the basic unit of chromatin, is an octamer composed of four core histones (H3, H4, H2A, H2B) surrounded by 147 DNA base pairs. A primary component of chromatin that plays an essential role in this regulation is the modification of histones. The core histones are mainly globular except for their unstructured N-terminal “tails.” A distinctive feature of histones, especially their tails, is the large number and type of modified residues they possess.

The gene expression regulation in euchromatin requires the delivery of chromatin-modifying enzymes by DNA-bound transcription factors (TFs). Following the external cues, TFs bind to the specific genes' promoter and promote the gene expression or silencing. So there will be activation-related and repression-related modification for the purposes of transcription. Histone modifications are involved in affecting gene expression. The mono-methylations of H3K27, H3K9, H4K20, H3K79, and H2BK5 are related to gene activation, whereas trimethylations of H3K27, H3K9, and H3K79 are linked to repression.

More and more evidences show that N-terminal tail of histone is subjected to covalent and reversible post-translational modifications, such as methylation and acetylation (18, 19). These modifications are used to modify chromatin compress or create anchoring sites for other transcriptional regulators (18). The methylation and acetylation status of chromatin is regulated by writers and erasers. The most studied representative of writers and erases are histone methyltransferases (HMT) and histone deacetylases (HDACs) (19).

Emerging evidences reveal a role for histone acetylation or methylation to modulate pathogenic gene expression in AF patients. GWAS of AF patients (n = 60,620) and controls (n = 970,216) indicated active enhancers as indicated by acetylation of histone H3 lysine 27 (H3K27ac) in right atrium, which has demonstrated that AF-associated risk variants fell near the genes acting via structural cardiac remodeling (20). Enhancer of zeste homolog2 (EZH2), a histone-lysine N-methyltransferase enzyme encoded by the EZH2 gene, is participating in histone methylation by binding to H3K27me3 (trimethylation of lysine 27 on histone 3). The expression of EZH2 and H3K27me3 is upregulated in permanent AF patients with atrial fibrosis (21). Compared to the people with sinus rhythm (SR), HDAC3 protein expression and activity levels were increased in paroxysmal AF (PAF), persistent AF (PeAF) (n = 5), and long-standing PeAF (n = 7) (22). The phosphorylated-HDAC5 levels were correlated with significantly increased BNP gene expression in PeAF patients (22). Similarly, the remarkable elevated expression and activity of HDAC6 were correlated with PeAF duration (23). However, whether the overall acetylation level of protein or the total HDAC activity had no dramatically difference between patients with PeAF/PAF and subjects in SR (23).

Furthermore, the mechanisms in epigenetic regulation of AF, especially HDACs inhibition, have been investigated in vivo and in vitro. The use of HDAC6 inhibitor tubastatin A in vivo can protect dogs with atrial tachycardia pacing from electrical remodeling, while the dominant negative HDAC6 mutant can completely rescue the systolic dysfunction induced by tachycardia pacing (23). HDACs may regulate transcriptional reprogramming in AF. The homeo-domain-only protein (hopx) transgenic mice were recruited into chromatin to induce serum response factor (SRF)—dependent transcription and myocardial hypertrophy. They were administered with or without pan HDAC inhibitor trichostatin A (TSA). TSA treated mice have protective effects on atrial arrhythmia and fibrosis induced by rapid pacing (24). Pan-HDAC inhibitor TSA and class I HDAC inhibitor MPT0E014 reduce the onset of AF by reducing calcium spark through normalize the expression of NCX1 and ryanodine receptors in rabbit pulmonary vein cardiomyocytes (25). Hopx transgenic mice with atrial remodeling and dogs with atrial tachypacing treated with HDAC inhibitor (CI-994) showed no significant effects on cardiac function but decrease of the total time of fibrillation and atrial fibrosis in atrial tachypacing-induced sustained AF (26).

