Perrier et al. Clinical Epigenetics (2019) 11:57 https://doi.org/10.1186/s13148-019-0637-x METHODOLOGY Open Access Association of leukocyte DNA methylation changes with dietary folate and alcohol intake in the EPIC study F. Perrier1, V. Viallon1, S. Ambatipudi2,3, A. Ghantous2, C. Cuenin2, H. Hernandez-Vargas2, V. Chajès4, L. Baglietto5, M. Matejcic4,6, H. Moreno-Macias7, T. Kühn8, H. Boeing9, A. Karakatsani10,11, A. Kotanidou10,12, A. Trichopoulou10, S. Sieri13, S. Panico14, F. Fasanelli15, M. Dolle16, C. Onland-Moret17, I. Sluijs17, E. Weiderpass18,19,20,21, J. R. Quirós22, A. Agudo23, J. M. Huerta24,25, E. Ardanaz24,25,26,27, M. Dorronsoro28, T. Y. N. Tong29, K. Tsilidis30, E. Riboli30, M. J. Gunter4, Z. Herceg2, P. Ferrari1*† and I. Romieu4† Abstract Background: There is increasing evidence that folate, an important component of one-carbon metabolism, modulates the epigenome. Alcohol, which can disrupt folate absorption, is also known to affect the epigenome. We investigated the association of dietary folate and alcohol intake on leukocyte DNA methylation levels in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Leukocyte genome-wide DNA methylation profiles on approximately 450,000 CpG sites were acquired with Illumina HumanMethylation 450K BeadChip measured among 450 women control participants of a case-control study on breast cancer nested within the EPIC cohort. After data preprocessing using surrogate variable analysis to reduce systematic variation, associations of DNA methylation with dietary folate and alcohol intake, assessed with dietary questionnaires, were investigated using CpG site- specific linear models. Specific regions of the methylome were explored using differentially methylated region (DMR) analysis and fused lasso (FL) regressions. The DMR analysis combined results from the feature-specific analysis for a specific chromosome and using distances between features as weights whereas FL regression combined two penalties to encourage sparsity of single features and the difference between two consecutive features. Results: After correction for multiple testing, intake of dietary folate was not associated with methylation level at any DNA methylation site, while weak associations were observed between alcohol intake and methylation level at CpG sites cg03199996 and cg07382687, with qval = 0.029 and qval = 0.048, respectively. Interestingly, the DMR analysis revealed a total of 24 and 90 regions associated with dietary folate and alcohol, respectively. For alcohol intake, 6 of the 15 most significant DMRs were identified through FL. Conclusions: Alcohol intake was associated with methylation levels at two CpG sites. Evidence from DMR and FL analyses indicated that dietary folate and alcohol intake may be associated with genomic regions with tumor suppressor activity such as the GSDMD and HOXA5 genes. These results were in line with the hypothesis that epigenetic mechanisms play a role in the association between folate and alcohol, although further studies are warranted to clarify the importance of these mechanisms in cancer. Keywords: DNA methylation, Dietary folate, Alcohol intake, DMR, Fused lasso, EPIC cohort * Correspondence: ferrarip@iarc.fr † Ferrari P and Romieu I are joint senior authors. 1 Nutritional Methodology and Biostatistics Group, International Agency for Research on Cancer (IARC), World Health Organization, 150, cours Albert Thomas, 69372 Lyon CEDEX 08, France Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Perrier et al. Clinical Epigenetics (2019) 11:57 Page 2 of 13 Introduction renal excretion of folate, and inhibiting methionine syn- DNA methylation is a crucial epigenetic mechanism in- thase, the key enzyme in the generation of the methyl volved in regulating important cellular processes, includ- donor in the OCM [32, 33]. This antagonistic effect of alco- ing gene expression, cell differentiation, genomic hol on folate could plausibly increase the need of folate in- imprinting, and preservation of chromosome stability. take. Inadequate folate levels may result in abnormal DNA DNA methylation refers to the addition of methyl synthesis due to a reduced availability of SAM [27] and dis- groups (–CH3) to the carbon-5 position of cytosine resi- rupted DNA repair and may, hence, influence cancer risk, dues in a cytosine-guanine DNA sequence (CpG) by including breast cancer [4, 60]. DNA methyltransferases. DNA methylation changes can The epidemiological evidence linking dietary folate, al- be influenced by many factors including aging [17, 19] cohol intake, and epigenome modifications is, however, and environmental exposure such as smoking [1, 24] or not well documented. Therefore, we investigated the rela- specific dietary factors [35]. Experimental evidence sug- tionships between dietary folate and alcohol intake with gests a link between B vitamins, including folate (vitamin leukocyte DNA methylation patterns in the controls from B9), and epigenetic modifications [3]. B vitamins, espe- the European Prospective Investigation into Cancer and cially folate, are essential components of one-carbon Nutrition (EPIC) study on breast cancer. We complemen- metabolism (OCM), the network of interrelated bio- ted standard regression analysis with techniques for the chemical reaction in which a one-carbon unit is received identification of relevant methylated regions. from methyl donor nutrients and transferred into bio- chemical and molecular pathways essential for DNA Methods replication and repair. Modifications in OCM can sig- Study population nificantly impact gene expression and thereby cellular EPIC is a multicenter study that recruited over 521,000 function [53]. participants, between 1992 and 2000 in 23 regional or Absorbed folate, circulating in the bloodstream, enters national centers in 10 European countries (Denmark, the OCM cycle in the liver where it is metabolized to France, Germany, Greece, Italy, The Netherlands, 5-methyltetrahydrofolate (5-methylTHF) and converted Norway, Spain, Sweden, and the UK) [43]. Among the into S-adenosylmethionine (SAM) after several successive 367,903 women recruited in EPIC, and after exclusion of transformation steps (Fig. 1). SAM is the methyl donor for 19,583 participants with prevalent cancers at recruit- numerous methylation reactions including the methyla- ment (except non-melanoma skin cancer), first malig- tion of DNA, RNA, and proteins. The potential role of nant primary BC occurred for 10,713 women during specific dietary factors including micronutrients such as follow-up between 1992 and 2010. Within a nested folate, alcohol, and soya intake, in modifying breast cancer case-control study that included 2491 invasive BC cases risk via epigenetic mechanisms, has been proposed [54], [34], a subsample of 960 women who completed dietary although evidence is still scarce and inconsistent. and lifestyle questionnaires and provided blood samples Alcohol intake affects epigenetic profiles [32]. Ethanol at recruitment (480 cases and 480 matched controls) metabolism generates toxins that may directly lead to from Germany, Greece, Italy, The Netherlands, Spain, OCM dysfunction by reducing folate absorption, increasing and the UK was selected for the DNA methylation Fig. 1 Diagram of the one-carbon metabolism pathway. MS methionine synthase, MTHFR methylenetetrahydrofolate reductase, THF tetrahydrofolate, SAH S-adenosylhomocysteine, SAM S-adenosylmethionine Perrier et al. Clinical Epigenetics (2019) 11:57 Page 3 of 13 analyses [2]. The present study included analysis of 450 Houseman’s estimation method [20] and included as controls only originally enrolled in this case-control covariates in the analysis. study on breast cancer (BC) nested within the EPIC study. Lifestyle and dietary exposures Data on dietary habits were collected at recruitment through validated center- or country-specific dietary ques- Methylation acquisition tionnaires (DQ) [43]. Northern Italy (Florence, Turin, and Genome-wide DNA methylation profiles in buffy coat Varese), UK, Germany, and The Netherlands used samples were quantified using the Illumina Infinium self-administered extensive quantitative food-frequency HumanMethylation 450K (HM450K) BeadChip assay [5] questionnaires (FFQs), whereas Southern Italy (Naples in 960 biospecimens from women included in the BC and Ragusa), Spain, and