Early Pregnancy Maternal Blood DNA Methylation in Repeat Pregnancies and Change in Gestational Diabetes Mellitus Status—A Pilot Study

Abstract

Repeat pregnancies with different perinatal outcomes minimize underlying maternal genetic diversity and provide unique opportunities to investigate nongenetic risk factors and epigenetic mechanisms of pregnancy complications. We investigated gestational diabetes mellitus (GDM)-related differential DNA methylation in early pregnancy peripheral blood samples collected from women who had a change in GDM status in repeat pregnancies. Six study participants were randomly selected from among women who had 2 consecutive pregnancies, only 1 of which was complicated by GDM (case pregnancy) and the other was not (control pregnancy). Epigenome-wide DNA methylation was profiled using Illumina HumanMethylation 27 BeadChips. Differential Identification using Mixture Ensemble and false discovery rate (<10%) cutoffs were used to identify differentially methylated targets between the 2 pregnancies of each participant. Overall, 27 target sites, 17 hypomethylated (fold change [FC] range: 0.77-0.99) and 10 hypermethylated (FC range: 1.01-1.09), were differentially methylated between GDM and control pregnancies among 5 or more study participants. Novel genes were related to identified hypomethylated (such as NDUFC1, HAPLN3, HHLA3, and RHOG) or hypermethylated sites (such as SEP11, ZAR1, and DDR). Genes related to identified sites participated in cell morphology, cellular assembly, cellular organization, cellular compromise, and cell cycle. Our findings support early pregnancy peripheral blood DNA methylation differences in repeat pregnancies with change in GDM status. Similar, larger, and repeat pregnancy studies can enhance biomarker discovery and mechanistic studies of GDM.

Introduction

Repeat pregnancies with different perinatal outcomes minimize underlying maternal genetic diversity and provide unique opportunities to investigate nongenetic risk factors and epigenetic mechanisms of pregnancy complications, such as investigations of DNA methylation, an epigenetic regulatory mechanism that influences gene expression and gene function.¹ Variations in DNA methylation have been associated with the pathogenesis of metabolic diseases, including type 2 diabetes (T2D).^2,3 Gestational diabetes mellitus (GDM), abnormal glucose metabolism first diagnosed during pregnancy, is a complex disorder that shares risk factors and pathophysiological characteristics with T2D.^4,5 Therefore, DNA methylation may play important roles in GDM risk development. We and others have previously shown that pathophysiologic changes occur early in pregnancies that are later complicated by GDM and/or other related disorders.^6,7 These pathophysiologic changes include systemic changes (including chronic systemic inflammation and dyslipidemia) that are reflected in peripheral blood.^6,7 Peripheral blood DNA methylation has also been targeted in diabetes-related research.^3,8 Hence, early pregnancy maternal peripheral blood DNA methylation profiling holds the promise of enhancing our understanding of GDM pathogenesis and risk prediction. Further, since epigenetic changes can play causal role in disease pathogenesis and are potentially reversible (eg, through dietary and behavioral intervention), it may help identify preventative or therapeutic targets.⁹

Therefore, we conducted epigenome-wide methylation profiling study of early pregnancy (16 weeks of gestation, on average) peripheral blood collected from participants with repeat pregnancies, first and second completed pregnancies (P0 and P1), only one of which is complicated by GDM, selected from a large prospective pregnancy cohort.

Methods

Study Setting

The study was conducted among participants of the Omega study.⁶ The Omega study was designed to investigate risk factors for pregnancy complications. Participants were recruited among attendants of prenatal clinics affiliated with Swedish Medical Center (Seattle, Washington) and Tacoma General Hospital (Tacoma, Washington). Omega study participants were pregnant women who initiated prenatal care before 20 weeks of gestation. Eligible participants were >18 years old, were able to speak and read English, plan to carry the pregnancy to term, and planned to deliver at either of the 2 research hospitals. All study participants provided informed consent. Study protocols were approved by institutional review boards of the participating institutions (IRB Number #2505).

