Identification of Novel Regulatory Regions Induced by Intrauterine Growth Restriction in Rat Islets

Abstract

Intrauterine growth restriction (IUGR) leads to the development of type 2 diabetes in adulthood, and the permanent alterations in gene expression implicate an epigenetic mechanism. Using a rat model of IUGR, we performed TrueSeq-HELP Tagging to assess the association of DNA methylation changes and gene dysregulation in islets. We identified 511 differentially methylated regions (DMRs) and 4377 significantly altered single CpG sites. Integrating the methylome and our published transcriptome data sets resulted in the identification of pathways critical for islet function. The identified DMRs were enriched with transcription factor binding motifs, such as Elk1, Etv1, Foxa1, Foxa2, Pax7, Stat3, Hnf1, and AR. In silico analysis of 3-dimensional chromosomal interactions using human pancreas and islet Hi-C data sets identified interactions between 14 highly conserved DMRs and 35 genes with significant expression changes at an early age, many of which persisted in adult islets. In adult islets, there were far more interactions between DMRs and genes with significant expression changes identified with Hi-C, and most of them were critical to islet metabolism and insulin secretion. The methylome was integrated with our published genome-wide histone modification data sets from IUGR islets, resulting in further characterization of important regulatory regions of the genome altered by IUGR containing both significant changes in DNA methylation and specific histone marks. We identified novel regulatory regions in islets after exposure to IUGR, suggesting that epigenetic changes at key transcription factor binding motifs and other gene regulatory regions may contribute to gene dysregulation and an abnormal islet phenotype in IUGR rats.

Poor fetal growth resulting from intrauterine growth restriction (IUGR) is strongly associated with an increased risk of type 2 diabetes (T2D) (1, 2). We developed a rat model of uteroplacental insufficiency induced by bilateral uterine artery ligation to elucidate the mechanisms by which IUGR results in the development of T2D later in life. The adverse IUGR milieu leads to immune cell infiltration of islets, decreased islet capillary density, and impaired development of pancreatic islets in the fetus (2-4). Furthermore, IUGR rats exhibit impaired insulin secretion and reduced islet vascularity at birth and have permanent β-cell dysfunction with reduced β-cell mass, mild-fasting hyperglycemia, and diminished glucose- and leucine-stimulated insulin secretion in adulthood (2, 4-6). Altered expression of more than 3000 genes in pathways regulating islet function, such as nutrient metabolism and transport, insulin secretion, innervation, extracellular matrix, and inflammation, was observed in the transcriptome from young and adult IUGR rats (7). However, there were major differences in the transcriptomic profiles from islets isolated from 2-week-old compared to adult IUGR rats (7). The changes in the transcriptome and phenotype in adult IUGR rats could not necessarily be predicted by the transcriptome changes observed at age 2 weeks.

Epigenetic modifications of the genome are heritable and reversible changes that modulate gene expression and contribute to the development of abnormal phenotypes and metabolic disorders, including diabetes, obesity, and cardiovascular diseases (8). There are at least 3 distinct categories of epigenetic modifications: DNA methylation, histone modifications, and noncoding RNAs.

Previously using the microarray-based HELP (HpaII tiny fragment enrichment by ligation-mediated PCR) assay with methylation-sensitive restriction enzyme HpaII and methylation-insensitive MspI, we demonstrated that IUGR alters DNA methylation at approximately 1400 loci in adult rat islets, and identified that many of these changes were associated with nearby gene expression changes (9).

We have also recently shown that IUGR induces significant genome-wide changes in enrichment of H3K4 trimethylation (H3K4me3), H3K27 trimethylation (H3K27me3), and H3K27 acetylation (H3K27Ac) marks in islets early in life, many of which persist into adulthood (10). These histone mark changes are enriched at critical transcription factor binding motifs and correlated with expression changes of genes critical for normal islet function that were disrupted in IUGR islets. These findings suggest that IUGR induces significant genome-wide epigenetic modifications in islets that are associated with changes in gene expression.

It is not clear how DNA methylation and histone modifications interact to regulate gene expression in IUGR islets. To identify additional epigenetic mechanisms as well as to determine if epigenetic modifications in early life can predict later phenotypic changes in adulthood that are associated with islet dysfunction in our model of IUGR-induced T2D, we took a multiomics approach, which has the power to identify the control mechanisms underpinning complex traits. Here we demonstrate that alterations in DNA methylation and histone modifications interact at key transcription factor binding motifs and in other gene regulatory regions in early life that are associated with changes in gene expression that may contribute to abnormal islet phenotypes that occur later in adult IUGR rats. Further, with in silico analysis of 3-dimensional (3D) chromosomal interactions, we identified novel differentially methylated regions (DMRs) and differentially methylated single CpG sites with potential distal regulatory functions.

Materials and Methods

Animal Model

The animals and procedures used in this study were approved by the animal care committee of the Children’s Hospital of Philadelphia and the Perelman School of Medicine at the University of Pennsylvania. Our IUGR animal model has been previously described (2). Briefly, bilateral uterine artery ligation was performed on pregnant Sprague-Dawley rats (Charles River) at embryonic day 18.5. Sham surgery was performed in controls. Pups were weighed the day of delivery to confirm IUGR, and litters were randomly culled to 8 to equilibrate postnatal nutrient availability. Dams and offspring were given ad libitum access to water and standard rodent chow.

Islet Isolation

Pancreata were excised at age 2 weeks and followed by islet isolation. Following the previous phenotypic studies using pooled samples from both sexes (4, 7), pancreata from one male and one female rat of the same litter were pooled for each sample. Pancreatic islets were isolated as previously described (4). Briefly, pancreata were digested with Collagenase P (Millipore Sigma) in Hank’s balanced salt solution supplemented with 4-mM NaCO₃ and 1% bovine serum albumin for 15 minutes at 37 °C. Digested tissues were then washed in cold supplemented Hank’s balanced salt solution without collagenase. Islets were isolated by histopaque gradient centrifugation.

Library Preparation for TrueSeq-HELP Tagging Assay

Genomic DNA was extracted from islets, and libraries for the TrueSeq-HELP Tagging assay, a sequencing-based HELP assay with Illumina Tru-Seq adapter, were prepared using a previously published protocol (11). Briefly, genomic DNA was digested with methylation-sensitive restriction enzymes HpaII or MspI (New England Biolabs), and then precipitated with isopropanol. Samples were ligated with index adapter TS_AE, digested with EcoP15I (New England Biolabs), and then end-repaired. After adding dA tailing, samples were ligated with TS_AS adapters, and in vitro–transcribed using MEGAshortscript (Thermo Fisher Scientific). Libraries were gel-purified by Qiagen MinElute Gel Extraction kit (Qiagen) after reverse transcription and amplification. The quality of libraries was assessed using a 2100 BioAnalyzer (Agilent) and Qubit 2.0 fluorometer (Thermo Fisher Scientific).

TrueSeq-HELP Tagging Assay and Data Analysis

Three biological replicates of multiplexed TrueSeq-HELP Tagging libraries for each group were single-end sequenced to 50 bp on an Illumina hiSeq 2500 in Center for Epigenomics at Albert Einstein College of Medicine. The raw data were processed and analyzed as previously described (11). The reads were mapped to rat genome assembly rn6. All sequencing tracks were made by HOMER. One IUGR sample had suboptimal reads and when unsupervised analysis was performed, this IUGR was a significant outlier (Supplementary Fig. S1) (12), thus this sample was removed from data analysis. DMRs were identified by Defiant, an in-house bioinformatic analysis program (13), using the following criteria: 1) minimum coverage of 10 in all samples, 2) an absolute percentage change in methylation at the DMR of greater than or equal to 10%, 3) a maximum distance between 2 consecutive CpGs of 5Kb and at least 4 CpGs per region, 4) a minimum of 2 differentially methylated CpGs in the DMR, and 5) a Q value less than .05 over the full DMR. The Benjamini-Hochberg approach was applied for multiple testing to obtain false discovery rate (FDR, Q values). Nearest genes were assigned to each DMR and significantly differentially methylated single CpGs. Functional analysis was performed using Qiagen’s Ingenuity Pathway Analysis (IPA). Enrichment analysis of transcription factor binding motifs within DMR and single CpG site ± 200 bp was carried out using HOMER. Sequence data have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE151684. Principal component analysis was performed using the 1000 most differentially methylated loci.