Greece’s centers used interview nested case-control study. A total of 20 biospecimens methods. Usual consumption of alcoholic beverages with replicates used to compare technical inter- and (number of glasses per day or week) per type of alcoholic intra-assay batch effects and then excluded from the beverage (wine, beer, spirits, and liquors) during the 12 main analysis together with 19 matched pairs, i.e., 38 months before the administration of dietary question- samples, where at least one of the two samples had a naires was collected at recruitment. In addition, 24-h diet- low-quality bisulfite conversion efficiency (intensity sig- ary recall (R) harmonized across EPIC countries was nal< 4000) or did not pass all of the Illumina GenomeS- collected from a random sample (n = 36,900) in each cen- tudio quality control steps, which were based on built-in ter to be used as reference measurements [50]. R measure- control probes for staining, hybridization, extension, and ments were used to improve estimation of alcohol content specificity [23]. To prevent collider bias [11], as both al- per specific alcoholic beverages using a country-specific cohol intake and folate intake and DNA methylation estimation of average of glass volume [48]. Dietary folate profiles are all potentially associated with causes of BC, intake (μg/day) was estimated using the updated EPIC among the 902 remaining samples from the original Nutrient Data Base (ENDB) [49], obtained after case-control study on BC nested within EPIC study, only harmonization from country-specific food composition ta- cancer-free women were selected for the present study. bles [7]. No specific information on the use of folate sup- For the 451 controls sample, probes with detection p plements was available. values higher than 0.05 were assigned “missing” value. After the exclusion of 14,548 cross-reactive probes [10], Statistical analyses 47,963 probes overlapping known SNPs with minor al- After exclusion of one outlier value of dietary folate lele frequency (MAF) greater than 5% in the overall (value larger than the third quartile plus 10 times the population (European ancestry) [10] and 1483 inter-quartile range of the distribution), a total of 450 low-quality probes (i.e., missing in more than 5% of the observations from controls only were retained for statis- samples), 421,583 probes were left for the statistical ana- tical analyses. lyses [2]. The association between dietary folate, alcohol intake, For each probe, β values were calculated as the ratio of and methylation levels was evaluated via (i) CpG methylated intensity over the overall intensity, defined as site-specific analysis, (ii) identification of differentially the sum of methylated and unmethylated intensities. methylated regions (DMRs) [41], and (iii) fused lasso The following preliminary adjustment steps were applied (FL) regression [57]. to β values: (i) color bias normalization using smooth quantile normalization [13], (ii) quantile normalization CpG site-specific models [6], and (iii) type I and type II bias correction using the M values expressing methylation levels at each CpG beta-mixture quantile normalization (BMIQ) [56]. Then, βvalues were linearly regressed on dietary folate (log-trans- M values, defined as M values ¼ log2 ð1−β Þ, were com- formed to reduce skewness) and alcohol intake. Models values puted [14]. Surrogate variable analysis (SVA) [30, 31] were adjusted for recruitment center, age at recruitment was used to remove systematic variation due to the pro- (year), menopausal status (pre- or post-menopause), and cessing of the biospecimens during methylation acquisi- white blood cell counts (proportions of T cells, natural tion such as batch, indicating groups of samples killer cells, B cells, and monocytes in blood). False dis- processed at the same time, and the position of the sam- covery rate (FDR) was used to control statistical tests for ples within the chip [40]. Then M values were standard- multiple testing. ized to have an identical variance of 1. For the two CpG sites that were associated with alco- The percentage of white blood cell counts, i.e., T cells hol intake, based on q values, the percentage of methyla- (CD8+T and CD4+T), natural killer (NK) cells, B cells, tion change for 1 standard deviation (SD) increase of monocytes, and granulocytes, was quantified using alcohol intake was calculated as follows: Perrier et al. Clinical Epigenetics (2019) 11:57 Page 4 of 13 Methylation values in site j were log-transformed and at least two significant sites separated by a maximal dis- regressed on alcohol intake (Ai), for each site j, and for i tance λ of 1000 base pairs. In line with [41], t statistics tk = 1, … , n, as: were obtained from regression models using an empir-   ical Bayes method to shrink the CpG site variance [51], log M ij ¼ α0 j þ α1 j Ai þ γ j T Z i as implemented in the limma package [52]. For each DMR, the minimum q value, the minimum and max- where α1j estimate the regression coefficient, Zi is a imum coefficients (in absolute value) of the sites in- vector of confounding factors related to methylation cluded in the region were presented as qDMR , βmin, DMR , levels through a vector of regression coefficients γj. The and βmax, DMR. ratio of any two log-transformed methylation values log(Mij1) and log(Mij0) with a difference of alcohol intake ^1 j σ^ alc . Therefore, the of 1 SD ( σ^ alc ) was predicted as α Fused lasso regression average percentage of methylation change for an in- Multivariate penalized regression provides an alternative crease of 1-SD in alcohol intake was estimated as: to DMRs. We implemented a fused lasso (FL) regression Mij1     [57], which is better suited than the standard lasso when ¼ exp α^1 j σ^ alc −1  100 covariates (CpGs) are naturally ordered and the objective Mij0 is to identify regions on the chromosome of differentially methylated CpG sites. FL is particularly useful when the DMR models number of features (p) is way larger than the sample size Differentially methylated region (DMRs) analyses were (n), a situation classically known as p ≫ n. identified with the DMRcate package [41]. The rationale FL is a multivariable regression method combining of this method is to use kernel smoothing to replace the two penalties: (i) the lasso penalty, which introduces t test statistics at a given CpG site by a weighted average sparsity of the parameter vector, i.e., many elements of of t test statistics across its neighboring sites on the the estimated vector are encouraged to be set to zero, same chromosome. More precisely, let pc express the and (ii) the fused penalty, which encourages sparsity of number of sites located on a given chromosome c with the difference between two consecutive components in c ∈ {1, … , 23} (the 23rd chromosome is chromosome X). the parameter vector, thus introducing smoothness of For any site k on this chromosome, with k = 1, … , pc, the parameter estimates in adjacent CpG sites [57]. term tk2 indicates the square of the t test statistics ob- To mimic the DMR analysis, a FL analysis was imple- tained in site-specific analyses. For each site j on mented where dietary folate and alcohol were, in turn, chromosome c, tj2 is replaced by the term ^t j , defined as 2 regressed on CpG methylation levels within each Ppc chromosome. The vector of methylation coefficient esti- ^t j ¼ k¼1 K jk t k 2: 2 mates β ^ obtained by fused lasso regression was defined where the terms Kjk express weights, with larger values as for sites k closer to j. Let xk express the position of site k on the chromosome, i.e., its chromosomal coordinate in base pairs, these weights are defined using a Gaussian X X 2 β ¼ arg minf i yi − M ij β j −γ T Z i kernel, as j Xpc   Xpc    2 !    