Data Collection

Data were collected using in-person interviews, shortly after enrollment. Information was collected on sociodemographic characteristics (eg, age, race/ethnicity, and education), height, weight, physical activity, and family history of chronic hypertension or diabetes. In addition, early pregnancy maternal blood specimen (16 weeks, on average) was collected around the time of interview from participants. We used last menstrual period-based dating (confirmed by early pregnancy ultrasound-based dating) to estimate date of conception and gestational age. At the end of pregnancy, medical records were abstracted to obtain information on course and outcomes of the pregnancy. Study participants underwent a screening test, a 50-g 1-hour oral glucose challenge test, at 24 to 28 weeks of gestation. For participants who failed this test (blood glucose ≥140 mg/dL), a follow-up test consisting of a 100-g, 3-hour oral glucose tolerance test (OGTT) was administered. Based on the then current American Diabetes Association (ADA) recommendations, women were diagnosed with GDM if 2 or more results of the OGTT exceeded the ADA criteria as follows: fasting ≥95 mg/dL, 1-hour ≥180 mg/dL, 2-hour ≥ 155 mg/dL, and 3-hour ≥140 mg/dL.¹⁰

Study Participant and GDM/Control Pregnancy Selection

Study participants were selected from among women (N = 173) who had 2 consecutive pregnancies included in the Omega study and provided early pregnancy peripheral blood samples (in both pregnancies) that were stored in the Omega Study Biorepository. Among these participants with repeat pregnancies, we identified those who experienced GDM at either of the pregnancies. For the current study, 6 women were randomly selected among those who were initially nulliparous and had 2 pregnancies, only one of which was complicated by GDM (case pregnancy) and the other was not (control pregnancy), in such a way that 3 of the women experienced GDM in their first completed pregnancy (P0), while the other 3 experienced GDM in their second completed pregnancy (P1).

Sample Preparation and Epigenome-Wide Methylation Profiling

Maternal peripheral blood buffy coat specimens were prepared from whole blood collected in early pregnancy. DNA, for methylation profiling, was extracted from maternal buffy coat samples using the Gentra PureGene Cell kit for DNA preparations (Qiagen, Valencia, California). Samples were processed for array analyses using the Infinium methylation assay, per manufacturer’s protocol (Illumina, San Diego, California). Briefly, 500 ng genomic DNA was bisulfite treated using the EZ DNA Methylation Kit (Zymo Research Corporation, Irvine, California), chemically denatured and neutralized using 0.1 N NaOH, and amplified at 37°C for 20 to 24 hours. The amplified products were then enzymatically fragmented. Fragmented DNA was precipitated using 100% 2-propanol at 4°C for 30 minutes and centrifuged (3000g) at 4°C for 20 minutes to collect a tight pellet. The supernatant was decanted by quick inversion, and pellets were allowed to dry at room temperature (22°C) for 1 hour. The pellets were resuspended in 46 µL Illumina’s custom hybridization buffer and incubated at 48°C for 1 hour followed by 95°C for 20 minutes. Each sample of 15 µL was then loaded on to the Illumina HumanMethylation 27 BeadChips (Illumina), a platform for epigenome-wide DNA methylation profiling, and hybridized at 48°C for 24 hours. The platform contains probes to assess methylation of >27 000 CpG sites located within proximal promoter regions of transcription start sites (TSSs) of 14 475 consensus coding sequences in the National Center for Biotechnology Information Database (Genome Build 36). BeadChips were scanned using the Illumina iScan Reader, and image data were then transferred to Illumina GenomeStudio for data procession, validation of assay controls, and report generation using the methylation module. The level of methylation for each site (interrogated locus) was calculated as the ratio of the methylated fluorescence signal to the total signal of the locus. All samples were processed and hybridized at the same time.

