Screening and functional prediction of differentially expressed circRNAs in proliferative human aortic smooth muscle cells.

Abstract Vascular smooth muscle cell (VSMC) proliferation is the pathological base of vascular remodelling diseases. Circular RNAs (circRNAs) are important regulators involved in various biological processes. However, the function of circRNAs in VSMC proliferation regulation remains largely unknown. This study was conducted to identify the key differentially expressed circRNAs (DEcircRNAs) and predict their functions in human aortic smooth muscle cell (HASMC) proliferation. To achieve this, DEcircRNAs between proliferative and quiescent HASMCs were detected using a microarray, followed by quantitative real‐time RT‐PCR validation. A DEcircRNA‐miRNA‐DEmRNA network was constructed, and functional annotation was performed using Gene Ontology (GO) and KEGG pathway analysis. The function of hsa_circ_0002579 in HASMC proliferation was analysed by Western blot. The functional annotation of the DEcircRNA‐miRNA‐DEmRNA network indicated that the four DEcircRNAs might play roles in the TGF‐β receptor signalling pathway, Ras signalling pathway, AMPK signalling pathway and Wnt signalling pathway. Twenty‐seven DEcircRNAs with coding potential were screened. Hsa_circ_0002579 might be a pro‐proliferation factor of HASMC. Overall, our study identified the key DEcircRNAs between proliferative and quiescent HASMCs, which might provide new important clues for exploring the functions of circRNAs in vascular remodelling diseases.

and 3′ tails, which make them resistant to degradation by RNA exonuclease and much more stable than linear RNAs. 6 Recently, studies suggested that circRNAs might be involved in various biological processes via different functional models, including binding microRNAs (miRNAs) as competitive endogenous RNAs (ceRNAs), 7,8 interacting with RNA binding proteins (RBPs) 9,10 or coding proteins/ peptides. [11][12][13][14] However, little is known about the roles of circRNAs in vascular biology. The first study showed that circular antisense non-coding RNA in the INK4 locus (cANRIL) influences the polycomb group (PcG)-mediated repression of the human INK4a/ARF locus, which is associated with atherosclerosis risk. 15 CircACTA2 competitively binds miR-548f-5p, which suppresses the expression of α-SMA in VSMCs, and is involved in hypertension. 16 Our group demonstrated that circ-Sirt1 controls NF-κB activation via a sequence-specific interaction with p65 and enhancement of SIRT1 expression by sponging miR-132/212 in the inflammatory phenotypic transformation of VSMCs. 17 Liu et al 18 found that the expression of circRNA-ZNF609, a sponge for miR-615-5p, has a negative correlation with the velocity of vascular endothelial cell migration and tube formation and is associated with vascular endothelial dysfunction. Besides, other evidence has indicated that the aberrant expression of circRNAs usually accompanies by vascular dysfunction. 19,20 This study aimed to screen key differentially expressed circRNAs

| RNA labelling and array hybridization
Sample labelling and array hybridization were performed according to the Agilent One-Color Microarray-Based Gene Expression Analysis protocol (Agilent Technologies) with minor modifications.
Briefly, total RNA was digested with RNase R (Epicentre) to remove linear RNAs and enrich circular RNAs. Then, each sample was amplified and transcribed into fluorescent cRNAs along the entire length of the transcripts without 3′ bias utilizing a random priming method.
The labelled cRNAs were purified by the RNeasy Mini Kit (QIAGEN).
The concentration and specific activity of the labelled cRNAs (pmol Cy3/μg cRNA) were measured by a NanoDrop ND-1000 (Thermo Fisher Scientific). Each labelled cRNA (1 μg) was fragmented by adding 5 μL 10 × blocking agent and 1 μL 25 × fragmentation buffer before being heated at 60°C for 30 minutes, followed by the addition of 25 μL 2 × GE Hybridization buffer. Hybridization solution (50 μL) was dispensed into the gasket slide and assembled on the circRNA expression microarray (8x15K, Arraystar) slide. The slides were incubated for 17 hours at 65°C in an Agilent Hybridization Oven. The hybridized arrays were washed, fixed and scanned using the Agilent DNA Microarray Scanner (G2505C). The analysis was conducted by Kangchen Bio-tech.
Agilent Feature Extraction software (version 11.0.1.1) was used to analyse the acquired array images. Quantile normalization and subsequent data processing were performed using the GeneSpring GX v11.5.1 software package (Agilent Technologies). After quantile normalization of the raw data, circRNAs with at least one out of two samples having flags in Present or Marginal ('All Targets Value') were chosen for further data analysis. DEcircRNAs were identified through fold change filtering. Heat map and hierarchical clustering were performed using Agilent GeneSpring GX software (version 11.5.1).

