Article | Open | Published:

DNA-Methylation and Body Composition in Preschool Children: Epigenome-Wide-Analysis in the European Childhood Obesity Project (CHOP)-Study

Scientific Reportsvolume 7, Article number: 14349 (2017) | Download Citation

Abstract

Adiposity and obesity result from the interaction of genetic variation and environmental factors from very early in life, possibly mediated by epigenetic processes. Few Epigenome-Wide-Association-Studies have identified DNA-methylation (DNAm) signatures associated with BMI and body composition in children. Body composition by Bio-Impedance-Analysis and genome-wide DNAm in whole blood were assessed in 374 pre-school children from four European countries. Associations were tested by linear regression adjusted for sex, age, centre, education, 6 WBC-proportions according to Houseman and 30 principal components derived from control probes. Specific DNAm variants were identified to be associated with BMI (212), fat-mass (230), fat-free-mass (120), fat-mass-index (24) and fat-free-mass-index (15). Probes in genes SNED1(IRE-BP1), KLHL6, WDR51A(POC1A), CYTH4-ELFN2, CFLAR, PRDM14, SOS1, ZNF643(ZFP69B), ST6GAL1, C3orf70, CILP2, MLLT4 and ncRNA LOC101929268 remained significantly associated after Bonferroni-correction of P-values. We provide novel evidence linking DNAm with (i) altered lipid and glucose metabolism, (ii) diabetes and (iii) body size and composition in children. Both common and specific epigenetic signatures among measures were also revealed. The causal direction with phenotypic measures and stability of DNAm variants throughout the life course remains unclear and longitudinal analysis in other populations is required. These findings give support for potential epigenetic programming of body composition and obesity.

Introduction

There is mounting evidence that the risk of obesity extends far beyond simple energy imbalance arising from a combination of overeating and sedentary lifestyle. This is supported by the variable incidence of associated comorbidities in overweight and obese individuals (e.g. cardiovascular disease (CVD), type 2 diabetes (T2D) and non-alcoholic fatty liver disease (NAFLD)). A greater understanding of aetiology is imperative for prevention and appropriate intervention in both, children and adults1,2,3.

Conceptualising obesity as a metabolic disease, that is developmentally programmed and thus conditioned on developmental pathways and plasticity, substantially broadens the focus beyond the obvious areas of excess food intake, lack of physical activity and excess weight gain. This is key to unravelling the functional and programmed aetiologies of metabolic dysregulation that begin early in life in response to genetic and environmental (including nutritional) factors and are potentially mediated by epigenetic mechanisms over the life course3,4,5,6,7,8.

Epigenetic ‘plasticity’ is now widely thought to underpin many developmental processes in humans in response to a ‘perceived’ postnatal environment(s), with many observational studies in humans implicating this process in the programming of the developmental and metabolic pathways towards obesity5,8,9,10,11. Building on the key observations of Barker and others, there is now a considerable body of evidence that in utero and early life factors are critical in influencing subsequent metabolic risk, although the underlying mechanisms are largely unexplored. It is also clear that key metabolic risk factors, including body mass index (BMI), blood pressure and cholesterol, track from late childhood into adulthood12,13 (reviewed in14,15,16). However, the development of metabolic risk in the period between infancy and adolescence is poorly understood. Considerable successes have been reported in adults through Epigenome-wide Association Studies (EWAS), linking epigenetic variants (particularly DNAm) to specific metabolic phenotypes including obesity, BMI and T2D17,18,19,20,21,22,23. Some candidate gene studies have been carried out in children24,25,26,27. However, only few comprehensive EWAS analyses in childhood cohorts, with high quality body composition and metabolic measures, have been published specifically in early childhood28,29,30,31.

This study, part of the multicentre European Childhood Obesity Project (CHOP)32,33,34, aimed at investigating the association between measures of body size (absolute and WHO standardised BMI) and body composition (absolute and height standardised fat mass and fat free mass measures) with DNAm on a genome-wide scale in 374 pre-school children. Functional pathway analyses were used to elucidate the likely metabolic and biological functions of genes showing altered DNAm to build evidence of a plausible link with obesity and adiposity related traits.

Results

DNAm variation, BMI and WHO-Standardised BMI

A total of 212 differentially methylated probes (DMPs) located in or near 181 genes were significantly associated with BMI after accounting for multiple testing by FDR (Supplementary Table S1). The top ten are listed below (Table 1). Methylation levels at two probes (cg13850887, cg01706498), located in genes SNED1 and KLHL6, remained significantly associated with BMI even after Bonferroni correction.

Table 1 Change in BMI, ZBMI, FM, FMI, FFM and FFMI for the top ten differentially methylated probes per percent change in DNA-methylation.

Methylation level at only four sites (located in SNED1, KLHL6, FARP1, or 10kb upstream of ZDHHC17) was significantly associated (Table 1, Supplementary Table S2) with WHO standardised BMI (ZBMI) as an outcome. The top two (cg13850887, cg01706498) were consistent irrespective of BMI measure (Table 1), while variants in FARP1, TBCD, ZNF750 and CILP2 remained among the top ten significant probes in both comparisons.

A summary of P-values (Manhattan plot) for each probe association with BMI and WHO standardised BMI is provided in Figs 1 and 2.

Figure 1
Figure 1

Manhattan plot of all HM450K probe P-values for the association of BMI (kg/m2) and methylation sites in preschool children. Top 10 sites are annotated. Red line indicates Bonferroni genome-wide significance and the blue line FDR significance. Chromosomal location of all HM450K probes is listed on the x-axis.

Figure 2
Figure 2

Manhattan plot of all HM450K probe P-values for the association of WHO-standardised BMI (z-score) and methylation sites in preschool children. Top 10 sites are annotated. Red line indicates Bonferroni genome-wide significance and the blue line FDR significance. Chromosomal location of all HM450K probes is listed on the x-axis.

DNAm variation, FM and FMI

A total of 230 and 24 DMPs were significantly associated with absolute fat mass (FM, kg) and fat mass index (FMI, kg/m²), after FDR correction for multiple testing (Supplementary Tables S3, S4). These are associated with 199 and 22 gene regions respectively. Probes in SNED1, WDR51A, CYTH4-ELFN2, CFLAR and PRDM14 remained significantly associated with FM, while those in SNED1, SOS1, CFLAR, ZNF643 and ST6GAL1 remained associated with FMI following Bonferroni correction (Table 1). The top DMP in both comparisons was cg13850887, located just downstream of a CpG island (South Shore) in the gene body of SNED1 on chromosome 2.

A summary of P-values (Manhattan plot) for each probe association with FM and FMI is provided in Figs 3 and 4).

Figure 3
Figure 3

Manhattan plot of all HM450K probe P-values for the association of FM (kg) and methylation sites in preschool children. Top 10 sites are annotated. Red line indicates Bonferroni genome-wide significance and the blue line FDR significance. Chromosomal location of all HM450K probes is listed on the x-axis.

Figure 4
Figure 4

Manhattan plot of all HM450K probe P-values for the association of FMI (kg/m2) and methylation sites in preschool children. Top 10 sites are annotated. Red line indicates Bonferroni genome-wide significance and the blue line FDR significance. Chromosomal location of all HM450K probes is listed on the x-axis.

