Hypertension is a chronic noncommunicable disease, and approximately 90% of cases are considered as essential hypertension without a definitive cause. In China, there is a high prevalence of hypertension (24.7%) in adults aged 18−69 years, but the awareness, treatment and control of hypertension are low (Zhang et al., Reference Zhang, Shi, Zhou, Huang, Zhao, Li, Zhang, Han, Peng, Li, Wang, Ezzati, Wang and Li2023). Hypertension has now become a considerable risk factor for reducing life expectancy, being a huge burden for the nationwide health system (Wang et al., Reference Wang, Peng, Nie, Chen, Weldon, Zhang, Xiao and Cai2016).
As a complex phenotype, hypertension is controlled by the interplay between genetic and environmental factors mediated by epigenetic modification. At present, although considerable research progress into the genetics on hypertension has been achieved (J. Chen et al., Reference Chen, Wang, Li, Xu, Tian and Zhang2021; Franceschini et al., Reference Franceschini, Fox, Zhang, Edwards, Nalls, Sung, Tayo, Sun, Gottesman, Adeyemo, Johnson, Young, Rice, Duan, Chen, Li, Tang, Fornage, Keene, Andrews and Zhu2013; Ganesh et al., Reference Ganesh, Tragante, Guo, Guo, Lanktree, Smith, Johnson, Castillo, Barnard, Baumert, Chang, Elbers, Farrall, Fischer, Franceschini, Gaunt, Gho, Gieger, Gong and Asselbergs2013; Kato et al., Reference Kato, Loh, Takeuchi, Verweij, Wang, Zhang, Kelly, Saleheen, Lehne, Leach, Drong, Abbott, Wahl, Tan, Scott, Campanella, Chadeau-Hyam, Afzal, Ahluwalia and Chambers2015; Wain et al., Reference Wain, Verwoert, O’Reilly, Shi, Johnson, Johnson, Bochud, Rice, Henneman, Smith, Ehret, Amin, Larson, Mooser, Hadley, Dörr, Bis, Aspelund and Esko2011), the specific pathogenesis underlying hypertension still needs to be further elucidated. It has been reported that obesity may play an important role in the pathogenesis of hypertension. One of the mechanisms is that the increase of visceral adiposity could lead to a proinflammatory, pro-oxidative milieu that promotes insulin resistance, and the insulin resistance in skeletal muscle tissue could action on the endothelium by vascular rarefaction, reductions in vascular relaxation, and vascular remodeling, thus further influencing blood pressure (BP) levels (Rao et al., Reference Rao, Pandya and Whaley-Connell2015). Another mechanism by which obesity may affect the development of hypertension may be related to an imbalance or lack of number of and function of immune cells, such as the neutrophils, in the adipose tissues of obese individuals (J. Li et al., Reference Li, Yin, Chen, Li, Ding, Wu, Yang and Xu2023). Although obesity is now an established risk factor for hypertension, the mechanisms are yet partially understood.
In recent years, increasing evidence has supported the significant role of epigenetic modification, such as DNA methylation (DNAm) variation underlying complex phenotypes. Currently, a functional role of DNAm variation in the development of hypertension or regulation of BP has gradually been demonstrated (Demura & Saijoh, Reference Demura and Saijoh2017; Liang, Reference Liang2018; Wang, Yao, Li et al., Reference Wang, Yao, Li, Wu, Duan, Xu, Tian, Li, Tan and Zhang2023). Moreover, the association of DNAm variation with obesity has also been widely studied (van Dijk et al., Reference van Dijk, Molloy, Varinli, Morrison and Muhlhausler2015). Furthermore, obesity could induce insulin resistance via changes in the DNAm profile of insulin pathway genes, and the DNAm profile changed depending on BMI (Małodobra-Mazur et al., Reference Małodobra-Mazur, Alama, Bednarska-Chabowska, Pawelka, Myszczyszyn and Dobosz2019), while the relationship between insulin resistance and hypertension has previously been proved (Rao et al., Reference Rao, Pandya and Whaley-Connell2015). However, to date, very few studies have investigated whether obesity (or BMI)-related DNAm variation can mediate the association between BMI and BP.
