Tatsuya Ehara,1,2 Yasutomi Kamei,3 Xunmei Yuan,3 Mayumi Takahashi,3 Sayaka Kanai,3 Erina Tamura,1

Kazutaka Tsujimoto,1 Takashi Tamiya,1 Yoshimi Nakagawa,4 Hitoshi Shimano,4,5 Takako Takai-Igarashi,6

Izuho Hatada,7 Takayoshi Suganami,3,8 Koshi Hashimoto,9 and Yoshihiro Ogawa1

Ligand-Activated PPARa-Dependent DNA DemethylationRegulates the Fatty Acidb-Oxidation Genes in thePostnatal LiverDiabetes 2015;64:775–784 | DOI: 10.2337/db14-0158

The metabolic function of the liver changes sequentiallyduring early life in mammals to adapt to the markedchanges in nutritional environment. Accordingly, hepaticfatty acid b-oxidation is activated after birth to produceenergy from breast milk lipids. However, how it is in-duced during the neonatal period is poorly understood.Here we show DNA demethylation and increased mRNAexpression of the fatty acid b-oxidation genes in thepostnatal mouse liver. The DNA demethylation doesnot occur in the fetal mouse liver under the physiologiccondition, suggesting that it is specific to the neonatalperiod. Analysis of mice deficient in the nuclear receptorperoxisome proliferator–activated receptor a (PPARa)and maternal administration of a PPARa ligand duringthe gestation and lactation periods reveal that the DNAdemethylation is PPARa dependent. We also find thatDNA methylation of the fatty acid b-oxidation genes arereduced in the adult human liver relative to the fetal liver.This study represents the first demonstration that theligand-activated PPARa-dependent DNA demethylation

regulates the hepatic fatty acid b-oxidation genes dur-ing the neonatal period, thereby highlighting the role ofa lipid-sensing nuclear receptor in the gene- and life-stage–specific DNA demethylation of a particular meta-bolic pathway.

Birth is a principal change in nutritional environment,when the major source of energy switches from glucose inthe cord blood to lipids in breast milk. The metabolicfunction of the liver changes sequentially during early lifein mammals, so that it becomes matured as a majormetabolic organ, thereby adapting to the marked changesin nutritional environment. After birth, metabolic path-ways such as gluconeogenesis, fatty acid b-oxidation, andde novo lipogenesis are known to be upregulated in theliver (1–3), whereas hepatic extramedullary hematopoiesisis downregulated (4). Indeed, hepatic fatty acid b-oxidationis activated progressively during the neonatal period toproduce energy from breast milk lipids (2,3). It should be

1Department of Molecular Endocrinology and Metabolism, Graduate School ofMedical and Dental Sciences, Tokyo Medical and Dental University, Bunkyo-ku,Tokyo, Japan2Nutrition Research Department, Nutritional Science Institute, Morinaga Milk In-dustry Co. Ltd., Zama, Kanagawa, Japan3Department of Organ Network and Metabolism, Graduate School of Medicaland Dental Sciences, Tokyo Medical and Dental University, Bunkyo-ku, Tokyo,Japan4Department of Internal Medicine (Metabolism and Endocrinology), Faculty ofMedicine, University of Tsukuba, Tsukuba, Ibaraki, Japan5International Institute for Integrative Sleep Medicine (WPI-IIIS), University ofTsukuba, Tsukuba, Ibaraki, Japan6Department of Health Record Informatics, Tohoku Medical Megabank Orga-nization, Aoba-ku, Sendai, Miyagi, Japan7Genome Science, Institute for Molecular and Cellular Regulation, GunmaUniversity, Maebashi, Gunma, Japan

8Japan Science and Technology Agency, PRESTO, Goban-cho Chiyoda-ku, Tokyo,Japan9Department of Preemptive Medicine and Metabolism, Graduate School of Medicaland Dental Sciences, Tokyo Medical and Dental University, Bunkyo-ku, Tokyo, Japan

Corresponding author: Yoshihiro Ogawa, ogawa.mem@tmd.ac.jp.

Received 29 January 2014 and accepted 21 September 2014.

This article contains Supplementary Data online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db14-0158/-/DC1.

Y.K. is currently affiliated with the Laboratory of Molecular Nutrition, GraduateSchool of Life and Environmental Sciences, Kyoto Prefectural University,Shimogamo, Sakyo-ku, Kyoto, Japan.

© 2015 by the American Diabetes Association. Readers may use this article aslong as the work is properly cited, the use is educational and not for profit, andthe work is not altered.

Diabetes Volume 64, March 2015 775

METABOLISM

critically involved in the regulation of metabolic homeosta-sis; deficiency of certain fatty acid b-oxidation enzyme(s)results in sudden infant death due to fasting intolerance(5). However, how a particular metabolic pathway is acti-vated during the liver maturation is poorly understood.

The nuclear receptor-type transcription factor peroxi-some proliferator–activated receptor a (PPARa) is consid-ered as a key regulator of lipid metabolism in the liver(6,7). Mice with targeted disruption of PPARa develophigh-fat diet– and fasting-induced fatty liver (8). PPARa,when activated by various lipid ligands, can stimulatetranscription of its target genes in the fatty acid b-oxidationpathway (6,9). During the suckling period, it may be acti-vated in response to breast milk–derived lipid ligand(s)(10). It is, therefore, conceivable that PPARa serves as amaster transcriptional regulator of the hepatic fatty acidb-oxidation pathway in early life.