þ^λ1 ω  β  þ ^ λ ν  β −β j−1 g; −x j −xk  j¼1 j j 2 j¼2 j j K jk ¼ exp 2ðλ=C Þ2 where yi indicates, in turn, alcohol and dietary folate where parameters λ and C represent the bandwidth and values for sample i = 1, … , n, Mij is the methylation the scaling factor, respectively. Here, we used λ = 1000 levels at CpG site j, βj is the associated regression coeffi- and C = 2, respectively, as recommended in [41]. cient, Zi is a vector of confounding factors, consistently Under the null hypothesis of no association between with linear regression and DMR analyses described P ^t j 2 pc K jk above, γ is the corresponding non-penalized vector of site j and alcohol (or folate), the distribution of Pn k 2 k K jk coefficients, and ωj and νj are the weights associated can be approximated by a χ distribution [41] with 2 with lasso penalty and fused penalty, respectively. Pp 2 Pp ð k c K jk Þ = k c K jk 2 degrees of freedom [45]. Accord- Following the rationale of the adaptive lasso [61] and ingly, p values were obtained for each site separately in the iterated lasso [8], the FL procedure was run for the each chromosome and q values were computed using first time with weights ωj and νj set to 1, which returned FDR correction on all the p values to control for mul- ^ , an initial estimate of β. β ^ The final estimates β ^ were 0 tiple testing. Then, DMRs were defined as regions with obtained after running a second FL procedure with Perrier et al. Clinical Epigenetics (2019) 11:57 Page 5 of 13 weights defined as ω j ¼ ^ 1 and ν j ¼ ^ ^1 , Table 1 Characteristics of the study population (n = 450) jβ0; j jþε jβ0; j −β0; j−1 jþε Mean (SD) Min-Max with ε = 10−4. Age at blood collection (years) 52 (9) 26–73 The FL procedure was implemented on a predefined Weight (kg) 66 (11) 40–103 grid of 50 × 50 = 2500 values for the pair of parameters (λ1, λ2). More precisely, the grid for λ1 consisted of 50 Height (cm) 161 (7) 143–196 λ1; max BMI (kg/m2) 26 (4) 16–43 equally spaced values (on a log scale) between 1000 and λ1, max, where λ1, max was the lowest λ1 value for which Alcohol intake (g/day) 8 (12) 0–72 FL returned a null β ^ vector for λ2=0, a situation where Blood folate level (nmol/L) 15 (10) 1–89 FL reduces to a standard lasso. For each value λ1on this Dietary folate (μg/day) 270 (106) 91–1012 grid, the grid for λ2consisted of 50 equally spaced values Cd8t (%) 7.5 (4) 0–23 λ ðλ Þ (on a log scale) between 2; 1000 max 1 and λ2, max(λ1), where Cd4t (%) 13.5 (5) 0–34 λ2, max(λ1) was the lowest λ2 value for which FL returned Natural killer (%) 6.7 (5) 0–27 a vector β^ with all components equal. The optimal pair B cells (%) 6.1 (2) 0–17 of tuning parameters (λ1, λ2) was selected as the one Monocytes (%) 5.7 (3) 0–17 minimizing the prediction error estimated by 5-fold Granulocytes (%) 60.8 (9) 27–85 cross-validation [16], whose principle can be summa- N % rized as follows. The original sample is first partitioned into 5 equally sized subsamples. One subsample is held Menopausal status as the test set while the other 4 are used as a training Pre-menopause 186 41.3 set, on which FL estimates are computed for the 2500 Post-menopause 264 58.7 values for (λ1, λ2). The prediction error is computed on Smoking status the test set, and the process is repeated 5 times, and for Never 250 55.6 each of the 2500 values of (λ1, λ2). The prediction error Former 93 20.7 is defined as the averaged prediction error on the 5 test sets. FL analysis was implemented using the FusedLasso Smoker 104 23.1 package. Missing 3 0.7 Preprocessing steps and statistical analyses were car- Physical activity index [58] ried out using the R software (https://www.r-project.org/ Inactive 99 22.0 ) and the Bioconductor packages [21], including lumi, Moderate inactive 187 41.5 wateRmelon, and sva [29] for the preprocessing steps. Moderate active 75 16.7 The nominal level of statistical significance was set to 5%. Active 78 10.7 Missing 11 2.4 Results SD standard deviation, reported for continuous variables only Study population characteristics Detailed characteristics of the 450 women included in sites (data not shown). Alcohol intake was inversely the study are shown in Table 1. The average age at blood associated with the cg07382687 CpG site (qval = collection was 52 years (range 26–73). Participants had 0.048) and positively associated with the cg03199996 an average body mass index (BMI) of 26 kg/m2 (range site (qval = 0.029) (Table 2). Both sites were located in 16–43) and were mostly post-menopausal (59%), an open sea region, i.e., a genomic region of isolated never-smokers (56%), and moderately physically inactive CpGs. cg07382687 was within the body region of (42%). The average daily intake of dietary folate was gene CREB3L2, and cg03199996 was within the body 270 μg/day (range 91–1012), and alcohol daily intake region of gene FAM65C. was 8 g/day (range 0–72). Non-alcohol consumers, de- fined as participants consuming less than 0.1 g/day of al- DMR analysis cohol at recruitment, represented 15% of the population. A total of 24 regions associated with dietary folate were Most participants were from the Italian and the German identified, which included 190 CpG sites over-repre- EPIC centers (Additional file 1: Figure S1). sented in the TSS1500 and 1st exon regions and under-represented in the body regions and regions out- CpG site-specific models side any gene regions (Fig. 2a). The 15 most significant After FDR correction, dietary folate intake was not sig- regions are described in Table 3 and the whole list pro- nificantly associated with methylation levels at any CpG vided in Additional file 2: Table S1. Among the 24 Perrier et al. Clinical Epigenetics (2019) 11:57 Page 6 of 13 Table 2 CpG site-specific model results for the significant CpG sites for alcohol intake (adjusted for recruitment center, age at recruitment, menopausal status, and level of different lymphocyte subtypes) CpG names Alcohol intake CpG characteristics β(1SD)1 qval2 % change 3 Associated genes Gene region4 Island5 Chr 1 cg03199996 0.263 0.029 9.7 FAM65C Body Open sea 20 2 cg07382687 − 0.257 0.048 − 10.3 CREB3L2 Body Open sea 7 1 Coefficients for 1 standard deviation alcohol intake (SD = 11.8) 2 False discovery rate (FDR) adjusted p values 3 Percentage of methylation change for an increase of 1 SD increase of alcohol intake 4 Gene region feature category describing the CpG position, from UCSC. TSS200 200 bases upstream of the transcriptional start site (TSS); TSS1500 1500 bases upstream of the TSS; 5′UTR within the 5′ untranslated region, between the TSS and the ATG start site; body between the ATG and stop codon irrespective of the presence of introns, exons, TSS, or promoters; 3′UTR between the stop codon and poly A signal 5 The location of the CpG relative to the CpG island. Shore 0–2 kb from island, Shelf 2–4 kb from island, N upstream (5′) of CpG island, S downstream (3′) of CpG island, open sea isolated CpGs in the genome DMRs, 54% showed an inverse association with dietary DMRs. The most significant DMR associated with alcohol folate, i.e., had a βmax, DMR < 0. The DMR most signifi- consumption was DMR.A1, 9 sites within the GSDMD cantly associated with dietary folate (qDMR = 1.3E gene, (qDMR = 4.7E−14, βmax, DMR = 0.020). −13, βmax, DMR = 0.019) was DMR.F1 in chromosome 7, Methylation levels of each CpG site located in the two including 49 CpG sites, related to HOXA5 and HOXA6 most significant DMRs for folate and alcohol, genes. DMR.F5 was associated with HOXA4, another i.e.DMR.F1, DMR.F2, DMR.A1 and DMR.A2, are pre- gene of the homeobox family, (qDMR = 5.8E−4, βmax, sented in Additional file 4: Figure S2 by tertiles of diet- DMR = − 0.016). ary folate and alcohol intake, respectively. Correlation Alcohol intake was associated with methylation levels in heatmaps of CpG sites in DMR.A1, DMR.A2, DMR.F1, 90 DMRs, including 550 CpG sites over-represented in and DMR.F2 are displayed in Additional file 5: Figure TSS200, 1st exon, and 5′ untranslated regions (5′UTR) S3, showing high levels of correlation among methyla- and under-represented in the body regions and the re- tion levels within the DMR.F2 of dietary folate and the gions outside any gene regions (Fig. 2b). The 15 most sig- DMR.A2 of alcohol. Other regions showed less correl- nificant DMRs are detailed in Table 4, and the full list is ation, including the DMR.A1 of alcohol intake. described in Additional file 3: Table S2. Alcohol intake was positively associated with methylation levels in 66% of Fused lasso regression the 90 DMRs. The two sites associated with alcohol intake For dietary folate, we identified 71 FL regions, 50 present- in the CpG site-specific analyses were not included in any ing a positive association and 21 an inverse association. Fig. 2 Repartition of gene regions (gene region feature category describing the CpG position, from UCSC. TSS200, 200 bases upstream of the transcriptional start site (TSS); TSS1500, 1500 bases