Nowadays, the trait or disease-discordant monozygotic (MZ) twin design has been a powerful tool for epigenetic studies while controlling for individual genetic components (W. Li et al., Reference Li, Christiansen, Hjelmborg, Baumbach and Tan2018; Tan et al., Reference Tan, Christiansen, Hjelmborg and Christensen2015). The Chinese population differ in the genetic make-up, environmental exposure and lifestyles from other ethnic populations worldwide. Considering that BMI and BP are separately important indicators of obesity and hypertension, we here aimed to evaluate the mediating effect of BMI-related DNAm on the relationship of BMI and BP by a causal inference test (CIT) and mediation analysis in a sample of Chinese MZ twins.
Materials and Methods
The primary materials and methods of this study were in accord with those reported in our previously published studies (Wang, Li et al., Reference Wang, Li, Duan, Xu, Tian, Li, Tan and Zhang2023; Wang, Yao, Li et al., Reference Wang, Yao, Li, Wu, Duan, Xu, Tian, Li, Tan and Zhang2023; Wang, Yao, Tan et al., Reference Wang, Yao, Tan, Li, Duan, Tian, Xu and Zhang2023).
Participants
Participants’ recruitment and collection have been described previously (Duan et al., Reference Duan, Ning, Zhang, Wang, Zhang, Tan, Tian and Pang2013). The participants who were pregnant or breastfeeding, or who suffered from heart disease, stroke and/or tumor, or who were regularly taking any medications within one month before participation were excluded. Meanwhile, participants who were unable or unwilling to cooperate were also excluded. In the analysis with systolic blood pressure (SBP) as the outcome, we excluded 8 pairs of twins. In the analysis with diastolic blood pressure (DBP) as the outcome, we excluded 10 pairs of twins. A total of 60 SBP-discordant and 58 DBP-discordant MZ twin pairs were finally included, with a median of absolute values of intrapair trait difference of 17 mmHg (95% range: 1, 50) for SBP and 10 mmHg (95% range: 2, 25) for DBP respectively.
All co-twin pairs completed a questionnaire and undertook a health examination after a 10−12 hour overnight fast. The twins’ zygosity was determined by a combination of sex, ABO blood types, and DNA testing (Wang, Yao, Li et al., Reference Wang, Yao, Li, Wu, Duan, Xu, Tian, Li, Tan and Zhang2023). The measurement of BMI and BP has been described in our previous studies (Z. Li et al., Reference Li, Wang, Tian, Duan, Xu and Zhang2021; Wang, Yao, Li et al., Reference Wang, Yao, Li, Wu, Duan, Xu, Tian, Li, Tan and Zhang2023). The ethical principles of the Helsinki Declaration were followed, and prior written informed consent was achieved.
Reduced Representation Bisulfite Sequencing Experiment
As described in our previous studies (Wang, Li et al., Reference Wang, Li, Duan, Xu, Tian, Li, Tan and Zhang2023; Wang, Yao, Li et al., Reference Wang, Yao, Li, Wu, Duan, Xu, Tian, Li, Tan and Zhang2023), total DNA extracted from whole blood was tested by the reduced representation bisulfite sequencing (RRBS) experiment (Biomarker Biological Technology in Beijing, China). Briefly, genomic DNA was first digested to generate short fragments that were then chosen and bisulfite-converted. Subsequently, a cDNA library was constructed and sequenced to obtain raw sequencing data that covered 551,447 CpGs across the genome of each individual. The raw sequencing data was mapped to the human GRCh37 by Bismark (Krueger & Andrews, Reference Krueger and Andrews2011) and then imported to BiSeq to smooth the methylation level (Hebestreit et al., Reference Hebestreit, Dugas and Klein2013). After quality control, a total of 248,593 CpGs for SBP and 249,300 CpGs for DBP remained for subsequent analyses. The methylation β value was transformed to M value by applying log2 transformation with M = log2(β/[1− β]) (Du et al., Reference Du, Zhang, Huang, Jafari, Kibbe, Hou and Lin2010).