Quality Control and Data Preprocessing

Within each array, we used sample-independent and sample-dependent controls to evaluate the quality of the assay and samples, per the Infinium HD Assay Methylation Protocol (Illimina). Briefly, sample-independent controls were used to evaluate the quality of specific steps (staining, extension, target removal, and hybridization) in the process flow. On the other hand, sample-dependent controls were used to evaluate assay performance across samples and included bisulfite-conversion, specificity, nonpolymorphic, and negative controls. All samples and assay controls met acceptable inclusion criteria. Probes (N = 17) that had missing data for at least 1 of the samples were excluded.

Statistical Analysis

Participant characteristics were described using mean (standard deviation) and number (percentage, %) for continuous and categorical variables, respectively. Data were normalized by global LOESS smoothing using the R package “limma.”¹¹ The underlying analyses approach was based on within-person comparisons of methylation in the pregnancy that was complicated by GDM (case pregnancy) with methylation in the pregnancy that was not complicated by GDM (control pregnancy), among each study participant. The difference in percentage of methylation at each target site (represented by the probe) between each participant’s case and control pregnancy was calculated. Differences in methylation were standardized into Z scores calculated from individual means and standard deviations for each participant. Differential methylation was then determined by using an R-package of Differential Identification using Mixture Ensemble (DIME).^12,13 Briefly, DIME considers an ensemble of finite mixture models and chooses the best model to determine differential methylation with a local false discovery rate (FDR). A FDR <0.10 was used as a threshold to identify differentially methylated targets for each paired case versus control pregnancies for each participant.

In comparisons of GDM case versus control pregnancies, target sites that were significantly (based on the FDR <0.10 criteria) hypo- or hypermethylated in the same direction among 5 or more participants (of the total 6) constituted our set of GDM-related differentially methylated sites. Genes that are related to these differentially methylated sites by proximity to the TSSs were evaluated for their function and functional relationships using the Ingenuity Pathway Analysis (IPA) software (Ingenuity Systems, Redwood City, California). In these analyses, biological networks that were overrepresented by genes in our set were identified using network score based on a modified Fisher exact test. Analyses were conducted using the R statistical package and IPA software.

Results

Characteristics of study participants are presented in Table 1. A total of 27 target sites were significantly (using the FDR <0.10 cutoff) hypomethylated (N = 17) or hypermethylated (N = 10) when comparing GDM-affected pregnancies to control pregnancies among 5 or more of the 6 repeat pregnancies evaluated in this study (Table 2).

Table 1.

Selected Characteristic of Study Participants.

	NH-White	HS	FHD	FHH	PA		Age		BMI, kg/m2		GA, weeks		GDM
					P0	P1	P0	P1	P0	P1	P0	P1	P0	P1
Participant 1	Yes	Yes	No	No	Yes	Yes	33	36	23.2	26.4	16.0	15.6	No	Yes
Participant 2	Yes	Yes	No	No	Yes	No	36	39	23.5	26.6	18.9	18.6	No	Yes
Participant 3	Yes	Yes	No	Yes	Yes	Yes	36	39	27.0	28.7	14.9	13.3	No	Yes
Participant 4	Yes	Yes	No	Yes	No	Yes	33	34	21.0	21.6	16.9	17.6	Yes	No
Participant 5	Yes	Yes	No	Yes	Yes	Yes	32	35	21.0	23.3	12.7	15.7	Yes	No
Participant 6	Yes	Yes	No	Yes	Yes	Yes	38	40	20.7	21.1	14.3	16.7	Yes	No

Abbreviations: NH-white, non-Hispanic white; HS, posthigh school; PA, physical activity; FHH, family history of hypertension; FHD, family history of diabetes; P0, first completed pregnancy; P1, second completed pregnancy; BMI, body mass index; GA, gestational age at blood collection; GDM, gestational diabetes mellitus pregnancy.

Table 2.

Differentially Methylated Target Sites (N = 27) and Related Genes.