| Quantitative real-time PCR validation
Total RNA was isolated from proliferative and quiescent HASMCs using TRIzol reagent (Thermo Fisher Scientific), and its quantity and quality were examined by a NanoDrop ND-1000 (Thermo Fisher Scientific) and 1% agarose gel electrophoresis. Then, total RNA was reverse-transcribed using reverse transcriptase with random primers according to the manufacturer's instructions of the M-MLV First Strand Kit (TaKaRa). The expression of 13 DEcircRNAs was tested by qRT-PCR using SYBR Green assays of SuperReal PreMix Plus (TaKaRa), and the primers used for validating up-or down-regulated circRNAs are displayed in Table S1. The qRT-PCR conditions were as follows: a denaturation step of 3 minutes at 95°C, followed by 40 cycles of 30 seconds at 93°C, 30 seconds at 55°C (adjusted with the Tm of different circRNAs), 20 seconds at 72°C and a final step of 5 minutes at 72°C. All samples in this study were normalized to the internal control β-actin. The comparative CT (2 −ΔΔCT ) method was used to calculate the fold change of circRNA expression levels, and Student's t test was used to test its statistical significance. All primer pairs for the detection of circRNAs were designed against the circRNA-specific back-splice sites (Table S1).

| Gene ontology and pathway analyses
Gene ontology and pathway analyses were performed using the standard enrichment computation method, which was used to determine the potential roles of DEcircRNAs in cellular activities.
Based on the online software GeneCodis (http://genec odis.cnb. csic.es/analysis), GO classification and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were conducted for the target genes that might be regulated by these five DEcircRNAs as miRNA sponges in the proliferative and quiescent HASMCs. Gene ontology and pathway analyses are effective methods to uncover the underlying biological function in response to abnormally expressed genes and proteins. 23 This analysis was used to determine the biological pathways that had significant enrichment of differentially expressed targeted genes.

| SiRNA transfection
SiRNA-hsa_circ_0002579 and siRNA-control were designed and purchased from GenePharma, China. The sequence of siRNA-hsa_ circ_0002579 was as follows: 5′-UCCUGUUCAUGGGCUCCAUTT-3′ and 5′-AUGGAGCCCAUGAACAGGATT-3′ and that of siRNA-control was as follows: 5′-UUCUCCGAACGUGUCACGUTT-3′ and 5′-ACGUGACACGUUCGGAGAATT-3′. SiRNA-hsa_circ_0002579 or siRNA-control was diluted by DEPC water to form a solution of 20 μM for 30 minutes at 4°C. According to the protocol of the HiperFect Transfection Kit (QIAGEN), HiperFect transfection reagent or siRNA was diluted in a serum-free medium for 5 minutes at room temperature and then HiperFect transfection reagent and siRNA were equal-ratio mixed for 30 minutes at room temperature to form a polyplex. The polyplex was finally adjusted to a proper volume by the same medium and then added into HASMCs. Following 6 hours of transfection in a humidified incubator at 37°C containing an atmosphere of 95% air, 5% CO 2 , each well containing HASMCs was replaced with the same volume of fresh complete SMCM.

| Western blot
Human aortic smooth muscle cells after siRNA-treated for 48 hours were harvested, and total protein was extracted using RIPA buffer

| Statistical analysis
All data were analysed using SPSS 21.0 software (IBM). The mean ± SD and independent-samples t test were used in the statistical analysis. A value of P < .05 was considered significant.

| Analysis of circRNA expression profiling
The microarray was applied to detect circRNAs in proliferative and quiescent HASMCs. For the distribution of the datasets, there were no distinct differences among the samples ( Figure 1A). The circR- were classified into three types: 84% were exonic, 11% were intronic and 5% were intragenic ( Figure 1E). What is more, we also used circbank 24 to analyse the conservation of the circRNAs of 134 DEcircRNAs and obtained 41 DEcircRNAs with conservation in humans and mice (Table S2).

| Construction of the DEcircRNA-miRNA-DEmRNA crosstalk network
As many as 58 062 circRNAs from the circBase database were found to have AGO-bound regions, 25 indicating that circRNAs might function as miRNA sponges. Among the 134 DEcircRNAs, 77 DEcircRNAs were predicted to bind with AGO ( Figure 3A,

| Functional annotation of DEmRNAs targeted by DEcircRNAs
Gene ontology and KEGG pathway analyses were employed to annotate the functions of DEmRNAs that shared the same MREs with five AGO-bound DEcircRNAs in proliferative and quiescent HASMCs ( Figure 3C).