Since total DNA was extracted from whole blood, the methylation profiles of distinct cell-types might cause false discoveries (Jaffe & Irizarry, Reference Jaffe and Irizarry2014). The top five components of ReFACTor method were chosen as covariates to correct for the cell-type composition effect on DNAm in this analysis (Rahmani et al., Reference Rahmani, Zaitlen, Baran, Eng, Hu, Galanter, Oh, Burchard, Eskin, Zou and Halperin2016). The top five components of ReFACTor method for SBP and DBP are separately provided in Table S1 and Table S2.
Statistical Analysis
Mediation analyses
We first used a partial correlation analysis model to examine the correlation between intrapair BMI difference and intrapair blood pressure difference, with adjusting for sex and age. Then we applied the CIT method and mediation analysis to evaluate the mediating effect of DNAm on the association between BMI and BP levels (Millstein et al., Reference Millstein, Zhang, Zhu and Schadt2009). For each individual CpG, four models were separately fitted in the CIT method: (1) Model 1 assessed the association of BMI and BP by using generalized estimating equation (GEE) through geeglm function in R-package geepack, with BP as the outcome and BMI as the predictor, adjusting for age and sex. (2) Model 2 assessed the association of BMI and DNAm of each CpG by conducting a genomewide DNA methylation analysis. We fitted this model by using the GEE while controlling for BP, with DNAm of each CpG as the outcome and BMI as the predictor, adjusting for age, sex, cell-type composition as well as BP. The Manhattan plot was drawn. To correct for multiple testing, we calculated false discovery rate (FDR) (Benjamini & Hochberg, Reference Benjamini and Hochberg2018) and defined FDR < 0.05 as genomewide significance. For CpGs with FDR ≥ 0.05, we defined p value < 1×10−6 as suggestive significance and 1×10−6 ≤ p value < 1×10−5 as weaker-than-suggestive significance (S. Li et al., Reference Li, Wang, Zhang, Li, Lund, Kruse, Mengel-From, Christensen and Tan2021). The CpGs with p value < 1×10−4 were reported as top CpGs in our study and were selected for subsequent analysis (Mohammadnejad et al., Reference Mohammadnejad, Soerensen, Baumbach, Mengel-From, Li, Lund, Li, Christiansen, Christensen, Hjelmborg and Tan2021; Wang, Li et al., Reference Wang, Li, Duan, Xu, Tian, Li, Tan and Zhang2023). (3) Model 3 assessed the association of DNAm of each CpG with BP by using the GEE while controlling for BMI, with BP as the outcome and DNAm of each CpG as the predictor, adjusting for age, sex, cell-type composition as well as BMI. The CpGs with p values of association < .05 were chosen for subsequent analysis (Wang, Li et al., Reference Wang, Li, Duan, Xu, Tian, Li, Tan and Zhang2023). (4) This model assessed the independence of BMI and BP by using R-package cit, adjusting for age, sex and cell-type composition as well as DNAm of each CpG identified in Model 3. If the median of all p values was more than .05 after conducting 1000 simulations using the Permutation method, it was considered that the BMI and SBP/DBP was insignificant. In the above four CIT models, in order to address the paired structure of twin data, we included a vector that identified the clusters of twins within a pair into the GEE model.
For the candidate CpG mediators obtained through the CIT method, we further verified their mediating effect on the association of BMI with BP by using R-package mediation (Tingley et al., Reference Tingley, Yamamoto, Hirose, Keele and Imai2014). We used 5000 simulations to calculate estimates and 95% confidence intervals (95% CIs). For each CpG, the average causal mediation effect (ACME), average direct effect (ADE), and proportion mediated (or absolute value of ratio of indirect effect to direct effect) were calculated. If the lower limit (LL) to the upper limit (UL) in the 95% CIs for the ACME of a particular CpG did not include zero, this CpG was considered to significantly mediate the association between BMI and BP. Additionally, we also tested the parallel mediating effects of DNAm mediators between BMI and BP by employing a bootstrap method using PROCESS macro in SPSS program and reported the total mediation effect of DNAm mediators. This approach allows researchers to investigate multiple mediator variables simultaneously, rather than focusing on a single mediator’s impact. The above analyses were also performed stratified by sex. Moreover, whether BMI could also mediate the association between DNAm of CpGs and BP was further evaluated.
