TY - JOUR
AU - Michels, Karin, B
AB - Abstract Despite the many advances made in the diagnosis and management of preeclampsia, this syndrome remains a leading cause of maternal mortality and life-long morbidity, as well as adverse fetal outcomes. Successful prediction and therapeutic intervention require an improved understanding of the molecular mechanisms, which underlie preeclampsia pathophysiology. We have used an integrated approach to discover placental genetic and epigenetic markers of preeclampsia and validated our findings in an independent cohort of women. We observed the microRNA, MIR138, to be upregulated in singleton preeclamptic placentas; however, this appears to be a female infant sex-specific effect. We did not identify any significant differentially methylated positions (DMPs) in singleton pregnancies, indicating that DNA methylation changes in mild forms of the disease are likely limited. However, we identified infant sex-specific preeclampsia-associated differentially methylated regions among singletons. Disease-associated DMPs were more obvious in a limited sampling of twin pregnancies. Interestingly, 2 out of the 10 most significant changes in methylation over larger regions overlap between singletons and twins and correspond to NAPRT1 and ZNF417. Introduction Preeclampsia, a hypertensive disorder, which occurs de novo during pregnancy, is a major contributor to maternal and fetal morbidity and mortality (1). Its clinical manifestations are often multisystemic and may include renal, liver and cerebral involvement. Despite a relatively high incidence of 3–8% of all pregnancies, the pathophysiology of preeclampsia remains insufficiently understood. Beyond the immediate risks posed to the mother and the infant, the risk for development of cardiovascular and metabolic disease in later life is increased for both the mother and her offspring, consistent with the Developmental Origins of Health and Disease (DOHaD) hypothesis (2). Hence, preeclampsia presents a substantial short- and long-term burden of disease. However, the heterogeneous nature of this syndrome impedes the discovery of sensitive and specific biomarkers that can be used in the clinic for its prediction and early detection. In addition, because the disease mechanisms remain ill-defined, continued investigation into molecular markers associated with preeclampsia is needed. Preeclampsia is a disease of abnormal placentation. It occurs as a consequence of the inability of the extravillous trophoblasts of fetal origin to fully invade the maternal decidua and remodel its spiral arteries. Remodeling of the spiral arteries is normally finalized by the beginning of the second trimester and achieves the low-resistance, high-flow blood vessels required to support fetal growth to term. Failure of this process results in decreased maternal blood flow to the intervillous space, thus reducing the exchange of gases and nutrients (3). This, likely, explains some of the early molecular hallmarks of preeclampsia reminiscent of ischemic injury, such as increased placental xanthine oxidase/dehydrogenase system expression and activity, increased generation of reactive oxygen species, impaired antioxidant mechanisms and increased apoptosis. Multiple molecular pathways have been implicated in preeclampsia including those related to oxidative and endoplasmic reticulum stress, endothelial dysfunction, intravascular inflammation, angiogenic imbalance and platelet and thrombin activation. These pathogenic mechanisms are reviewed by Chaiworapongsa et al. (4) Epidemiologic studies suggest an association between fetal sex and pregnancy complications, such as preeclampsia (5–7). There is also accumulating evidence that the human placenta is sexually dimorphic, underscoring the importance of considering infant sex in the design and analysis of association studies (8). Maternal serum levels of placental biomarkers in the first trimester of pregnancy depend on fetal sex (9). In addition, infant sex-specific modulation of placental gene expression at term has now been described (10). Whole genome DNA methylation analysis of the healthy human placenta has also revealed sex-specific differentially methylated regions (DMRs), not all of which are located on the X-chromosome (11). In the context of preeclampsia, infant sex-specific disease-associated changes of DNA methylation at individual CpGs in placental tissue have been previously reported (12). The importance of epigenetic modifications to trophoblast differentiation and invasion, and placental functioning in vivo, has been conclusively demonstrated in animal models. Disruption of either the de novo DNA methylation machinery (Dnmt3a/3b) (13) or the histone lysine demethylase—Lsd1 (14) affects trophoblast differentiation, adhesion, migration and invasive potential. Recently, a comprehensive DNA methylation profile of the full-term healthy human placenta has become available, revealing several unique features usually seen to date in cancers, including global hypomethylation and the presence of long stretches of DNA with low-to-intermediate levels of methylation, the so-called partially methylated domains (15). DNA methylation is essential for the normal function of the human extraembryonic tissue, and its physiological signatures in the developing placenta, established in response to increasing gestational age and environmental influences (16), can be disrupted under pathological conditions and exposures (17). Such disease/exposure-associated epigenetic changes have been reported by many groups, including ours, for gestational diabetes mellitus (18), preeclampsia, phthalate and phenol exposure (19) and maternal smoking (20), among others. However, dissecting the functional significance of the differences between epigenetic signatures in the healthy human placenta and pregnancy-associated pathologies remains challenging. An integrated analysis of transcriptomic and epigenomic markers is required to best capture the complexity of syndromes, such as preeclampsia. A number of large studies have confirmed both maternal and fetal contributions to preeclampsia (21–23). However, it is the placenta that is required for the disease to develop, not the fetus itself, as evidenced by the occurrence of preeclampsia in women with molar pregnancies (24). Ultimately, the clinical manifestations of preeclampsia, such as maternal hypertension, constitute an adaptive response to ischemia occurring at the materno-fetal interface. The decidua basalis or maternal side of the placenta, thus, should carry both gene expression and DNA methylation signatures reflective of this process. In order to discover placental transcriptomic and epigenomic markers associated with preeclampsia, we used samples from women enrolled in the Harvard Epigenetic Birth Cohort (HEBC) (25). We selected tissues harvested from the maternal side of the placenta directly opposite the site of umbilical cord insertion and designed a case–control study, nested within the HEBC, wherein we matched on criteria that are known risk factors for preeclampsia, including maternal age and pre-pregnancy body mass index (BMI). DMRs were validated in samples from women enrolled in the Barwon Infant Study (BIS) (26). Results Expression changes associated with preeclampsia in singletons For transcriptomic analysis, we used the Affymetrix Human Transcriptome Array (HTA) 2.0, a microarray with >6 million distinct probes covering both protein coding and non-coding transcripts (27). This microarray gives information on gene- and exon-level expression, while also providing a comprehensive view of transcript isoforms generated as a result of alternative splicing. Only RNA samples of sufficient quality (RIN ≥ 6) were included on the microarray for a final sample size of n = 42 placenta tissues, from n = 38 distinct mothers (n = 34 singleton and n = 4 twin pregnancies). Subject characteristics are summarized in Table 1. Cases were mothers who either had chart reported preeclampsia or had chart reported pregnancy-induced hypertension and proteinuria diagnosed by a physician. Each case was matched with a control by age (within 5 years), mode of conception, ethnicity, smoking status (all non-smokers), infant sex and pre-pregnancy BMI. When we compare preeclampsia cases and controls among singleton pregnancies, in addition to the aforementioned characteristics, we also match for the plurality of the pregnancy. It should be noted that the average gestational age at the time of delivery, approximately 38 weeks, indicates the presence of mild, late-onset, disease. Table 1 Patient characteristics of the samples included on the HTA 2.0 expression array Characteristic . Cases (n = 19) . Controls (n = 19) . P-value . Pre-pregnancy BMI (kg/m2) 28.38 (8.89) 26.65 (6.57) P = 0.4982 Maternal age (years) 32.05 (6.91) 31.68 (6.01) P = 0.8618 Gravidity P = 0.8822  0 7 (36.84%) 8 (42.11%)  1 5 (26.32%) 3 (15.79%)  2 5 (26.32%) 5 (26.32%)  ≥3 2 (10.53%) 3 (15.79%) Parity P = 1  0 10 (52.63%) 9 (47.37%)  1 4 (21.05%) 5 (26.32%)  2 5 (26.32%) 5 (26.32%) Smoke during pregnancy P = 1  No 19 (100%) 19 (100%) Infant sexa P = 1  Males 10 (47.62%) 10 (47.62%)  Females 11 (52.38%) 11 (52.38%) Ethnicity P = 0.8126  Non-hispanic Caucasian 11 (57.89%) 12 (63.16%)  Hispanic or Latino 3 (15.79%) 4 (21.05%)  Black/African-American 5 (26.32%) 3 (15.79%) Conception P = 1  Spontaneous planned 11 (57.89%) 12 (63.16%)  Spontaneous unplanned 7 (36.84%) 6 (31.58%)  IVF 1 (5.26%) 1 (5.26%) Birth weight (g) 3268.29 (500.6) 3387.09 (549.47) P = 0.4682 Gestational age (weeks) 38.13 (2.22) 38.94 (1.43) P = 0.1778 Characteristic . Cases (n = 19) . Controls (n = 19) . P-value . Pre-pregnancy BMI (kg/m2) 28.38 (8.89) 26.65 (6.57) P = 0.4982 Maternal age (years) 32.05 (6.91) 31.68 (6.01) P = 0.8618 Gravidity P = 0.8822  0 7 (36.84%) 8 (42.11%)  1 5 (26.32%) 3 (15.79%)  2 5 (26.32%) 5 (26.32%)  ≥3 2 (10.53%) 3 (15.79%) Parity P = 1  0 10 (52.63%) 9 (47.37%)  1 4 (21.05%) 5 (26.32%)  2 5 (26.32%) 5 (26.32%) Smoke during