In Silico Analysis of 3-Dimensional Chromosomal Interaction

DMR sequences were first extracted from rat genome (rn6) and aligned to the human genome (hg38) using both Blast (https://blast.ncbi.nlm.nih.gov/Blast.cgi) and BLAT (14). Cutoff values of 30% query coverage and 50% identity were applied, and if multiple hits were reported, only the highest one was included in downstream analysis. Similar to DMR alignment, sequences from regions of single CpG sites ± 100 bp were extracted and aligned against the human genome with cutoff values of 50% query coverage and 50% identity. Once ortholog sequences were obtained, we identified Hi-C loops or topologically associating domains in the published human pancreas and islet data (15-19) that may contain one or more of the ortholog regions using BEDTools (20), and reported genes that were affected under either partner of the loops or topologically associating domains.

Integrating DNA Methylation With Chromatin Immunoprecipitation Sequencing Data

Peaks of histone modification sites from 2-week-old IUGR and control islets were obtained from our previously published data set (10). Histone marker differential enrichment analysis was conducted using diffReps. DMR or CpG ±5-kb regions were used to compare with differential peaks using BedTools. DMR and CpG sites that had at least 50% covered by peaks were reported.

Results

Genome-wide Assessment of DNA Methylome in Islets at Age 2 Weeks

In our previous studies assessing epigenetic changes in IUGR islets, we interrogated genome-wide DNA methylation only at age 7 weeks (9). Because changes in the transcriptome between ages 2 and 7 weeks in IUGR islets were significantly different (9), we measured global DNA methylation in pancreatic islets isolated from 2-week-old control and IUGR Sprague-Dawley rats. DNA methylation at randomly selected loci were compared between control and IUGR samples to generate a representative heat map that showed no overall shift of the genome-wide pattern of DNA methylation toward hypomethylation or hypermethylation in the IUGR group (Fig. 1A) similar to what we have previously shown in adulthood (9). We identified 511 DMRs (Fig. 1B, Supplementary Table S1) (12), including 52% intergenic, 17% intronic, and 10% exonic regions (Fig. 1C). Approximately 16% of DMRs were located in promoter regions. The 511 identified DMRs were associated with 481 proximal genes, 22 of which were associated with a gain in DNA methylation, and 489 were associated with a loss of DNA methylation in IUGR compared to control (see Supplementary Table S1) (12). Interestingly, our genome-wide DNA methylation analysis identified more than 4000 significantly altered single CpG sites. The absolute percentage changes in methylation at these single CpG sites were 70% or greater in IUGR islets compared with control islets (Supplementary Table S2) (12).

Figure 1.

DNA methylome of 2-week-old rat islets. A, An unsupervised clustering heat map of randomly selected loci showing an absence of bias in global methylation between control (C) and intrauterine growth restriction (IUGR) islets. Each row in the heat map corresponds to data from a single locus. The branching dendrogram at the top corresponds to the relationships among samples. Hypermethylation and hypomethylation are shown on a continuum from blue to red, respectively. B, Heat map of differentially methylated regions (DMRs) showing significant differences in cytosine methylation between C and IUGR islets. Each row in the heat map corresponds to a data point from a single locus, whereas columns correspond to individual samples. The branching dendrogram corresponds to the relationships among samples, as determined by clustering using the 512 identified DMRs. Hypermethylation and hypomethylation are shown on a continuum from blue to red, respectively. C, Pie chart representing the location and proportion of DMRs. The gene body included exons and introns. The promoter was limited to 3 kb upstream from the transcriptional start site. The 5′-untranslated region began at the transcription start site and ended before the initiation sequence. The 3′-untranslated region began immediately after the translation termination sequence. The intergenic region is composed of the regions not included in the previously defined regions.

Changes in DNA Methylation Early in Development Is a Potential Epigenetic Mechanism Contributing to Dysregulation of Genes and Changes in Islet Development and Function Later in Life

To identify potential molecular pathways disrupted in IUGR islets, IPA was used to identify function networks potentially regulated by DMRs. Gene lists were generated by identifying the gene most proximal to each DMR. IPA revealed that more than 150 canonical pathways were disrupted in 2-week-old IUGR islets and were enriched for pathways critical for islet function, such as T2D, insulin receptor signaling, mechanistic target of rapamycin signaling, ceramide signaling, protein kinase A signaling, nuclear factor κB signaling, Gβγ protein signaling, androgen signaling, and estrogen signaling (Table 1). There was also an enrichment of pathways that regulate innervation and neuronal function, such as ephrin receptor, CREB, and axonal guidance, and pathways modulating angiogenesis and vascular function, including relaxin and nitric oxide signaling (see Table 1). Pathways regulating the immune system were also altered in 2-week-old IUGR islets, including FcγRIIB signaling in B lymphocytes, PKCθ signaling in T lymphocytes, and iCOS-iCOSL signaling in T-helper cells. Of note, these findings are highly correlated with IUGR phenotypes in islets including impaired insulin secretion, mitochondrial dysfunction, islet capillary rarefaction, and inflammation later in life (2-6).

Table 1.

Top canonical pathways implicated by differentially methylated regions in 2-week-old intrauterine growth restriction islets