The methylation of cytosine residues in DNA is amajor heritable and reversible epigenetic modification.DNA methylation of the promoter region is a well-studiedrepressive modification for gene expression (11). It is es-sential for embryonic development; DNA methylationcontributes to a variety of critical biological processessuch as genome imprinting and X-chromosome inactiva-tion (12–14). However, the functional role of DNA meth-ylation after organogenesis has not been fully addressedexcept for carcinogenesis. Recent studies showed thatgene- and life-stage–specific changes in DNA methylationoccur in certain metabolic genes during the fetal periodand even after birth (15,16). We have reported that expres-sion of the lipogenic glycerol-3-phosphate acyltransferase 1gene (Gpam) is induced in the postnatal mouse period viaa DNA demethylation mechanism (16). On the other hand,genome-wide analysis revealed marked changes in DNAmethylation in numerous genes in the fetal and neonatallivers in both the mouse and human (17,18). However, thedetailed molecular mechanism and physiologic implicationof the gene- and life-stage–specific changes in DNA meth-ylation has not been fully addressed.

Here we show that the ligand-activated PPARa-dependentDNA demethylation occurs in the promoter regions of its tar-get genes, which are mapped onto the fatty acid b-oxidationpathway in the postnatal mouse liver. We also demonstratethat DNA demethylation occurs specifically in the fatty acidb-oxidation genes from the fetal to adult human liver. Thedata of this study highlight the role of a lipid-sensing nu-clear receptor in the gene- and life-stage–specific DNAdemethylation of a particular metabolic pathway.

RESEARCH DESIGN AND METHODS

AnimalsAll animal experiments were approved by the TokyoMedical and Dental University Committee on AnimalResearch (0090041). Eight-week-old male and femaleC57BL6 (Japan SLC Inc., Hamamatsu, Japan) mice werecrossbred, and the offspring were used for analysis. Weobtained the liver samples at 18.5 days after fertilization

(e18.5, before birth), 2 days after birth (D2, just afterbirth), 16 days after birth (D16, preweaning), and 28 daysafter birth (D28, postweaning).

Since it was reported that differentiated hepatocytesexpressing specific metabolic enzymes are observed onembryonic day 15.5 (e15.5) (19), maternal administrationof PPARa ligand Wy14643 (Wy) was performed at 40mg/kg/day from gestation day 15 to 18 and from day 3to 16 of the lactation period by intraperitoneal injection.The PPARa-deficient (KO) mice (19) (strain B6.129S4-Pparatm1Gonz/J) were purchased from The Jackson Labo-ratory (Bar Harbor, ME). Homozygous male and femalePPARa-KO mice were crossed to obtain the PPARa-KOoffspring. The age-matched C57BL6 mice were used as wild-type (WT) control. The detailed protocol of Wy-administrationexperiments is illustrated in Supplementary Fig. 1.

DNA Methylation ProfilingThe mouse liver genomic DNA was extracted by thestandard proteinase K method. Human fetal (male 18 and22 weeks and female 18 weeks after fertilization) andadult (male 20, 44, and 59 years old) liver genomic DNAswere purchased from BioChain Institute Inc. (Newark,CA). The Microarray-Based Integrated Analysis of Meth-ylation by Isoschizomers (MIAMI) analysis, a genome-wide analysis of DNA methylation using a gene promoterarray and methylation-sensitive restriction enzyme, wasperformed as described (20–22). Detailed experimentalprocedure is described in the Supplementary Data.

cDNA Microarray AnalysisRNA was isolated from the livers of sex-matched mice ate18.5, D2, D16, and D28 (five samples for each group arecombined). Human fetal (male 20 weeks and female 38weeks after fertilization) and adult (male 26 and 44 yearsold) liver RNAs were purchased from BioChain InstituteInc. They were labeled with a cyanine 3-CTP using a LowInput Quick Amp Labeling Kit (Agilent Technologies Inc.,Santa Clara, CA) and hybridized to the Agilent wholemouse genome 4 3 44 K microarray or Human GeneExpression 4 3 44 K version 2 MicroArray. Signal detec-tion was performed according to manufacturer’s manual,and fluorescence signals obtained were normalized withthe global normalization method. Signals .1.5-fold and,1.5-fold were considered as increase and decrease, re-spectively. To describe the correlation between changes inDNA methylation and gene expression, scatter plot graphswere depicted with the log-transformed methylation differ-ences from MIAMI data and the log-transformed foldchanges from cDNA microarray data.

Gene Ontology AnalysisWe used DAVID version 6.7 (http://david.abcc.ncifcrf.gov/) (23) for the gene ontology (GO) analysis. The func-tional annotation chart and clustering of DAVID wereapplied to group similar GO terms. Benjamini-Hochbergmultiple testing corrections were performed, and the cor-rected P values were used for judging the candidate genes

776 DNA Demethylation of b-Oxidation Genes Diabetes Volume 64, March 2015

(P , 0.05 were considered significant). For the GO piechart (Fig. 1C), we counted the total number of genesannotated to each GO cluster.

Analysis of Transcriptional Factor Binding MotifsWe used the MATCH software (24) (BIOBASE GmbH,Worfenbuettel, Germany) to explore transcriptional

binding motifs in the promoter regions (from 21,000to +100 of each gene, taking the first nucleotide of exon1 as +1) of the genes of interest.