Region-based analysis
As described previously (Wang, Li et al., Reference Wang, Li, Duan, Xu, Tian, Li, Tan and Zhang2023; Wang, Li, Jiang et al., Reference Wang, Li, Jiang, Lin, Wu, Wen, Xu, Tian, Li, Tan and Zhang2021; Wang, Li, Wu et al., Reference Wang, Li, Wu, Tian, Duan, Li, Tan and Zhang2021; Wang, Yao, Li et al., Reference Wang, Yao, Li, Wu, Duan, Xu, Tian, Li, Tan and Zhang2023), we applied the comb-p tool to detect the BMI-associated differentially methylated regions (DMRs) using the CpGs with p values < .05 identified in Model 2 (Pedersen et al., Reference Pedersen, Schwartz, Yang and Kechris2012). The significant enriched DMRs were determined by slk corrected p value < .05.
Ontology enrichments analysis
We submitted the CpGs with p values < .05 identified in Model 2 and the candidate CpG mediators identified in CIT method to Genomic Regions Enrichment of Annotations Tool (GREAT) online to analyze the functional ontology enrichments (McLean et al., Reference McLean, Bristor, Hiller, Clarke, Schaar, Lowe, Wenger and Bejerano2010). The significant ontology enrichments were judged by FDR < 0.05.
Study power estimation
We have published a computer simulation study on the power of methylation analysis using twin design (W. Li et al., Reference Li, Christiansen, Hjelmborg, Baumbach and Tan2018). According to this study, if one trait or disease had a heritability (h 2) of .6 and there was a low correlation between environmental factors and DNAm (R 2 M, E = .1), the sample size required for statistical power to exceed 80% in a trait/disease-discordant twin design would range from 22 (when the correlation within twin pair due to either shared genetic background or common environment, denoted as ρ ε = .8) to 63 (when ρ ε = .1) pairs, which was an immense improvement over the ordinary case-control design. The heritability was about .76 for BMI (Duan et al., Reference Duan, Pang, Zhang, Li, Kruse, Kyvik, Christensen and Tan2011) and .60 for BP (Wu et al., Reference Wu, Zhang, Pang, Jiang, Wang, Li, von Bornemann Hjelmborg and Tan2015) in the same twin population as our study; thus we speculated that our study based on nearly 60 twin pairs would get a statistical power of about 80%.
Results
Characteristics of Participants
As shown in Table 1, a total of 60 twin pairs for SBP and 58 twin pairs for DBP were included in this study, with the medians of 134.00 mmHg (95% range: 102.05, 184.90) for SBP and 80.00 mmHg (95% range: 62.00, 105.03) for DBP respectively. The mean of BMI was all about 25 kg/m2 (SD 3.5). Most of the clinical indicators, including serum uric acid, fasting glucose, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol and triglyceride showed intrapair statistically correlated, implying the co-twin design beneficial.
Table 1. Basic characteristics of the participants
Note: Continuous variables were presented as mean (standard deviation (SD)) or median (P2.5, P97.5); Categorical variables were presented as numbers with percentages.Intrapair correlations were tested by Pearson or Spearman correlation method.
BMI, body mass index; BP, blood pressure; CHOL, total cholesterol; DBP, diastolic blood pressure; GLU, fasting glucose; HDLC, high-density lipoprotein cholesterol; LDLC, low-density lipoprotein cholesterol; SBP, systolic blood pressure; SUA, serum uric acid; TG, triglyceride
Mediation Analyses
Mediating effect of DNAm on association between BMI and SBP
Association of BMI and SBP
The intrapair BMI difference in twins was positively correlated with intrapair SBP difference (r = .338, p = .010). BMI was positively associated with SBP levels (β = 1.86, p = .0004) in twins using the GEE in Model 1.