pregnancy P = 1  No 19 (100%) 19 (100%) Infant sexa P = 1  Males 10 (47.62%) 10 (47.62%)  Females 11 (52.38%) 11 (52.38%) Ethnicity P = 0.8126  Non-hispanic Caucasian 11 (57.89%) 12 (63.16%)  Hispanic or Latino 3 (15.79%) 4 (21.05%)  Black/African-American 5 (26.32%) 3 (15.79%) Conception P = 1  Spontaneous planned 11 (57.89%) 12 (63.16%)  Spontaneous unplanned 7 (36.84%) 6 (31.58%)  IVF 1 (5.26%) 1 (5.26%) Birth weight (g) 3268.29 (500.6) 3387.09 (549.47) P = 0.4682 Gestational age (weeks) 38.13 (2.22) 38.94 (1.43) P = 0.1778 aTotal infants from n = 34 singleton and n = 4 twin pregnancies. Open in new tab Table 1 Patient characteristics of the samples included on the HTA 2.0 expression array Characteristic . Cases (n = 19) . Controls (n = 19) . P-value . Pre-pregnancy BMI (kg/m2) 28.38 (8.89) 26.65 (6.57) P = 0.4982 Maternal age (years) 32.05 (6.91) 31.68 (6.01) P = 0.8618 Gravidity P = 0.8822  0 7 (36.84%) 8 (42.11%)  1 5 (26.32%) 3 (15.79%)  2 5 (26.32%) 5 (26.32%)  ≥3 2 (10.53%) 3 (15.79%) Parity P = 1  0 10 (52.63%) 9 (47.37%)  1 4 (21.05%) 5 (26.32%)  2 5 (26.32%) 5 (26.32%) Smoke during pregnancy P = 1  No 19 (100%) 19 (100%) Infant sexa P = 1  Males 10 (47.62%) 10 (47.62%)  Females 11 (52.38%) 11 (52.38%) Ethnicity P = 0.8126  Non-hispanic Caucasian 11 (57.89%) 12 (63.16%)  Hispanic or Latino 3 (15.79%) 4 (21.05%)  Black/African-American 5 (26.32%) 3 (15.79%) Conception P = 1  Spontaneous planned 11 (57.89%) 12 (63.16%)  Spontaneous unplanned 7 (36.84%) 6 (31.58%)  IVF 1 (5.26%) 1 (5.26%) Birth weight (g) 3268.29 (500.6) 3387.09 (549.47) P = 0.4682 Gestational age (weeks) 38.13 (2.22) 38.94 (1.43) P = 0.1778 Characteristic . Cases (n = 19) . Controls (n = 19) . P-value . Pre-pregnancy BMI (kg/m2) 28.38 (8.89) 26.65 (6.57) P = 0.4982 Maternal age (years) 32.05 (6.91) 31.68 (6.01) P = 0.8618 Gravidity P = 0.8822  0 7 (36.84%) 8 (42.11%)  1 5 (26.32%) 3 (15.79%)  2 5 (26.32%) 5 (26.32%)  ≥3 2 (10.53%) 3 (15.79%) Parity P = 1  0 10 (52.63%) 9 (47.37%)  1 4 (21.05%) 5 (26.32%)  2 5 (26.32%) 5 (26.32%) Smoke during pregnancy P = 1  No 19 (100%) 19 (100%) Infant sexa P = 1  Males 10 (47.62%) 10 (47.62%)  Females 11 (52.38%) 11 (52.38%) Ethnicity P = 0.8126  Non-hispanic Caucasian 11 (57.89%) 12 (63.16%)  Hispanic or Latino 3 (15.79%) 4 (21.05%)  Black/African-American 5 (26.32%) 3 (15.79%) Conception P = 1  Spontaneous planned 11 (57.89%) 12 (63.16%)  Spontaneous unplanned 7 (36.84%) 6 (31.58%)  IVF 1 (5.26%) 1 (5.26%) Birth weight (g) 3268.29 (500.6) 3387.09 (549.47) P = 0.4682 Gestational age (weeks) 38.13 (2.22) 38.94 (1.43) P = 0.1778 aTotal infants from n = 34 singleton and n = 4 twin pregnancies. Open in new tab To control for technical variation within the array, we analyzed two samples in triplicate (Supplementary Material, Fig. S1). In the first instance, we carried out the gene expression analysis without separating the samples by infant sex. Principal component analysis (PCA) of expression data separated three samples with unusual expression profiles that could not be explained by known characteristics. As such, these were excluded in downstream analysis (Fig. 1A). Preeclampsia was not significantly associated with any of the first 10 principal components of gene expression variation, which accounted for approximately 70% of the total gene expression variation in the dataset. Multivariable linear regression models were used to identify significant associations between specific gene expression and preeclampsia, with adjustment for pre-pregnancy BMI (kg/m2), maternal age (years) and ethnicity associated with slight variation in matched pairs. Singleton and twin pregnancies were analyzed separately. Among singletons, only one gene was differentially expressed in placental tissue from preeclamptic versus control individuals after correcting for multiple comparisons (corrected P < 0.05). Expression of transcript TC03002313.hg.1 (chr3: 44 046 930–44 163 857; genome build hg19) was 1.84-fold greater among preeclamptic placenta samples relative to the control pregnancies (Fig. 1B; Supplementary Material, Table S1). This upregulated transcript region overlaps with two non-coding elements, MIR138–1 (chr3: 44 155 704–44 155 802) and piR-59 342 (chr3: 44 159 798–44 163 857). The second largest change was a 1.75-fold increase in TC03000234.hg.1, a transcript that only captures piR-59 342 (Supplementary Material, Table S1). Expression of these two transcripts was highly correlated (R = 0.91), suggesting that the variation in piR-59 342 expression accounted for most of the association between case-status and TC03002313.hg.1. Accordingly, the magnitude of the association between preeclampsia and TC03002313.hg.1 was reduced by nearly 30% (1.30-fold reduction, P = 0.002) after adjusting for TC03000234.hg.1 expression, as well as pre-pregnancy BMI, maternal age and ethnicity. Figure 1 Open in new tabDownload slide Gene-level expression analysis of singletons using the HTA 2.0 microarray. (A) PCA of gene expression results for singleton samples. (B) Boxplots of the top gene-level expression changes (No = control; Yes = preeclampsia case). The two boxplots correspond to MIR138-1 probe TC03002313.hg.1 (P-adjusted = 0.001) and piR-59 342 probe TC03000234.hg.1 (P-adjusted = 0.178) expression. Figure 1 Open in new tabDownload slide Gene-level expression analysis of singletons using the HTA 2.0 microarray. (A) PCA of gene expression results for singleton samples. (B) Boxplots of the top gene-level expression changes (No = control; Yes = preeclampsia case). The two boxplots correspond to MIR138-1 probe TC03002313.hg.1 (P-adjusted = 0.001) and piR-59 342 probe TC03000234.hg.1 (P-adjusted = 0.178) expression. Infant sex-specific gene expression associated with preeclampsia In order to identify preeclampsia-associated differentially expressed genes modulated by infant sex, we analyzed gene expression according to infant sex—male or female (Fig. 2). In placentas from female infants (Fig. 2C), we observed a 2.29-fold upregulation of TC03002313.hg.1 (P-adjusted = 0.0084) in conjunction with upregulation of TC03000234.hg.1, as described previously for all samples (fold change = 2.28; P-adjusted = 0.045). Interestingly, these two transcripts are not differentially expressed in placentas from male infants. This suggests that their potential association with preeclampsia might be specific to females. Among females (Fig. 2C), we also identified a significant association between case-status and Signaling Lymphocytic Activation Molecule Family Member 1 (SLAMF1; TC01003406.hg.1; fold change = 0.62; P-adjusted = 0.0174), which was slightly downregulated in preeclampsia compared to controls, and two genes, which were modestly upregulated—Tetratricopeptide Repeat Domain 25 (TTC25; TC17000519.hg.1; fold change = 1.79; P-adjusted = 0.0335) and Glycoprotein VI Platelet (GP6; TC19002721.hg.1; fold change = 1.58; P-adjusted = 0.0335) (Fig. 2C). Two of these genes, SLAMF1 (28) and TTC25 (29), have been previously reported as differentially expressed in preeclamptic placentas. We did not observe genome-wide significant associations between expression of any gene and case-status in placentas with male infants (Fig. 2D). Figure 2 Open in new tabDownload slide Infant sex-specific preeclampsia-associated gene-level expression analysis in singletons using the HTA 2.0 microarray. (A) PCA in samples analyzed separately depending on infant sex. Six most significant associations between gene expression and preeclampsia across (B) all infants, (C) females and (D) males. Sex-specific estimates were obtained from models integrating a product term between preeclampsia and sex. Figure 2 Open in new tabDownload slide Infant sex-specific preeclampsia-associated gene-level expression analysis in singletons using the HTA 2.0 microarray. (A) PCA in samples analyzed separately depending on infant sex. Six most significant associations between gene expression and preeclampsia across (B) all infants, (C) females and (D) males. Sex-specific estimates were obtained from models integrating a product term between preeclampsia and sex. Expression changes associated with preeclampsia in twins Of the four total sets of twins analyzed, two were from preeclamptic pregnancies (cases) and two were controls. One of the sets was a monozygotic preeclamptic twin pair; the three other sets of twins were dizygotic. All of the pregnancies were dichorionic, with a sample extracted from each placenta (n = 8 samples). Each set of twins were the same sex (4 males, 4 females), and all mothers were non-Hispanic Caucasian. As was the case with the singleton samples, control twin pregnancies were matched to preeclamptic twin pregnancies based on maternal age, pre-pregnancy BMI, method of conception (IVF versus spontaneous), infant sex and plurality of the pregnancy. Preeclampsia was not associated with any of the principal components of expression after adjusting for maternal age and matched pair. The expression levels of five transcripts were associated with preeclampsia in adjusted models after correcting for multiple comparisons. These shifts in expression tended to be driven by a specific set of twins, with relatively similar intra-pair expression levels (Supplementary Material, Fig. S2). The differentially expressed transcripts included: CYP1B1 (chr2: 38 294 116–38 297 840), MIR1973 (chr4: 117 220 881–117 220 924), DEFA1B (chr8: 6 873 391–6 875 823), a lincRNA (chr14: 27 342 339–27 383 949) and a transcript of uncertain coding potential (chr19: 302 198–304 460) with all genomic coordinates referring to genome build hg19. Genome-wide DNA methylation analysis in singletons Genome-wide DNA methylation of 39 singleton placenta samples (including 20 cases) was assayed using the Illumina Infinium Human Methylation 450 (450K) BeadChip array. The 450K BeadChip includes >480 000 individual CpGs located in 99% of RefSeq genes and covers a variety of regions, such as promoters, 5′- and 3′-UTRs, as well as gene bodies. PCA of the resulting methylation data revealed that preeclampsia was