Epigenome-wide association study of triglyceride postprandial responses to a high-fat dietary challenge
2016; Elsevier BV; Volume: 57; Issue: 12 Linguagem: Inglês
10.1194/jlr.m069948
ISSN1539-7262
AutoresChao‐Qiang Lai, Mary K. Wojczynski, Laurence D. Parnell, Bertha Hidalgo, Marguerite R. Irvin, Stella Aslibekyan, Michael A. Province, Devin Absher, Donna K. Arnett, José M. Ordovás,
Tópico(s)Genetic Associations and Epidemiology
ResumoPostprandial lipemia (PPL), the increased plasma TG concentration after consuming a high-fat meal, is an independent risk factor for CVD. Individual responses to a meal high in fat vary greatly, depending on genetic and lifestyle factors. However, only a few loci have been associated with TG-PPL response. Heritable epigenomic changes may be significant contributors to the unexplained inter-individual PPL variability. We conducted an epigenome-wide association study on 979 subjects with DNA methylation measured from CD4+ T cells, who were challenged with a high-fat meal as a part of the Genetics of Lipid Lowering Drugs and Diet Network study. Eight methylation sites encompassing five genes, LPP, CPT1A, APOA5, SREBF1, and ABCG1, were significantly associated with PPL response at an epigenome-wide level (P < 1.1 × 10−7), but no methylation site reached epigenome-wide significance after adjusting for baseline TG levels. Higher methylation at LPP, APOA5, SREBF1, and ABCG1, and lower methylation at CPT1A methylation were correlated with an increased TG-PPL response. These PPL-associated methylation sites, also correlated with fasting TG, account for a substantially greater amount of phenotypic variance (14.9%) in PPL and fasting TG (16.3%) when compared with the genetic contribution of loci identified by our previous genome-wide association study (4.5%). In summary, the epigenome is a large contributor to the variation in PPL, and this has the potential to be used to modulate PPL and reduce CVD. Postprandial lipemia (PPL), the increased plasma TG concentration after consuming a high-fat meal, is an independent risk factor for CVD. Individual responses to a meal high in fat vary greatly, depending on genetic and lifestyle factors. However, only a few loci have been associated with TG-PPL response. Heritable epigenomic changes may be significant contributors to the unexplained inter-individual PPL variability. We conducted an epigenome-wide association study on 979 subjects with DNA methylation measured from CD4+ T cells, who were challenged with a high-fat meal as a part of the Genetics of Lipid Lowering Drugs and Diet Network study. Eight methylation sites encompassing five genes, LPP, CPT1A, APOA5, SREBF1, and ABCG1, were significantly associated with PPL response at an epigenome-wide level (P < 1.1 × 10−7), but no methylation site reached epigenome-wide significance after adjusting for baseline TG levels. Higher methylation at LPP, APOA5, SREBF1, and ABCG1, and lower methylation at CPT1A methylation were correlated with an increased TG-PPL response. These PPL-associated methylation sites, also correlated with fasting TG, account for a substantially greater amount of phenotypic variance (14.9%) in PPL and fasting TG (16.3%) when compared with the genetic contribution of loci identified by our previous genome-wide association study (4.5%). In summary, the epigenome is a large contributor to the variation in PPL, and this has the potential to be used to modulate PPL and reduce CVD. Postprandial lipemia (PPL) refers to the changes in plasma lipoproteins following food consumption. PPL is highly correlated with fasting plasma TG concentrations, yet some evidence suggests that PPL may be an independent risk factor for CVD (1.Patsch W. Esterbauer H. Foger B. Patsch J.R. Postprandial lipemia and coronary risk.Curr. Atheroscler. Rep. 2000; 2: 232-242Crossref PubMed Scopus (35) Google Scholar, 2.Zilversmit D.B. Atherogenesis: a postprandial phenomenon.Circulation. 