Analysis of changes in circular RNA expression and construction of ceRNA networks in human dilated cardiomyopathy

Abstract Dilated cardiomyopathy (DCM) is a severe life‐threatening disease worldwide, and the underlying mechanisms remain unclear. Circular RNAs (circRNAs) have been reported to play important roles in various cardiovascular diseases and can function as competitive endogenous RNAs (ceRNAs). However, their role in human DCM has not been fully elucidated. In the present study, heart samples from DCM patients and healthy controls were used to identify circRNAs by RNA sequencing. Real‐time quantitative reverse transcription‐polymerase chain reaction (qRT‐PCR) was conducted to validate differentially expressed circRNAs and mRNAs. A total of 9585 circRNAs and 22050 mRNAs were detected in the two groups. Overall, 213 circRNAs and 617 mRNAs were significantly up‐regulated in the DCM group compared with the control group. Similarly, 85 circRNAs and 1125 mRNAs were significantly down‐regulated. According to the ceRNA theory, circRNAs can indirectly interact with mRNAs by directly binding to microRNAs (miRNAs), and circRNAs and mRNAs should be concurrently either up‐regulated or down‐regulated. Based on this theory, we constructed two circRNA‐miRNA‐mRNA networks by using the RNA sequencing data and prediction by proprietary software. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed to probe the potential functions of differentially expressed circRNAs. In conclusion, this study revealed that the expression of cardiac circRNAs was altered in human DCM and explored the potential functions of circRNAs by constructing ceRNA networks. These findings provide a foundation for future studies of circRNAs in DCM.


| INTRODUC TI ON
Dilated cardiomyopathy (DCM), characterized by ventricular dilation and systolic dysfunction, is a common disease that results in arrhythmia, heart failure and sudden cardiac death. Many factors have been reported to be closely related to DCM, including coronary artery disease, endocrine and metabolic abnormalities, viral infections, autoimmunity and gene mutations. 1 However, the exact mechanisms involved in the development of DCM have not yet been elucidated.
At present, there is no effective treatment for DCM, and the mortality rate is increasing year by year. Therefore, studies focused on finding novel therapeutic targets for DCM are urgently required.
With the rapid development of sequencing technologies, an increasing number of non-coding RNAs (ncRNAs), such as microRNA (miRNA), long non-coding RNA (lncRNA) and circular RNA (circRNA), have been found. Although messenger RNAs (mRNAs) have been studied extensively, traditional protein-coding RNAs account for only a minority of all RNAs, and ncRNAs actually account for most of the transcriptome. 2 In the past, ncRNAs were considered evolutionary junk, but increasing studies have indicated that these non-coding transcripts play important roles via epigenetic, post-transcriptional and translational mechanisms and have considerable impacts on biological processes. [3][4][5] Therefore, ncRNAs have been assessed as potential diagnostic candidates and therapeutic targets for various diseases, including cardiovascular diseases. 6,7 CircRNAs, unlike linear RNAs, which contain 5′ and 3′ ends, have covalently linked ends that form a closed continuous loop. Due to their circular structure, circRNAs are resistant to RNase R activity and are more stable than other RNAs. 8 Currently, circRNAs can be categorized as exonic circRNAs (ecircRNAs), circular intronic RNAs (ciRNAs) and exon-intron circRNAs (EIciRNAs) based on the back-splicing mechanism and the rearrangement of exons and/or introns in precursor messenger RNA (pre-mRNA). In addition, circRNAs can also be classified into five types, 'exonic', 'intronic', 'intergenic', 'antisense' and 'sense overlapping' according to their location relationship with adjacent coding RNA. [9][10][11] Emerging evidence has shown that the vast majority of circRNAs are derived from exons and are primarily localized in the cytoplasm, while only a small portion of circRNAs, particularly ciRNAs, reside in the cell nucleus. 9,12 Based on their subcellular localization, cir-cRNAs play important roles in regulating gene expression at the transcriptional, post-transcriptional, translational and post-translational levels. [13][14][15] Recent studies have revealed many functions of circRNAs, among which their function as competing endogenous RNAs (ceRNAs) has become a research hotspot. As ceRNAs, circRNAs contain shared miRNA response elements and can competitively bind to miRNAs, resulting in a reduction of miRNAs and an up-regulation of the expression of miRNA target genes. As circRNAs can 'absorb' miRNAs like a sponge, they are often referred to as miRNA sponges. 16 Studies have revealed that circRNAs are abundant in human tissues, including the heart, and most are tissue-specific. 17,18 Interestingly, a number of cir-cRNAs are generated from genes, such as TTN and RYR2, which are associated with cardiovascular diseases. 19,20 In addition, circRNAs are differentially expressed between healthy and diseased human hearts and peripheral blood, suggesting that they may play important roles in cardiac physiology and the initiation and development of cardiovascular diseases. 21,22 Recently, studies have identified the microarray profile of ln-cRNAs and miRNAs in human DCM and constructed a lncRNA-miR-NA-mRNA network. 23,24 As the changes in circRNA expression and the potential circRNA-miRNA-mRNA network remain unclear in human DCM, the present study evaluated circRNA expression in heart samples from DCM patients and healthy controls and constructed two ceRNA networks based on the ceRNA theory. The results provide a new understanding of the mechanisms involved in the development of DCM. Furthermore, the circRNA-miRNA-mRNA network indicated that circRNAs may become potential therapeutic targets for DCM.