DNAm variation, FFM and FFMI

Absolute fat free mass (FFM, kg) and fat free mass index (FFMI, kg/m²) was significantly associated with 120 and 15 methylation variants (Supplementary Tables S5, S6), located in 109 and 14 gene regions, respectively. After Bonferroni correction for multiple testing, only three variants remained significantly associated with FFM, located in the non-coding RNA LOC101929268, upstream of MLLT4 and in the uncharacterised C3orf70 gene, while a single probe in gene CILP2 was associated with FFMI (Table 1).

A summary of P-values (Manhattan plot) for each probe association with FFM and FFMI is provided in Figs 5 and 6).

Figure 5
Figure 5

Manhattan plot of all HM450K probe P-values for the association of FFM (kg) and methylation sites in preschool children. Top 10 sites are annotated. Red line indicates Bonferroni genome-wide significance and the blue line FDR significance. Chromosomal location of all HM450K probes is listed on the x-axis.

Figure 6
Figure 6

Manhattan plot of all HM450K probe P-values for the association of FFMI (kg/m2) and methylation sites in preschool children. Top 10 sites are annotated. Red line indicates Bonferroni genome-wide significance and the blue line FDR significance. Chromosomal location of all HM450K probes is listed on the x-axis.

Overlap of hits

DMPs commonly associated with BMI, FMI and FFMI are presented in Fig. 7a. In total, two DMPs, one upstream of gene ZDHHC17 (cg21525627) and another in KLHL6 (cg01706498), were associated with all three measures with the same direction of effect. Of 24 DMPs associated with FMI, 19 were also associated with BMI. Out of 15 associated with FFMI, 14 were also associated with BMI.

Figure 7
Figure 7

Overlap of differentially methylated probes (DMPs) associated with BMI, FMI and FFMI (a) and with FM and FFM (b) in preschool children. A full list of DMPs is provided in Supplementary Tables S1 to S6. Genes in brackets indicate manual annotation for closest genes using the UCSC genome browser on human GRCh37/hg19 assembly.

Figure 7b shows the 36 DMPs associated with both absolute FM and FFM among the 230 and 120 hits, respectively. Among these, 24 are annotated to protein-coding gene bodies or their promoters. The remaining 12 are situated in gene-poor regions. One probe (cg26995653) is located in a long non-coding RNA LINC01115 and the other (cg14795409) close to LINC01565. Despite some overlap in differential methylation, the majority of probes significantly associated with FM or FFM are specific for each measure.

Functional Pathway Analysis

The resulting gene ontology (GO) terms for associations of DNAm with body size and composition measures are summarised in Supplementary Table S7. BMI-associations were over-represented for ‘Ras, MAPK, SHC-mediated signalling cascades by fibroblast growth factor receptors (FGFR1 to 4)’, and ‘Neurotrophin signalling pathways’. ‘Regulation of cytoskeleton organization’, ‘cardiac and striated muscle cell proliferation’ and ‘neuronal system’ were also enriched. Genes showing differential methylation in association with FM were over-represented for ‘Ras and Neurotrophin signalling pathways’, ‘spleen development’, ‘muscle cell proliferation’ and ‘pro-B cell differentiation’, while those associated with FFM included ‘Hippo signalling’, ‘negative regulation of growth’, ‘odontogenesis’, ‘cartilage development’ and ‘gliogenesis and neuron fate commitment’.

Discussion

In this study of peripheral blood from 374 European pre-school children with highly accurate measures of body composition, we identified 212 differentially methylated probes (DMPs) associated with BMI, 230 DMPs with FM, 120 DMPs with FFM, 24 DMPs with FMI and 15 DMPs with FFMI. These probes measure DNAm in CpG sites that generally reside in protein-coding genes and non-coding RNAs previously linked with inflammation, glucose and lipid metabolism, browning of white fat, obesity and diabetes. Thus, many of the associations potentially make sense from a biological perspective; though only limited information is currently available in several instances (see below). Our finding that some DMPs are specifically associated with only one of the body size or composition measures BMI, FM, FMI, FFM, and FFMI whereas others are common to some or all measures e.g. FM and FFM, may be worthwhile investigating in more depth in future longitudinal studies to gain further insights into the process of body composition and obesity development and the role such potential epigenetic markers may play.

In the following paragraphs we provide some information on the 13 genes in which DNAm variants significantly associated after Bonferroni correction are located or are close by to add to the biological meaning and plausibility of our findings. These genes are SNED1(IRE-BP1), KLHL6, WDR51A(POC1A), CYTH4-ELFN2, CFLAR, PRDM14, SOS1, ZNF643(ZFP69B), ST6GAL1, C3orf70, MLLT4, CILP2 and the ncRNA LOC101929268.

SNED1 (Sushi, nidogen and EGF like domains), also known as IRE-BP1 (insulin-responsive sequence DNA-binding protein 1), activates insulin responsive genes IGF-I, IGFBP-1 and IGFBP-3 35. In is expressed in insulin-responsive tissues such as fat and muscle35 and in hypothalamic regions involved in control of appetite and energy balance36. In animal models, overexpression of SNED1 reduces37, but in some instances also increases38, hyperglycaemia and diabetes associated phenotypes. In our study, DMP cg13850887 at SNED1 was inversely associated with BMI, FM and FMI after Bonferroni correction.

KLHL6 (Kelch like family member 6) regulates B cell differentiation and potentially plays a role in diabetes, as it appears up-regulated in islet cells of mice with type 1 diabetes (T1D)39, in children with pre-T1D40 and in post-mortem pancreatic tissue of adult T1D patients41. In our study, DMP cg01706498 within this gene was positively associated with BMI after Bonferroni correction and with ZBMI, FM, FMI, FFM and FFMI according to FDR.

WDR51A (WD repeat domain 51A), also known as POC1A (centriolar protein A) have been found to show mutations in the centrosomal gene for individuals with primordial dwarfism and extreme insulin resistance42. We found a positive association between DNAm for DMP cg08692210 at this region with FM after Bonferroni correction and with BMI and FMI according to FDR.

CYTH4-ELFN2 gene region (cytohesin 4, extracellular leucine-rich repeat and fibronectin type III domain containing 2) was closest to the DMP cg14518658 associated with FM that we found in our study after Bonferroni correction and with BMI and FMI according to FDR. This region overlaps with binding sites of many transcription factors including the insulator protein CTCF and is close to a cluster of piwi-interacting RNAs (piRNAs) that have been linked to nutritional status and paternal inheritance of obesity43,44. Interestingly, expression of the two flanking genes is affected by diet: CYTH4, involved in vesicular secretion becomes upregulated in brown adipose tissue upon diet-induced obesity45 whereas ELFN2, a protein phosphatase regulator, becomes upregulated upon switch to Nordic diet46.

CFLAR (CASP8 and FADD like apoptosis regulator), a gene involved in regulating the pro-apoptotic protein CFLAR, is upregulated in peripheral blood mononuclear cells of obese children and adolescents and normalises after BMI reduction47. We found an inverse association of DNAm for DMP cg26627956 within this gene with FM and FMI after Bonferroni correction and with BMI according to FDR.