Probe IDa	Gene Symbol	Gene Accession ID	Distance to TSS	Gene Name	Mean Methylation		Fold Changeb	Hypo-/Hypermethylated in GDM
					GDM	No GDM
cg08348496	HAPLN3	NM_178232.2	(−) 99	Hyaluronan and proteoglycan link protein 3	0.229	0.299	0.77	Hypo
cg09735905	HHLA3	NM_007071.1	(+) 315	HERV-H LTR-associating 3	0.173	0.224	0.77	Hypo
cg17732521	RHOG	NM_001665.2	(−) 863	ras homolog family member G	0.361	0.467	0.77	Hypo
cg12504957	MDM2	NM_002392.2	(+) 80	MDM2 oncogene, E3 ubiquitin protein ligase	0.268	0.319	0.84	Hypo
cg13456653	DNAJB6	NM_005494.2	(+) 355	DnaJ (Hsp40) homolog, subfamily B, member 6	0.398	0.454	0.88	Hypo
cg23512958	IL7	NM_000880.2	(−) 18	Interleukin 7	0.210	0.237	0.89	Hypo
cg01442799	YAP1	NM_006106.2	(+) 281	Yes-associated protein 1	0.408	0.451	0.90	Hypo
cg19764407	CDKN2B	NM_004936.3	(−) 420	Cyclin-dependent kinase inhibitor 2B	0.415	0.459	0.90	Hypo
cg27187881	NAGA	NM_000262.1	(−) 428	N-Acetylgalactosaminidase, α-	0.422	0.469	0.90	Hypo
cg16491909	C1orf43	NM_015449.1	(−) 635	chromosome 1 open reading frame 43	0.351	0.376	0.94	Hypo
Cg15010390	NDUFC1c	NM_002494.2	(−) 0	NADH dehydrogenase (ubiquinone) 1, subcomplex unknown	0.233	0.263	0.89	Hypo
Cg22141781	NDUFA12	NM_018838.3	(−) 193	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 12	0.541	0.587	0.92	Hypo
cg21615127	TMCO4	NM_181719.2	(−) 465	Transmembrane and coiled-coil domains 4	0.294	0.319	0.92	Hypo
cg21669679	DCC	NM_005215.1	(+) 610	Deleted in colorectal carcinoma	0.460	0.485	0.95	Hypo
cg05250458	ZNF177	NM_003451.1	(+) 131	Zinc finger protein 177	0.741	0.749	0.99	Hypo
cg09018040	VCX	NM_013452.2	(+) 50	Variable charge, X-linked	0.987	0.991	0.99	Hypo
cg12572827	PDE6H	NM_006205.1	(+) 469	Phosphodiesterase 6H, cGMP-specific, cone, γ	0.982	0.982	0.99	Hypo
cg04176254	AGRP	NM_001138.1	(−) 287	Agouti-related protein	0.953	0.948	1.01	Hyper
cg14308452	MGC24975	NM_153359.1	(−) 426	Proline rich, 22	0.956	0.944	1.01	Hyper
cg25953146	MGC50273	NM_214461.1	(−) 295	Chromosome 2 open reading frame 27B	0.817	0.810	1.01	Hyper
cg02500392	PBOV1	NM_021635.1	(−) 256	Prostate and breast cancer overexpressed 1	0.950	0.935	1.02	Hyper
cg21168884	C6orf122	NM_207502.1	(−) 50	Long intergenic nonprotein coding RNA 242	0.942	0.923	1.02	Hyper
cg00229387	TJAP1	NM_080604.1	(+) 631	Tight junction-associated protein 1	0.666	0.648	1.03	Hyper
cg25226891	ZMYM3	NM_005096.2	(−) 251	Zinc finger, MYM-type 3	0.591	0.569	1.04	Hyper
cg18342279	ZAR1	NM_175619.1	(+) 35	Zygote arrest 1	0.151	0.139	1.09	Hyper
cg11977634	DDR1	NM_001954.3	(+) 838	Discoidin domain receptor tyrosine kinase 1	0.157	0.149	1.05	Hyper
cg00899086	SEPT11	NM_018243.2	(+) 399	Septin 11	0.202	0.186	1.09	Hyper

Abbreviations: cGMP, cyclic guanosine monophosphate; DIME, Differential Identification using Mixture Ensemble; GDM, gestational diabetes mellitus; HERV-H LTR, human endogenous retrovirus-H long terminal repeat; Hsp40, heat shock protein 40; MDM2, murine double minute; NADH, nicotinamide adenine dinucleotide; TSS, transcription start site and strand (−/+).