not associated with the top 20 components of the variation in DNA methylation (70% of total variation) after adjusting for pre-pregnancy BMI, maternal age, ethnicity and array batch effects. Given that preeclampsia was not associated with genome-wide shifts in DNA methylation, we next investigated whether, or not, it was associated with more localized changes in placental epigenetic regulation. We first examined site-specific associations between DNA methylation and preeclampsia using an empirical Bayes method (30) and adjusting for the covariates listed above. Using the Benjamini and Hochberg method (31) to correct for the false discovery rate (FDR), we did not detect any significant site-specific associations with preeclampsia (Supplementary Material, Table S2). Furthermore, the top 10 changes in DNA methylation between cases and controls were relatively small, ranging from 0.023 to 0.072 (~2.3–7.2%), with methylation higher among preeclampsia samples in all cases. Exploiting the correlation structure between proximal CpG loci on the microarray, we next utilized a ‘bump-hunting’ approach to identify potentially larger DMRs between preeclamptic placentas and controls. Clusters were defined as contiguous CpG loci within 500 bps that had a difference in DNA methylation between the cases and controls in adjusted models that was greater than the 99th percentile of those estimated across the array. Top regions were sorted based on the bootstrapped empirical probability of observing a regional change in methylation of that magnitude (Table 2 and Fig. 3). This approach prioritizes larger shifts in methylation between cases and controls in adjusted models over large groupings of CpG loci. Accordingly, the 10 top ranked regions consisted of 7–11 CpG loci with absolute differences in methylation ranging from approximately 10 to 20% at each individual site. The distribution of methylation at these sites was clearly ‘bimodal’ in some instances, with each mode being distinctly enriched for either case or control samples (Fig. 3). Several of the top ranked regions were within close proximity to a gene transcription start site (TSS). In addition to potentially indicating programming of proximal gene expression, this enrichment partially reflects the high density of these sites chosen in the microarray design. Table 2 The top-ranked regional changes in methylation among singletons on the 450K Region . Number of CpG loci . Area . P-value . Gene in proximity . chr12: 53 693 322–53 694 018 11 1.693 6.38E − 05 C12ORF10 (MYG1) chr5: 1 594 282–1 594 863 10 1.528 1.05E – 04 SDHAP3 chr5: 150 284 302–150 284 796 9 1.230 2.88E − 04 ZNF300 chr11: 9 113 152–9 113 458 9 1.080 4.78E − 04 SCUBE2 chr8: 144 659 831–144 660 772 9 1.062 5.10E − 04 NAPRT1 chr5: 111 093 197–111 093 755 9 1.053 5.26E − 04 NREP chr17: 15 847 977–15 848 264 9 1.041 5.50E − 04 ADORA2B chr18: 53 446 528–53 447 841 7 0.951 7.62E − 04 AK127787 chr19: 52 074 293–52 074 501 8 0.949 7.70E − 04 ZNF175 chr19: 58 427 707–58 428 186 8 0.832 1.21E – 03 ZNF417 Region . Number of CpG loci . Area . P-value . Gene in proximity . chr12: 53 693 322–53 694 018 11 1.693 6.38E − 05 C12ORF10 (MYG1) chr5: 1 594 282–1 594 863 10 1.528 1.05E – 04 SDHAP3 chr5: 150 284 302–150 284 796 9 1.230 2.88E − 04 ZNF300 chr11: 9 113 152–9 113 458 9 1.080 4.78E − 04 SCUBE2 chr8: 144 659 831–144 660 772 9 1.062 5.10E − 04 NAPRT1 chr5: 111 093 197–111 093 755 9 1.053 5.26E − 04 NREP chr17: 15 847 977–15 848 264 9 1.041 5.50E − 04 ADORA2B chr18: 53 446 528–53 447 841 7 0.951 7.62E − 04 AK127787 chr19: 52 074 293–52 074 501 8 0.949 7.70E − 04 ZNF175 chr19: 58 427 707–58 428 186 8 0.832 1.21E – 03 ZNF417 Open in new tab Table 2 The top-ranked regional changes in methylation among singletons on the 450K Region . Number of CpG loci . Area . P-value . Gene in proximity . chr12: 53 693 322–53 694 018 11 1.693 6.38E − 05 C12ORF10 (MYG1) chr5: 1 594 282–1 594 863 10 1.528 1.05E – 04 SDHAP3 chr5: 150 284 302–150 284 796 9 1.230 2.88E − 04 ZNF300 chr11: 9 113 152–9 113 458 9 1.080 4.78E − 04 SCUBE2 chr8: 144 659 831–144 660 772 9 1.062 5.10E − 04 NAPRT1 chr5: 111 093 197–111 093 755 9 1.053 5.26E − 04 NREP chr17: 15 847 977–15 848 264 9 1.041 5.50E − 04 ADORA2B chr18: 53 446 528–53 447 841 7 0.951 7.62E − 04 AK127787 chr19: 52 074 293–52 074 501 8 0.949 7.70E − 04 ZNF175 chr19: 58 427 707–58 428 186 8 0.832 1.21E – 03 ZNF417 Region . Number of CpG loci . Area . P-value . Gene in proximity . chr12: 53 693 322–53 694 018 11 1.693 6.38E − 05 C12ORF10 (MYG1) chr5: 1 594 282–1 594 863 10 1.528 1.05E – 04 SDHAP3 chr5: 150 284 302–150 284 796 9 1.230 2.88E − 04 ZNF300 chr11: 9 113 152–9 113 458 9 1.080 4.78E − 04 SCUBE2 chr8: 144 659 831–144 660 772 9 1.062 5.10E − 04 NAPRT1 chr5: 111 093 197–111 093 755 9 1.053 5.26E − 04 NREP chr17: 15 847 977–15 848 264 9 1.041 5.50E − 04 ADORA2B chr18: 53 446 528–53 447 841 7 0.951 7.62E − 04 AK127787 chr19: 52 074 293–52 074 501 8 0.949 7.70E − 04 ZNF175 chr19: 58 427 707–58 428 186 8 0.832 1.21E – 03 ZNF417 Open in new tab Figure 3 Open in new tabDownload slide Genome-wide DNA methylation analysis of all singletons using 450K. DMRs with highest rank in paired models (No = control; Yes = preeclampsia case). (A) C12ORF10 (MYG1; chr12: 53 693 322–53 694 018); (B) SDHAP3 (chr5: 1 594 282–1 594 863); (C) ZNF300 (chr5: 150 284 302–150 284 796); (D) SCUBE2 (chr11: 9 113 152–9 113 458); (E) NAPRT1 (chr8: 144 659 831–144 660 772); (F) NREP (chr5: 111 093 197–111 093 755). Figure 3 Open in new tabDownload slide Genome-wide DNA methylation analysis of all singletons using 450K. DMRs with highest rank in paired models (No = control; Yes = preeclampsia case). (A) C12ORF10 (MYG1; chr12: 53 693 322–53 694 018); (B) SDHAP3 (chr5: 1 594 282–1 594 863); (C) ZNF300 (chr5: 150 284 302–150 284 796); (D) SCUBE2 (chr11: 9 113 152–9 113 458); (E) NAPRT1 (chr8: 144 659 831–144 660 772); (F) NREP (chr5: 111 093 197–111 093 755). Table 3 Ten most significant regional changes in methylation in singleton pregnancies according to infant-sex Region . Number of CpG loci . Area . P-value . Gene in proximity . Significant male-specific preeclampsia-associated DMRs chr17: 15 847 977–15 848 264 9 1.823 2.70E − 04 ADORA2B chr5: 1 594 282–1 594 863 10 1.597 4.56E − 04 SDHAP3 chr21: 37 442 104–37 442 777 6 1.445 6.70E − 04 CBR1 chr19: 52 074 293–52 074 501 8 1.246 1.13E − 03 ZNF175 chr18: 53 446 528–53 448 189 8 1.224 1.19E − 03 AK127787 chr1: 173 638 690–173 639 135 6 1.195 1.30E − 03 ANKRD45 chr7: 23 720 005–23 720 838 7 1.127 1.55E − 03 FAM221A chr19: 12 876 846–12 877 188 4 0.902 2.92E − 03 HOOK2 chr17: 75 524 696–75 525 151 5 0.806 3.89E − 03 BC040189 chr19: 40 324 850–40 324 993 6 0.793 4.04E − 03 DYRK1B Significant female-specific preeclampsia-associated DMRs chr12: 53 693 322–53 694 018 11 2.160 1.49E − 04 C12ORF10 (MYG1) chr4: 80 246 713–80 247 524 10 1.820 3.19E − 04 NAA11 chr8: 144 659 627–144 660 772 10 1.789 3.41E − 04 NAPRT1 chr5: 111 093 197–111 093 755 9 1.477 6.94E − 04 NREP chr7: 101 005 910–101 006 308 8 1.460 7.21E − 04 COL26A1 chr5: 1 594 282–1 594 808 8 1.347 9.57E − 04 SDHAP3 chr2: 63 283 939–63 284 768 8 1.259 1.21E − 03 OTX1 chr19: 58 220 370–58 220 837 9 1.197 1.41E − 03 ZNF417 chr5: 148 960 831–148 961 311 8 1.135 1.66E − 03 ARHGEF37 chr16: 68 572 892–68 573 721 7 1.109 1.77E − 03 ZFP90 Region . Number of CpG loci . Area . P-value . Gene in proximity . Significant male-specific preeclampsia-associated DMRs chr17: 15 847 977–15 848 264 9 1.823 2.70E − 04 ADORA2B chr5: 1 594 282–1 594 863 10 1.597 4.56E − 04 SDHAP3 chr21: 37 442 104–37 442 777 6 1.445 6.70E − 04 CBR1 chr19: 52 074 293–52 074 501 8 1.246 1.13E − 03 ZNF175 chr18: 53 446 528–53 448 189 8 1.224 1.19E − 03 AK127787 chr1: 173 638 690–173 639 135 6 1.195 1.30E − 03 ANKRD45 chr7: 23 720 005–23 720 838 7 1.127 1.55E − 03 FAM221A chr19: 12 876 846–12 877 188 4 0.902 2.92E − 03 HOOK2 chr17: 75 524 696–75 525 151 5 0.806 3.89E − 03 BC040189 chr19: 40 324 850–40 324 993 6 0.793 4.04E − 03 DYRK1B Significant female-specific preeclampsia-associated DMRs chr12: 53 693 322–53 694 018 11 2.160 1.49E − 04 C12ORF10 (MYG1) chr4: 80 246 713–80 247 524 10 1.820 3.19E − 04 NAA11 chr8: 144 659 627–144 660 772 10 1.789 3.41E − 04 NAPRT1 chr5: 111 093 197–111 093 755 9 1.477 6.94E − 04 NREP chr7: 101 005 910–101 006 308 8 1.460 7.21E − 04 COL26A1 chr5: 1 594 282–1 594 808 8 1.347 9.57E − 04 SDHAP3 chr2: 63 283 939–63 284 768 8 1.259 1.21E − 03 OTX1 chr19: 58 220 370–58 220 837 9 1.197 1.41E − 03 ZNF417 chr5: 148 960 831–148 961 311 8 1.135 1.66E − 03 ARHGEF37 chr16: 68 572 892–68 573 721 7 1.109 1.77E − 03 ZFP90 Open in new tab Table 3 Ten most significant regional changes in methylation in singleton pregnancies according to infant-sex Region . Number of CpG loci . Area . P-value . Gene in proximity . Significant male-specific preeclampsia-associated DMRs chr17: 15 847 977–15 848 264 9 1.823 2.70E − 04 ADORA2B chr5: 1 594 282–1 594 863 10 1.597 4.56E − 04 SDHAP3 chr21: 37 442 104–37 442 777 6 1.445 6.70E − 04 CBR1 chr19: 52 074 293–52 074 501 8 1.246 1.13E − 03 ZNF175 chr18: 53 446 528–53 448 189 8 1.224 1.19E − 03 AK127787 chr1: 173 638 690–173 639 135 6 1.195 1.30E − 03 ANKRD45 chr7: 23 720 005–23 720 838 7 1.127 1.55E − 03 FAM221A chr19: 12 876 846–12 877 188 4 0.902 2.92E − 03 HOOK2 chr17: 75 524 696–75 525 151 5 0.806 3.89E − 03 BC040189 chr19: 40 324 850–40 324 993 6 0.793 4.04E − 03 DYRK1B Significant female-specific preeclampsia-associated DMRs chr12: 53 693 322–53 694 018 11 2.160 1.49E − 04 C12ORF10 (MYG1) chr4: 80 246 713–80 247 524 10 1.820 3.19E − 04 NAA11 chr8: 144 659 627–144 660 772 10 1.789 3.41E − 04 NAPRT1 chr5: 111 093 197–111 093 755 9 1.477 6.94E − 04 NREP chr7: 101 005 910–101 006 308 8 1.460 7.21E − 04 COL26A1 chr5: 1 594 282–1 594 808 8 1.347 9.57E − 04 SDHAP3 chr2: 63 283 939–63 284 768 8 1.259 1.21E − 03 OTX1 chr19: 58 220 370–58 220 837 9 1.197 1.41E − 03 ZNF417 chr5: 148 960 831–148 961 311 8 1.135 1.66E − 03 ARHGEF37 chr16: 68 572 892–68 573 721 7 1.109 1.77E − 03 ZFP90 Region . Number of CpG loci . Area . P-value . Gene in proximity . Significant male-specific preeclampsia-associated DMRs chr17: 15 847 977–15 848 264 9 1.823 2.70E − 04 ADORA2B chr5: 1 594 282–1 594 863 10 1.597 4.56E − 04 SDHAP3 chr21: 37 442 104–37 442 777 6 1.445 6.70E − 04 CBR1 chr19: 52 074 293–52 074 501 8 1.246 1.13E − 03 ZNF175 chr18: 53 446 528–53 448 189 8 1.224 1.19E − 03 AK127787 chr1: 173 638 690–173 639 135 6 1.195 1.30E − 03 ANKRD45 chr7: 23 720 005–23 720 838 7 1.127 1.55E − 03 FAM221A chr19: 12 876 846–12 877 188 4 0.902 2.92E − 03 HOOK2 chr17: 75 524 696–75 525 151 5 0.806 3.89E − 03 BC040189 chr19: 40 324 850–40 324 993 6 0.793 4.04E − 03 DYRK1B Significant female-specific preeclampsia-associated DMRs chr12: 53 693 322–53 694 018 11 2.160 1.49E − 04 C12ORF10 (MYG1) chr4: 80 246 713–80 247 524 10 1.820 3.19E − 04 NAA11 chr8: 144 659 627–144 660 772 10 1.789 3.41E − 04 NAPRT1 chr5: 111 093 197–111 093 755 9 1.477 6.94E − 04 NREP chr7: 101 005 910–101 006 308 8 1.460 7.21E − 04 COL26A1 chr5: 1 594 282–1 594 808 8 1.347 9.57E − 04 SDHAP3 chr2: 63 283 939–63 284 768 8 1.259 1.21E − 03 OTX1 chr19: 58 220 370–58 220 837 9 1.197 1.41E − 03 ZNF417 chr5: 148 960 831–148 961 311 8 1.135 1.66E − 03 ARHGEF37 chr16: 68 572 892–68 573 721 7 1.109 1.77E − 03 ZFP90 Open in new tab Infant sex-specific DNA methylation associated with preeclampsia Analysis of differentially methylated positions revealed little overlap in the top disease-associated DMPs depending on infant sex (Supplementary Material, Table S3). When we investigated DMRs in placenta samples stratified by infant sex, we observed that SDHAP3 is the only preeclampsia-associated DMR shared among males and females. In contrast, ADORA2B (chr17: 15 847 977–15 848 264) and AK127787 (chr18: 53 446 528–53 448 189) appear to be male-specific preeclampsia-associated DMRs, while C12ORF10 (MYG1; chr12: 53 693 322–53 694 018), NREP (chr5: 111 093 197–111 093 755), and NAPRT1 (chr8: 144 659 627–144 660 772) appear to be female-specific (Table 3). The ADORA2B and NAPRT1 regions that we observe overlap with previously reported preeclampsia-associated DMRs (32). Genome-wide DNA methylation analysis in twins The site-specific variation in methylation between preeclamptic and control placentas from twin pregnancies was more pronounced, compared to singletons. We detected a significant difference in methylation (FDR; P-adjusted < 0.05) at three CpG loci after adjusting for maternal pre-pregnancy BMI, age, ethnicity and estimated batch effects (Supplementary Material, Fig. S4). Beta-values were −0.36 (36% lower) among the preeclamptic samples relative to the control samples within a CpG island close to the TSS of LHX9 (cg16353957) and −0.48 (48% lower) within an island located in the last exon of LACTBL1 (cg17974166). In contrast, methylation was 17% higher among the preeclamptic samples within an intron of NRIP3 (cg02864070). Similar to the singleton analysis, we also investigated possible regional variation in methylation levels using a bump-hunting approach. The top regional shifts in methylation again showed a clear ‘bimodal’ distribution (as observed in Fig. 3). While methylation levels within the most significant regions tended to be higher among the preeclamptic relative to control pregnancies in singletons (Fig. 3), among multiples, methylation levels tended to be higher among the control pregnancies (Supplementary Material, Fig. S5 and Supplementary Material, Table S4). None of the significant site-specific differences within twins were among the 10 most significant regional changes. Interestingly, 2 of the top 10 regional changes in methylation among singletons and twins overlapped, specifically chr8: 144 659 883–144 660 772 (singleton rank = 5; twin rank = 4) and chr19: 58 427 707–58 428 186 (singleton rank = 10; twin rank = 6). These two DMRs correspond to the promoter CpG islands (CGIs) of NAPRT1 and ZNF417, respectively. Limited association between gene expression changes discovered by HTA 2.0 and DMRs observed from the 450K methylation array Among the 28 singleton samples (including 15 cases) with both expression and methylation data, we investigated whether the change in TC03002313.hg.1 expression correlates with the methylation of CpG loci within 100 kb of the transcript start or end. Expression of this gene was not significantly correlated with methylation at any CpG site in relatively close proximity after correcting for multiple testing (FDR). Similarly, we evaluated whether the average methylation level across the top 10 ranked DMRs was correlated with the expression of any genes within 100 kb. Average methylation across the top-ranked DMR, which was located within the promoter of C12ORF10, had an inverse correlation with the expression of this gene that was marginally significant (R = −0.47, P-adjusted = 0.058). Average methylation across the fifth-ranked DMR, which was within the promoter of NAPRT1 (Table 2), was also inversely correlated with gene expression (R = −0.66, P-adjusted = 7.05e − 04; Fig. 4A). Methylation within the promoter of NREP, the sixth-ranked DMR, was inversely correlated with expression as well (R = −0.59, P-adjusted = 8.99e-04). Methylation within the CGIs of ZNF175 and ZNF417, the ninth- and tenth-ranked DMRs, respectively (Table 2), was inversely correlated with expression also. For ZNF175, the correlation was moderate (R = −0.42; P-adjusted = 0.0246), while for ZNF417 it was stronger (R = −0.62; P-adjusted = 9.47e − 04; Fig. 4A). Despite these correlations with promoter DNA methylation, preeclampsia was not clearly associated with changes in the expression of any of these genes after adjusting for pre-pregnancy BMI, maternal age and ethnicity (not shown). As noted in the Materials and Methods, none of the surrogate variables estimated from the expression data were significantly associated with batch. This might suggest that the variation in methylation across these regions that was correlated with expression is independent of the variation associated with preeclampsia or that the sample size was not sufficient to detect smaller changes in gene expression associated with the disorder. None of the other top 10 ranked regional changes were correlated with the expression of any genes in close proximity. Infant sex did not have a major effect on the gene expression—DNA methylation correlation. We observed a significant correlation between NAPRT1, ZNF154, NAA11 and NREP gene expression and DNA methylation at DMRs in proximity to these genes in females (Fig. 4B). A similar correlation was observed only for FAM221A in males (Fig. 4C). When we looked at the correlation between site-specific DNA methylation and gene expression, we observed that they were also different depending on infant sex (Supplementary Material, Fig. S6). Figure 4 Open in new tabDownload slide Boxplots of association between gene-level expression and DNA methylation variation in proximity. Significant (q < 0.05) correlations between the top-ranked DMRs and gene expression within 100 kb. Plotting associations that were significant across (A) All infants. Associations are shown for NAPRT1 (R = −0.66, P-adjusted = 7.05e − 04); NREP (R = −0.59; P-adjusted = 8.99e − 04); ZNF175 (R = −0.42; P-adjusted = 0.0246) and ZNF417 (R = −0.62; P = 9.47e − 04). (B) Among females; (C) Among males. Figure 4 Open in new tabDownload slide Boxplots of association between gene-level expression and DNA methylation variation in proximity. Significant (q < 0.05) correlations between the top-ranked DMRs and gene expression within 100 kb. Plotting associations that were significant across (A) All infants. Associations are shown for NAPRT1 (R = −0.66, P-adjusted = 7.05e − 04); NREP (R = −0.59; P-adjusted = 8.99e − 04); ZNF175 (R = −0.42; P-adjusted = 0.0246) and ZNF417 (R = −0.62; P = 9.47e − 04). (B) Among females; (C) Among males. Similarly, we examined the association between the top-ranked differences in methylation and expression among the four sets of twins with methylation and expression data. Expression of DEFA1B was positively correlated with the methylation of one CpG site within 1500 bp of the TSS of DEFA1B (cg04178011; R = 0.93, P-adjusted = 0.032). None of the other changes in expression were associated with methylation within 100 kb of the transcript after correcting for multiple comparisons. The three significant changes in site-specific DNA methylation were not associated with any proximal expression. Only two of the top 10 differences in regional methylation among twins were correlated with the expression of a gene within 100 kb. Average methylation in the most significant regional difference (chr6: 31 783 240–31 783 545) was negatively correlated with the expression of L2M2 (TC06001545.hg.1; R = −0.87, P-adjusted = 0.0261), a gene nearly 10 kb downstream. Methylation within the fifth-ranked regional shift (chr11: 43 902 112–43 902 458) was negatively correlated with an unclassified transcript (TC11001574.hg.1; R = −0.93, P-adjusted = 0.00182). Validation of differential DNA methylation An informative epigenome-wide association study (EWAS) requires validation of the results in a separate cohort using a targeted bisulfite sequencing approach (33). In order to validate the results from the 450K DNA methylation analysis, we obtained genomic DNA isolated from full-thickness placenta samples harvested from women enrolled in the BIS (26). The validation was performed using n = 16 samples from preeclamptic placentas and the same number of matched controls, obtained from singleton pregnancies (total n = 32). We again matched cases and controls according to known risk factors and possible confounders, including maternal age, pre-pregnancy BMI, gestational age, smoking status and infant sex. The majority of the participants in this validation cohort were Caucasian with European ancestry. For a complete list of BIS patient characteristics, see Table 4. Table 4 Characteristics of patients from the Barwon Infant Study cohort used for validation of DNA methylation results Characteristic . Cases (n = 16) . Controls (n = 16) . Pre-pregnancy BMI (kg/m2) 26.20 (17.67–35.16) 27.47 (18.18–49.95) BMI category Missing n = 1  Underweight (<18.5) 1 (6.25%) 1 (6.25%)  Normal (18.5–<25) 4 (25%) 5 (31.25%)  Overweight (25–<30) 6 (37.5%) 6 (37.5%)  Obese (30+) 4 (25%) 4 (25%) Maternal age (years) 34.22 33.31 Smoke during pregnancy  No 14 (87.5%) 14 (87.5%)  Yes 2 (12.5%) 2 (12.5%) Infant sex  Males 9 (56.25%) 9 (56.25%)  Females 7 (43.75%) 7 (43.75%) Mode of birth  Elective 2 (12.5%) 4 (25%)  Vaginal 5 (31.25%) 11 (68.75%)  Emergency 9 (56.25%) 1 (6.25%) Gestational age (weeks) 37.38 (34–40) 38.50 (36–40) Ethnicity of the child  British/Irish or European ancestrya 11 (68.75%) 12 (75%)  Other 5 (31.25%) 4 (25%) Characteristic . Cases (n = 16) . Controls (n = 16) . Pre-pregnancy BMI (kg/m2) 26.20 (17.67–35.16) 27.47 (18.18–49.95) BMI category Missing n = 1  Underweight (<18.5) 1 (6.25%) 1 (6.25%)  Normal (18.5–<25) 4 (25%) 5 (31.25%)  Overweight (25–<30) 6 (37.5%) 6 (37.5%)  Obese (30+) 4 (25%) 4 (25%) Maternal age (years) 34.22 33.31 Smoke during pregnancy  No 14 (87.5%) 14 (87.5%)  Yes 2 (12.5%) 2 (12.5%) Infant sex  Males 9 (56.25%) 9 (56.25%)  Females 7 (43.75%) 7 (43.75%) Mode of birth  Elective 2 (12.5%) 4 (25%)  Vaginal 5 (31.25%) 11 (68.75%)  Emergency 9 (56.25%) 1 (6.25%) Gestational age (weeks) 37.38 (34–40) 38.50 (36–40) Ethnicity of the child  British/Irish or European ancestrya 11 (68.75%) 12 (75%)  Other 5 (31.25%) 4 (25%) aAll four grandparents are of British/Irish or European ancestry. Open in new tab Table 4 Characteristics of patients from the Barwon Infant Study cohort used for validation of DNA methylation results Characteristic . Cases (n = 16) . Controls (n = 16) . Pre-pregnancy BMI (kg/m2) 26.20 (17.67–35.16) 27.47 (18.18–49.95) BMI category Missing n = 1  Underweight (<18.5) 1 (6.25%) 1 (6.25%)  Normal (18.5–<25) 4 (25%) 5 (31.25%)  Overweight (25–<30) 6 (37.5%) 6 (37.5%)  Obese (30+) 4 (25%) 4 (25%) Maternal age (years) 34.22 33.31 Smoke during pregnancy  No 14 (87.5%) 14 (87.5%)  Yes 2 (12.5%) 2 (12.5%) Infant sex  Males 9 (56.25%) 9 (56.25%)  Females 7 (43.75%) 7 (43.75%) Mode of birth  Elective 2 (12.5%) 4 (25%)  Vaginal 5 (31.25%) 11 (68.75%)  Emergency 9 (56.25%) 1 (6.25%) Gestational age (weeks) 37.38 (34–40) 38.50 (36–40) Ethnicity of the child  British/Irish or European ancestrya 11 (68.75%) 12 (75%)  Other 5 (31.25%) 4 (25%) Characteristic . Cases (n = 16) . Controls (n = 16) . Pre-pregnancy BMI (kg/m2) 26.20 (17.67–35.16) 27.47 (18.18–49.95) BMI category Missing n = 1  Underweight (<18.5) 1 (6.25%) 1 (6.25%)  Normal (18.5–<25) 4 (25%) 5 (31.25%)  Overweight (25–<30) 6 (37.5%) 6 (37.5%)  Obese (30+) 4 (25%) 4 (25%) Maternal age (years) 34.22 33.31 Smoke during pregnancy  No 14 (87.5%) 14 (87.5%)  Yes 2 (12.5%) 2 (12.5%) Infant sex  Males 9 (56.25%) 9 (56.25%)  Females 7 (43.75%) 7 (43.75%) Mode of birth  Elective 2 (12.5%) 4 (25%)  Vaginal 5 (31.25%) 11 (68.75%)  Emergency 9 (56.25%) 1 (6.25%) Gestational age (weeks) 37.38 (34–40) 38.50 (36–40) Ethnicity of the child  British/Irish or European ancestrya 11 (68.75%) 12 (75%)  Other 5 (31.25%) 4 (25%) aAll four grandparents are of British/Irish or European ancestry. Open in new tab To assess DNA methylation at CpGs within the DMRs of C12ORF10 (MYG1), ZNF300, SDHAP3 and NAPRT1, we designed pyrosequencing assays listed in Supplementary Material, Table S5. We chose to validate these targets because they were among the 10 top-ranked DMRs identified from the 450K methylation array. In addition, they were either completely demethylated in some preeclamptic samples compared to controls in what we refer to as ‘bimodal’ distribution in methylation (MYG1 and SDHAP3) or exhibited a sizable methylation difference between cases and controls (ZNF300), or the regional changes in methylation overlapped between singletons and twins (NAPRT1). When we plotted percent methylation at individual CpGs for each gene, we observed the same pattern of completely demethylated CpGs in mostly preeclampsia-affected samples in MYG1 and SDHAP3. This ‘bimodal’ distribution was evident when we averaged methylation levels across all sites for each gene and plotted methylation differences within each case-control pair (Fig. 5). For ZNF300 and NAPRT1 (Fig. 5), DNA methylation levels appeared more heterogeneous among samples with no clear distinction between preeclampsia cases and controls. Surprisingly, we observed a difference in DNA methylation not only within matched case–control pairs but also between pairs, which was not associated with differences in other characteristics such as maternal pre-pregnancy BMI (Supplementary Material, Fig. S7A), maternal age (Supplementary Material, Fig. S7B), gestational age (Supplementary Material, Fig. S7C) or mode of birth (not shown). Note that the line of best fit in each respective plot is mostly driven by the outlier matched pairs. The observation that there is DNA methylation variation between case–control pairs that cannot be accounted for by known sample characteristics might suggest the presence of an additional, unidentified, confounder. Figure 5 Open in new tabDownload slide Pair-wise comparison of DNA methylation as measured using pyrosequencing and averaged among multiple CpGs. Pairs consist of matched controls and cases. C12ORF10 (MYG1) pyrosequencing measurement averaged across four CpGs; SDHAP3 pyrosequencing measurement averaged across six CpGs; ZNF300 pyrosequencing measurement averaged across six CpGs; NAPRT1 pyrosequencing measurement averaged across five CpGs. Figure 5 Open in new tabDownload slide Pair-wise comparison of DNA methylation as measured using pyrosequencing and averaged among multiple CpGs. Pairs consist of matched controls and cases. C12ORF10 (MYG1) pyrosequencing measurement averaged across four CpGs; SDHAP3 pyrosequencing measurement averaged across six CpGs; ZNF300 pyrosequencing measurement averaged across six CpGs; NAPRT1 pyrosequencing measurement averaged across five CpGs. Next, we plotted the distribution of methylation level values at each individual CpG site that we investigated across all four genes of interest. We observed that for NAPRT1 (Fig. 6) in all investigated CpG positions the median DNA methylation levels were lower in preeclampsia cases compared to controls. The box and whisker plot for SDHAP3 was most striking with an obvious difference in distribution of methylation values between preeclampsia cases and controls despite similar median values (Fig. 6). For MYG1 and ZNF300, we could not detect a significant difference neither in distribution nor in median DNA methylation values between cases and controls (Fig. 6). Figure 6 Open in new tabDownload slide Box and whisker plots depicting distribution of DNA methylation levels at individual CpGs as measured using pyrosequencing. Each plot depicts minimum, maximum, median, first and third quartile values for DNA methylation measured in C12ORF10 (MYG1), SDHAP3, ZNF300 and NAPRT1. Figure 6 Open in new tabDownload slide Box and whisker plots depicting distribution of DNA methylation levels at individual CpGs as measured using pyrosequencing. Each plot depicts minimum, maximum, median, first and third quartile values for DNA methylation measured in C12ORF10 (MYG1), SDHAP3, ZNF300 and NAPRT1. Finally, we performed a non-parametric paired analysis for methylation change comparing methylation levels averaged across all CpGs for each gene. There was no compelling evidence of a difference in DNA methylation in matched pairs (α level = 0.05). Discussion The placenta is generally considered to be both necessary and sufficient for the development of preeclampsia, making it the appropriate tissue to use when investigating preeclampsia (PE) pathogenesis. We have designed an EWAS, coupled with a gene expression microarray, in order to discover novel differentially methylated positions and/or regions associated with preeclampsia and their effect on expression levels of genes in proximity. The majority of previous studies exploring molecular markers associated with preeclampsia have focused on the fetal side of the placenta (32,34–38). In contrast, we opted to use the maternal side since preeclampsia is, essentially, an adaptive response of the mother to a hypoperfused placenta. One previous study had focused on the maternal side of the placenta; however, it investigated DNA methylation only at approximately 27 000 CpG sites (12). For the current EWAS, we used the 450K BeadChip, which remains the most popular format for cost-effective DNA methylation analysis. While this array provides comprehensive coverage of the genome, CpG dense regions tend to be overrepresented. In general, we did not identify extensive epigenetic changes associated with case-status. Apart from a few genes, there was little overlap between significant differentially methylated sites/regions observed in our study and previously published data. This may be due to the target population, and the cohort's structure and size. It can also be attributed to discrepancies in terms of study design and bioinformatic methods used for analysis. With one exception (12), previous studies also did not investigate infant sex-specific preeclampsia-associated differences in DNA methylation and gene expression, irrespective of the fact that placental sexual dimorphism is known to exist. Interestingly, in their study, Chu et al. found that only in placental samples from pregnancies in which the infant sex was female can they observe statistically significant DMPs associated with preeclampsia. Using a bump-hunting approach applied to the 450K methylation array, we discovered regions of differential DNA methylation between controls and preeclampsia-affected term placentas at CpG islands within the promoters of C12ORF10 (MYG1, ranked first), SDHAP3 (rank 2), ZNF300 (rank 3), SCUBE2 (rank 4) and NAPRT1 (rank 5) in singletons (Fig. 3). While the SDHAP3 DMR is detected in placentas from both female and male infants, NAPRT1 and MYG1 DMRs appear to be female-specific. NAPRT1 is particularly