PRDM14 (PR/SET domain 14, previously PR domain containing 14) is a pluripotency gene belonging to the PRDM family of transcriptional regulators. This family includes PRDM16, which promotes differentiation of adult skeletal muscle stem cells48 and PRDM4, which stimulates browning of white adipose tissue49. We found a positive association of DMP cg00384539 in the PRDM14 gene with FM after Bonferroni correction and with BMI and FMI according to FDR. Interestingly, we found further associations within the PRDM family: DMP cg14282798 in PRDM15 was associated with BMI and FFM and DMP cg01360054 in PRDM6 with BMI according to FDR.

SOS1 (SOS Ras/Rac guanine nucleotide exchange factor 1) was recently reported to be transcriptionally upregulated in association with BMI in young adult monozygotic BMI-discordant Finnish twins50. A DMP (cg04010122) within this gene was identified in our study, which was associated, according to Bonferroni, with FMI and according to FDR with FM.

ZNF643(Zinc finger protein 643 also known as ZFP69B), a putative transcription factor gene, is situated next to ZFP69, which has been linked to pathogenesis of human diabetes, as its allelic variation associates with impaired lipid storage in white adipose tissue51. We found an inverse association of DNAm at cg26401512 in this gene with FMI according to Bonferroni correction and with FM according to FDR.

ST6GAL1 (ST6 beta-galactosamide alpha-2,6-sialyltranferase 1) codes for a membrane-bound glycosyltransferase and has been found to have a strong linkage with a high impact genetic variant in the adiponectin gene ADIPOQ that affects adiponectin plasma levels in Hispanics52. In our study DMP cg15026574 was inversely associated with FMI according to Bonferroni correction and with BMI and FM according to FDR.

C3orf70 (Chromosome 3 open reading frame 70). Gene polymorphisms in that gene code for an unknown protein and has also been found to have a strong linkage with a high impact genetic variant in adiponectin gene ADIPOQ that affects adiponectin plasma levels in Hispanics52. In our study, DMP cg 24332767 was positively associated with FFM after Bonferroni correction, and with BMI and FM according to FDR. Interestingly, KLHL6 gene is also located around 3 Mb upstream of this region on chromosome 3q26-27.

LOC101929268 (uncharacterised noncoding RNA). This ncRNA gene was closest to the DMP cg27038634 that we found in our study to be associated with FFM after Bonferroni correction and with BMI and FM according to FDR. LOC101929268 is upstream of the EFCAB1 gene for which altered methylation was detected in adipose tissue of individuals with type 2 diabetes53.

MLLT4 (myeloid/lymphoid or mixed-lineage leukaemia, translocated to, 4, also known as AFDN = afadin, adherens junction formation factor). This gene encodes a multi-domain protein involved in signalling and organisation of cell junctions during embryogenesis according to a pubmed/gene and HGNC search (http://www.genenames.org/) and was originally detected in relation to mouse development54. In our study, DMP cg08074767 near this gene was inversely associated with FFM after Bonferroni correction and with BMI, FM and FFMI according to FDR.

CILP2 (Cartilage intermediate layer protein 2). Genetic variation in CILP2 antagonizes insulin-like growth factor (IGF-I) activity in chondrocytes55 and has been linked with serum lipid levels56 and type 2 diabetes57. Methylation at CILP2 (cg17935297) was the top hit association for FFMI in our study after Bonferroni correction and was associated with BMI, FM and FFM according to FDR.

Two DNAm variants cg01706498 in ZDHHC17 and cg21525627 in KLHL6 were in our EWAS associated with all six (BMI, ZBMI, FM, FMI, FFM and FFMI) body size and composition measures after applying FDR threshold were found. ZDHHC17 encodes for a palmitoyltransferase that regulates glucose homeostasis in adipocytes and has been proposed as a candidate gene for type 1 diabetes58,59,60.

It is also of interest that we found in this EWAS of the CHOP study, a methylation variant located in an unknown non-coding RNA LINC01115 located approx. 100 kb downstream of the ‘obesity gene’ TMEM18 that is expressed in the hypothalamus and has been linked with childhood obesity61, early onset extreme obesity62 and adult obesity6.

None of the identified differently methylated positions (DMP) in our study were found in respective EWAS in children and adolescence28,29,31. However, DMP in the same genes often just in some 1000 bp distance of each other were found in an EWAS conducted in West Australian obese 12 year old boys28. Associations with phenotypes at DMP in common genes include the genes ANKRD11 (BMI, FM), BAT2 (BMI, FM), BCL11A (FM), C2orf85 (BMI), C4orf22 (BMI), CCNL2 (BMI), CCR6 (BMI, FM, FFM), CDH13 (BMI), CNGA3 (BMI, FM), CYFIP1 (FM, FMI), DIABLO (BMI), DUSP5 (FMI), FAM188B (BMI), FGFR2 (BMI), FOXK2 (BMI, FM, FFM), GNA12 (BMI), HIPK2 (BMI), IGF2R (BMI), INPP5A (BMI, FFM), JARID2 (FM), MACROD1 (BMI), MAD1L1 (FFM), MCF2L (FM), MEGF11 (BMI, FM), MLLT4 (FM, FFM, FFMI), MXD3 (FFM), NPSR1 (BMI, FM, FMI), PRKDC (FM, FMI), PTPRN2 (FFM), RAB5C (BMI, FM), RERE (FM), RPS6KA2 (FM, FFM), SFMBT2 (BMI, FM, FMI), TBCD (BMI, FM), TNXB (FM, FFM), TRAPPC9 (FM), and TRIM39 (BMI, FM). Supplementary Tables 8 to 12 list all FDR-significant DMPs of the CHOP study for the phenotypes separately and those DMPs of the Australian study that have at least a 5% difference in methylation among obese boys and an age-matched control group located within the above mentioned common genes28.

Several genes in our study have also been found differentially methylated in relation to BMI and obesity in previous EWAS in adults18,21,63.

An EWAS in an Arab population found a genome-wide significant inverse association of BMI with a DMP (cg17501210) in gene RPS6KA2 (Chr6:166970252) in whole blood of adults18. We found in the same gene at DMP (cg01577646, Chr6:166911121) an inverse association with FM and FFM according to FDR criteria.

In an EWAS in African American adults, a genome-wide significant positive association with BMI was found at DMP (cg09664445, Chr17: 2612406) in gene KIAA0664 21. We also found a positive association with BMI in the same gene at DMP (cg09927637, Chr17: 2606848) according to FDR criteria.

In a very recent EWAS conducted in adult European and Indian Asian populations, a large number (187 markers) of associations of DNAm and BMI were identified and replicated in several populations and some in adipose tissue63. In our study, we found several DMP located in the same genes associated with BMI or other phenotypes. Associations with phenotypes at DMP in common genes include ANKRD11(BMI, FM), JARID2 (FM), MAD1L1 (FFM), USP22 (BMI), RPS6KA2 (FM, FFM), SLC41A1 (FM), SMC3 (FMI) and ZC3H3 (BMI, FM, FFM). Supplementary Tables S13 to S16 list all overlaps in methylated genes of our study in European children and the EWAS in European or Indian Asian adults for the phenotypes separately63.

However, in contrast to several previous studies21,22,26,29, we did not detect differential methylation at HIF3A gene locus that has been detected in obese adults and children. In fact in our study, the smallest uncorrected P-value for an association of phenotype BMI, ZBMI, FM, FMI, FFM and FFMI with a methylation-site in gene HIF3A was P > 0.037, P > 0.048, P > 0.022, P > 0.018, P > 0.181 and P > 0.051 respectively for cg26749414 (cg02879662 for FFM). For all other methylation sites in HIF3A associations showed uncorrected P-values well above P > 0.11 (see Supplementary Tables S17 and S18). Nevertheless, we cannot rule out that this finding is due to a lack in power resulting from our sample size of 374 children, as some of the EWAS in adults are based on substantially larger samples.