Ingenuity canonical pathways	P	Differentially methylated genes in the pathway
Relaxin signaling	2.19E-05	GNAZ, GNA13, VEGFA, NFKBIA, ENPP6, PDE8B, PDE6A, PDE4D, GRB2, PRKAG1, FGFR4, GUCY2G, PIK3C3
Astrocyte and oligodendrocyte cell proliferation signaling	1.12E-04	PRKCE, TGFA, CALM1, PDGFRB, RBL2, CAMK2G, GRB2, FGFR4, RRAS, PIK3C3
CREB signaling in neurons	1.17E-04	PRKCE, GNAZ, GNA13, CALM1, CAMK2G, GRB2, CACNA1C, CACNA1G, CACNA2D2, PRKAG1, FGFR4, POLR2G, RRAS, PIK3C3
3-Phosphoinositide biosynthesis	2.51E-04	PDGFRB,PTPA,PTPN1,INPP5B,DUSP5,GRB2,PTPRF,FGFR4,PTPN13,PTPN20,SGPP2,PIK3C3,NUDT14
Nitric oxide signaling	5.01E-04	PRKCE, CALM1, VEGFA, GRB2, CACNA1C, BDKRB2, PRKAG1, FGFR4, PIK3C3
G β γ signaling	5.37E-04	PRKCE, GNAZ, GNA13, GRB2, CACNA1C, CACNA1G, CACNA2D2, PRKAG1, RRAS
Insulin receptor signaling	6.31E-04	INPP5B, GRB2, ASIC2, SCNN1A, PTPRF, PRKAG1, FGFR4, RRAS, PIK3C3, PTPN1
Ephrin receptor signaling	9.33E-04	GNAZ, GNA13, ITSN1, VEGFA, GRB2, EFNA5, PTPN13, EPHB1, RRAS, KALRN, EFNB3
FcγRIIB signaling in B lymphocytes	1.12E-03	GRB2, CACNA1C, CACNA1G, CACNA2D2, FGFR4, RRAS, PIK3C3
PKCθ signaling in T lymphocytes	1.58E-03	CAMK2G, GRB2, CACNA1C, CACNA1G, CACNA2D2, FGFR4, NFATC1, NFKBIA, RRAS, PIK3C3
G-protein coupled receptor signaling	1.70E-03	PRKCE, NFKBIA, ENPP6, PDE8B, CAMK2G, PDE6A, PDE4D, CHRM5, GRB2, PRKAG1, FGFR4, RRAS, PIK3C3, HTR2A
Androgen signaling	2.00E-03	PRKCE, GNAZ, CALM1, GNA13, CACNA1C, CACNA1G, CACNA2D2, PRKAG1, POLR2G
NRF2-mediated oxidative stress response	2.04E-03	PRKCE, FKBP5, GSTT2/GSTT2B, ACTG1, BACH1, GRB2, MAF, FGFR4, RRAS, ENC1, PIK3C3
Paxillin signaling	2.19E-03	ITGA1, ITGB5, ACTG1, ITGAE, GRB2, FGFR4, RRAS, PIK3C3
Gap junction signaling	2.24E-03	PRKCE, CSNK1E, ACTG1, GRB2, PRKAG1, FGFR4, TUBA3E, RRAS, GUCY2G, PIK3C3, HTR2A
Axonal guidance signaling	2.51E-03	PRKCE, GNAZ, GNA13, ITSN1, SRGAP3, VEGFA, ECEL1, EFNA5, NFATC1, EPHB1, NTNG2, GRB2, PRKAG1, FGFR4, TUBA3E, RRAS, PIK3C3, KALRN, EFNB3
mTOR signaling	2.82E-03	PRKCE, RPS10, VEGFA, PTPA, GRB2, PRKAG1, FGFR4, RHOJ, RRAS, PIK3C3, MAPKAP1
Ceramide signaling	2.88E-03	KSR1, PTPA, GRB2, FGFR4, RRAS, CNKSR1, PIK3C3
Caveolar-mediated endocytosis signaling	3.24E-03	ITGA1, ITGB5, ACTG1, ITSN1, ITGAE, PTPN1
Protein kinase A signaling	3.72E-03	PRKCE, GNA13, NFATC1, NFKBIA, ADD2, PTPN1, CALM1, DUSP5, ENPP6, PDE8B, CAMK2G, PDE6A, MYLK, PDE4D, PTPRF, PRKAG1, PTPN13
iCOS-iCOSL signaling in T-helper cells	3.72E-03	CALM1, CAMK2G, GRB2, IL2RB, FGFR4, NFATC1, NFKBIA, PIK3C3
NF-κB signaling	4.07E-03	TGFA, PDGFRB, IL1R1, GRB2, AZI2, FGFR4, NFKBIA, TRAF3, RRAS, PIK3C3
Type 2 diabetes mellitus signaling	4.37E-03	PRKCE, GRB2, CACNA1C, CACNA1G, CACNA2D2, PRKAG1, FGFR4, NFKBIA, PIK3C3
BMP signaling pathway	4.47E-03	NKX2-5, ZNF423, GRB2, PRKAG1, RRAS, RUNX2
Estrogen receptor signaling	4.68E-03	NCOR2, MED21, GRB2, TAF6, MED10, POLR2G, RRAS, RUNX2

Abbreviations: mTOR, mechanistic target of rapamycin; NF-κB, nuclear factor κB.

Next, we sought to determine whether IUGR-induced changes in DNA methylation in 2-week-old islets correlated with gene expression changes. We compared the methylome data set (511 DMRs) with our previously published transcriptome data set (> 1300 differentially expressed genes with Q < .05) obtained from 2-week-old islets (7). Changes in the expression of 30 genes overlapped with changes in DNA methylation, including Pah, Pde8b, Pfkp, Camkk2, Nfkbia, Htr2a, Col5a3, and Itsn1 (Table 2). Although all 30 genes showed decreased DNA methylation in 2-week-old IUGR islets, the expression of 13 genes was upregulated, including Pfkb, Htr2a, and Col5a3, and the expression of the remaining 17 genes was downregulated, including Pah, Pde8b, and Camkk2. While DNA methylation at gene promoters is strongly associated with gene silencing, DNA methylation in intronic regions may mark enhancers or repressors and can be associated with increased or decreased gene expression (21, 22). Pfkp is phosphofructokinase, a key regulatory enzyme in glycolysis, and plays a critical role in glucose metabolism (23). 5-Hydroxytrptamine receptor 2A (Htr2a) is a receptor for neurotransmitter serotonin that regulates both insulin and glucagon secretion. Overexpression of Htr2a is associated with islet dysfunction and T2D (24). Collagen type V α-3 (Col5A3) plays a critical role in glucose homeostasis. Col5a3-deficient mice have reduced islet numbers, decreased β-cell function, and increased β-cell death (25). Phosphodiesterase 8B (Pde8b) regulates insulin response and secretion in islets via modulating the 3′,5′-cyclic adenosine 5′-monophosphate (cAMP) pool (26). Calcium/calmodulin-dependent protein kinase kinase 2 (Camkk2) is important for calcium-mediated response to glucose signaling in β cells and regulates insulin release and insulin sensitivity (27). It can also phosphorylate adenosine monophosphate–activated protein kinase C (AMPK) to regulate insulin secretion and β-cell survival.

Table 2.

Correlation between differentially methylated regions in old intrauterine growth restriction islets and proximal gene expression at age 2 weeks

Gene name	Symbol	DMRs		RNA-seq
		CpGs location	ΔMethylationa,b, %	LogFCa,b
SH3 domain binding kinase 1	Sbk1	Intergenic	–34.6	–1.27
Phenylalanine hydroxylase	Pah	Intron 3	–12.6	–1.93
Chymotrypsin-like elastase 1	Cela1	Intron 4-6	–12.8	–3.02
Calmodulin-regulated spectrin-associated protein family, member 3	Camsap3	Intron 1	–23.1	–1.07
Glycoprotein 2	Gp2	Intergenic	–35.1	–2.63
Solute carrier family 4 member 8	Slc4a8	Intergenic	–22.3	–1.34
Phosphodiesterase 8B	Pde8b	Intron 4-5	–24.2	–1.39
LBH regulator of WNT signaling pathway	Lbh	Intergenic	–18.0	–1.09
Cysteine and serine-rich nuclear protein 1	Csrnp1	Exon 4	–11.2	1.13
Histocompatibility minor 13	Hm13	Intergenic	–16.5	–0.89
Tubulin tyrosine ligase-like 10	Ttll10	Intron 3-5	–20.5	–1.45
Connector enhancer of kinase suppressor of Ras 1	Cnksr1	Intergenic	–23.0	–0.91
Phosphofructokinase, platelet	Pfkp	Intergenic	–21.4	0.69
Gasdermin C	Gsdmc	Intergenic	–29.3	3.77
Carboxypeptidase A4	Cpa4	Intergenic	–28.3	–3.56
RNA binding protein, mRNA processing factor	Rbpms	Promoter	–24.8	0.63
Echinoderm microtubule associated protein-like 1	Eml1	Intron 2	–31.1	0.66
Echinoderm microtubule associated protein-like 1	Eml1	Intron 4	–13.5	0.66
Calcium/calmodulin-dependent protein kinase 2	Camkk2	Intron 1	–14.0	–0.71
NFKB inhibitor α	Nfkbia	Intergenic	–26.1	0.90
Olfactomedin 1	Olfm1	Intergenic	–12.7	1.17
Erythrocyte membrane protein band 4.1-like 4A	Epb41l4a	Intergenic	–24.0	–0.58
Annexin A1	Anxa1	Intergenic	–36.2	0.87
5-Hydroxytryptamine receptor 2A	Htr2a	Intergenic	–21.8	1.01
Collagen type V α 3 chain	Col5a3	Exon 36-41	–21.4	0.84
Docking protein 5	Dok5	Intron 1	–24.2	1.67
Calmin	Clmn	Intergenic	–20.7	–0.81
HtrA serine peptidase 1	Htra1	Intron 3-5	–18.9	0.66
Intersectin 1	Itsn1	Intron and exon 30	–14.8	0.54
BTB domain and CNC homolog 1	Bach1	Intergenic	–35.3	–0.59
Ventricular zone-expressed PH domain-containing 1	Veph1	Intergenic	–24.8	–1.34

Abbreviations: DMR, differentially methylated region; mRNA, messenger RNA; RNA-seq, RNA sequencing.

