The subjects enrolled in this study were recruited from the endocrinology department of Wuhan Union Hospital from January 2019 to December 2019. This study has been approved by the Institutional Research Ethics Committee of Union Hospital, Huazhong University of Science and Technology. A prior written informed consent was obtained from each participant in this study. The subjects were dividedinto three groups: normal control (NC group, n=43), diabetes without cardiovascular diseases (DM group, n=60), and diabetes with cardiovascular diseases (CVD group, n=51). The inclusion criteria of subjects in the NC group were: no history of diabetes or cardiovascular diseases, normal glucose tolerance in OGTT, normal coronary angiography, no history of carotid artery disease or peripheral arterial disease. Diabetes were diagnosed according to the World Health Organization (WHO, 1999) (18). Data on vital signs, anthropometric factors, medical history and behaviors as well as physical activity were collected. Additional data, including weight, height, BMI, waist circumference, hip circumference, glycemia and glycated hemoglobin were obtained for each individual. Diabetes patients in CVD group were evidenced as follows: ischemic heart disease defined by clinical history, and/or ischemic electrocardiographic alterations; peripheral vascular disease including atherosclerosis obliterans and cerebrovascular disease based on history, physical examinations and Doppler velocimetry.

Genomic DNA (gDNA) was extracted from nine blood samples (3 from NC group, 3 from DM group, and 3 from CVD group) by a DNeasy Blood & Tissue Kit (Qiagen, Fremont, CA). Using Biomag™ magnetic beads that coupled with the mouse monoclonal antibody against 5-methylcytidine, methylated DNA was immunoprecipitated, which was then eluted and purified by QiagenMinElute columns (Qiagen, Fremont, CA). The total input genomic DNA and immunoprecipitated DNA were labeled with Cy3- and Cy5-fluorophere, respectively. They were then hybridized to NimbleGen Human DNA Methylation 3×720K CpG Island Plus RefSeq Promoter Microarray. Scanning was performed with the Axon GenePix 4000B microarray scanner according to the manufacturer’s guidelines detailed in NimbleGen MeDIP-Chip protocol (NimbleGen Systems Inc., Madison, USA). The MeDIP-chip data were submitted to Gene Expression Omnibus (GEO) data repository (http://www.ncbi.nlm.nih.gov/projects/geo/) under accession number GSE188395.

The log₂ ratio obtained from raw data values has been normalized to ensure technical invariability and evaluate methylation differences. Median centering, quantile normalization, and linear smoothing were then performed using Bioconductor packages Ringo, Limma and MEDME (19). We then applied a sliding-window (750bp) peak-finding algorithm which was provided by NimbleScan v2.5 (Roche-NimbleGen) to analysis the MeDIP-chip data. Furthermore, to confirm whether the probes drawnwere from a more positive distribution of intensity log₂ ratio, a one-sided Kolmogorov-Smirnov (KS) test was applied. Each probe received a -log₁₀ P scores from the windowed KS test around that probe. The region of which adjacent probes significantly rose above a set threshold was assigned to an enrichment peak (EP). The peaks were detected by searching for at least two probes that above a P-value minimum cut-off (-log₁₀) of 2. Peaks that within 500bp of each other are merged.

The Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were applied to investigate the potential biological processes and the DNA methylation function (20, 21). The DAVID Functional Annotation Tool (http://david.abcc.ncifcrf.gov/) was adopted for GO enrichment, and KEGG pathway analysis with a 0.05 cut-off for Benjamini adjusted P value (Q-value) (22). According to GO annotation, the differentially methylated genes were divided into cellular component (CC), molecular function (MF), and biological process (BP). The source genes used to analysis are freely available at https://github.com/hsqhaha/2021-10-08.

In order to confirm the MeDIP-chip data, MSP was applied to identify the methylation status of specific genes (23). Peripheral blood of the subjects (40 from NC group, 57 from DM group, and 48 from CVD group) were collected. The MSP primers are designed by online MethPrimer software (http://www.urogene.org/methprimer). The sequences and product length of each primer used for MSP analysis were listed in Table 1 .

qPCR was performed by SYBR Premix Ex Tap™ (Takara, Japan) on the LightCycler^® 480 II (Roche Diagnostics Ltd., Basel, Switzerland) following the manufacturer’s instructions. GAPDH was used as the reference genes to normalize the relative mRNA expression. The primer sequences are listed in Table 2 . The relative quantification analysis was calculated using the 2^-△△Ct method (24). The primers of qPCR were designed using the Primer-BLAST program (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) or software packages Primer Premier 5.0.

The expressions of PLCB1, DRD5, PLGF, FATP4, and VEGFB in serum were analyzed by enzyme‐linked immunosorbent assay (Blue Gene, Shanghai, China). The measurement was performed following the manufacturer’s instructions. All samples were assessed in triplicate.

Statistical analysis was performed by SPSS 22.0 software and GraphPad 5. The significance was calculated by one-way analysis variance (ANOVA) or Student t-test. All experiments were performed at least three times independently. The results are presented as the mean ± standard error. Differences were statistically significant if P< 0.05.

