ORIGINAL RESEARCH published: 07 May 2021 doi: 10.3389/fgene.2021.665769 Distinct Associations of BMI and Fatty Acids With DNA Methylation in Fasting and Postprandial States in Men Azucena Pescador-Tapia 1† , Guillermo A. Silva-Martínez 1,2† , Nicolás Fragoso-Bargas 1 , Dalia Rodríguez-Ríos 1 , Manel Esteller 3,4,5,6 , Sebastian Moran 7 , Silvio Zaina 8* and Gertrud Lund 1* 1 Department of Genetic Engineering, CINVESTAV Irapuato Unit, Irapuato, Mexico, 2 Celaya Technological Institute, Celaya, Edited by: Mexico, 3 Josep Carreras Leukemia Research Institute (IJC), Barcelona, Spain, 4 Centro de Investigación Biomédica en Red Bruno Ramos-Molina, Cancer (CIBERONC), Madrid, Spain, 5 Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain, 6 Biomedical Research Institute Physiological Sciences Department, School of Medicine and Health Sciences, University of Barcelona (UB), Barcelona, of Murcia (IMIB), Spain Spain, 7 IDIBELL, Barcelona, Spain, 8 Department of Medical Sciences, Division of Health Sciences, Leon Campus, University of Guanajuato, Leon, Mexico Reviewed by: Neil A. Youngson, Institute of Hepatology London, We have previously shown that blood global DNA methylation (DNAm) differs Foundation for Liver Research, United Kingdom between postprandial state (PS) and fasting state (FS) and is associated with BMI Keiko Matsubara, and polyunsaturated fatty acid (PUFA) (negatively and positively, respectively) in 12 National Center for Child Health metabolically healthy adult Mexican men (AMM cohort) equally distributed among and Development (NCCHD), Japan conventional BMI classes. Here, we detailed those associations at CpG dinucleotide *Correspondence: Gertrud Lund level by exploiting the Infinium methylation EPIC array (Illumina). We sought differentially gertrud.lund@cinvestav.mx methylated CpG (dmCpG) that were (1) associated with BMI (BMI-dmCpG) and/or fatty Silvio Zaina szaina@ugto.mx acids (FA) (FA-dmCpG) in FS or PS and (2) different across FS and PS within a BMI class. † These authors have contributed BMI-dmCpG and FA-dmCpG were more numerous in FS compared to PS and largely equally to this work and share first prandial state-specific. For saturated and monounsaturated FA, dmCpG overlap was authorship higher across than within the respective saturation group. Several BMI- and FA-dmCpG Specialty section: mapped to genes involved in metabolic disease and in some cases matched published This article was submitted to experimental data sets. Notably, SETDB1 and MTHFS promoter dmCpG could explain Epigenomics and Epigenetics, the previously observed associations between global DNAm, PUFA content, and BMI a section of the journal Frontiers in Genetics in FS. Surprisingly, overlap between BMI-dmCpG and FA-dmCpG was limited and the Received: 08 February 2021 respective dmCpG were differentially distributed across functional genomic elements. Accepted: 23 March 2021 BMI-dmCpG showed the highest overlap with dmCpG of the saturated FA palmitate, Published: 07 May 2021 monounsaturated C20:1 and PUFA C20:2. Of these, selected promoter BMI-dmCpG Citation: Pescador-Tapia A, showed opposite associations with palmitate compared to C20:1 and C20:2. As for Silva-Martínez GA, Fragoso-Bargas N, the comparison between FS and PS within BMI classes, dmCpG were strikingly more Rodríguez-Ríos D, Esteller M, abundant and variably methylated in overweight relative to normoweight or obese Moran S, Zaina S and Lund G (2021) Distinct Associations of BMI and Fatty subjects (∼70–139-fold, respectively). Overweight-associated dmCpG-hosting genes Acids With DNA Methylation were significantly enriched in targets for E47, SREBP1, and RREB1 transcription in Fasting and Postprandial States in Men. Front. Genet. 12:665769. factors, which are known players in obesity and lipid homeostasis, but none doi: 10.3389/fgene.2021.665769 overlapped with BMI-dmCpG. We show for the first time that the association of BMI Frontiers in Genetics | www.frontiersin.org 1 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles and FA with methylation of disease-related genes is distinct in FS and PS and that limited overlap exists between BMI- and FA-dmCpG within and across prandial states. Our study also identifies a transcriptional regulation circuitry in overweight that might contribute to adaptation to that condition or to transition to obesity. Further work is necessary to define the pathophysiological implications of these findings. Keywords: prandial state, BMI, fatty acid, DNA methylation, epigenomics INTRODUCTION range. Thus, confounders such as genetic variants and age could be ruled out. Furthermore, AMM was homogeneous with respect The associations of specific DNA methylation (DNAm) profiles to gender, metabolic profiles, and occupational and lifestyle- with BMI, lipids, and fatty acids (FA) have been documented related factors. The main finding of the original AMM study was by observational studies, and in some instances, causality has an inverse correlation between global DNAm and BMI that was been established. In the case of BMI, Mendelian randomization conserved across prandial states. Conversely, polyunsaturated approaches have been used in epigenome-wide association FA (PUFA) were positively associated with DNAm in both studies (EWAS) to assess causality in large cohorts with complex prandial states. In particular, eicosapentaenoic acid was positively phenotypic stratification. Conflicting results were obtained, as associated with DNAm in both FS and PS, while AA only showed both BMI-to-DNAm and the opposite causality direction were a positive association with DNAm in PS (de la Rocha et al., 2016). demonstrated (Huan et al., 2019; Sun et al., 2019; Vehmeijer The present study describes CpG methylation data obtained et al., 2020). As for lipid-DNAm causality, circulating lipids cause in the AMM cohort with the Infinium methylation EPIC DNAm and not vice versa (Dekkers et al., 2016). Among FA, array (Illumina) (Moran et al., 2016). In particular, we identify arachidonic (AA), oleic (OA), elaidic (EA), and palmitic acid differentially methylated CpG (dmCpG) sites associated with (PA) elicit changes in DNAm in experimental models (Barrès BMI and/or FA, within or across FS and PS. et al., 2009; Hall et al., 2014; Flores-Sierra et al., 2016; Silva- Martínez et al., 2016; González-Becerra et al., 2019). Associations of FA with DNAm have also been uncovered in EWAS and global DNAm studies (Aslibekyan et al., 2014; de la Rocha et al., 2016). MATERIALS AND METHODS Despite the wealth of data on DNAm and metabolism, the A flow diagram of the study is shown in Supplementary Figure 1. existence of rapid—i.e., within hours—effects of the diet on DNAm in humans is little explored, particularly in the context of obesity. It is conceivable that as part of the normal metabolic Study Participants adaptation, epigenetic profiles are dynamically reset during The AMM cohort was previously described (de la Rocha fasting to postprandial conditions and that obesity may lead to a et al., 2016). Briefly, AMM consisted of 12 adult male subjects loss of resetting ability (Longo and Mattson, 2014). Western-style equally distributed across the three conventional BMI classes, obesogenic diets might further exacerbate such loss of flexibility. normal weight (N), overweight (Ow), and obese (BMI bins This hypothesis of transition of epigenetic profiles from flexible 18.5–24.9, 25–29.9, and > 30, respectively). All participants in pre-obesity to fixed in obesity is conceptually similar to the were metabolically healthy (normal blood glucose, triglycerides, model that identifies dynamic reprogramming during embryonic cholesterol, and blood pressure), lived in the city of Leon, and development as sensitive to environmental stimuli that can result were employed as janitors at the University of Guanajuato, in the acquisition of stable phenotypic traits in later life (Kwong Mexico. On one experimental day, subjects who had undergone et al., 2000). Therefore, insights into the epigenetic effects of overnight fasting and abstained from exercise for 48 h received fasting/postprandial cycles might help in defining novel strategies a standard Western-style meal (one each of McDonald’s against obesity. McChicken, medium-size fries, medium-size Coke) at 10 am, To address that issue, we previously exploited a cohort of immediately after a venous blood draw which was followed 12 adult Mexican men (AMM cohort) distributed across the by four additional draws every 2 h (postprandial state or PS). three conventional BMI classes. This study design offered several On a different day, the same subjects underwent complete unique advantages. The same subjects underwent fasting (FS) and fasting under the same blood draw scheme (fasting state or postprandial states (PS) within a 1–2-week period, and DNAm FS). PS and FS procedures were separated by 7–14 days. and individual FA content was measured within a 1–8 h time All biochemical and molecular determinations were carried out in blood obtained from the same draw. The study was Abbreviations: AA, arachidonic acid; AMM, adult Mexican men cohort; BMI, carried out in accordance with the recommendations of the body mass index; CpG, 50 -CG-30 dinucleotide; DGLA, dihomo-γ-linolenic Mexican Government guidelines (NOM-012-SSA3-2012) and acid; dmCpG, differentially methylated CpG; DNAm, DNA methylation; EA, the Ethical Committee of the Department of Medical Sciences, elaidic acid; EDA, eicosadienoic acid; FA, fatty acid; FS, fasting state; MUFA, University of