1979; 60: 473-485Crossref PubMed Scopus (1436) Google Scholar, 3.Pirillo A. Norata G.D. Catapano A.L. Postprandial lipemia as a cardiometabolic risk factor.Curr. Med. Res. Opin. 2014; 30: 1489-1503Crossref PubMed Scopus (88) Google Scholar, 4.Sarwar N. Danesh J. Eiriksdottir G. Sigurdsson G. Wareham N. Bingham S. Boekholdt S.M. Khaw K.T. Gudnason V. Triglycerides and the risk of coronary heart disease: 10,158 incident cases among 262,525 participants in 29 Western prospective studies.Circulation. 2007; 115: 450-458Crossref PubMed Scopus (1103) Google Scholar) in ways distinct from fasting TG (5.Nordestgaard B.G. Benn M. Schnohr P. Tybjaerg-Hansen A. Nonfasting triglycerides and risk of myocardial infarction, ischemic heart disease, and death in men and women.JAMA. 2007; 298: 299-308Crossref PubMed Scopus (1607) Google Scholar, 6.Bansal S. Buring J.E. Rifai N. Mora S. Sacks F.M. Ridker P.M. Fasting compared with nonfasting triglycerides and risk of cardiovascular events in women.JAMA. 2007; 298: 309-316Crossref PubMed Scopus (1234) Google Scholar, 7.Stalenhoef A.F. de Graaf J. Association of fasting and nonfasting serum triglycerides with cardiovascular disease and the role of remnant-like lipoproteins and small dense LDL.Curr. Opin. Lipidol. 2008; 19: 355-361Crossref PubMed Scopus (94) Google Scholar). The risk of CVD that can be independently attributed to elevated or elongated PPL (factors relating to amount and time) is supported by the fact that modern humans spend most of their lifetimes in the postprandial state (2.Zilversmit D.B. Atherogenesis: a postprandial phenomenon.Circulation. 1979; 60: 473-485Crossref PubMed Scopus (1436) Google Scholar). A persistent and elevated presence of TG-rich lipoprotein (TRL) particles in the circulation promotes thrombotic processes, which augment the risk for CVD events (3.Pirillo A. Norata G.D. Catapano A.L. Postprandial lipemia as a cardiometabolic risk factor.Curr. Med. Res. Opin. 2014; 30: 1489-1503Crossref PubMed Scopus (88) Google Scholar). PPL varies greatly among individuals, being defined, in addition to the characteristics of the food ingested, by age, sex, genetic variation, and environmental exposures (1.Patsch W. Esterbauer H. Foger B. Patsch J.R. Postprandial lipemia and coronary risk.Curr. Atheroscler. Rep. 2000; 2: 232-242Crossref PubMed Scopus (35) Google Scholar, 3.Pirillo A. Norata G.D. Catapano A.L. Postprandial lipemia as a cardiometabolic risk factor.Curr. Med. Res. Opin. 2014; 30: 1489-1503Crossref PubMed Scopus (88) Google Scholar). Several genome-wide association studies (GWASs) for fasting TG have been performed, revealing over 30 related loci (8.Teslovich T.M. Musunuru K. Smith A.V. Edmondson A.C. Stylianou I.M. Koseki M. Pirruccello J.P. Ripatti S. Chasman D.I. Willer C.J. et al.Biological, clinical and population relevance of 95 loci for blood lipids.Nature. 2010; 466: 707-713Crossref PubMed Scopus (2781) Google Scholar, 9.Lewis G.F. Xiao C. Hegele R.A. Hypertriglyceridemia in the genomic era: a new paradigm.Endocr. Rev. 2015; 36: 131-147Crossref PubMed Scopus (104) Google Scholar). However, only two GWASs have been conducted for PPL-TG (10.Wojczynski M.K. Parnell L.D. Pollin T.I. Lai C.Q. Feitosa M.F. O'Connell J.R. Frazier-Wood A.C. Gibson Q. Aslibekyan S. Ryan K.A. et al.Genome-wide association study of triglyceride response to a high-fat meal among participants of the NHLBI Genetics of Lipid Lowering Drugs and Diet Network (GOLDN).Metabolism. 