Identification of Circular RNA-MicroRNA-Messenger RNA Regulatory Network in Atrial Fibrillation by Integrated Analysis

Background Circular RNA (circRNA) is a noncoding RNA that forms a closed-loop structure, and its abnormal expression may cause disease. We aimed to find potential network for circRNA-related competitive endogenous RNA (ceRNA) in atrial fibrillation (AF). Methods The circRNA, miRNA, and mRNA expression profiles in the heart tissue from AF patients were retrieved from the Gene Expression Omnibus database and analyzed comprehensively. Differentially expressed circRNAs (DEcircRNAs), differentially expressed miRNAs (DEmiRNAs), and differentially expressed mRNAs (DEmRNAs) were identified, followed by the establishment of DEcircRNA-DEmiRNA-DEmRNA regulatory network. Functional annotation analysis of host gene of DEcircRNAs and DEmRNAs in ceRNA regulatory network was performed. In vitro experiment and electronic validation were used to validate the expression of DEcircRNAs, DEmiRNAs, and DEmRNAs. Results A total of 1611 DEcircRNAs, 51 DEmiRNAs, and 1250 DEmRNAs were identified in AF. The DEcircRNA-DEmiRNA-DEmRNA network contained 62 circRNAs, 14 miRNAs, and 728 mRNAs. Among which, two ceRNA regulatory pairs of hsa-circRNA-100053-hsa-miR-455-5p-TRPV1 and hsa-circRNA-005843-hsa-miR-188-5p-SPON1 were identified. In addition, six miRNA-mRNA regulatory pairs including hsa-miR-34c-5p-INMT, hsa-miR-1253-DDIT4L, hsa-miR-508-5p-SMOC2, hsa-miR-943-ACTA1, hsa-miR-338-3p-WIPI1, and hsa-miR-199a-3p-RAP1GAP2 were also obtained. MTOR was a significantly enriched signaling pathway of host gene of DEcircRNAs. In addition, arrhythmogenic right ventricular cardiomyopathy, dilated cardiomyopathy, and hypertrophic cardiomyopathy were remarkably enriched signaling pathways of DEmRNAs in DEcircRNA-DEmiRNA-DEmRNA regulatory network. The expression validation of hsa-circRNA-402565, hsa-miR-34c-5p, hsa-miR-188-5p, SPON1, DDIT4L, SMOC2, and WIPI1 was consistent with the integrated analysis. Conclusion We speculated that hsa-circRNA-100053-hsa-miR-455-5p-TRPV1 and hsa-circRNA-005843-hsa-miR-188-5p-SPON1 interaction pairs may be involved in AF.


Introduction
Atrial fibrillation (AF) is one of the most common arrhythmias and associated with heart failure [1][2][3][4]. Age, gender, obesity, and heart valve abnormalities are important factors of AF [4][5][6]. AF can also lead to heart failure hospitalization and death [7]. However, current treatment of AF may have adverse reactions [8,9]. The pathogenesis of AF remains unclear. Further study of the underlying mechanisms of AF may provide new treatments for AF [10].
Circular RNAs (circRNAs) (with a covalent closed-loop structure) are considered to be the key to pathogenesis of heart disease, providing a new perspective for the pathogenesis of AF [11]. circRNA plays a crucial role in several pathophysiological processes [4,12]. In heart disease, circRNAs function as the regulator of miRNA levels. circRNAs may be the potential biomarker. Moreover, bioinformatics analysis provides a novel perspective on circRNAs involved in AF and establishes the foundation for future research of the potential roles of circRNAs in AF [13]. miRNAs play a variety of roles in atrial fibrillation, including regulation of electrical remodeling and modulation of structural remodeling of cardiac tissue. Different miRNAs were confirmed to be up-or downregulated in AF patients [14]. Jiang GSM1821359  GSM1821358  GSM1821355  GSM1821354  GSM1673405  GSM1673403  GSM1673402  GSM1673401  GSM1673400  GSM1673399  GSM1673398  GSM1673397  GSM1673396  GSM1673395  GSM1673394  GSM1673393  GSM1673392  GSM1673391  GSM1673390  GSM1673389  GSM1673388  GSM1673387  GSM1673386  GSM1673385  2 BioMed Research International circRNA_100612-has-miR-133b-KCNIP1/JPH2/ADRB1 and has_circRNA_405917/hsa_circRNA_008132/hsa_circRNA_ 104052/hsa_circRNA_101021/hsa_circRNA_101020/hsa_cir-cRNA_102341-has-miR-892b-GJA1 in the heart tissue of AF patients [4]. However, the potential mechanism of AF remains to be studied. In this study, we performed integrated analysis based on GEO datasets to further identify dysregulated cir-cRNAs in AF.

Data Collection.
We obtained the expression profiles of circRNA, miRNA, and mRNA from GEO datasets by searching keywords ("Atrial fibrillation" [All Fields]) AND ("Homo sapiens" [porgn] AND "gse" [Filter]). We selected data according to the following criteria: (1) the selected dataset must be genome-wide circRNA/miRNA/mRNA transcriptome data; (2) these data were obtained from the heart tissues of the patients in the AF group and the normal control (NC) group (without drug stimulation or transfection); (3) standardized or original datasets were considered in this study. One circRNA expression dataset (GSE129409), two miRNA expression datasets (GSE68475 and GSE70887), and one mRNA expression dataset (GSE31821) were selected (Supplementary Table 1).