Guanajuato; the latter also approved the protocol. monounsaturated fatty acid; N, normoweight subject; OA, oleic acid; Ob, obese subject; Ow, overweight subject; PA, palmitic acid; PS, postprandial state; PUFA, All subjects gave written informed consent in accordance with polyunsaturated fatty acid; SFA, saturated fatty acid. the Declaration of Helsinki. Frontiers in Genetics | www.frontiersin.org 2 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles DNA Methylation Data Gene functional enrichment was tested with the DAVID resource The five peripheral blood samples corresponding to each with FDR p < 0.05 as significance threshold2 . subject were pooled, resulting in 12 samples for each experimental day. Whole-blood DNA was profiled with the Infinium methylation EPIC array (Illumina) according to RESULTS the manufacturer’s instructions (Moran et al., 2016). Arrays were imaged using BeadArray Reader (Illumina HiSeq 2000) Prandial State-Dependent and using the standard recommended Illumina scanner setting. Independent Associations of DNAm With The IDAT files were normalized (control normalization) and background corrected using the methylation module BMI We first sought to identify dmCpG associated with BMI (??) available on GenomeStudio (v2011.1) software from (BMI-CpG). To that end, we performed a correlation analysis Illumina. Probes with the following characteristics were filtered between CpG methylation (M-values) and BMI across N, out: SNPs within CpG, probes, or single-base extensions Ow, and Ob samples, either in FS or PS. A BMI-dmCpG (list compiled by Pidsley et al., 2016); X or Y chromosome was called if average 1β > 0.10 between N and Ob and probes; probes with detection p > 0.01; and those significantly DNAm M value correlation significance of p < 0.05. Although associated with age by Pearson’s or Spearman’s test depending carbohydrates might affect DNAm, we excluded glucose from on the normality of distribution (see section “Statistics”). this and the following association analyses, as glycemia was A total of 736,741 probes passed those filters. All genomic below the 100-mg/dL normal reference limit in all subjects. coordinates refer to the GRCh37/hg19 assembly. The raw We identified a total of 107 and 41 BMI-dmCpG in FS and datasets generated for this study were deposited in the GEO PS, respectively. Of these, 15 were conserved across prandial repository1 , ID no. GSE140692. states, i.e., 37 and 14% of FS and PS BMI-dmCpG, respectively (Figure 1A and Supplementary Table 1). Notably, the 1β Fatty Acid Profiles of BMI-dmCpG conserved between FS and PS were highly Whole-blood total (i.e., free and bound) FA profiles were correlated (r = 0.996, p < 10−6 ), indicating high between- obtained previously (de la Rocha et al., 2016) and included array technical consistency. While a slight majority (57.9%) the following FA: saturated (SFA): myristic (C14:0), PA and of FS BMI-dmCpG was inversely associated (hypomethylated) stearic acid (C18:0); monounsaturated (MUFA): palmitoleic with BMI, hypermethylation was predominant (78.0%) of acid (C16:1n-9), EA, OA, and a mix of C20:1 isomers; and PS BMI-dmCpG (Figure 1A and Supplementary Table 1). PUFA: linoleic, eicosadienoic (EDA), dihomo-γ-linolenic That trend resembled the global DNAm trends previously (DGLA), AA, eicosapentaenoic (C20:5), docosatetraenoic published in FS and in N relative to Ow individuals in PS (C22:4n-6), docosapentaenoic (C22:5n-3), and docosahexaenoic (de la Rocha et al., 2016). (C22:6n-3) acids. C18:3 was determined but excluded No significantly enriched pathway or transcription factor as one or more data points could not be determined. target was identified of BMI-dmCpG-harboring genes, and none Throughout the text, the most widely known FA will of BMI-dmCpG overlapped with those reported in EWAS of be referred to by acronym, the remaining FA by short BMI (Aslibekyan et al., 2015; Demerath et al., 2015; Dhana chemical annotation. et al., 2018; Geurts et al., 2018). However, two BMI-dmCpG— cg16733226 in the CAMK1D gene body and cg21273765 in Statistics the HIF3A distal promoter—were associated with BMI in Methylation M-values were used for statistics, while the more peripheral blood in the Strong Heart Study (Crocker et al., intuitive β-values were used in data presentation (Du et al., 2020). Focusing on BMI-dmCpG (specific to or shared across 2010). Probes with one or more missing data points were prandial states) that were located within promoter regions, not considered. The statistical tests applied were dependent on we identified single or clustered dmCpG within a number of whether data fitted a normal distribution as assessed by the obesity and metabolism-related genes (Table 1). Examples of Shapiro–Wilk test, i.e., Pearson’s or Spearman’s test, respectively, the latter were the detoxification gene GSTM5 in FS and the in correlations with BMI or FA, and paired t-test or Wilcoxon transcription factor ZNF718 in PS. Notably, akin to the findings signed-rank test, respectively, in FS versus PS comparisons. of the Strong Heart Study (Crocker et al., 2020), cg21273765 Frequencies were compared by the Chi-square test. DNAm in HIF3A was positively associated with BMI. The direction of was implicitly considered as the dependent variable, based on association of promoter BMI-dmCpG was similar to the overall previous biological evidence (Barrès et al., 2009; Hall et al., 2014; trend described previously, i.e., the predominance of negative de la Rocha et al., 2016; Silva-Martínez et al., 2016). FA-values (hypomethylated) and positive (hypermethylated) associations in were percent normalized. P-values were corrected for blood FS and PS, respectively (Table 1 and Supplementary Table 1). cell composition using the minfi package in R. Nominal (not Of the 15 dmCpG that were shared between FS and PS, three genome-wide corrected) p < 0.05 was considered significant. were within promoter regions of the following genes: MTHFS, which encodes an enzyme of the folate metabolism pathway; 1 2 www.ncbi.nlm.nih.gov/geo/ david.ncifcrf.gov Frontiers in Genetics | www.frontiersin.org 3 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles FIGURE 1 | DmCpG in associations with BMI or FA. (A) Prandial state-dependent associations of DNA methylation with BMI, SFA, MUFA, and PUFA. In pie charts, white and black indicate hypomethylated (negatively associated) and hypermethylated (positively associated) BMI-dmCpG, SFA-dmCpG, MUFA-dmCpG, and PUFA-dmCpG sets, respectively. Hypomethylated dmCpG percentage is indicated in each pie chart. DmCpG counts are indicated above the respective pie chart. (B) Shared dmCpG among SFA, MUFA, and PUFA. r-values show correlations between shared dmCpG across saturation groups. TABLE 1 | Selected relevant promoter BMI-dmCpG. Gene CpG ID r, FS r, PS Function/relevance FS ANKLE2 cg26577667 0.81 SIRT2 pathway, nuclear envelope assembly (Kaufmann et al., 2016) ASAH1 cg22061769 −0.83 Ceramide metabolism, NAFLD (Presa et al., 2019) ATP13A4 cg23159337 0.70 Obesity (Biamino et al., 2016) CCNL1 cg24725175 0.66 Genetic variants associated with leptin (Kilpeläinen et al., 2016) GSTM5 cg25210835 −0.63 Detoxification, obesity, NAFLD (Stepanova et al., 2010) cg24467349 −0.63 cg12858902 −0.58 cg23719124 −0.62 cg22864244 −0.59 HIF3A cg21273765 0.87 Hypermethylated promoter in diabetes and obesity (Zhang et al., 2019; Crocker et al., 2020) HOPX cg25456368 −0.65 Cardiomyocyte differentiation (Friedman et al., 2018) cg04085076 −0.75 VEGFC cg24597353 −0.74 Obesity and insulin resistance (Tinahones et al., 2012) PS DCLK1 cg05639937 0.67 Inflammation, epithelium repair (Yi et al., 2019) ZNF718 cg20320494 0.68 Hypermethylated promoter in asthma (Wysocki et al., 2015) cg19819404 0.70 cg22831726 0.65 cg12717203 0.66 cg21149944 0.69 cg10197305 0.70 PS and FS HOPX cg00493422 −0.78 −0.78 Adipocyte differentiation (Hng et al., 2020) MTHFS cg04124281 0.69 0.79 Folate metabolism and purine synthesis (Anguera et al., 2003; Misselbeck et al., 2019) SPIRE2 cg07402062 −0.81 −0.80 Fatty acid synthase regulation (Song et al., 2020) NAFLD, nonalcoholic fatty liver disease. Frontiers in Genetics | www.frontiersin.org 4 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles the homeobox-binding HOPX; and SPIRE2, a player in actin PS, respectively. With a few exceptions (C14:0, C18:0, AA, cytoskeleton dynamics and intracellular transport. C22:4n-6), individual FA showed opposite associations with CpG methylation in FS and PS (Supplementary Figure 2). Within each saturation group, dmCpG profiles were consistent across Prandial State-Dependent Associations individual PUFA with the exception of AA but heterogeneous of DNAm With FA across individual SFA and MUFA. However, the most abundant Next, we sought associations between CpG methylation M-values individual FA in blood of each group based on percentage (OA, and FA grouped by degree of saturation, i.e., SFA, MUFA, PA, and linoleic acid, respectively) largely mirrored the overall and PUFA. Criteria for dmCpG calling were as described methylation associations of each saturation group, with the above for BMI. Ranking by total dmCpG number was exception of OA in PS (compare Figure 1A and Supplementary SFA≈PUFA > MUFA, and most dmCpG were prandial state- Figure 2). In FS, we corroborated the previously documented specific, i.e., on average only 2 and 6% were conserved across tendency of EPA and EA to elicit hypermethylation and the prandial states as proportions of FS and PS dmCpG, respectively hypomethylating effect of OA (de la Rocha et al., 2016; Flores- (Figure 1A and Supplementary Table 2). For SFA-dmCpG Sierra et al., 2016). Notably, the limited dmCpG overlap between and PUFA-dmCpG, the relative proportions of negative and the MUFA isomers OA and EA is in line with their previously positive associations with DNAm (i.e., hypomethylated and reported divergent effects on DNAm (Flores-Sierra et al., 2016). hypermethylated, respectively) were opposite between prandial Indeed, the paucity of shared dmCpG within a saturation states, whereas MUFA-dmCpG were mostly hypomethylated group compared to dmCpG shared among saturation group in either prandial state (Figure 1A). MUFA-dmCpG profiles was particularly obvious for SFA-dmCpG and MUFA-dmCpG, mirrored that of BMI-dmCpG in FS, while neither SFA, nor i.e., overlap within those saturation groups was on average MUFA, nor PUFA-dmCpG profiles mirrored PS BMI-dmCpG. 