An Epigenome-Wide Association Study of Obesity-Related Traits

An Epigenome-Wide Association Study of Obesity-Related Traits Abstract We conducted an epigenome-wide association study on obesity-related traits. We used data from 2 prospective, population-based cohort studies: the Rotterdam Study (RS) (2006–2013) and the Atherosclerosis Risk in Communities (ARIC) Study (1990–1992). We used the RS (n = 1,450) as the discovery panel and the ARIC Study (n = 2,097) as the replication panel. Linear mixed-effect models were used to assess the cross-sectional associations between genome-wide DNA methylation in leukocytes and body mass index (BMI) and waist circumference (WC), adjusting for sex, age, smoking, leukocyte proportions, array number, and position on array. The latter 2 variables were modeled as random effects. Fourteen 5′-C-phosphate-G-3′ (CpG) sites were associated with BMI and 26 CpG sites with WC in the RS after Bonferroni correction (P < 1.07 × 10−7), of which 12 and 13 CpGs were replicated in the ARIC Study, respectively. The most significant novel CpGs were located on the Musashi RNA binding protein 2 gene (MSI2; cg21139312) and the leucyl-tRNA synthetase 2, mitochondrial gene (LARS2; cg18030453) and were associated with both BMI and WC. CpGs at BRDT, PSMD1, IFI44L, MAP1A, and MAP3K5 were associated with BMI. CpGs at LGALS3BP, MAP2K3, DHCR24, CPSF4L, and TMEM49 were associated with WC. We report novel associations between methylation at MSI2 and LARS2 and obesity-related traits. These results provide further insight into mechanisms underlying obesity-related traits, which can enable identification of new biomarkers in obesity-related chronic diseases. body mass index, cohort studies, DNA methylation, epigenome-wide association studies, obesity, waist circumference Obesity is an important risk factor for cardiovascular disease, diabetes, some cancers, and musculoskeletal disorders (1–3). Evidence suggests that obesity not only is dependent on lifestyle factors but is a result of interactions between genes and lifestyle (4, 5). Epigenetics has been proposed as a molecular mechanism that can affect the expression of genes by environmental influences and potentially could describe further the link between obesity and its complications (6). Nevertheless, unlike genetics, DNA methylation is dynamic over time; therefore, change in DNA methylation could also be a consequence of obesity. Epigenetics is the study of heritable variation in gene function that is not a result of a change in DNA sequence (7). One of the best studied epigenetic mechanisms is DNA methylation, the attachment of a methyl group to a cytosine nucleotide of 5′-C-phosphate-G-3′ (CpG) dinucleotides. DNA methylation has varying functions at different locations in the human genome, including regulation of gene expression (8). To date, epigenome-wide association studies (EWAS) have identified several differentially methylated CpG regions related to body mass index (BMI)—the most widely used measure of obesity—and waist circumference (WC) (9–12). These few studies were performed in either patient populations, specific ethnic groups, or young adults. However, information on the older adults from population-based studies is scarce. In older adults and elderly persons, biological mechanisms involved in body weight and body composition may be different from those in younger adults (13). Therefore, it is crucial to explore the relationship of obesity to epigenetic variation in older adults. We performed a cross-sectional EWAS of DNA methylation in blood leukocytes for BMI and WC in subjects from the Rotterdam Study (RS) and replicated our findings in the Atherosclerosis Risk in Communities (ARIC) Study. METHODS Study population The RS is a large prospective, population-based cohort study aimed at assessing the occurrence of and risk factors for chronic diseases (cardiovascular, endocrine, hepatic, neurological, ophthalmic, psychiatric, dermatological, oncological, and respiratory) in the elderly (14). The study comprises 14,926 subjects in total, living in the well-defined Ommoord district in the city of Rotterdam, the Netherlands. In 1989, the first RS cohort, Rotterdam Study I (RS-I), was established and comprised 7,983 subjects aged 55 years or above. In 2000, the second cohort, RS-II, was established, with 3,011 subjects who had reached the age of 55 years since 1989. In 2006, the third cohort, RS-III, was further included, with 3,932 subjects aged 45 years or above. The discovery panel for the current analysis consisted of a random sample of 1,450 participants from the first and second study visits of the third cohort (RS-III-1, RS-III-2) and the third study visit of the second cohort (RS-II-3). We sought replication of the identified CpG sites in the ARIC Study, which is described in detail elsewhere (15). Briefly, the ARIC Study is a prospective cohort study of cardiovascular disease in adults. Between 1987 and 1989, 7,082 men and 8,710 women aged 45–64 years were recruited from 4 US communities (15). Methylation data were available in a subset of 2,097 African-American participants (10). The RS and ARIC Study protocols were approved by institutional review boards at each participating university, and all participants provided written informed consent. Anthropometric measures and covariates Height and weight were measured with the participant standing without shoes and heavy outer garments. WC was measured at the level midway between the lower rib margin and the iliac crest, with the participant in a standing position without heavy outer garments and with emptied pockets, breathing out gently. Hip circumference was recorded as the maximum circumference of the hips over the buttocks. BMI was calculated as weight (kg) divided by the square of height (m), and waist:hip ratio was calculated as WC divided by hip circumference (16). Information on current and past smoking behavior was acquired from questionnaires. DNA methylation data DNA was extracted from whole peripheral blood (stored in ethylenediaminetetraacetic acid tubes) by means of standardized salting-out methods. Genome-wide DNA methylation levels were measured using the Illumina Infinium Human Methylation 450K BeadChip array (Illumina, Inc., San Diego, California) (17). In short, samples (500 ng of DNA per sample) were first treated with bisulfite using the Zymo EZ-96 DNA methylation kit (Zymo Research, Irvine, California). Next, samples were hybridized to the arrays according to the manufacturer’s protocol. The methylation proportion of a CpG site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). Data preprocessing was additionally performed in both data sets using an R programming pipeline that is based on the pipeline developed by Touleimat and Tost (18), which includes additional parameters and options for preprocessing and normalizing methylation data directly from image data (IDAT) files. We excluded probes which had a detection P value greater than 0.01 in more than 95% of samples. We excluded 11,648 probes at X and Y chromosomes to avoid sex bias. The raw β values were then background-corrected and normalized using the “dasen” option in the wateRmelon R package (R Foundation for Statistical Computing, Vienna, Austria) (19). Per individual probe, participants with methylation levels higher than 3 times the interquartile range were excluded. Statistical analyses The characteristics of the discovery and replication population are presented as mean values for continuous variables and proportions for categorical variables. In the discovery stage, we modeled cross-sectional associations between “dasen”-normalized β values of the CpG sites as the outcome and BMI or WC as the exposure, using linear mixed-effect models adjusting for age, sex, smoking, white blood cell proportion, array number (65 arrays), and position on array (12 positions; a combination of row number and column number). We performed an independent analysis in persons from RS-III-1, RS-III-2, and RS-II-3. We then performed a fixed-effects meta-analysis on the estimates for these 3 RS groups using the inverse-variance–weighted method implemented in METAL, combining the RS-III-1 participants with the RS-III-2 and RS-II-3 participants (20). Technical covariates (array number and position on array) were modeled as random effects. For the RS-III-1 group, we estimated leukocyte proportions (B cells, CD4-positive T cells, CD8-positive T cells, granulocytes, monocytes, and natural killer cells) by means of a formula developed by Houseman et al. (21) and implemented in the minfi package in R (22). For the RS-II-3 and RS-III-2 groups, we used white blood cell counts (i.e., lymphocytes, monocytes, and granulocytes), which were assessed with a Coulter Ac·T diff2 Hematology Analyzer (Beckman Coulter, Brea, California). We corrected for multiple testing using a robust Bonferroni-corrected P value of 1.07 × 10−7 as the threshold for significance (0.05/463,456 probes). The probes identified in the discovery analysis were tested for replication in the independent samples from the ARIC Study. A Bonferroni-corrected P value of 0.05 divided by the number of significant findings in the discovery study was used as a threshold level of significant replication. Finally, we checked all identified CpG sites for cross-reaction or polymorphism (23). A CpG site was considered polymorphic when a single nucleotide polymorphism with a minor allele frequency greater than 0.01 resided at the position of the cytosine or guanine nucleotide or within 10 base pairs of the CpG site within the probe binding site (24). Methylation risk score A methylation risk score was calculated based on CpG sites that were associated with the phenotypes. The effect estimates were used to build the methylation risk score using data from the discovery panel. Linear regression analyses were performed using BMI or WC as the outcome variable and the included CpG sites as exposure variables. With the use of linear regression models, we calculated the variance in lipid levels explained by the methylation risk score. RESULTS Table 1 summarizes the characteristics of participants in the studies. The RS is entirely comprised of Europeans, whereas the ARIC Study participants comprised only African Americans. Compared with RS participants (mean age = 63.7 (standard deviation, 8.1) years), the participants in ARIC were younger, on average (mean age = 56.2 (standard deviation, 5.7) years) and comprised more women (64% in ARIC vs. 56% in the RS)). The respective mean values for BMI in the RS and the ARIC Study were 27.7 and 30.1. The mean values for WC were 93.7 cm and 101.3 cm in the RS and ARIC, respectively. Table 1. Characteristics of Participants in the Rotterdam Study (Rotterdam, the Netherlands; 2006–2013) and the Atherosclerosis Risk in Communities Study (4 US Communities; 1990–1992) Selected for an Epigenome-Wide Association Study on Obesity-Related Traits Characteristic RS (n = 1,450) ARIC Study (n = 2,097) No. of Persons % No. of Persons % Age, yearsa 63.7 (8.1) 56.2 (5.7) Female sex 811 55.9 1,334 63.6 Race, %  White 1,450 100 0 0  African-American 0 0 2,097 100 BMIa,b,c 27.7 (4.4) 30.1 (6.1) BMI statusc  Normal-weight 419 28.9 388 18.5  Overweight 674 46.5 788 37.6  Obese 357 24.6 918 43.8 WC, cma 93.7 (12.9) 101.3 (15.1) Smoking status  Current smoker 273 18.8 512 24.4  Current nonsmoker 1,177 81.2 1,585 75.6 Diabetes mellitus 160 11 545 26.0 Characteristic RS (n = 1,450) ARIC Study (n = 2,097) No. of Persons % No. of Persons % Age, yearsa 63.7 (8.1) 56.2 (5.7) Female sex 811 55.9 1,334 63.6 Race, %  White 1,450 100 0 0  African-American 0 0 2,097 100 BMIa,b,c 27.7 (4.4) 30.1 (6.1) BMI statusc  Normal-weight 419 28.9 388 18.5  Overweight 674 46.5 788 37.6  Obese 357 24.6 918 43.8 WC, cma 93.7 (12.9) 101.3 (15.1) Smoking status  Current smoker 273 18.8 512 24.4  Current nonsmoker 1,177 81.2 1,585 75.6 Diabetes mellitus 160 11 545 26.0 Abbreviations: ARIC, Atherosclerosis Risk in Communities; BMI, body mass index; RS, Rotterdam Study; WC, waist circumference. a Data are presented as mean (standard deviation). b BMI was calculated as weight (kg)/height (m)2. c Data on BMI were missing for 3 ARIC participants. Table 1. Characteristics of Participants in the Rotterdam Study (Rotterdam, the Netherlands; 2006–2013) and the Atherosclerosis Risk in Communities Study (4 US Communities; 1990–1992) Selected for an Epigenome-Wide Association Study on Obesity-Related Traits Characteristic RS (n = 1,450) ARIC Study (n = 2,097) No. of Persons % No. of Persons % Age, yearsa 63.7 (8.1) 56.2 (5.7) Female sex 811 55.9 1,334 63.6 Race, %  White 1,450 100 0 0  African-American 0 0 2,097 100 BMIa,b,c 27.7 (4.4) 30.1 (6.1) BMI statusc  Normal-weight 419 28.9 388 18.5  Overweight 674 46.5 788 37.6  Obese 357 24.6 918 43.8 WC, cma 93.7 (12.9) 101.3 (15.1) Smoking status  Current smoker 273 18.8 512 24.4  Current nonsmoker 1,177 81.2 1,585 75.6 Diabetes mellitus 160 11 545 26.0 Characteristic RS (n = 1,450) ARIC Study (n = 2,097) No. of Persons % No. of Persons % Age, yearsa 63.7 (8.1) 56.2 (5.7) Female sex 811 55.9 1,334 63.6 Race, %  White 1,450 100 0 0  African-American 0 0 2,097 100 BMIa,b,c 27.7 (4.4) 30.1 (6.1) BMI statusc  Normal-weight 419 28.9 388 18.5  Overweight 674 46.5 788 37.6  Obese 357 24.6 918 43.8 WC, cma 93.7 (12.9) 101.3 (15.1) Smoking status  Current smoker 273 18.8 512 24.4  Current nonsmoker 1,177 81.2 1,585 75.6 Diabetes mellitus 160 11 545 26.0 Abbreviations: ARIC, Atherosclerosis Risk in Communities; BMI, body mass index; RS, Rotterdam Study; WC, waist circumference. a Data are presented as mean (standard deviation). b BMI was calculated as weight (kg)/height (m)2. c Data on BMI were missing for 3 ARIC participants. Tables 2 and 3 list the CpG sites associated with BMI and WC in both populations. Using the Bonferroni-corrected statistical significance level of 1.07 × 10−7, we identified 14 CpG sites associated with BMI (see Web Table 1, available at https://academic.oup.com/aje) and 26 CpG sites associated with WC (Web Table 2) in the RS. In the ARIC Study, we successfully replicated 12 of the 14 BMI-related CpG sites (P < 3.57 × 10−3) (Table 2) and 13 of the 26 WC-related CpG sites (P < 1.92 × 10−3) (Table 3). Among these, 8 BMI-related CpG sites and 11 WC-related CpG sites were novel. The most significant novel CpG sites were located on the Musashi RNA binding protein 2 gene (MSI2; cg21139312) and the leucyl-tRNA synthetase 2, mitochondrial gene (LARS2; cg18030453) for both BMI and WC. For every unit increase in BMI (kg/m2), there were increases in MSI2 and LARS2 methylation of 0.0009 (P = 4.5 × 10−10) and 0.0009 (P = 4.5 × 10−9), respectively. For every unit increase in WC (cm), there were increases in MSI2 and LARS2 methylation of 0.0004 (P = 5.9 × 10−12) and 0.0003 (P = 8.8 × 10−8), respectively. Additionally, for BMI, other novel CpG sites were located in the bromodomain testis associated gene (BRDT; cg03421440) and the microtubule associated protein 1A gene (MAP1A; cg15159104). For WC, the other top novel CpG sites were located in the transmembrane protein 49 gene (TMEM49; cg24174557) and the galectin 3 binding protein gene (LGALS3BP; cg04927537). Table 2. 