So various HDACs inhibitors are emerging as interesting druggable targets for AF (27, 28). RGFP966, a specific inhibitor of HDAC3, could prevent systolic dysfunction in cardiomyocyte for AF (29). ACY-1215 (ricolinostat), one of HDAC6 inhibitors, will be a potential candidate drug for trials in patients with AF. Because it is undergoing Phase I and II clinical trials for the treatment of multiple myeloma at present with no reports of serious side effects so far (30). Valproic acid (VPA), an HDAC class I/IIA inhibitor, alleviated atrial remodeling in transgenic mice, animal AF models, and human AF (31). And the researchers used ChIP to identify 9 VPA-downregulated genes (Atp5l, Ces1d, Myl7, Ndufa12, Ndufa8, Ndufs7, Pdha1, Tnni3, Uqcr10) (31).

Most AF-associated GWAS variants reside in the non-coding genome. The non-coding RNAs (ncRNAs) can be classified as short-chain ncRNAs (<200 nucleotides) and long-ncRNA (lncRNAs, >200 nucleotides) (32, 33). The short-chain ncRNAs include microRNAs (miRNAs), transfer RNAs, small nuclearRNAs, small nucleolar RNAs, piwi-interacting RNAs, telomerase RNAs, and other endogenous RNA species (32). CircularRNAs (circRNAs) emerge as novel non-coding RNAs that differ from traditional linear RNAs (33).

The transition from paroxysmal to permanent AF is characterized by a pattern of dysregulated miRNAs, which involve in atrial remodeling or fibrosis, electrical remodeling, calcium signaling, and dysregulation of transcription factors (TFs) (34–42). Mir-133, mir-590, mir-21, miR30, and mir-29b regulated the genes related to arial fibrosis or fibroblast remodeling in atrial fibrillation (34–37). There was upregulation of transforming growth factor (TGF) TGF-β1 and TGFβRII at the protein level and downregulation of miR-133 and miR-590 in the levels of miRNAs in nicotine stimulated atrial fibrosis of AF in dog (34). The angiotensin II (AngII)-induced upregulation of miR-21 and repression of Spry1 was showed in neonatal cardiac fibroblasts (35). Connective tissue growth factor (CTGF) is a secreted protein as a powerful inducer of extracellular matrix (ECM) synthesis. Its levels are substantially increased in hypertension-induced left ventricular hypertrophy (LVH) and transverse aortic constriction (TAC)-induced LVH rodent models (36). The expression of miR29b decreased and the expression of miR29b ECM target-genes collagen-1A1 (COL1A1), collagen-3A1 (COL3A1) increased significantly in congestive heart failure (CHF) atrial fibroblasts (37, 38). The regulation of miR-1, miR-26, miR-208a, miR-328, and miR-499 on genes associated with electrical remodeling promotes reentry circuits by shortening of the action potential duration and effective refractory period (37–40). The increased inward-rectifier K+ current (I_K1), along with increased expression of the principal underlying subunit KCNJ2 mRNA and its encoded K_ir2.1 protein, has been demonstrate to exhibit a pro-AF-related atrial electrical remodeling. Girmatsion et al. reported that miR-1 levels decreased significantly in human AF and led to increased I_K1 possibly by up-regulating of KCNJ2 (39), miR-26 was found to be downregulated accompanied by upregulation of I_K1/K_ir2.1 protein in atrial samples of AF patients (40). The slowly activating delayed rectifier potassium currents (I_Ks) and atrial effecitve refractory period (AERP) shortening play an important role in the electronic remodeling of AF. The expression of miR-1 was upregulated to induce AERP shortening by targeting I_K1 channel genes (KCNE1, KCNB2) in right atrial tachypacing in New Zealand white rabbits (41). MicroRNAs have also been shown to control the expression of genes encoding important Ca²⁺ processing and signaling proteins in AF (42). Studies have shown that the atrial level of ryanodine receptor type-2 (RyR2) protein is elevated in paroxysmal AF (pAF) patients, suggesting that post-transcriptional regulation of RyR2 might contribute to the pathogenesis of AF (43). Members of the miR-106b-25 cluster, such as miR-106b and miR-93, could suppress the translation of RyR2 by binding to its 3'-untransted region (UTR) (43). The downregulation of the miR-106b-25 cluster and upregulation of RyR2-mediated sarcoplasmic reticulum Ca²⁺ leak were demonstratedin atria of pAF patients (44). The miR-328 level was elevated by more than 3-fold in AF patients and AF dogs compared to non-AF subjects. Overexpression of miR-328 diminished the L-type Ca²⁺ currents by targeting CACNA1C and CACNB1, which encode its α1c and β1subunits (44). In addtion, miRNA loss-of-function could develop sinoatrial node dysfunction to increase susceptibility of AF (45). miR-17-92 and miR-106b-25 directly supress the genes required for sinoatrial node function, such as Shox2 and Tbx3 (45). Both miR-17-92 and miR-106b-25 inactivation exhibited pacing-induced AF in mice (45).