| Coding potential of DEcircRNAs
Through the circRNADb, 26 we screened 27 DEcircRNAs with internal ribosomal entry sites (IRES) and open reading frames (ORF), suggesting that they might code proteins (Table 1).

| Comparison of expression levels between DEcircRNAs and corresponding linear mRNAs
The main two categories of DEcircRNAs were exonic and intronic.
No matter in which portion, there were less than half of the corresponding mRNAs whose expression was changing along with the differential expression of the circRNAs. In the part of exonic circRNAs, there were 62 circRNAs irrelated with their host genes, 26 co-expression (co-up or co-down) and 23 inverse-expression (circRNAs downward while corresponding mRNAs upward or cir-cRNAs upward while corresponding mRNAs downward) (Figure 4).
Only a small moiety of the DEcircRNAs derived from introns and half of them have nothing to do with the corresponding mRNAs ( Figure 4).

| Knockdown of hsa_circ_0002579 inhibited the proliferation of HASMCs
From the data obtained by microarray, we found that hsa_ circ_0002579 was up-regulated in the proliferative HASMCs.
Gene ontology and pathway analyses showed that 35 DEmRNAs co-expressed with hsa_circ_0002579 ( Figure 3B) were enriched in the TGF-β receptor signalling pathway, Ras signalling pathway and AMPK signalling pathway ( Figure 3C). Interestingly, high mobility group AT-hook 2 (HMGA2) is one of the DEmRNAs that coexpressed with hsa_circ_0002579, 30 shared miRNAs were found between hsa_circ_0002579 and HMGA2 3′-UTR ( Figure 5A), indicating that hsa_circ_0002579 might target HMGA2 as ceRNA.
SiRNA for hsa_circ_0002579 was designed, which could decrease 65% hsa_circ_0002579 levels (data not shown). Compared to control, knockdown of hsa_circ_0002579 reduced the expression level of HMGA2 and PCNA and increased the expression level of SM22a ( Figure 5B).
Taken together, these data showed that knockdown of hsa_ circ_0002579 could inhibit the proliferation of HASMCs. There was a hint that hsa_circ_0002579 could be a promising ceRNA, and HMGA2 was one of its downstream targets.

