目的 利用双生子人群分别探索与空腹血糖(fasting plasma glucose,FPG)或糖化血红蛋白(glycated haemoglobin, HbA1c)存在相关性的胞嘧啶-磷酸-鸟嘌呤(cytidine-phosphate-guanosine site,CpG)位点。方法 研究对象来自中国双生子登记系统2013年6月至12月和2017年6月至2018年10月于山东青岛、浙江、江苏、四川及黑龙江五地募集的338人(169对)同卵双生子。甲基化检测分别为Illumina Infinium HumanMethylation450 BeadChip和Illumina Infinium MethylationEPIC BeadChip两种芯片,采用混合效应模型,分别将空腹血糖及HbA1c作为主效应,将甲基化水平(β值)作为因变量,将年龄、体重指数(body mass index, BMI)、血压、血细胞组成成分、用SVA包生成的代理变量等连续变量,及性别、吸烟、饮酒、是否服用降糖药等分类变量作为协变量纳入固定效应模型,将双生子对编号纳入随机效应模型,将截距设置为随机,进行回归分析,找出分别与空腹血糖或HbA1c相关的CpG位点。结果 本研究最终纳入同卵双生子338人(169对),CpG位点412 459个,其中男性同卵双生子114对、女性55对,平均年龄(48.2±11.9)岁。在调整年龄、性别、BMI、血压、吸烟、饮酒、血细胞成分等协变量及多重比较校正后,发现7个与空腹血糖相关的CpG位点(cg19693031、cg01538969、cg08501915、cg04816311、ch.8.1820050F、cg06721411、cg26608667), 其中3个位点(cg08501915、ch.8.1820050F、cg26608667)为本研究新发现的位点;发现10个与HbA1c相关的CpG位点(cg19693031、cg04816311、cg01538969、cg01339781、cg01676795、cg24667115、cg09029192、cg20697417、ch.4.1528651F、cg16097041),其中4个位点(cg01339781、cg24667115、cg20697417、ch.4.1528651F)为本研究新发现的位点。本研究发现,位于TXNIP基因上的cg19693031位点在DNA甲基化与空腹血糖及HbA1c相关分析中均是P值最小的位点(P_FPG=2.42×10^-19, FD {R_{\mathrm{FPG}}}_{} < 0.001; P_HbA1c=1.72×10^-19, FD {R_{\mathrm{HbA}1\mathrm{c}}}_{} < 0.001)。 结论 在验证既往研究发现的同时,本研究利用双生子的天然优势,发现新的与空腹血糖和HbA1c相关的CpG位点,为血糖相关指标的DNA甲基化机制提供一定的线索,但尚需外部样本进一步验证。

Objective: To explore the cytidine-phosphate-guanosine (CPG) sites associated with fas-ting plasma glucose (FPG) and glycated haemoglobin (HbA1c) in twins.Methods: In the study, 169 pairs of monozygotic twins were recruited in Qingdao, Zhejiang, Jiangsu, Sichuan and Heilongjiang in June to December of 2013 and June 2017 to October 2018. The methylation was detected by Illumina Infinium HumanMethylation450 BeadChip and Illumina Infinium MethylationEPIC BeadChip. According to the Linear Mixed Effect model (LME model), fasting plasma glucose and HbA1c were taken as the main effects, the methylation level (β value) was taken as the dependent variable, continuous variables, such as age, body mass index (BMI), blood pressure, components of blood cells, surrogate variables generated by SVA, and categorical variables, such as gender, smoking and drinking status, hypoglycemic drugs taking, were included in the fixed effect model as covariates, and the identity numbers (ID) of the twins was included in the random effect model. The intercept was set as a random. Regression analysis was carried out to find out the CpG sites related to fasting blood glucose or HbA1c, respectively.Results: In this study, 338 monozygotic twins (169 pairs) were included, with 412 459 CpG loci. Among them, 114 pairs were male, and 55 pairs were female, with an average age of (48.2±11.9) years. After adjustment of age, gender, BMI, blood pressure, smoking, drinking, blood cell composition, and other covariates, and multiple comparison test, 7 CpG sites (cg19693031, cg01538969, cg08501915, cg04816311, ch.8.1820050F, cg06721411, cg26608667) were found related to fasting blood glucose, 3 of which (cg08501915, ch.8.1820050f, cg26608667) were the newly found sites in this study; whereas 10 CpG sites (cg19693031, cg04816311, cg01538969, cg01339781, cg01676795, cg24667115, cg09029192, cg20697417, ch.4.1528651F, cg16097041) were found related to HbA1c, and 4 of which(cg01339781, cg24667115, cg20697417, and ch.4.1528651f) were new. We found that cg19693031 in TXNIP gene was the lowest P-value site in the association analysis between DNA methylation and fas-ting plasma glucose and HbA1c (P_FPG=2.42×10^-19, FDR_FPG<0.001; P_HbA1c=1.72×10^-19, FDR_HbA1c<0.001).Conclusion: In this twin study, we found new CpG sites related to fasting blood glucose and HbA1c, and provided some clues that partly revealed the potential mechanism of blood glucose metabolism in terms of DNA methylation, but it needed further verification in external larger samples.