2015; 64: 1359-1371Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar, 11.Pollin T.I. Damcott C.M. Shen H. Ott S.H. Shelton J. Horenstein R.B. Post W. McLenithan J.C. Bielak L.F. Peyser P.A. et al.A null mutation in human APOC3 confers a favorable plasma lipid profile and apparent cardioprotection.Science. 2008; 322: 1702-1705Crossref PubMed Scopus (519) Google Scholar), and only four genetic variants, which are also associated with fasting TG, have been identified in relation to PPL (9.Lewis G.F. Xiao C. Hegele R.A. Hypertriglyceridemia in the genomic era: a new paradigm.Endocr. Rev. 2015; 36: 131-147Crossref PubMed Scopus (104) Google Scholar, 10.Wojczynski M.K. Parnell L.D. Pollin T.I. Lai C.Q. Feitosa M.F. O'Connell J.R. Frazier-Wood A.C. Gibson Q. Aslibekyan S. Ryan K.A. et al.Genome-wide association study of triglyceride response to a high-fat meal among participants of the NHLBI Genetics of Lipid Lowering Drugs and Diet Network (GOLDN).Metabolism. 2015; 64: 1359-1371Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar, 11.Pollin T.I. Damcott C.M. Shen H. Ott S.H. Shelton J. Horenstein R.B. Post W. McLenithan J.C. Bielak L.F. Peyser P.A. et al.A null mutation in human APOC3 confers a favorable plasma lipid profile and apparent cardioprotection.Science. 2008; 322: 1702-1705Crossref PubMed Scopus (519) Google Scholar, 12.Lai C.Q. Arnett D.K. Corella D. Straka R.J. Tsai M.Y. Peacock J.M. Adiconis X. Parnell L.D. Hixson J.E. Province M.A. et al.Fenofibrate effect on triglyceride and postprandial response of apolipoprotein A5 variants: the GOLDN study.Arterioscler. Thromb. Vasc. Biol. 2007; 27: 1417-1425Crossref PubMed Scopus (108) Google Scholar). This may result from the limited sample sizes of these studies combined with the strong environmental influence on this phenotype (13.Ordovas J.M. Genetics, postprandial lipemia and obesity.Nutr. Metab. Cardiovasc. Dis. 2001; 11: 118-133PubMed Google Scholar). Environmental factors exert significant effects on gene expression, in some instances through epigenetic mechanisms (14.Tammen S.A. Friso S. Choi S.W. Epigenetics: the link between nature and nurture.Mol. Aspects Med. 2013; 34: 753-764Crossref PubMed Scopus (258) Google Scholar). Thus, epigenetic marks can be considered "fingerprints" of that communication between the environment and the genome and some experimental evidence indicates that diet-induced epigenetic changes can be transmitted through several generations (15.Huypens P. Sass S. Wu M. Dyckhoff D. Tschop M. Theis F. Marschall S. de Angelis M.H. Beckers J. Epigenetic germline inheritance of diet-induced obesity and insulin resistance.Nat. Genet. 2016; 48: 497-499Crossref PubMed Scopus (218) Google Scholar, 16.Sabet J.A. Park L.K. Iyer L.K. Tai A.K. Koh G.Y. Pfalzer A.C. Parnell L.D. Mason J.B. Liu Z. Byun A.J. et al.Paternal B vitamin intake is a determinant of growth, hepatic lipid metabolism and intestinal tumor volume in female Apc1638N mouse offspring.PLoS One. 2016; 11: e0151579PubMed Google Scholar). As such, the environment, including habitual diet, may contribute to the health status of the individual and his/her descendants (17.Ling C. Groop L. Epigenetics: a molecular link between environmental factors and type 2 diabetes.Diabetes. 2009; 58: 2718-2725Crossref PubMed Scopus (441) Google Scholar). We hypothesize that individuals respond to environmental exposures by modulation of the epigenome, which elicits changes in the PPL response that could alter CVD risk. Of different forms of epigenetic modification, DNA methylation is the most extensively studied for its technical feasibility at the epigenome-wide scale, cost-effectiveness, well-established standard analysis platform, and its apparent relation to nutrition (18.Anderson O.S. Sant K.E. Dolinoy D.C. Nutrition and epigenetics: an interplay of dietary methyl donors, one-carbon metabolism and DNA methylation.J. Nutr. Biochem. 