| D ISCUSS I ON
CircRNAs, mainly produced by precursor mRNA backsplicing of exons, have recently been predicted as a novel class of gene expression potential regulatory factors. CircRNAs often show celltype-specific, tissue-specific and spatiotemporal-specific patterns, including the cardiovascular system, 27 and multitudinous studies suggested that circRNAs are associated with CVD. 15,16,[18][19][20] However, direct evidence about circRNAs in VSMCs in terms of proliferation regulation is still absent. Since the ceRNA hypothesis was proposed, 28 emerging evidence has indicated that circRNAs could act as miRNA sponges to regulate the stability or translation of mRNAs. In this study, we aimed to explore the roles of circRNAs acting as ceRNAs to mediate the proliferation of VSMCs. Here, we performed microarray analysis to identify 134 DEcircRNAs in proliferative HASMCs compared to quiescent HASMCs, and we predicted 77 AGO-bound DEcircRNAs that were promising ceRNA molecules.
According to recent significant research, many miRNAs, including miR-24-3p, 29 miR-424-5p, 30 miR-1298-5p, 31 miR-204-5p 32 and let-7a-5p, 33 are involved in VSMC proliferation and down-regulated in proliferative VSMCs. In the HASMCs, the MREs of the miRNAs mentioned above were predicted in DEcircRNAs. We expected that Here, we found that hsa_circ_0002579 shared the same MREs as hsa-let-7a-5p with GAB2, TGFBR3 and ACVR1B. GAB2 was annotated in the Ras signalling pathway, an important pathway involved in cell proliferation. Thus, hsa_circ_0002579 might be a promising mediator of HASMC proliferation via involvement in the Ras signalling pathway. The well-known TGF-β signalling pathway is involved in many cellular processes of VSMCs, including cell growth. TGFBR3 is the most abundant TGF-β type II receptor, and the impact of TGFBR3 on TGF-β signalling depends on the circumstances. 34  DVL3 shared the MREs of hsa-miR-24-3p and hsa-miR-424-5p with hsa_circ_0006371 in HASMCs. One signalling modulator, Rspo3, of the Wnt signalling pathway was identified as a potential regulator of coronary stem proliferation. 37 Additionally, the Wnt signalling pathway was shown to be involved in the proliferation and migration of pulmonary arteriolar smooth muscle cells. 38 Consequently, we conjecture that hsa_circ_0006371 would be a positive conductor of Wnt signals isolating hsa-miR-24-3p and hsa-miR-424-5p to regulate HASMC proliferation.
In vitro, the overexpression of miR-424-5p inhibited VSMC proliferation, 30 which implies that miR-424-5p might be induced as a signal to counteract the proliferation. The MREs of hsa-miR-424-5p were predicted in hsa_circ_0006371 and hsa_circ_0004872, which were also predicted in the DEmRNA ABL2. ABL2 was annotated in the Ras signalling pathway. Some researchers have identified ABL2 as the target of many regulators to facilitate the proliferation process. 39 Here, we proposed that hsa_circ_0006371 and hsa_circ_0004872 could sponge hsa-miR-424-5p to up-regulate ABL2 to be involved in the proliferation of HASMCs via the Ras signalling pathway.
Moreover, hsa_circ_0006371 shared the same MREs of hsa-miR-24-3p, hsa-miR-424-5p and hsa-miR-1298-5p with MAP3K13 (namely LZK). The JNK signalling pathway was downstream of MAP3K13. A specific role of miR-1298 in regulating VSMC proliferation was demonstrated. 31 The MAPK/JNK pathway also might regulate VSMC proliferation. 40 In addition, another DEmRNA annotated in the MAPK pathway was ELK4, an ERK-regulated cofactor of the SRF transcription factor, which has been shown to act as a target of the ERK, JNK and p38 MAPK families. 41 indicating the coding potential of these circRNAs (Table 1).
In one previous study, the results showed that circRNAs were typically generated at the cost of canonical mRNA isoforms and even more abundant than the linear counterparts. 47 Identifying hundreds of head-to-tail junction reads in published data sets of chromatin-bound (nascent) RNA from fly heads, 48 circRNAs were considered as being generated co-transcriptionally and could compete with linear splicing mutually. 49 However, some studies beg to differ. The expression of some circRNAs is found to be independent of related linear isoforms. 50,51 The specifically overexpressed circRNAs in the multiple system atrophy brain have been determined, and the expression levels of linear transcripts are not significantly altered and thus do not follow the pattern of their circular counterparts. 52 With the alteration of circRNAs, only a small percentage of the corresponding mRNAs also make a change, which is consistent with that only a few circRNAs show co-regulation with their host genes, and circRNAs exhibited changes independent of the cognate mRNA. 53 In our study, we also failed to find the clear-cut relationship between circRNAs and the cognate linear transcripts. And we expected that the production of most circRNAs is regulated, which might be indispensable in the mediation of gene expression. 54 Moreover, circRNAs generated from one locus could be several and different. Thus, the relationship of homology circRNAs, as well as respective functions, required further exploration.
By DEcircRNAs-DEmRNAs co-expression subnetwork analysis, we found has_circ_0002579 was co-expression with 35 DEmRNAs in the proliferative HASMCs ( Figure 3B), and 30 miRNAs shared by hsa_circ_0002579 and HMGA2 3′-UTR ( Figure 5A). Studies have shown that HMGA2 is overexpressed in cancer cells 55,56 and promotes angiogenesis. 57 And miR-4500, 58 miR-98-5p 59 and miR-485-5p 60 have been reported to directly target the 3′-UTR of HMGA2 in other cell-type studies. These cases profile the bright prospects of hsa_circ_0002579, which acts as a ceRNA molecule interacting with miRNAs to regulate the proliferation of HASMCs by targeting HMGA2. We found that hsa_circ_0002579 was up-regulated in the proliferative HASMCs (Figure 1), and knockdown of hsa_circ_0002579 could decrease the expression level of HMGA2 and PCNA ( Figure 5B). We speculated that hsa_circ_0002579 might cross-interact with miRNAs to increase the expression of HMGA2, ultimately positively regulating the proliferation of VSMCs.
However, further experimental work remains to be done to validate those forecasting results.
In our study, we investigated the DEcircRNAs that were found in the proliferative and quiescent conditions of HASMCs to help elucidate their function. Next, the DEcircRNA-miRNA-DEmRNA crosstalk network was constructed for further investigations.
Then, we analysed the coding potential of DEcircRNA and the ef- Province (14393001D to SS).

CO N FLI C T O F I NTE R E S T
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All authors read and approved the final manuscript.

E TH I C A L A PPROVA L
The human aortic smooth muscle cells (HASMCs) we used in this study were commercial cells purchased from ScienCell.

DATA AVA I L A B I L I T Y S TAT E M E N T
Raw sequence reads have been deposited in the GEO database (GSE77278 and GSE77279).