5′-C-Phosphate-G-3′ Methylation Sites Associated With Body Mass Indexa at the Level of Genome-Wide Significance in the Rotterdam Study (2006–2013) and Successfully Replicated in the Atherosclerosis Risk in Communities Study (1990–1992) Probe ID No. Chromosome Gene Mean (SD) Methylationb Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βc P Valued βc P Valued cg00574958 11 CPT1A 0.19 (0.04) −0.0011 6.2 × 10−15 −0.0029 3.2 × 10−12 1.9 cg00851028 1 N/A 0.72 (0.04) 0.0010 5.4 × 10−8 0.0038 9.0 × 10−4 0.4 cg03421440 1 BRDT 0.71 (0.07) −0.0015 3.2 × 10−8 −0.0043 1.3 × 10−3 0.3 cg06096336 2 PSMD1 0.64 (0.05) 0.0016 4.3 × 10−8 0.0058 5.5 × 10−4 1.0 cg06500161 21 ABCG1 0.71 (0.03) 0.0011 1.7 × 10−9 0.0081 1.5 × 10−13 0.3 cg06872964 1 IFI44L 0.62 (0.06) 0.0015 4.8 × 10−8 0.0100 4.3 × 10−7 0.4 cg11024682 17 SREBF1 0.55 (0.04) 0.0013 6.6 × 10−15 0.0068 9.6 × 10−9 0.1 cg15159104 15 MAP1A 0.48 (0.05) 0.0010 3.2 × 10−8 0.0048 5.1 × 10−6 0.0 cg15903032 10 N/A 0.57 (0.04) 0.0010 7.6 × 10−8 0.0037 2.8 × 10−3 0.2 cg18030453 3 LARS2 0.72 (0.04) 0.0009 4.5 × 10−9 0.0028 1.7 × 10−3 0.1 cg21139312 17 MSI2 0.89 (0.03) 0.0009 4.5 × 10−10 0.0028 1.2 × 10−6 2.0 cg21506299 6 MAP3K5 0.23 (0.06) −0.0010 3.5 × 10−8 −0.0019 2.8 × 10−3 1.1 Probe ID No. Chromosome Gene Mean (SD) Methylationb Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βc P Valued βc P Valued cg00574958 11 CPT1A 0.19 (0.04) −0.0011 6.2 × 10−15 −0.0029 3.2 × 10−12 1.9 cg00851028 1 N/A 0.72 (0.04) 0.0010 5.4 × 10−8 0.0038 9.0 × 10−4 0.4 cg03421440 1 BRDT 0.71 (0.07) −0.0015 3.2 × 10−8 −0.0043 1.3 × 10−3 0.3 cg06096336 2 PSMD1 0.64 (0.05) 0.0016 4.3 × 10−8 0.0058 5.5 × 10−4 1.0 cg06500161 21 ABCG1 0.71 (0.03) 0.0011 1.7 × 10−9 0.0081 1.5 × 10−13 0.3 cg06872964 1 IFI44L 0.62 (0.06) 0.0015 4.8 × 10−8 0.0100 4.3 × 10−7 0.4 cg11024682 17 SREBF1 0.55 (0.04) 0.0013 6.6 × 10−15 0.0068 9.6 × 10−9 0.1 cg15159104 15 MAP1A 0.48 (0.05) 0.0010 3.2 × 10−8 0.0048 5.1 × 10−6 0.0 cg15903032 10 N/A 0.57 (0.04) 0.0010 7.6 × 10−8 0.0037 2.8 × 10−3 0.2 cg18030453 3 LARS2 0.72 (0.04) 0.0009 4.5 × 10−9 0.0028 1.7 × 10−3 0.1 cg21139312 17 MSI2 0.89 (0.03) 0.0009 4.5 × 10−10 0.0028 1.2 × 10−6 2.0 cg21506299 6 MAP3K5 0.23 (0.06) −0.0010 3.5 × 10−8 −0.0019 2.8 × 10−3 1.1 Abbreviations: ABCG1, ATP binding cassette, subfamily G, member 1 gene; ARIC, Atherosclerosis Risk in Communities; ATP, adenosine triphosphate; BMI, body mass index; BRDT, bromodomain testis associated gene; CPT1A, carnitine palmitoyltransferase 1A gene; ID, identification; IFI44L, interferon induced protein 44 like gene; LARS2, leucyl-tRNA synthetase 2, mitochondrial gene; MAP1A, microtubule associated protein 1A gene; MAP3K5, mitogen-activated protein kinase kinase 5 gene; MSI2, Musashi RNA binding protein 2 gene; N/A, not annotated; PSMD1, proteasome 26s subunit, non-ATPase 1 gene; RS, Rotterdam Study; SD, standard deviation; SREBF1, sterol regulatory element binding transcription factor 1 gene. a BMI was calculated as weight (kg)/height (m)2. b The methylation proportion of a 5′-C-phosphate-G-3′ site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). c Regression coefficient based on a linear mixed model and reflecting difference in the methylation β value per unit increase in BMI (kg/m2). Models adjusted for age, sex, current smoking, leukocyte proportions, array number, and position on array. d In the RS we corrected for multiple testing using a robust Bonferroni-corrected P value of 1.07 × 10−7, and in the ARIC Study the level of significance for replication was P < 3.57 × 10−3. Table 2. 5′-C-Phosphate-G-3′ Methylation Sites Associated With Body Mass Indexa at the Level of Genome-Wide Significance in the Rotterdam Study (2006–2013) and Successfully Replicated in the Atherosclerosis Risk in Communities Study (1990–1992) Probe ID No. Chromosome Gene Mean (SD) Methylationb Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βc P Valued βc P Valued cg00574958 11 CPT1A 0.19 (0.04) −0.0011 6.2 × 10−15 −0.0029 3.2 × 10−12 1.9 cg00851028 1 N/A 0.72 (0.04) 0.0010 5.4 × 10−8 0.0038 9.0 × 10−4 0.4 cg03421440 1 BRDT 0.71 (0.07) −0.0015 3.2 × 10−8 −0.0043 1.3 × 10−3 0.3 cg06096336 2 PSMD1 0.64 (0.05) 0.0016 4.3 × 10−8 0.0058 5.5 × 10−4 1.0 cg06500161 21 ABCG1 0.71 (0.03) 0.0011 1.7 × 10−9 0.0081 1.5 × 10−13 0.3 cg06872964 1 IFI44L 0.62 (0.06) 0.0015 4.8 × 10−8 0.0100 4.3 × 10−7 0.4 cg11024682 17 SREBF1 0.55 (0.04) 0.0013 6.6 × 10−15 0.0068 9.6 × 10−9 0.1 cg15159104 15 MAP1A 0.48 (0.05) 0.0010 3.2 × 10−8 0.0048 5.1 × 10−6 0.0 cg15903032 10 N/A 0.57 (0.04) 0.0010 7.6 × 10−8 0.0037 2.8 × 10−3 0.2 cg18030453 3 LARS2 0.72 (0.04) 0.0009 4.5 × 10−9 0.0028 1.7 × 10−3 0.1 cg21139312 17 MSI2 0.89 (0.03) 0.0009 4.5 × 10−10 0.0028 1.2 × 10−6 2.0 cg21506299 6 MAP3K5 0.23 (0.06) −0.0010 3.5 × 10−8 −0.0019 2.8 × 10−3 1.1 Probe ID No. Chromosome Gene Mean (SD) Methylationb Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βc P Valued βc P Valued cg00574958 11 CPT1A 0.19 (0.04) −0.0011 6.2 × 10−15 −0.0029 3.2 × 10−12 1.9 cg00851028 1 N/A 0.72 (0.04) 0.0010 5.4 × 10−8 0.0038 9.0 × 10−4 0.4 cg03421440 1 BRDT 0.71 (0.07) −0.0015 3.2 × 10−8 −0.0043 1.3 × 10−3 0.3 cg06096336 2 PSMD1 0.64 (0.05) 0.0016 4.3 × 10−8 0.0058 5.5 × 10−4 1.0 cg06500161 21 ABCG1 0.71 (0.03) 0.0011 1.7 × 10−9 0.0081 1.5 × 10−13 0.3 cg06872964 1 IFI44L 0.62 (0.06) 0.0015 4.8 × 10−8 0.0100 4.3 × 10−7 0.4 cg11024682 17 SREBF1 0.55 (0.04) 0.0013 6.6 × 10−15 0.0068 9.6 × 10−9 0.1 cg15159104 15 MAP1A 0.48 (0.05) 0.0010 3.2 × 10−8 0.0048 5.1 × 10−6 0.0 cg15903032 10 N/A 0.57 (0.04) 0.0010 7.6 × 10−8 0.0037 2.8 × 10−3 0.2 cg18030453 3 LARS2 0.72 (0.04) 0.0009 4.5 × 10−9 0.0028 1.7 × 10−3 0.1 cg21139312 17 MSI2 0.89 (0.03) 0.0009 4.5 × 10−10 0.0028 1.2 × 10−6 2.0 cg21506299 6 MAP3K5 0.23 (0.06) −0.0010 3.5 × 10−8 −0.0019 2.8 × 10−3 1.1 Abbreviations: ABCG1, ATP binding cassette, subfamily G, member 1 gene; ARIC, Atherosclerosis Risk in Communities; ATP, adenosine triphosphate; BMI, body mass index; BRDT, bromodomain testis associated gene; CPT1A, carnitine palmitoyltransferase 1A gene; ID, identification; IFI44L, interferon induced protein 44 like gene; LARS2, leucyl-tRNA synthetase 2, mitochondrial gene; MAP1A, microtubule associated protein 1A gene; MAP3K5, mitogen-activated protein kinase kinase 5 gene; MSI2, Musashi RNA binding protein 2 gene; N/A, not annotated; PSMD1, proteasome 26s subunit, non-ATPase 1 gene; RS, Rotterdam Study; SD, standard deviation; SREBF1, sterol regulatory element binding transcription factor 1 gene. a BMI was calculated as weight (kg)/height (m)2. b The methylation proportion of a 5′-C-phosphate-G-3′ site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). c Regression coefficient based on a linear mixed model and reflecting difference in the methylation β value per unit increase in BMI (kg/m2). Models adjusted for age, sex, current smoking, leukocyte proportions, array number, and position on array. d In the RS we corrected for multiple testing using a robust Bonferroni-corrected P value of 1.07 × 10−7, and in the ARIC Study the level of significance for replication was P < 3.57 × 10−3. Table 3. 5′-C-Phosphate-G-3′ Methylation Sites Associated With Waist Circumference at the Level of Genome-Wide Significance in the Rotterdam Study (2006–2013) and Successfully Replicated in the Atherosclerosis Risk in Communities Study (1990–1992) Probe ID No. Chromosome Gene Mean (SD) Methylationa Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βb P Valuec βb P Valuec cg00574958 11 CPT1A 0.19 (0.04) −0.0005 1.2 × 10−17 −0.0034 5.8 × 10−17 3.3 cg00851028 1 N/A 0.72 (0.04) 0.0004 6.0 × 10−9 0.0043 1.2 × 10−4 0 cg04927537 17 LGALS3BP 0.57 (0.05) 0.0006 7.0 × 10−8 0.0093 7.0 × 10−8 1.6 cg05899984 12 N/A 0.84 (0.03) 0.0003 8.1 × 10−8 0.0038 5.7 × 10−6 2.9 cg06500161 21 ABCG1 0.71 (0.03) 0.0005 2.4 × 10−12 0.0096 4.4 × 10−19 0.8 cg11024682 17 SREBF1 0.55 (0.04) 0.0005 2.9 × 10−15 0.0080 3.5 × 10−12 1.2 cg13139542 2 N/A 0.89 (0.02) 0.0002 6.0 × 10−8 0.0029 4.7 × 10−6 0 cg15416179 17 MAP2K3 0.14 (0.03) −0.0002 9.1 × 10−8 −0.0019 2.6 × 10−4 3.6 cg17901584 1 DHCR24 0.68 (0.07) −0.0005 1.7 × 10−8 −0.0080 8.3 × 10−8 2.0 cg18030453 3 LARS2 0.72 (0.04) 0.0003 8.8 × 10−8 0.0029 8.6 × 10−4 0 cg18772573 17 CPSF4L 0.85 (0.03) 0.0003 7.3 × 10−8 0.0039 2.8 × 10−5 0 cg21139312 17 MSI2 0.89 (0.03) 0.0004 5.9 × 10−12 0.0028 6.1 × 10−7 8.2 cg24174557 17 TMEM49 0.38 (0.07) −0.0005 1.1 × 10−8 −0.0059 5.3 × 10−5 0 Probe ID No. Chromosome Gene Mean (SD) Methylationa Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βb P Valuec βb P Valuec cg00574958 11 CPT1A 0.19 (0.04) −0.0005 1.2 × 10−17 −0.0034 5.8 × 10−17 3.3 cg00851028 1 N/A 0.72 (0.04) 0.0004 6.0 × 10−9 0.0043 1.2 × 10−4 0 cg04927537 17 LGALS3BP 0.57 (0.05) 0.0006 7.0 × 10−8 0.0093 7.0 × 10−8 1.6 cg05899984 12 N/A 0.84 (0.03) 0.0003 8.1 × 10−8 0.0038 5.7 × 10−6 2.9 cg06500161 21 ABCG1 0.71 (0.03) 0.0005 2.4 × 10−12 0.0096 4.4 × 10−19 0.8 cg11024682 17 SREBF1 0.55 (0.04) 0.0005 2.9 × 10−15 0.0080 3.5 × 10−12 1.2 cg13139542 2 N/A 0.89 (0.02) 0.0002 6.0 × 10−8 0.0029 4.7 × 10−6 0 cg15416179 17 MAP2K3 0.14 (0.03) −0.0002 9.1 × 10−8 −0.0019 2.6 × 10−4 3.6 cg17901584 1 DHCR24 0.68 (0.07) −0.0005 1.7 × 10−8 −0.0080 8.3 × 10−8 2.0 cg18030453 3 LARS2 0.72 (0.04) 0.0003 8.8 × 10−8 0.0029 8.6 × 10−4 0 cg18772573 17 CPSF4L 0.85 (0.03) 0.0003 7.3 × 10−8 0.0039 2.8 × 10−5 0 cg21139312 17 MSI2 0.89 (0.03) 0.0004 5.9 × 10−12 0.0028 6.1 × 10−7 8.2 cg24174557 17 TMEM49 0.38 (0.07) −0.0005 1.1 × 10−8 −0.0059 5.3 × 10−5 0 Abbreviations: ABCG1, ATP binding cassette, subfamily G, member 1; ARIC, Atherosclerosis Risk in Communities; ATP, adenosine triphosphate; CPSF4L, cleavage and polyadenylation specific factor 4 like gene; CPT1A, carnitine palmitoyltransferase 1A gene; DHCR24, 24-dehydrocholesterol reductase gene; ID, identification; LARS2, leucyl-TRNA synthetase 2, mitochondrial gene; LGALS3BP, galectin 3 binding protein gene; MAP2K3, mitogen-activated protein kinase kinase 3 gene; MSI2, Musashi RNA binding protein 2 gene; N/A, not annotated; RS, Rotterdam Study; SD, standard deviation; SREBF1, sterol regulatory element binding transcription factor 1 gene; TMEM49, transmembrane protein 49 gene. a The methylation proportion of a 5′-C-phosphate-G-3′ site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). b Regression coefficient based on a linear mixed model and reflecting difference in the methylation β value per unit increase in waist circumference (cm). Models adjusted for age, sex, current smoking, leukocyte proportions, array number, and position on array. c In the RS we corrected for multiple testing using a robust Bonferroni-corrected P value of 1.08 × 10−7, and in the ARIC Study the level of significance for replication was P < 1.92 × 10−3. Table 3. 