This EWAS study has several strengths. It is one of the first to compare genome-wide DNAm among several body size and body composition measures in pre-school children with a reasonably large sample size. With the exception of one EWAS study29 in children that focused in their publication mainly on BMI - no EWAS in children investigated methylation in more than one body composition measure and in fat-mass or fat-free mass measured by bio-impedance analysis as in our study. This EWAS study provides novel evidence to the relatively less investigated field of epigenetics of childhood body composition. As demonstrated above, our findings are plausible from a biological point of view despite the necessarily explorative character intrinsic in EWAS analyses with a cross-sectional design. Our finding of specific and common methylation of body composition measures - if replicable in other populations – implies a promising potential for future research into the ways of epigenetic programming of obesity.

This EWAS study also has some limitations. First and foremost, a replication of our findings in further pre-school populations is required, despite the fact that some results were in line with previous studies in older children28 and in adult populations18,21,63. Although our sample size is larger than most of the few EWAS on body size or composition in children and adolescents28,29,31 a much larger sample size would be required to be sure that our null–association of DNAm variants within the HIF3A gene and BMI – shown previously in adults21,22 and children26 is not just a power problem. Moreover, we cannot rule out that other associated methylation sites, due to sample size, were missed.

A further limitation of our EWAS study is the cross-sectional design. Therefore we cannot determine the causal nature of our findings. In particular we cannot rule out that the found methylation levels are a consequence rather than a cause of our body composition phenotypes63. Moreover, the causal pathways may be even more complicated by genetic and other environmental influences, epigenetic mediation, modification and mechanisms for gene-environment interactions64,65,66. The list of corresponding methylation trait loci (mQTL) of our findings with those of the ARIES study, listed in Supplementary Tables S19 to S22, points in that direction66.

The issue of tissue specific methylation is a further limitation of our study – evaluating DNA-methylation in children’s blood and not in adipose tissue (for practical and ethical reasons) may have missed the true methylation markers associated with body composition67,68. In EWAS in adults it has been shown that DNAm assessed from subcutaneous adipose tissue was associated with BMI, DXA-measured centrally located fat and body fat distribution, whereas DNAm derived from blood was not17. Moreover, in a small EWAS of adult monozygotic twins discordant for obesity, it was shown that genome-wide DNAm variants determined from blood were not different among the twin-pairs. However, when stratifying the twin-pairs by level of liver fat accumulation, epigenetically different signatures were observed if the heavier co-twins had excessive liver fat23. On the other hand, this methods study reveals that if the metabolic status can be accounted for by metabolic measurements often associated with excess adipose tissue (higher levels of fasting glucose and insulin, higher-low-density lipoprotein, C-reactive protein, or higher diastolic blood pressure) differences in epigenetic profile relevant to the phenotype and tissue of interest may be detected from blood23.

A further issue is the high inflation according to genomic control lambda69 ranging from 1.24 to 146 for the 6 body size and composition measure models. High lambda values may be considered as casting some doubt on the number of methylation associations found. However, a recent methods study demonstrated that the usual inflation measure lambda is often an overestimate of the true inflation, as lambda is substantially increased dependent on the number of significant EWAS associations70. According to this recent method of “bacon” inflation, values in our study were substantially smaller, ranging from 1.10 to 1.17.

Using estimates of the 6 white blood cell types derived from adult blood according to the Houseman method71 and adjusting by this reference based approach for cell mix heterogeneity in pre-school children, could potentially be seen as a limitation of our analysis as blood cell mix levels from younger children differ somehow from those of adults. In fact, there is a vivid discussion on how to account for cell heterogeneity in EWAS71,72,73,74,75,76,77. However, adjustment with other reference based methods like Bakulski’s cord blood reference72 did not improve the genomic inflation lambda values in our study and most of the top 10 associations of the original analysis showed up as well. Using a reference based model to account for cell heterogeneity may be considered as a limitation in any case, as references derived from blood may not reflect the tissue-specificity of methylation of adipose tissue as discussed already above. However, using Houseman’s new reference free model “RefFreeEWAS”74,75 with K = 9 latent dimensions to account for potential collinearity of the phenotype and or covariates within the methylation value matrix, resulted in our study in an over-adjustment indicated by very low genomic inflation values (0.74 to 0.84) and extremely low “bacon”-inflation (0.10–0.21).

An implicit related limitation of our EWAS approach is that we cannot rule out that methylation values in our dataset are confounded due to collinearity with the body composition phenotypes. The reference free model potentially accounts for such collinearity. However, as mentioned above, applying the reference free method with the determined number of K = 9 estimated latent factors resulted in over-adjustment. Moreover, using just K = 8 or K = 7 latent adjustment factors resulted in quite different top 10 associations among the 3 models, despite the fact that genomic inflation lambdas were still below 1 even for K = 7. According to our knowledge it is still an open methodological research question to determine, which approach is best to use to account for cell heterogeneity in children’s EWAS – reference-based or reference-free methods. Therefore we have placed the results from the reference free approach in Supplementary Tables S23 to S28.

In the reported reference-based EWAS analyses maternal smoking during pregnancy was not adjusted for. This could be seen as a potential biasing factor. However, a repetition of the main EWAS analyses with this additional adjustment show that such a bias is limited. For all six body-size and composition measures almost all of the originally found top 10 associations with DNAm variants were also found within the top 10 in these additional adjusted analyses or were at least under the top 17. Moreover, these estimates did not differ substantially from those not adjusted for maternal smoking as documented in Supplementary Tables S29 to S34.

A further limitation of this EWAS is that no RNA data is available to study the associations of differential methylation at the found loci with gene expression. However, this may be worthwhile to investigate in further and larger EWAS studies.

In summary, we provide novel evidence linking DNAm at SNED1(IRE-BP1), KLHL6, WDR51A(POC1A), intergenic CYTH4-ELFN2, CFLAR, PRDM14, SOS1, ZNF643(ZFP69B), ST6GAL1, C3orf70, MLLT4, CILP2 genes and noncoding RNA LOC101929268 with altered lipid and glucose metabolism and differential body size and body composition measures in children. We also found methylation variation related to body composition in 39 genes previously found mostly in severely obese children but some also in adults. We also provide novel evidence about common and different epigenetic signatures between fat mass and fat free mass. The causal direction with phenotypic measures and stability of the DNAm variants throughout the life course remains unclear and longitudinal analysis in other populations is required. These findings give support for potential epigenetic programming of body composition and obesity and contribute to an emerging body of work linking specific exposures to variation in epigenetic profile and metabolic phenotypes in humans.

Methods

Study design and participants

This study is based on a subset of 374 children out of 543 children aged 5.5 years from study centres in Germany, Belgium, Italy and Spain of the European Childhood Obesity Trial Study (CHOP) registered at clinicaltrials.gov as NCT00338689 and URL: http://clinicaltrials.gov/ct2/show/NCT00338689?term=NCT00338689&rank=1. Details on the study have been published previously32,33,34. Inclusion criteria for this analysis were availability of blood buffy coats, valid exposure data on measured weight, height and Bio-Impedance Analysis (BIA) derived measurements of body composition (fat mass, fat free mass); all collected at 5.5 years of age in the children and information on basic characteristics of the offspring (sex, age of blood draw, country of study centre). Characteristics of the analysed study population are listed in Table 2.