interesting as it has previously been shown to be downregulated (1.56-fold) in placenta from first trimester pregnancies with high-resistance uterine artery blood flow, which is predictive of later complications such as preeclampsia, among others (39). Nicotinate phosphoribosyltransferase (NAPRT1) is a ubiquitously expressed gene, which is necessary for the conversion of nicotinic acid to nicotinamide adenine dinucleotide and can be epigenetically regulated (40). Indeed, the NAPRT1 TSS/5′-UTR region (chr8: 144 659 627–144 661 051) has been previously reported as hypomethylated (Δβ = 16%) in preeclampsia by Yeung et al. who investigated genome-wide DNA methylation in n = 8 preeclampsia cases (32). The direction and magnitude of the difference in DNA methylation was similar to what we detected for both singletons and twins in an overlapping region (chr8: 144 659 831–144 660 772). Because we also conducted transcriptomic analysis, we observed an inverse association between NAPRT1 gene expression and DNA methylation. Although validation by pyrosequencing did not reveal a difference in DNA methylation in an independent cohort, the median DNA methylation levels of NAPRT1 were markedly lower in preeclampsia cases when compared to controls. Collectively, these data indicate that NAPRT1 and its pathway might be an important target for therapeutic intervention. Melanocyte proliferating gene 1 (MYG1) is a ubiquitously expressed protein with a high degree of sequence conservation across phyla, suggestive of an essential role in the eukaryotic cell. MYG1 was initially reported as a DHH-motif containing phosphoesterase (41) and was only recently demonstrated to function as a novel 3′–5′ RNA exonuclease (42) localized in the nucleolus, nucleoplasm and mitochondrial matrix. It has emerged as a global regulator of cell protein translation, which participates in ribosome assembly and maturation by processing of ribosomal RNAs encoded by both the nuclear and mitochondrial genomes. In addition, Myg1 appears to be required for the proper functioning of mitochondria through its role in the processing and turnover of nuclear-encoded mitochondrial transcripts and by trimming mitochondrial mRNAs in the mitochondrial matrix itself. Oxygen consumption rate, a measure of cellular respiration, is drastically reduced following siRNA-mediated knockdown of Myg1 in mouse melanoma cells and upon knockout of its orthologous gene in yeast, indicating a reduction in oxidative phosphorylation. Thus, Myg1 has been observed as a critical mediator of the crosstalk between the nucleus and the mitochondria and may be implicated in diseases in which mitochondrial dysfunction accompanies disease progression (42). This gene’s only known disease association, so far, is with vitiligo (43) where MYG1 is downregulated in patient-derived epidermal lesions with concomitant upregulation of genes for ribosomal proteins and mitoribosomes, OXPHOS and mitochondrial proteins (42). Two adjacent DMRs covering exon1/TSS of HSPA1L were ranked first (chr6: 31 783 240–31 783 545, P = 3.56e − 05) and third (chr6: 31 782 711–31 783 029, P = 1.27e − 04), respectively, among twins. Notably, ZNF300 and HSPA1L were previously identified as differentially methylated in placentas from severe growth-discordant monochorionic twins (44). In that study, a gain of DNA methylation within the promoters of ZNF300 (chr5: 150 284 416–150 284 796) and HSPA1L (chr6: 31 783 322–31 783 482) was reported in the intrauterine growth restriction (IUGR)-affected twin, a relevant finding because IUGR is a frequent adverse outcome of preeclampsia. These regions overlap with the DMRs we have identified as hypermethylated in preeclamptic placentas, namely ZNF300 chr5: 150 284 302–150 284 796 and HSPA1L chr6: 31 783 240–31 783 545. However, we were unable to validate differential DNA methylation associated with preeclampsia in an independent cohort using pyrosequencing of ZNF300. Similar to what we observed from the 450K methylation array, for SDHAP3 pyrosequencing revealed that some samples, mostly preeclampsia cases but also some controls, were completely demethylated at the analyzed CpGs. SDHAP3 is a pseudogene of the flavoprotein gene SDHA, which belongs to the mitochondrial succinate dehydrogenase complex (45). Until recently, pseudogenes were considered as mere relics of genes, which have undergone changes to their open reading frames and/or promoter activity rendering them non-functional genomic sequences. However, it has become evident that pseudogenes can be ubiquitously expressed or cell type specific, as well as differentially expressed in certain pathological conditions. They can function as competing endogenous RNAs or are translated into functional peptides (46). Although a function for SDHAP3 has yet to be described, recently, hypomethylation of the promoter-associated CpG island of SDHAP3 was reported in the prefrontal cortex of the fetal brain following in utero exposure to maternal smoking (47). However, in that study, there was no corresponding change in SDHAP3 expression between exposed and non-exposed samples. Similarly, we did not observe an association between SDHAP3 expression and preeclampsia. A recent EWAS investigated epigenetic signatures of pre-term birth (PTB), which persist later in life in identical adult twins (48). Birth-related information on these samples are recorded as part of the Danish Medical Birth Register (DMBR), which, among other characteristics, includes information on maternal medical conditions and pregnancy complications (49). Using the 450K BeadChip and a bump-hunting approach similar to our approach, this study identified three DMRs in whole blood that were associated with PTB. Among them, the most significant one was the promoter region of SDHAP3, which the authors observed to be hypomethylated in twins born prematurely (before 37 weeks of gestational age). The loss of DNA methylation within this DMR was up to 80% (mean loss of 50%). This DMR is identical to the one that we report on here (chr5: 1 594 282–1 594 863) and that was differentially methylated in our preeclamptic placenta samples, irrespective of infant sex. It is notable that a multi-fetal gestation is a risk factor for preeclampsia; however, preeclamptic pregnancies were not explicitly excluded from Tan et al.’s study design. This makes it difficult to assess whether promoter hypomethylation of SDHAP3 is an epigenetic biomarker of preeclampsia, of PTB, or both. In order to complement the DNA methylation analysis, we used a comprehensive microarray to capture gene-level and exon-level expression changes associated with preeclampsia. The non-coding MIR138-1 was modestly upregulated in preeclamptic versus control placentas in singleton pregnancies. Aside from its function in a variety of cancers, this miRNA has previously been found to regulate hypoxia-induced endothelial cell dysfunction (50,51). Exposure of endothelial cells to proinflammatory cytokines leads to stabilization of hypoxia-inducible factor-1a (HIF1a), followed by upregulation of MIR138-1 and repression of S100A1, thus triggering endothelial dysfunction. Functional analysis has also revealed that HIF1a is a direct target for repression by MIR138-1. Finally, in HeLa cells, MIR138 functions as a negative regulator of general miRNA expression through inhibition of pre-miRNA nuclear export (52). It appears that MIR138 is a multifunctional regulator of many cellular processes, some of which are relevant to the pathophysiology of preeclampsia such as cell migration, epithelial-to-mesenchymal transition, cell cycle progression and cell differentiation (53). A limitation of this study is the relatively small sample size, which reduces the probability of discovering DMPs with genome-wide significance; however, among studies which evaluate preeclamptic pregnancies, ours is still one of the largest to integrate DNA methylation and gene expression profiling. Further, the samples we used are from women who have experienced a mild form of preeclampsia, as demonstrated by the late gestational age at delivery. We note that the validation samples were from full-thickness placenta obtained from the region close to the cord. Differences in sampling depth and the region of the placenta from which the sample was obtained (cord, middle, or periphery) can result in altered cell composition. Such differences can manifest themselves as intraplacental variability in gene expression and DNA methylation; however, this appears to be mostly gene-specific. Indeed, such variability has been observed for some genes (54), but not others (55,56). Finally, as with other EWAS, our investigation is performed months after disease onset and as such can only capture markers associated with the maternal adaptive response to a pathological situation, depriving them of predictive value. An important consideration for EWAS design is tissue heterogeneity. In many instances, it is appropriate to control for cellular composition because DNA methylation signatures from irrelevant cell types can obscure true disease-associated variation. However, it is becoming increasingly apparent that many diseases are caused by changes in the normal cellular proportions and composition within the tissue. Thus, studies in mixed cell populations are informative (57). This might be particularly relevant to preeclampsia, where the underlying early cellular defect is extravillous trophoblast differentiation gone awry. The later clinical manifestations are a maladaptation occurring as a consequence. In this context, for observable changes in DNA methylation signals associated with preeclampsia to be informative, they should reflect these anticipated changes in placental physiology. In conclusion, we have performed a comprehensive investigation into molecular variation (DNA methylation and gene expression) associated with preeclampsia. We rigorously matched cases and controls, exhaustively modeling the results from the DNA methylation analysis in order to describe changes