Pathway AnalysisFor pathway analysis, we used DAVID 6.7. Pathways forthe genes undergoing DNA methylation change were

Figure 1—Genome-wide DNA methylation analysis of the mouse liver. A: MIAMI analysis comparing e18.5 (labeled with Cy3) with D2 (Cy5),D2 (Cy3) with D16 (Cy5), and D16 (Cy3) with D28 (Cy5). Plots of log-transformed values of HpaII (methylation sensitive, horizontal axis) andMspI signal difference (methylation insensitive, vertical axis) between samples; a single plot shows a difference in DNA methylation statusof corresponding HpaII site between samples, and increase and decrease along the horizontal axis denotes DNA hypermethylation andhypomethylation of Cy5-labeled sample, respectively. The values of HpaII signal difference/MspI signal difference judged as increased anddecreased DNA methylation are above 2 and below 0.5, respectively. B: Correlation between DNA methylation and gene expression of thegenes with increased or decreased DNA methylation in A. Horizontal axis is log2-transformed values of methylation difference (HpaII/MspIsignal difference) based on the MIAMI data. Vertical axis is log2-transformed values of fold changes in mRNA expression based on thecDNA microarray data. C: GO pie graphs of the genes undergoing DNA hypermethylation with decreased mRNA expression (left) and DNAhypomethylation with increased mRNA expression (right) in the D16 liver relative to the D2 liver. (a) Hematocyte function; (b) hematocytefunction and metabolic function (genes with both functions); (c) metabolic function; (d) others (genes with other categories of function); and(e) genes without GO annotations. D: Mapping of the genes with DNA hypomethylation from D2 to D16 (B, middle panel) onto the fatty acidb-oxidation pathway, where 2,4-dienoyl-CoA reductase and 3,2-Enoyl-CoA isomerase are the enzymes involved in the oxidation of un-saturated fatty acids. Cpt1a, carnitine palmitoyltransferase 1a; Acox1, acyl-CoA oxidase 1; Decr1, 2,4-dienoyl CoA reductase 1; Decr2,2,4-dienoyl CoA reductase 2; Peci, enoyl-CoA delta isomerase 2; Ehhadh, enoyl-CoA, hydratase/3-hydroxyacyl CoA dehydrogenase;Echs1, enoyl-CoA hydratase, short chain, 1; Acaa1a, acetyl-CoA acyltransferase 1A; Acaa1b, acetyl-CoA acyltransferase 1B. E: Bisulfitesequencing analysis of Cpt1a, Acox1, and Ehhadh on D2 and D16. F: Relative mRNA levels of Cpt1a, Acox1, and Ehhadh on D2 and D16.Values for D2 are set at 100. ***P< 0.001 vs. D2; n = 5. G: Serum TG concentrations and hepatic TG contents in D2 and D16. **P< 0.01 vs.D2; n = 5. Data are expressed as the mean 6 SEM.

diabetes.diabetesjournals.org Ehara and Associates 777

enriched by the corrected P values from Benjamini-Hochberg multiple testing corrections (P , 0.05 were con-sidered significant). R heatmap.2 (function of gplot package)was used to create a heat map from the pathway analysisresults (Fig. 4C). The logarithms of the P values to base 10were converted into the heat map on blue-red scale.

Quantitative RT-PCR AnalysisQuantitative RT-PCR analysis was performed according tothe previous report (16) using the following primers: Acox1forward GCCTTTGTTGTCCCTATCCGT, reverse CGATATCCCCAACAGTGATGC; Cpt1a forward CCTGCATTCCTTCCCATTTG, reverse TGCCCATGTCCTTGTAATGTG; Ehhadhforward CCAATGCAAAGGCTCGTGTT, reverse GGTAGAAGCTGCGTTCCTCTTG.

Bisulfite DNA Methylation AnalysisBisulfite DNA methylation analysis was performedaccording to the previous report (16) using the followingspecific primers: Acox1 forward AATTGTGGGAGAGGGTGGGTTA, reverse AAACCATATCTCCAACCCCCTTATAT;Cpt1a forward GGATATTTTTATTTTTTGGTGGGAATAG,reverse ACTCCATAATCATTCTCTCTAACCTCCT; Ehhadhforward GGTTTAAGTTAAATTTGTAGTAGTTTGGTTGG;reverse AAAACAAAATCTAAATAAATCAACAATTCCT. Weused a web-based quantification tool for the bisulfite se-quencing analysis (http://quma.cdb.riken.jp/) (25).

Western Blot AnalysisWestern blot analysis was performed as described previously(26). Anti-ACOX1 (sc-98499; Santa Cruz Biotechnology, Inc.,Santa Cruz, CA), anti-CPT1A (ab128568; Abcam, Inc., Cam-bridge, MA), and anti-b-actin (sc-81178; Santa Cruz Biotech-nology, Inc.) were used as primary antibodies.

Hepatic Triacylglycerol AnalysisAfter extraction with 2:1 (vol/vol) chloroform/methanol,hepatic triacylglycerol (TG) contents were measured bythe enzymatic colorimetric method using triglyceride Etests (Wako Pure Chemicals, Osaka, Japan) according tothe previous report (16).

Hepatic Acetyl-CoA AnalysisHepatic acetyl-CoA levels were measured by Pico-ProbeAcetyl-CoA Fluorometric Assay kit (BioVision, Milpitas,CA).

Analysis of Serum Metabolic ParametersSerum TG, total cholesterol, nonesterified fatty acid(NEFA), and glucose concentrations were measured byTest Wako Kits (Wako). Serum insulin and total ketonebody concentrations were measured by Mouse InsulinELISA Kit (Morinaga Institute of Biological Science Inc.,Yokohama, Japan) and EnzyChrom Ketone Body AssayKit (BioAssay Systems, Hayward, CA), respectively.

Statistical AnalysisStatistical analysis was performed using Student t testand ANOVA followed by Scheffé test. Data are expressedas the mean 6 SEM. P , 0.05 was considered significant.

RESULTS

Characterization of DNA Methylation Status inPostnatal Mouse LiverWe analyzed the genome-wide methylation changes in thepostnatal mouse liver, comparing e18.5 with D2, D2 withD16, and D16 with D28. By the MIAMI analysis, we foundmarked changes in DNA methylation; 1,721 genes thatgained DNA methylation and 568 genes that lost DNAmethylation from D2 to D16 (Fig. 1A).