Table 2 Characteristics of analysed study population.

Ethics statement

The CHOP study was conducted according to the principles expressed in the Declaration of Helsinki. The local ethics committees of each study centre approved all study procedures: Belgium (Comitè d’Ethique de L’Hopital Universitaire des Enfants Reine Fabiola; no. CEH 14/02), Germany (Bayerische Landesärztekammer Ethik-Kommission; no. 02070), Italy (Azienda Ospedaliera San Paolo Comitato Etico; no. 14/2002), Poland (Instytut Pomnik–Centrum Zdrowia Dziecka Komitet Etyczny; no 243/KE/2001), and Spain (Comité ético de investigación clinica del Hospital Universitario de Tarragona Joan XXIII). Written informed parental consent was obtained for each infant.


Procedures

Weight and height was measured to the nearest 100 g and 0.1 cm during physical exam at age 5.5 years by a trained study personal according to strict SOPs using SECA 702 electronic scales (SECA, Hamburg; Germany) and SECA 242 stadiometer (SECA, Hamburg; Germany). BMI was calculated from these anthropometric measurements as child’s weight in kg divided by height in metres squared (BMI = (weight (kg)/height (m²)). BMI was also standardised according to WHO age- and sex–specific child growth standards (ZBMI) using a macro downloaded from http://www.who.int/childgrowth/en. Fat Mass in kg (FM) and Fat Free Mass in kg (FFM) were determined by the built-in equation of the Bio-Impedance-Analysis device (Tanita BC-418MA, Segmental Body Composition Analyser, Tanita Europe, Sindelfingen Germany). Fat mass index and fat free mass index were derived from these measures: FMI = FM (kg)/Height (m2) and FFMI = FFM (kg)/Height (m2).

During the child’s physical examination at age 5.5 years, blood was drawn to collect peripheral blood cells from buffy coats. These were used to determine DNA methylation (DNAm) status at a genome-wide level by using the Infinium HumanMethylation450 BeadChip (HM450K). DNA extraction, bisulfite conversion and methylation analysis were performed at the Genome Analysis Center of Helmholtz Zentrum Muenchen, Munich, Germany. Details were described previously34. In brief, genomic DNA was extracted using a standard precipitation procedure. Bisulfite conversion was performed using the EZ-96 DNA Methylation Kit (Zymo Research, Irvine, Ca; USA) and converted DNA samples were hybridised on the Infinium HumanMethylation450 BeadChip (HM450K) according to the manufactures instructions (Illumina Inc., San Diego, USA). Data pre-processing and normalisation were performed by the first author according to the approach of Touleimat and Tost78 with some adaptions e.g. the beta-mixture quantile normalization (BMIQ) step of the ß-values79 and exclusion of cross-reactive probes80 as described in detail previously34.

The final data set comprised DNA-methylation values (DNAm) at 431313 CpG sites in each of the 374 children for EWAS analysis.


Statistical Analysis (EWAS)

Potential associations between each of the measures of body size (BMI, ZBMI) or body composition (FM, FMI, FFM, FFMI) and quality controlled but untransformed methylation values at each of 431313 CpG sites were determined by standard linear regression models with adjustment for child’s sex, age at blood draw (months), study centre (Germany (DE), Italy (IT), Spain(ES), reference Belgium (BE)), parental education (High = both parents 12 + yrs. of schooling, reference Middle = both parents 10 − <12 yrs. of schooling or only one parent 12 + years of schooling and Low = both parents achieved basic schooling only), the estimated proportions of six major white blood cell types (WBC), CD4 + T cells, CD8 + T cells, B cells, NK cells, monocytes and granulocytes according to Houseman’s method and the top 30 principal components (PC) derived from control probes on the HM450K platform71,81. All statistical analyses were performed with R-software version R3.3.2 of the R-project (https://www.r-project.org/) at the mainframe of the Leibniz-Rechenzentrum (LRZ) in Garching/Munich.

An association of a differently methylated CpG site (beta-values) with the respective body size or composition measure was considered significant at the epigenome-wide level, if the false discovery rate (FDR) was q < 0.0582. The estimates listed in tables in the result section or the appendix are the adjusted regression coefficient of the model described above where a CpG site enters the equation as methylation fraction (range 0 to 1) and is scaled to percent methylation by dividing the CpG related regression coefficient by 100. Therefore the shown estimates can be interpreted as the respective change in the analysed outcome for a one percent change in methylation of the respective CpG (predictor). Inflation of P-values for the assessed associations in this EWAS were computed according to both the usual genomic control inflation factor lambda69 and according to the recent “bacon” method that account for the number of significant associations70. Both inflation estimates are depicted in Q-Q-plots for each of the models of the body size and composition measures (see Supplementary Fig. S1).


Functional characterization of differentially methylated CpGs

The identified CpG sites were annotated according to Illumina (www.illumina.com, HumanMethylation450_15017482_v1.csv) or manually using the UCSC genome browser, GRCh37/h19 assembly for non-annotated CpGs (https://genome.ucsc.edu). Ontology analyses were conducted using a fixed set gene enrichment analysis approach performed with g:Profiler (http://biit.cs.ut.ee/gprofiler/index.cgi)83. Pathway analysis included Gene Ontology (Biological Process, Cellular Component and Molecular Function), KEGG and Reactome gene-set databases. The analysis was performed on ranked gene lists (ranked according to P-value from EWAS regression analysis) with advanced options ‘Size of functional category’: 3 (min) to 500 (max) and ‘Size of Q&T’: min of 2 using the gSCS threshold that is more stringent than FDR.


Data availability

Results are extensively documented in the supplement. To protect patient confidentiality, data is available upon request. Future interested researchers can make requests to Prof Dr Berthold Koletzko, email: office.koletzko@med.lmu.de

Additional information

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  1. 1.

    Lobstein, T. et al. Child and adolescent obesity: part of a bigger picture. Lancet 385, 2510–20 (2015).

  2. 2.

    McAllister, E. J. et al. Ten putative contributors to the obesity epidemic. Crit Rev Food Sci Nutr 49, 868–913 (2009).

  3. 3.

    Stenvinkel, P. Obesity–a disease with many aetiologies disguised in the same oversized phenotype: has the overeating theory failed? Nephrol Dial Transplant 30, 1656–64 (2015).

  4. 4.

    Hagg, S. et al. Gene-based meta-analysis of genome-wide association studies implicates new loci involved in obesity. Hum Mol Genet 24, 6849–60 (2015).

  5. 5.

    Hanson, M. A. & Gluckman, P. D. Early developmental conditioning of later health and disease: physiology or pathophysiology? Physiol Rev 94, 1027–76 (2014).

  6. 6.

    Locke, A. E. et al. Genetic studies of body mass index yield new insights for obesity biology. Nature 518, 197–206 (2015).

  7. 7.

    Wahlqvist, M. L. et al. Early-life influences on obesity: from preconception to adolescence. Ann N Y Acad Sci 1347, 1–28 (2015).

  8. 8.

    Waterland, R. A. Epigenetic mechanisms affecting regulation of energy balance: many questions, few answers. Annu Rev Nutr 34, 337–55 (2014).