3% (Supplementary Figure 3 and Supplementary Table 3) However, SFA and PUFA profiles showed opposite tendencies but ∼4–5-fold higher across saturation groups (15 and 12%, in either prandial state. Functional analysis revealed that SFA- respectively; p < 0.01 in either case, Mann–Whitney U-test) dmCpG-harboring genes were significantly enriched for targets (Supplementary Table 4). PUFA-dmCpG followed a different of nuclear factor kappa B (NFκB), a master regulator of the trend, as overlap either within PUFA or with SFA/MUFA groups inflammatory response (FDR < 0.007) (Lawrence, 2009). was comparatively high (27 and 14%, respectively; p > 0.05). As for dmCpG overlap between FA saturation groups, 53% of No significant functional enrichment was detected of SFA-dmCpG were specific to that group. Conversely, the majority individual FA-dmCpG-harboring genes, and none overlapped of MUFA-dmCpG and PUFA-dmCpG were shared with one with loci associated with n-3 PUFA intake in EWAS (Aslibekyan and/or the other saturation group (Figure 1B). For example, 90% et al., 2014). However, several FA-dmCpG mapped within and 79% of MUFA-dmCpG and PUFA-dmCpG, respectively, promoter regions of pathobiologically relevant genes. All overlapped with other saturation groups. In particular, in either were prandial state-specific, with the exception of dmCpG in prandial state highest and lowest overlap was between SFA- DTX2 and LCLAT1, and several were shared across individual dmCpG and PUFA-dmCpG and SFA-dmCpG and MUFA- FA from different saturation groups (Table 2). Furthermore, dmCpG, respectively. Notably, for the former the direction of we uncovered some general patterns with respect to the association was negative, while the latter showed a positive direction of association across individual FA from distinct association. A total of seven and five dmCpG overlapped in all saturation groups: (1) an opposite direction between SFA PA- three saturation groups in FS and PS, respectively. Of these, two dmCpG on the one hand, and MUFA C20:1-dmCpG and were located within regulatory regions, i.e., in SETDB1 in FS individual PUFA-dmCpG on the other, and (2) a concordance and LASS3 in PS. In either gene, CpG methylation correlated between SFA PA-dmCpG and MUFA OA-dmCpG. For each negatively with SFA and MUFA but positively with PUFA, akin of these FA, the described associations mirrored their overall to the previously published associations between FA saturation dmCpG profiles (Supplementary Figure 2). The most notable and global DNAm (de la Rocha et al., 2016). prandial state-specific profile was a 10-dmCpG region in To understand whether dmCpG of a given saturation group the 50 region of LHX6, encoding a pleiotropic DNA-binding were representative of all individual FA-dmCpG within that protein that controls craniofacial development (Zhou et al., group, we probed for associations between CpG methylation 2015; Table 2). These 10 dmCpG (collectively referred to as values and individual SFA, MUFA, and PUFA (three, four, and dmLHX6) were specific for FS C20:1-dmCpG, EDA-dmCpG, eight FA, respectively). Individual FA-dmCpG count ranged and DGLA-dmCpG and overlapped with the binding sites of between 2 and 58 in FS and between 0 and 32 in PS, with the transcriptional regulator EZH2 (chromosome 9:124,987,559– EDA and C16:1 being associated with most and least dmCpG, 124,990,236), which is implicated in adipogenesis, among respectively. PA, C20:1, EDA, and EPA were associated with most other phenomena (Wang et al., 2010). Interestingly, three dmCpG within SFA, MUFA, n-6, and n-3 PUFA, respectively dmLHX6 CpG—cg04282082, cg13571460, and cg21469772— (Supplementary Table 3 and Supplementary Figure 2). As were hypermethylated, albeit only by 3–5%, in human monocytes observed for BMI and FA saturation groups, individual FA- stimulated with increasing AA doses in the 1–10-µM range dmCpG were significantly more abundant in FS than in PS (Silva-Martínez et al., 2016). The PUFA AA was associated with (average FS/PS ratio = 3.5; p = 0.009), with only a minority dmLXH6, although significance was lost after correcting for cell shared between prandial states, i.e., 5 and 12% in FS and type composition. Therefore, the associations of dmLHX6 with Frontiers in Genetics | www.frontiersin.org 5 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles TABLE 2 | Selected 50 region dmCpG associated with multiple FA. Gene, CpG ID FA, association r Relevance SFA MUFA PUFA FS PS FS PS FS PS CBLN1 Highly expressed in adipose tissue, lipid-binding activity (Ye et al., 2020) cg09792926 PA, 0.65 OA, 0.64 LA, −0.62 C20:1, −0.63 EDA, −0.73 DGLA, −0.69 CCAR1 Transcription regulation (Kim et al., 2008) cg13515095 C14:0, −0.75 EDA, −0.78 PA, 0.64 DGLA, −0.61 EPA, −0.59 C22:5n-3, −0.61 DTX2 Notch signaling (Kishi et al., 2001) cg03509467 C20:1, −063 EA, −0.66 EDA, −0.62 EDA, −0.60 C20:1, −0.62 DHA, −0.67 DGLA, −0.75 cg06619150 PA, 0.59 C20:1, −0.59 AA, −0.71 DHA, −0.64 HIVEP3 Inflammation (Hicar et al., 2001; Oukka et al., 2002) cg11769885 EPA, −0.66 cg23762517 PA, 0.59 PA, 0.58 OA, 0.60 EPA, −0.73 EDA, −0.65 cg26038582 PA, 0.63 OA, 0.65 EPA, −0.73 LCLAT1 Cardiac homeostasis (Bravo-San Pedro et al., 2017) cg15652532 C20:1, 0.70 EDA, 0.60 EDA, 0.67 DGLA, 0.69 LHX6 Craniofacial development (Zhou et al., 2015) cg00142257 C20:1, 0.64 EDA, 0.58 cg03363289 C20:1, 0.60 cg04282082 C20:1, 0.69 EDA, 0.68 cg05037505 C20:1, 0.84 EDA, 0.78 DGLA, 0.61 cg05136264 C20:1, 0.65 EDA, 0.62 cg11479503 C20:1, 0.70 EDA, 0.71 C22:4n-6, 0.66 DHA, 0.87 cg13571460 C20:1, 0.74 EDA, 0.78 C22:4n-6, 0.73 DHA, 0.85 cg13862711 C20:1, 0.67 EDA, 0.66 cg15124400 C20:1, 0.80 EDA, 0.76 C22:4n-6, 0.61 cg21469772 C20:1, 0.70 EDA, 0.74 C22:4n-6, 0.64 MIR487B Ischemia, vascular remodeling (Nossent et al., 2013; Van Der Kwast et al., 2018) cg18863119 PA, −0.71 C20:1, 0.58 EDA, 0.76 DGLA, 0.63 MIR539 Heart failure (Muthusamy et al., 2014) cg14667980 PA, −0.64 C20:1, 0.58 EDA, 0.68 DGLA, 0.62 POU2AF1 (OCAB) Brown fat tissue accumulation, insulin resistance (Carter et al., 2018) cg24049888 C20:1, 0.76 EDA, 0.68 DGLA, 0.61 SETDB1 Regulation of chromatin, lipid storage (Zheng et al., 2018) cg10589385 PA, −0.66 AA, 0.64 EPA, 0.62 C22:4n-6, 0.79 Frontiers in Genetics | www.frontiersin.org 6 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles C20:1, EDA, and DGLA are likely causal. Another notable case FA-dmCpG, 18.2%; p = 0.029) but not in FS (Supplementary are dmCpG within HIVEP3 or Schnurri-3, encoding a ligand Table 4). A list of overlapping BMI-dmCpG and FA-dmCpG of the kappaB enhancer motif that regulates the inflammatory located within promoter regions is shown in Table 3. Of note, response, which were positively associated with PA and OA the directions of association with BMI were discordant with the but negatively with selected PUFA (Hicar et al., 2001; Oukka MUFA C20:1 and the PUFA EDA and DGLA, but concordant et al., 2002; Supplementary Table 3, Table 2). Interestingly, with the SFA PA. A further notable example is six ZNF718 we previously found that HIVEP3 cg23762517 and cg26038582 dmCpG positively associated with BMI (Table 1) but negatively methylation correlated inversely with expression in THP-1 with linoleic acid (Supplementary Table 3). cells stimulated with OA in the 1–10-µM range (r = −0.61; To further compare prandial state-related features of CpG Affymetrix Human Exon 1.0 ST data and 450K data), thus methylation associations with BMI and FA, we determined the mirroring the positive association of OA with HIVEP3 promoter genomic distributions of dmCpG relative to gene compartments dmCpG observed here (Silva-Martínez et al., 2016). Additional and CpG islands. In the latter distribution, the genome is divided dmCpG in regulatory regions of pathobiologically relevant genes into CpG islands, shores (2-kb regions adjacent to either side included PA-dmCpG within the ACBD5 that is involved in long- of a CpG island), shelves (2-kb regions adjacent to shores, chain FA metabolism (Ferdinandusse et al., 2017); CD34, an on either side of a CpG island), and open sea (rest of the adipocyte progenitor marker (Hui et al., 2007); and CDH18 genome). CpG islands are mostly unmethylated regions that that hosts variants associated with the metabolic syndrome can undergo hypermethylation in aging and atherosclerosis, (Zhang et al., 2013). Importantly, the previously documented while aberrant methylation of shores has been linked to cancer associations of FA with PDK4 and HDAC4 regulatory regions in (Issa et al., 1994; Irizarry et al., 2009; Castillo-Díaz et al., AMM (de la Rocha et al., 2016) showed the same tendency as 2010). With respect to individual