Identification of Differentially
Expressed circRNAs, miRNAs, and mRNAs. Firstly, the probes corresponding to multiple circRNAs/miRNAs/mRNAs were removed. Only the single probe with the largest average expression was retained in multiple probes corresponding to circRNAs/miR-NAs/mRNAs. After this treatment, qualified circRNAs/miR-NAs/mRNAs were used for further analysis. Then, LIMMA package analysis was used to identify differentially expressed circRNAs and mRNAs. The metaMA package analysis was used to identify differentially expressed miRNAs. P value < 0.05 was the screening criteria for differentially expressed cir-cRNAs, miRNAs, and mRNAs. Detailed data analysis process was performed as previously described [15].

Functional Annotation.
To assess the functional annotations of host gene of DEcircRNAs and DEmRNAs in ceRNA regulatory network, Gene Ontology (GO) classification and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis were conducted based on the online software GeneCodis3. Statistical significance was with the cutoff criteria of P value < 0.05.  Firstly, starBase v3.0 was used to establish DEcircRNA-DEmiRNA regulatory network. Then, miRWalk 3.0 (http://mirwalk.umm.uniheidelberg.de/) was utilized to find the target differentially expressed mRNAs of differentially expressed miRNAs. Finally, the DEcircRNA-DEmiRNA regulatory network was fused with the DEmiRNA-DEmRNA regulatory network to further construct the ceRNA (DEcircRNA-DEmiRNA-DEmRNA) regulatory through Cytoscape (version 3.6.1) software.
2.5. In Vitro Validation. The blood samples from 10 patients with AF and 10 healthy individuals were obtained for quantitative real-time polymerase chain reaction (qRT-PCR) validation. The inclusion criteria of AF patients were as follows: (1) patients were diagnosed according to the 2014 American Heart Association (AHA)/American College of Cardiology (ACC)/American Heart Rhythm Society (HRS) AF guidelines and confirmed by electrocardiogram (ECG) or Holter monitor, (2) the onset of AF in the patients was at least once a month, (3) patients were under the age of 85, (4) patients with normal liver and kidney function, and (5) patients visited the doctor at least 2 times and took medicine regularly. The exclusion criteria of AF patients were as follows: (1) patients with AF with hemodynamic disorders or malignant arrhythmia; (2) patients were 85 years old and above; (3) patients with AF with chronic cardiac insufficiency (grades II-IV); (4) patients with other severe systemic diseases, or liver and kidney failure; and (5) patients with  ischemic heart disease, valvular disease, cardiomyopathy, rheumatic heart disease, primary pulmonary hypertension, and connective tissue disease. There were no statistically significant differences in age, sex, and body mass index (BMI) between AF patients and the normal individuals. We obtained the written informed consent and the approval from the ethics committee of the Second Hospital of Hebei Medical University. Total RNA was isolated with the TRIzol reagent following the manufacturer's protocol. Based on SuperReal PreMix Plus in ABI 7500 Real-Time PCR Detection System, the qRT-PCR reactions were performed. In the 2 -ΔΔCt method, relative circRNA/miRNA/mRNA expression was determined. Human ACTB and GAPDH were used as endogenous controls for mRNA. In addition, human GAPDH and U6 were used as endogenous controls for circRNA and miRNA expression, respectively.
2.6. Electronic Validation. The GSE135445 dataset (involving 15 AF patients and 15 normal controls) was used to validate the expression of identified differentially expressed mRNAs. The result was presented as box plots. Statistical significance was ascribed to P value < 0.05.
3.5. Electronic Validation of Selected mRNAs. In this study, two differentially expressed mRNAs (SMOC2 and WIPI1) and three host genes of circRNAs (MFN2, ZNF880, and LRBA) were randomly selected for validation ( Figure 6). The result showed that SMOC2 and WIPI1 were downregulated in AF, which was consisted with our integrated analysis. The expression of MFN2, ZNF880, and LRBA was upregulated in AF. However, MFN2, ZNF880, and LRBA were not differentially expressed mRNAs in this study. Further study of MFN2, ZNF880, and LRBA in AF is needed.