Association of BMI and DNAm while controlling for SBP
The Manhattan plot of genomewide DNA methylation analysis on BMI is shown in Figure S1(a). We identified 85 BMI-related top CpGs (p < 1×10−4), of which 18 CpGs showed weaker-than-suggestive significance (1×10−6 ≤ p < 1×10−5) and 5 CpGs reached genomewide significance (FDR < 0.05). In the top 20 CpGs shown in Table S3, the five strongest associations were detected for the CpGs located at CAMK4 (β = −0.74) and OGDHL (β = 0.04−0.05). These top 20 CpGs were located at or near six genes, including CAMK4, OGDHL, KCNQ1, BTBD19, FEZF2 and AXDND1.
We found 11 DMRs potentially related to BMI (Table 2). As illustrated in Figure 1(a), most DMRs were hypermethylated with increasing BMI, whereas the methylation levels of 3 DMRs (DMR-5, 6, 11) were negatively correlated with BMI. But the associations of another DMR (DMR-9) with BMI were uncertain. Interestingly, one DMR within OGDHL also covered the strongest DNAm signals listed in Table S3.
Table 2. The results of annotation to significant differentially methylated regions for BMI
Note: bp, base pair; DMR, differentially methylated region; DBP, diastolic blood pressure; Length, the number of CpGs in each DMR; SBP, systolic blood pressure; slk, Stouffer-Liptak-Kechris; NA, not available.
Figure 1. Circular Manhattan plots for genomewide DNA methylation analysis of BMI. (a) BMI-systolic blood pressure; (b) BMI-diastolic blood pressure. The numbers of chromosome and the −log10 of p values for statistical significance are shown. The dots represent the observed CpGs.
Many GREAT ontology enrichments potentially related to BMI were found for the identified CpGs with p values < .05, including p53 pathway by glucose deprivation, hedgehog signaling pathway, notch signaling pathway and so forth (Figure S2).
Association of DNAm with SBP while controlling for BMI
The DNAm at 9 CpGs mapped to C10orf71-AS1, NDUFB5P1, KRT80, BAI2, ABCA2, PEX11G, and FGF4 was significantly associated with SBP levels (p < .05) in Model 3 (Table S4). These CpGs were mainly involved in voltage-gated chloride channel activity and regulation of cholesterol transport (Figure S3).
Correlation of BMI and SBP while controlling DNAm
The medians of p values for CpGs identified in Model 3 were all more than .05, indicating that BMI was independent of SBP while controlling for DNAm.
Mediation effect of candidate CpG mediators
The results of testing mediation effects are shown in Table 3. We found that two CpGs mapped to ABCA2 could mediate the association between BMI and SBP by using R-package mediation. The DNAm of individual CpG was with an ADE of about 1.97 (95% CI [0.98, 2.96])-mmHg greater SBP levels and an ACME of about 0.24 (95% CI [0.65, 0.01])-mmHg lower SBP levels. However, the total mediation effect (−0.23, 95% CI [−0.66, 0.05]) of these two DNAm mediators did not show statistically significance in the parallel mediation model. Interestingly, BMI also negatively mediated the association between the DNAm of these two CpGs and SBP, with an ACME of −1.6748 (95% CI [−3.6883, −0.2662]) and −2.3493 (95% CI [−4.9707, −0.4547]) respectively (Table S5).
Table 3. DNA methylation mediators for the association of BMI and blood pressure in single and parallel mediation model
Note: DBP, diastolic blood pressure; SBP, systolic blood pressure; β, regression coefficient.
* Given that the direct effects and mediated effects showed the opposite direction, the absolute value of ratio of indirect effect to total effect was calculated.
# BMI also mediated the association between DNAm of CpGs and blood pressure.
† Given the characteristics of PROCESS macro in SPSS program, choosing 10 CpGs in the parallel mediation model on the basis of absolute value of regression coefficients.