^a ΔMethylation values are means from DNA methylome, and LogFC (fold change) values are means from transcriptome.

^b Q values less than.05 for both DMR ΔMethylation and RNA-seq LogFC.

Despite an overlap of only 30 genes having DNA methylation and gene expression changes, when we compared pathways that were altered in each data set, there was substantial overlap. For example, insulin signaling, peroxisome proliferator–activated receptor, and AMPK signaling, gap junction and tight junction signaling, endocytosis signaling, as well as pathways regulating extracellular matrix (Table 3) were altered in both data sets. Thus, the transcriptomic data set corroborates the methylome data and further highlights the disruption of pathways critical for β-cell function.

Table 3.

Overlapping canonical pathways of 2-week-old DNA methylome and transcriptome data sets

Ingenuity canonical pathways	Methylome DMR analysis		Transcriptome analysis
	P	Differentially methylated genes	P	Differentially expressed genes
Gap junction signaling	2.24E-03	PRKCE, CSNK1E, ACTG1, GRB2, PRKAG1, FGFR4, TUBA3E, RRAS, Gucy2g, PIK3C3, HTR2A	1.23E-02	TUBB3, NOV, PRKAR2B, TUBB6, ITPR3, TUBB2A, CAV1, GUCY1A2, EGF, GUCY1B3, ACTA1, TUBB2B
Caveolar-mediated endocytosis signaling	3.24E-03	ITGA1, ITGB5, ACTG1, ITSN1, ITGAE, PTPN1	1.02E-02	FLNC, CAV1, EGF, INSR, PTRF, ACTA1, ITGB3
Type 2 diabetes mellitus signaling	4.37E-03	PRKCE, GRB2, CACNA1C, CACNA1G, CACNA2D2, PRKAG1, FGFR4, NFKBIA, PIK3C3	2.95E-02	NFKBIA, TNFRSF1A, MAPK10, ADIPOQ, PRKAA2, INSR, TNF, ACSL1, TNFRSF11B
ILK signaling	7.24E-03	MYH3, ITGB5, ACTG1, VEGFA, PTPA, GRB2, FGFR4, RHOJ, TMSB10/TMSB4X, PIK3C3	3.55E-02	MYL9, FN1, RHOB, CFL2, TNFRSF1A, FLNC, MAPK10, VIM, VCL, TNF, ACTA1, ITGB3
Calcium signaling	9.77E-03	TRPC7, MYH3, Calm1, CAMK2G, CACNA1C, CACNA1G, CAMKK2, CACNA2D2, PRKAG1, NFATC1	4.07E-02	MYL9, PRKAR2B, Tpm4, CAMK1D, HDAC11, ITPR3, Tpm1, HDAC10, Tpm2, CAMKK2, ACTA1
tRNA splicing	1.02E-02	ENPP6, PDE8B, PDE6A, PDE4D	4.27E-02	SMPDL3A, PDE4C, PDE4A, PDE8B
PPAR signaling	1.29E-02	NCOR2, PDGFRB, IL1R1, GRB2, NFKBIA, RRAS	4.37E-03	PPARA, NFKBIA, NR0B2, TNFRSF1A, PPARD, IL1B, INSR, TNF, TNFRSF11B
Fibrosis/Stellate cell activation	1.45E-02	TGFA, MYH3, PDGFRB, COL23A1, IL1R1, VEGFA, IFNGR2, COL5A3, TIMP2	1.74E-04	IGFBP4, FN1, TNFRSF1A, COL12A1, SMAD3, KLF6, EGF, COL20A1, MYL9, CXCL3, COL5A3, IL1B, EDNRA, COL18A1, TNF, AGT, TNFRSF11B
LPS/IL-1–mediated inhibition of RXR Function	1.58E-02	HS3ST3B1, Sult1c2, NDST2, GSTT2/GSTT2B, Fmo9, IL1R1, HS6ST1, ABCG1, UST, ABCC3	3.89E-03	PPARA, TNFRSF1A, IRAK1, ALDH1L1, ALDH1A1, ALDH1L2, MGST2, NR0B2, SMOX, IL1B, NR5A2, FABP5, TNF, ALDH6A1, ACSL1, TNFRSF11B
cGMP-induced vasodilation signaling	1.78E-02	MYH3, Calm1, ACTG1, MYLK, PDE4D, CACNA1C, PRKAG1	1.32E-02	MYL9, PRKAR2B, PDE4C, PABPC4, ITPR3, PDE4A, GUCY1A2, MYL12B, GUCY1B3, ACTA1
Tight junction signaling	2.14E-02	MYH3, ACTG1, PTPA, CLDN23, MYLK, PRKAG1, CPSF6, CPSF2	2.75E-02	MYL9, CLDN10, TJP3, PRKAR2B, TNFRSF1A, CGN, VCL, TNF, ACTA1, OCLN, TNFRSF11B
Germ cell-Sertoli cell junction signaling	2.57E-02	ACTG1, JUP, GRB2, FGFR4, RHOJ, TUBA3E, RRAS, PIK3C3	6.31E-03	TUBB3, TNFRSF1A, MAP3K13, TUBB2A, TUBB2B, RHOB, TUBB6, CFL2, MAPK10, MAP3K8, VCL, TNF, ACTA1
AMPK signaling	3.31E-02	PFKP, PTPA, AK8, CHRM5, GRB2, CAMKK2, FGFR4, PRKAG1, PIK3C3	2.04E-03	TSC1, ULK1, ARID1A, ACACB, ADIPOQ, CFTR, PFKP, MAPK13, PRKAR2B, PCK2, ADRA2A, TBC1D1, PRKAA2, INSR, CAMKK2
Superpathway of methionine degradation	3.89E-02	PRMT8, AHCYL1, SUOX	8.13E-05	PCCA, CBS/CBSL, BHMT, CDO1, CTH, MTR, AHCY

Abbreviations: AMPK, adenosine monophosphate–activated protein kinase C; DMR, differentially methylated region; PPAR, peroxisome proliferator–activated receptor; tRNA, transfer RNA.

To investigate the potential long-term effects of DNA methylation changes early in life on gene expression later in life, we integrated the 2-week-old methylome and 10-week-old transcriptome data sets (7). There were more proximal genes associated with altered expression in 10-week-old IUGR islets, including many key islet genes, such as glycoprotein 2 (Gp2), collagen type V α 3 chain (Col5a3), tissue inhibitor of metalloproteinase-2 (Timp2), and MAF bZIP transcription factor (Maf) (25, 28). Among them, 25 genes were upregulated, including Mdga2, Rpia, and Sphkap, and 72 genes were downregulated, including Cpa4, Ang2, Col5a3, Timp2, Maf, and Gp2 (Supplementary Table S3) (12). This finding suggested a potential contribution of DNA methylation changes early in development to the disruption in islet development and function later in life in IUGR animals.