5′-C-Phosphate-G-3′ Methylation Sites Associated With Waist Circumference at the Level of Genome-Wide Significance in the Rotterdam Study (2006–2013) and Successfully Replicated in the Atherosclerosis Risk in Communities Study (1990–1992) Probe ID No. Chromosome Gene Mean (SD) Methylationa Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βb P Valuec βb P Valuec cg00574958 11 CPT1A 0.19 (0.04) −0.0005 1.2 × 10−17 −0.0034 5.8 × 10−17 3.3 cg00851028 1 N/A 0.72 (0.04) 0.0004 6.0 × 10−9 0.0043 1.2 × 10−4 0 cg04927537 17 LGALS3BP 0.57 (0.05) 0.0006 7.0 × 10−8 0.0093 7.0 × 10−8 1.6 cg05899984 12 N/A 0.84 (0.03) 0.0003 8.1 × 10−8 0.0038 5.7 × 10−6 2.9 cg06500161 21 ABCG1 0.71 (0.03) 0.0005 2.4 × 10−12 0.0096 4.4 × 10−19 0.8 cg11024682 17 SREBF1 0.55 (0.04) 0.0005 2.9 × 10−15 0.0080 3.5 × 10−12 1.2 cg13139542 2 N/A 0.89 (0.02) 0.0002 6.0 × 10−8 0.0029 4.7 × 10−6 0 cg15416179 17 MAP2K3 0.14 (0.03) −0.0002 9.1 × 10−8 −0.0019 2.6 × 10−4 3.6 cg17901584 1 DHCR24 0.68 (0.07) −0.0005 1.7 × 10−8 −0.0080 8.3 × 10−8 2.0 cg18030453 3 LARS2 0.72 (0.04) 0.0003 8.8 × 10−8 0.0029 8.6 × 10−4 0 cg18772573 17 CPSF4L 0.85 (0.03) 0.0003 7.3 × 10−8 0.0039 2.8 × 10−5 0 cg21139312 17 MSI2 0.89 (0.03) 0.0004 5.9 × 10−12 0.0028 6.1 × 10−7 8.2 cg24174557 17 TMEM49 0.38 (0.07) −0.0005 1.1 × 10−8 −0.0059 5.3 × 10−5 0 Probe ID No. Chromosome Gene Mean (SD) Methylationa Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βb P Valuec βb P Valuec cg00574958 11 CPT1A 0.19 (0.04) −0.0005 1.2 × 10−17 −0.0034 5.8 × 10−17 3.3 cg00851028 1 N/A 0.72 (0.04) 0.0004 6.0 × 10−9 0.0043 1.2 × 10−4 0 cg04927537 17 LGALS3BP 0.57 (0.05) 0.0006 7.0 × 10−8 0.0093 7.0 × 10−8 1.6 cg05899984 12 N/A 0.84 (0.03) 0.0003 8.1 × 10−8 0.0038 5.7 × 10−6 2.9 cg06500161 21 ABCG1 0.71 (0.03) 0.0005 2.4 × 10−12 0.0096 4.4 × 10−19 0.8 cg11024682 17 SREBF1 0.55 (0.04) 0.0005 2.9 × 10−15 0.0080 3.5 × 10−12 1.2 cg13139542 2 N/A 0.89 (0.02) 0.0002 6.0 × 10−8 0.0029 4.7 × 10−6 0 cg15416179 17 MAP2K3 0.14 (0.03) −0.0002 9.1 × 10−8 −0.0019 2.6 × 10−4 3.6 cg17901584 1 DHCR24 0.68 (0.07) −0.0005 1.7 × 10−8 −0.0080 8.3 × 10−8 2.0 cg18030453 3 LARS2 0.72 (0.04) 0.0003 8.8 × 10−8 0.0029 8.6 × 10−4 0 cg18772573 17 CPSF4L 0.85 (0.03) 0.0003 7.3 × 10−8 0.0039 2.8 × 10−5 0 cg21139312 17 MSI2 0.89 (0.03) 0.0004 5.9 × 10−12 0.0028 6.1 × 10−7 8.2 cg24174557 17 TMEM49 0.38 (0.07) −0.0005 1.1 × 10−8 −0.0059 5.3 × 10−5 0 Abbreviations: ABCG1, ATP binding cassette, subfamily G, member 1; ARIC, Atherosclerosis Risk in Communities; ATP, adenosine triphosphate; CPSF4L, cleavage and polyadenylation specific factor 4 like gene; CPT1A, carnitine palmitoyltransferase 1A gene; DHCR24, 24-dehydrocholesterol reductase gene; ID, identification; LARS2, leucyl-TRNA synthetase 2, mitochondrial gene; LGALS3BP, galectin 3 binding protein gene; MAP2K3, mitogen-activated protein kinase kinase 3 gene; MSI2, Musashi RNA binding protein 2 gene; N/A, not annotated; RS, Rotterdam Study; SD, standard deviation; SREBF1, sterol regulatory element binding transcription factor 1 gene; TMEM49, transmembrane protein 49 gene. a The methylation proportion of a 5′-C-phosphate-G-3′ site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). b Regression coefficient based on a linear mixed model and reflecting difference in the methylation β value per unit increase in waist circumference (cm). Models adjusted for age, sex, current smoking, leukocyte proportions, array number, and position on array. c In the RS we corrected for multiple testing using a robust Bonferroni-corrected P value of 1.08 × 10−7, and in the ARIC Study the level of significance for replication was P < 1.92 × 10−3. In addition to these novel findings, we confirmed previously reported associations of CpG sites with BMI and WC, including those on the carnitine palmitoyltransferase 1A gene (CPT1A), the adenosine triphosphate (ATP) binding cassette, subfamily G, member 1, gene (ABCG1), and the sterol regulatory element binding transcription factor 1 gene (SREBF1). Scatterplots of the associations between the replicated CpG sites and BMI and WC are shown in Web Figures 1 and 2, respectively. Figure 1 shows successfully replicated findings for BMI and WC and highlights the overlapping loci, including those on ABCG1, MSI2, LARS2, SREBF1, and CPT1A. To test for genomic inflation, we calculated the λ value for the EWAS on BMI and WC and created Q-Q plots. The λ values for the EWAS on BMI and WC were 1.487 and 1.556, respectively. The Q-Q plots for BMI and WC are shown in Web Figures 3 and 4, respectively. Figure 1. View largeDownload slide 5′-C-Phosphate-G-3′ sites for body mass index (BMI) and waist circumference (WC) identified in the Rotterdam Study (Rotterdam, the Netherlands; 2006–2013) and replicated in the Atherosclerosis Risk in Communities Study (4 US communities; 1990–1992) and their overlap. ABCG1, ATP binding cassette, subfamily G, member 1, gene; ATP, adenosine triphosphate; BRDT, bromodomain testis associated gene; CPSF4L, cleavage and polyadenylation specific factor 4 like gene; CPT1A, carnitine palmitoyltransferase 1A gene; DHCR24, 24-dehydrocholesterol reductase gene; IFI44L, interferon induced protein 44 like gene; LARS2, leucyl-tRNA synthetase 2, mitochondrial gene; LGALS3BP, galectin 3 binding protein gene; MAP1A, microtubule associated protein 1A gene; MAP2K3, mitogen-activated protein kinase kinase 3 gene; MAP3K5, mitogen-activated protein kinase kinase 5 gene; MSI2, Musashi RNA binding protein 2 gene; PSMD1, proteasome 26s subunit, non-ATPase 1 gene; SREBF1, sterol regulatory element binding transcription factor 1 gene. Figure 1. View largeDownload slide 5′-C-Phosphate-G-3′ sites for body mass index (BMI) and waist circumference (WC) identified in the Rotterdam Study (Rotterdam, the Netherlands; 2006–2013) and replicated in the Atherosclerosis Risk in Communities Study (4 US communities; 1990–1992) and their overlap. ABCG1, ATP binding cassette, subfamily G, member 1, gene; ATP, adenosine triphosphate; BRDT, bromodomain testis associated gene; CPSF4L, cleavage and polyadenylation specific factor 4 like gene; CPT1A, carnitine palmitoyltransferase 1A gene; DHCR24, 24-dehydrocholesterol reductase gene; IFI44L, interferon induced protein 44 like gene; LARS2, leucyl-tRNA synthetase 2, mitochondrial gene; LGALS3BP, galectin 3 binding protein gene; MAP1A, microtubule associated protein 1A gene; MAP2K3, mitogen-activated protein kinase kinase 3 gene; MAP3K5, mitogen-activated protein kinase kinase 5 gene; MSI2, Musashi RNA binding protein 2 gene; PSMD1, proteasome 26s subunit, non-ATPase 1 gene; SREBF1, sterol regulatory element binding transcription factor 1 gene. We calculated a methylation risk score based on the 12 CpG sites for BMI and 14 CpG sites for WC that were identified and replicated in the current study. For BMI, 2.0% of the variance was explained by the methylation risk score, whereas for WC the variance explained was 6.4%. DISCUSSION In this study, we used an EWAS approach to identify novel differentially methylated genes for obesity-related traits in older adults. The EWAS analysis in the RS data identified numerous novel loci associated with BMI and WC, of which many findings were successfully replicated in the ARIC data. The most significant CpG sites associated with both BMI and WC were located on the MSI2 and LARS2 genes. Additionally, CpG sites at BRDT and MAP1A were associated with BMI, and CpG sites at TMEM49 and LGALS3BP were associated with WC. Moreover, we confirmed previous findings that methylation at CPT1A, ABCG1, and SREBF1 is associated with BMI and WC. Previous EWAS on obesity traits were conducted in population-based studies, including ARIC and the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) Study (9, 10), and in persons with a history of myocardial infarction or healthy blood donors from the Cardiogenics Consortium (11). Similarly to our findings, ARIC and GOLDN investigators reported an inverse association between a CpG site at CPT1A and BMI (9, 10) and positive associations of CpG sites at ABCG1 and SREBF1 with BMI and WC (10). In the Cardiogenics Consortium, however, Dick et al. (11) reported only a positive association between 3 CpG sites on the hypoxia inducible factor 3, α subunit, gene (HIF3A) and BMI in both blood and adipose tissue DNA in European adults. These CpG sites at HIF3A did not achieve the threshold for statistical significance in our study. However, CpG sites at HIF3A were replicated in the ARIC Study in DNA blood (10). This discrepancy may be due to differences in the prevalences of obesity and comorbidity between our study (25% obesity, 11% diabetes, 7% coronary heart disease), the ARIC Study (44% obesity, 26% diabetes), and the Cardiogenics Consortium (4% diabetes, 52% myocardial infarction). The genes known to be associated with obesity—CPT1A, ABCG1, and SREBF1—are involved in regulation of lipids, lipoprotein metabolism, and insulin sensitivity (25–27). Specifically, CPT1A encodes for carnitine palmitoytransferase-1, which is a mitochondrial protein involved in fatty acid metabolism (28) and lipoprotein subfraction profile (26). ABCG1 encodes for the ATP binding cassette, subfamily G, member 1, protein and is involved in the transport of cholesterol and phospholipids in macrophages (29). Finally, SREBF1 encodes for sterol regulatory element binding transcription factor 1, which is known to promote adipocyte differentiation and signaling of insulin action (27). Although it has been shown previously that these loci are associated with obesity-related traits, it is still important to replicate these findings across different study populations. Since the EWAS approach is hypothesis-free, findings are prone to being false-positive. By replicating previously reported results, we can say with more certainty that these CpG sites are true-positive findings. In addition to confirming these previously identified loci, we have identified and replicated novel CpG sites located on the MSI2 (cg21139312) and LARS2 (cg18030453) genes, which were associated with both BMI and WC. The CpG site on MSI2 explained 2.2% of the variation in BMI and 8.2% of the variation in WC. MSI2 encodes RNA-binding proteins and plays a central role in posttranscriptional gene regulation (30). A genome-wide association study in pigs suggested that MSI2 is associated with eating behaviors, including number of visits to the feeder per day (31). Moreover, in another study performed in mice, Sakakibara et al. (32) reported that MSI2 is linked with the proliferation and maintenance of stem cells in the central nervous system. This study suggested that during neurogenesis, MSI2 expression persisted in a subset of neuronal lineage cells, such as parvalbumin-containing γ-aminobutyric acid (GABA) neurons in the neocortex (30, 32). GABA receptors are involved in controlling feeding behavior, reinforcing the role of MSI2 in obesity. The other novel locus associated with both BMI and WC, LARS2, encodes an enzyme that catalyzes aminoacylation of mitochondrial tRNALeu (33). A previous postmortem study showed that LARS2 expression (human leucyl-tRNA synthetase 2, mitochondrial NM015340) was increased in brain tissue of patients with bipolar disorder as compared with controls (34). Considering that bipolar disorder is associated with obesity, overweight, and abdominal obesity (35), methylation of MSI2 and LARS2 could play a role in disturbances in eating behaviors, and consequently in BMI and WC. However, further studies are warranted to establish the temporality and pathway of the associations. Although previous investigators have studied the association between DNA methylation and anthropometric characteristics, this study is the first, to our knowledge, to have found an association between DNA methylation of several CpG sites, including those at MSI2 and LARS2, and BMI and WC. One possible explanation for discrepancies between the findings of our current study and those of previous similar studies is the difference in population characteristics. Study populations in previous studies consisted of mixed ethnic groups, participants of younger ages, or persons at high disease risk (9–12). Considering that our discovery cohort consisted of an ethnically homogenous group of older adults from the general population of Rotterdam, underlying mechanisms may differ from those in other population groups. In this study, we conducted an EWAS in a European population and replicated the findings in African Americans. Epidemiologists have reported large disparities across racial/ethnic groups in the development of obesity (36). For example, in the current study, the rates of obesity were significantly lower in Europeans (24.6%) than in African Americans (43.8%). However, despite the differences in ethnicity and prevalence of obesity between our studies, most of our CpG sites (86.7%) were successfully replicated in ARIC. This may indicate that, in contrast to genetic studies, where replication across ethnic groups is challenging due to differences in linkage disequilibrium pattern, epigenetic findings could more easily be translated across ethnic groups. The strengths of the current study include the large size of the sample with available data on DNA methylation and the ability to replicate our findings in different ethnic populations. However, the results of this study must be interpreted in light of several limitations. We used whole blood samples for the quantification of DNA methylation, whereas adipose tissue may be a more relevant tissue for examining obesity. In this case, important CpG sites may not have been identified in our study. Unlike in genetic studies, unraveling the direction of the association between DNA methylation and phenotypes in epigenetic epidemiology remains challenging. Due to the cross-sectional study design and the nature of our variables, which are responsive to the environment and dynamic over time, a temporal direction in the association between DNA methylation and anthropometric measures cannot be determined. Because previous studies have shown that change in DNA methylation is a consequence of BMI for the majority of CpG sites, this may be the most likely direction for the associations observed in the current study as well (12). However, longitudinal studies are required to confirm the direction of the associations between DNA methylation and anthropometric factors. Another possibility is that our findings could be explained by third common factors. For instance, associations may have been confounded by differences in cell type proportion. In order to avoid this source of confounding, we adjusted all analyses for cell type proportions. However, as in any observational study, residual confounding due to various lifestyle factors still remains an issue. Furthermore, the Q-Q plots showed