associated with preeclampsia and validated our findings in an independent cohort of individuals. Finally, we have integrated two layers of gene expression regulation, namely transcriptional and epigenetic, in order to describe gene-level dysregulation on the maternal side of the placenta, which is associated with preeclampsia, and its potential correlation with differential DNA methylation. We have identified targets that are plausible early markers of the disease, such as NAPRT1, and others that likely reflect placental maladaptation, such as MIR138. Materials and Methods Study population Our primary study population consisted of women enrolled in the Harvard Epigenetic Birth Cohort (HEBC), which includes a total of 1941 mother–child dyads. Biospecimens were collected from June 2007 through June 2009 on the labor and delivery floor of the Department of Obstetrics, Gynecology and Reproductive Biology at Brigham and Women’s Hospital in Boston, Massachusetts, USA. Cases were mothers with either chart reported preeclampsia or had chart reported pregnancy-induced hypertension and proteinuria. Preeclampsia was defined as per International Society for the Study of Hypertension in Pregnancy (ISSHP) criteria as onset of high blood pressure (>140/90 mm Hg) and proteinuria (>0.3 g/24 h) after 20 weeks of gestation (58). Maternal ages of the controls were within 5 years of the case, with the same method of conception (spontaneous planned, spontaneous unplanned, in vitro fertilization), ethnicity (non-Hispanic Caucasian, Hispanic/Latino, Black/African-American), smoking during pregnancy (yes/no), infant sex, and the closest pre-pregnancy BMI (kg/m2). For difficult to match cases with very high pre-pregnancy BMI, we relaxed the matching criteria, ignoring infant sex and method of conception. Permission to use these samples for the study of epigenetic and transcriptomic determinants of preeclampsia was granted following review by Partners Human Research IRB Committee at Brigham and Women’s Hospital in Boston, Massachusetts (2006P001665/BWH). Biospecimen collection In HEBC, placental tissue samples were harvested from (i) the upper (fetal) side of the placenta near the umbilical cord; (ii) the placental perimeter on the same, fetal, side; (iii) the opposite lower (maternal) side near the cord and (iv) the placental perimeter on the same, maternal, side. These were collected for DNA and RNA isolation. Placental tissues for DNA extraction were snap-frozen and stored in liquid nitrogen. Tissues for RNA extraction were stored in RNAlater (Ambion, Carlsbad, CA) at −20°C until further processing. For the present study, we used placental tissue from the lower (maternal) side of the placenta directly opposite the site of umbilical cord insertion. Validation cohort In order to validate DMRs identified from the Illumina 450K BeadChip, we used samples from the BIS, a cohort assembled in the south-east of Australia using an unselected sampling frame (26). We obtained genomic DNA from a total of n = 32 full-thickness placentas from singleton pregnancies, obtained from the region close to the cord. The samples were case–control pairs (n = 16), matched on maternal age, gestational age, fetal sex, pre-pregnancy BMI (or BMI category) and smoking status, but not ethnicity. Preeclampsia was defined as above. The study was approved by Barwon Health Human Research and Ethics Committee (HREC 10/24). Transcriptomic analysis RNA was isolated using the RNeasy Mini kit (Qiagen). Quality control was performed using the Agilent 2200 TapeStation. We identified any sample with a RIN ≥ 6 as appropriate for use on the expression array. Global gene expression was assessed using the Gene Chip® Human Transcriptome Array 2.0 (Affymetrix). The microarray workflow was carried out at Dana Farber Cancer Institute Molecular Biology Core Facility. Briefly, the Affymetrix Gene Chip WT Plus Reagent kit was used to convert 100 ng of total RNA into biotinylated sense-strand DNA targets, which were hybridized onto Affymetrix Human Transcriptome Arrays 2.0. The arrays were washed and stained on Fluidics Stations 450 using the Affymetrix Hybridization/Wash/Stain buffer. Finally, the arrays were scanned using Gene Chip Scanner 3000 7G. DNA isolation Genomic DNA was isolated using DNeasy Blood and Tissue kit (Qiagen) from the maternal side of placenta. Briefly, a small tissue biopsy was incubated at 56°C in 180 uL lysis buffer supplemented with 20 uL Proteinase K solution for 2 h or until the tissue was completely dissolved. As per the manufacturer’s protocol, 200uL AL Buffer was pre-mixed with 100% ethanol in 1:1 ratio and added to each sample. The samples were then loaded onto the supplied columns and washed. DNA quantity and quality were measured on a NanoDrop 1000 spectrophotometer. Genome-wide DNA methylation Genome-wide profiling of DNA methylation was performed using the Infinium Human Methylation 450K Bead Chip at the University of Southern California Norris Comprehensive Cancer Center Molecular Genomics Core Facility. Briefly, 1 μg genomic DNA for each sample was treated with sodium bisulfite, recovered using the Zymo EZ DNA Methylation kit (Zymo Research, Irvine, CA) according to the manufacturer’s specifications and eluted in 18 uL volume. An aliquot (3 uL) was removed for MethyLight-based quality control testing of bisulfite conversion completeness and the amount of bisulfite converted DNA available for the Infinium Methylation assay. All samples that passed the QC tests entered into the Infinium DNA methylation assay data production pipeline. Pyrosequencing Validation of DMRs was done by bisulfite pyrosequencing on a PyroMark Q24 instrument. Genomic DNA samples for validation were isolated from full thickness placental biopsies collected from participants in the BIS. Primers for all pyrosequencing assays were designed using the PyroMark Assay Design software 2.0 and purchased from Integrated DNA Technologies (IDT). All biotinylated primers were HPLC purified. Pyrosequencing was carried out according to the manufacturer’s instructions (Qiagen). Briefly, biotin-labeled PCR products were incubated with Streptavidin Sepharose High Performance Beads (GE Healthcare Life Sciences) in the presence of PyroMark Binding Buffer for 5–10 min at room temperature and constant agitation. The beads were then captured by the vacuum tool of the PyroMark workstation and washed consecutively in 70% ethanol, PyroMark Denaturation Solution and 1X PyroMark Wash Buffer. The biotin-tagged single-strand DNA was released onto the wells of PyroMark Q24 plates containing the respective sequencing primer diluted to a final concentration of 300 mm in PyroMark Annealing Buffer. The plate was then incubated at 80°C/2 min and allowed to cool to room temperature before being placed into the PyroMark Q24 instrument. Pyrosequencing was performed using nucleotides, enzyme and substrate solutions provided in PyroMark Gold Q24 Reagents, which were loaded in a PyroMark Q24 cartridge. Statistical analysis Expression intensities were processed prior to analysis using the Affymetrix Expression Console Software (Affymetrix). Background correction was performed using the Robust Multichip Analysis (RMA) algorithm, to minimize the variance seen across arrays (59). These probe values were then quantile normalized and summarized into one gene-level expression measure using median polish (59,60). The technical triplicates for two random samples were highly reproducible (Supplementary Material, Fig. S1). Methylation microarray preprocessing was performed using minfi (61). Background correction was performed using ‘out-of-band’ signal from type I probes, assuming a normal-exponential convolution of signal and background (‘noob’ method) (62). To remove additional unwanted technical variation, background correction was followed by functional normalization (63). Probes containing a SNP at the target CpG (minor allele frequency > 0.05) were removed prior to analysis. The fluorescence intensities from methylated (M) and unmethylated (U) alleles were then converted to methylation levels, ranging from 0 to 1, given by β = M/(M + U + 100). Independent surrogate variable analysis (ISVA) was used to estimate batch effects from both the expression and methylation array data (64). This batch correction approach was chosen to reduce the likelihood of controlling for latent noise reflecting biologically relevant variation, given the suspected impact of preeclampsia on trophoblast development. Unlike classical surrogate variable analysis, ISVA restricts to surrogate variables associated with pre-specified confounders, potentially measured with error/uncertainty. None of the surrogate variables estimated from the expression data were associated with our indicator of batch (plate). Therefore, we did not adjust for any of the estimated surrogate variables in the expression models. In contrast, two surrogate variables estimated from the methylation data were associated with methylation chip and bisulfite conversion plate and were adjusted for in all subsequent methylation models. Singletons and twins were analyzed separately. The association between gene-level expression and preeclampsia among singletons was analyzed using the empirical Bayes method implemented in the R limma package to moderate standard errors (30). Singleton models were adjusted for maternal age (years), pre-pregnancy BMI (kg/m2) and ethnicity. Due to a small sample size, analyses of twins were only adjusted for maternal age and matched pair. Limma was also used to model the association between site-specific methylation and preeclampsia in singletons and twins, adjusting for maternal age, pre-pregnancy BMI and ethnicity, as well as the estimated batch effects. Regional changes in methylation were identified using a ‘bump-hunting’ approach (65). The coefficients from our adjusted models to identify DMRs among contiguous loci within 500 bp that exceed the 99th percentile of genome-wide changes, summarizing these regions based on the effect size. A