  9. 9.

    Foley, D. L. et al. Prospects for epigenetic epidemiology. Am J Epidemiol 169, 389–400 (2009).

  10. 10.

    Kappil, M., Wright, R. O. & Sanders, A. P. Developmental Origins of Common Disease: Epigenetic Contributions to Obesity. Annu Rev Genomics Hum Genet (2016).

  11. 11.

    Koletzko, B. et al. Early nutrition programming of long-term health. Proc Nutr Soc 71, 371–8 (2012).

  12. 12.

    Jarvisalo, M. J. et al. Increased aortic intima-media thickness: a marker of preclinical atherosclerosis in high-risk children. Circulation 104, 2943–7 (2001).

  13. 13.

    Juhola, J. et al. Tracking of serum lipid levels, blood pressure, and body mass index from childhood to adulthood: the Cardiovascular Risk in Young Finns Study. J Pediatr 159, 584–90 (2011).

  14. 14.

    Kones, R. Primary prevention of coronary heart disease: integration of new data, evolving views, revised goals, and role of rosuvastatin in management. A comprehensive survey. Drug Des Devel Ther 5, 325–80 (2011).

  15. 15.

    Kwiterovich, P. O. & Gidding, S. S. Universal screening of cholesterol in children. Clin Cardiol 35, 662–4 (2012).

  16. 16.

    Sun, C. et al. Effects of early-life environment and epigenetics on cardiovascular disease risk in children: highlighting the role of twin studies. Pediatr Res 73, 523–30 (2013).

  17. 17.

    Agha, G. et al. Adiposity is associated with DNA methylation profile in adipose tissue. Int J Epidemiol 44, 1277–87 (2015).

  18. 18.

    Al Muftah, W. A. et al. Epigenetic associations of type 2 diabetes and BMI in an Arab population. Clin Epigenetics 8, 13 (2016).

  19. 19.

    Aslibekyan, S. et al. Epigenome-wide study identifies novel methylation loci associated with body mass index and waist circumference. Obesity (Silver Spring) 23, 1493–501 (2015).

  20. 20.

    Benton, M. C. et al. An analysis of DNA methylation in human adipose tissue reveals differential modification of obesity genes before and after gastric bypass and weight loss. Genome Biol 16, 8 (2015).

  21. 21.

    Demerath, E. W. et al. Epigenome-wide association study (EWAS) of BMI, BMI change and waist circumference in African American adults identifies multiple replicated loci. Human Molecular Genetics (2015).

  22. 22.

    Dick, K. J. et al. DNA methylation and body-mass index: a genome-wide analysis. Lancet 383, 1990–8 (2014).

  23. 23.

    Ollikainen, M. et al. Genome-wide blood DNA methylation alterations at regulatory elements and heterochromatic regions in monozygotic twins discordant for obesity and liver fat. Clin Epigenetics 7, 39 (2015).

  24. 24.

    Godfrey, K. M. et al. Epigenetic gene promoter methylation at birth is associated with child’s later adiposity. Diabetes 60, 1528–34 (2011).

  25. 25.

    Groom, A. et al. Postnatal growth and DNA methylation are associated with differential gene expression of the TACSTD2 gene and childhood fat mass. Diabetes 61, 391–400 (2012).

  26. 26.

    Pan, H. et al. HIF3A association with adiposity: the story begins before birth. Epigenomics 7, 937–50 (2015).

  27. 27.

    Relton, C. L. et al. DNA methylation patterns in cord blood DNA and body size in childhood. PLoS One 7, e31821 (2012).

  28. 28.

    Huang, R. C. et al. Genome-wide methylation analysis identifies differentially methylated CpG loci associated with severe obesity in childhood. Epigenetics 10, 995–1005 (2015).

  29. 29.

    Richmond, R. C. et al. DNA Methylation and BMI: Investigating Identified Methylation Sites at HIF3A in a Causal Framework. Diabetes 65, 1231–44 (2016).

  30. 30.

    Sharp, G. C. et al. Maternal pre-pregnancy BMI and gestational weight gain, offspring DNA methylation and later offspring adiposity: findings from the Avon Longitudinal Study of Parents and Children. Int J Epidemiol (2015).

  31. 31.

    Rounge, T. B. et al. Genome-wide DNA methylation in saliva and body size of adolescent girls. Epigenomics 8, 1495–1505 (2016).

  32. 32.

    Koletzko, B. et al. Lower protein in infant formula is associated with lower weight up to age 2 y: a randomized clinical trial. Am J Clin Nutr 89, 1836–45 (2009).

  33. 33.

    Weber, M. et al. Lower protein content in infant formula reduces BMI and obesity risk at school age: follow-up of a randomized trial. Am J Clin Nutr (2014).

  34. 34.

    Rzehak, P. et al. Maternal Smoking during Pregnancy and DNA-Methylation in Children at Age 5.5 Years: Epigenome-Wide-Analysis in the European Childhood Obesity Project (CHOP)-Study. PLoS One 11, e0155554 (2016).

  35. 35.

    Chahal, J. et al. Regulation of insulin-response element binding protein-1 in obesity and diabetes: potential role in impaired insulin-induced gene transcription. Endocrinology 149, 4829–36 (2008).

  36. 36.

    Herwig, A. et al. A thyroid hormone challenge in hypothyroid rats identifies T3 regulated genes in the hypothalamus and in models with altered energy balance and glucose homeostasis. Thyroid 24, 1575–93 (2014).

  37. 37.

    Villafuerte, B. C. & Kaytor, E. N. An insulin-response element-binding protein that ameliorates hyperglycemia in diabetes. J Biol Chem 280, 20010–20 (2005).

  38. 38.

    Villafuerte, B. C. et al. Transgenic expression of insulin-response element binding protein-1 in beta-cells reproduces type 2 diabetes. Endocrinology 150, 2611–7 (2009).

  39. 39.

    Berry, G. J., Frielle, C., Brucklacher, R. M., Salzberg, A. C. & Waldner, H. Identifying type 1 diabetes candidate genes by DNA microarray analysis of islet-specific CD4 + T cells. Genom Data 5, 184–8 (2015).

  40. 40.

    Elo, L. L. et al. Early suppression of immune response pathways characterizes children with prediabetes in genome-wide gene expression profiling. J Autoimmun 35, 70–6 (2010).

  41. 41.

    Planas, R., Pujol-Borrell, R. & Vives-Pi, M. Global gene expression changes in type 1 diabetes: insights into autoimmune response in the target organ and in the periphery. Immunol Lett 133, 55–61 (2010).

  42. 42.

    Chen, J. H. et al. Truncation of POC1A associated with short stature and extreme insulin resistance. J Mol Endocrinol 55, 147–58 (2015).

  43. 43.

    Donkin, I. et al. Obesity and Bariatric Surgery Drive Epigenetic Variation of Spermatozoa in Humans. Cell Metab 23, 369–78 (2016).

  44. 44.

    Grandjean, V. et al. RNA-mediated paternal heredity of diet-induced obesity and metabolic disorders. Sci Rep 5, 18193 (2015).

  45. 45.

    McGregor, R. A. et al. Time-course microarrays reveal modulation of developmental, lipid metabolism and immune gene networks in intrascapular brown adipose tissue during the development of diet-induced obesity. Int J Obes (Lond) 37, 1524–31 (2013).