Intrauterine Growth Restriction Altered DNA Methylation at Critical Transcription Factor Binding Motifs

To investigate whether the DMRs identified in IUGR islets at age 2 weeks were located near transcription factor binding sites where changes in DNA methylation may facilitate or block transcription factor binding and result in dysregulation of gene expression in IUGR, we performed HOMER binding motif analysis for 511 DMRs. The top binding motifs enriched in DMRs were many well-known transcription factors critical for β-cell function, including forkhead box protein A1 (Foxa1), forkhead box protein A2 (Foxa2), androgen receptor (AR), signal transducer and activator of transcription 3 (Stat3), hepatocyte nuclear factor 1 homeobox A (Hnf1), nuclear factor κB, activating transcription factor 4 (Atf4), SRY-box transcription factor 4 (Sox4), and 2 ETS family transcription factors ETS-like 1 (Elk1) and ETS variant 1 (Etv1) (Fig. 2).

Figure 2.

Top transcription factor binding motifs enriched within differentially methylated regions (DMRs). These transcription factors are also identified as upstream regulators in transcriptome. TF, transcription factor. Activation z score is to infer the activation states of predicted transcriptional regulators. Positive values indicate activation, and negative values indicate inhibition.

Importantly, changes in DNA methylation near the transcription factor binding motif were associated with changes in the expression of 149 target genes in IUGR islets, including Cll3l3, Tnf, Vim, Fabp5, Smad3, and Spp1 (Supplementary Table S4) (12). Consistent with the HOMER binding motif analysis, the same transcription factors identified through binding motif analysis within the DMRs were also identified as critical upstream regulators modulating gene expression changes in IUGR islets from our transcriptome data set (see Fig. 2, Supplemental Table S4) (12).

Furthermore, the list of single CpGs with a 70% or greater change in DNA methylation was associated with 167 differentially expressed genes (see Supplementary Table S2) (12). In addition, the locations of multiple single CpG sites with significant changes in DNA methylation were close (< 200 bp) to transcription factor binding motifs, including ras-responsive element-binding protein 1 (Rreb1), Sp1, Sp2, Krüppel-like factor 5 (Klf5), Klf9, and nuclear receptor subfamily 5, group A, member 2 (Nr5a2, also known as liver receptor homolog-1) (Table 4). The genes with multiple significantly altered single CpG sites included several genes important for islet development and function, such as Sox9, Agtr1b, Fabp5, and Slc6a11; genes regulating neuronal growth and function, like Gap43, Mapk10, and Drgx; and genes modulating immune function, including Ccl1 and Litaf (see Table 4).

Table 4.

Differentially expressed genes with multiple significantly altered single CpG sites and associated binding motifs

	2-wk-old Transcriptome		DNA methylomea
Genes	LogFC	FDR	Single CpG No.	Binding motifsb
Adra2a	0.72	4.69E-02	2	ZNF263, RREB1, EWSR1-FLI1, ZNF384
Agtr1b	–2.00	2.88E-02	4	RREB1, ZNF263
Ano1	–0.75	3.64E-02	2	ZNF263, KLF5, KLF9
Anxa1	0.87	3.87E-02	2	ZNF263
Aqp8	–3.00	1.00E-04	3	EWSR1-FLI1, ZNF263, SP2
Arhgef28	–0.78	2.45E-02	2	EWSR1-FLI1, ZNF263, NR5A2
B4galt5	0.63	2.55E-02	2	RREB1, SP1, SP2, ZNF263, KLF9
Calcoco1	–0.51	4.42E-02	2	ZNF263
Ccl1	4.39	3.74E-02	2	ZNF263, Pparg::Rxra, RREB1, NR2C2, Nr2f6, KLF9, ZNF384
Dok5	1.67	4.36E-02	5	ZNF263, EWSR1-FLI1, KLF9
Drgx	–2.10	2.21E-02	2	REST, ZNF384, RREB1, SP1, ZNF263, KLF5, KLF9
Elavl2	0.85	4.25E-02	2	ZNF263
Eml1	0.66	3.39E-02	2	ZNF263, CTCF
Erbb4	–2.25	8.00E-04	2	ZNF740
Fabp5	0.87	4.46E-02	2	EWSR1-FLI1, SP2, ZNF263, ZNF384
Fam160a1	–1.15	5.80E-03	2	RREB1, ZNF263, KLF9, ZNF384
Fam46a	0.93	1.44E-02	2	FOXD3, ZNF263, ZNF384
Fbln1	0.79	1.50E-03	2	SP1, SP2, ZNF263, KLF5, RREB1, EWSR1-FLI1, ZNF740, KLF9
Fkbp1b	1.58	2.15E-02	2	RREB1, EWSR1-FLI1, ZNF263
Fut8	–0.91	5.00E-04	2	EWSR1-FLI1, ZNF263
Gap43	1.17	3.21E-02	2	ZNF263, RREB1, EWSR1-FLI1, KLF5
Hm13	–0.89	1.84E-02	2	RREB1, ZNF263
Id3	0.82	1.77E-02	2	RREB1, EWSR1-FLI1, ZNF263, ZNF384, KLF9
Igsf11	–0.58	2.64E-02	2	RREB1, ZNF263
Litaf	–0.77	2.19E-02	3	FOXD3, RREB1, SP2, ZNF263, KLF5,
LOC257642	–1.91	2.63E-02	2	ZNF263, ZNF384
Lrrc7	–1.96	1.00E-04	2	ZNF263
Map3k8	0.94	4.35E-02	2	SP1, EWSR1-FLI1, SP2, ZNF263,
Mapk10	1.04	8.90E-03	2	ZNF384, FOXD3
Mob3b	–1.41	1.30E-03	3	ZNF384
Nuak1	0.83	2.12E-02	2	RREB1, KLF9, ZNF384, SP1, SP2, ZNF263, KLF5
Phka1	–0.52	3.35E-02	3	RREB1, ZNF384
Pkia	0.65	2.79E-02	2	FOXD3, RREB1, KLF9, SP2, ZNF263
Pla2g4b	–0.73	1.67E-02	2	RREB1, SP2, ZNF263, KLF5
Pls3	0.61	1.15E-02	3	SP1, EWSR1-FLI1, SP2, ZNF263, KLF5, ZNF384, NR5A2
Pros1	0.64	4.79E-02	2	ZNF263, EWSR1-FLI1
Schip1	0.61	4.04E-02	4	SP1, SP2, ZNF740
Sdc4	0.65	1.44E-02	2	CTCF, EWSR1-FLI1, ZNF263
Slc6a11	–2.98	1.00E-04	3	ZNF263, KLF5, EWSR1-FLI1, KLF9
Snx18	0.49	3.30E-02	2	ZNF384
Sox9	–1.52	4.90E-02	3	KLF9, RREB1, ZNF263, SP1, KLF5
Veph1	–1.34	4.86E-02	3	ZNF263, KLF9, EWSR1-FLI1, ZNF384
Zbed3	–0.70	1.11E-02	3	RREB1, SP1, SP2, ZNF263, KLF5, KLF9

Abbreviations: FC, fold change; FDR, false discovery rate.

^a DNA methylation changes for single CpG sites have P less than .05.

^b Putative binding motifs are within the regions of single CpG ± 200 bp, and from multiple single CpG sites for the gene as indicated.

Differentially Methylated Regions and Single CpG Sites With Significant Changes in DNA Methylation May Regulate the Expression of Distal Interacting Genes

To determine if the identified DMRs represented regulatory regions for non-proximal genes, we performed in silico Hi-C analysis to assess 3D genomic interactions between conserved DMRs and other regions of the genome. For this analysis we used published human pancreas and islet Hi-C data sets (15-19). We identified 1171 potential interacting genes for 21 conserved DMRs. Integrating with the transcriptome, we identified interactions between 14 highly conserved DMRs and 35 genes with differential expression in 2-week-old IUGR islets, including Eif2ak3, Pah, Dram1, Chad, and Ccl1 (Table 5). A total of 13 genes were upregulated and 22 genes were downregulated in 2-week-old IUGR islets. Many of the genes with changes in expression at age 2 weeks were also found to have changes in gene expression at age 10 weeks, such as Muc1, Gjb1, Prss16, Dram1, and Pah (see Table 5). In 10-week-old islets, there were far more Hi-C interaction genes associated with altered expression and most of these were key islet genes, including Eif2ak3, Nup37, Igf1, Yipf6, Chrnb2, and Rab3c (Supplementary Table S5) (12).