To identify genes with an inverse correlation betweenDNA methylation and mRNA expression, we combinedthe DNA methylation status of the genes with theirmRNA expression. Figure 1B shows the relationship be-tween the DNA methylation and mRNA expression levels.From D2 to D16, 433 genes gained DNA methylation withdecreased mRNA expression (25.2% of 1,721 genes) and249 genes lost DNA methylation with increased mRNAexpression (43.8% of 568 genes). Correlation analysisrevealed that negative correlation between methylationand expression is only observed in the genes that lostDNA methylation from D2 to D16 (R = 20.11; P =0.013) but not in those that gained DNA methylationfrom D2 to D16 (R = 0.0036; P = 0.89) or those thatlost DNA methylation from D16 to D28 (R = 0.0039;P = 0.97). The number of genes that underwent DNAmethylation change are summarized in SupplementaryTable 1.

GO Analysis of Gene Set with Inverse CorrelationBetween DNA Methylation and mRNA ExpressionWe next performed GO analysis. From D2 to D16, 10clusters were produced for genes that gained DNAmethylation with decreased mRNA expression and thosethat lost DNA methylation with increased mRNA expres-sion (Supplementary Table 2). The GO analysis data sug-gest that the genes with marked changes in DNAmethylation and mRNA expression are related to meta-bolic and hematocyte functions (Supplementary Table 2).In this study, the genes that lost DNA methylation withincreased mRNA expression from D16 to D28 did notproduce any significant GO terms.

Identification of Genes Classified with Hematocyteand/or Metabolic FunctionsWe illustrated the existence ratio of the genes classifiedwith hematocyte and/or metabolic functions in those withmarked difference in DNA methylation between D2 andD16 (Fig. 1C). The genes with metabolic function werepredominant among those that lost DNA methylationwith increased mRNA expression from D2 to D16 [indi-cated as (c) in Fig. 1C, pie graph on the right], suggestingthat some metabolic processes are activated in parallelwith DNA demethylation in the postnatal mouse liver.By contrast, many genes with increased DNA methylationand decreased mRNA expression were related to hematocytefunction [indicated as (a) in Fig. 1C, pie graph on the left],which is consistent with reduced hepatic extramedullaryhematopoiesis upon birth (4).

778 DNA Demethylation of b-Oxidation Genes Diabetes Volume 64, March 2015

Identification of Transcription Factor Binding Motifsand Pathway AnalysisWe investigated transcriptional factor binding motifs inthe promoter regions of the genes shown in Fig. 1C. Eightmotifs were significantly enriched in the genes that lostDNA methylation from D2 to D16 (Supplementary Table3). Several potential binding motifs of nuclear receptorPPARs (V$PPARG, V$PPAR_DR1, V$PPARA) were high-lighted. By contrast, analysis of the genes that gainedDNA methylation revealed no particular binding motifs.These observations suggest that the genes that lost DNAmethylation with increased mRNA expression are poten-tial targets of PPARs.

We next conducted a pathway analysis using the geneset shown in Fig. 1C. Several pathways were enriched,with the “PPAR signaling pathway” and “fatty acid metab-olism” being significant (Supplementary Table 4). Becauseamong PPARs PPARa is considered as a key regulator oflipid metabolism in the liver (6,9), we hereafter focusedon PPARa. We found that 9 out of 249 genes that lostDNA methylation from D2 to D16 are mapped onto thefatty acid b-oxidation pathway (Fig. 1D).

As the MIAMI analysis evaluates DNA methylation ofthe HpaII digestion site, we performed the detailed bisul-fite sequencing analysis of representative b-oxidationgenes, Acox1, Ehhadh, and Cpt1a, at several CpG sites,including the HpaII sites. In all cases, the HpaII sitesand neighboring CpG sites showed a concordant decreasein DNA methylation (Fig. 1E). We also examined mRNAexpression and protein levels from D2 to D16 (Fig. 1F andSupplementary Fig. 2A and B). There were significant neg-ative correlations between DNA methylation and mRNAexpression levels in all the genes highlighted in Fig. 1D.PPARa mRNA expression was also increased from D2 toD16. Serum TG concentrations and hepatic TG contentswere significantly decreased in D16 relative to D2 (Fig.1G), which is consistent with the upregulation of the fattyacid b-oxidation in the liver. As previously reported (27),serum glucose and insulin concentrations were signifi-cantly increased and decreased in D16 relative to D2, re-spectively. Serum total cholesterol concentration wassignificantly decreased in D16 (Supplementary Fig. 2C).

Genome-Wide Analysis of DNA Methylation inPPARa-KO MiceWe next performed genome-wide DNA methylationanalysis of the livers from PPARa-KO mice. There wasonly a slight difference in DNA methylation betweenWT and PPARa-KO mice on D2 (25 genes with DNAhypermethylation and 14 genes with DNA hypomethyla-tion in the KO liver relative to WT liver) (Fig. 2A, left). Bycontrast, there were more genes with DNA hypermeth-ylation in PPARa-KO mice on D16 (110 genes with DNAhypermethylation and 72 genes with DNA hypomethyla-tion) (Fig. 2A, right). The PPAR binding motif wasenriched in the promoter regions of the genes withDNA hypermethylation (Supplementary Table 3), which

are involved in fatty acid metabolism (Supplementary Ta-ble 4). Consistently, many of the genes with DNA hyper-methylation were mapped onto the fatty acid b-oxidationpathway (Fig. 2B). Decreased DNA methylation of Acox1and Ehhadh observed in the WT liver were partially butsignificantly attenuated in the KO liver from D2 to D16(Fig. 2C). Consistently, mRNA expression of Acox1 andEhhadh was not different between the WT and PPARa-KO mice on D2, whereas it was significantly decreased inPPARa-KO mice on D16 (Fig. 2D). In addition, serum TG,total cholesterol, and NEFA concentrations and hepaticTG contents were significantly increased in PPARa-KOmice relative to WT on D16 (Fig. 2E and SupplementaryFig. 3). There was no significant difference in serum glu-cose and insulin concentrations between WT and KO miceon D16 (Supplementary Fig. 3).