2012; 23: 853-859Crossref PubMed Scopus (481) Google Scholar). The objective of this study was to conduct an epigenome-wide association analysis in order to identify DNA methylation sites that were associated with PPL-TG concentrations in response to a high-fat meal in the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) study. In addition, we have characterized DNA sequence variation in the significant epigenome-wide association study (EWAS) regions in relation to PPL-TG responses. The GOLDN study, as a part of the National Heart, Lung, and Blood Institute Family Heart Study, recruited participants (n = 1,327) from families of European descent at two field centers: Minneapolis, MN and Salt Lake City, UT. GOLDN was designed as an intervention study to identify genetic factors that determine lipid responses to two interventions: 1) a high-fat meal test; and 2) a 3 week treatment of fenofibrate (160 mg). Participants were requested to stop the use of lipid-lowering medication for at least 4 weeks and to refrain from alcohol for 24 h prior to their study visit. Diet history questionnaires were used to collect demographic, lifestyle, and dietary data (19.Smith C.E. Arnett D.K. Tsai M.Y. Lai C.Q. Parnell L.D. Shen J. Laclaustra M. Junyent M. Ordovas J.M. Physical inactivity interacts with an endothelial lipase polymorphism to modulate high density lipoprotein cholesterol in the GOLDN study.Atherosclerosis. 2009; 206: 500-504Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar). The study protocol was approved by the Institutional Review Boards at Tufts University, the University of Minnesota, the University of Utah, and the University of Alabama at Birmingham. All participants provided written consent for the study. The current study comprised a total of 979 participants for whom complete PPL and epigenome data exist. Postprandial TG responses were calculated based on the growth curve models of TG as the function of times, as described (10.Wojczynski M.K. Parnell L.D. Pollin T.I. Lai C.Q. Feitosa M.F. O'Connell J.R. Frazier-Wood A.C. Gibson Q. Aslibekyan S. Ryan K.A. et al.Genome-wide association study of triglyceride response to a high-fat meal among participants of the NHLBI Genetics of Lipid Lowering Drugs and Diet Network (GOLDN).Metabolism. 2015; 64: 1359-1371Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar). Briefly, the postprandial phenotypes were estimated as four measurements: uptake, clearance, area under the whole curve (AUC), and area under the curve increase (AUI). Uptake was estimated as the slope of the TG response from 0 to 3.5 h after the meal consumption, a time at which most fat from the meal has been absorbed (3.Pirillo A. Norata G.D. Catapano A.L. Postprandial lipemia as a cardiometabolic risk factor.Curr. Med. Res. Opin. 2014; 30: 1489-1503Crossref PubMed Scopus (88) Google Scholar). Clearance was defined as the downward slope of the TG level from 3.5 to 6 h after meal consumption, which measures the speed of the metabolic process that metabolizes the excess fat from the plasma (3.Pirillo A. Norata G.D. Catapano A.L. Postprandial lipemia as a cardiometabolic risk factor.Curr. Med. Res. Opin. 2014; 30: 1489-1503Crossref PubMed Scopus (88) Google Scholar, 20.Cohn J.S. Postprandial lipemia and remnant lipoproteins.Clin. Lab. Med. 2006; 26: 773-786Abstract Full Text Full Text PDF PubMed Scopus (47) Google Scholar). The AUC was calculated as the total AUC according to the trapezoid method, and the AUI was estimated by subtracting the baseline area from the AUC (10.Wojczynski M.K. Parnell L.D. Pollin T.I. Lai C.Q. Feitosa M.F. O'Connell J.R. Frazier-Wood A.C. Gibson Q. Aslibekyan S. Ryan K.A. et al.Genome-wide association study of triglyceride response to a high-fat meal among participants of the NHLBI Genetics of Lipid Lowering Drugs and Diet Network (GOLDN).Metabolism. 