high genomic inflation. Many EWAS studies have had high genomic inflation (37). Adjustment for potential confounders such as technical covariates could decrease the inflation. Lehne et al. (38) have suggested that the correlation between CpG sites and the large number of findings in EWAS explain the residual inflation. In this study, we performed adequate adjustment for technical covariates. Moreover, the replication of our results in an independent population provides further evidence for the robustness of our findings. In conclusion, we have reported a novel association of increased methylation in the MSI2 and LARS2 genes with increased BMI and WC in older adults. Moreover, we confirmed 3 previously identified methylation loci (CPT1A, ABCG1, and SREBF1) suggested to be associated with obesity. Further investigations using repeatedly measured genome-wide DNA methylation and obesity-related traits are needed to assess causality and to further evolve the growing field of epigenetic epidemiology toward novel therapeutic and preventative approaches to obesity and related noncommunicable disorders. ACKNOWLEDGMENTS Author affiliations: Department of Epidemiology, Erasmus University Medical Center (Erasmus MC), Rotterdam, the Netherlands (Klodian Dhana, Kim V. E. Braun, Jana Nano, Trudy Voortman, Andre G. Uitterlinden, Albert Hofman, Oscar H. Franco, Abbas Dehghan); Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Klodian Dhana, Kim V. E. Braun); Rotterdam Intergenerational Ageing Research Center (ErasmusAGE), Rotterdam, the Netherlands (Kim V. E. Braun, Trudy Voortman, Oscar H. Franco); Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, Minneapolis, Minnesota (Ellen W. Demerath); Division of Biostatistics, School of Public Health, University of Minnesota, Minneapolis, Minnesota (Weihua Guan); Human Genetics Center, School of Public Health, University of Texas Health Sciences Center at Houston, Houston, Texas (Myriam Fornage); Human Genome Sequencing Center, Baylor College of Medicine, Houston, Texas (Myriam Fornage); Department of Internal Medicine, Erasmus University Medical Center (Erasmus MC), Rotterdam, the Netherlands (Joyce B. J. van Meurs, Andre G. Uitterlinden); Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Albert Hofman); and Department of Epidemiology, Imperial College London, London, United Kingdom (Abbas Dehghan). K.D. and K.V.E.B. contributed equally to this article. The Rotterdam Study is funded by Erasmus University Medical Center (Erasmus MC) and Erasmus University Rotterdam; the Netherlands Organization for Health Research and Development; the Research Institute for Diseases in the Elderly; the Netherlands Ministry of Education, Culture and Science; the Netherlands Ministry of Health, Welfare and Sport; the European Commission (Directorate-General XII); and the Municipality of Rotterdam. The ARIC Study is funded by the US National Heart, Lung, and Blood Institute. Generation and management of the Infinium Human Methylation 450K BeadChip array data for the Rotterdam Study was conducted by the Human Genotyping Facility of the Genetic Laboratory of the Department of Internal Medicine, Erasmus MC; funding was provided by Erasmus MC and the Netherlands Organization for Scientific Research (project 184021007). The data were made available as part of Rainbow Project 3 of the Biobanking and Biomolecular Resources Research Infrastructure–Netherlands. We thank Michael Verbiest, Mila Jhamai, Sarah Higgins, Marijn Verkerk, and Dr. Lisette Stolk for their help in creating the methylation database. We are grateful to the staff of and participants in the Rotterdam and ARIC studies and to all of the general practitioners and pharmacists involved. The funding organizations played no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, and approval of the manuscript. The Rotterdam Intergenerational Ageing Research Center (ErasmusAGE), a center for aging research across the life course, is supported by Nestlé Nutrition (Nestec Ltd., Lausanne, Switzerland) and Metagenics, Inc. (Aliso Viejo, California). Conflict of interest: none declared. Abbreviations ABCG1 ATP binding cassette, subfamily G, member 1, gene ARIC Atherosclerosis Risk in Communities ATP adenosine triphosphate BMI body mass index CpG 5′-C-phosphate-G-3′ CPT1A carnitine palmitoyltransferase 1A gene EWAS epigenome-wide association study GABA γ-aminobutyric acid GOLDN Genetics of Lipid Lowering Drugs and Diet Network LARS2 leucyl-tRNA synthetase 2, mitochondrial gene MSI2 Musashi RNA binding protein 2 gene RS Rotterdam Study SREBF1 sterol regulatory element binding transcription factor 1 gene WC waist circumference REFERENCES 1 World Health Organization . Obesity: Preventing and Managing the Global Epidemic. Report of a WHO Consultation. (WHO Technical Report Series, no. 894). Geneva, Switzerland: World Health Organization; 2000 . 2 Berrington de Gonzalez A , Hartge P , Cerhan JR , et al. . Body-mass index and mortality among 1.46 million white adults . N Engl J Med . 2010 ; 363 ( 23 ): 2211 – 2219 . Google Scholar CrossRef Search ADS PubMed 3 NCD Risk Factor Collaboration . Trends in adult body-mass index in 200 countries from 1975 to 2014: a pooled analysis of 1698 population-based measurement studies with 19.2 million participants . Lancet . 2016 ; 387 ( 10026 ): 1377 – 1396 . CrossRef Search ADS PubMed 4 Marti A , Martinez-Gonzalez MA , Martinez JA . Interaction between genes and lifestyle factors on obesity . Proc Nutr Soc . 2008 ; 67 ( 1 ): 1 – 8 . Google Scholar CrossRef Search ADS PubMed 5 Vattikuti S , Guo J , Chow CC . Heritability and genetic correlations explained by common SNPs for metabolic syndrome traits . PLoS Genet . 2012 ; 8 ( 3 ): e1002637 . Google Scholar CrossRef Search ADS PubMed 6 Bollati V , Baccarelli A . Environmental epigenetics . Heredity (Edinb) . 2010 ; 105 ( 1 ): 105 – 112 . Google Scholar CrossRef Search ADS PubMed 7 Mill J , Dempster E , Caspi A , et al. . Evidence for monozygotic twin (MZ) discordance in methylation level at two CpG sites in the promoter region of the catechol-O-methyltransferase (COMT) gene . Am J Med Genet B Neuropsychiatr Genet . 2006 ; 141B ( 4 ): 421 – 425 . Google Scholar CrossRef Search ADS PubMed 8 Jones PA . Functions of DNA methylation: islands, start sites, gene bodies and beyond . Nat Rev Genet . 2012 ; 13 ( 7 ): 484 – 492 . Google Scholar CrossRef Search ADS PubMed 9 Aslibekyan S , Demerath EW , Mendelson M , et al. . Epigenome-wide study identifies novel methylation loci associated with body mass index and waist circumference . Obesity (Silver Spring) . 2015 ; 23 ( 7 ): 1493 – 1501 . Google Scholar CrossRef Search ADS PubMed 10 Demerath EW , Guan W , Grove ML , et al. . Epigenome-wide association study (EWAS) of BMI, BMI change and waist circumference in African American adults identifies multiple replicated loci . Hum Mol Genet . 2015 ; 24 ( 15 ): 4464 – 4479 . Google Scholar CrossRef Search ADS PubMed 11 Dick KJ , Nelson CP , Tsaprouni L , et al. . DNA methylation and body-mass index: a genome-wide analysis . Lancet . 2014 ; 383 ( 9933 ): 1990 – 1998 . Google Scholar CrossRef Search ADS PubMed 12 Wahl S , Drong A , Lehne B , et al. . Epigenome-wide association study of body mass index, and the adverse outcomes of adiposity . Nature . 2017 ; 541 ( 7635 ): 81 – 86 . Google Scholar CrossRef Search ADS PubMed 13 Mendelson MM , Marioni RE , Joehanes R , et al. . Association of body mass index with DNA methylation and gene expression in blood cells and relations to cardiometabolic disease: a Mendelian randomization approach . PLoS Med . 2017 ; 14 ( 1 ): e1002215 . Google Scholar CrossRef Search ADS PubMed 14 Ikram MA , Brusselle GGO , Murad SD , et al. . The Rotterdam Study: 2018 update on objectives, design and main results . Eur J Epidemiol . 2017 ; 32 ( 9 ): 807 – 850 . Google Scholar CrossRef Search ADS PubMed 15 The ARIC Investigators . The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives . Am J Epidemiol . 1989 ; 129 ( 4 ): 687 – 702 . CrossRef Search ADS PubMed 16 Physical status: the use and interpretation of anthropometry. Report of a WHO Expert Committee . World Health Organ Tech Rep Ser . 1995 ; 854 : 1 – 452 . PubMed 17 Sandoval J , Heyn H , Moran S , et al. . Validation of a DNA methylation microarray for 450,000 CpG sites in the human genome . Epigenetics . 2011 ; 6 ( 6 ): 692 – 702 . Google Scholar CrossRef Search ADS PubMed 18 Touleimat N , Tost J . Complete pipeline for Infinium® Human Methylation 450K BeadChip data processing using subset quantile normalization for accurate DNA methylation estimation . Epigenomics . 2012 ; 4 ( 3 ): 325 – 341 . Google Scholar CrossRef Search ADS PubMed 19 Pidsley R , Wong CC , Volta M , et al. . A data-driven approach to preprocessing Illumina 450K methylation array data . BMC Genomics . 2013 ; 14 : 293 . Google Scholar CrossRef Search ADS PubMed 20 Willer CJ , Li Y , Abecasis GR . METAL: fast and efficient meta-analysis of genomewide association scans . Bioinformatics . 2010 ; 26 ( 17 ): 2190 – 2191 . Google Scholar CrossRef Search ADS PubMed 21 Houseman EA , Accomando WP , Koestler DC , et al. . DNA methylation arrays as surrogate measures of cell mixture distribution . BMC Bioinformatics . 2012 ; 13 : 86 . Google Scholar CrossRef Search ADS PubMed 22 Aryee MJ , Jaffe AE , Corrada-Bravo H , et al. . Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays . Bioinformatics . 2014 ; 30 ( 10 ): 1363 – 1369 . Google Scholar CrossRef Search ADS PubMed 23 Chen YA , Lemire M , Choufani S , et al. . Discovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray . Epigenetics . 2013 ; 8 ( 2 ): 203 – 209 . Google Scholar CrossRef Search ADS PubMed 24 Barfield RT , Almli LM , Kilaru V , et al. . Accounting for population stratification in DNA methylation studies . Genet Epidemiol . 2014 ; 38 ( 3 ): 231 – 241 . Google Scholar CrossRef Search ADS PubMed 25 Braun KV , Voortman T , Dhana K , et al. . The role of DNA methylation in dyslipidaemia: a systematic review . Prog Lipid Res . 2016 ; 64 : 178 – 191 . Google Scholar CrossRef Search ADS PubMed 26 Frazier-Wood AC , Aslibekyan S , Absher DM , et al. . Methylation at CPT1A locus is associated with lipoprotein subfraction profiles . J Lipid Res . 2014 ; 55 ( 7 ): 1324 – 1330 . Google Scholar CrossRef Search ADS PubMed 27 Kim JB , Spiegelman BM . ADD1/SREBP1 promotes adipocyte differentiation and gene expression linked to fatty acid metabolism . Genes Dev . 1996 ; 10 ( 9 ): 1096 – 1107 . Google Scholar CrossRef Search ADS PubMed 28 Gobin S , Thuillier L , Jogl G , et al. . Functional and structural basis of carnitine palmitoyltransferase 1A deficiency . J Biol Chem . 2003 ; 278 ( 50 ): 50428 – 50434 . Google Scholar CrossRef Search ADS PubMed 29 Klucken J , Büchler C , Orsó E , et al. . ABCG1 (ABC8), the human homolog of the drosophila white gene, is a regulator of macrophage cholesterol and phospholipid transport . Proc Natl Acad Sci USA . 2000 ; 97 ( 2 ): 817 – 822 . Google Scholar CrossRef Search ADS PubMed 30 Sakakibara S , Nakamura Y , Satoh H , et al. . RNA-binding protein Musashi2: developmentally regulated expression in neural precursor cells and subpopulations of neurons in mammalian CNS . J Neurosci . 2001 ; 21 ( 20 ): 8091 – 8107 . Google Scholar CrossRef Search ADS PubMed 31 Do DN , Strathe AB , Ostersen T , et al. . Genome-wide association study reveals genetic architecture of eating behavior in pigs and its implications for humans obesity by comparative mapping . PLoS One . 2013 ; 8 ( 8 ): e71509 . Google Scholar CrossRef Search ADS PubMed 32 Sakakibara S , Nakamura Y , Yoshida T , et al. . RNA-binding protein Musashi family: roles for CNS stem cells and a subpopulation of ependymal cells revealed by targeted disruption and antisense ablation . Proc Natl Acad Sci USA . 2002 ; 99 ( 23 ): 15194 – 15199 . Google Scholar CrossRef Search ADS PubMed 33 Sohm B , Sissler M , Park H , et al. . Recognition of human mitochondrial tRNALeu(UUR) by its cognate leucyl-tRNA synthetase . J Mol Biol . 2004 ; 339 ( 1 ): 17 – 29 . Google Scholar CrossRef Search ADS PubMed 34 Munakata K , Iwamoto K , Bundo M , et al. . Mitochondrial DNA 3243A>G mutation and increased expression of LARS2 gene in the brains of patients with bipolar disorder and schizophrenia . Biol Psychiatry . 2005 ; 57 ( 5 ): 525 – 532 . Google Scholar CrossRef Search ADS PubMed 35 McElroy SL , Kotwal R , Malhotra S , et al. . Are mood disorders and obesity related? A review for the mental health professional . J Clin Psychiatry . 2004 ; 65 ( 5 ): 634 – 651 , quiz 730. Google Scholar CrossRef Search ADS PubMed 36 Zhang Q , Wang Y , Huang ES . Changes in racial/ethnic disparities in the prevalence of type 2 diabetes by obesity level among U.S. adults . Ethn Health . 2009 ; 14 ( 5 ): 439 – 457 . Google Scholar CrossRef Search ADS PubMed 37 Joubert BR , den Dekker HT , Felix JF , et al. . Maternal plasma folate impacts differential DNA methylation in an epigenome-wide meta-analysis of newborns . Nat Commun . 2016 ; 7 : 10577 . Google Scholar CrossRef Search ADS PubMed 38 Lehne B , Drong AW , Loh M , et al. . A coherent approach for analysis of the Illumina Human Methylation450 BeadChip improves data quality and performance in epigenome-wide association studies . Genome Biol . 2015 ; 16 : 37 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Journal of Epidemiology Oxford University Press

Loading next page...
 
/lp/ou_press/an-epigenome-wide-association-study-of-obesity-related-traits-m03Je5S0P0
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0002-9262
eISSN
1476-6256
D.O.I.