Table 5.

Differentially methylated regions and expression data of 2-week-old islets from potential distally regulated genes based on in silico Hi-C analysis

Rat genome						Human genome liftedc			Proximal gene: RNA-Seq			Hi-C gene interaction: RNA-Seq_2wk
Chr	DMR start-end	Total CpG No.a	Diff. CpG No.b	Methyl %	Q	Chr	Start-end	UCSC regulatory region	Name	Log FC	Q	Name	Log FC	Q
1	54517159-54517367	4	2	18.6	1.68E-02	1	45205972-45206189	Strong promoter/enhancer/CpG island	Vom2r7	NA	NA	MKNK1	–1.48	7.00E-06
												MUTYH	–0.90	2.10E-02
												LURAP1	–1.60	5.38E-04
												TSPAN1	–1.02	4.24E-02
2	188712546-188713832	5	3	–22.6	2.14E-02	1	155008964-155007652	Intron 1/strong promoter/enhancer	Zbtb7b	–0.31	4.28E-01	PBXIP1	0.49	3.97E-02
												KRTCAP2	–0.84	4.04E-02
												MUC1	–2.08	4.70E-03
												TRIM46	–1.38	5.38E-04
4	99133877-99134143	6	3	16.3	2.80E-02	2	88055678-88055537	Intron 1/strong promoter/enhancer/CpG island	Krcc1	0.23	4.71E-01	EIF2AK3	–0.85	1.37E-02
6	137718907-137719738	5	2	11.6	1.97E-02	14	105150856-105151199	CpG island	Nudt14	–0.48	2.51E-01	CKB	0.79	6.81E-04
												CRIP1	0.98	3.31E-03
												CRIP2	0.66	2.32E-02
												CEP170B	–0.84	9.34E-03
												ZBTB42	–1.04	3.20E-02
7	28086897-28087393	6	2	–12.6	2.99E-02	12	102885945-102885440	Intron 3	Pah	-1.93	1.43E-03	PAH	–1.93	1.43E-03
												DRAM1	–1.42	1.41E-03
												NT5DC3	–0.87	9.12E-03
10	84306601-84308067	4	2	–21.7	2.90E-02	17	48433288-48433974	Potential enhancer	Skap1	0.56	5.98E-01	PPP1R9B	0.55	4.62E-02
												CHAD	–1.60	9.87E-03
												PRR15L	–1.41	3.20E-02
												ITGB3	0.93	4.15E-02
10	69737405-69737941	4	2	–19.9	1.78E-02	17	34580905-34581199	Exon 1/strong enhancer/ CpG island	Tmem132e	0.66	5.43E-01	CCL3	1.66	4.92E-02
												CCL1	4.39	3.74E-02
												CCL7	1.53	4.46E-02
												MMP28	1.09	2.13E-02
10	84502238-84502586	4	2	13.9	1.71E-02	17	48242776-48242430	Intron 4/strong enhancer	Skap1	0.56	5.98E-01	PPP1R9B	0.55	4.62E-02
10	65065958-65068052	7	2	–15.2	1.27E-02	17	29166304-29167027	Exon 2/CpG island	Myo18a	–0.37	1.49E-01	CORO6	–1.39	7.91E-04
11	32055157-32055478	4	2	–14.8	1.77E-02	21	33865194-33865342	Exon 31	Itsn1	0.54	4.77E-02	ITSN1	0.54	4.77E-02
15	108915755-108917499	6	3	–36.1	1.83E-02	13	99989121-99989724	CpG island	Zic2	0.98	7.23E-01	GGACT	–1.01	1.47E-02
												PCCA	–1.08	2.78E-03
20	3912891-3915417	4	3	–16.8	4.43E-02	6	32976005-32976988	Exon 5-7	Brd2	0.03	1.00E+00	DDAH2	0.55	4.21E-02
												CLIC1	0.53	2.76E-02
20	4896106-4896500	5	3	–33.7	4.43E-02	6	29887993-29888234	CpG island/weak promoter/enhancer	RT1-CE14	0.17	1.00E+00	PRSS16	–2.20	6.44E-04
									RT1-CE5	0.40	5.02E-01	TRIM40	–2.65	1.58E-02
X	71126718-71127368	4	3	37.8	1.52E-02	X	71067838-71068454	Promoter/enhancer	Snx12	0.18	5.76E-01	GJB1	–1.99	1.01E-02
												OGT	–0.63	8.10E-03

Genes shown in bold indicate changes in gene expression persisted until 10 weeks.

Abbreviations: Chr, chromosome; FC, fold change; DMR, differentially methylated region; NA, not available; RNA-seq, RNA sequencing.

^a Total number of CpG within the DMR.

^b Number of differentially methylated CpG within the DMR.

^c Corresponding location of DMR in human genome hg38 assembly.

Furthermore, we also identified 660 potential interacting genes for 134 conserved single CpG sites with significant DNA methylation changes in Hi-C analysis. A total of 24 and 85 interaction genes were differentially expressed in 2-week-old and 10-week-old IUGR islets, respectively (Supplementary Table S6) (12). Genes with expression changes in 2-week-old islets that persisted until adulthood included Pde8b, Kctd14, Mrap2, and Hpn.

DNA Methylation May Interact With Histone Modifications to Regulate Expression of Proximal and Hi-C Interaction Genes

There is a complex interplay between DNA methylation and histone modifications, as well as among different histone modifications, to maintain normal chromatin function and regulate gene expression. To investigate the potential interactions between DNA methylation and histone modifications in regulating Hi-C interacting genes, we integrated the methylome with our previously published chromatin immunoprecipitation sequencing (ChIP-seq) data sets of H3K4me3, H3K27me3, and H3K27Ac modifications from 2-week-old control and IUGR islets (10). We identified significant histone modification changes within 5 kb upstream or downstream of many conserved DMRs and single CpG sites that may act as potential regulators of distal genes (Tables 6 and 7). Multiple DMRs and single CpG sites were associated with more than one histone mark change. Alterations of multiple histone modifications near conserved DMRs and potential single CpG regulators were correlated with DNA methylation changes. We further integrated changes in histone marks with all 511 identified DMRs and single CpG sites that were associated with proximal gene expression changes. Significant histone modification alterations were identified within 5 kb of 155 DMRs and 17 single CpG sites (Supplementary Tables S7 and S8) (12). Among them, approximately 70% of the changes in histone marks were highly correlated and consistent with alterations in DNA methylation. These results suggest DNA methylation may interact with histone modifications in regulating proximal and distal gene expression in IUGR islets.

Table 6.