2015; 64: 1359-1371Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar). Different cell types in whole blood may have contrasting methylation patterns. Thus, to minimize the confounding effect of cell type differences and increase the consistency of methylation measures across samples, we restricted DNA methylome analysis to CD4+ T cells, which represent the most common lymphocytes in whole blood (21.Aslibekyan S. Demerath E.W. Mendelson M. Zhi D. Guan W. Liang L. Sha J. Pankow J.S. Liu C. Irvin M.R. et al.Epigenome-wide study identifies novel methylation loci associated with body mass index and waist circumference.Obesity (Silver Spring). 2015; 23: 1493-1501Crossref PubMed Scopus (122) Google Scholar, 22.Irvin M.R. Zhi D. Joehanes R. Mendelson M. Aslibekyan S. Claas S.A. Thibeault K.S. Patel N. Day K. Jones L.W. et al.Epigenome-wide association study of fasting blood lipids in the Genetics of Lipid-lowering Drugs and Diet Network study.Circulation. 2014; 130: 565-572Crossref PubMed Scopus (158) Google Scholar). Using CD4+ specific antigen magnetic beads (Invitrogen, Carlsbad, CA), CD4+ T cells were isolated from frozen buffy coat samples that were collected at visit 2 (baseline) before fenofibrate intervention. DNA was extracted from these cells using DNeasy kits (Qiagen, Venlo, The Netherlands) (23.Absher D.M. Li X. Waite L.L. Gibson A. Roberts K. Edberg J. Chatham W.W. Kimberly R.P. Genome-wide DNA methylation analysis of systemic lupus erythematosus reveals persistent hypomethylation of interferon genes and compositional changes to CD4+ T-cell populations.PLoS Genet. 2013; 9: e1003678Crossref PubMed Scopus (242) Google Scholar). Genome-wide DNA methylation of all DNA samples was quantified as described (23.Absher D.M. Li X. Waite L.L. Gibson A. Roberts K. Edberg J. Chatham W.W. Kimberly R.P. Genome-wide DNA methylation analysis of systemic lupus erythematosus reveals persistent hypomethylation of interferon genes and compositional changes to CD4+ T-cell populations.PLoS Genet. 2013; 9: e1003678Crossref PubMed Scopus (242) Google Scholar) using Illumina Infinium human methylation 450K arrays (Illumina, San Diego, CA), which contain over 485,000 probe sets to measure DNA methylation of over 450,000 CpG sites across the human genome. Using Illumina's GenomeStudio package, we estimated the proportion of total signal of methylation for each probe as the β score, and detection P value as the probability that the total intensity for a given probe falls within the background signal intensity. Methylation signals were then further filtered out if CpG sites met one of the following criteria: 1) detection P > 0.01 and 1.5% of samples have missing data; or 2) >10% of samples have no adequate intensity (23.Absher D.M. Li X. Waite L.L. Gibson A. Roberts K. Edberg J. Chatham W.W. Kimberly R.P. Genome-wide DNA methylation analysis of systemic lupus erythematosus reveals persistent hypomethylation of interferon genes and compositional changes to CD4+ T-cell populations.PLoS Genet. 2013; 9: e1003678Crossref PubMed Scopus (242) Google Scholar). For adjustment of the batch effect across samples, the filtered β scores were normalized separately for Infinium I and II probe sets using the ComBat package for R (22.Irvin M.R. Zhi D. Joehanes R. Mendelson M. Aslibekyan S. Claas S.A. Thibeault K.S. Patel N. Day K. Jones L.W. et al.Epigenome-wide association study of fasting blood lipids in the Genetics of Lipid-lowering Drugs and Diet Network study.Circulation. 