^aIllumina probe ID.

^bFold changes comparing average percentage of methylation in all GDM pregnancies with average percentage of methylation in all control pregnancies.

^cAll 6 comparisons were significant (based on the DIME-based false discovery rate <10% criteria) for the NDUFC1 gene while for the other 26 genes, 5 of the 6 comparisons were significant. For all 27 genes, the direction (hyper- or hypomethylation) of differential methylation in GDM pregnancies was similar for all participants.

Fold change comparing average percentage of methylation in GDM pregnancies with average percentage of methylation in control pregnancies at target sites ranged from 0.77 (hypomethylated) to 1.09 (hypermethylated). A target site associated with the NDUFC1 gene was hypomethylated (11% less methylated, on average) in GDM pregnancies, relative to the respective control pregnancy, among all 6 study participants. Methylation in target sites related to SEPT11, ZAR1, and DDR genes was 5% to 9% higher in GDM pregnancies, while methylation in target sites related to HAPLN3, HHLA3, and RHOG genes was 23% lower in GDM pregnancies, compared with control pregnancies. Several of these differentially methylated genes, including NDUFC1 and SEPT11, are novel.

Function and functional relationship-based network analyses of genes represented by differentially methylated target sites identified 2 biological networks that were significantly overrepresented, network 1 with a score (negative log of P value of Fisher exact test) of 26 and network 2 with a score of 19 (Table 3). These networks of genes had functions that involve cell morphology, cellular assembly, cellular organization, cellular compromise, and cell cycle.

Table 3.

Networks^a Overrepresented by Genes Related to Differentially Methylated Sites.

Genes in Network	Score	Focus Genes	Functions
AGRB, ALKBH1, C1orf43, CAMK1, CDKN2AIP, DNAJB6, DOCK4, DUSP5, EID2, GLTSCR2, GRB2, HAPLN3, HCST, HTT, MC1R, MELK, NAGA, NDUFA12, NDUFC1, POLR1D, PTEN, SEPT5, SEPT6, SEPT7, SEPT8, SEPT9, SEPT11, SMAD4, TJAP1, TM4SF1, UBC, VCX3A (includes others), VGF, WDR45, ZMYM3	26	11	Cell morphology, cellular comprise, cellular assembly, and organization
Akt, AMOTL1, CCR9, CDK4/6, CDKN2B, DCC, Dcc dimer, DDR1, DLC1, DSCAM, ERK1/2, GNL2, GNL3L, Ga12/13, HCST, Iglv2, IL7, Jnk, MCF2, MDM2, NDUFA12, Nfat (family), NFkB (complex), P38 MAPK, PAK6, PAK1IP1, PDE6H, PI3K (complex), PLAC8, Ras homolog, RHOG, RpI5, SESN2, YAP1	19	10	Cell cycle, cell morphology, cellular assembly, and organization

^aThe networks were generated through the use of Ingenuity Pathways Analysis (Ingenuity Systems, www.ingenuity.com). Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base (IPKB). These genes were overlaid onto a global molecular network developed from information contained in the IPKB. Network enrichment is then assessed using a network score (negative log of P values of Fisher tests). Focus genes (in bold) are genes identified in our list of differentially methylated genes. Networks shown here are those with network scores >3.0.