  46. 46.

    Kolehmainen, M. et al. Healthy Nordic diet downregulates the expression of genes involved in inflammation in subcutaneous adipose tissue in individuals with features of the metabolic syndrome. Am J Clin Nutr 101, 228–39 (2015).

  47. 47.

    Rainone, V. et al. Upregulation of inflammasome activity and increased gut permeability are associated with obesity in children and adolescents. Int J Obes (Lond) 40, 1026–33 (2016).

  48. 48.

    Seale, P. et al. PRDM16 controls a brown fat/skeletal muscle switch. Nature 454, 961–7 (2008).

  49. 49.

    Song, N. J. et al. Prdm4 induction by the small molecule butein promotes white adipose tissue browning. Nat Chem Biol 12, 479–81 (2016).

  50. 50.

    Pietilainen, K. H. et al. DNA methylation and gene expression patterns in adipose tissue differ significantly within young adult monozygotic BMI-discordant twin pairs. Int J Obes (Lond) 40, 654–61 (2016).

  51. 51.

    Scherneck, S. et al. Positional cloning of zinc finger domain transcription factor Zfp69, a candidate gene for obesity-associated diabetes contributed by mouse locus Nidd/SJL. PLoS Genet 5, e1000541 (2009).

  52. 52.

    Hellwege, J. N. et al. Empirical characteristics of family-based linkage to a complex trait: the ADIPOQ region and adiponectin levels. Hum Genet 134, 203–13 (2015).

  53. 53.

    Nilsson, E. et al. Altered DNA methylation and differential expression of genes influencing metabolism and inflammation in adipose tissue from subjects with type 2 diabetes. Diabetes 63, 2962–76 (2014).

  54. 54.

    Ikeda, W. et al. Afadin: A key molecule essential for structural organization of cell-cell junctions of polarized epithelia during embryogenesis. J Cell Biol 146, 1117–32 (1999).

  55. 55.

    Johnson, K., Farley, D., Hu, S. I. & Terkeltaub, R. One of two chondrocyte-expressed isoforms of cartilage intermediate-layer protein functions as an insulin-like growth factor 1 antagonist. Arthritis Rheum 48, 1302–14 (2003).

  56. 56.

    Tai, E. S. et al. Polymorphisms at newly identified lipid-associated loci are associated with blood lipids and cardiovascular disease in an Asian Malay population. J Lipid Res 50, 514–20 (2009).

  57. 57.

    Matsuba, R. et al. Replication Study in a Japanese Population to Evaluate the Association between 10 SNP Loci, Identified in European Genome-Wide Association Studies, and Type 2 Diabetes. PLoS One 10, e0126363 (2015).

  58. 58.

    Ren, W., Sun, Y. & Du, K. DHHC17 palmitoylates ClipR-59 and modulates ClipR-59 association with the plasma membrane. Mol Cell Biol 33, 4255–65 (2013).

  59. 59.

    Ren, W., Sun, Y. & Du, K. Glut4 palmitoylation at Cys223 plays a critical role in Glut4 membrane trafficking. Biochem Biophys Res Commun 460, 709–14 (2015).

  60. 60.

    Berchtold, L. A. et al. Huntingtin-interacting protein 14 is a type 1 diabetes candidate protein regulating insulin secretion and beta-cell apoptosis. Proc Natl Acad Sci U S A 108, E681–8 (2011).

  61. 61.

    Comuzzie, A. G. et al. Novel genetic loci identified for the pathophysiology of childhood obesity in the Hispanic population. PLoS One 7, e51954 (2012).

  62. 62.

    Scherag, A. et al. Two new Loci for body-weight regulation identified in a joint analysis of genome-wide association studies for early-onset extreme obesity in French and german study groups. PLoS Genet 6, e1000916 (2010).

  63. 63.

    Wahl, S. et al. Epigenome-wide association study of body mass index, and the adverse outcomes of adiposity. Nature 541, 81–86 (2017).

  64. 64.

    Ladd-Acosta, C. & Fallin, M. D. The role of epigenetics in genetic and environmental epidemiology. Epigenomics 8, 271–83 (2016).

  65. 65.

    Pan, H., Holbrook, J. D., Karnani, N. & Kwoh, C. K. Gene, Environment and Methylation (GEM): a tool suite to efficiently navigate large scale epigenome wide association studies and integrate genotype and interaction between genotype and environment. BMC Bioinformatics 17, 299 (2016).

  66. 66.

    Gaunt, T. R. et al. Systematic identification of genetic influences on methylation across the human life course. Genome Biol 17, 61 (2016).

  67. 67.

    Crujeiras, A. B. et al. DNA methylation map in circulating leukocytes mirrors subcutaneous adipose tissue methylation pattern: a genome-wide analysis from non-obese and obese patients. Sci Rep 7, 41903 (2017).

  68. 68.

    Zhang, B. et al. Functional DNA methylation differences between tissues, cell types, and across individuals discovered using the M&M algorithm. Genome Res 23, 1522–40 (2013).

  69. 69.

    Devlin, B. & Roeder, K. Genomic control for association studies. Biometrics 55, 997–1004 (1999).

  70. 70.

    van Iterson, M., van Zwet, E. W., Consortium, B. & Heijmans, B. T. Controlling bias and inflation in epigenome- and transcriptome-wide association studies using the empirical null distribution. Genome Biol 18, 19 (2017).

  71. 71.

    Houseman, E. A. et al. DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinformatics 13, 86 (2012).

  72. 72.

    Bakulski, K. M. et al. DNA methylation of cord blood cell types: Applications for mixed cell birth studies. Epigenetics 11, 354–62 (2016).

  73. 73.

    Cardenas, A. et al. Validation of a DNA methylation reference panel for the estimation of nucleated cells types in cord blood. Epigenetics 11, 773–779 (2016).

  74. 74.

    Houseman, E. A. et al. Reference-free deconvolution of DNA methylation data and mediation by cell composition effects. BMC Bioinformatics 17, 259 (2016).

  75. 75.

    Houseman, E. A., Molitor, J. & Marsit, C. J. Reference-free cell mixture adjustments in analysis of DNA methylation data. Bioinformatics 30, 1431–9 (2014).

  76. 76.

    Shiwa, Y. et al. Adjustment of Cell-Type Composition Minimizes Systematic Bias in Blood DNA Methylation Profiles Derived by DNA Collection Protocols. PLoS One 11, e0147519 (2016).

  77. 77.

    Yousefi, P. et al. Estimation of blood cellular heterogeneity in newborns and children for epigenome-wide association studies. Environ Mol Mutagen 56, 751–8 (2015).

  78. 78.

    Touleimat, N. & Tost, J. Complete pipeline for Infinium((R)) Human Methylation 450K BeadChip data processing using subset quantile normalization for accurate DNA methylation estimation. Epigenomics 4, 325–41 (2012).

  79. 79.

    Teschendorff, A. E. et al. A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data. Bioinformatics 29, 189–96 (2013).

  80. 80.

    Chen, Y. A. et al. Discovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray. Epigenetics 8, 203–9 (2013).

  81. 81.

    Lehne, B. et al. A coherent approach for analysis of the Illumina HumanMethylation450 BeadChip improves data quality and performance in epigenome-wide association studies. Genome Biol 16, 37 (2015).