2014; 130: 565-572Crossref PubMed Scopus (158) Google Scholar, 24.Chen C. Grennan K. Badner J. Zhang D. Gershon E. Jin L. Liu C. Removing batch effects in analysis of expression microarray data: an evaluation of six batch adjustment methods.PLoS One. 2011; 6: e17238Crossref PubMed Scopus (334) Google Scholar, 25.Hidalgo B. Irvin M.R. Sha J. Zhi D. Aslibekyan S. Absher D. Tiwari H.K. Kabagambe E.K. Ordovas J.M. Arnett D.K. Epigenome-wide association study of fasting measures of glucose, insulin, and HOMA-IR in the Genetics of Lipid Lowering Drugs and Diet Network study.Diabetes. 2014; 63: 801-807Crossref PubMed Scopus (119) Google Scholar). At the end, 464,005 CpG sites passed quality control and these were used for statistical analysis in this study. To control for heterogeneity of CD4+ T cells across all samples, principal components (PCs) based on the β scores of all autosomal CpG sites that passed quality control were calculated using the prcomp function in R (v12.12.1). Four PCs were used in the EWAS. The detailed procedure of genome-wide genotyping in GOLDN has been described (10.Wojczynski M.K. Parnell L.D. Pollin T.I. Lai C.Q. Feitosa M.F. O'Connell J.R. Frazier-Wood A.C. Gibson Q. Aslibekyan S. Ryan K.A. et al.Genome-wide association study of triglyceride response to a high-fat meal among participants of the NHLBI Genetics of Lipid Lowering Drugs and Diet Network (GOLDN).Metabolism. 2015; 64: 1359-1371Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar, 26.Aslibekyan S. Kabagambe E.K. Irvin M.R. Straka R.J. Borecki I.B. Tiwari H.K. Tsai M.Y. Hopkins P.N. Shen J. Lai C.Q. et al.A genome-wide association study of inflammatory biomarker changes in response to fenofibrate treatment in the Genetics of Lipid Lowering Drug and Diet Network.Pharmacogenet. Genomics. 2012; 22: 191-197Crossref PubMed Google Scholar). In this study, we used the hybrid genotype data of 2,543,887 SNPs, among which 484,029 were genotyped using the Affymetrix Genome-wide 6.0 Array (Affymetrix, Santa Clara, CA). The remaining SNPs were imputed using MaCH software (version 1.0.16) with human genome build 36 as reference, and genotyped SNPs that met the following criteria (27.Kraja A.T. Borecki I.B. Tsai M.Y. Ordovas J.M. Hopkins P.N. Lai C.Q. Frazier-Wood A.C. Straka R.J. Hixson J.E. Province M.A. et al.Genetic analysis of 16 NMR-lipoprotein fractions in humans, the GOLDN study.Lipids. 2013; 48: 155-165Crossref PubMed Scopus (31) Google Scholar): call rate >96%, minor allele frequency >1%, and Hardy-Weinberg equilibrium test P > 10−6. As the epigenome is known as the fingerprint of individuals in response to lifetime exposures up to the time point when the samples were collected, an individual epigenotype depends on local environments. Therefore, individuals with the same genotype may have different epigenotypes under different environments. For such reasons, the best discovery and replication in EWASs should be done within the same population. We randomly split the whole population (n = 979) into two-thirds as a discovery sample (n = 653) and one-third as a replicate sample (n = 326) using Proc Surveyselect in SAS v9.4 (Cary, NC) while holding the distributions of baseline TG, BMI, and sex similar between the two samples. To examine the differences in clinical characteristics between the sexes among the discovery and replication, we performed a t-test. In the discovery stage, we modeled the association between methylation β score at each CpG site and PPL response measures using a linear mixed model (28.Kang H.M. Sul J.H. Service S.K. Zaitlen N.A. Kong S.Y. Freimer N.B. Sabatti C. Eskin E. Variance component model to account for sample structure in genome-wide association studies.Nat. Genet. 