upstream of the TSS; 5′UTR, within the 5′ untranslated region, between the TSS and the ATG start site; body, between the ATG and stop codon; irrespective of the presence of introns, exons, TSS, or promoters; 3′UTR, between the stop codon and poly A signal) among DMRs compare to their repartition within the Illumina 450K (the repartition of CpG sites was done among the 421,583 sites included in this study). a DMRs significant for folate. b DMRs significant for alcohol. c Illumina 450K Perrier et al. Clinical Epigenetics (2019) 11:57 Page 7 of 13 Table 3 The 15 most significant DMRs associated with dietary folate out of 24 significant DMRs (adjusted for recruitment center, age at recruitment, menopausal status, and level of different lymphocyte subtypes) DMR characteristics CpG characteristics Fused lasso Associated genes Gene regions hg19coord Sites1 qDMR2 βmin, DMR 3 βmax, DMR3 Overlap4 βFL5 F1 HOXA5,HOXA6 1stExon, 5′UTR, TSS200, chr7:27183133-27185512 49 1.3E−13 0.0002 0.019 TSS1500, 3′UTR, body F2 GDF7 Body chr2:20869434-20871401 8 1.4E−08 − 0.016 − 0.033 7/8 − 0.0029 F3 CYP1A1 TSS1500 chr15:75018731-75019376 13 2.4E−05 0.0004 0.014 F4 PRSS50 Body, 1stExon, 5′UTR, chr3:46759096-46759698 9 2.4E−04 − 0.002 − 0.020 4/4 − 0.0069 TSS200, TSS1500 F5 HOXA4 1stExon, 5′UTR, TSS200, chr7:27170241-27171154 14 5.8E−04 − 0.005 − 0.016 TSS1500 F6 SYNGAP1 Body chr6:33401192-33401542 6 1.0E−03 0.004 0.008 F7 ZNF833 TSS1500, TSS200, body chr19:11784514-11785337 13 1.1E−03 − 0.002 − 0.012 F8 LAMB2 1stExon, 5′UTR, TSS200, chr3:49170496-49170849 6 3.1E−03 − 0.008 − 0.012 TSS1500 F9 GPR19 5′UTR, 1stExon, TSS200, chr12:12848977-12849588 9 3.7E−03 0.001 0.023 6/7 0.0076 TSS1500 F10 MTMR15 TSS1500, TSS200, 5′UTR, chr15:31195612-31196075 7 4.0E−03 − 0.003 − 0.017 1stExon F11 KCNE1 5′UTR, 1stExon, TSS200, chr21:35831871-35832364 8 4.2E−03 0.007 0.019 TSS1500 F12 TNXB Body chr6:32054659-32055474 20 7.2E−03 0.0002 − 0.013 F13 TERT Body chr5:1269992-1270152 3 7.2E−03 0.008 0.011 F14 C2orf27A 5′UTR chr2:132481613-132481826 2 1.7E−02 0.010 0.031 F15 ANKRD44 Body chr2:198029141-198029332 3 2.1E−02 − 0.005 − 0.018 1 Number of sites located in DMRs significant for dietary folate 2 Minimum dietary folate q values of sites located in the DMRs (FDR correction) 3 Absolute minimum and maximum of dietary folate coefficient of sites located in the DMRs, for 1 standard deviation of log-transformed diet folate (SD = 0.346) 4 Number of sites from the FL region overlapping the DMR/number of sites in the FL region 5 Dietary folate changes for an increase of 1 standard deviation of methylation levels of sites located in the FL region Three FL regions were overlapping the 15 most sig- Graphical representations of the DMRs, the FL re- nificant DMRs (Table 3). Seven out of 8 sites from a gions, and their overlap are illustrated for each chromo- FL region within the GDF7 gene were included in the some in Additional file 6: Figure S4 for dietary folate DMR.F2 (βFL = − 0.0029). All sites from a FL region and Additional file 7: Figure S5 for alcohol intake. For associated with the PRSS50 gene were part of the dietary folate, most of FL regions were located in DMR.F4 (βFL = − 0.0069). Six out of 7 sites from the chromosome 3, chromosome 22, and chromosome X. A FL region within the GPR19 gene were within the maximum of four DMRs located in the same chromo- DMR.F9 (βFL = 0.0076). None of the 68 other FL re- some was observed for chromosomes 2 and 3. As for al- gions were overlapping any folate-related DMRs. cohol intake, DMR and FL showed overlap mostly in For alcohol consumption, we identified 133 FL re- chromosomes 6 and 22, with, respectively, 4 and 3 gions, 71 regions presenting a positive association and DMRs overlapping FL regions. 62 an inverse association. Twenty-one regions were in- cluded in alcohol-related DMRs. Among them, 9 were Discussion overlapping 6 of the 15 most significant DMRs (Table 4). In this study of women from a large prospective cohort, The situation where two close FL regions were part of we investigated the association of dietary folate and alco- the same DMR was observed 3 times in the 15 most sig- hol intake with leukocyte DNA methylation via three nificant alcohol-related DMRs. In particular, four and different approaches. The site-specific analysis aimed at three sites from two FL regions located in chromosome identifying single CpG sites independently from each 22 were included in DMR.A11, associated with genes other, whereas DMR and FL analyses aimed at identify- SMC1B and RIBC2. All the 9 sites from a FL region were ing regions of CpG sites using the inter-correlation be- included in DMR.A9 (βFL = − 0.474). tween methylation levels in close sites, thus exploiting Perrier et al. Clinical Epigenetics (2019) 11:57 Page 8 of 13 Table 4 The 15 most significant DMRs associated with alcohol out of 90 significant DMRs (adjusted for recruitment center, age at recruitment, menopausal status, and level of different lymphocyte subtypes) DMRs characteristics CpG characteristics Fused lasso Associated genes Gene regions hg19coord Sites1 qDMR 2 βmin, DMR3 βmax, DMR3 Overlap4 βFL5 A1 GSDMD TSS1500, TSS200, 5′UTR, chr8:144635260-144636462 9 4.7E−14 0.0060 0.020 1stExon A2 chr6:31650735-31651362 21 1.8E−13 0.0049 0.018 2/2, 2/2 0.390 A3 TRIM4 Body, 1stExon, 5′UTR, chr7:99516603-99517509 14 3.0E−06 − 0.0007 0.018 TSS200, TSS1500 A4 RGL3 Body chr19:11517079-11517436 5 3.3E−06 0.0041 0.020 A5 COL9A3 TSS1500 chr20:61446962-61447992 32 4.8E−06 − 0.0004 − 0.012 4/4 − 1.027 A6 ADAM32 TSS1500, TSS200, 1stExon, chr8:38964500-38965492 10 1.3E−04 0.0019 0.014 5′UTR, Body A7 C21orf56 5′UTR, 1stExon, TSS1500 chr21:47604052-47605174 8 1.5E−04 0.0191 0.032 A8 chr2:118616155-118616576 5 1.9E−04 0.0143 0.019 5/7 0.514 A9 LTB4R2, LTB4R, CIDEB Body, 1stExon, TSS1500, chr14:24780404-24780926 10 2.3E−04 − 0.0031 − 0.012 9/9 − 0.474 5′UTR, TSS200 A10 PTDSS2 Body chr11:457256-457304 3 3.0E−04 0.0044 0.011 A11 SMC1B, RIBC2 Body, TSS1500, 1stExon, chr22:45808669-45810043 16 3.0E−04 0.0009 0.019 4/4, 3/3 0.332 TSS200, 5′UTR A12 chr10:72013286-72013397 2 8.4E−04 − 0.0087 − 0.014 A13 TRAF3 Body chr14:103366987-103367858 5 1.4E−03 − 0.0044 0.013 A14 C22orf27 TSS1500, TSS200, body chr22:31317764-31318546 12 1.4E−03 0.0016 0.015 4/4, 2/2 0.641 A15 S100A13, S100A1 5′UTR, 1stExon, TSS1500, chr1:153599479-153600156 8 3.0E−03 0.0076 0.019 TSS200 1 Number of sites located in DMRs significant for alcohol 2 Minimum alcohol q values of sites located in the DMRs (FDR correction) 3 Absolute minimum and maximum of alcohol coefficient of sites located in the DMRs, for 1 standard deviation of alcohol intake (SD = 11.8) 4 Number of sites from the FL region overlapping the DMR/number of sites in the FL region, appears twice if two FL regions are included in the DMR 5 Alcohol changes for an increase of 1 standard deviation of methylation levels of sites located in the FL region or average of alcohol change if two FL regions are included in a DMR the potential of specific regions of the epigenome to associated with methylation levels in DMR.A64 related show methylation activity related to lifestyle factors. to the MLH1 gene, which is frequently mutated in her- While site-specific analysis showed a lack of associ- editary nonpolyposis colon cancer (HNPCC) [39]. A ation between dietary folate, alcohol intake, and individ- positive association between alcohol intake and methyla- ual CpG sites, DMR and FL analyses identified regions tion in the DMR.A79 was related to the TSPAN32 (tetra- of the epigenome associated with dietary folate or alco- spanin 32) gene, also known as the TSSC6 gene, which hol intake. These two sites are located within the body is one of the several tumor suppressor genes located at region of the genes FAMB65C and CREB3L2. The locus 11p15.5 in the imprinted gene domain of chromo- FAMB65C gene, also named RIPOR3, is a non-annotated some 11 [28]. This locus has been associated with adre- gene. The CREB3L2 gene encodes a transcriptional acti- nocortical carcinoma, lung, ovarian, and breast cancers. vator protein and plays a critical role in cartilage devel- Methylations within DMR.A1 were positively associated opment by activating the transcription of SEC23A [18]. with alcohol intake, and the related GSDMD gene has Translocation of CREB3L2 gene, located on chromo- also been suggested to act as a tumor