FA, we restricted the analysis previously observed (de la Rocha et al., 2016), albeit none of the to the aforementioned FA that had > 15 dmCpG in either corresponding dmCpG passed the calling filter. Taken together, prandial state (i.e., PA, C20:1, EDA, and DGLA). Supplementary the data indicate that associations of FA with DNAm are better Figure 4 shows a summary of dmCpG distributions and their captured by analyzing individual FA rather than FA grouped comparisons with either the 736,741 Infinium methylation EPIC by degree of saturation, in particular of SFA and MUFA. Also, array reference probes, between FS and PS, or among dmCpG experimental data sets support a causal role for FA at least in the sets. In either prandial state, BMI-dmCpG distribution relative case of LHX6 and HIVEP3 promoter dmCpG. to gene compartments or CpG islands was not significantly different from the distribution of the array reference probes. Conversely, all FA-dmCpG—with the exception of FS C20:1- Comparison Between BMI and Individual dmCpG—differed from reference probe distribution in the FA-Associated DNAm Profiles majority of cases, with stronger significance in PS relative A general comparison of the associations with CpG methylation to FS. FA-dmCpG and BMI-dmCpG distribution across gene in FS and PS identified a number of patterns. Firstly, prandial compartments was more similar in FS than in PS. Likewise, only state-associated BMI-dmCpG were ∼2–35-fold more abundant FA-dmCpG distribution differed significantly across prandial compared to individual FA-dmCpG. Shared dmCpG between FS states. Also, across FA-dmCpG, 50 -regulatory regions were and PS as percentage of PS dmCpG followed a similar trend (BMI: more frequent in PS compared to FS (p < 10−6 ), but CpG 36.6%; all FA: average 10.5%) (Figure 1A and Supplementary islands were less represented (∼1.4–1.6-fold; p < 0.01). In all Figure 2). Secondly, dmCpG tended to be shared between BMI cases, CpG islands were mostly outside promoters (gene bodies and FA more frequently in FS compared to PS. For example, and intergenic). the SFA PA-dmCpG, MUFA C20:-dmCpG, PUFA EDA-dmCpG, Next, we asked whether dmCpG were differentially distributed and DGLA-dmCpG were shared between BMI and FA on average among quartiles of β-values in N samples. The rationale for this ∼10-fold more frequently in FS compared to PS (16.0% and 1.5% analysis was that regulatory regions across the genome display of the corresponding FA-dmCpG, respectively; p = 1.9 × 10−11 , characteristic DNAm levels. In particular, partially methylated Chi-square test) (Figure 2). Of note, these particular FA were regions are prone to dynamic changes in DNAm in response to chosen based on an arbitrary threshold of > 15 dmCpG in regulatory DNA-binding proteins and undergo global changes each prandial state for statistical robustness. Also, pathobiological in cancer (Stadler et al., 2011; Berman et al., 2012). In both FS relevance has also clearly been established for some of those FA: and PS, BMI-dmCpG were most abundant in N sample upper- PA is a proinflammatory FA that induces DNA hypermethylation intermediate β-values (0.5–0.75) and did not significantly differ in cultured cells (Barrès et al., 2009; Hall et al., 2014), and between FS and PS (Supplementary Figure 5). Interestingly, 15 EDA can be converted to AA and DGLA, two precursors of out of 16 CpG island BMI-dmCpG had β > 0.25 (average: 0.57; inflammation modulators (Brenner, 1974; Huang et al., 2011). range: 0.29–0.76) in FS N samples, and the majority (11 out of 16) Thirdly, dmCpG were shared to a lesser extent between BMI underwent further hypermethylation with increasing BMI. In PS, and FA than across FA in FS (16.0% and 41.6%, respectively; that pattern was less obvious, as ∼50% of CpG island dmCpG p = 0.044, Mann–Whitney U-test) and even more markedly so were highly methylated in N samples. With respect to specific in PS (1.5% and 63.1%, respectively; p = 7.9 × 10−7 ) (Figure 2). FA, PA-dmCpG and EDA-dmCpG were significantly different This trend was also observed when all FA were considered in between FS and PS, tending to be shifted toward intermediate- PS (BMI-dmCpG and FA-dmCpG overlap, 7.5%; overlap across low β in PS (p = 1.1 × 10−5 and p = 7.8 × 10−9 , respectively). Frontiers in Genetics | www.frontiersin.org 7 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles characteristics of the overall Ow dataset or from nonbiological factors such as sample handling or microarray technical issues. Yet, the comparison between Ow and N or between Ob and N in either FS or PS yielded predictable profiles, i.e., dmCpG counts were ∼2-fold larger in the Ob-N comparison relative to Ow-N in FS and were similar in the two comparisons in PS—i.e., no excess dmCpG count was observed in Ow (Figure 3B). DmCpG were identified by the same filtering criteria as above (p < 0.05 in M value comparisons, 1β > 0.10 between Ow or Ob and N). Also, the extent of differential methylation as measured by average 1β was similar across comparisons and indeed lower in the Ow-N comparison relative to Ob-N in PS (0.11 ± 0.02 and 0.14 ± 0.06, respectively; p = 0.016). Furthermore, BMI was not more variable in Ow: SD expressed as percentage of average FIGURE 2 | Shared dmCpG between BMI and FA. Percentage of shared increased with BMI and was 1.4, 5.7, and 8.8 in N, Ow, and Ob, dmCpG between BMI and FA, or among FA in FS (solid bars) and PS (open respectively, in FS. The same trend was observed in PS: 4.6, 5.7, bars). Percentages were calculated of FA dmCpG counts in BMI-FA and 9.1, respectively. Furthermore, as a possible explanation for comparisons, or of the lowest dmCpG counts in FA-FA comparisons. FS and PS, fasting and postprandial state, respectively. Significance of differences the distinct behavior of Ow-dmCpG, we sought associations of between the percentage of dmCpG shared by BMI and FA and shared by FA the 139 Ow-dmCpG with all available metabolic (triglycerides, is shown in FS and PS (Mann–Whitney U-test). cholesterol, lipoproteins, glucose), somatometric (height, waist and hip circumference), and vascular (systolic and diastolic blood pressure) parameters. No significant association was detected in any case. We conclude that the observed difference between The Prandial State Exerts BMI FS and PS Ow is specifically imposed by the combination of Class-Specific Effects on DNAm Profiles prandial state and BMI class, although the underlying mechanism Finally, we sought differential DNAm between FS and PS. is unclear. Notably, Ow-dmCpG-hosting genes were significantly To that end, we compared CpG methylation profiles within enriched in targets of known transcription factors involved in each of the three BMI classes—i.e., N in FS against N in PS, lipid metabolism and obesity (p < 0.05, FDR; Supplementary and the corresponding comparisons for Ow and Ob. In this Table 6). No overlap of Ow-dmCpG-hosting genes with any of case, dmCpG were called when the difference between average the published studies indicated above was observed. β > 0.10 and M-value paired comparison p < 0.05. Strikingly, Ow-dmCpG showed a slight trend toward hypomethylation the comparison between Ow yielded a total of 139 dmCpG (Ow- in PS relative to FS (81 of 139 Ow-dmCpG or 58.3%). Ow- dmCpG), whereas the comparisons between N and between Ob dmCpG genomic distribution relative to gene compartments or yielded only two and one dmCpG, respectively (Figure 3A). The CpG islands was significantly skewed in favor of promoters and dmCpG list is provided in Supplementary Table 5. This large first exons (∼1.8-fold enrichment, p = 0.006; Chi-square test) difference in dmCpG counts could result either from peculiar and CpG islands (∼2.3-fold, p = 9.7 × 10−8 ), where enrichment TABLE 3 | Promoter dmCpG associated with BMI and FA. Gene, CpG ID Association r Relevance BMI PA C20:1 EDA DGLA FS PS FS PS FS PS FS PS FS PS CRISP2 Associated with atherosclerosis (Istas et al., 2017) cg01076129 −0.85 −0.76 0.62 0.80 cg14997592 0.61 0.64 cg15509430 −0.73 −0.59 0.70 cg25390787 0.68 0.62 cg26715042 −0.70 −0.83 0.76 0.69 cg12440062 −0.79 ESPNL Marker of alcohol exposure (Clark et al., 2018) cg15871206 0.61 −0.69 −0.66 cg19391697 0.60 ZNF260 Elevated in hypertension (Debrus et al., 2005) cg06920214 −0.82 −0.69 0.62 0.71 Frontiers in Genetics | www.frontiersin.org 8 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles FIGURE 3 | Overview of prandial state-induced and BMI category-associated DNA methylation changes. (A) Within each BMI class in PS relative to FS. (B) Across BMI classes. Solid and open bars, hypermethylated and hypomethylated, respectively. The number above each bar indicates the average absolute 1β. N, Ow, and Ob, normoweight, overweight, and obese, respectively. *, p < 0.05. t-test between M-values. was highest of any dmCpG set, compared to the distribution of but peaked in the middle-high methylation β quartile (0.5- the 736,741 Infinium methylation EPIC array reference probes 0.75) in the case of the hypomethylated set (Supplementary (Supplementary Figure 4). That enrichment was contributed to Figure 5). Since 79.1% of promoter/50 UTR/first exon and 82.4% by the Ow-dmCpG fraction that was hypermethylated in PS, of CpG island Ow-dmCpG had β < 0.5 in FS, we conclude which was ∼4.5-fold enriched in CpG islands (p = 8.7 × 10−67 ) that hypermethylation of the promoter and CpG island with low and ∼5-fold in promoters, while ∼3-fold and ∼21-fold baseline methylation was a distinct feature of Ow-dmCpG. No depleted in gene bodies and