Mediation analyses in different sexes
In male twins, BMI was positively associated with SBP level (β = 1.68, p = .002) in Model 1. The association between BMI and DNAm of 261 CpGs reached p < 1×10−5 level in Model 2, of which the DNAm at 99 CpGs was significantly associated with SBP level (p < .05) in Model 3. The CpGs mapped to ZNF579, FOXQ1 and ENSG00000271788 negatively mediated the association between BMI and SBP, with a total negative mediation effect of −2.18 (95% CI [−3.99, −1.07]) (Table S6).
In female twins, BMI was also positively associated with SBP level (β = 2.03, p = .021). The association between BMI and DNAm of 473 CpGs reached p < 1×10−5 level, of which the DNAm of 203 CpGs was significantly associated with SBP level (p < .05). The CpGs within RSPH6A, CPZ, UFC1, RN7SL831P, SMAD6 and TUBB4BP6 could negatively mediate the association between BMI and SBP in female twins, with a total mediation effect of −1.61 (95% CI [−3.54, −0.17]) (Table S6).
Mediating effect of DNAm on association between BMI and DBP
Association of BMI and DBP
The intrapair BMI difference in twins was positively correlated with intrapair DBP difference (r =.400, p = .002). BMI was positively associated with DBP levels (β = 0.60, p = .0495) in twins using the GEE in Model 1.
Association of BMI and DNAm while controlling for DBP
The Manhattan plot of genomewide DNA methylation analysis on BMI is shown in Figure S1(b). We identified 193 BMI-related top CpGs (p < 1×10−4), of which 55 CpGs reaching genomewide significance (FDR < 0.05). As shown in Table S3, the strongest associations were detected for the CpGs located at OGDHL (β = 0.06). The top 20 CpGs were located at or near OGDHL, PEX11G, RSPH6A, RXRA, RNF126, MED27 and POU4F2.
We found 25 DMRs potentially related to BMI (Table 2). As shown in Figure 1(b), most DMRs were hypermethylated with increasing BMI, whereas the methylation levels of seven DMRs (DMR-3, 5, 6, 7, 8, 10, 12) were negatively correlated with BMI. However, the relationship between another five DMRs (DMR-11, 13, 15, 21, 22) and BMI were unclear. Interestingly, two DMRs falling into OGDHL and POU4F2 also covered the top DNAm signals listed in Table S3.
Many GREAT ontology enrichments, such as p53 pathway by glucose deprivation and PI3 kinase pathway, potentially related to BMI, were found for the identified CpGs (p value < .05) (Figure S2).
Association of DNAm with DBP while controlling for BMI
The DNAm at 33 CpGs mapped to ABCA2, ADORA2B, CTNNBIP1, KDM4B, NAA60, RSPH6A, SLC25A19 and STIL, etc. was significantly associated with DBP levels (p < .05) in Model 3 (Table S4). These CpGs were mainly involved in regulation of chronic inflammatory response and ABCA transporters in lipid homeostasis. (Figure S3).
Correlation of BMI and DBP while controlling DNAm
The medians of p values for all CpGs identified in Model 3 were more than .05, indicating that BMI was independent of DBP while controlling for DNAm.
Mediation effect of candidate CpG mediators
As shown in Table 3, the DNAm at 12 CpGs could negatively mediate the association between BMI and DBP, with a total mediation effect of −0.66 (95% CI [−1.07, −0.30]) in the parallel mediation model. The DNAm of 3 CpGs mapped at ABCA2 was with an ACME of about −0.19 (95% CI [−0.45, −0.02]), 2 CpGs mapped at KDM4B with an ACME of about −0.17 (95% CI [−0.39, −0.02]), and 2 CpGs mapped at SLC25A19 with an ACME of about −0.14 (95% CI [−0.34, −0.0006]), and so forth. Interestingly, BMI also negatively mediated the association between the DNAm of most CpGs mentioned above and DBP, with an ACME ranging −0.9111 to −0.3485 (Table S5).