suppressor [44]. some 7, and the FUS gene (fused in sarcoma) located on Alcohol intake was also positively associated with the chromosome 16 has been found in some tumors, in- DMR.A6 related to the gene ADAM32, which encodes a cluding skin cancer and soft tissue sarcoma [37, 38]. protein involved in diverse biological processes, such as Alcohol is known to alter DNA methylation, mostly brain development, fertilization, tumor development, because it contributes to deregulation of folate absorp- and inflammation [36]. tion, which can lead to a dysfunction of OCM [27]. In Several genes, associated with the 24 DMRs identified our study, alcohol intake was associated with 90 DMRs, in our study for dietary folate, were possibly involved in some of which may have a role in specific carcinogenesis biological processes leading to carcinogenesis. For ex- processes. For example, alcohol intake was inversely ample, dietary folate was positively associated with Perrier et al. Clinical Epigenetics (2019) 11:57 Page 9 of 13 methylation in DMR.F16 related to the RTKN (rhotekin) included in Joubert’s meta-analysis against 450 in our gene, which interacts with GTP-bound Rho proteins. study. Then, different methods were used to assess folate Rho proteins regulate many important cellular processes, intake, i.e., plasma folate against dietary folate. including cell growth and transformation, cytokinesis, An intervention study was conducted to evaluate the transcription, and smooth muscle contraction. Dysregu- effects of long-term supplementation with folic acid and lation of the Rho signal transduction pathway has been vitamin B12 on white blood cell DNA methylation in eld- implicated in many forms of cancer such as bladder can- erly subjects [26]. After the intervention of 2 years, 162 cer, gastric cancer, and breast cancer [9, 15]. Dietary fol- sites were significantly differentially methylated com- ate was also associated with methylation levels in pared to baseline, versus 6 sites only for the placebo DMR.F1 and DMR.F5 within the HOXA4, HOXA5, and group. Folate and vitamin B12 were not significantly as- HOXA6 genes, members of the HOX family, known to sociated with methylation level in any CpG sites. Within be associated with cellular differentiation [46]. Perturbed the same study, 173 and 425 DMRs were identified for HOX gene expression has been implicated in multiple folate and vitamin B12, respectively. The gene HOX4, cancer types [47]. In addition, HOXA5 may also regulate which was inversely associated with dietary folate in our gene expression and morphogenesis. Methylation of this study in DMR.F5, was the only region overlapping with gene may result in the loss of its expression and, since the first 10 DMRs found in the intervention study [26]. the encoded protein upregulates the tumor suppressor However, a higher level of folic acid was observed in the p53, may play an important role in tumorigenesis [55]. intervention study: averages blood folate of 52 and 23 Results from site-specific and DMR analyses were gen- nmol/L in the intervention and placebo groups, respect- erated with different analytical strategies: methylation ively, compared to an average blood folate of 15 nmol/L levels in different sites were assumed independent in the in our study which might partly explain the different former, with linear regression models fitted separately in findings. each CpG site, while in the latter, the physical proximity Within a recent meta-analysis including 9643 partici- of CpGs was exploited to identify specific regions of the pants of European ancestry, aged 42 to 76 years with 54% epigenome with similar methylation activity, under the women [32], 363 CpG sites were significantly associated assumption that neighboring CpG sites may share rele- with alcohol consumption, with 87% of these sites show- vant epigenetic information. FL analysis revealed some ing inverse associations. In our study, site cg02711608 was overlaps with DMRs, particularly for alcohol intake, part of the 363 identified sites and was also included in where 9 FL regions were observed within the 15 most DMR.A25 associated with gene SLC1A5. SLC1A5 gene en- significant DMRs. Yet, the overlap between DMR and codes a protein which is a sodium-dependent amino acid FL analyses is relatively low and their results deserve transporter [42]. The important difference in the number cautious interpretations as they have differences in ana- of significant sites between the meta-analysis and the lytical strategies. Unlike DMRs, FL does not take into ac- present study might mostly be explained by the larger count the physical distance between consecutive sites, study population size and the larger levels of alcohol in- but rather introduce smoothness of parameters esti- take observed in the meta-analysis [32]. Indeed, in the mated in adjacent mutually adjusted CpG sites. Methyla- meta-analysis, composed of 46% of men, the medians of tion levels within a chromosome were mutually adjusted alcohol intake ranged from 0 to 14 g/day in the 10 Euro- in FL regression, while in DMR analysis t test statistics pean cohorts, while with a median of 3.5 g/day, alcohol in- were based on independent associations of methylation take was quite low in our study, which included only levels with folate and alcohol. women. Lastly, cohort-specific approaches were used in The association between folate and DNA methylation the meta-analysis to remove technical variability, while the has been investigated at different stages of human life, in SVA approach was used in our study, which was shown to particular during fetal development and elderly, where produce conservative findings compared to other normal- folate is especially needed. A meta-analysis of izing techniques [40]. mother-offspring pairs estimated the association be- In our study, the sample size was relatively low (n = tween maternal plasma folate during pregnancy and 450), and women only were included. With a median DNA methylation in cord blood [25]. After FDR correc- value of 3.5 g/day, a 95th percentiles equal to 31 g/day, tion, maternal plasma folate was positively associated and a percentage of non-consumers equal to 15%, alco- with methylation level at 27 CpG sites and inversely as- hol intake displayed limited variability which potentially sociated with methylation level at 416 CpG sites. None constrained the power of the study. In addition, ques- of these sites was observed in any of the 24 DMRs re- tionnaire measurements used to assess dietary folate and lated to dietary folate in the present study. This might alcohol intake are prone to exposure misclassification, be explained by the lack of power to identify specific which likely attenuated associations between lifestyle ex- sites due to the sample size: over 2000 samples were posures and methylation levels. These elements may Perrier et al. Clinical Epigenetics (2019) 11:57 Page 10 of 13 alone explain the lack of significant associations in our not only to be associated with breast cancer but also to study. Further studies including men and women, pos- have a mutually antagonistic role in the one-carbon me- sibly with larger sample size, are needed to further inves- tabolism. In some regions identified by DMRs or FL tigate the relationship between dietary folate, alcohol analysis, mapped genes are known to act as tumor sup- intake, and DNA methylation. pressors such as the GSDMD and HOXA5 genes. These A major strength of this study was the use of ad hoc results were in line with the hypothesis that folate- and methodology for normalization of methylation data. alcohol-deregulated epigenetic mechanisms might have a Technical management of samples likely introduces sys- role in the pathogenesis of cancer. tematic technical variability in methylation measure- ments that might compromise the accuracy of the Additional files acquisition process and, if not properly taken into ac- count, could introduce bias in the estimation of the as- Additional file 1: Figure S1. Sample size by recruitment centers. sociation of interest. The population used in this study (PDF 10 kb) included European women from the UK, Germany, Italy, Additional file 2: Table S1. DMRs associated with dietary folate (log). (DOCX 19 kb) Greece, The Netherlands, and Spain, implying a diversity Additional file 3: Table S2. DMRs