2010; 42: 348-354Crossref PubMed Scopus (1619) Google Scholar), adjusting for sex, age, age2, age3, study site, and the first four PCs for T cell impurity as fixed effects, and kinship as a random effect. The kinship matrix was generated based on family pedigree (29.Wright S. Coefficients of inbreeding and relationship.Am. Nat. 1922; 56: 330-338Crossref Google Scholar). The analysis was implemented in SNP and VARIATION SUITE 8.4.3 (GoldenHelix Inc., Bozeman, MT). In addition, we conducted an EWAS adjusted for an additional covariate of baseline TG. We applied the Bonferroni correction, setting epigenome-wide significance at 1.1 × 10−7 (25.Hidalgo B. Irvin M.R. Sha J. Zhi D. Aslibekyan S. Absher D. Tiwari H.K. Kabagambe E.K. Ordovas J.M. Arnett D.K. Epigenome-wide association study of fasting measures of glucose, insulin, and HOMA-IR in the Genetics of Lipid Lowering Drugs and Diet Network study.Diabetes. 2014; 63: 801-807Crossref PubMed Scopus (119) Google Scholar). We subsequently fitted the identical model in the replication sample (n = 326) for the CpG sites that were statistically significant in the discovery set for PPL measures. We corrected the threshold for significance in the replication stage for multiple testing using the Bonferroni approach, P = 0.05/number of replicated sites. Applying the concept of meta-analysis, we then combined the discovery and replication samples (i.e., the entire sample n = 979) and repeated the analysis using the same models (with or without adjusting for the baseline TG) as in the discovery and replication stages. Because PPL-TG response traits are strongly correlated with fasting TG, we also conducted an EWAS for fasting TG with the entire population using the same method and model (without adjusting for baseline TG). Variance contribution of individual methylation sites was estimated using efficient mixed-model association while controlling for normalized kinship (28.Kang H.M. Sul J.H. Service S.K. Zaitlen N.A. Kong S.Y. Freimer N.B. Sabatti C. Eskin E. Variance component model to account for sample structure in genome-wide association studies.Nat. Genet. 2010; 42: 348-354Crossref PubMed Scopus (1619) Google Scholar) that was calculated based on family pedigree (29.Wright S. Coefficients of inbreeding and relationship.Am. Nat. 1922; 56: 330-338Crossref Google Scholar). This procedure was implemented in the Mixed Linear Model Analysis tools of SNP and VARIATION SUITE 8.4.3 (GoldenHelix Inc., Bozeman, MT). As the identified methylation sites were not totally independent from each other, the combined variance contribution of all methylation sites was estimated with the option of Multi-Locus Mixed Model of the Mixed Linear Model Analysis while controlling for family relationship and covariates. The variance contribution of the previously identified genetic variants (rs964184 and rs10243693) that were associated with AUC (10.Wojczynski M.K. Parnell L.D. Pollin T.I. Lai C.Q. Feitosa M.F. O'Connell J.R. Frazier-Wood A.C. Gibson Q. Aslibekyan S. Ryan K.A. et al.Genome-wide association study of triglyceride response to a high-fat meal among participants of the NHLBI Genetics of Lipid Lowering Drugs and Diet Network (GOLDN).Metabolism. 2015; 64: 1359-1371Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar) was calculated using the same method in participants (n = 707) for whom the genotype data was available. For the CpG sites that showed a significant association, we further examined their correlations with loci previously identified (10.Wojczynski M.K. Parnell L.D. Pollin T.I. Lai C.Q. Feitosa M.F. O'Connell J.R. Frazier-Wood A.C. Gibson Q. Aslibekyan S. Ryan K.A. et al.Genome-wide association study of triglyceride response to a high-fat meal among participants of the NHLBI Genetics of Lipid Lowering Drugs and Diet Network (GOLDN).Metabolism. 