intergenic regions, respectively shared dmCpG among the three PS-to-FS comparisons, with BMI (p = 2.9 × 10−53 ), compared with the hypomethylated fraction. or FA, was observed. Ow-dmCpG distribution was significantly different from BMI- The above data suggest that the postprandial dynamic dmCpG and any FA-dmCpG, either in FS or PS. Distribution remodeling of DNAm in Ow might establish Ob-specific across FS β quartiles was highly skewed toward unmethylated or epigenetic profiles upon transition from overweight to obesity. very low methylation regions for hypermethylated Ow-dmCpG To assess that hypothesis, we examined the β-values of four Frontiers in Genetics | www.frontiersin.org 9 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles promoter dmCpG mapping to relevant genes and with the of dmCpG between prandial states within the same BMI class is highest absolute 1β between PS and FS Ow, across BMI classes. dramatically different among BMI classes. The data indicated a larger variability in Ow compared to N or Comparisons of associations with BMI and with FA reveal that Ob. However, no Ob-specific profile could be observed, as N and BMI-dmCpG are more abundant than FA-dmCpG, when either Ob profiles were indistinguishable (Figure 4). prandial state-specific or shared between prandial states dmCpG are considered. These observations underline the relevance of the BMI, either as causal factor or covariable, for the DISCUSSION epigenome. The paucity of shared dmCpG also indicates that associations of BMI and FA are mechanistically distinct. This We reveal for the first time two distinct associations among conclusion is reinforced by the observation that BMI-dmCpG prandial state, BMI, FA, and CpG methylation in human and FA-dmCpG map to genomic regions with distinct baseline peripheral blood. Firstly, the associations of CpG methylation methylation levels and positions relative to CpG islands or gene with BMI and FA differ in fasting and postprandial states, in terms compartments: BMI-dmCpG are not significantly enriched in of number and identity of targeted CpG, and direction of CpG any genomic context and have mostly intermediate–high baseline methylation change. Also, BMI-dmCpG and most FA-dmCpG methylation level. Conversely, PA-dmCpG and EDA-dmCpG are are more numerous in FS relative to PS. Secondly, the number in distinct baseline methylation level bins in FS and PS, and FIGURE 4 | DNA methylation trend in fasting and postprandial states across BMI classes. Open squares and solid circles, fasting and postprandial state, respectively. Gene name and EPIC array Ow-dmCpG ID are indicated above each graph. N, Ow, and Ob, normal weight, overweight and obese, respectively. Frontiers in Genetics | www.frontiersin.org 10 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles FA-dmCpG are generally differentially distributed relative to point to the association of HIF3A promoter hypermethylation CpG islands and gene compartments in FS and PS. Also, CpG with adverse metabolic outcomes such as obesity and insulin island FS BMI-dmCpG, C20:1-dmCpG, and EDA-dmCpG are resistance, in omentum of gestational diabetes mellitus patients highly methylated and undergo further hypermethylation with and in human peripheral blood cells analyzed by EWAS (Zhang increasing BMI or FA. This pattern is unusual, as CpG islands are et al., 2019; Crocker et al., 2020). Another relevant finding is mostly unmethylated or methylated at low levels throughout the differential methylation of CRISP2. The data are more difficult genome (Bird et al., 1985; Edwards et al., 2017). Corresponding to interpret in this case, as hypermethylation of the CRISP2 PS CpG island dmCpG do not show such polarized features. promoter CpG cg26715042 in peripheral blood cells is associated Our study design strongly suggests causality of the prandial with atherosclerosis, but the opposite association is observed state with respect to DNAm, possibly via interaction with here with BMI (Istas et al., 2017). It is possible that our data genetics or environmental/dietary variables. Yet, the mechanism represent compensatory responses to the establishment of obesity underlying prandial state-specific associations of DNAm with in metabolically normal subjects, whereas the literature points BMI or FA can only be speculated. BMI and FA might to phenomena occurring in metabolically heterogeneous obese differentially affect DNAm depending on the postprandial or samples. A further caveat is that previous studies were based on fasting metabolic conditions or might be inert covariates of the a variety of biological models and may not be readily comparable causal factor. Our CpG filtering method rules out diet-dependent with the approach present here. It is therefore not surprising that blood cell composition effects. Also, it is unlikely that results our findings did not replicate any of the published BMI or FA were affected by statistical artifacts due to the difference in FA EWAS, with the exception of the Strong Heart Study (Crocker pool composition between FS and PS (de la Rocha et al., 2016). et al., 2020). A likely reason for those discrepancies is the Specifically, the paucity of associations in PS could be due to specific characteristics of our cohort, particularly its homogeneity more homogeneous distributions in that group: yet, taking PA in terms of gender, age, race, occupational and environmental as example, correlation r is in the −0.85–+0.85 range in FS and exposure, and absence of any complication of obesity. Taken PS. Also, PA tends to increase with BMI and actually does so together, several considerations suggest that our findings are more markedly in PS than in FS (105.0% and 57.9% increase biologically plausible, despite the two main limitations of this in Ob compared to N, respectively). Additionally, dmCpG were study: small sample size and lack of information on dietary intake often specific to FS or PS. In fact, FS and PS shared relatively prior to sampling. few FA-associated DNAm profiles, and a number of dmCpG In addition to prandial state-specific profiles, our study switched the direction of associations depending on the nature uncovered BMI-associated DNAm profiles that overlapped of the individual FA, within FS or PS. Another consideration between FS and PS and thus are testable candidate causal loci of is that the directions of association were consistent for most later complications such as metabolic syndrome, CVD, diabetes, PUFA on the one hand, and for BMI, OA, and PA on the other, and fatty liver disease. Of particular interest was the positive in accordance with the accepted pathophysiological properties association between promoter methylation of MTHFS and BMI respectively assigned to PUFA and SFA. Notably, the position of in FS and PS. Notably, upregulation of MTHFS lowers cellular the MUFA OA is ambiguous, as is its association with disease risk folate, an essential cofactor for methylation reactions, which (Steffen et al., 2018). could potentially explain the inverse correlation between global Distinct patterns emerged also at the individual CpG level. For DNAm and BMI previously reported in FS and PS (Anguera et al., some dmCpG, causal effects of BMI or FA on DNAm are strongly 2003; de la Rocha et al., 2016). In that study, we also showed supported either by direct mechanistic data such as dmLHX6 that in both prandial states SFA and MUFA were inversely profiles and HIVEP3 expression in cultured cells stimulated with associated with global DNAm, while the opposite was true for AA or OA or by independent functional and gene expression PUFA. Interestingly, SETDB1 promoter methylation mirrored studies of promoter regions of LHX6, ASAH1, MIR487B, the above-described associations in FS. SETDB1 is an important and HIF3A as follows: (1) LHX6 controls embryonic skeletal regulator of lipid storage (Zheng et al., 2018), in addition to development, and the relevance of dmLHX6 is supported by the being the most important driver of transposon methylation known involvement of PUFA and their metabolites in craniofacial in embryonic stem cells and somatic tissues. SETDB1 is development in animal models (Greene and Garbarino, 1984; upregulated during early embryogenesis when the mammalian Kay et al., 1988). DmLHX6 are uniquely associated with three C20 genome undergoes a global hypomethylation event (Fukuda FA which are precursors or intermediates of the same metabolic and Shinkai, 2020). This might suggest that fasting-induced pathway, thus pointing to specific biological activity. (2) The SETBD1 promoter hypomethylation represents a compensatory negative association of ASAH1 promoter methylation with BMI mechanism for transposon silencing during fasting-induced is in accordance with higher ASAH1 expression in a mouse BMI-dependent global DNA hypomethylation. Furthermore, NAFLD model (Presa et al., 2019). (3) Hypomethylation of the the ceramide synthase LASS3 promoter methylation may be MIR487B promoter with increasing PA, but hypermethylation predicted to decrease global DNAm given the recent finding with increasing C20:1, EDA, and DGLA, is in agreement of the DNA hypermethylating potential of ceramides in PS with its previously documented adverse effect of MIR487B on (Castro et al., 2019). insulin signaling and upregulation in hypertension in vivo in Also, the comparison of our previous global DNAm survey a rat model, validated in human cultured cells (Nossent et al., in AMM and the present study reveals differences in association 2013). 