Successful establishment of the diabetic cardiomyopathy rat model
After the STZ induction, blood sugar level and body weight of rats were measured and documented in a certain time interval. A distinct difference between DCM and normal control rats was observed. In the DCM group, the rat had significantly higher blood sugar levels and lower body weight, which were inconsistent with the typical manifestations and symptoms of diabetes mellitus (Figure1A, B).. This experiment reflects left ventricular systolic function in rats through left ventricular ejection fraction (LVEF) and short-axis shortening rate (FS). LVEF values were significantly decreased in the DCM group (56.33 ± 1.662%) compared with the normal control group (74.06 ± 2.631%). (Fig1C).. In diabetic cardiomyopathy, the increased deposition of collagen and interstitial fibrosis in the myocardium is one of the characteristic structural changes within it. In this study, LV collagen deposition and fibrosis ratio were markedly increased in DCM rats compared to normal controls (5.74 ± 0.3472 vs 0.94 ± 0.07924), using Masson staining method (Figure1D)..
Identification of differential expressed circRNAs
To investigate the circRNA expression profile of diabetic cardiomyopathy, we have completed the Arraystar Rat circRNA Array analysis of specimen from 3 DCM samples and 3 normal controls. A total of 14139 circRNAs were detected by Arraystar Rat circRNA Microarray. When comparing two groups of profile differences, the “fold change” between the groups for each circRNA is computed. The statistical significance of the difference is estimated by t-test. CircRNAs having |fold changes| 2.0 and p-values 0.05 are selected as the significantly differentially expressed. Eventually, 171 differentials expressed circRNAs, including 89 upregulated and 82 downregulated circRNAs were identified. A Scatter plot (Fig 2A) and a volcano plot (Fig 2B) showed the distributions of circRNAs more directly. The top 20 dysregulated circRNAs based on fold changes were summarized in Table 1. The result of hierarchical clustering showed a distinguishable circRNA expression profiling among samples. The data suggested that the circRNA in DCM was different in normal control samples (Fig 2C)..
Real-time qPCR validation of 6 differentially expressed circRNAs
After the comparison and the analysis of the differentially expressed multiple circRNAs in the microarray experiments, we randomly verified 6 circRNAs (rno_circRNA_000466, rno_circRNA_000964, rno_circRNA_003395, rno_circRNA_000173, rno_circRNA_013989, and rno_circRNA_003643), using GAPDH as the internal control. There were significant differences in all of the selected circRNAs (P<0.05) except for rno_circRNA_003643 (P = 0.06361). Among the 6 selected circRNAs, rno_circRNA_000466 was founded to be the highest increase of nearly 28-fold change. rno_circRNA_000964 was down-regulated with a 0.51-fold decrease. The results are shown in Fig 3. There was a great consistency between the real-time qPCR results and microarray analysis data, which demonstrated the high reliability of the microarray expression data.
Functional profiles of the parental genes of differentially expressed circRNAs in DCM
circRNA cis-regulates the expression of parental genes. On the one hand, circRNA can combine with RNA binding proteins to affect the expression of the parental gene mRNA. On the other hand, competitive complementary pairing between introns during the formation of circular RNA can achieve a balance between linear RNA and affect mRNA expression, even protein translation. Parental genes from 171 differentially expressed circRNAs were subjected to GO and KEGG pathway enrichment analyses. The top 10 classes of GO enrichment terms and the top 10 classes of KEGG pathway enrichment terms are presented (Figure4).Compared to the normal heart sample, the data demonstrated that the gene expression profile of linear counterparts of differentially expressed circRNAs in DCM involves in the insulin signaling pathway, autophagy, HIF–1 signaling pathway, inflammatory mediator regulation of TRP channels, insulin secretion, regulation of lipolysis in adipocytes and so on. As for the insulin signaling pathway, CALM3, CBLB, INSR, PIK3R3, PRKACA, SORBS1 were dysregulated in DCM heart tissue. And ATP1B1, CALM3, CAMK2A, GRIA2, PIK3R3, PRKACA in the cAMP signaling pathway were down in DCM heart tissue.In the autophagy pathway, differentially expressed IGF1R, MTMR3, PIK3R3 PRKACA, SH3GLB1 were found to be altered. To be noted, many genes participated in cell membrane biological activity, including regulation of transmembrane transport, regulation of potassium ion transmembrane transporter activity, etc.
circRNA‐microRNA interaction network construction
CircRNA is generally produced by RNA alternative splicing and circRNA has been found to play the role of miRNA sponge. As a competitive endogenous RNA (ceRNA) binding to intracellular miRNA, blocking the inhibition of miRNA on its target gene is its main regulatory mode. The circRNA/microRNA interaction was predicted with Arraystar’s home-made miRNA target prediction software based on TargetScan and miRanda, and the differentially expressed circRNAs within all the comparisons were annotated in detail with the circRNA/miRNA interaction information. All the differentially expressed circRNAs prediction was based on the principle of complementary sequence pairing, and targeted miRNAs were ranked according to their mirSVR scores, and five miRNAs with the highest mirSVR score were identified as MREs for each circRNAs. In the topological network (Fig 5A), the orange oval nodes represent upregulated circRNAs and the yellow rectangle nodes represent down-regulated circRNAs. It can be founded in the network that one circRNA interacts with a couple of miRNAs and some miRNA can also interact with multiple circRNAs. Furthermore, we using CytoHubba plugin to screen the top 20 nodes in network ranked by the Degree method, to assess the key modules that significantly regulated the pathogenesis of DCM. The co-expression network reveals that rno-miR–22, rno-miR–27, rno-miR–28, rno-miR–466, rno-miR–320 may play important roles in the development of DCM. And the magnified network was visualized using Cytoscape (v.3.8.0) as well (Fig 5B).
The construction of rno_circRNA_000466-associated ceRNA networks
In our previous array and qRT-PCR test, rno_circRNA_000466 was highly expressed in DCM heart tissue with a 28-fold change. And based on the hypothesis that circRNAs may function through the ceRNA mechanism, we used Arraystar’s home-made miRNA target prediction software to predict the interaction miRNA of rno_circRNA_000466. Five miRNAs, including rno-miR–1306–3p, rno-miR–665, rno-miR–3068–5p, rno-miR–223–3p, rno-miR–3593–5p, were identified most likely to be interacted with rno_circRNA_000466. And then we use DIANA microT-CDS online tools to predict the target of the five miRNAs. At last, 139 targets were found with the threshold set to 0.9. The topical relationship was visualized through Cytoscape software. The functional annotation was performed using the CluGO and CluPedia plugins. The results show that some of the target genes were enriched in the TGF-β signaling pathway, regulation of glucose transmembrane transport, endocrine process and response to lipoprotein particle, which are related to metabolic remodeling and cardiac fibrogenesis in DCM. These results were presented in Fig 6.