We found no significant difference in mRNA expres-sion of Cpt1a between WT and KO mice on both D2 andD16 (Fig. 2D), suggesting that DNA methylation andmRNA expression of Cpt1a do not depend upon PPARaduring the neonatal period. We therefore focused onAcox1 and Ehhadh for further analysis.

Effect of Maternal Administration of a PPARa Ligandon DNA MethylationTo explore if DNA demethylation of the fatty acid b-oxidationgenes depends upon PPARa activation, we adminis-tered a PPARa ligand Wy14643 (Wy) to dams in the lategestation and lactation periods (Supplementary Fig. 1A).DNA demethylation and increased mRNA expression ofAcox1 and Ehhadh was observed in the liver of the D16offspring derived from the Wy-administered dams (Fig.3A and B). In this study, PPARa mRNA expression wassignificantly increased (Supplementary Fig. 4A). We alsoexamined parameters related to the fatty acid b-oxidation.(Fig. 3C). Although hepatic TG contents were unchanged,serum TG and NEFA concentrations were significantlydecreased, whereas hepatic acetyl-CoA levels were signif-icantly increased, in the Wy-administered offspringon D16 (Fig. 3C and Supplementary Fig. 4B). Moreover,serum total ketone body concentrations were increased,although not significantly, in the Wy-administeredoffspring.

Notably, the Wy-induced DNA demethylation andincreased mRNA expression of Acox1 and Ehhadh alsooccurred, although slightly, in the liver of D2 offspring(Fig. 3D and E), which may be referred to as prematureDNA demethylation. The Wy-induced DNA demethylationwas also observed in other genes in the liver of D2 (32genes with DNA hypomethylation and 19 genes with DNAhypermethylation) (Fig. 3F). Pathway analysis of thegenes undergoing DNA demethylation identified thePPARa-related pathways (Supplementary Table 4).

We next examined DNA methylation and mRNAexpression in the liver of both D2 and D16 offspringderived from the Wy-administered PPARa-KO dams (Sup-plementary Fig. 1B). The Wy-induced DNA demethylation

diabetes.diabetesjournals.org Ehara and Associates 779

and increased mRNA expression of Acox1 and Ehhadh (Fig.3A, B, D, and E) were not observed in the Wy-adminis-tered PPARa-KO offspring (Fig. 3G and H). Consistentwith this, hepatic acetyl-CoA levels and serum ketonebody concentrations were not increased or serum TG con-centrations were not decreased in the Wy-administeredPPARa-KO offspring. (Supplementary Fig. 4C).

Genome-Wide DNA Methylation Analysis of Fetal andAdult Human LiverTo investigate if the DNA methylation changes in thepostnatal mouse liver (Fig. 1A–D) are conserved inhumans, we performed a genome-wide analysis of DNAmethylation and cDNA microarray analysis in the fetaland adult human liver samples. The MIAMI analysis iden-tified numerous genes with marked changes in DNAmethylation (529 genes with DNA hypomethylation and299 genes with DNA hypermethylation) (Fig. 4A) from

the fetus to adult human liver. Among those with DNAhypomethylation, 122 genes showed increased mRNA ex-pression. Pathway analysis highlighted the PPARa-relatedpathways in the genes with decreased DNA methylationand increased mRNA expression (Supplementary Table 4and Fig. 4B). A heat map generated with the pathwayanalysis data confirmed that the decreased DNA methyl-ation with increased mRNA expression in the PPAR sig-naling and fatty acid metabolism pathways occur in boththe mouse and human (Fig. 4C, compare D2 vs. D16 andfetus vs. adult). Interestingly, PPARa-dependent DNAmethylation changes were only observed in the genesmapped onto the two pathways (Fig. 4C, compare D2vs. D16, WT vs. KO, and vehicle vs. Wy).

DISCUSSION

Epigenetic modification of the genome such as DNAmethylation may play an important role in the organ

Figure 2—Genome-wide DNA methylation analysis of PPARa-KO mice. A: MIAMI analysis comparing PPARa-KO mice (labeled with Cy5)with WT mice (Cy3) on D2 and D16. Detailed information on the scatter plots is described in the legend to Fig. 1A. B: Mapping of the geneswith DNA hypermethylation (A, right panel) onto the fatty acid b-oxidation pathway in D16 KO mice relative to WT mice. The methylationdifference (HpaII/MspI signal difference) between KO and WT mice judged as increased DNA methylation is above 1.5. Ech, enoyl-CoAhydratase. C: Bisulfite sequencing analysis of Acox1 and Ehhadh in KO mice and WT mice. D: Relative mRNA levels of Acox1 and Ehhadhin KO and WT D2 and D16 mice. Values for D2 WT are set at 100. ***P < 0.001 vs. WT; N.S., not significant; n = 5. E: Serum TGconcentrations and hepatic TG contents in D16 WT and KO mice. *P < 0.05 vs. WT; n = 5. Data are expressed as the mean 6 SEM.

780 DNA Demethylation of b-Oxidation Genes Diabetes Volume 64, March 2015

development and maturation during the fetal and neo-natal periods. A previous study has suggested a role forDNA methylation in early postnatal liver development(18). Moreover, hepatic metabolic genes such as Tat (ty-rosine amino transferase) and Gpam are upregulated inparallel with DNA demethylation before or even afterbirth (15,16). Here we demonstrate that many of the fattyacid b-oxidation genes undergo DNA demethylation withincreased mRNA expression in the postnatal mouse liver.