2015; 64: 1359-1371Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar) that were associated with AUC in participants for whom both epigenome and genome data were available (n = 707). In addition, we further examined the association of SNPs within a 50 kb region of each CpG site associating with AUC. Data from previous genetic association studies were retrieved from the GWAS catalog (30.Welter D. MacArthur J. Morales J. Burdett T. Hall P. Junkins H. Klemm A. Flicek P. Manolio T. Hindorff L. et al.The NHGRI GWAS catalog, a curated resource of SNP-trait associations.Nucleic Acids Res. 2014; 42: D1001-D1006Crossref PubMed Scopus (1997) Google Scholar) and gene-environment interactions from CardioGxE (31.Parnell L.D. Blokker B.A. Dashti H.S. Nesbeth P.D. Cooper B.E. Ma Y. Lee Y.C. Hou R. Lai C.Q. Richardson K. et al.CardioGxE, a catalog of gene-environment interactions for cardiometabolic traits.BioData Min. 2014; 7: 21Crossref PubMed Scopus (49) Google Scholar). The TG-related characteristics of the discovery and replication samples are listed in Table 1. There were no significant differences between the discovery and replication samples for the TG and PPL response-related phenotypes (Table 1). However, there were equivalent significant differences in the baseline TG and TG AUC between sexes within each sample.TABLE 1Characteristics of discovery and replicate samples in GOLDNDiscovery Sample (n = 653)Replication Sample (n = 326)MenWomenBothMenWomenBothn313340653156170326Age, years48.1 (15.9)47.9 (16.4)48.0 (16.2)50.2 (17.3)47.6 (16.4)48.8 (16.9)BMI, kg/m228.3 (4.6)28.2 (6.4)28.2 (5.6)28.5 (5.0)28.2 (6.4)28.3 (5.8)Waist, inches100.0 (13.8)93.2 (17.8)96.5 (16.4)100.6 (13.6)93.1 (16.9)96.7 (15.8)TG at baseline, mg/dl149.1 (111.0)aP value for differences between men and women (within sample) significant with P < 0.05.125.7 (82.2)136.9 (97.7)144.4 (90.1)127.2 (87.3)135.4 (89.0)TG uptake slope0.18 (0.03)0.17 (0.03)0.18 (0.03)0.18 (0.03)0.18 (0.03)0.18 (0.03)TG clearance slope−0.06 (0.05)−0.07 (0.05)−0.07 (0.05)−0.05 (0.05)−0.07 (0.05)−0.06 (0.05)TG AUC31.6 (3.4)bP value for differences between men and women (within sample) significant with P < 0.001.30.4 (3.3)30.9 (3.4)31.6 (3.3)aP value for differences between men and women (within sample) significant with P < 0.05.30.3 (3.3)30.9 (3.4)TG AUI2.5 (0.5)2.4 (0.5)2.4 (0.6)2.6 (0.6)2.4 (0.5)2.5 (0.6)Values are means (standard deviations).a P value for differences between men and women (within sample) significant with P < 0.05.b P value for differences between men and women (within sample) significant with P < 0.001. Open table in a new tab Values are means (standard deviations). We first conducted epigenome-wide association tests for each of four PPL-TG traits (Table 2) in the discovery sample (n = 653). For AUC, we identified four methylation sites in three genes (CPT1A, APOA5, and SREBF1) that reached epigenome-wide significance at P ≤ 1.1 × 10−7 (Table 2). However, when adjusted for baseline TG, no methylation site reached epigenome-wide significance (supplemental Table S1). For the other three TG response traits (uptake, clearance, and AUI), we did not find any methylation sites that reached epigenome-wide significance either with or without adjusting for baseline TG. We then replicated the findings from the discovery stage in the replication sample (n = 326; Table 2). All four sites that were associated with AUC in the discovery sample replicated in the second sample after correction for multiple testing (P < 0.05/4 = 0.0125). After adjustment for baseline TG, only two methylation sites (cg00574958 and cg17058475 at CPT1A) were replicated in the replication sample (supplemental Table S1).TABLE 2CpG sites associated with AUC in response to a high-fat meal in discovery and replication samplesMarkerChr:PositionaGenomic position was based on genome build 37.GeneDiscovery (n = 653)Replication (n = 326)β (SE)Pβ (SE)Pcg0057495811:68607622CPT1A−33.24 (4.91)3.
Referência(s)