Discussion
Mitofusin 2 (MFN2) is the host gene of hsa_circRNA_ 100053. MFN2 plays a key role in normal cardiac development [16]. MFN2 could regulate heart failure-related mitophagy by altering the mitochondrial membrane potential [17]. Chen et al. found that deletion of MFN2 leads to a spontaneous lethal dilated cardiomyopathy in mice [18]. In this study, we found that hsa_circRNA_100053 was downregulated in the heart tissue of AF patients. Moreover, downregulated hsa-miR-455-5p and target upregulated transient receptor potential vanilloid 1 (TRPV1) were under the regulation of hsa_circRNA_100053. Huang et al. reported that hsa-miR-455-5p was related to hypoxia-induced cardiomyocytes injury [19]. TRPV1 is a nonselective ion channel that preferentially obtains calcium from painful stimuli. In addition to traditional pain activation of TRPV1, TRPV1 can also be used as a universal sensor for cell damage including hypoxia. Direct activation of TRPV1 has been shown to produce cardioprotective effects on ischemia and reperfusion injury [20]. In addition, blocking TRPV1 limits the long-term preconditioning-induced cardioprotection of laparotomy [20]. An experimental study showed that TRPV1 inhibition blocked ischemic preconditioning-(IPC-) induced myocardial protection [21]. These reports suggested that hsa_cir-cRNA_100053, hsa-miR-455-5p, and TRPV1 may play an important role in heart protection. Our study indicated that the interaction of hsa_circRNA_100053-hsa-miR-455-5p-TRPV1 may be involved in the process of AF.
Recently, there was no report about the association between hsa_circRNA_005843 and AF. Interestingly, we found that hsa_circRNA_005843 was downregulated in the heart tissue of patients with AF. Furthermore, upregulated hsa-miR-188-5p and target downregulated spondin 1 (SPON1) were regulated by hsa_circRNA_005843. hsa-miR-188-5p plays an important regulation role in the reninangiotensin system [22]. It has been demonstrated that hsa-miR-188-5p is involved in murine cardiomyocyte biogenesis [23]. Decreased expression of hsa-miR-188-5p is found in hyperhomocysteinemia cardiomyocytes [24]. SPON1, a member of antiangiogenic family, is a sensitive plasma biomarker for early myocardial injury [25]. This indicated that hsa-miR-188-5p and SPON1 play roles in angiogenesis, which is associated with cardiomyocyte biogenesis. Our result suggested that the interaction between hsa_circRNA_005843, hsa-miR-188-5p, and SPON1 could be associated with AF.
In addition, we found that hsa-circRNA-402565 was upregulated in the heart tissue of AF patients. Interestingly, qRT-PCR validated the expression of hsa-circRNA-402565. hsa-circRNA-402565 is downregulated in patients with ventricular septal defect [40]. It is noted that hsa-circRNA-405811 and hsa-circRNA-103752 were, respectively, the most upregulated and downregulated circRNAs in AF. hsa-miR-99a was the most downregulated miRNA in AF. Zinc finger protein 880 (ZNF880) is the host gene of hsa-circRNA-405811. The inactivating mutation of ZNF880 is found in isolated cardiac myxoma tissue samples [41]. LPS responsive beige-like anchor (LRBA) protein is the host gene of hsa-circRNA-103752. It is reported that LRBA is involved in signal transduction and vesicle trafficking in cardiogenesis [42]. The expression of LRBA is decreased 2 d after myocardial infarction [43]. It is suggested that hsa-miR-99a is involved in cardioprotective in postinfarction left ventricular remodelling [44]. Thus, it can be seen that these circRNAs involve in cardiogenesis and cardioprotection. Our result indicated that hsa-cir-cRNA-402565, hsa-circRNA-405811, hsa-circRNA-103752, and hsa-miR-99a may be associated with the pathology of AF.
According to the functional annotation analysis, we found that mTOR was one of the most enriched signaling pathways of the host genes of DEcircRNAs. MTOR is a protein kinase that acts as an interface to a variety of metabolic pathways and is widely found in many species. MTOR plays  BioMed Research International a vital role in cellular metabolism [45]. Kinases are activated by extracellular growth factor signaling, enhancing cytoplasmic translation processes and protein synthesis. The mTOR pathway is involved in the steady-state process of the heart against stress [46]. The mTOR pathway contributes to the proliferation and survival of cardiomyocytes. In aged mice, knocking out and inhibiting mTOR can prolong survival and inhibit cardiac hypertrophy [47]. This indicated that the mTOR signaling pathway plays an important role in the heart against stress and cardiomyocyte survival, which may be associated with the development of AF. In addition, arrhythmogenic right ventricular cardiomyopathy, dilated cardiomyopathy, and hypertrophic cardiomyopathy were three remarkably enriched signaling pathways of   DEmRNAs in ceRNA regulatory network. Arrhythmogenic right ventricular cardiomyopathy is a rare inherited cardiomyopathy characterized by fibro-fatty replacement of cardiomyocytes [48,49]. Sudden cardiac death and ventricular enlargement are the most common clinical manifestations [50]. Arrhythmogenic right ventricular cardiomyopathy mainly involves the left and right ventricle during disease progression. The enlarged left atrial is associated with the incidence of risk for death in dilated cardiomyopathy patients [51][52][53][54][55][56][57][58][59][60]. In addition, extensive atrial fibrosis is observed at autopsy in dilated cardiomyopathy patients [61]. AF is an arrhythmia often complicating the course of hypertrophic cardiomyopathy. It is found that patients with hypertrophic cardiomyopathy have a higher risk (20%) for AF [62,63]. These reports suggested that arrhythmogenic right ventricular cardiomyopathy, dilated cardiomyopathy, and hypertrophic cardiomyopathy may be involved in the process of AF.

Data Availability
All data are available in the article.

Conflicts of Interest
The authors declare that there is no conflict of interest.