Several studies have shown that lncRNAs are participanting in basic mechanism of AF development, including electrical remodeling, atrial remodeling, and metabolic remodeling (46). IncRNA TCONS-00106987 was confirmed to promote electrical remodeling of increasing inward-rectifier K⁺ current (I_K1) through endogenous competition with microRNA-26 (miR-26) by luciferase reporter assays and whole-cell patch-clamp recording in AF rabbit model (47). Silencing of TCONS_00075467 can shorten the atrial effective refractory period in vivo and reduce duration of the L-type calcium current and action potential in vitro, which may also play important roles in electrical remodeling regulation during AF (48). The expression of lncRNAs nuclear-enriched abundant transcript 1 (NEAT1) in atrial tissue of AF patients was up-regulated, and NEAT1 knockdown could improve Ang II-induced atrial fibrosis via the miR-320-NPAS2 axis in mice (49). Recent evidence suggested that lncRNAs in epicardial adipose tissue (EAT) may modulate atrial remodeling (50, 51). Seventeen upregulated lncRNAs and 40 downregulated lncRNAs were differentially expressed in EAT samples collected from persistent non-valvular AF and sinus rhythm (SR) (P < 0.05; fold change>1.5) (50). These differentially expressed lncRNAs were mainly related with stress response and metabolic remodeling, which indicating pathogenesis of AF (50). The biological function predictions for the RNA sequencing data collecting from atrial EAT samples of AF and SR revealed that TNF signaling pathway was the most frequent pathway that the lncRNAs might involve in (51).

Recently, more attention has been paid on circRNAs and their association with miRNA and lncRNAs in AF intiation and perpetuation. The differentially expression of circRNAs (Table 1) were verified between persistent AF patients and SR people (52). One hundred and forty-six different circRNAs were found between control and rapid atrial pacing (RAP) dogs (53). And analysis showed that the differentially expressed circRNAs may be involved in the process of “cytoskeleton structural composition and ion channel activity” as well as extensive interaction among different circRNAs and AF related miRNAs and mRNAs (53–55). Integrated analysis speculated that circRNA-microRNA interaction pairs and intricate cross-talk may be involved in AF (Table 1) (56, 57).

Circular RNAs show higher stability than other RNAs. They also exhibit more functional patterns, such as sponging microRNAs (58). The sponging activities of the circRNAs could be responsible for the down-regulation of specific miRNAs in establishment of a permanent AF condition (59).

In addtion, the dysregulated circRNAs may be enriched in participation of inflammatory response in AF (60, 61). 250 up- and 126 down-regulated circRNAs were differentially expressed between AF subjects and healthy donors. It should be noted that the enrichment analysis identified five circRNAs showing the highest significance. Among them, four were enriched in cytokine-cytokine receptor interaction (60). A well-recognized participant for atrial fibrosis association with AF was TGF-beta signaling pathway (61). Differential expression of 14,215 circRNAs were detected in AF patients and healthy controls. Among them, hsa_circ_0000075 and hsa_circ_0082096 was exhibited to be invovled in TGF-beta signaling pathway of the AF pathogenesis (61).

All these researches make ncRNAs promising candidates for the development of latent diagnostic or prognostic biomarkers and even therapeutic targets for AF.