10.1093/aje/kwy025
Publisher site
See Article on Publisher Site

Abstract

Abstract We conducted an epigenome-wide association study on obesity-related traits. We used data from 2 prospective, population-based cohort studies: the Rotterdam Study (RS) (2006–2013) and the Atherosclerosis Risk in Communities (ARIC) Study (1990–1992). We used the RS (n = 1,450) as the discovery panel and the ARIC Study (n = 2,097) as the replication panel. Linear mixed-effect models were used to assess the cross-sectional associations between genome-wide DNA methylation in leukocytes and body mass index (BMI) and waist circumference (WC), adjusting for sex, age, smoking, leukocyte proportions, array number, and position on array. The latter 2 variables were modeled as random effects. Fourteen 5′-C-phosphate-G-3′ (CpG) sites were associated with BMI and 26 CpG sites with WC in the RS after Bonferroni correction (P < 1.07 × 10−7), of which 12 and 13 CpGs were replicated in the ARIC Study, respectively. The most significant novel CpGs were located on the Musashi RNA binding protein 2 gene (MSI2; cg21139312) and the leucyl-tRNA synthetase 2, mitochondrial gene (LARS2; cg18030453) and were associated with both BMI and WC. CpGs at BRDT, PSMD1, IFI44L, MAP1A, and MAP3K5 were associated with BMI. CpGs at LGALS3BP, MAP2K3, DHCR24, CPSF4L, and TMEM49 were associated with WC. We report novel associations between methylation at MSI2 and LARS2 and obesity-related traits. These results provide further insight into mechanisms underlying obesity-related traits, which can enable identification of new biomarkers in obesity-related chronic diseases. body mass index, cohort studies, DNA methylation, epigenome-wide association studies, obesity, waist circumference Obesity is an important risk factor for cardiovascular disease, diabetes, some cancers, and musculoskeletal disorders (1–3). Evidence suggests that obesity not only is dependent on lifestyle factors but is a result of interactions between genes and lifestyle (4, 5). Epigenetics has been proposed as a molecular mechanism that can affect the expression of genes by environmental influences and potentially could describe further the link between obesity and its complications (6). Nevertheless, unlike genetics, DNA methylation is dynamic over time; therefore, change in DNA methylation could also be a consequence of obesity. Epigenetics is the study of heritable variation in gene function that is not a result of a change in DNA sequence (7). One of the best studied epigenetic mechanisms is DNA methylation, the attachment of a methyl group to a cytosine nucleotide of 5′-C-phosphate-G-3′ (CpG) dinucleotides. DNA methylation has varying functions at different locations in the human genome, including regulation of gene expression (8). To date, epigenome-wide association studies (EWAS) have identified several differentially methylated CpG regions related to body mass index (BMI)—the most widely used measure of obesity—and waist circumference (WC) (9–12). These few studies were performed in either patient populations, specific ethnic groups, or young adults. However, information on the older adults from population-based studies is scarce. In older adults and elderly persons, biological mechanisms involved in body weight and body composition may be different from those in younger adults (13). Therefore, it is crucial to explore the relationship of obesity to epigenetic variation in older adults. We performed a cross-sectional EWAS of DNA methylation in blood leukocytes for BMI and WC in subjects from the Rotterdam Study (RS) and replicated our findings in the Atherosclerosis Risk in Communities (ARIC) Study. METHODS Study population The RS is a large prospective, population-based cohort study aimed at assessing the occurrence of and risk factors for chronic diseases (cardiovascular, endocrine, hepatic, neurological, ophthalmic, psychiatric, dermatological, oncological, and respiratory) in the elderly (14). The study comprises 14,926 subjects in total, living in the well-defined Ommoord district in the city of Rotterdam, the Netherlands. In 1989, the first RS cohort, Rotterdam Study I (RS-I), was established and comprised 7,983 subjects aged 55 years or above. In 2000, the second cohort, RS-II, was established, with 3,011 subjects who had reached the age of 55 years since 1989. In 2006, the third cohort, RS-III, was further included, with 3,932 subjects aged 45 years or above. The discovery panel for the current analysis consisted of a random sample of 1,450 participants from the first and second study visits of the third cohort (RS-III-1, RS-III-2) and the third study visit of the second cohort (RS-II-3). We sought replication of the identified CpG sites in the ARIC Study, which is described in detail elsewhere (15). Briefly, the ARIC Study is a prospective cohort study of cardiovascular disease in adults. Between 1987 and 1989, 7,082 men and 8,710 women aged 45–64 years were recruited from 4 US communities (15). Methylation data were available in a subset of 2,097 African-American participants (10). The RS and ARIC Study protocols were approved by institutional review boards at each participating university, and all participants provided written informed consent. Anthropometric measures and covariates Height and weight were measured with the participant standing without shoes and heavy outer garments. WC was measured at the level midway between the lower rib margin and the iliac crest, with the participant in a standing position without heavy outer garments and with emptied pockets, breathing out gently. Hip circumference was recorded as the maximum circumference of the hips over the buttocks. BMI was calculated as weight (kg) divided by the square of height (m), and waist:hip ratio was calculated as WC divided by hip circumference (16). Information on current and past smoking behavior was acquired from questionnaires. DNA methylation data DNA was extracted from whole peripheral blood (stored in ethylenediaminetetraacetic acid tubes) by means of standardized salting-out methods. Genome-wide DNA methylation levels were measured using the Illumina Infinium Human Methylation 450K BeadChip array (Illumina, Inc., San Diego, California) (17). In short, samples (500 ng of DNA per sample) were first treated with bisulfite using the Zymo EZ-96 DNA methylation kit (Zymo Research, Irvine, California). Next, samples were hybridized to the arrays according to the manufacturer’s protocol. The methylation proportion of a CpG site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). Data preprocessing was additionally performed in both data sets using an R programming pipeline that is based on the pipeline developed by Touleimat and Tost (18), which includes additional parameters and options for preprocessing and normalizing methylation data directly from image data (IDAT) files. We excluded probes which had a detection P value greater than 0.01 in more than 95% of samples. We excluded 11,648 probes at X and Y chromosomes to avoid sex bias. The raw β values were then background-corrected and normalized using the “dasen” option in the wateRmelon R package (R Foundation for Statistical Computing, Vienna, Austria) (19). Per individual probe, participants with methylation levels higher than 3 times the interquartile range were excluded. Statistical analyses The characteristics of the discovery and replication population are presented as mean values for continuous variables and proportions for categorical variables. In the discovery stage, we modeled cross-sectional associations between “dasen”-normalized β values of the CpG sites as the outcome and BMI or WC as the exposure, using linear mixed-effect models adjusting for age, sex, smoking, white blood cell proportion, array number (65 arrays), and position on array (12 positions; a combination of row number and column number). We performed an independent analysis in persons from RS-III-1, RS-III-2, and RS-II-3. We then performed a fixed-effects meta-analysis on the estimates for these 3 RS groups using the inverse-variance–weighted method implemented in METAL, combining the RS-III-1 participants with the RS-III-2 and RS-II-3 participants (20). Technical covariates (array number and position on array) were modeled as random effects. For the RS-III-1 group, we estimated leukocyte proportions (B cells, CD4-positive T cells, CD8-positive T cells, granulocytes, monocytes, and natural killer cells) by means of a formula developed by Houseman et al. (21) and implemented in the minfi package in R (22). For the RS-II-3 and RS-III-2 groups, we used white blood cell counts (i.e., lymphocytes, monocytes, and granulocytes), which were assessed with a Coulter Ac·T diff2 Hematology Analyzer (Beckman Coulter, Brea, California). We corrected for multiple testing using a robust Bonferroni-corrected P value of 1.07 × 10−7 as the threshold for significance (0.05/463,456 probes). The probes identified in the discovery analysis were tested for replication in the independent samples from the ARIC Study. A Bonferroni-corrected P value of 0.05 divided by the number of significant findings in the discovery study was used as a threshold level of significant replication. Finally, we checked all identified CpG sites for cross-reaction or polymorphism (23). A CpG site was considered polymorphic when a single nucleotide polymorphism with a minor allele frequency greater than 0.01 resided at the position of the cytosine or guanine nucleotide or within 10 base pairs of the CpG site within the probe binding site (24). Methylation risk score A methylation risk score was calculated based on CpG sites that were associated with the phenotypes. The effect estimates were used to build the methylation risk score using data from the discovery panel. Linear regression analyses were performed using BMI or WC as the outcome variable and the included CpG sites as exposure variables. With the use of linear regression models, we calculated the variance in lipid levels explained by the methylation risk score. RESULTS Table 1 summarizes the characteristics of participants in the studies. The RS is entirely comprised of Europeans, whereas the ARIC Study participants comprised only African Americans. Compared with RS participants (mean age = 63.7 (standard deviation, 8.1) years), the participants in ARIC were younger, on average (mean age = 56.2 (standard deviation, 5.7) years) and comprised more women (64% in ARIC vs. 56% in the RS)). The respective mean values for BMI in the RS and the ARIC Study were 27.7 and 30.1. The mean values for WC were 93.7 cm and 101.3 cm in the RS and ARIC, respectively. Table 1. Characteristics of Participants in the Rotterdam Study (Rotterdam, the Netherlands; 2006–2013) and the Atherosclerosis Risk in Communities Study (4 US Communities; 1990–1992) Selected for an Epigenome-Wide Association Study on Obesity-Related Traits Characteristic RS (n = 1,450) ARIC Study (n = 2,097) No. of Persons % No. of Persons % Age, yearsa 63.7 (8.1) 56.2 (5.7) Female sex 811 55.9 1,334 63.6 Race, %  White 1,450 100 0 0  African-American 0 0 2,097 100 BMIa,b,c 27.7 (4.4) 30.1 (6.1) BMI statusc  Normal-weight 419 28.9 388 18.5  Overweight 674 46.5 788 37.6  Obese 357 24.6 918 43.8 WC, cma 93.7 (12.9) 101.3 (15.1) Smoking status  Current smoker 273 18.8 512 24.4  Current nonsmoker 1,177 81.2 1,585 75.6 Diabetes mellitus 160 11 545 26.0 Characteristic RS (n = 1,450) ARIC Study (n = 2,097) No. of Persons % No. of Persons % Age, yearsa 63.7 (8.1) 56.2 (5.7) Female sex 811 55.9 1,334 63.6 Race, %  White 1,450 100 0 0  African-American 0 0 2,097 100 BMIa,b,c 27.7 (4.4) 30.1 (6.1) BMI statusc  Normal-weight 419 28.9 388 18.5  Overweight 674 46.5 788 37.6  Obese 357 24.6 918 43.8 WC, cma 93.7 (12.9) 101.3 (15.1) Smoking status  Current smoker 273 18.8 512 24.4  Current nonsmoker 1,177 81.2 1,585 75.6 Diabetes mellitus 160 11 545 26.0 Abbreviations: ARIC, Atherosclerosis Risk in Communities; BMI, body mass index; RS, Rotterdam Study; WC, waist circumference. a Data are presented as mean (standard deviation). b BMI was calculated as weight (kg)/height (m)2. c Data on BMI were missing for 3 ARIC participants. Table 1. Characteristics of Participants in the Rotterdam Study (Rotterdam, the Netherlands; 2006–2013) and the Atherosclerosis Risk in Communities Study (4 US Communities; 1990–1992) Selected for an Epigenome-Wide Association Study on Obesity-Related Traits Characteristic RS (n = 1,450) ARIC Study (n = 2,097) No. of Persons % No. of Persons % Age, yearsa 63.7 (8.1) 56.2 (5.7) Female sex 811 55.9 1,334 63.6 Race, %  White 1,450 100 0 0  African-American 0 0 2,097 100 BMIa,b,c 27.7 (4.4) 30.1 (6.1) BMI statusc  Normal-weight 419 28.9 388 18.5  Overweight 674 46.5 788 37.6  Obese 357 24.6 918 43.8 WC, cma 93.7 (12.9) 101.3 (15.1) Smoking status  Current smoker 273 18.8 512 24.4  Current nonsmoker 1,177 81.2 1,585 75.6 Diabetes mellitus 160 11 545 26.0 Characteristic RS (n = 1,450) ARIC Study (n = 2,097) No. of Persons % No. of Persons % Age, yearsa 63.7 (8.1) 56.2 (5.7) Female sex 811 55.9 1,334 63.6 Race, %  White 1,450 100 0 0  African-American 0 0 2,097 100 BMIa,b,c 27.7 (4.4) 30.1 (6.1) BMI statusc  Normal-weight 419 28.9 388 18.5  Overweight 674 46.5 788 37.6  Obese 357 24.6 918 43.8 WC, cma 93.7 (12.9) 101.3 (15.1) Smoking status  Current smoker 273 18.8 512 24.4  Current nonsmoker 1,177 81.2 1,585 75.6 Diabetes mellitus 160 11 545 26.0 Abbreviations: ARIC, Atherosclerosis Risk in Communities; BMI, body mass index; RS, Rotterdam Study; WC, waist circumference. a Data are presented as mean (standard deviation). b BMI was calculated as weight (kg)/height (m)2. c Data on BMI were missing for 3 ARIC participants. Tables 2 and 3 list the CpG sites associated with BMI and WC in both populations. Using the Bonferroni-corrected statistical significance level of 1.07 × 10−7, we identified 14 CpG sites associated with BMI (see Web Table 1, available at https://academic.oup.com/aje) and 26 CpG sites associated with WC (Web Table 2) in the RS. In the ARIC Study, we successfully replicated 12 of the 14 BMI-related CpG sites (P < 3.57 × 10−3) (Table 2) and 13 of the 26 WC-related CpG sites (P < 1.92 × 10−3) (Table 3). Among these, 8 BMI-related CpG sites and 11 WC-related CpG sites were novel. The most significant novel CpG sites were located on the Musashi RNA binding protein 2 gene (MSI2; cg21139312) and the leucyl-tRNA synthetase 2, mitochondrial gene (LARS2; cg18030453) for both BMI and WC. For every unit increase in BMI (kg/m2), there were increases in MSI2 and LARS2 methylation of 0.0009 (P = 4.5 × 10−10) and 0.0009 (P = 4.5 × 10−9), respectively. For every unit increase in WC (cm), there were increases in MSI2 and LARS2 methylation of 0.0004 (P = 5.9 × 10−12) and 0.0003 (P = 8.8 × 10−8), respectively. Additionally, for BMI, other novel CpG sites were located in the bromodomain testis associated gene (BRDT; cg03421440) and the microtubule associated protein 1A gene (MAP1A; cg15159104). For WC, the other top novel CpG sites were located in the transmembrane protein 49 gene (TMEM49; cg24174557) and the galectin 3 binding protein gene (LGALS3BP; cg04927537). Table 2. 