Mediation analyses in different sexes
In male twins, BMI was positively associated with DBP level (β = 1.23, p =.004) in Model 1. The association between BMI and DNAm of 207 CpGs reached p < 1×10−5 level in Model 2, of which the DNAm at 160 CpGs was significantly associated with DBP level (p < .05) in Model 3. Five CpGs mapped to MAU2 and CHRNE negatively mediated the association between BMI and DBP in male twins, with a total negative mediation effect of −0.3 (95% CI [−1.65, −1.42]) (Table S6). However, the association of BMI with DBP level was not statistically significant (β = 0.02, p = .953) in female twins, thus the subsequent mediation analysis was not performed.
Discussion
Obesity is an established risk factor for hypertension. Here, based on a sample of BMI and BP-discordant MZ twins, we observed that the DNAm of 2 and 12 CpGs could separately negatively mediate the association of BMI with SBP and DBP, suggesting the negative mediating role of DNAm variation. Interestingly, we found that BMI could also negatively mediate the association between the DNAm of several CpGs and SBP/DBP, highlighting the need to consider whether exposure and mediator were moderators of each other in epigenetic studies.
We found several CpG mediators annotated to ABCA2, KDM4B, SLC25A19, LMNB2, PTPRN2 and DNASE1 could partially negatively mediate the relationship between BMI and BP. (1) ABCA2. The protein encoded by the ABCA2 gene was involved in cholesterol metabolism and cholesterol transport (Mack et al., Reference Mack, Townsend, Beljanski and Tew2007), while dyslipidemia was an independent risk factor for atherosclerotic cardiovascular disease (Miao et al., Reference Miao, Fiehn, Chen, Zhang, Umans, Lee, Howard, Roman, Devereux and Zhao2023), thus further influencing BP levels. (2) KDM4B. It was reported that the protein encoded by KDM4B in adipose tissues played a critical role in energy balance, oxidation, lipolysis and thermogenesis, and might serve as a therapeutic target for treatment of obesity (Cheng et al., Reference Cheng, Yuan, Vergnes, Rong, Youn, Li, Yu, Liu, Cai, Lin, Tontonoz, Hong, Reue and Wang2018). KDM4B was also involved in estrogen-mediated signaling pathway that played a role underlying the vasculature during hypertension and other cardiovascular diseases (Visniauskas et al., Reference Visniauskas, Kilanowski-Doroh, Ogola, McNally, Horton, Imulinde Sugi and Lindsey2023). (3) SLC25A19. The protein encoded by the SLC25A19 gene was a member of the solute carrier family and was involved in vitamin B1 metabolism process, while the vitamin B1 level was reported to be negatively related to obesity (Fu et al., Reference Fu, Zhu, Huang, He, Zhang, Li, Tan and Rong2023). Moreover, the vitamin B1 deficiency was also related to some cardiovascular traits, such as elevated cardiac stroke volume, decreased vascular resistance, and elevated pressor responses, suggesting a possible role in influencing BP levels (DiNicolantonio et al., Reference DiNicolantonio, Liu and O’Keefe2018). (4) LMNB2. The role of LMNB2 protein in regulating polyploidization of cardiomyocyte nuclei and myocardial regeneration has been reported in an animal experiment (Han et al., Reference Han, Choudhury, Mich-Basso, Ammanamanchi, Ganapathy, Suresh, Khaladkar, Singh, Maehr, Zuppo, Kim, Eberwine, Wyman, Wu and Kühn2020). The RNA m6A methylation of LMNB2 was reported to be strongly associated with diabetes (C. Chen et al., Reference Chen, Xiang, Liu, Liang, Yang and Tao2021), but the association of LMNB2 with obesity is still unknown. (5) PTPRN2. This gene encodes a protein with a sequence similarity to receptor-like protein tyrosine phosphatases. It was reported that the PTPRN2 in the blood of newborn term infants born to mothers with obesity showed differential methylation compared to infants from mothers with a normal BMI (Sasaki et al., Reference Sasaki, Murphy, Briollais, McGowan and Matthews2022). In addition, some DNAm variants of PTPRN2 might be related to the renin-angiotensin-aldosterone system (RAAS), which plays a role in the pathophysiology underlying hypertension (van der Linden et