4) Three independent studies including ours consistently directions for BMI and FA, particularly in PS. These discrepancies Frontiers in Genetics | www.frontiersin.org 11 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles suggest a markedly different behavior of repeated and non- data. Based on transcription factor-binding site enrichment repeated genomic sequences. data, E47/SREBP and RREB1 might be master regulators of Lastly, we demonstrate that the prandial state affects CpG Ow-specific transcriptional profiles. In particular, our data methylation in a BMI class-dependent fashion. Surprisingly, suggest a loop in RREB1 signaling, with reduced RREB1– dmCpG between PS and FS are vastly more abundant in the target gene interaction, consistent with the enrichment in Ow group compared to N and Ob, in addition to showing hypermethylated promoters, secondary to an increase in RREB1 relatively high methylation variability. Thus, the combination promoter methylation induced by an unidentified mechanism of Ow and prandial state generates dmCpG, echoing evidence in PS. The functional roles for RREB1 and E47 are supported that BMI causes DNAm (Sun et al., 2019). Yet, no overlap by their respective documented beneficial effects on adipocyte with BMI-dmCpG exists. In addition, Ow-dmCpG were enriched browning and adiponectin expression (Doran et al., 2008; in CpG islands and promoters with low baseline methylation Brunmeir et al., 2016). At any rate, further studies with a level that undergo hypermethylation in PS compared to FS, a larger sample size are necessary for confirming the observed Ow distinct pattern in comparison with BMI-dmCpG (see above). epigenetic dynamicity. The dynamic epigenetic profiles of Ow likely represent the initial drift during the transition to obesity, or an adaptation to overweight. A testable hypothesis is that the intrinsic DATA AVAILABILITY STATEMENT Ow epigenetic dynamicity and prandium-induced differential methylation transiently modifies the expression of critical genes, The datasets presented in this study can be found in online thus setting a transcriptional program in motion that proceeds repositories. The names of the repository/repositories and autonomously once obesity is reached. Ow-dmCpG mapped accession number(s) can be found below: https://www.ncbi.nlm. to a number of genes involved in obesity and metabolism. nih.gov/geo/, GSE140692. Notable examples were MAP4K2 or GCK, close to an SNP associated with uric acid and a candidate player in inflammation ETHICS STATEMENT (Chuang et al., 2016; Lee et al., 2018); IRX2, associated with BMI in women and expressed in an adipose tissue depot- The studies involving human participants were reviewed dependent fashion (Karastergiou et al., 2013; Ng et al., 2017)— and approved by Ethical Committee of the Department IRX2 promotes cellular differentiation in leukemia and could of Medical Sciences, University of Guanajuato. The therefore exert currently unappreciated obesity-related effects patients/participants provided their written informed consent to in blood cells, for instance by altering the leukocyte repertoire participate in this study. (Scalea et al., 2020); KLF3, a driver of diet-induced obesity in mice which could act by promoting the expression of the adipose browning factor in selected white blood cells AUTHOR CONTRIBUTIONS (Bell-Anderson et al., 2013; Knights et al., 2020); RREB1, promoting adipose tissue browning (Brunmeir et al., 2016)— AP-T and GAS-M performed most of the bioinformatics analysis. the RREB1-altered expression in peripheral blood is associated NF-B assisted in the bioinformatics analysis. DR-R performed with pediatric obesity (Plaza-Florido et al., 2020); KLHL32, most of the molecular biology work. ME and SM generated associated with BMI in a blood-based genome-wide association the microarray data. SZ participated in designing the study study (Monda et al., 2013); PLOD3 (aka LH3), involved in and supervising the experimental work and wrote the initial adiponectin production (Zhang et al., 2015)—peripheral blood manuscript version. GL designed the study, supervised the PLOD3 expression is altered in obstructive sleep apnea, a experimental work, and wrote the final version of the manuscript. condition linked to obesity (Perry et al., 2013); and HVCN1, All the authors read and approved the final manuscript version. a driver of reactive oxygen species production in B cells (Capasso et al., 2010). The most significant enrichment was of targets for E47, an activator of adiponectin expression (Doran FUNDING et al., 2008). E47 regulates hematopoiesis, and its expression in This work was generously supported by Kellogg’s Institute for blood cells could therefore determine to establish obesity-related Nutrition and Health, Mexico, APIN 2016 grant to GL and lineages (Meyer et al., 2020). Other relevant target enrichments by the Mexican National Council for Research and Technology were for RREB1, which is also encoded by an Ow-dmCpG- (CONACyT) “Ciencia Básica” grant no. A1-S-51654 to GL. AP-T hosting gene, and for SREBP1, a partner of E47 in adiponectin is a Ph.D. Fellowship recipient from CONACyT. GAS-M is a expression activation and an extensively studied regulator of FA “Cátedra CONACyT” Fellowship recipient. metabolism (Doran et al., 2008; Jump, 2011). Methylation of selected SREBP1 CpG is also associated with adiposity (Justice et al., 2020). Also, SREBP1 has been implicated in establishing SUPPLEMENTARY MATERIAL a proatherogenic lipid profile in monocytes (Bowman et al., 2020). In summary, several of the indicated genes have well- The Supplementary Material for this article can be found online characterized functions in adipocytes, but evidence for regulation at: https://www.frontiersin.org/articles/10.3389/fgene.2021. of blood cell homeostasis supports our differential methylation 665769/full#supplementary-material Frontiers in Genetics | www.frontiersin.org 12 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles REFERENCES blood: an exploration of the utility of peripheral blood as proxy tissue for brain in alcohol methylation studies. Alcohol. Clin. Exp. Res. 42, 2360–2368. Anguera, M. C., Suh, J. R., Ghandour, H., Nasrallah, I. M., Selhub, J., and Stover, doi: 10.1111/acer.13905 P. J. (2003). Methenyltetrahydrofolate synthetase regulates folate turnover and Crocker, K. C., Domingo-Relloso, A., Haack, K., Fretts, A. M., Tang, W.-Y., accumulation. J. Biol. Chem. 278, 29856–29862. doi: 10.1074/jbc.M302883200 Herreros, M., et al. (2020). DNA methylation and adiposity phenotypes: an Aslibekyan, S., Demerath, E. W., Mendelson, M., Zhi, D., Guan, W., Liang, L., epigenome-wide association study among adults in the strong heart study. Int. et al. (2015). Epigenome-wide study identifies novel methylation loci associated J. Obes. 44, 2313–2322. doi: 10.1038/s41366-020-0646-z with body mass index and waist circumference. Obesity 23, 1493–1501. doi: de la Rocha, C., Pérez-Mojica, E., León, S. Z., Cervantes-Paz, B., Tristán-Flores, 10.1002/oby.21111 F. E., Rodríguez-Ríos, D., et al. (2016). Associations between whole peripheral Aslibekyan, S., Wiener, H. W., Havel, P. J., Stanhope, K. L., O’Brien, D. M., blood fatty acids and DNA methylation in humans. Sci. Rep. 6:25867. doi: Hopkins, S. E., et al. (2014). DNA methylation patterns are associated with n-3 10.1038/srep25867 fatty acid intake in Yup’ik people. J. Nutr. 144, 425–430. doi: 10.3945/jn.113. Debrus, S., Rahbani, L., Marttila, M., Delorme, B., Paradis, P., and Nemer, M. 187203 (2005). The zinc finger-only protein Zfp260 is a novel cardiac regulator and Barrès, R., Osler, M. E., Yan, J., Rune, A., Fritz, T., Caidahl, K., et al. (2009). Non- a nuclear effector of α1-adrenergic signaling. Mol. Cell. Biol. 25, 8669–8682. CpG methylation of the PGC-1alpha promoter through DNMT3B controls doi: 10.1128/mcb.25.19.8669-8682.2005 mitochondrial density. Cell Metab. 10, 189–198. doi: 10.1016/j.cmet.2009. Dekkers, K. F., van Iterson, M., Slieker, R. C., Moed, M. H., Bonder, M. J., van 07.011 Galen, M., et al. (2016). Blood lipids influence DNA methylation in circulating Bell-Anderson, K. S., Funnell, A. P., Williams, H., Jusoh, H. M., cells. Genome Biol. 17:138. doi: 10.1186/s13059-016-1000-6 Scully, T., Lim, W. F., et al. (2013). Loss of Krüppel-Like Factor Demerath, E. W., Guan, W., Grove, M. L., Aslibekyan, S., Mendelson, M., Zhou, 3 (KLF3/BKLF) leads to upregulation of the insulin-sensitizing Y.-H., et al. (2015). Epigenome-wide Association Study (EWAS) of BMI, BMI factor adipolin (FAM132A/CTRP12/C1qdc2). Diabetes 62, 2728–2737. change, and waist circumference in african american adults identifies multiple doi: 10.2337/db12-1745 replicated loci. Hum. Mol. Genet. 24, 4464–4479. doi: 10.1093/hmg/ddv161 Berman, B. P., Weisenberger, D. J., Aman, J. F., Hinoue, T., Ramjan, Z., Liu, Dhana, K., Braun, K. V. E., Nano, J., Voortman, T., Demerath, E. W., Guan, W., Y., et al. (2012). Regions of focal DNA hypermethylation and long-range et al. (2018). An Epigenome-Wide Association Study (EWAS) of obesity-related hypomethylation in colorectal cancer coincide with nuclear laminag-associated traits. Am. J. Epidemiol. 187, 1662–1669. doi: 10.1093/aje/kwy025 domains. Nat. Genet. 44, 40–46. doi: 10.1038/ng.969 Doran, A. C., Meller, N., Cutchins, A., Deliri, H., Slayton, R. P., Oldham, S. N., Biamino, E., di Gregorio, E., Belligni, E. F., Keller, R., Riberi, E., Gandione, M., et al. et al. (2008). The helix-loop-helix factors Id3 and E47 are novel regulators of (2016). A novel 3q29 deletion associated with autism, intellectual disability, adiponectin. Circ. Res. 103, 624–634. doi: 10.1161/CIRCRESAHA.108.175893 psychiatric disorders, and obesity. Am. J. Med. Genet. Part B Neuropsychiatr. Du, P., Zhang, X., Huang, C.