5′-C-Phosphate-G-3′ Methylation Sites Associated With Body Mass Indexa at the Level of Genome-Wide Significance in the Rotterdam Study (2006–2013) and Successfully Replicated in the Atherosclerosis Risk in Communities Study (1990–1992) Probe ID No. Chromosome Gene Mean (SD) Methylationb Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βc P Valued βc P Valued cg00574958 11 CPT1A 0.19 (0.04) −0.0011 6.2 × 10−15 −0.0029 3.2 × 10−12 1.9 cg00851028 1 N/A 0.72 (0.04) 0.0010 5.4 × 10−8 0.0038 9.0 × 10−4 0.4 cg03421440 1 BRDT 0.71 (0.07) −0.0015 3.2 × 10−8 −0.0043 1.3 × 10−3 0.3 cg06096336 2 PSMD1 0.64 (0.05) 0.0016 4.3 × 10−8 0.0058 5.5 × 10−4 1.0 cg06500161 21 ABCG1 0.71 (0.03) 0.0011 1.7 × 10−9 0.0081 1.5 × 10−13 0.3 cg06872964 1 IFI44L 0.62 (0.06) 0.0015 4.8 × 10−8 0.0100 4.3 × 10−7 0.4 cg11024682 17 SREBF1 0.55 (0.04) 0.0013 6.6 × 10−15 0.0068 9.6 × 10−9 0.1 cg15159104 15 MAP1A 0.48 (0.05) 0.0010 3.2 × 10−8 0.0048 5.1 × 10−6 0.0 cg15903032 10 N/A 0.57 (0.04) 0.0010 7.6 × 10−8 0.0037 2.8 × 10−3 0.2 cg18030453 3 LARS2 0.72 (0.04) 0.0009 4.5 × 10−9 0.0028 1.7 × 10−3 0.1 cg21139312 17 MSI2 0.89 (0.03) 0.0009 4.5 × 10−10 0.0028 1.2 × 10−6 2.0 cg21506299 6 MAP3K5 0.23 (0.06) −0.0010 3.5 × 10−8 −0.0019 2.8 × 10−3 1.1 Probe ID No. Chromosome Gene Mean (SD) Methylationb Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βc P Valued βc P Valued cg00574958 11 CPT1A 0.19 (0.04) −0.0011 6.2 × 10−15 −0.0029 3.2 × 10−12 1.9 cg00851028 1 N/A 0.72 (0.04) 0.0010 5.4 × 10−8 0.0038 9.0 × 10−4 0.4 cg03421440 1 BRDT 0.71 (0.07) −0.0015 3.2 × 10−8 −0.0043 1.3 × 10−3 0.3 cg06096336 2 PSMD1 0.64 (0.05) 0.0016 4.3 × 10−8 0.0058 5.5 × 10−4 1.0 cg06500161 21 ABCG1 0.71 (0.03) 0.0011 1.7 × 10−9 0.0081 1.5 × 10−13 0.3 cg06872964 1 IFI44L 0.62 (0.06) 0.0015 4.8 × 10−8 0.0100 4.3 × 10−7 0.4 cg11024682 17 SREBF1 0.55 (0.04) 0.0013 6.6 × 10−15 0.0068 9.6 × 10−9 0.1 cg15159104 15 MAP1A 0.48 (0.05) 0.0010 3.2 × 10−8 0.0048 5.1 × 10−6 0.0 cg15903032 10 N/A 0.57 (0.04) 0.0010 7.6 × 10−8 0.0037 2.8 × 10−3 0.2 cg18030453 3 LARS2 0.72 (0.04) 0.0009 4.5 × 10−9 0.0028 1.7 × 10−3 0.1 cg21139312 17 MSI2 0.89 (0.03) 0.0009 4.5 × 10−10 0.0028 1.2 × 10−6 2.0 cg21506299 6 MAP3K5 0.23 (0.06) −0.0010 3.5 × 10−8 −0.0019 2.8 × 10−3 1.1 Abbreviations: ABCG1, ATP binding cassette, subfamily G, member 1 gene; ARIC, Atherosclerosis Risk in Communities; ATP, adenosine triphosphate; BMI, body mass index; BRDT, bromodomain testis associated gene; CPT1A, carnitine palmitoyltransferase 1A gene; ID, identification; IFI44L, interferon induced protein 44 like gene; LARS2, leucyl-tRNA synthetase 2, mitochondrial gene; MAP1A, microtubule associated protein 1A gene; MAP3K5, mitogen-activated protein kinase kinase 5 gene; MSI2, Musashi RNA binding protein 2 gene; N/A, not annotated; PSMD1, proteasome 26s subunit, non-ATPase 1 gene; RS, Rotterdam Study; SD, standard deviation; SREBF1, sterol regulatory element binding transcription factor 1 gene. a BMI was calculated as weight (kg)/height (m)2. b The methylation proportion of a 5′-C-phosphate-G-3′ site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). c Regression coefficient based on a linear mixed model and reflecting difference in the methylation β value per unit increase in BMI (kg/m2). Models adjusted for age, sex, current smoking, leukocyte proportions, array number, and position on array. d In the RS we corrected for multiple testing using a robust Bonferroni-corrected P value of 1.07 × 10−7, and in the ARIC Study the level of significance for replication was P < 3.57 × 10−3. Table 2. 5′-C-Phosphate-G-3′ Methylation Sites Associated With Body Mass Indexa at the Level of Genome-Wide Significance in the Rotterdam Study (2006–2013) and Successfully Replicated in the Atherosclerosis Risk in Communities Study (1990–1992) Probe ID No. Chromosome Gene Mean (SD) Methylationb Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βc P Valued βc P Valued cg00574958 11 CPT1A 0.19 (0.04) −0.0011 6.2 × 10−15 −0.0029 3.2 × 10−12 1.9 cg00851028 1 N/A 0.72 (0.04) 0.0010 5.4 × 10−8 0.0038 9.0 × 10−4 0.4 cg03421440 1 BRDT 0.71 (0.07) −0.0015 3.2 × 10−8 −0.0043 1.3 × 10−3 0.3 cg06096336 2 PSMD1 0.64 (0.05) 0.0016 4.3 × 10−8 0.0058 5.5 × 10−4 1.0 cg06500161 21 ABCG1 0.71 (0.03) 0.0011 1.7 × 10−9 0.0081 1.5 × 10−13 0.3 cg06872964 1 IFI44L 0.62 (0.06) 0.0015 4.8 × 10−8 0.0100 4.3 × 10−7 0.4 cg11024682 17 SREBF1 0.55 (0.04) 0.0013 6.6 × 10−15 0.0068 9.6 × 10−9 0.1 cg15159104 15 MAP1A 0.48 (0.05) 0.0010 3.2 × 10−8 0.0048 5.1 × 10−6 0.0 cg15903032 10 N/A 0.57 (0.04) 0.0010 7.6 × 10−8 0.0037 2.8 × 10−3 0.2 cg18030453 3 LARS2 0.72 (0.04) 0.0009 4.5 × 10−9 0.0028 1.7 × 10−3 0.1 cg21139312 17 MSI2 0.89 (0.03) 0.0009 4.5 × 10−10 0.0028 1.2 × 10−6 2.0 cg21506299 6 MAP3K5 0.23 (0.06) −0.0010 3.5 × 10−8 −0.0019 2.8 × 10−3 1.1 Probe ID No. Chromosome Gene Mean (SD) Methylationb Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βc P Valued βc P Valued cg00574958 11 CPT1A 0.19 (0.04) −0.0011 6.2 × 10−15 −0.0029 3.2 × 10−12 1.9 cg00851028 1 N/A 0.72 (0.04) 0.0010 5.4 × 10−8 0.0038 9.0 × 10−4 0.4 cg03421440 1 BRDT 0.71 (0.07) −0.0015 3.2 × 10−8 −0.0043 1.3 × 10−3 0.3 cg06096336 2 PSMD1 0.64 (0.05) 0.0016 4.3 × 10−8 0.0058 5.5 × 10−4 1.0 cg06500161 21 ABCG1 0.71 (0.03) 0.0011 1.7 × 10−9 0.0081 1.5 × 10−13 0.3 cg06872964 1 IFI44L 0.62 (0.06) 0.0015 4.8 × 10−8 0.0100 4.3 × 10−7 0.4 cg11024682 17 SREBF1 0.55 (0.04) 0.0013 6.6 × 10−15 0.0068 9.6 × 10−9 0.1 cg15159104 15 MAP1A 0.48 (0.05) 0.0010 3.2 × 10−8 0.0048 5.1 × 10−6 0.0 cg15903032 10 N/A 0.57 (0.04) 0.0010 7.6 × 10−8 0.0037 2.8 × 10−3 0.2 cg18030453 3 LARS2 0.72 (0.04) 0.0009 4.5 × 10−9 0.0028 1.7 × 10−3 0.1 cg21139312 17 MSI2 0.89 (0.03) 0.0009 4.5 × 10−10 0.0028 1.2 × 10−6 2.0 cg21506299 6 MAP3K5 0.23 (0.06) −0.0010 3.5 × 10−8 −0.0019 2.8 × 10−3 1.1 Abbreviations: ABCG1, ATP binding cassette, subfamily G, member 1 gene; ARIC, Atherosclerosis Risk in Communities; ATP, adenosine triphosphate; BMI, body mass index; BRDT, bromodomain testis associated gene; CPT1A, carnitine palmitoyltransferase 1A gene; ID, identification; IFI44L, interferon induced protein 44 like gene; LARS2, leucyl-tRNA synthetase 2, mitochondrial gene; MAP1A, microtubule associated protein 1A gene; MAP3K5, mitogen-activated protein kinase kinase 5 gene; MSI2, Musashi RNA binding protein 2 gene; N/A, not annotated; PSMD1, proteasome 26s subunit, non-ATPase 1 gene; RS, Rotterdam Study; SD, standard deviation; SREBF1, sterol regulatory element binding transcription factor 1 gene. a BMI was calculated as weight (kg)/height (m)2. b The methylation proportion of a 5′-C-phosphate-G-3′ site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). c Regression coefficient based on a linear mixed model and reflecting difference in the methylation β value per unit increase in BMI (kg/m2). Models adjusted for age, sex, current smoking, leukocyte proportions, array number, and position on array. d In the RS we corrected for multiple testing using a robust Bonferroni-corrected P value of 1.07 × 10−7, and in the ARIC Study the level of significance for replication was P < 3.57 × 10−3. Table 3. 5′-C-Phosphate-G-3′ Methylation Sites Associated With Waist Circumference at the Level of Genome-Wide Significance in the Rotterdam Study (2006–2013) and Successfully Replicated in the Atherosclerosis Risk in Communities Study (1990–1992) Probe ID No. Chromosome Gene Mean (SD) Methylationa Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βb P Valuec βb P Valuec cg00574958 11 CPT1A 0.19 (0.04) −0.0005 1.2 × 10−17 −0.0034 5.8 × 10−17 3.3 cg00851028 1 N/A 0.72 (0.04) 0.0004 6.0 × 10−9 0.0043 1.2 × 10−4 0 cg04927537 17 LGALS3BP 0.57 (0.05) 0.0006 7.0 × 10−8 0.0093 7.0 × 10−8 1.6 cg05899984 12 N/A 0.84 (0.03) 0.0003 8.1 × 10−8 0.0038 5.7 × 10−6 2.9 cg06500161 21 ABCG1 0.71 (0.03) 0.0005 2.4 × 10−12 0.0096 4.4 × 10−19 0.8 cg11024682 17 SREBF1 0.55 (0.04) 0.0005 2.9 × 10−15 0.0080 3.5 × 10−12 1.2 cg13139542 2 N/A 0.89 (0.02) 0.0002 6.0 × 10−8 0.0029 4.7 × 10−6 0 cg15416179 17 MAP2K3 0.14 (0.03) −0.0002 9.1 × 10−8 −0.0019 2.6 × 10−4 3.6 cg17901584 1 DHCR24 0.68 (0.07) −0.0005 1.7 × 10−8 −0.0080 8.3 × 10−8 2.0 cg18030453 3 LARS2 0.72 (0.04) 0.0003 8.8 × 10−8 0.0029 8.6 × 10−4 0 cg18772573 17 CPSF4L 0.85 (0.03) 0.0003 7.3 × 10−8 0.0039 2.8 × 10−5 0 cg21139312 17 MSI2 0.89 (0.03) 0.0004 5.9 × 10−12 0.0028 6.1 × 10−7 8.2 cg24174557 17 TMEM49 0.38 (0.07) −0.0005 1.1 × 10−8 −0.0059 5.3 × 10−5 0 Probe ID No. Chromosome Gene Mean (SD) Methylationa Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βb P Valuec βb P Valuec cg00574958 11 CPT1A 0.19 (0.04) −0.0005 1.2 × 10−17 −0.0034 5.8 × 10−17 3.3 cg00851028 1 N/A 0.72 (0.04) 0.0004 6.0 × 10−9 0.0043 1.2 × 10−4 0 cg04927537 17 LGALS3BP 0.57 (0.05) 0.0006 7.0 × 10−8 0.0093 7.0 × 10−8 1.6 cg05899984 12 N/A 0.84 (0.03) 0.0003 8.1 × 10−8 0.0038 5.7 × 10−6 2.9 cg06500161 21 ABCG1 0.71 (0.03) 0.0005 2.4 × 10−12 0.0096 4.4 × 10−19 0.8 cg11024682 17 SREBF1 0.55 (0.04) 0.0005 2.9 × 10−15 0.0080 3.5 × 10−12 1.2 cg13139542 2 N/A 0.89 (0.02) 0.0002 6.0 × 10−8 0.0029 4.7 × 10−6 0 cg15416179 17 MAP2K3 0.14 (0.03) −0.0002 9.1 × 10−8 −0.0019 2.6 × 10−4 3.6 cg17901584 1 DHCR24 0.68 (0.07) −0.0005 1.7 × 10−8 −0.0080 8.3 × 10−8 2.0 cg18030453 3 LARS2 0.72 (0.04) 0.0003 8.8 × 10−8 0.0029 8.6 × 10−4 0 cg18772573 17 CPSF4L 0.85 (0.03) 0.0003 7.3 × 10−8 0.0039 2.8 × 10−5 0 cg21139312 17 MSI2 0.89 (0.03) 0.0004 5.9 × 10−12 0.0028 6.1 × 10−7 8.2 cg24174557 17 TMEM49 0.38 (0.07) −0.0005 1.1 × 10−8 −0.0059 5.3 × 10−5 0 Abbreviations: ABCG1, ATP binding cassette, subfamily G, member 1; ARIC, Atherosclerosis Risk in Communities; ATP, adenosine triphosphate; CPSF4L, cleavage and polyadenylation specific factor 4 like gene; CPT1A, carnitine palmitoyltransferase 1A gene; DHCR24, 24-dehydrocholesterol reductase gene; ID, identification; LARS2, leucyl-TRNA synthetase 2, mitochondrial gene; LGALS3BP, galectin 3 binding protein gene; MAP2K3, mitogen-activated protein kinase kinase 3 gene; MSI2, Musashi RNA binding protein 2 gene; N/A, not annotated; RS, Rotterdam Study; SD, standard deviation; SREBF1, sterol regulatory element binding transcription factor 1 gene; TMEM49, transmembrane protein 49 gene. a The methylation proportion of a 5′-C-phosphate-G-3′ site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). b Regression coefficient based on a linear mixed model and reflecting difference in the methylation β value per unit increase in waist circumference (cm). Models adjusted for age, sex, current smoking, leukocyte proportions, array number, and position on array. c In the RS we corrected for multiple testing using a robust Bonferroni-corrected P value of 1.08 × 10−7, and in the ARIC Study the level of significance for replication was P < 1.92 × 10−3. Table 3. 