Discussion

In this proof-of-concept pilot DNA methylation profiling study among women with repeat pregnancies (P0-P1 pregnancies), we identified differential methylation (both hypo- and hypermethylation) of several target sites in relation to change in GDM status. A number of novel genes were associated with identified hypomethylated (such as NDUFC1, HAPLN3, HHLA3, and RHOG) or hypermethylated sites (such as SEP11, ZAR1, and DDR). Genes related to these differentially methylated sites play critical roles in cellular morphology, organization, assembly, or compromise, as well as, cell cycle.

To our knowledge, no prior study investigated maternal peripheral blood DNA methylation during pregnancy in relation to GDM status. However, investigators have evaluated peripheral blood DNA methylation in men and nonpregnant women with T2D as well as GDM-related DNA methylation in other tissues.^3,14–20 Available evidence indicates that DNA methylation is involved in maintaining gene expression patterns associated with insulin resistance in T2D, a condition that is similar to GDM.^14,16 Several genes involved in glucose metabolism (including GLUT4) have been shown to exhibit differential DNA methylation in their promoter regions.^14,15 A possible general defect in DNA methylation in diabetes is suggested by the recent observation that S-adenosylmethionine, the main physiologic donor of methyl groups, is decreased in erythrocytes of diabetic patients, and, that the decrease is associated with disease progression.^14,16 In a recent peripheral blood based epigenome-wide investigation, a decrease in DNA methylation at a CpG site in the first intron of the FTO gene was associated with incident T2D, indicating the utility of peripheral blood-based DNA methylation investigations in diabetes.³ Further, investigations that examine methylation changes as direct causes of abnormal glucose tolerance (ie, beyond associations) are needed to clarify causal relationships.

Prior pregnancy-related glucose metabolism and DNA methylation studies investigated differential methylation of sites in candidate genes in placental or offspring (cord blood) tissues.^17,18 Bouchard et al reported that placental hypomethylation in the promoter region of ADIPOQ gene was associated with higher insulin resistance index (homeostasis model assessment [HOMA] of insulin resistance) during the second and third trimesters of pregnancy (Spearman correlation coefficients ≤−.27, P value <.05) and higher maternal circulating adiponectin levels throughout pregnancy (Spearman correlation coefficients ≤−.27, P value <.05).¹⁸ Similarly, statistically significant associations between placental DNA methylation in the LEPTIN gene and maternal glucose levels were reported by the same group of investigators (Spearman correlation coefficient = .53, P value = .009).¹⁹ Recently, El Hajj et al investigated DNA methylation in both placental and cord blood tissues and reported hypomethylation of maternally imprinted MEST gene (4%-7%), the nonimprinted NR3C1 gene (1%-2%), and interspersed ALU repeats (1%) among women who had insulin treated and diet-controlled GDM cases compared with women who had normoglycemic pregnancies.¹⁷

Investigators have also used study designs that involved paired samples, with similar genetic make-up, to investigate DNA methylation and glucose metabolism.²⁰ Using samples collected from 84 monozygotic twin pairs, Zhao et al investigated global DNA methylation in peripheral blood leukocytes and its association with insulin resistance (HOMA).²⁰ A 10% increase in the difference in mean Alu methylation was associated with an increase of 4.54 units (0.34-8.71; P value = .036) in the difference in HOMA.²⁰ Such pair-based investigations, including our current study, can help highlight the important role of epigenetics, and more specifically DNA methylation, in explaining mechanisms through which nongenetic (eg, behavioral and environmental) risk factors may increase disease incidence and/or severity.