-C., Jafari, N., Kibbe, W. A., Hou, L., et al. (2010). Genet. 171, 290–299. doi: 10.1002/ajmg.b.32406 Comparison of Beta-value and M-value methods for quantifying methylation Bird, A., Taggart, M., Frommer, M., Miller, O. J., and Macleod, D. (1985). A fraction levels by microarray analysis. BMC Bioinform. 11:587. doi: 10.1186/1471-2105- of the mouse genome that is derived from islands of nonmethylated, CpG-rich 11-587 DNA. Cell 40, 91–99. Edwards, J. R., Yarychkivska, O., Boulard, M., and Bestor, T. H. (2017). DNA Bowman, E. R., Cameron, C. M., Richardson, B., Kulkarni, M., Gabriel, J., Cichon, methylation and DNA methyltransferases. Epigenet. Chromat. 10:23. doi: 10. M. J., et al. (2020). Macrophage maturation from blood monocytes is altered in 1186/s13072-017-0130-8 people with HIV, and is linked to serum lipid profiles and activation indices: Ferdinandusse, S., Falkenberg, K. D., Koster, J., Mooyer, P. A., Jones, R., van a model for studying atherogenic mechanisms. PLoS Pathog. 16:e1008869. doi: Roermund, C. W. T., et al. (2017). ACBD5 deficiency causes a defect in 10.1371/journal.ppat.1008869 peroxisomal very long-chain fatty acid metabolism. J. Med. Genet. 54, 330–337. Bravo-San Pedro, J. M., Kroemer, G., and Galluzzi, L. (2017). Autophagy and doi: 10.1136/jmedgenet-2016-104132 mitophagy in cardiovascular disease. Circ. Res. 120, 1812–1824. doi: 10.1161/ Flores-Sierra, J., Arredondo-Guerrero, M., Cervantes-Paz, B., Rodríguez-Ríos, D., CIRCRESAHA.117.311082 Alvarado-Caudillo, Y., Nielsen, F. C., et al. (2016). The trans fatty acid elaidate Brenner, R. R. (1974). The oxidative desaturation of unsaturated fatty acids in affects the global DNA methylation profile of cultured cells and in vivo. Lipids animals. Mol. Cell. Biochem. 3, 41–52. doi: 10.1007/BF01660076 Health Dis. 15:75. doi: 10.1186/s12944-016-0243-2 Brunmeir, R., Wu, J., Peng, X., Kim, S. Y., Julien, S. G., Zhang, Q., et al. (2016). Friedman, C. E., Nguyen, Q., Lukowski, S. W., Helfer, A., Chiu, H. S., Miklas, J., Comparative transcriptomic and epigenomic analyses reveal new regulators of et al. (2018). Single-cell transcriptomic analysis of cardiac differentiation from murine brown Adipogenesis. PLoS Genet. 12:e1006474. doi: 10.1371/journal. human PSCs reveals HOPX-dependent cardiomyocyte maturation. Cell Stem pgen.1006474 Cell 23, 586–598.e8. doi: 10.1016/j.stem.2018.09.009 Capasso, M., Bhamrah, M. K., Henley, T., Boyd, R. S., Langlais, C., Cain, K., Fukuda, K., and Shinkai, Y. (2020). SETDB1-mediated silencing of retroelements. et al. (2010). HVCN1 modulates BCR signal strength via regulation of BCR- Viruses 12:596. doi: 10.3390/v12060596 dependent generation of reactive oxygen species. Nat. Immunol. 11, 265–272. Geurts, Y. M., Dugué, P.-A., Joo, J. E., Makalic, E., Jung, C.-H., Guan, W., et al. doi: 10.1038/ni.1843 (2018). Novel associations between blood DNA methylation and body mass Carter, S., Miard, S., Caron, A., Sallé-Lefort, S., St-Pierre, P., Anhê, F. F., et al. index in middle-aged and older adults. Int. J. Obes. 42, 887–896. doi: 10.1038/ (2018). Loss of OcaB prevents age-induced fat accretion and insulin resistance ijo.2017.269 by altering B-lymphocyte transition and promoting energy expenditure. González-Becerra, K., Ramos-Lopez, O., Barrón-Cabrera, E., Riezu-Boj, J. I., Diabetes 67, 1285–1296. doi: 10.2337/db17-0558 Milagro, F. I., Martínez-López, E., et al. (2019). Fatty acids, epigenetic Castillo-Díaz, S. A., Garay-Sevilla, M. E., Hernández-González, M. A., Solís- mechanisms and chronic diseases: a systematic review. Lipids Health Dis. Martínez, M. O., and Zaina, S. (2010). Extensive demethylation of normally 18:178. doi: 10.1186/s12944-019-1120-6 hypermethylated CpG islands occurs in human atherosclerotic arteries. Int. J. Greene, R. M., and Garbarino, M. P. (1984). Role of cyclic AMP, prostaglandins, Mol. Med. 26, 691–700. doi: 10.3892/ijmm-00000515 and catecholamines during normal palate development. Curr. Top. Dev. Biol. Castro, K., Ntranos, A., Amatruda, M., Petracca, M., Kosa, P., Chen, E. Y., et al. 19, 65–79. doi: 10.1016/s0070-2153(08)60395-6 (2019). Body mass index in multiple sclerosis modulates ceramide-induced Hall, E., Volkov, P., Dayeh, T., Bacos, K., Rönn, T., Nitert, M. D., et al. (2014). DNA methylation and disease course. eBio Med. 43, 392–410. doi: 10.1016/j. Effects of palmitate on genome-wide mRNA expression and DNA methylation ebiom.2019.03.087 patterns in human pancreatic islets. BMC Med. 12:103. doi: 10.1186/1741-7015- Chuang, H. C., Wang, X., and Tan, T. H. (2016). MAP4K family kinases in 12-103 immunity and inflammation. Adv. Immunol. 129, 277–314. doi: 10.1016/bs.ai. Hicar, M. D., Liu, Y., Allen, C. E., and Wu, L. C. (2001). Structure of the human zinc 2015.09.006 finger protein HIVEP3: molecular cloning, expression, exon-intron structure, Clark, S. L., Costin, B. N., Chan, R. F., Johnson, A. W., Xie, L., Jurmain, J. L., and comparison with paralogous genes HIVEP1 and HIVEP2. Genomics 71, et al. (2018). A whole Methylome study of ethanol exposure in brain and 89–100. doi: 10.1006/geno.2000.6425 Frontiers in Genetics | www.frontiersin.org 13 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles Hng, C. H., Camp, E., Anderson, P., Breen, J., Zannettino, A., and Gronthos, S. Longo, V. D., and Mattson, M. P. (2014). Fasting: molecular mechanisms and (2020). HOPX regulates bone marrow-derived mesenchymal stromal cell fate clinical applications. Cell Metab. 19, 181–192. doi: 10.1016/j.cmet.2013.12.008 determination via suppression of adipogenic gene pathways. Sci. Rep. 10:11345. Meyer, A., Herkt, S., Kunze-Schumacher, H., Kohrs, N., Ringleb, J., Schneider, L., doi: 10.1038/s41598-020-68261-2 et al. (2020). The transcription factor TAL1 and miR-17-92 create a regulatory Huan, T., Joehanes, R., Song, C., Peng, F., Guo, Y., Mendelson, M., et al. loop in hematopoiesis. Sci. Rep. 10:21438. doi: 10.1038/s41598-020-78629-z (2019). Genome-wide identification of DNA methylation QTLs in whole blood Misselbeck, K., Marchetti, L., Priami, C., Stover, P. J., and Field, M. S. (2019). The 5- highlights pathways for cardiovascular disease. Nat. Commun. 10:4267. doi: formyltetrahydrofolate futile cycle reduces pathway stochasticity in an extended 10.1038/s41467-019-12228-z hybrid-stochastic model of folate-mediated one-carbon metabolism. Sci. Rep. Huang, Y. S., Huang, W. C., Li, C. W., and Te Chuang, L. (2011). Eicosadienoic acid 9:4322. doi: 10.1038/s41598-019-40230-4 differentially modulates production of pro-inflammatory modulators in murine Monda, K. L., Chen, G. K., Taylor, K. C., Palmer, C., Edwards, T. L., Lange, macrophages. Mol. Cell. Biochem. 358, 85–94. doi: 10.1007/s11010-011-0924-0 L. A., et al. (2013). A meta-analysis identifies new loci associated with body Hui, R. Z., Kim, E. K., Kim, H., and Claycombe, K. J. (2007). Obesity-associated mass index in individuals of African ancestry. Nat. Genet. 45, 690–696. doi: mouse adipose stem cell secretion of monocyte chemotactic protein-1. Am. 10.1038/ng.2608 J. Physiol. Endocrinol. Metab. 293, E1153–E1158. doi: 10.1152/ajpendo.00186. Moran, S., Arribas, C., and Esteller, M. (2016). Validation of a DNA methylation 2007 microarray for 850,000 CpG sites of the human genome enriched in enhancer Irizarry, R. A., Ladd-Acosta, C., Wen, B., Wu, Z., Montano, C., Onyango, P., sequences. Epigenomics 8, 389–399. doi: 10.2217/epi.15.114 et al. (2009). The human colon cancer methylome shows similar hypo- and Muthusamy, S., DeMartino, A. M., Watson, L. J., Brittian, K. R., Zafir, A., hypermethylation at conserved tissue-specific CpG island shores. Nat. Genet. Dassanayaka, S., et al. (2014). MicroRNA-539 is up-regulated in failing heart, 41, 178–186. doi: 10.1038/ng.298 and suppresses O-GlcNAcase expression. J. Biol. Chem. 289, 29665–29676. doi: Issa, J. P. J., Ottaviano, Y. L., Celano, P., Hamilton, S. R., Davidson, N. E., and 10.1074/jbc.M114.578682 Baylin, S. B. (1994). Methylation of the oestrogen receptor CpG island links Ng, M. C. Y., Graff, M., Lu, Y., Justice, A. E., Mudgal, P., Liu, C. T., et al. (2017). ageing and neoplasia in human colon. Nat. Genet. 7, 536–540. doi: 10.1038/ Discovery and fine-mapping of adiposity loci using high density imputation ng0894-536 of genome-wide association studies in individuals of African ancestry: African Istas, G., Declerck, K., Pudenz, M., Szic, K. S., Vel Lendinez-Tortajada, V., Leon- ancestry anthropometry genetics consortium. PLoS Genet. 13:e1006719. doi: Latre, M., et al. (2017). Identification of differentially methylated BRCA1 and 10.1371/journal.pgen.1006719 CRISP2 DNA regions as blood surrogate markers for cardiovascular disease. Nossent, A. Y., Eskildsen, T. V., Andersen, L. B., Bie, P., Brønnum, H., Schneider, Sci. Rep. 7:5120. doi: 10.1038/s41598-017-03434-0 M., et al. (2013). The 14q32 MicroRNA-487b targets the antiapoptotic insulin Jump, D. B. (2011). Fatty acid regulation of hepatic lipid metabolism. Curr. Opin. receptor substrate 1 in hypertension-induced remodeling of the aorta. Ann. Clin. Nutr. Metab. Care 14, 115–120. doi: 10.1097/MCO.0b013e328342991c Surg. 258, 743–753. doi: 10.1097/SLA.0b013e3182a6aac0 Justice, A. E., Chittoor, G., Gondalia, R., Melton, P. E., Lim, E., Grove, M. L., Oukka, M., Kim, S. T., Lugo, G., Sun, J., Wu, L. C., and Glimcher, L. H. (2002). et al. (2020). Methylome-wide association study of central adiposity implicates A mammalian homolog of Drosophila schnurri, KRC, regulates TNF receptor- genes involved in immune and endocrine systems. Epigenomics 12, 1483–1499. driven responses and interacts with TRAF2. Mol. Cell 9, 121–131. doi: 10.1016/ doi: 10.2217/epi-2019-0276 S1097-2765(01)00434-8 Karastergiou, K., Fried, S. K., Xie, H., Lee, M. J., Divoux, A., Rosencrantz, M. A., Perry, J. C., Guindalini, C., Bittencourt, L., Garbuio, S., Mazzotti, D. R., and Tufik, et al. (2013). Distinct developmental signatures of human abdominal and S. (2013). Whole blood hypoxia-related gene expression reveals novel pathways gluteal subcutaneous adipose tissue depots. J. Clin. Endocrinol. Metab. 