The DNA demethylation and increased mRNA expressionare accompanied by increased protein expression. Notably,the DNA demethylation seems to be neonatal-stage–specific; it does not occur before birth under the physio-logic condition. We also find that DNA methylation isreduced in the fatty acid b-oxidation genes in the adulthuman liver relative to the fetal liver, suggesting that itis functionally conserved across species. This study is thefirst demonstration of gene- and life-stage–specific DNA

Figure 3—DNA methylation analysis of the offspring derived from the Wy-administered dams. (A) Bisulfite sequencing analysis and (B)relative mRNA levels of Acox1 and Ehhadh in the D16 offspring from the vehicle- and Wy-administered dams. Values for the vehicle are setat 100. ***P < 0.001 vs. vehicle; n = 5. C: Parameters of lipid metabolism in D16 offspring. Serum TG and total ketone body concentrations(left panel) and hepatic TG contents and acetyl-CoA levels (right panel) are shown. *P < 0.05 vs. vehicle; N.S., not significant; n = 5. (D)Bisulfite sequencing analysis and (E ) relative mRNA levels of Acox1 and Ehhadh in the D2 offspring from the vehicle- and Wy-administereddams. Values for the vehicle are set at 100; *P< 0.05 vs. vehicle; n = 5. F: MIAMI analysis of the D2 offspring from the vehicle- (labeled withCy3) and Wy- (Cy5) administered dams. Detailed information on the scatter plots is described in the legend to Fig. 1A. The methylationdifference (HpaII/MspI signal difference) between vehicle and Wy judged as decreased DNA methylation is below 0.667. (G) Bisulfitesequencing analysis and (H) relative mRNA levels of Acox1 and Ehhadh in the D2 and D16 PPARa-KO offspring from the vehicle orWy-administered dams. Values for the vehicle are set at 100. *P < 0.05 vs. vehicle; N.S., not significant; n = 5 (D2) and 8 (D16). Data areexpressed as the mean 6 SEM.

diabetes.diabetesjournals.org Ehara and Associates 781

demethylation of a particular metabolic pathway. It isconceivable that DNA demethylation of the fatty acidb-oxidation genes leads to their coordinate induction inthe neonatal liver, thereby representing an adaptativemechanism to the marked changes in nutritional envi-ronment after birth.

Because known factors involved in DNA demethylationpossess no DNA sequence specificity, how the genesspecific to a particular metabolic pathway are coordinatelydemethylated is an important issue to be addressed. Inthis study, analysis of the genes with DNA demethylationreveal PPAR binding motifs, suggesting that they aretranscriptional targets of PPARs. Notably, DNA demeth-ylation of the fatty acid b-oxidation genes during the neo-natal period is suppressed in PPARa-KO mice. On theother hand, maternal administration of a PPARa ligandaccelerates the DNA demethylation in the Wy-administeredoffspring. These observations suggest that ligand-activated PPARa triggers DNA demethylation of the fattyacid b-oxidation genes in the liver, thereby leading toincreased mRNA expression during the neonatal period.

In this regard, previous studies showed that transcriptionfactors and their DNA target sequences are critical deter-minants of the DNA methylation state in vitro (28,29).There is also evidence that thymine DNA glycosylase,which is considered as a critical regulator in the multipleenzymatic reactions of DNA demethylation (30), binds toseveral nuclear receptors (31,32). It is, therefore, interest-ing to speculate that the ligand-activated PPARa recruitsfactors involved in DNA demethylation to the promoterregions of the fatty acid b-oxidation genes via the PPAR-responsive element, thus leading to the gene-specific DNAdemethylation. Further studies are required to under-stand how PPARa triggers DNA demethylation in agene-specific manner.

The detailed mechanism of life-stage–specific DNAdemethylation of the fatty acid b-oxidation genes in theliver is poorly understood. In this study, we find that DNAdemethylation of the fatty acid b-oxidation genes doesnot occur in the fetal liver, where PPARa is expressedabundantly. It is, therefore, unlikely that increasedPPARa mRNA expression affects DNA demethylation

Figure 4—Genome-wide DNA methylation analysis of the fetal and adult human livers. A: MIAMI analysis of the fetal (labeled with Cy3) andadult (Cy5) human livers (left panel) and correlation between DNA methylation and expression of the genes with DNA methylation change(right panel). A representative result of three independent MIAMI analysis with similar results is shown. Detailed information on the scatterplots is described in the legend to Fig. 1A. The methylation difference (HpaII/MspI signal difference) between samples judged as decreasedDNA methylation is below 0.2. B: Mapping of the genes with DNA hypomethylation with increased mRNA expression in the adult liver ontothe fatty acid b-oxidation pathway. ACSL1, acyl-CoA synthetase long-chain family member 1. C: A heat map comparing the genome-wideDNA methylation of the metabolic pathways. Pathways with decreased DNA methylation from D2 to D16 are listed in the leftmost columnwith the corresponding P values. The DNA methylation changes in other pairs of comparison are displayed in each column with thecorresponding P values.

782 DNA Demethylation of b-Oxidation Genes Diabetes Volume 64, March 2015

and mRNA expression of PPARa target genes. On theother hand, maternal administration of Wy during thegestation period results in premature DNA demethylationin the liver of WT but not PPARa-KO offspring just afterbirth, indicating that DNA demethylation of the fatty acidb-oxidation genes in the liver is ligand-activated PPARadependent. We speculate that the PPARa-dependent DNAdemethylation does not occur in the fetal liver, wherePPARa ligands are unavailable under the physiologicalcondition. PPARa is known to be activated by lipid ligands(33), which, during the suckling period, are supplied bybreast milk (10). This discussion is supported by our ob-servation that there is no significant difference in theDNA methylation state of the fatty acid b-oxidation genesbetween PPARa-KO and WT mice on D2. These observa-tions suggest that the neonatal-stage–specific DNA de-methylation of the fatty acid b-oxidation genes is nota programmed process of liver maturation; it may be in-duced in response to the changes in nutritional environ-ment, when PPARa lipid ligands are available. Furtherstudies are required to identify the natural occurringPPARa ligands in breast milk lipids.