associated with alcohol intake. of diet and lifestyle habits. Three approaches were used (DOCX 34 kb) to evaluate the relationship between dietary folate, alco- Additional file 4: Figure S2. Graphical representation of the most 2 hol intake, and DNA methylation. The comparison be- significant DMR of dietary folate and alcohol intake. The x-axis represents tween DMR and FL analyses was particularly relevant to the position (hg 19 coordinates) of the CpGs included in the plotted DMR. Each tertile of dietary folate, alcohol intake, or their interaction is identify regions of the genome associated with dietary represented by different colors: green for T1, blue for T2, and red for T3. folate and alcohol intake. For all the CpGs included in the plotted DMR, the dashed lines are their Alcohol was classified as group 1 carcinogen in 2012 1st and 3rd quartiles of methylation levels and the points represent their median values. (PDF 33 kb) by the IARC Monograph [22] and was associated with Additional file 5: Figure S3. Correlation heatmap of methylation levels cancer of the upper aero-digestive tract, female breast, inside the two most significant DMR of folate and alcohol. (PDF 43 kb) liver, and colorectum. Dietary folate has been recently Additional file 6: Figure S4. DMRs and FL regions of folate in each inversely associated with the risk of breast cancer in chromosome. Dark blue rectangles represent DMRs and light blue FL EPIC [12], although the evidence is not conclusive [59]. regions. Overlaps between the two methods are represented by red points. Positive coefficients of the two methods are represented on the Among the DMRs identified in this study for dietary fol- top part of each graphic, and negative coefficients are on the bottom ate or alcohol intake, several regions were associated part. Positive (negative) coefficients of DMRs were set to 0.5 (− 0.5) and with genes potentially implicated in cancer development, positive (negative) coefficients of FL regions were set to 1 (− 1) to clearly differentiate DMRs from FL regions. The x-axis represents the rank of CpG such as RTKN, the HOX family of genes, and the two sites according to their position on the chromosome. (PDF 12 kb) tumor suppressor genes GSDMD and TSPAN32. Our Additional file 7: Figure S5. DMRs and FL regions of alcohol in each study provides some evidence that dietary folate and al- chromosome. Dark blue rectangles represent DMRs and light blue FL cohol intakes may be associated with carcinogenesis regions. Overlaps between the two methods are represented by red points. Positive coefficients of the two methods are represented on the through a deregulation of epigenetic mechanisms, al- top part of each graphic and negative coefficients are on the bottom though our findings need to be replicated in future part. Positive (negative) coefficients of DMRs were set to 0.5 (− 0.5), and evaluations. positive (negative) coefficients of FL regions were set to 1 (− 1) to clearly differentiate DMRs from FL regions. The x-axis represents the rank of CpG In this study, site-specific analyses served as a basis to sites according to their position on the chromosome. (PDF 58 kb) explore more complex evaluations. By addressing the high dimensionality and complexity of DNA methyla- Abbreviations tion, statistical techniques used in this work may prove BC: Breast cancer; BMI: Body mass index; DMR: Differentially methylated useful for future epigenetic studies focusing on the rela- region; EPIC: European Prospective Investigation into Cancer and Nutrition; FDR: False discovery rate; FL: Fused lasso; HM450K: Illumina Infinium tionship between lifestyle exposures, DNA methylation, HumanMethylation 450K; MAF: Minor allele frequency; NK: Natural killer; and the occurrence of disease outcomes. These tools OCM: One-carbon metabolite; SAM: S-Adenosylmethionine; SVA: Surrogate presented may be adapted to suit specific features of variable analysis other -omics data. Acknowledgements The authors would like to thank the financial support provided by La Conclusion Fondation de France for a doctoral fellowship. They are also grateful for all the women who participated in the EPIC cohort and without whom this Weak associations between alcohol intake and methyla- work would not have been possible. tion levels at two CpG sites were observed. DMR and FL analyses provided evidence that specific regions of CpG Funding sites were associated with dietary folate and alcohol in- This work was supported by a doctoral fellowship from ‘Fondation de France’ (grant number 2015 00060737) to FP and the grants from the Institut take, assuming that neighboring features share relevant National du Cancer (INCa, France, 2012-070 to IR and ZH), la Ligue nationale epigenetic information. Folate and alcohol are known contre le cancer (to Z. Herceg). ZH was supported by the European Perrier et al. Clinical Epigenetics (2019) 11:57 Page 11 of 13 Commission (EC) Seventh Framework Programme (FP7) Translational Cancer Medicine, University of Southern California/Norris Comprehensive Cancer Research (TRANSCAN) Framework, the Fondation Association pour la Center, Los Angeles, CA, USA. 7Universidad Autonoma Metropolitana, Mexico Recherche contre le Cancer (ARC, France). In addition, this study was City, Mexico. 8Division of Cancer Epidemiology, German Cancer Research supported by postdoctoral fellowship to SA from the International Agency Center (DKFZ), Heidelberg, Germany. 9Department of Epidemiology, German for Research on Cancer, partially supported by the EC FP7 Marie Curie Institute of Human Nutrition (DIfE), Potsdam-Rehbrücke, Germany. 10Hellenic Actions – People – Co-funding of regional, national and international Health Foundation, Athens, Greece. 112nd Pulmonary Medicine Department, programmes (COFUND). SA’s work is supported by Cancer Research UK School of Medicine, National and Kapodistrian University of Athens, (grant number: C18281/A19169). SA work in the Medical Research Council “ATTIKON” University Hospital, Haidari, Greece. 121st Department of Critical Integrative Epidemiology Unit at the University of Bristol which is supported Care Medicine and Pulmonary Services, University of Athens Medical School, by the Medical Research Council and the University of Bristol (grant number: Evangelismos Hospital, Athens, Greece. 13Epidemiology and Prevention Unit, MC_UU_00011/1, MC_UU_00011/4 and MC_UU_00011/5). Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 14Dipartimento The coordination of EPIC is financially supported by the European di Medicina Clinica e Chirurgia, Federico II University, Naples, Italy. 15Cancer Commission (DG-SANCO) and the International Agency for Research on Epidemiology Unit, Department of Medical Sciences, University of Turin, Via Cancer. The national cohorts are supported by German Cancer Aid, German Santena 7, Turin, Italy. 16National Institute of Public Health and the Cancer Research Center (DKFZ), Federal Ministry of Education and Research Environment (RIVM), Centre for Health Protection (pb12), Bilthoven, The (BMBF), Deutsche Krebshilfe, Deutsches Krebsforschungszentrum and Federal Netherlands. 17Department of Epidemiology, Julius Center Research Program Ministry of Education and Research (Germany); the Hellenic Health Cardiovascular Epidemiology, Utrecht, The Netherlands. 18Department of Foundation (Greece); Associazione Italiana per la Ricerca sul Cancro-AIRC-Italy Research, Cancer Registry of Norway, Institute of Population-Based Cancer and National Research Council (Italy); Dutch Ministry of Public Health, Welfare Research, Oslo, Norway. 19Department of Medical Epidemiology and and Sports (VWS), Netherlands Cancer Registry (NKR), LK Research Funds, Biostatistics, Karolinska Institutet, Stockholm, Sweden. 20Genetic Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Epidemiology Group, Folkhälsan Research Center and Faculty of Medicine, Cancer Research Fund (WCRF), Statistics Netherlands (The Netherlands); University of Helsinki, Helsinki, Finland. 21Department of Community Health Research Fund (FIS), PI13/00061 to Granada, PI13/01162 to EPIC- Medicine, University of Tromsø, The Arctic University of Norway, Tromsø, Murcia, PI13/02633 to EPIC-Navarra), Regional Governments of Andalucía, Norway. 22Public Health Directorate, Asturias, Spain. 