| Samples and RNA isolation
DCM heart samples were collected from the left ventricular wall of explanted hearts of patients diagnosed with DCM (clinical data are presented in Appendix S1). Control heart samples were collected from healthy donors (accident victims). Total RNA was isolated using TRIzol and purified with the RNeasy mini kit (Qiagen) according to the manufacturer's instructions. RNA quantity and quality were measured by a NanoDrop ND-1000 instrument (Thermo Fisher Scientific). RNA integrity was determined by gel electrophoresis (Appendix S2). The study protocol was approved by the Medical Ethics Committee of Zhongshan Hospital of Fudan University, and informed consent forms were signed by the subjects recruited in the study or by their immediate family members.

| RNA high-throughput sequencing
The removal of rRNAs from total RNA was performed using the

| CircRNA and mRNA sequencing analysis
Paired-end reads were obtained from the HiSeq 4000 system and were quality controlled by Q30. Next, 3′ adaptor trimming and low-quality reads removal were performed using Cutadapt software (V1.9.3). The high-quality trimmed reads were used to analyse cir-cRNAs and mRNAs. The high-quality clean reads of circRNAs were mapped to the reference genome using STAR software (v2.5.1b).

| Identification of differentially expressed circRNAs and mRNAs
Edger software (v3.16.5) was used to normalize the data and perform differentially expressed circRNA analysis. Cuffdiff software (v2.2.1, part of Cufflinks) was used to obtain the fragments per kilobase of exon per million (FPKM) for the expression profiles of mRNAs, and fold change and P value were calculated based on the FPKM values. CircRNAs and mRNAs that exhibited fold change ≥2 or ≤0.5 with P value < .05 were considered significantly differentially expressed.

| Construction of ceRNA networks
CircRNA-miRNA interactions were predicted using miRNA target gene prediction software. The identification of miRNAbinding sites and target mRNA prediction were performed using proprietary software based on miRanda and TargetScan. The circRNA-miRNA-mRNA network was constructed based on the ggalluvial package in R.

| Statistical analysis
The data are expressed as the mean ± SEM and analysed with SPSS 21.0 (SPSS Inc). Student's t test was used to determine the statistical significance. P values < .05 were considered statistically significant.  Figure 1B). The majority of the circRNAs in the two groups ranged in length from 80 to 2500 bp ( Figure 1C).

| Distribution profiles of circRNAs
The 9585 circRNAs were distributed across all chromosomes ( Figure 1D).

