3.1 Characteristics of the cases and controls
The Chinese BMI obese criterion was applied to divide the participants into cases and controls. General characteristics of cases and controls were shown in Table 1. There was no difference between cases and controls in gender, height, smoking proportion or drinking proportion. Nevertheless, the cases had higher levels of weight, WC, BMI, FBG, TC, TG, LDL-C, SBP, DBP, HTN proportion and lower levels of age, HDL-C as compared with controls (P<0.05) (Table 1).
3.2 Association between CASZ1 methylation levels and BMI
We obtained methylation levels for 17 CpG sites in CASZ1. Multiple linear regression models were used to evaluate the association of CASZ1 methylation levels with BMI. In Table 2, after adjustment for age, sex, smoking and drinking status, CpG51, 67, 70, 72, 94, 98, 116, 157, 165 were associated with BMI, respectively (P<0.05). The associations between methylation levels of CpG72 and 165 and BMI were significant after considering multiple testing (Bonferroni correction, P < 0.003 ≈ 0.05/17). In model 2, adjusting for age, sex, smoking and drinking status, SBP, FBG and TC, there were associations between CpG51, 70, 72, 98, 157, 165 and BMI (P<0.05). The remaining CpG sites were not associated with BMI(P>0.05). The association between methylation level of CpG 165 and BMI was significant after considering multiple testing in model 2 (P = 0.001). The median of all of the tested CpG sites in CASZ1 was associated with BMI both in model 1 (P=0.002) and model 2 (P=0.032) (Table 2).
3.3 Association between CASZ1 methylation and overweight/obesity
As shown in figure 1, the CASZ1 methylation levels of 17 CpG sites were expressed as the median. There was a significant difference of CASZ1 DNA methylation values for CpG58, 72, 98, 116, 157, 163, 165 and median of all CpG sites (P<0.05). The median methylation level of these CpG sites of the cases were higher than those of the controls (Table 3). The level of CpG165 had a significant difference between cases and controls both in male and female as well (P=0.004 and 0.003, respectively) (figure 2). In table 3, after adjustment for age, sex, smoking and drinking status, CpG70, 72, 98, 116, 163, 165 were associated with overweight/obesity cases which were defined by BMI obesity criterion markedly (P<0.05). Besides, CpG72, 98, 163, 165 were related to overweight/obesity (P<0.05) after adjustment for age, sex, smoking and drinking status, SBP, FBG and TC. In the two models, the median of methylation level at CpG sites was associated with overweight/obesity divided by BMI obese criterion (P<0.05). The associations between methylation level of CpG98 (OR=1.06, 95%CI: 1.02-1.10) and CpG 165 (OR=1.15, 95%CI: 1.07-1.42) and overweight/obesity were significant after considering multiple testing (P < 0.003).
3.4 The additive interaction effects on BMI with adjustment for environmental factors and CpG sites
We tested the interaction effects of methylation levels of CASZ1 CpG sites and risk factors on BMI using an additive model. Interaction effects between the CpG74, 98 and age on BMI were observed after adjustment for age, sex, smoking and drinking status, SBP, FBG and TC (β= -0.004, -0.004, P-interaction = 0.031, 0.040, respectively). There was a similar interaction effect between CpG163 and sex on BMI( β= 0.177，P-interaction= 0.043). CpG163 and drinking status had an significant additive interaction effect on BMI as well(β= -0.285，P-interaction= 0.001). Besides, CpG70 and SBP had an interaction effect on BMI(β= 0.002，P-interaction= 0.045). There were also interaction effects between CpG35, 94 and FBG on BMI (β= -0.046 and -0.072, P-interaction = 0.019 and 0.033, respectively). Interactions of CpG163 and TG, CpG72 and HDL-C were found for decreasing BMI (β= -0.080 and -0.200, P-interaction = 0.002 and 0.011, respectively) (Table 4). The multiplicative interaction effects between the CASZ1 CpG sites and phenotypes on overweight/obesity were conducted after adjusting for age, sex, smoking and drinking status, SBP, FBG and TC as well. However, these interactions were not significant after considering multiple testing.