There was no significant difference in age, gender, systolic blood pressure (SBP), and triglycerides (TG) among the three groups (P> 0.05) as Table 3 shown. Compared with NC group, waist-hip ratio (WHR), diastolic blood pressure (DBP), total cholesterol (TC), HbA1c, and fasting plasma glucose (FPG) were significant higher in CVD and DM group (P< 0.05). Meanwhile, no obvious difference was observed in those clinical parameters between DM and CVD groups (P> 0.05). Consistent with the previous findings, BMI in DM group was significantly higher than that in NC group (P< 0.05). Compared with NC group, there was an increased tendency of BMI of diabetic patients in CVD group (P> 0.05).

The global cytosine methylated peaks in peripheral blood were obtained to assess the DNA methylation variations among the three groups. A total number of 57273 methylated peaks were identified. Among these peaks, 19677 (34.3%) were in NC group, and 20552 (35.9%) methylated peaks were in DM group, and the remaining 17044 (29.8%) methylated peaks were in CVD group.

As shown in Figure 1A , the CpG lands (CGIs) were grouped into three classes depending on the distance of sequence, which contains high-frequency CG sites to RefSeq genes. They are promoter islands (from about -10kb to +0.5kb around the transcription starting site), intragenic islands (from +0.5kb around the transcription start site to the transcription end site) and intergenic islands (the remainder that does not fall in either promoter or intragenic). The numbers of methylated enrichment peaks (Eps) among different groups were depicted in Figure 1B . A number of 13086, 13452, and 11461 methylated EPs were detected in NC, DM, and CVD group, respectively, in promoter CGIs, while the number of EPs in intragenic islands is 2196, 2418 and 1901 in NC, DM and CVD group respectively.

Based on the ratio of CpG, GC contents and the length of CpG-rich region, the promoter regions were subdivided into three classes: high CpG density promoter (HCP), low CpG density promoter (LCP) and intermediate CpG density promoter (ICP). As shown in Figure 1C , compared with the number of methylated peaks in ICP and LCP types, HCP type had the highest number among the three groups. Furthermore, both DM and CVD groups have more methylated differential enrichment peaks in HCP than those in ICP and LCP, regardless of the promoters that were either up- or down-methylated ( Figures 1D, E ).

Compared with the NC group, 179 and 251 genes were methylated in the promoter regions of DM and CVD group, respectively, whereas 54 genes overlapped between these two groups ( Figure 2A ). Of these genes, a total of 211 genes were hypermethylated while 166 were hypomethylated, in which 22 genes and 32 genes were overlapped between DM and CVD group, respectively ( Figures 2B, C ). Meanwhile, the number of methylated genes in CVD group was more than that in DM group, and each group had its special aberrant methylated genes apart from the overlapped genes. Additionally, CVD group had 149 differentially methylated genes compared to the DM, which includes 79 up-regulated genes and 70 down-regulated genes. The above results indicate that DNA methylation modification was widespread in diabetes and its cardiovascular complications. DNA methylation of specific genes might play a role in the outcome of cardiovascular diseases in diabetes.

To analyze the potential role of differentially methylated genes, we performed gene ontology (GO) and pathway enrichment analysis using the DAVID online tool. The threshold was set asP< 0.05. In Figure 2D , KEGG pathway analysis showed that the calcium signaling pathway and the PI3K-Akt signaling pathway ranked the top 2 pathways according to the enrichment score, which are the critical signaling pathways in the development of DM and its macrovascular complications. GO analysis revealed the functional pathways of abnormally methylated genes in DM and CVD groups, compared with NC group ( Table 4 ). Genes participated in the fatty acid metabolic process such as fatty acid transport protein 3 (FATP3) and fatty acid transport protein 4 (FATP4) were screened out to be down-methylated. Meanwhile, genes involved in the vascular endothelial growth factor receptor (VEGFR) signaling pathway such as vascular endothelial growth factor B (VEGFB), placental growth factor (PLGF), and coagulation factor II thrombin receptor (F2R) were found to be down-methylated.

Several studies have demonstrated the important role of calcium signaling pathway in atherosclerosis (25–27). Additionally, dysfunction of vascular endothelial function could enhance atherosclerotic plaque progression (28). Based on the results of bioinformatics analysis, we also found several significantly hypomethylated genes (PLCB1, CAMK1D, and DRD5) participated in the calcium signaling pathway, while several hypomethylated genes (VEGFB, PLGF, F2R, FATP3, and FATP4) participated in VEGFR signaling pathway. Then we performed MSP evaluation using leukocytes from peripheral blood of the subjects. In general, a good correlation between the MeDIP-chip and MSP results was observed for PLCB1, VEGFB, PLGF, and FATP4, except for DRD5. However, MSP of FATP3, CAMK1D, and F2R did not show significant differences among the three groups ( Figures 3 , 4 ). Compared with NC group, DM and CVD groups had a significantly lower DNA methylation in PLCB1, VEGFB, PLGF, and FATP4 (P< 0.05). Furthermore, the levels of DNA methylation in FATP4 and VEGFB in CVD group were significantly lower than those in DM group (P< 0.05). There was no significant difference in PLCB1 and PLGF between DM and CVD group (P> 0.05).

To further verify the differential methylated genes mentioned above, we performed qPCR to detect the mRNA expression in serum of three groups. The results showed that the levels of PLGF, PLCB1, FATP4, and VEGFB in the DM or CVD group were significantly higher than those in NC group (P < 0.05) ( Figures 5A, E, G, I ). Elisa results also showed the similar patterns among three groups ( Figures 5A-J ). A good consistency among the DNA methylation, mRNA and protein levels was observed for PLCB1, PLGF, FATP4, and VEGFB genes.