5′-C-Phosphate-G-3′ Methylation Sites Associated With Waist Circumference at the Level of Genome-Wide Significance in the Rotterdam Study (2006–2013) and Successfully Replicated in the Atherosclerosis Risk in Communities Study (1990–1992) Probe ID No. Chromosome Gene Mean (SD) Methylationa Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βb P Valuec βb P Valuec cg00574958 11 CPT1A 0.19 (0.04) −0.0005 1.2 × 10−17 −0.0034 5.8 × 10−17 3.3 cg00851028 1 N/A 0.72 (0.04) 0.0004 6.0 × 10−9 0.0043 1.2 × 10−4 0 cg04927537 17 LGALS3BP 0.57 (0.05) 0.0006 7.0 × 10−8 0.0093 7.0 × 10−8 1.6 cg05899984 12 N/A 0.84 (0.03) 0.0003 8.1 × 10−8 0.0038 5.7 × 10−6 2.9 cg06500161 21 ABCG1 0.71 (0.03) 0.0005 2.4 × 10−12 0.0096 4.4 × 10−19 0.8 cg11024682 17 SREBF1 0.55 (0.04) 0.0005 2.9 × 10−15 0.0080 3.5 × 10−12 1.2 cg13139542 2 N/A 0.89 (0.02) 0.0002 6.0 × 10−8 0.0029 4.7 × 10−6 0 cg15416179 17 MAP2K3 0.14 (0.03) −0.0002 9.1 × 10−8 −0.0019 2.6 × 10−4 3.6 cg17901584 1 DHCR24 0.68 (0.07) −0.0005 1.7 × 10−8 −0.0080 8.3 × 10−8 2.0 cg18030453 3 LARS2 0.72 (0.04) 0.0003 8.8 × 10−8 0.0029 8.6 × 10−4 0 cg18772573 17 CPSF4L 0.85 (0.03) 0.0003 7.3 × 10−8 0.0039 2.8 × 10−5 0 cg21139312 17 MSI2 0.89 (0.03) 0.0004 5.9 × 10−12 0.0028 6.1 × 10−7 8.2 cg24174557 17 TMEM49 0.38 (0.07) −0.0005 1.1 × 10−8 −0.0059 5.3 × 10−5 0 Probe ID No. Chromosome Gene Mean (SD) Methylationa Discovery Panel (RS) Replication Panel (ARIC Study) % of Variance Explained βb P Valuec βb P Valuec cg00574958 11 CPT1A 0.19 (0.04) −0.0005 1.2 × 10−17 −0.0034 5.8 × 10−17 3.3 cg00851028 1 N/A 0.72 (0.04) 0.0004 6.0 × 10−9 0.0043 1.2 × 10−4 0 cg04927537 17 LGALS3BP 0.57 (0.05) 0.0006 7.0 × 10−8 0.0093 7.0 × 10−8 1.6 cg05899984 12 N/A 0.84 (0.03) 0.0003 8.1 × 10−8 0.0038 5.7 × 10−6 2.9 cg06500161 21 ABCG1 0.71 (0.03) 0.0005 2.4 × 10−12 0.0096 4.4 × 10−19 0.8 cg11024682 17 SREBF1 0.55 (0.04) 0.0005 2.9 × 10−15 0.0080 3.5 × 10−12 1.2 cg13139542 2 N/A 0.89 (0.02) 0.0002 6.0 × 10−8 0.0029 4.7 × 10−6 0 cg15416179 17 MAP2K3 0.14 (0.03) −0.0002 9.1 × 10−8 −0.0019 2.6 × 10−4 3.6 cg17901584 1 DHCR24 0.68 (0.07) −0.0005 1.7 × 10−8 −0.0080 8.3 × 10−8 2.0 cg18030453 3 LARS2 0.72 (0.04) 0.0003 8.8 × 10−8 0.0029 8.6 × 10−4 0 cg18772573 17 CPSF4L 0.85 (0.03) 0.0003 7.3 × 10−8 0.0039 2.8 × 10−5 0 cg21139312 17 MSI2 0.89 (0.03) 0.0004 5.9 × 10−12 0.0028 6.1 × 10−7 8.2 cg24174557 17 TMEM49 0.38 (0.07) −0.0005 1.1 × 10−8 −0.0059 5.3 × 10−5 0 Abbreviations: ABCG1, ATP binding cassette, subfamily G, member 1; ARIC, Atherosclerosis Risk in Communities; ATP, adenosine triphosphate; CPSF4L, cleavage and polyadenylation specific factor 4 like gene; CPT1A, carnitine palmitoyltransferase 1A gene; DHCR24, 24-dehydrocholesterol reductase gene; ID, identification; LARS2, leucyl-TRNA synthetase 2, mitochondrial gene; LGALS3BP, galectin 3 binding protein gene; MAP2K3, mitogen-activated protein kinase kinase 3 gene; MSI2, Musashi RNA binding protein 2 gene; N/A, not annotated; RS, Rotterdam Study; SD, standard deviation; SREBF1, sterol regulatory element binding transcription factor 1 gene; TMEM49, transmembrane protein 49 gene. a The methylation proportion of a 5′-C-phosphate-G-3′ site is reported as a β value ranging between 0 (no methylation) and 1 (full methylation). b Regression coefficient based on a linear mixed model and reflecting difference in the methylation β value per unit increase in waist circumference (cm). Models adjusted for age, sex, current smoking, leukocyte proportions, array number, and position on array. c In the RS we corrected for multiple testing using a robust Bonferroni-corrected P value of 1.08 × 10−7, and in the ARIC Study the level of significance for replication was P < 1.92 × 10−3. In addition to these novel findings, we confirmed previously reported associations of CpG sites with BMI and WC, including those on the carnitine palmitoyltransferase 1A gene (CPT1A), the adenosine triphosphate (ATP) binding cassette, subfamily G, member 1, gene (ABCG1), and the sterol regulatory element binding transcription factor 1 gene (SREBF1). Scatterplots of the associations between the replicated CpG sites and BMI and WC are shown in Web Figures 1 and 2, respectively. Figure 1 shows successfully replicated findings for BMI and WC and highlights the overlapping loci, including those on ABCG1, MSI2, LARS2, SREBF1, and CPT1A. To test for genomic inflation, we calculated the λ value for the EWAS on BMI and WC and created Q-Q plots. The λ values for the EWAS on BMI and WC were 1.487 and 1.556, respectively. The Q-Q plots for BMI and WC are shown in Web Figures 3 and 4, respectively. Figure 1. View largeDownload slide 5′-C-Phosphate-G-3′ sites for body mass index (BMI) and waist circumference (WC) identified in the Rotterdam Study (Rotterdam, the Netherlands; 2006–2013) and replicated in the Atherosclerosis Risk in Communities Study (4 US communities; 1990–1992) and their overlap. ABCG1, ATP binding cassette, subfamily G, member 1, gene; ATP, adenosine triphosphate; BRDT, bromodomain testis associated gene; CPSF4L, cleavage and polyadenylation specific factor 4 like gene; CPT1A, carnitine palmitoyltransferase 1A gene; DHCR24, 24-dehydrocholesterol reductase gene; IFI44L, interferon induced protein 44 like gene; LARS2, leucyl-tRNA synthetase 2, mitochondrial gene; LGALS3BP, galectin 3 binding protein gene; MAP1A, microtubule associated protein 1A gene; MAP2K3, mitogen-activated protein kinase kinase 3 gene; MAP3K5, mitogen-activated protein kinase kinase 5 gene; MSI2, Musashi RNA binding protein 2 gene; PSMD1, proteasome 26s subunit, non-ATPase 1 gene; SREBF1, sterol regulatory element binding transcription factor 1 gene. Figure 1. View largeDownload slide 5′-C-Phosphate-G-3′ sites for body mass index (BMI) and waist circumference (WC) identified in the Rotterdam Study (Rotterdam, the Netherlands; 2006–2013) and replicated in the Atherosclerosis Risk in Communities Study (4 US communities; 1990–1992) and their overlap. ABCG1, ATP binding cassette, subfamily G, member 1, gene; ATP, adenosine triphosphate; BRDT, bromodomain testis associated gene; CPSF4L, cleavage and polyadenylation specific factor 4 like gene; CPT1A, carnitine palmitoyltransferase 1A gene; DHCR24, 24-dehydrocholesterol reductase gene; IFI44L, interferon induced protein 44 like gene; LARS2, leucyl-tRNA synthetase 2, mitochondrial gene; LGALS3BP, galectin 3 binding protein gene; MAP1A, microtubule associated protein 1A gene; MAP2K3, mitogen-activated protein kinase kinase 3 gene; MAP3K5, mitogen-activated protein kinase kinase 5 gene; MSI2, Musashi RNA binding protein 2 gene; PSMD1, proteasome 26s subunit, non-ATPase 1 gene; SREBF1, sterol regulatory element binding transcription factor 1 gene. We calculated a methylation risk score based on the 12 CpG sites for BMI and 14 CpG sites for WC that were identified and replicated in the current study. For BMI, 2.0% of the variance was explained by the methylation risk score, whereas for WC the variance explained was 6.4%. DISCUSSION In this study, we used an EWAS approach to identify novel differentially methylated genes for obesity-related traits in older adults. The EWAS analysis in the RS data identified numerous novel loci associated with BMI and WC, of which many findings were successfully replicated in the ARIC data. The most significant CpG sites associated with both BMI and WC were located on the MSI2 and LARS2 genes. Additionally, CpG sites at BRDT and MAP1A were associated with BMI, and CpG sites at TMEM49 and LGALS3BP were associated with WC. Moreover, we confirmed previous findings that methylation at CPT1A, ABCG1, and SREBF1 is associated with BMI and WC. Previous EWAS on obesity traits were conducted in population-based studies, including ARIC and the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) Study (9, 10), and in persons with a history of myocardial infarction or healthy blood donors from the Cardiogenics Consortium (11). Similarly to our findings, ARIC and GOLDN investigators reported an inverse association between a CpG site at CPT1A and BMI (9, 10) and positive associations of CpG sites at ABCG1 and SREBF1 with BMI and WC (10). In the Cardiogenics Consortium, however, Dick et al. (11) reported only a positive association between 3 CpG sites on the hypoxia inducible factor 3, α subunit, gene (HIF3A) and BMI in both blood and adipose tissue DNA in European adults. These CpG sites at HIF3A did not achieve the threshold for statistical significance in our study. However, CpG sites at HIF3A were replicated in the ARIC Study in DNA blood (10). This discrepancy may be due to differences in the prevalences of obesity and comorbidity between our study (25% obesity, 11% diabetes, 7% coronary heart disease), the ARIC Study (44% obesity, 26% diabetes), and the Cardiogenics Consortium (4% diabetes, 52% myocardial infarction). The genes known to be associated with obesity—CPT1A, ABCG1, and SREBF1—are involved in regulation of lipids, lipoprotein metabolism, and insulin sensitivity (25–27). Specifically, CPT1A encodes for carnitine palmitoytransferase-1, which is a mitochondrial protein involved in fatty acid metabolism (28) and lipoprotein subfraction profile (26). ABCG1 encodes for the ATP binding cassette, subfamily G, member 1, protein and is involved in the transport of cholesterol and phospholipids in macrophages (29). Finally, SREBF1 encodes for sterol regulatory element binding transcription factor 1, which is known to promote adipocyte differentiation and signaling of insulin action (27). Although it has been shown previously that these loci are associated with obesity-related traits, it is still important to replicate these findings across different study populations. Since the EWAS approach is hypothesis-free, findings are prone to being false-positive. By replicating previously reported results, we can say with more certainty that these CpG sites are true-positive findings. In addition to confirming these previously identified loci, we have identified and replicated novel CpG sites located on the MSI2 (cg21139312) and LARS2 (cg18030453) genes, which were associated with both BMI and WC. The CpG site on MSI2 explained 2.2% of the variation in BMI and 8.2% of the variation in WC. MSI2 encodes RNA-binding proteins and plays a central role in posttranscriptional gene regulation (30). A genome-wide association study in pigs suggested that MSI2 is associated with eating behaviors, including number of visits to the feeder per day (31). Moreover, in another study performed in mice, Sakakibara et al. (32) reported that MSI2 is linked with the proliferation and maintenance of stem cells in the central nervous system. This study suggested that during neurogenesis, MSI2 expression persisted in a subset of neuronal lineage cells, such as parvalbumin-containing γ-aminobutyric acid (GABA) neurons in the neocortex (30, 32). GABA receptors are involved in controlling feeding behavior, reinforcing the role of MSI2 in obesity. The other novel locus associated with both BMI and WC, LARS2, encodes an enzyme that catalyzes aminoacylation of mitochondrial tRNALeu (33). A previous postmortem study showed that LARS2 expression (human leucyl-tRNA synthetase 2, mitochondrial NM015340) was increased in brain tissue of patients with bipolar disorder as compared with controls (34). Considering that bipolar disorder is associated with obesity, overweight, and abdominal obesity (35), methylation of MSI2 and LARS2 could play a role in disturbances in eating behaviors, and consequently in BMI and WC. However, further studies are warranted to establish the temporality and pathway of the associations. Although previous investigators have studied the association between DNA methylation and anthropometric characteristics, this study is the first, to our knowledge, to have found an association between DNA methylation of several CpG sites, including those at MSI2 and LARS2, and BMI and WC. One possible explanation for discrepancies between the findings of our current study and those of previous similar studies is the difference in population characteristics. Study populations in previous studies consisted of mixed ethnic groups, participants of younger ages, or persons at high disease risk (9–12). Considering that our discovery cohort consisted of an ethnically homogenous group of older adults from the general population of Rotterdam, underlying mechanisms may differ from those in other population groups. In this study, we conducted an EWAS in a European population and replicated the findings in African Americans. Epidemiologists have reported large disparities across racial/ethnic groups in the development of obesity (36). For example, in the current study, the rates of obesity were significantly lower in Europeans (24.6%) than in African Americans (43.8%). However, despite the differences in ethnicity and prevalence of obesity between our studies, most of our CpG sites (86.7%) were successfully replicated in ARIC. This may indicate that, in contrast to genetic studies, where replication across ethnic groups is challenging due to differences in linkage disequilibrium pattern, epigenetic findings could more easily be translated across ethnic groups. The strengths of the current study include the large size of the sample with available data on DNA methylation and the ability to replicate our findings in different ethnic populations. However, the results of this study must be interpreted in light of several limitations. We used whole blood samples for the quantification of DNA methylation, whereas adipose tissue may be a more relevant tissue for examining obesity. In this case, important CpG sites may not have been identified in our study. Unlike in genetic studies, unraveling the direction of the association between DNA methylation and phenotypes in epigenetic epidemiology remains challenging. Due to the cross-sectional study design and the nature of our variables, which are responsive to the environment and dynamic over time, a temporal direction in the association between DNA methylation and anthropometric measures cannot be determined. Because previous studies have shown that change in DNA methylation is a consequence of BMI for the majority of CpG sites, this may be the most likely direction for the associations observed in the current study as well (12). However, longitudinal studies are required to confirm the direction of the associations between DNA methylation and anthropometric factors. Another possibility is that our findings could be explained by third common factors. For instance, associations may have been confounded by differences in cell type proportion. In order to avoid this source of confounding, we adjusted all analyses for cell type proportions. However, as in any observational study, residual confounding due