Histone modification changes near conserved differentially methylated regions

DMR location		ΔMethylation, %	Proximal gene	Histone mark	Distance to mid-DMR, bpa	Histone mark changes
Chr	DMR start-end					LogFC	Q
Chr2	40968165-40971954	–17.6	Pde4d	H3K4me3	2592	–0.56	1.05E-03
Chr2	188712546-188713832	–22.6	Zbtb7b	H3K27Ac	562	0.76	4.95E-05
				H3K27Ac	5112	–0.32	1.89E-03
Chr3	165021192-165021321	–21.9	Kcng1	H3K27Ac	3345	0.94	1.04E-03
Chr4	99133877-99134143	16.3	Krcc1	H3K27Ac	–859	–0.44	3.69E-04
Chr7	28086897-28087393	–12.6	Pah	H3K27Ac	–2044	–1.29	4.93E-06
Chr10	65065958-65068052	–15.2	Myo18a	H3K27Ac	5696	0.87	4.49E-05
Chr10	84502238-84502586	13.9	Skap1	H3K4me3	1189	2.77	1.61E-03
Chr20	3912891-3915417	–16.8	Brd2	H3K27Ac	–803	–0.94	1.57E-03
Chr20	4896106-4896500	–33.7	RT1-CE5	H3K27Ac	1448	–0.36	2.37E-04
				H3K27me3	1398	–0.53	4.18E-05
				H3K27me3	3348	–0.71	1.34E-04
				H3K4me3	4248	1.49	2.81E-03
				H3K27me3	–1702	–0.82	1.25E-03
				H3K27me3	–4302	–1.02	9.52E-04
				H3K4me3	–1752	2.35	2.06E-04

Numbers in bold indicate the direction of histone modification changes consistent with DNA methylation changes.

Abbreviations: Chr, chromosome; DMR, differentially methylated region; FC, fold change.

^a Distance between the midpoint of chromatin immunoprecipitation sequencing peak to midpoint of DMR. Negative value represents upstream and positive value represents downstream.

Table 7.

Histone modification changes near conserved single CpG sites

Single CpG location		ΔMethylation, %	Proximal gene	Histone mark	Distance to CpG, bpa	Histone mark changes
Chr	CpG position					LogFC	Q
Chr1	12824609	–72.12	Cited2	H3K27Ac	4392	0.82	1.29E-03
				H3K27Ac	–458	0.39	7.81E-05
				H3K27Ac	–2308	0.41	7.86E-04
Chr2	188711200	90.78	Zbtb7b	H3K27Ac	2551	0.76	4.95E-05
Chr2	205551131	–69.52	Csde1	H3K4me3	4170	3.33	7.38E-04
Chr3	3354092	–85.44	Kcnt1	H3K27me3	4259	1.54	4.13E-08
Chr4	119697358	–88.95	Rab43	H3K4me3	1243	2.38	1.62E-03
				H3K27Ac	–4307	0.77	2.78E-05
Chr7	70818805	–89.70	Ndufa4l2	H3K4me3	246	–0.38	2.44E-03
Chr8	53006201	–97.15	Zbtb16	H3K27me3	1700	0.77	1.13E-03
Chr8	57974978	–74.89	Kdelc2	H3K4me3	1023	3.07	2.71E-04
Chr8	97580672	–87.89	Tbc1d2b	H3K27Ac	–71	1.05	3.05E-04
Chr9	88085225	–72.60	Irs1	H3K4me3	3876	–0.42	6.45E-04
				H3K4me3	–124	–0.39	1.54E-04
				H3K27me3	–1924	0.75	7.36E-05
Chr10	6975258	91.88	Usp7	H3K4me3	–857	–0.31	3.39E-02
Chr13	81215769	–95.17	Prrx1	H3K27me3	–1668	–0.70	1.25E-02
Chr14	104885154	–71.70	Sertad2	H3K27Ac	–953	0.86	1.12E-04

Numbers in bold indicate the direction of histone modification changes consistent with DNA methylation changes.

Abbreviations: Chr, chromosome; FC, fold change.

^a Distance between the midpoint of chromatin immunoprecipitation sequencing peak to single CpG. Negative value represents upstream and positive value represents downstream.

Overlap of Methylome in Intrauterine Growth Restriction Rat Islets and Human Islets From Donors With Type 2 Diabetes

IUGR is strongly associated with increased risk of T2D (1, 2). Indeed, our findings in 2-week-old IUGR islets highly overlap with the results of a DNA methylation study in islets from humans with T2D (29). The pathways with enrichment of genes that had DMRs in both data sets include insulin signaling, metabolic pathways, mitogen-activated protein kinase signaling, transforming growth factor–β signaling, Jak-STAT signaling, and pathways regulating neuronal function, immune function, and cytoskeleton and extracellular matrix. Multiple genes with DNA methylation changes were also identified in islets both from IUGR rats and T2D donors, including MGLL (monoglyceride lipase), FAM19A5 (an adipocyte-derived adipokine), SRGAP3 (cytoskeleton regulation), NEBL (cytoskeleton), PDE8B (catalyzes hydrolysis of cAMP), CADPS (regulates recruitment of insulin granules and β-cell function), DISP1 (regulation of hedgehog pattern signaling), and EPB41L3 (cytoskeleton component inhibiting cell proliferation) (Fig. 3A; Supplementary Table S9) (12, 29). Furthermore, many genes with DNA methylation and expression changes in T2D islets were differentially expressed in 2-wk and 10-wk IUGR islets (Fig. 3B; see Supplementary Table S9) (7, 12, 29).

Figure 3.

Venn diagrams. A, Overlap of genes with DNA methylation changes in islets from both intrauterine growth restriction (IUGR) rat and type 2 diabetes (T2D) donors. B, Overlap between genes with DNA methylation and expression changes in T2D islets and differentially expressed genes in 2-week-old and 10-week-old IUGR transcriptomes.

Discussion

The major finding of this study is that IUGR induces genome-wide epigenetic modifications that interact and are associated with changes in expression of key genes regulating islet function. More than 500 DMRs and 4000 single CpG sites were significantly altered in 2-week-old IUGR islets in the present study that were associated with changes in histone modifications and were located in close proximity to important transcription factor binding motifs. We identified novel genomic regulatory regions in IUGR rat islets that are characterized by IUGR-induced changes in DNA methylation and histone modifications that also contain key transcription factor binding motifs. Additionally, these unique data sets allowed us to make important comparisons between early epigenetic modifications in islets and gene expression changes later in life that were associated with changes in phenotype.

Our findings in 2-week-old IUGR islets were highly correlated with the results of a DNA methylation study in human T2D islets by Volkov et al (29). Multiple differentially expressed genes with DNA methylation changes and pathways with enrichment of genes that had DMRs in T2D islets were also identified in islets from 2-week-old IUGR rats. Far more genes with DNA methylation changes in T2D islets overlapped with differentially expressed genes in 10-week-old IUGR transcriptome (7). These results support the use of the IUGR model to determine mechanisms that are altered early in life before the onset of T2D that contribute to the pathogenesis of this disease.

IPA analysis identified multiple pathways altered both in the methylome and transcriptome from 2-week-old IUGR islets, although the individual genes involved in the pathways were dissimilar in the methylome and transcriptome. Multiple studies by us and others also show limited overlap between the methylome and transcriptome (29-33). Many factors can drive the transcription in islets other than DNA methylation (34). For example, the expression of Pdx1, a homeobox transcription factor critically important for pancreatic β-cell function and development, is epigenetically regulated. However, DNA methylation regulates Pdx1 expression only in adult islets (35). We have shown that Pdx1 expression in IUGR islets at age 2 weeks was regulated by histone modifications (35, 36). Consistently, DNA methylation of the Pdx1 gene was not changed in 2-week-old IUGR islets in our present study, but H3K4me3 modification was decreased in 2-week-old IUGR islets in our previous ChIP-seq study (10).