Transcriptional activation of the PPARa-target genesdepends upon the amount and kinds of fatty acid ligands(33,34). In this study, maternal administration of Wyduring the lactation period triggers DNA demethylationof the fatty acid b-oxidation genes, suggesting that DNAdemethylation in the neonatal liver can be controlled bythe PPARa ligands delivered to the pup via breast milk.Evidence has suggested that the fatty acid composition ofbreast milk is affected by maternal nutritional condition(35,36). It is reported that PPARa serves as an adaptivemechanism to nutritional changes such as an excess oflipid intake and fasting (8). In this study, analysis ofserum and hepatic metabolic parameters revealed thathepatic lipid metabolism is impaired in PPARa-KO miceand that maternal administration of Wy accelerates thefatty acid b-oxidation of WT but not of the PPARa-KOoffspring. It is conceivable that PPARa-dependent DNAdemethylation of the fatty acid b-oxidation genes in theneonatal liver is an adaptive response to the changes innutritional environment at the onset of lactation.

Given that the DNA methylation status established inearly life remains relatively stable throughout life and thusaffects hepatic lipid metabolism and susceptibility tometabolic diseases in later life (37), development of for-mula milk with the differing amount and kinds of PPARaligands may offer therapeutic strategies in early life to pre-vent the development of metabolic diseases in adulthood.

This study is the first demonstration of gene- andlife-stage–specific DNA demethylation on the fatty acidb-oxidation pathway, which is triggered by the ligand-activated PPARa during the liver maturation (Supplemen-tary Fig. 5). Before birth, when glucose, a major source ofenergy during the fetal period, is provided via the cord blood,expression of all the fatty acid b-oxidation genes examinedmay be suppressed in a DNAmethylation-dependent manner,

which is partly because PPARa ligands are unavailable.After birth, activation of hepatic PPARa by milk lipidligands may result in the activation of the fatty acidb-oxidation pathway via a DNA demethylation mechanism.It is, therefore, conceivable that during the neonatal period,milk lipids serve as a nutrient signal as well as nutrients sothat they can be oxidized efficiently as an energy source.

Acknowledgments. The authors thank Dr. Kyoichiro Tsuchiya of TokyoMedical and Dental University for critical reading of the manuscript.Funding. This work was supported in part by Grants-in-Aid for ScientificResearch from the Ministry of Education, Culture, Sports, Science, and Technol-ogy of Japan and the Ministry of Health, Labor, and Welfare of Japan. This workwas also supported by research grants from Ono Medical Foundation, The NaitoFoundation, and Nestlé Nutrition Council, Japan.Duality of Interest. No potential conflicts of interest relevant to this articlewere reported.Author Contributions. T.E. and Y.K. designed and performed research,analyzed data, and wrote the manuscript. X.Y., M.T., S.K., E.T., K.T., and T.T.performed research. Y.N., H.S., and I.H. contributed new reagents/analytic tools.T.T.-I. contributed new reagents/analytic tools and analyzed data. T.S. and K.H.analyzed data. Y.O. designed research and wrote the manuscript. Y.O. is theguarantor of this work and, as such, had full access to all the data in the studyand takes responsibility for the integrity of the data and the accuracy of the dataanalysis.

References1. Sekine K, Chen YR, Kojima N, Ogata K, Fukamizu A, Miyajima A. Foxo1 linksinsulin signaling to C/EBPalpha and regulates gluconeogenesis during liver de-velopment. EMBO J 2007;26:3607–36152. Perez-Castillo A, Schwartz HL, Oppenheimer JH. Rat hepatic mRNA-S14and lipogenic enzymes during weaning: role of S14 in lipogenesis. Am J Physiol1987;253:E536–E5423. Périchon R, Bourre JM. Peroxisomal beta-oxidation activity and catalaseactivity during development and aging in mouse liver. Biochimie 1995;77:288–2934. Brauner P, Nibbelink M, Flachs P, et al. Fast decline of hematopoiesis anduncoupling protein 2 content in human liver after birth: location of the protein inKupffer cells. Pediatr Res 2001;49:440–4475. Boles RG, Buck EA, Blitzer MG, et al. Retrospective biochemical screening offatty acid oxidation disorders in postmortem livers of 418 cases of sudden deathin the first year of life. J Pediatr 1998;132:924–9336. Lefebvre P, Chinetti G, Fruchart JC, Staels B. Sorting out the roles of PPARalpha in energy metabolism and vascular homeostasis. J Clin Invest 2006;116:571–5807. Kersten S, Desvergne B, Wahli W. Roles of PPARs in health and disease.Nature 2000;405:421–4248. Kersten S, Seydoux J, Peters JM, Gonzalez FJ, Desvergne B, Wahli W.Peroxisome proliferator-activated receptor alpha mediates the adaptive responseto fasting. J Clin Invest 1999;103:1489–14989. Rakhshandehroo M, Knoch B, Muller M, Kersten S. Peroxisome proliferator-activated receptor alpha target genes. PPAR Res 2010;201010. Hondares E, Rosell M, Gonzalez FJ, Giralt M, Iglesias R, Villarroya F. HepaticFGF21 expression is induced at birth via PPARalpha in response to milk intakeand contributes to thermogenic activation of neonatal brown fat. Cell Metab2010;11:206–21211. Maunakea AK, Nagarajan RP, Bilenky M, et al. Conserved role of intragenicDNA methylation in regulating alternative promoters. Nature 2010;466:253–25712. Okano M, Bell DW, Haber DA, Li E. DNA methyltransferases Dnmt3a andDnmt3b are essential for de novo methylation and mammalian development. Cell1999;99:247–257