23Unit of Nutrition and Asturias, Basque Country, Murcia and Navarra, ISCIII RETIC (RD06/0020) Cancer, Cancer Epidemiology Research Program, Catalan Institute of (Spain); Cancer Research UK (14136 to EPIC-Norfolk; C570/A16491 and Oncology-IDIBELL, L’Hospitalet de Llobregat, Barcelona, Spain. 24Department C8221/A19170 to EPIC-Oxford), Medical Research Council (1000143 to EPIC- of Epidemiology, Murcia Regional Health Council, IMIB-Arrixaca, Murcia, Norfolk, MR/M012190/1 to EPIC-Oxford) (UK). Spain. 25CIBER Epidemiology and Public Health CIBERESP, Madrid, Spain. 26 The funders of the study had no role in study design, data collection, data Navarra Public Health Institute, Pamplona, Spain. 27IdiSNA, Navarra Institute analysis, data interpretation or writing of the manuscript. for Health Research, Pamplona, Spain. 28Public Health Direction and Biodonostia Research Institute and CIBERESP, Basque Regional Health Availability of data and materials Department, San Sebastian, Spain. 29Cancer Epidemiology Unit, Nuffield For information on how to submit an application for gaining access to EPIC Department of Population Health, University of Oxford, Oxford, UK. 30 data and/or biospecimens, please follow the instructions at http://epic.iarc.fr/ Department of Epidemiology and Biostatistics, School of Public Health, access/index.php Imperial College London, London, UK. Authors’ contributions Received: 12 November 2018 Accepted: 20 February 2019 FP performed the statistical data analysis and drafted the manuscript. IR and PF developed the concept of the study with FP and contributed to draft the manuscript. SA and CC were responsible for the technical aspects of DNA methylation acquisition. IR and ZH conceived the epigenetics study in the References nested case-control study on breast cancer and critically reviewed the 1. Ambatipudi S, Cuenin C, Hernandez-Vargas H, Ghantous A, Le Calvez-Kelm manuscript. SA, AG, and HHV contributed to the interpretation of the results. F, Kaaks R, Herceg Z. Tobacco smoking-associated genome-wide DNA LB, CV, MM, and MJG were involved in the data interpretation. All authors methylation changes in the EPIC study. Epigenomics. 2016;8(5):599–618. contributed to draft the final versions of the manuscript. All authors read and https://doi.org/10.2217/epi-2016-0001. approved the final manuscript. 2. Ambatipudi S, Horvath S, Perrier F, Cuenin C, Hernandez-Vargas H, Le Calvez-Kelm F, Herceg Z. DNA methylome analysis identifies accelerated Ethics approval and consent to participate epigenetic ageing associated with postmenopausal breast cancer The study was approved by the Ethical Review Board of the International susceptibility. Eur J Cancer. 2017;75:299–307. https://doi.org/10.1016/j.ejca. Agency for Research on Cancer, and by the local Ethics Committees in the 2017.01.014 participating centres. This study was also conducted in accordance with the 3. Ba Y, Yu H, Liu F, Geng X, Zhu C, Zhu Q, Zhang Y. Relationship of folate, IARC Ethic Committee (Project No 10-22). vitamin B12 and methylation of insulin-like growth factor-II in maternal and cord blood. Eur J Clin Nutr. 2011;65(4):480–85. https://doi.org/10.1038/ejcn. Consent for publication 2010.294. Not applicable. 4. Baglietto L, English DR, Gertig DM, Hopper JL, Giles GG. Does dietary folate intake modify effect of alcohol consumption on breast cancer risk? Competing interests Prospective cohort study. Bmj. 2005;331(7520):807. https://doi.org/10.1136/ The authors declare that they have no competing interests. bmj.38551.446470.06. 5. Bibikova M, Barnes B, Tsan C, Ho V, Klotzle B, Le JM. High density DNA methylation array with single CpG site resolution. Genomics. 2011;98. Publisher’s Note https://doi.org/10.1016/j.ygeno.2011.07.007. Springer Nature remains neutral with regard to jurisdictional claims in published 6. Bolstad B. M. Probe level quantile normalization of high density maps and institutional affiliations. oligonucleotide array data. 2001. Retrieved from http://bmbolstad.com/ stuff/qnorm.pdf Author details 7. Bouckaert KP, Slimani N, Nicolas G, Vignat J, Wright AJ, Roe M, Finglas PM. 1 Nutritional Methodology and Biostatistics Group, International Agency for Critical evaluation of folate data in European and international databases: Research on Cancer (IARC), World Health Organization, 150, cours Albert recommendations for standardization in international nutritional studies. Thomas, 69372 Lyon CEDEX 08, France. 2Epigenetics Group, IARC, Lyon, Mol Nutr Food Res. 2011;55(1):166–80. https://doi.org/10.1002/mnfr. France. 3MRC Integrative Epidemiology Unit, Bristol Medical School, 201000391. University of Bristol, Bristol, UK. 4Nutritional Epidemiology Group, IARC, Lyon, 8. Candès EJ, Wakin MB, Boyd SP. Enhancing sparsity by reweighted ℓ 1 France. 5Department of Clinical and Experimental Medicine, University of minimization. J Fourier Anal Appl. 2008;14(5):877–905. https://doi.org/10. Pisa, Pisa, Italy. 6Department of Preventive Medicine, Keck School of 1007/s00041-008-9045-x. Perrier et al. Clinical Epigenetics (2019) 11:57 Page 12 of 13 9. Chen M, Bresnick AR, O'Connor KL. Coupling S100A4 to Rhotekin alters Rho experiments. Bioinformatics. 2012;28(6):882–3. https://doi.org/10.1093/ signaling output in breast cancer cells. Oncogene. 2012;32:3754. https://doi. bioinformatics/bts034. org/10.1038/onc.2012.383 https://www.nature.com/articles/ 30. Leek JT, Storey JD. Capturing heterogeneity in gene expression studies by onc2012383#supplementary-information. surrogate variable analysis. PLoS Genet. 2007;3. https://doi.org/10.1371/ 10. Chen YA, Lemire M, Choufani S, Butcher DT, Grafodatskaya D, Zanke BW, journal.pgen.0030161. Weksberg R. Discovery of cross-reactive probes and polymorphic CpGs in 31. Leek JT, Storey JD. A general framework for multiple testing dependence. the Illumina Infinium HumanMethylation450 microarray. Epigenetics. 2013; Proc Natl Acad Sci U S A. 2008;105(48):18718–23. https://doi.org/10.1073/ 8(2):203–09. https://doi.org/10.4161/epi.23470. pnas.0808709105. 11. Cole SR, Platt RW, Schisterman EF, Chu H, Westreich D, Richardson D, Poole 32. Liu C, Marioni RE, Hedman AK, Pfeiffer L, Tsai PC, Reynolds LM, Levy D. A C. Illustrating bias due to conditioning on a collider. Int J Epidemiol. 2010; DNA methylation biomarker of alcohol consumption. Mol Psychiatry. 2016. 39(2):417–20. https://doi.org/10.1093/ije/dyp334. https://doi.org/10.1038/mp.2016.192. 12. de Batlle J, Ferrari P, Chajes V, Park J. Y, Slimani N, McKenzie F, Romieu I. 33. Mason JB, Choi S-W. Effects of alcohol on folate metabolism: implications Dietary folate intake and breast cancer risk: European prospective for carcinogenesis. Alcohol. 2005;35(3):235–41. https://doi.org/10.1016/j. investigation into cancer and nutrition. J Natl Cancer Inst. 2015;107(1):367. alcohol.2005.03.012. https://doi.org/10.1093/jnci/dju367. 34. Matejcic M, de Batlle J, Ricci C, Biessy C, Perrier F, Huybrechts I, Chajes V. 13. Du P, Kibbe WA, Lin SM. Lumi: a pipeline for processing Illumina microarray. Biomarkers of folate and vitamin B12 and breast cancer risk: report from the EPIC Bioinformatics. 2008;24(13):1547–8. https://doi.org/10.1093/bioinformatics/ cohort. Int J Cancer. 2017;140(6):1246–59. https://doi.org/10.1002/ijc.30536. btn224. 35. Niculescu MD, Haggarty P. Nutrition in epigenetics: Wiley; 2011. 14. Du P, Zhang X, Huang C-C, Jafari N, Kibbe WA, Hou L, Lin SM. Comparison 36. O'Leary NA, Wright MW, Brister JR, Ciufo S, Haddad D, McVeigh R, Pruitt KD. of beta-value and M-value methods for quantifying methylation levels by Reference sequence (RefSeq) database at NCBI: current status, taxonomic microarray analysis. BMC Bioinformatics. 2010;11:587. https://doi.org/10. expansion, and functional annotation. Nucleic Acids Res. 2016;44(D1):D733– 1186/1471-2105-11-587. D745. https://doi.org/10.1093/nar/gkv1189. 15. Fan J, Ma LJ, Xia SJ, Yu L, Fu Q, Wu CQ, Tang XD. Association between 37. Panagopoulos I, Storlazzi CT, Fletcher CD, Fletcher JA, Nascimento A, clinical characteristics and expression abundance of RTKN gene in human Domanski HA, Mertens F. The chimeric FUS/CREB3l2 gene is specific for bladder carcinoma tissues from Chinese patients. J Cancer Res Clin Oncol. low-grade fibromyxoid sarcoma. Genes Chromosomes Cancer. 2004;40(3): 2005;131(3):157–62. https://doi.org/10.1007/s00432-004-0638-8. 218–28. https://doi.org/10.1002/gcc.20037. 