| Identification of differentially expressed circRNAs
To assess the differentially expressed circRNAs, the criteria were set as a fold change of ≥2 or ≤0.5 and a P value of <.05. Compared with circRNAs in the control group, 298 dysregulated circRNAs were identified in patients with DCM, of which 231 were up-regulated and 85 were down-regulated ( Figure 2A). A Manhattan plot presented the circRNAs with fold change ≥2 or ≤0.5 between two groups. We found that differentially expressed circRNAs (with a P value of <.05) were scattered throughout all chromosomes ( Figure 2B). In addition, a scatter plot and a volcano plot were generated to identify differentially expressed circRNAs between the control and DCM groups ( Figure 2C,D). A heat map of differentially expressed circRNAs is presented in Figure 2E.

| Target miRNAs of differentially expressed circRNAs
Next, we predicted the target miRNAs of differentially expressed circRNAs (top 10 up-regulated and top 10 down-regulated cir-cRNAs). All the predicted miRNAs are presented in Appendix S5.

| Identification of differentially expressed mRNAs and validation of mRNA expression
RNA sequencing detected a total of 22050 mRNAs in the control and DCM groups. According to the criteria of a P value of <.05 and a fold change of ≥2 or ≤0.5, 1742 mRNAs were differentially expressed in the DCM group compared with the control group, of which 617 were up-regulated and 1125 were down-regulated ( Figure 5A). A scatter plot was generated to identify differentially expressed circRNAs between the control and DCM groups ( Figure 5B). Next, we predicted the target mRNAs of the selected miRNAs shown in Figure 4A,B and compared the predictions with our mRNA sequencing data. Fifty-four target mRNAs of the 13 selected miRNAs were up-regulated, and 66 target mRNAs of the 6 selected miRNAs were down-regulated ( Figure 5C). We selected three mRNAs from the 54 up-regulated mRNAs and three mRNAs from the 66 down-regulated mRNAs to verify their expression levels. NPR3 ( Figure 5D), CFL2 ( Figure 5E) and MLIP ( Figure 5F)   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21    analysis ( Figure 6C). The down-regulated mRNAs were found to be mostly enriched in the following terms: regulation of multicellular organismal process and regulation of signal transduction in the 'biological process' analysis ( Figure 7A); cytoplasm, nucleus and organelle in the 'cellular component' analysis ( Figure 7B); and protein binding in the 'molecular function' analysis ( Figure 7C).
KEGG pathway analysis revealed that the up-regulated mRNAs were mainly associated with Ras signalling pathway, regulation of actin cytoskeleton and cAMP signalling pathway ( Figure 6D), and the down-regulated mRNAs were mainly associated with MAPK signalling pathway, phospholipase D signalling pathway and Rap1 signalling pathway ( Figure 7D).

| Prediction of ceRNA networks
According to the ceRNA theory, circRNAs and mRNAs that share the same miRNA-binding sites can function as ceRNAs and should be both either up-regulated or down-regulated, so we constructed two ceRNA networks using dysregulated circRNAs and mRNAs. As shown in Figure 8A, a down-regulated circRNA was less effective at absorbing target miRNAs, resulting in the down-regulation of miRNA-targeted mRNAs. As shown in Figure 8B, an up-regulated circRNA could more effectively absorb target miRNAs, resulting in the up-regulation of miRNA-targeted mRNAs. Therefore, when circRNAs and mRNAs act as ceRNAs, both are up-regulated or down-regulated.

A B
the genesis and development of DCM. 26 Over the past few years, many researchers have made great efforts to explore the molecular mechanisms of DCM, and ncRNAs have been found to play important roles in the development of DCM. 23,27 Although circRNAs are extensively studied in animal disease models, their functions in humans remain to be elucidated.
In the present study, we performed an analysis of dysregulated cardiac circRNAs and mRNAs between patients with DCM and   was reported to act as an endogenous miRNA-223 sponge to inhibit the activity of miR-233 and thus inhibit cardiac hypertrophy and heart failure. 35 CircRNA CDR1as was reported to increase cardiac infarct size by regulating the expression of miR-7 target genes. 36 CircRNAs were also found to be associated with cardiac fibrosis, atherosclerosis, arrhythmia and so on. 21 In addition, com-

ACK N OWLED G EM ENTS
This work was supported by the National Natural Science Foundation of China (grant number 81873538).

CO N FLI C T O F I NTE R E S T S
The authors declare that there are no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
All sequence data discussed in this study are accessible through GEO Series accession number GSE162505.