diabetes.diabetesjournals.org Ehara and Associates 783

13. Kaneda M, Okano M, Hata K, et al. Essential role for de novo DNA meth-yltransferase Dnmt3a in paternal and maternal imprinting. Nature 2004;429:900–90314. Panning B. X-chromosome inactivation: the molecular basis of silencing. JBiol 2008;7:3015. Thomassin H, Flavin M, Espinás ML, Grange T. Glucocorticoid-induced DNAdemethylation and gene memory during development. EMBO J 2001;20:1974–198316. Ehara T, Kamei Y, Takahashi M, et al. Role of DNA methylation in theregulation of lipogenic glycerol-3-phosphate acyltransferase 1 gene expression inthe mouse neonatal liver. Diabetes 2012;61:2442–245017. Nagae G, Isagawa T, Shiraki N, et al. Tissue-specific demethylation in CpG-poor promoters during cellular differentiation. Hum Mol Genet 2011;20:2710–272118. Waterland RA, Kellermayer R, Rached MT, et al. Epigenomic profiling in-dicates a role for DNA methylation in early postnatal liver development. Hum MolGenet 2009;18:3026–303819. Lee SS, Pineau T, Drago J, et al. Targeted disruption of the alpha isoform ofthe peroxisome proliferator-activated receptor gene in mice results in abolish-ment of the pleiotropic effects of peroxisome proliferators. Mol Cell Biol 1995;15:3012–302220. Hatada I, Fukasawa M, Kimura M, et al. Genome-wide profiling of promotermethylation in human. Oncogene 2006;25:3059–306421. Takahashi M, Kamei Y, Ehara T, et al. Analysis of DNA methylation changeinduced by Dnmt3b in mouse hepatocytes. Biochem Biophys Res Commun 2013;434:873–87822. Kobayashi Y, Aizawa A, Takizawa T, et al. DNA methylation changes be-tween relapse and remission of minimal change nephrotic syndrome. PediatrNephrol 2012;27:2233–224123. Huang W, Sherman BT, Lempicki RA. Systematic and integrative analysis oflarge gene lists using DAVID bioinformatics resources. Nat Protoc 2009;4:44–5724. Kel AE, Gössling E, Reuter I, Cheremushkin E, Kel-Margoulis OV, WingenderE. MATCH: A tool for searching transcription factor binding sites in DNA se-quences. Nucleic Acids Res 2003;31:3576–357925. Kumaki Y, Oda M, Okano M. QUMA: quantification tool for methylationanalysis. Nucleic Acids Res 2008;36:W170-5

26. Kamei Y, Miura S, Suganami T, et al. Regulation of SREBP1c gene ex-pression in skeletal muscle: role of retinoid X receptor/liver X receptor andforkhead-O1 transcription factor. Endocrinology 2008;149:2293–230527. Mayor F, Cuezva JM. Hormonal and metabolic changes in the perinatalperiod. Biol Neonate 1985;48:185–19628. Stadler MB, Murr R, Burger L, et al. DNA-binding factors shape the mousemethylome at distal regulatory regions [published correction appears in Nature2012;484:550]. Nature 2011;480:490–49529. Lienert F, Wirbelauer C, Som I, Dean A, Mohn F, Schübeler D. Identificationof genetic elements that autonomously determine DNA methylation states. NatGenet 2011;43:1091–109730. Cortellino S, Xu J, Sannai M, et al. Thymine DNA glycosylase is essential foractive DNA demethylation by linked deamination-base excision repair. Cell 2011;146:67–7931. Chen D, Lucey MJ, Phoenix F, et al. T:G mismatch-specific thymine-DNAglycosylase potentiates transcription of estrogen-regulated genes through directinteraction with estrogen receptor alpha. J Biol Chem 2003;278:38586–3859232. Um S, Harbers M, Benecke A, Pierrat B, Losson R, Chambon P. Retinoic acidreceptors interact physically and functionally with the T:G mismatch-specificthymine-DNA glycosylase. J Biol Chem 1998;273:20728–2073633. Kliewer SA, Sundseth SS, Jones SA, et al. Fatty acids and eicosanoidsregulate gene expression through direct interactions with peroxisome proliferator-activated receptors alpha and gamma. Proc Natl Acad Sci U S A 1997;94:4318–432334. Nakatani T, Kim HJ, Kaburagi Y, Yasuda K, Ezaki O. A low fish oil inhibitsSREBP-1 proteolytic cascade, while a high-fish-oil feeding decreases SREBP-1mRNA in mice liver: relationship to anti-obesity. J Lipid Res 2003;44:369–37935. Francois CA, Connor SL, Wander RC, Connor WE. Acute effects of dietaryfatty acids on the fatty acids of human milk. Am J Clin Nutr 1998;67:301–30836. Arterburn LM, Hall EB, Oken H. Distribution, interconversion, and dose re-sponse of n-3 fatty acids in humans. Am J Clin Nutr 2006;83(Suppl.):1467S–1476S37. Gluckman PD, Hanson MA, Buklijas T, Low FM, Beedle AS. Epigeneticmechanisms that underpin metabolic and cardiovascular diseases. Nat Rev En-docrinol 2009;5:401–408

784 DNA Demethylation of b-Oxidation Genes Diabetes Volume 64, March 2015