98, to obstructive sleep apnea in humans. Respir. Physiol. Neurobiol. 189, 649–654. 362–371. doi: 10.1210/jc.2012-2953 doi: 10.1016/j.resp.2013.08.012 Kaufmann, T., Kukolj, E., Brachner, A., Beltzung, E., Bruno, M., Kostrhon, S., Pidsley, R., Zotenko, E., Peters, T. J., Lawrence, M. G., Risbridger, G. P., Molloy, et al. (2016). SIRT2 regulates nuclear envelope reassembly through ANKLE2 P., et al. (2016). Critical evaluation of the Illumina MethylationEPIC BeadChip deacetylation. J. Cell Sci. 129, 4607–4621. doi: 10.1242/jcs.192633 microarray for whole-genome DNA methylation profiling. Genome Biol. 17:208. Kay, E. D., Goldman, A. S., and Daniel, J. C. (1988). Arachidonic acid reversal doi: 10.1186/s13059-016-1066-1 of phenytoin-induced neural tube and craniofacial defects in vitro in mice. Plaza-Florido, A., Altmäe, S., Esteban, F. J., Cadenas-Sanchez, C., Aguilera, C. M., J. Craniofac. Genet. Dev. Biol. 8, 179–186. Einarsdottir, E., et al. (2020). Distinct whole-blood transcriptome profile of Kilpeläinen, T. O., Carli, J. F. M., Skowronski, A. A., Sun, Q., Kriebel, J., Feitosa, children with metabolic healthy overweight/obesity compared to metabolic M. F., et al. (2016). Genome-wide meta-analysis uncovers novel loci influencing unhealthy overweight/obesity. Pediatr. Res. doi: 10.1038/s41390-020-01276-7 circulating leptin levels. Nat. Commun. 7:10494. doi: 10.1038/ncomms10494 Presa, N., Clugston, R. D., Lingrell, S., Kelly, S. E., Merrill, A. H., Jana, Kim, J. H., Yang, C. K., Heo, K., Roeder, R. G., An, W., and Stallcup, M. R. (2008). S., et al. (2019). Vitamin E alleviates non-alcoholic fatty liver disease CCAR1, a key regulator of mediator complex recruitment to nuclear receptor in phosphatidylethanolamine N-methyltransferase deficient mice. Biochim. transcription complexes. Mol. Cell 31, 510–519. doi: 10.1016/j.molcel.2008. Biophys. Acta Mol. Basis Dis. 1865, 14–25. doi: 10.1016/j.bbadis.2018.10.010 08.001 Scalea, S., Maresca, C., Catalanotto, C., Marino, R., Cogoni, C., Reale, A., et al. Kishi, N., Tang, Z., Maeda, Y., Hirai, A., Mo, R., Ito, M., et al. (2001). Murine (2020). Modifications of H3K4 methylation levels are associated with DNA homologs of deltex define a novel gene family involved in vertebrate Notch hypermethylation in acute myeloid leukemia. FEBS J. 287, 1155–1175. doi: signaling and neurogenesis. Int. J. Dev. Neurosci. 19, 21–35. doi: 10.1016/S0736- 10.1111/febs.15086 5748(00)00071-X Silva-Martínez, G. A., Rodríguez-Ríos, D., Alvarado-Caudillo, Y., Vaquero, A., Knights, A. J., Vohralik, E. J., Houweling, P. J., Stout, E. S., Norton, L. J., Esteller, M., Carmona, F. J., et al. (2016). Arachidonic and oleic acid exert Alexopoulos, S. J., et al. (2020). Eosinophil function in adipose tissue is distinct effects on the DNA methylome. Epigenetics 11, 321–334. doi: 10.1080/ regulated by Krüppel-like factor 3 (KLF3). Nat. Commun. 11:2922. doi: 10.1038/ 15592294.2016.1161873 s41467-020-16758-9 Song, L. R., Li, D., Weng, J. C., Li, C. B., Wang, L., Wu, Z., et al. (2020). MicroRNA- Kwong, W. Y., Wild, A. E., Roberts, P., Willis, A. C., and Fleming, T. P. (2000). 195 functions as a tumor suppressor by directly targeting fatty acid synthase in Maternal undernutrition during the preimplantation period of rat development malignant meningioma. World Neurosurg. 136:e0355-64. doi: 10.1016/j.wneu. causes blastocyst abnormalities and programming of postnatal hypertension. 2019.12.182 Development 127, 4195–4202. Stadler, M. B., Murr, R., Burger, L., Ivanek, R., Lienert, F., Schöler, A., et al. (2011). Lawrence, T. (2009). The nuclear factor NF-kappaB pathway in inflammation. Cold DNA-binding factors shape the mouse methylome at distal regulatory regions. Spring Harb. Perspect. Biol. 1:a001651. doi: 10.1101/cshperspect.a001651 Nature 480, 490–495. doi: 10.1038/nature10716 Lee, J., Lee, Y., Park, B., Won, S., Han, J. S., and Heo, N. J. (2018). Genome-wide Steffen, B. T., Duprez, D., Szklo, M., Guan, W., and Tsai, M. Y. (2018). Circulating association analysis identifies multiple loci associated with kidney disease- oleic acid levels are related to greater risks of cardiovascular events and all- related traits in Korean populations. PLoS One 13:e0194044. doi: 10.1371/ cause mortality: the Multi-Ethnic Study of Atherosclerosis. J. Clin. Lipidol. 12, journal.pone.0194044 1404–1412. doi: 10.1016/j.jacl.2018.08.004 Frontiers in Genetics | www.frontiersin.org 14 May 2021 | Volume 12 | Article 665769 Pescador-Tapia et al. Prandial State-Associated DNA Methylation Profiles Stepanova, M., Hossain, N., Afendy, A., Perry, K., Goodman, Z. D., Baranova, Zhang, L., Li, M. M., Corcoran, M., Zhang, S., and Cooper, G. J. S. (2015). Essential A., et al. (2010). Hepatic gene expression of Caucasian and African-American roles of insulin, AMPK signaling and lysyl and prolyl hydroxylases in the patients with obesity-related non-alcoholic fatty liver disease. Obes. Surg. 20, biosynthesis and multimerization of adiponectin. Mol. Cell. Endocrinol. 399, 640–650. doi: 10.1007/s11695-010-0078-2 164–177. doi: 10.1016/j.mce.2014.09.014 Sun, D., Zhang, T., Su, S., Hao, G., Chen, T., Li, Q.-Z., et al. (2019). Body mass Zhang, Y., Chen, Y., Qu, H., and Wang, Y. (2019). Methylation of HIF3A promoter index drives changes in DNA Methylation: a longitudinal study. Circ. Res. 125, CpG islands contributes to insulin resistance in gestational diabetes mellitus. 824–833. doi: 10.1161/CIRCRESAHA.119.315397 Mol. Genet. Genom. Med. 7:e00583. doi: 10.1002/mgg3.583 Tinahones, F. J., Coín-Aragüez, L., Mayas, M. D., Garcia-Fuentes, E., Hurtado- Zhang, Y., Kent, J. W., Olivier, M., Ali, O., Cerjak, D., Broeckel, U., et al. Del-Pozo, C., Vendrell, J., et al. (2012). Obesity-associated insulin resistance (2013). A comprehensive analysis of adiponectin QTLs using SNP association, is correlated to adipose tissue vascular endothelial growth factors and SNP cis-effects on peripheral blood gene expression and gene expression metalloproteinase levels. BMC Physiol. 12:4. doi: 10.1186/1472-6793-12-4 correlation identified novel metabolic syndrome (MetS) genes with potential Van Der Kwast, R. V. C. T., Van Ingen, E., Parma, L., Peters, H. A. B., Quax, P. H. A., role in carcinogenesis and systemic inflammation. BMC Med. Genomics 6:14. and Nossent, A. Y. (2018). Adenosine-to-inosine editing of MicroRNA-487b doi: 10.1186/1755-8794-6-14 alters target gene selection after ischemia and promotes neovascularization. Zheng, Q., Cao, Y., Chen, Y., Wang, J., Fan, Q., Huang, X., et al. (2018). Senp2 Circ. Res. 122, 444–456. doi: 10.1161/CIRCRESAHA.117.312345 regulates adipose lipid storage by de-SUMOylation of Setdb1. J. Mol. Cell Biol. Vehmeijer, F. O. L., Küpers, L. K., Sharp, G. C., Salas, L. A., Lent, S., Jima, D. D., 10, 258–266. doi: 10.1093/JMCB/MJX055 et al. (2020). DNA methylation and body mass index from birth to adolescence: Zhou, C., Yang, G., Chen, M., He, L., Xiang, L., Ricupero, C., et al. (2015). Lhx6 and meta-analyses of epigenome-wide association studies. Genome Med. 12:105. Lhx8: cell fate regulators and beyond. FASEB J. 29, 4083–4091. doi: 10.1096/fj. doi: 10.1186/s13073-020-00810-w 14-267500 Wang, L., Jin, Q., Lee, J.-E., Su, I., and Ge, K. (2010). Histone H3K27 methyltransferase Ezh2 represses Wnt genes to facilitate adipogenesis. Proc. Conflict of Interest: The authors declare that the research was conducted in the Natl. Acad. Sci. U.S.A. 107, 7317–7322. doi: 10.1073/pnas.1000031107 absence of any commercial or financial relationships that could be construed as a Wysocki, K., Conley, Y., and Wenzel, S. (2015). Epigenome variation in severe potential conflict of interest. asthma. Biol. Res. Nurs. 17, 263–269. doi: 10.1177/1099800414553463 Ye, J. J., Bian, X., Lim, J., and Medzhitov, R. (2020). Adiponectin and related Copyright © 2021 Pescador-Tapia, Silva-Martínez, Fragoso-Bargas, Rodríguez- C1q/TNF-related proteins bind selectively to anionic phospholipids and Ríos, Esteller, Moran, Zaina and Lund. This is an open-access article distributed sphingolipids. Proc. Natl. Acad. Sci. U.S.A. 117, 17381–17388. doi: 10.1073/ under the terms of the Creative Commons Attribution License (CC BY). The pnas.1922270117 use, distribution or reproduction in other forums is permitted, provided the Yi, J., Bergstrom, K., Fu, J., Shan, X., McDaniel, J. M., McGee, S., et al. (2019). original author(s) and the copyright owner(s) are credited and that the original Dclk1 in tuft cells promotes inflammation-driven epithelial restitution and publication in this journal is cited, in accordance with accepted academic practice. mitigates chronic colitis. Cell Death Differ. 26, 1656–1669. doi: 10.1038/s41418- No use, distribution or reproduction is permitted which does not comply with 018-0237-x these terms. Frontiers in Genetics | www.frontiersin.org 15 May 2021 | Volume 12 | Article 665769