16. Hastie T. The elements of statistical learning: data mining, inference, and 38. Patel RM, Downs-Kelly E, Dandekar MN, Fanburg-Smith JC, Billings SD, prediction; 2009. Tubbs RR, Goldblum JR. FUS (16p11) gene rearrangement as detected by 17. Heyn H, Li N, Ferreira HJ, Moran S, Pisano DG, Gomez A. Distinct DNA fluorescence in-situ hybridization in cutaneous low-grade fibromyxoid methylomes of newborns and centenarians. Proc Natl Acad Sci U S A. 2012; sarcoma: a potential diagnostic tool. Am J Dermatopathol. 2011;33(2):140–3. 109. https://doi.org/10.1073/pnas.1120658109. https://doi.org/10.1097/IAE.0b013e318176de80. 18. Hino K, Saito A, Kido M, Kanemoto S, Asada R, Takai T, Imaizumi K. Master 39. Peltomaki P, de la Chapelle A. Mutations predisposing to hereditary regulator for chondrogenesis, Sox9, regulates transcriptional activation of nonpolyposis colorectal cancer. Adv Cancer Res. 1997;71:93–119. the endoplasmic reticulum stress transducer BBF2H7/CREB3L2 in 40. Perrier F, Novoloaca A, Ambatipudi S, Baglietto L, Ghantous A, Perduca V, chondrocytes. J Biol Chem. 2014;289(20):13810–820. https://doi.org/10.1074/ Ferrari, P. Identifying and correcting epigenetics measurements for jbc.M113.543322. systematic sources of variation. Clin Epigenetics. 2018;10(1):38. https://doi. 19. Horvath S, Raj K. DNA methylation-based biomarkers and the epigenetic org/10.1186/s13148-018-0471-6. clock theory of ageing. Nat Rev Genet. 2018;19(6):371–84. https://doi.org/10. 41. Peters TJ, Buckley MJ, Statham AL, Pidsley R, Samaras K, V Lord R, Molloy PL. 1038/s41576-018-0004-3. De novo identification of differentially methylated regions in the human 20. Houseman EA, Accomando WP, Koestler DC, Christensen BC, Marsit CJ, genome. Epigenetics Chromatin. 2015;8(1):1–16. https://doi.org/10.1186/ Nelson HH, Kelsey KT. DNA methylation arrays as surrogate measures of cell 1756-8935-8-6. mixture distribution. BMC Bioinformatics. 2012;13(1):1–16. https://doi.org/10. 42. Pochini L, Scalise M, Galluccio M, Indiveri C. Membrane transporters for the 1186/1471-2105-13-86. special amino acid glutamine: structure/function relationships and relevance 21. Huber W, Carey VJ, Gentleman R, Anders S, Carlson M, Carvalho BS, Morgan to human health. Frontiers Chem. 2014;2(61). https://doi.org/10.3389/fchem. M. Orchestrating high-throughput genomic analysis with bioconductor. Nat 2014.00061. Methods. (2015);12(2):115–21. https://doi.org/10.1038/nmeth.3252. 43. Riboli E, Hunt KJ, Slimani N, Ferrari P, Norat T, Fahey M, Saracci R. European 22. IARC Working Group on the Evaluation of Carcinogenic Risks to Humans. Prospective Investigation into Cancer and Nutrition (EPIC): study (2012). Personal habits and indoor combustions. Volume 100 E. A review of populations and data collection. Public Health Nutr. 2002;5(6B):1113–24. human carcinogens (1017–1606). Retrieved from https://www.ncbi.nlm.nih. https://doi.org/10.1079/phn2002394. gov/pubmed/23193840, https://www.ncbi.nlm.nih.gov/pmc/PMC4781577/. 44. Saeki N, Usui T, Aoyagi K, Kim DH, Sato M, Mabuchi T, Sasaki H. Distinctive 23. Illumina (Producer). (2011). GenomeStudio/BeadStudio Software Methylation expression and function of four GSDM family genes (GSDMA-D) in normal Module. and malignant upper gastrointestinal epithelium. Genes Chromosomes 24. Joehanes R, Just AC, Marioni RE, Pilling LC, Reynolds LM, Mandaviya PR. Cancer. 2009;48(3):261–71. https://doi.org/10.1002/gcc.20636. London SJ. Epigenetic signatures of cigarette smoking. Circ Cardiovasc 45. Satterthwaite FE. An approximate distribution of estimates of variance Genet. 2016;9(5):436–47. https://doi.org/10.1161/circgenetics.116.001506. components. Biometrics. 1946;2. https://doi.org/10.2307/3002019. 25. Joubert BR, den Dekker HT, Felix JF, Bohlin J, Ligthart S, Beckett E, London 46. Seifert A, Werheid DF, Knapp SM, Tobiasch E. Role of Hox genes in stem cell SJ. Maternal plasma folate impacts differential DNA methylation in an differentiation. World J Stem Cells. 2015;7(3):583–95. https://doi.org/10.4252/ epigenome-wide meta-analysis of newborns. Nat Commun. (2016);7:10577. wjsc.v7.i3.583. https://doi.org/10.1038/ncomms10577. 47. Shah N, Sukumar S. The Hox genes and their roles in oncogenesis. Nat Rev 26. Kok DE, Dhonukshe-Rutten RA, Lute C, Heil SG, Uitterlinden AG, van der Velde Cancer. 2010;10(5):361–71. https://doi.org/10.1038/nrc2826. N, Steegenga WT. The effects of long-term daily folic acid and vitamin B12 48. Sieri S, Agudo A, Kesse E, Klipstein-Grobusch K, San-Jose B, Welch AA, supplementation on genome-wide DNA methylation in elderly subjects. Clin Slimani N. Patterns of alcohol consumption in 10 European countries Epigenetics. 2015;7;121. https://doi.org/10.1186/s13148-015-0154-5. participating in the European Prospective Investigation into Cancer and 27. Kruman II, Fowler AK. Impaired one carbon metabolism and DNA Nutrition (EPIC) project. Public Health Nutr. 2002;5(6b):1287–96. https://doi. methylation in alcohol toxicity. J Neurochem. 2014;129(5):770–80. https:// org/10.1079/phn2002405. doi.org/10.1111/jnc.12677. 49. Slimani N, Deharveng G, Unwin I, Southgate DA, Vignat J, Skeie G, Riboli E. The 28. Lee MP, Brandenburg S, Landes GM, Adams M, Miller G, Feinberg AP. Two EPIC nutrient database project (ENDB): a first attempt to standardize nutrient novel genes in the center of the 11p15 imprinted domain escape genomic databases across the 10 European countries participating in the EPIC study. Eur imprinting. Hum Mol Genet. 1999;8(4):683–90. J Clin Nutr. 2007a;61(9):1037–56. https://doi.org/10.1038/sj.ejcn.1602679. 29. Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for 50. Slimani N, Kaaks R, Ferrari P, Casagrande C, Clavel-Chapelon F, Lotze G, removing batch effects and other unwanted variation in high-throughput Riboli E. European Prospective Investigation into Cancer and nutrition (EPIC) Perrier et al. Clinical Epigenetics (2019) 11:57 Page 13 of 13 calibration study: rationale, design and population characteristics. Public Health Nutr. 2007b;5(6b):1125–45. https://doi.org/10.1079/PHN2002395. 51. Smyth GK. Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet and Mol Biol. 2004;3. 52. Smyth GK. Limma: linear models for microarray data. In: Gentleman R, Carey V, Dudoit S, Irizarry R, Huber W, editors. Bioinformatics and computational biology solutions using R. and bioconductor. New York: Springer; 2005. 53. Szyf M. The implications of DNA methylation for toxicology: toward toxicomethylomics, the toxicology of DNA methylation. Toxicol Sci. 2011; 120(2):235–55. https://doi.org/10.1093/toxsci/kfr024. 54. Teegarden D, Romieu I, Lelievre SA. Redefining the impact of nutrition on breast cancer incidence: is epigenetics involved? Nutr Res Rev. 2012;25(1): 68–95. https://doi.org/10.1017/s0954422411000199. 55. Teo WW, Merino VF, Cho S, Korangath P, Liang X, Wu Rc, Sukumar S. HOXA5 determines cell fate transition and impedes tumor initiation and progression in breast cancer through regulation of E-cadherin and CD24. Oncogene. 2016;35(42):5539–51. https://doi.org/10.1038/onc.2016.95. 56. Teschendorff AE, Marabita F, Lechner M, Bartlett T, Tegner J, Gomez-Cabrero D, Beck S. A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data. Bioinformatics. 2013;29(2):189–96. https://doi.org/10.1093/bioinformatics/bts680. 57. Tibshirani R, Saunders M, Rosset S, Zhu J, Knight K. Sparsity and smoothness via the fused lasso. J. R. Stat. Soc. Ser. B Stat Methodol. 2005;67(1):91–108. https://doi.org/10.1111/j.1467-9868.2005.00490.x. 58. Wareham NJ, Jakes RW, Rennie KL, Schuit J, Mitchell J, Hennings S, Day NE. Validity and repeatability of a simple index derived from the short physical activity questionnaire used in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Public Health Nutr. 2003;6(4):407–13. https://doi.org/10.1079/phn2002439. 59. World Cancer Research Fund International, & American Institue for Cancer Research. (2017). Continuous update project report: diet, nutrition, physical activity and breast cancer. Retrieved from https://www.wcrf.org/sites/ default/files/Breast-Cancer-2017-Report.pdf 60. Zhang S, Hunter DJ, Hankinson SE, Giovannucci EL, Rosner BA, Colditz GA, Willett WC. A prospective study of folate intake and the risk of breast cancer. JAMA. 1999;281(17):1632–7. 61. Zou H. The adaptive lasso and its oracle properties. J Am Stat Assoc. 2006; 101(476):1418–29. https://doi.org/10.1198/016214506000000735.