In our study, we identified differential methylation of several GDM-associated sites that are related to novel genes. One of these novel genes, related to a hypomethylated site, is the NDUFC1 gene that encodes a protein constituting the first enzyme complex in the electron transport chain located in the inner mitochondrial membrane.²¹ Another is the NDUFA12 gene that encodes a protein which is part of the mitochondrial complex 1, part of the oxidative phosphorylation system in mitochondria.²² Mitochondrial function in tissues (including liver, muscle, adipose tissue, and pancreatic beta cells) is critical in cellular metabolism and maintenance of adaptive responses that balance oxidative activity and nutrient load.^23,24 The imbalance that follows failure of complete oxidation, due to suboptimal mitochondrial oxidative activity, leads to the accumulation of lipid intermediates, incomplete fatty acid oxidation products, and ROS, which may induce both insulin resistance and altered secretion.²³ Another novel gene, related to a hypermethylated site, is the SEPT11 gene encoding a protein that belongs to the conserved septin family of filament-forming cytoskeletal GTPases that are involved in a variety of cellular functions including cytokinesis and vesicle trafficking.^25–27 Septins have been associated with a diverse set of disease conditions including neoplasia, neurodegenerative diseases, infections, and exocytosis.^25–27 Along with NDUFC1, NDUFA12, and SEPT11, other genes identified in our study, such as the AGRB, HAPLN3, NAGA, and VCX31, were members of a gene set with functions related to cellular morphology, organization, assembly, or compromise, as well as, cell cycle. However, like many of the other novel genes identified in the current study, their role in glucose metabolism is currently unknown and need further investigations.

Our preliminary study addresses a significant gap in the literature on GDM-related DNA methylation. In addition, it highlights several opportunities of this research area. DNA-based investigations (such as DNA methylation profiling), due to stability of the DNA, can produce more reproducible information for research or clinical diagnostics purposes, compared with other markers (eg, gene expression).²⁸ Early pregnancy DNA methylation profiling before 26 to 28 weeks gestation (when standard glucose challenge tests are administered) can have significant preventive or early diagnostic clinical applications. For instance, early pregnancy interventions (eg, diet or lifestyle modifications) among high-risk populations identified by a specific methylation profile may help prevent GDM later in pregnancy. Metabolic consequences of behavioral and environmental risk factors (eg, high caloric diets and heavy metal exposure) on cellular/tissue functions may become “locked” by DNA methylation¹⁴ indicating the opportunity to use these investigations to understand gene–environment interactions. As described earlier, the design of our study, which employs repeat pregnancies (P0-P1 pregnancies), facilitates identification of these environmentally induced epigenetic (ie, methylation) changes.

Several limitations of our pilot study deserve mention. First, our pilot study was small in size, and power calculations were not for this pilot study. The results reported in this article can now be used to estimate sample size and power for future studies. DNA methylation profile of peripheral blood leukocytes may not be representative of target tissue (eg, liver and pancreatic β cell) DNA methylation profiles. However, accumulating evidence, including our prior work,^6,7 supports the use of peripheral blood as an easily accessible useful tissue, particularly for systemic conditions like GDM. We did not adjust for possible differences in cell proportion of samples. This is particularly important, given our observation that differences in GDM-related DNA methylation are associated with cell morphology, cellular compromise, and cell cycle, which are closely associated with type and number of cells that constitute peripheral blood. Advanced procedures being developed in other areas of research (eg, transcriptomics) that help determine cell content of samples will mitigate this concern in future studies.^29,30 Finally, the generalizability of findings needs to be enhanced by conducting similar experiments in more diverse study populations.

In this novel DNA methylation profiling study among women with repeat pregnancies and a change in GDM status, we identified several differentially methylated sites. These sites were related to novel genes participating in mitochondrial function, cell morphology, cellular compromise, and cell cycle. Future replication studies, particularly those designed to investigate genetic variations related to the genes, and studies of gene expression and downstream posttranscription regulation of represented genes are warranted. Similar studies of GDM cases identified using recently updated diagnostic criteria as well as other related conditions (such as impaired glucose tolerance) are important in this area. Further, investigations of the postpartum epigenetic profile of women who experienced GDM may help elucidate mechanisms underlying well-demonstrated relationships between GDM and subsequent T2D. A better understanding of epigenetic mechanisms that may account for gene–environment interactions in GDM pathogenesis will improve our efforts for the diagnosis, treatment, and possibly even prevention of a common pregnancy complication with wide-reaching implications on maternal and offspring health.