  82. 82.

    Benjamini, Y. & Hochberg, Y. Controlling the False Discovery Rate - a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series B-Methodological 57, 289–300 (1995).

  83. 83.

    Reimand, J. et al. g:Profiler—a web server for functional interpretation of gene lists (2016 update). Nucleic Acids Research 44, W83–W89 (2016).

Download references

Acknowledgements

The research reported herein was partially supported by the Commission of the European Community, specific RTD Programme “Quality of Life and Management of Living Resources,” within the 5th Framework Programme (research grant nos. QLRT-2001-00389 and QLK1-CT-2002-30582); the 6th Framework Programme contract no. 007036 (FP6-007036); the European Union’s Seventh Framework Programme Project EarlyNutrition under grant agreement no. 289346 (FP7-289346), the Horizon 2020 research and innovation programme DYNAHEALTH (no. 633595) and the European Research Council Advanced Grant META-GROWTH (ERC-2012-AdG – no. 322605). Additional support from the German Ministry of Education and Research, Berlin (Grant Nr. 01 GI 0825) and the University of Munich Innovative Research Priority Project MC-Health is gratefully acknowledged. This manuscript does not necessarily reflect the views of the Commission and in no way anticipates the future policy in this area. The funders of this study had no role in study design, data collection, data analysis, data interpretation, decision to publish, or preparation of the manuscript. RS is funded by a NHMRC Senior Research Fellowship and NHMRC: EU project grant and the MCRI by the Victorian Operational Infrastructure support scheme. The authors thank the participating families and all project partners for their enthusiastic support of the project work. Special thanks to Prof Dr Markus Ege, Dr von Hauner Children’s Hospital, Ludwig-Maximilians Universität München (LMU), Munich and Dr Christoph Bernau from the Leibniz-Rechenzentrum (LRZ) in Garching/Munich for their great support with performing epigenetic analyses on the supercomputing facilities of the Leibniz-Rechenzentrum (LRZ). We also like to acknowledge the The European Childhood Obesity Trial Study Group for their continuous and salient support of the CHOP project: Philippe Goyens, Clotilde Carlier, Joana Hoyos, Pascale Poncelet, and Elena Dain (Universite Libre de Bruxelles – (ULB) –Brussels, Belgium); Jean-Noel Van Hees (CHC St Vincent– Françoise Martin, Annick Xhonneux, Jean-Paul Langhendries, and Jean-Noel Van Hees - Liège-Rocourt, Belgium); Ricardo Closa-Monasterolo, Joaquin Escribano, Veronica Luque, Georgina Mendez, Natalia Ferre, and Marta Zaragoza-Jordana (Universitat Rovira i Virgili, Institut d’Investigacio´ Sanitaria Pere Virgili, Taragona, Spain); Marcello Giovannini, Enrica Riva, Carlo Agostoni, Silvia Scaglioni, Elvira Verduci, Fiammetta Vecchi, and Alice Re Dionigi (University of Milano, Milano, Italy); Jerzy Socha, Piotr Socha and Anna Stolarczyk (Children’s Memorial Health Institute, Department of Gastroenterology, Hepatology and Immunology, Warsaw, Poland); Anna Dobrzanska and Dariusz Gruszfeld (Children’s Memorial Health Institute, Neonatal Intensive Care Unit, Warsaw, Poland); Roman Janas (Children’s Memorial Health Institute, Diagnostic Laboratory, Warsaw, Poland); Emmanuel Perrin (Danone Research Centre for Specialized Nutrition, Schiphol, the Netherlands); Rudiger von Kries (Division of Pediatric Epidemiology, Institute of Social Pediatrics and Adolescent Medicine, Ludwig Maximilians University of Munich, Munich, Germany); Helfried Groebe, Anna Reith, and Renate Hofmann (Klinikum Nurnberg Sued, Nurnberg, Germany); and Berthold Koletzko, Veit Grote, Martina Weber, Peter Rzehak, Sonia Schiess, Jeannette Beyer, Michaela Fritsch, Uschi Handel, Ingrid Pawellek, Sabine Verwied-Jorky, Iris Hannibal, Hans Demmelmair, Gudrun Haile, and Melissa Theurich (Division of Nutritional Medicine and Metabolism, Dr von Hauner Childrens Hospital, Ludwig-Maximilians Universität München (LMU),, Munich, Germany).

Author information

Author notes

  1. Peter Rzehak, Marcela Covic and Richard Saffery contributed equally to this work.

Affiliations

  1. Division of Metabolic and Nutritional Medicine, Dr. von Hauner Children’s Hospital, Ludwig-Maximilians Universität München (LMU), Munich, Germany

    • Peter Rzehak
    • , Marcela Covic
    • , Veit Grote
    • , Martina Weber
    •  & Berthold Koletzko
  2. Cancer and Disease Epigenetics Research Group, Murdoch Childrens Research Institute, Royal Children’s Hospital, Flemington Road, Parkville, 3052, Victoria, Australia

    • Richard Saffery
  3. Research Unit of Molecular Epidemiology, Institute of Epidemiology II, Helmholtz Zentrum Muenchen, Munich, Germany

    • Eva Reischl
    •  & Simone Wahl
  4. CHC St Vincent, Liège-Rocourt, Belgium

    • Annick Xhonneux
    •  & Jean-Paul Langhendries
  5. Universitat Rovira I Virgili, Reus, Spain

    • Natalia Ferre
    • , Ricardo Closa-Monasterolo
    •  & Joaquin Escribano
  6. University of Milano, Milano, Italy

    • Elvira Verduci
    •  & Enrica Riva
  7. Children’s Memorial Health Institute, Warsaw, Poland

    • Piotr Socha
    •  & Dariusz Gruszfeld

Authors

  1. Search for Peter Rzehak in:

  2. Search for Marcela Covic in:

  3. Search for Richard Saffery in:

  4. Search for Eva Reischl in:

  5. Search for Simone Wahl in:

  6. Search for Veit Grote in:

  7. Search for Martina Weber in:

  8. Search for Annick Xhonneux in:

  9. Search for Jean-Paul Langhendries in:

  10. Search for Natalia Ferre in:

  11. Search for Ricardo Closa-Monasterolo in:

  12. Search for Joaquin Escribano in:

  13. Search for Elvira Verduci in:

  14. Search for Enrica Riva in:

  15. Search for Piotr Socha in:

  16. Search for Dariusz Gruszfeld in:

  17. Search for Berthold Koletzko in:

Contributions

P.R., B.K. conceived this study. P.R., M.C., R.S. drafted the report. E.R. generated the Illumina 450 K array DNA-methylation data. P.R. performed pre-processing and normalization of the Illumina 450 K data. P.R., M.C., S.W. performed or contributed to statistical analysis. B.K., V.G., M.W., J.P.L., A.X., N.F., R.C.M., J.E., E.V., E.R.I., P.S., D.G. conceived or contributed to the CHOP study design and data collection. All authors (P.R., M.C., R.S., E.R., S.W., V.G., M.W., A.X., J.P.L., N.F., R.C.M., J.E., E.V., E.R.I., P.S., D.G., B.K.) critically reviewed, discussed and revised the paper.

Competing Interests

The authors declare that they have no competing interests.

Corresponding author

Correspondence to Peter Rzehak.

Electronic supplementary material

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/s41598-017-13099-4

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.