al., Reference van der Linden, Halley, Meeks, Chilunga, Hayfron-Benjamin, Venema, Garrelds, Danser, van den Born, Henneman and Agyemang2022). Moreover, the variants at the PTPRN2 were also suggested to be associated with persistent pulmonary hypertension of the newborn (PPHN) (Nakwan et al., Reference Nakwan, Kunhapan, Chaiyasung, Satproedprai, Singkhamanan, Mahasirimongkol and Charalsawadi2023). (6) DNASE1. The study on the role of DNASE1 gene in the pathophysiology underlying obesity or hypertension was limited. It might be that DNASE1 was involved in regulation of acute inflammatory response (GO:0002673 in AmiGO 2 database) and then induced the occurrence of diseases. However, the association of TFCP2L1 and C10orf71-AS1 with obesity or hypertension is still unknown. In sum, the DNAm variation of most of the genes we observed might potentially negatively mediate the association of BMI with BP.
A major strength of our study is the implementation of a trait-discordant MZ twin model that was proven to be a powerful tool for methylation studies in linking the environmental basis of epigenetic modification variation to certain traits while controlling for individual genetic makeup (W. Li et al., Reference Li, Christiansen, Hjelmborg, Baumbach and Tan2018; Tan et al., Reference Tan, Christiansen, Hjelmborg and Christensen2015). Although a majority of the participants in our study did not have obesity or hypertension, we still observed certain CpG mediators that effectively associated BMI with BP. Additionally, considering that the genetic constitutions, environmental exposures and a multitude of lifestyles in various ethnic populations differed worldwide, our findings will specifically help elucidate the underlying pathogenesis of increased BP levels and hypertension in the Chinese population.
Several limitations should also be mentioned for this study. First, our sample size was relatively limited due to the challenges of recruiting and identifying qualified twin pairs. However, the trait-discordant twin design we adopted had a greater statistical power over the traditional cross-sectional or case-control design. For BMI and BP with a moderate and high heritability (Duan et al., Reference Duan, Pang, Zhang, Li, Kruse, Kyvik, Christensen and Tan2011; Wang et al., Reference Wang, Zhang, Xu, Wu, Duan, Li and Tan2018), this design would allow for large sample size reductions compared to the traditional designs according to the computer simulation study we published (W. Li et al., Reference Li, Christiansen, Hjelmborg, Baumbach and Tan2018). Hence, our EWAS had sufficient power of about 80% for detecting the CpG mediators. We will further validate the CpG mediators, genes and biological pathways in a community population. Second, because we performed this study in BMI- and BP-discordant MZ twins rather than obesity- or hypertension-discordant MZ twins, the DNAm mediators we found potentially played a negative mediating effect between BMI and BP. Third, we did not find any replicated results while comparing the CpGs identified in our study with other similar methylation studies because the latter were very limited in number.
In summary, we observed that BMI might interact with DNAm at specific loci, mediating the association with SBP/DBP. Our findings provide important clues to further elucidate the pathogenesis of elevated BP levels and helped to identify new diagnostic biomarkers and therapeutic targets for hypertension.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/thg.2024.3.
Acknowledgements
None.
Author contributions
Jie Yao: Methodology, Formal analysis, Writing-Original Draft. Feng Ning: Investigation, Resources, Data curation. Weijing Wang: Conceptualization, Methodology, Software, Funding acquisition. Dongfeng Zhang: Conceptualization, Methodology, Supervision, Funding acquisition.
Funding
This work is funded by the Natural Science Foundation of Shandong Province (ZR2020QH304), Qingdao Science and Technology Fund (21-1-4-rkjk-1-nsh), and China Postdoctoral Science Foundation (2020M682129).
Competing interests
The authors declare no competing financial interests.