Interestingly, many transcription factor binding motifs were enriched within the DMRs, suggesting DNA methylation changes near the transcription factor binding motif could potentially lead to a change in the expression of target genes. These transcription factors were also identified as key upstream regulators in our transcriptome study (7), including Foxa1, Foxa2, AR, Stat3, Hnf1, and Sox4. Foxa1 and Foxa2 play critical roles in islet development and differentiation, as well as in regulating insulin secretion and glucose homeostasis in mature β cells (37-39). Stat3 modulates insulin secretion, maintains normal islet architecture and microvascular network, and protects β cells from DNA damage (40-42). Hnf1a is associated with Maturity Onset of Diabetes of the Young 3 (43). It regulates expression of genes involved in insulin synthesis, glucose-stimulated insulin secretion, and β-cell differentiation (44-46). Atf4 plays an important role for β-survival during stress (47). Sox4 is critical for pancreas development, β-cell proliferation, and insulin secretion (48-50). The ETS transcription factor family, such as Elk1and Etv1, regulate many processes during development, including cell proliferation, differentiation, migration, apoptosis, and mesenchymal-epithelial interactions (51). These transcription factors also play critical roles during pancreatic development (52). Furthermore, although it is unclear whether the identified single CpG sites with significant changes in DNA methylation directly regulate proximal gene expression, putative transcription factor binding motifs, such as the critical regulator Sp1 that regulates key β-cell genes such as Pdx-1 (53), GCK (54), and Nur77 (55), and RREB1 whose variants are associated with T2D and gestational diabetes (56), were also identified within 200 bp upstream and downstream of single CpG sites. These findings suggest single CpG sites with significant DNA methylation changes may at least partially contribute to the gene expression changes.

Methylation at different genomic regions can have varied effects on gene transcriptional activity. While DNA methylation at promoter regions is relatively well studied and strongly associated with transcriptional silencing, methylation in intergenic regions and gene bodies has been less characterized and may have different functions (57). In the present study, differential methylation was highly enriched at intergenic regions (52%) in the IUGR islets. This outcome is similar to our previous findings in adult IUGR islets (9), in which approximately 65% of DMRs were located in intergenic regions, as well as to other models of early-life adverse environments (30, 33). These conserved intergenic regions may represent important cis- or trans-regulatory sites in regulating gene expression. Our data also showed that approximately 27% of DMRs in the IUGR islets were located in gene bodies. DNA methylation in gene bodies may define the exon boundaries, regulate alternative promoters in gene bodies, and regulate messenger RNA splicing and alternative splicing (58). However, our transcriptome analysis (7) was insufficient to probe such effects. Intragenic DMRs could also serve as potential gene enhancers or insulators (21, 22, 59).

Using in silico Hi-C analysis with published human pancreas and islet Hi-C data sets (15-19), we identified multiple highly conserved DMRs and single CpG sites with potential distal regulatory functions. The methylation changes at these loci were associated with differential expression of many distal genes critical for β-cell function such as Eif2ak3, Ddah2, and OGT both in 2-week-old and 10-week-old IUGR islets. Mutations in Eif2ak3 result in permanent neonatal diabetes and Wolcott-Rallison syndrome (60). Eif2ak3 is required for normal β-cell development and proliferation. Ddah2 regulated glucose-stimulated insulin secretion in the β cell in a pathway with Sirt1 and secretagonin (61). OGT is an anabolic nutrient sensor in the β cell, and acute stimulation promotes adaption to β-cell stress (62). Recent Hi-C studies of human islet T2D enhancer variants identified more than 1300 groups of islet enhancers, super-enhancers, and active promoters forming 3D hubs, and T2D genetic variation in these hubs can affect the expression of key genes regulating functions such as insulin secretion (19). Greenwald et al (17) also identified more than 3000 active enhancers in islets that formed loops to a gene promoter. They found that the promoter regions of 1000 genes have chromatin loops to multiple active enhancers, which are enriched with T2D risk variants. In addition, a significantly higher proportion of genes expressed in islets have at least one enhancer loop compared to nonislet-expressed genes, including genes regulating transport and secretion, cell communication, and carbohydrate homeostasis. Indeed, many of these genes, including Chad, Hoxb3, Cdk5r1, Pbxip1, Npr1, Chrnb2, Efna4, Cadps, and Irs1, were also identified as Hi-C interacting genes in our IUGR model that were differentially expressed in either 2-week-old or 10-week-old IUGR islets and potentially regulated by DMRs or significantly changed single CpG sites. Thus, DMRs and significantly differentially methylated single CpG sites identified in our study may represent distal regulatory regions that have the potential to modulate expression of genes critical for normal islet function.

By integrating DNA methylation and histone modifications data sets, our present study also provides novel mechanistic insights into how epigenetic modifications altered in IUGR islets contribute to changes in gene expression and hence phenotypic changes. Significant alterations in the enrichment of histone marks were found within 5 kb upstream and downstream of 155 DMRs and many differentially methylated single CpG sites, including multiple conserved DMRs and single CpG sites that may act as potential regulators of distal genes. Our results showed that about 70% of histone mark changes were highly correlated and consistent with DNA methylation changes, suggesting the interplay between DNA methylation and histone modifications is critically important in IUGR islets for regulating proximal and distal gene expression. The crosstalk between DNA methylation and histone modifications is complicated. For example, H3K4me3 might block the binding of DNA (cytosine-5)-methyltransferase 3-like (Dnmt3L) to H3K4 and prevent de novo DNA methylation (63), whereas mixed lineage leukemia 1 can recognize unmethylated CpG and methylate H3K4 (64). Thus, the histone marks and DNA methylation changes that were not correlated well at age 2 weeks may interact with each other and regulate gene expression later in life.

One limitation of our study is that only 2 IUGR samples were included in the analysis because 1 sample was identified as an outliner due to insufficient reads. Thus, the smaller-magnitude changes induced by IUGR may not be detected because of decreased statistical power, and the significance may be less representative of the general population. However, we used a more stringent criteria of Q less than .05 to identify DMRs to limit the possibility of false-negative or false-positive results. Another limitation of our study is the use of pooled samples from both sexes. To follow the previous phenotypic studies as well as to integrate the current data with our published transcriptome and histone modification data sets, we needed to use pooled samples. Since sexual dimorphism has been reported in the fetal programming and developmental origins of adult disease, the gene targets/pathways proposed in this study could be more important in regulating islets in one sex compared to the other. Despite these limitations, this study is the first genome-wide study of DNA methylation integrated with histone modifications in young rat islets. While serving as a valuable resource for elucidating chromatin landscapes, we have identified novel DMRs and single CpG sites with potential distal regulatory functions and suggest that epigenetic modifications at key transcription factor binding motifs and genes in early life may contribute to gene dysregulation and an abnormal islet phenotype that results in diabetes later in life.

Glossary

Abbreviations

3D

3-dimensional

AMPK

adenosine monophosphate–activated protein kinase C

cAMP

3′,5′-cyclic adenosine 5′-monophosphate

ChIP-Seq

chromatin immunoprecipitation sequencing

DMR

differentially methylated region

FDR

false discovery rate

H3K27Ac

histone H3 lysine K27 acetylation

H3K27me3

histone H3 lysine K27 trimethylation

H3K4me3

histone H3 lysine K4 trimethylation

HELP

HpaII tiny fragment enrichment by ligation-mediated PCR

IPA

Ingenuity Pathway Analysis

IUGR

intrauterine growth restriction

T2D

type 2 diabetes

Financial Support

This work was supported by the National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK grant Nos. R01 DK055704 and R01 DK114054 to R.A.S.). The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Author Contributions

Conceptualization, Y.C.L. and R.A.S.; methodology, Y.C.L. and S.E.P.; validation, Y.C.L.; formal analysis, X.M.L.; data curation, Y.C.L., X.M.L., and R.A.S.; writing—original draft preparation, Y.C.L.; writing—review and editing, S.E.P. and R.A.S.; supervision, R.A.S.; project administration, Y.C.L. and R.A.S.; funding acquisition, R.A.S. All authors have read and agreed to the published version of the manuscript.

Disclosures

The authors have nothing to disclose.

Data Availability

Sequence data have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE151684. All data generated or analyzed during this study are included in this article and in the data repository listed in “Reference.”