to various lifestyle factors still remains an issue. Furthermore, the Q-Q plots showed high genomic inflation. Many EWAS studies have had high genomic inflation (37). Adjustment for potential confounders such as technical covariates could decrease the inflation. Lehne et al. (38) have suggested that the correlation between CpG sites and the large number of findings in EWAS explain the residual inflation. In this study, we performed adequate adjustment for technical covariates. Moreover, the replication of our results in an independent population provides further evidence for the robustness of our findings. In conclusion, we have reported a novel association of increased methylation in the MSI2 and LARS2 genes with increased BMI and WC in older adults. Moreover, we confirmed 3 previously identified methylation loci (CPT1A, ABCG1, and SREBF1) suggested to be associated with obesity. Further investigations using repeatedly measured genome-wide DNA methylation and obesity-related traits are needed to assess causality and to further evolve the growing field of epigenetic epidemiology toward novel therapeutic and preventative approaches to obesity and related noncommunicable disorders. ACKNOWLEDGMENTS Author affiliations: Department of Epidemiology, Erasmus University Medical Center (Erasmus MC), Rotterdam, the Netherlands (Klodian Dhana, Kim V. E. Braun, Jana Nano, Trudy Voortman, Andre G. Uitterlinden, Albert Hofman, Oscar H. Franco, Abbas Dehghan); Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Klodian Dhana, Kim V. E. Braun); Rotterdam Intergenerational Ageing Research Center (ErasmusAGE), Rotterdam, the Netherlands (Kim V. E. Braun, Trudy Voortman, Oscar H. Franco); Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, Minneapolis, Minnesota (Ellen W. Demerath); Division of Biostatistics, School of Public Health, University of Minnesota, Minneapolis, Minnesota (Weihua Guan); Human Genetics Center, School of Public Health, University of Texas Health Sciences Center at Houston, Houston, Texas (Myriam Fornage); Human Genome Sequencing Center, Baylor College of Medicine, Houston, Texas (Myriam Fornage); Department of Internal Medicine, Erasmus University Medical Center (Erasmus MC), Rotterdam, the Netherlands (Joyce B. J. van Meurs, Andre G. Uitterlinden); Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Albert Hofman); and Department of Epidemiology, Imperial College London, London, United Kingdom (Abbas Dehghan). K.D. and K.V.E.B. contributed equally to this article. The Rotterdam Study is funded by Erasmus University Medical Center (Erasmus MC) and Erasmus University Rotterdam; the Netherlands Organization for Health Research and Development; the Research Institute for Diseases in the Elderly; the Netherlands Ministry of Education, Culture and Science; the Netherlands Ministry of Health, Welfare and Sport; the European Commission (Directorate-General XII); and the Municipality of Rotterdam. The ARIC Study is funded by the US National Heart, Lung, and Blood Institute. Generation and management of the Infinium Human Methylation 450K BeadChip array data for the Rotterdam Study was conducted by the Human Genotyping Facility of the Genetic Laboratory of the Department of Internal Medicine, Erasmus MC; funding was provided by Erasmus MC and the Netherlands Organization for Scientific Research (project 184021007). The data were made available as part of Rainbow Project 3 of the Biobanking and Biomolecular Resources Research Infrastructure–Netherlands. We thank Michael Verbiest, Mila Jhamai, Sarah Higgins, Marijn Verkerk, and Dr. Lisette Stolk for their help in creating the methylation database. We are grateful to the staff of and participants in the Rotterdam and ARIC studies and to all of the general practitioners and pharmacists involved. The funding organizations played no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, and approval of the manuscript. The Rotterdam Intergenerational Ageing Research Center (ErasmusAGE), a center for aging research across the life course, is supported by Nestlé Nutrition (Nestec Ltd., Lausanne, Switzerland) and Metagenics, Inc. (Aliso Viejo, California). Conflict of interest: none declared. Abbreviations ABCG1 ATP binding cassette, subfamily G, member 1, gene ARIC Atherosclerosis Risk in Communities ATP adenosine triphosphate BMI body mass index CpG 5′-C-phosphate-G-3′ CPT1A carnitine palmitoyltransferase 1A gene EWAS epigenome-wide association study GABA γ-aminobutyric acid GOLDN Genetics of Lipid Lowering Drugs and Diet Network LARS2 leucyl-tRNA synthetase 2, mitochondrial gene MSI2 Musashi RNA binding protein 2 gene RS Rotterdam Study SREBF1 sterol regulatory element binding transcription factor 1 gene WC waist circumference REFERENCES 1 World Health Organization . Obesity: Preventing and Managing the Global Epidemic. Report of a WHO Consultation. (WHO Technical Report Series, no. 894). Geneva, Switzerland: World Health Organization; 2000 . 2 Berrington de Gonzalez A , Hartge P , Cerhan JR , et al. . Body-mass index and mortality among 1.46 million white adults . N Engl J Med . 2010 ; 363 ( 23 ): 2211 – 2219 . Google Scholar CrossRef Search ADS PubMed 3 NCD Risk Factor Collaboration . Trends in adult body-mass index in 200 countries from 1975 to 2014: a pooled analysis of 1698 population-based measurement studies with 19.2 million participants . Lancet . 2016 ; 387 ( 10026 ): 1377 – 1396 . CrossRef Search ADS PubMed 4 Marti A , Martinez-Gonzalez MA , Martinez JA . Interaction between genes and lifestyle factors on obesity . Proc Nutr Soc . 2008 ; 67 ( 1 ): 1 – 8 . Google Scholar CrossRef Search ADS PubMed 5 Vattikuti S , Guo J , Chow CC . Heritability and genetic correlations explained by common SNPs for metabolic syndrome traits . PLoS Genet . 2012 ; 8 ( 3 ): e1002637 . Google Scholar CrossRef Search ADS PubMed 6 Bollati V , Baccarelli A . Environmental epigenetics . Heredity (Edinb) . 2010 ; 105 ( 1 ): 105 – 112 . Google Scholar CrossRef Search ADS PubMed 7 Mill J , Dempster E , Caspi A , et al. . Evidence for monozygotic twin (MZ) discordance in methylation level at two CpG sites in the promoter region of the catechol-O-methyltransferase (COMT) gene . Am J Med Genet B Neuropsychiatr Genet . 2006 ; 141B ( 4 ): 421 – 425 . Google Scholar CrossRef Search ADS PubMed 8 Jones PA . Functions of DNA methylation: islands, start sites, gene bodies and beyond . Nat Rev Genet . 2012 ; 13 ( 7 ): 484 – 492 . Google Scholar CrossRef Search ADS PubMed 9 Aslibekyan S , Demerath EW , Mendelson M , et al. . Epigenome-wide study identifies novel methylation loci associated with body mass index and waist circumference . Obesity (Silver Spring) . 2015 ; 23 ( 7 ): 1493 – 1501 . Google Scholar CrossRef Search ADS PubMed 10 Demerath EW , Guan W , Grove ML , et al. . Epigenome-wide association study (EWAS) of BMI, BMI change and waist circumference in African American adults identifies multiple replicated loci . Hum Mol Genet . 2015 ; 24 ( 15 ): 4464 – 4479 . Google Scholar CrossRef Search ADS PubMed 11 Dick KJ , Nelson CP , Tsaprouni L , et al. . DNA methylation and body-mass index: a genome-wide analysis . Lancet . 2014 ; 383 ( 9933 ): 1990 – 1998 . Google Scholar CrossRef Search ADS PubMed 12 Wahl S , Drong A , Lehne B , et al. . Epigenome-wide association study of body mass index, and the adverse outcomes of adiposity . Nature . 2017 ; 541 ( 7635 ): 81 – 86 . Google Scholar CrossRef Search ADS PubMed 13 Mendelson MM , Marioni RE , Joehanes R , et al. . Association of body mass index with DNA methylation and gene expression in blood cells and relations to cardiometabolic disease: a Mendelian randomization approach . PLoS Med . 2017 ; 14 ( 1 ): e1002215 . Google Scholar CrossRef Search ADS PubMed 14 Ikram MA , Brusselle GGO , Murad SD , et al. . The Rotterdam Study: 2018 update on objectives, design and main results . Eur J Epidemiol . 2017 ; 32 ( 9 ): 807 – 850 . Google Scholar CrossRef Search ADS PubMed 15 The ARIC Investigators . The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives . Am J Epidemiol . 1989 ; 129 ( 4 ): 687 – 702 . CrossRef Search ADS PubMed 16 Physical status: the use and interpretation of anthropometry. Report of a WHO Expert Committee . World Health Organ Tech Rep Ser . 1995 ; 854 : 1 – 452 . PubMed 17 Sandoval J , Heyn H , Moran S , et al. . Validation of a DNA methylation microarray for 450,000 CpG sites in the human genome . Epigenetics . 2011 ; 6 ( 6 ): 692 – 702 . Google Scholar CrossRef Search ADS PubMed 18 Touleimat N , Tost J . Complete pipeline for Infinium® Human Methylation 450K BeadChip data processing using subset quantile normalization for accurate DNA methylation estimation . Epigenomics . 2012 ; 4 ( 3 ): 325 – 341 . Google Scholar CrossRef Search ADS PubMed 19 Pidsley R , Wong CC , Volta M , et al. . A data-driven approach to preprocessing Illumina 450K methylation array data . BMC Genomics . 2013 ; 14 : 293 . Google Scholar CrossRef Search ADS PubMed 20 Willer CJ , Li Y , Abecasis GR . METAL: fast and efficient meta-analysis of genomewide association scans . Bioinformatics . 2010 ; 26 ( 17 ): 2190 – 2191 . Google Scholar CrossRef Search ADS PubMed 21 Houseman EA , Accomando WP , Koestler DC , et al. . DNA methylation arrays as surrogate measures of cell mixture distribution . BMC Bioinformatics . 2012 ; 13 : 86 . Google Scholar CrossRef Search ADS PubMed 22 Aryee MJ , Jaffe AE , Corrada-Bravo H , et al. . Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays . Bioinformatics . 2014 ; 30 ( 10 ): 1363 – 1369 . Google Scholar CrossRef Search ADS PubMed 23 Chen YA , Lemire M , Choufani S , et al. . Discovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray . Epigenetics . 2013 ; 8 ( 2 ): 203 – 209 . Google Scholar CrossRef Search ADS PubMed 24 Barfield RT , Almli LM , Kilaru V , et al. . Accounting for population stratification in DNA methylation studies . Genet Epidemiol . 2014 ; 38 ( 3 ): 231 – 241 . Google Scholar CrossRef Search ADS PubMed 25 Braun KV , Voortman T , Dhana K , et al. . The role of DNA methylation in dyslipidaemia: a systematic review . Prog Lipid Res . 2016 ; 64 : 178 – 191 . Google Scholar CrossRef Search ADS PubMed 26 Frazier-Wood AC , Aslibekyan S , Absher DM , et al. . Methylation at CPT1A locus is associated with lipoprotein subfraction profiles . J Lipid Res . 2014 ; 55 ( 7 ): 1324 – 1330 . Google Scholar CrossRef Search ADS PubMed 27 Kim JB , Spiegelman BM . ADD1/SREBP1 promotes adipocyte differentiation and gene expression linked to fatty acid metabolism . Genes Dev . 1996 ; 10 ( 9 ): 1096 – 1107 . Google Scholar CrossRef Search ADS PubMed 28 Gobin S , Thuillier L , Jogl G , et al. . Functional and structural basis of carnitine palmitoyltransferase 1A deficiency . J Biol Chem . 2003 ; 278 ( 50 ): 50428 – 50434 . Google Scholar CrossRef Search ADS PubMed 29 Klucken J , Büchler C , Orsó E , et al. . ABCG1 (ABC8), the human homolog of the drosophila white gene, is a regulator of macrophage cholesterol and phospholipid transport . Proc Natl Acad Sci USA . 2000 ; 97 ( 2 ): 817 – 822 . Google Scholar CrossRef Search ADS PubMed 30 Sakakibara S , Nakamura Y , Satoh H , et al. . RNA-binding protein Musashi2: developmentally regulated expression in neural precursor cells and subpopulations of neurons in mammalian CNS . J Neurosci . 2001 ; 21 ( 20 ): 8091 – 8107 . Google Scholar CrossRef Search ADS PubMed 31 Do DN , Strathe AB , Ostersen T , et al. . Genome-wide association study reveals genetic architecture of eating behavior in pigs and its implications for humans obesity by comparative mapping . PLoS One . 2013 ; 8 ( 8 ): e71509 . Google Scholar CrossRef Search ADS PubMed 32 Sakakibara S , Nakamura Y , Yoshida T , et al. . RNA-binding protein Musashi family: roles for CNS stem cells and a subpopulation of ependymal cells revealed by targeted disruption and antisense ablation . Proc Natl Acad Sci USA . 2002 ; 99 ( 23 ): 15194 – 15199 . Google Scholar CrossRef Search ADS PubMed 33 Sohm B , Sissler M , Park H , et al. . Recognition of human mitochondrial tRNALeu(UUR) by its cognate leucyl-tRNA synthetase . J Mol Biol . 2004 ; 339 ( 1 ): 17 – 29 . Google Scholar CrossRef Search ADS PubMed 34 Munakata K , Iwamoto K , Bundo M , et al. . Mitochondrial DNA 3243A>G mutation and increased expression of LARS2 gene in the brains of patients with bipolar disorder and schizophrenia . Biol Psychiatry . 2005 ; 57 ( 5 ): 525 – 532 . Google Scholar CrossRef Search ADS PubMed 35 McElroy SL , Kotwal R , Malhotra S , et al. . Are mood disorders and obesity related? A review for the mental health professional . J Clin Psychiatry . 2004 ; 65 ( 5 ): 634 – 651 , quiz 730. Google Scholar CrossRef Search ADS PubMed 36 Zhang Q , Wang Y , Huang ES . Changes in racial/ethnic disparities in the prevalence of type 2 diabetes by obesity level among U.S. adults . Ethn Health . 2009 ; 14 ( 5 ): 439 – 457 . Google Scholar CrossRef Search ADS PubMed 37 Joubert BR , den Dekker HT , Felix JF , et al. . Maternal plasma folate impacts differential DNA methylation in an epigenome-wide meta-analysis of newborns . Nat Commun . 2016 ; 7 : 10577 . Google Scholar CrossRef Search ADS PubMed 38 Lehne B , Drong AW , Loh M , et al. . A coherent approach for analysis of the Illumina Human Methylation450 BeadChip improves data quality and performance in epigenome-wide association studies . Genome Biol . 2015 ; 16 : 37 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

American Journal of EpidemiologyOxford University Press

Published: Aug 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off