bootstrapping (N replicates = 1000) approach was used to estimate the statistical significance. To estimate possible sex-specific associations between gene regulation and preeclampsia, an interaction term between preeclampsia and infant sex was then added to each of the above models. In the validation cohort, the association between percent methylation measured by pyrosequencing and preeclampsia was assessed using non-parametric paired tests (Wilcoxon signed-rank tests) to account for the matched design. All analysis was conducted in R 3.4.1 and visualized using ggplot2. Acknowledgements The authors would like to thank the participants in the Harvard Epigenetic Birth Cohort for their contribution to this project. We also thank current and past staff for their efforts in recruiting and maintaining the cohort and in obtaining and processing the data and biospecimens. We thank the BIS participants for the generous contribution they have made to this project. The establishment work and infrastructure for the BIS were provided by the Murdoch Children’s Research Institute, Deakin University and Barwon Health. Subsequent funding was secured from the National Health and Medical Research Council of Australia, The Shepherd Foundation, The Jack Brockhoff Foundation, the Scobie Trust, the Shane O’Brien Memorial Asthma Foundation, the Our Women’s Our Children’s Fund Raising Committee Barwon Health, the Rotary Club of Geelong, the Ilhan Food Allergy Foundation, GMHBA Ltd, The Gandel Foundation, The Percy Baxter Charitable Trust, Perpetual Trustees and the Gwenyth Raymond Trust. In-kind support was provided by the Cotton on Foundation and CreativeForce. Research at Murdoch Children’s Research Institute is supported by the Victorian Government’s Operational Infrastructure Support Program. A.-L. Ponsonby and P. Vuillermin receive NHMRC fellowship support. The BIS Investigator Group includes Mimi Tang and Len Harrison. Conflict of Interest Statement: The authors declare no conflict of interest. Author Contributions M.N.L. and K.B.M. conceived the study. K.B.M. provided the funding for this work. R.S., P.V. and A.L.P. provided the samples for DNA methylation validation. M.N.L. processed the samples for the microarray and 450K DNA methylation analysis, designed and performed the pyrosequencing assays. A.M.B. performed the transcriptomic and epigenetic data analysis. M.N.L., A.M.B. and K.B.M. wrote the manuscript with critical input from R.S., P.V. and A.L.P. All authors participated in revising the manuscript. References 1. Kuklina , E.V. , Ayala , C. and Callaghan , W.M. ( 2009 ) Hypertensive disorders and severe obstetric morbidity in the United States . Obstet. Gynecol. , 113 , 1299 – 1306 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Wadhwa , P.D. , Buss , C. , Entringer , S. and Swanson , J.M. ( 2009 ) Developmental origins of health and disease: brief history of the approach and current focus on epigenetic mechanisms . Semin. Reprod. Med. , 27 , 358 – 368 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Roberts , J.M. ( 2014 ) Pathophysiology of ischemic placental disease . Semin. Perinatol. , 38 , 139 – 145 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Chaiworapongsa , T. , Chaemsaithong , P. , Yeo , L. and Romero , R. ( 2014 ) Pre-eclampsia part 1: current understanding of its pathophysiology . Nat. Rev. Nephrol. , 10 , 466 – 480 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Vatten , L.J. and Skjaerven , R. ( 2004 ) Offspring sex and pregnancy outcome by length of gestation . Early Hum. Dev. , 76 , 47 – 54 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Shiozaki , A. , Matsuda , Y. , Satoh , S. and Saito , S. ( 2011 ) Impact of fetal sex in pregnancy-induced hypertension and preeclampsia in Japan . J. Reprod. Immunol. , 89 , 133 – 139 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Nielsen , H.S. , Steffensen , R. , Lund , M. , Egestad , L. , Mortensen , L.H. , Andersen , A.M. , Lidegaard , Ø. and Christiansen , O.B. ( 2010 ) Frequency and impact of obstetric complications prior and subsequent to unexplained secondary recurrent miscarriage . Hum. Reprod. , 25 , 1543 – 1552 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Clifton , V.L. ( 2010 ) Review: sex and the human placenta: mediating differential strategies of fetal growth and survival . Placenta , 31 ( suppl ), S33 – S39 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Brown , Z.A. , Schalekamp-Timmermans , S. , Tiemeier , H.W. , Hofman , A. , Jaddoe , V.W. and Steegers , E.A. ( 2014 ) Fetal sex specific differences in human placentation: a prospective cohort study . Placenta , 35 , 359 – 364 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Wang , Y. , Pringle , K.G. , Sykes , S.D. , Marques , F.Z. , Morris , B.J. , Zakar , T. and Lumbers , E.R. ( 2012 ) Fetal sex affects expression of renin-angiotensin system components in term human decidua . Endocrinology , 153 , 462 – 468 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Gong , S. , Johnson , M.D. , Dopierala , J. , Gaccioli , F. , Sovio , U. , Constância , M. , Smith , G.C. and Charnock-Jones , D.S. ( 2018 ) Genome-wide oxidative bisulfite sequencing identifies sex-specific methylation differences in the human placenta . Epigenetics , 13 , 228 – 239 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Chu , T. , Bunce , K. , Shaw , P. , Shridhar , V. , Althouse , A. , Hubel , C. and Peters , D. ( 2014 ) Comprehensive analysis of preeclampsia-associated DNA methylation in the placenta . PLoS One , 9 , e107318 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Branco , M.R. , King , M. , Perez-Garcia , V. , Bogutz , A.B. , Caley , M. , Fineberg , E. , Lefebvre , L. , Cook , S.J. , Dean , W. , Hemberger , M. et al. ( 2016 ) Maternal DNA methylation regulates early trophoblast development . Dev Cell , 36 , 152 – 163 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Zhu , D. , Hölz , S. , Metzger , E. , Pavlovic , M. , Jandausch , A. , Jilg , C. , Galgoczy , P. , Herz , C. , Moser , M. , Metzger , D. et al. ( 2014 ) Lysine-specific demethylase 1 regulates differentiation onset and migration of trophoblast stem cells . Nat. Commun. , 5 , 3174 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Schroeder , D.I. , Blair , J.D. , Lott , P. , Yu , H.O. , Hong , D. , Crary , F. , Ashwood , P. , Walker , C. , Korf , I. , Robinson , W.P. et al. ( 2013 ) The human placenta methylome . Proc. Natl. Acad. Sci. USA. , 110 , 6037 – 6042 . Google Scholar Crossref Search ADS WorldCat 16. Novakovic , B. , Yuen , R.K. , Gordon , L. , Penaherrera , M.S. , Sharkey , A. , Moffett , A. , Craig , J.M. , Robinson , W.P. and Saffery , R. ( 2011 ) Evidence for widespread changes in promoter methylation profile in human placenta in response to increasing gestational age and environmental/stochastic factors . BMC Genomics , 12 , 529 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Januar , V. , Desoye , G. , Novakovic , B. , Cvitic , S. and Saffery , R. ( 2015 ) Epigenetic regulation of human placental function and pregnancy outcome: considerations for causal inference . Am. J. Obstet. Gynecol. , 213 ( 4 Suppl ), S182 – S196 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Binder , A.M. , LaRocca , J. , Lesseur , C. , Marsit , C.J. and Michels , K.B. ( 2015 ) Epigenome-wide and transcriptome-wide analyses reveal gestational diabetes is associated with alterations in the human leukocyte antigen complex . Clin Epigenetics , 7 , 79 . Google Scholar Crossref Search ADS PubMed WorldCat 19. LaRocca , J. , Binder , A.M. , McElrath , T.F. and Michels , K.B. ( 2014 ) The impact of first trimester phthalate and phenol exposure on IGF2/H19 genomic imprinting and birth outcomes . Environ. Res. , 133 , 396 – 406 . Google Scholar Crossref Search ADS PubMed WorldCat 20. van Otterdijk , S.D. , Binder , A.M. and Michels , K.B. ( 2017 ) Locus-specific DNA methylation in the placenta is associated with levels of pro-inflammatory proteins in cord blood and they are both independently affected by maternal smoking during pregnancy . Epigenetics , 0 . OpenURL Placeholder Text WorldCat 21. Fong , F.M. , Sahemey , M.K. , Hamedi , G. , Eyitayo , R. , Yates , D. , Kuan , V. , Thangaratinam , S. and Walton , R.T. ( 2014 ) Maternal genotype and severe preeclampsia: a HuGE review . Am. J. Epidemiol. , 180 , 335 – 345 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Lie , R.T. , Rasmussen , S. , Brunborg , H. , Gjessing , H.K. , Lie-Nielsen , E. and Irgens , L.M. ( 1998 ) Fetal and maternal contributions to risk of pre-eclampsia: population based study . BMJ , 316 , 1343 – 1347 . Google Scholar Crossref Search ADS PubMed WorldCat 23. Esplin , M.S. , Fausett , M.B. , Fraser , A. , Kerber , R. , Mineau , G. , Carrillo , J. and Varner , M.W. ( 2001 ) Paternal and maternal components of the predisposition to preeclampsia . N. Engl. J. Med. , 344 , 867 – 872 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Acosta-Sison , H. ( 1957 ) Etiology of preeclampsia-eclampsia and the role of hydatidiform mole . Obstet. Gynecol. , 9 , 233 – 236 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 25. Michels , K.B. , Harris , H.R. and B.L. ( 2011 ) Birthweight, maternal weight trajectories and global DNA methylation of LINE-1 repetitive elements . PLoS One , 6 , e25254 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Vuillermin , P. , Saffery , R. , Allen , K.J. , Carlin , J.B. , Tang , M.L. , Ranganathan , S. , Burgner , D. , Dwyer , T. , Collier , F. , Jachno , K. et al. ( 2015 ) Cohort profile: the Barwon infant study . Int. J. Epidemiol. , 44 , 1148 – 1160 . Google Scholar Crossref Search ADS PubMed WorldCat 27. Xu , W. , Seok , J. , Mindrinos , M.N. , Schweitzer , A.C. , Jiang , H. , Wilhelmy , J. , Clark , T.A. , Kapur , K. , Xing , Y. , Faham , M. et al. ( 2011 ) Human transcriptome array for high-throughput clinical studies . Proc. Natl. Acad. Sci. USA. , 108 , 3707 – 3712 . Google Scholar Crossref Search ADS WorldCat 28. Ashar-Patel , A. , Kaymaz , Y. , Rajakumar , A. , Bailey , J.A. , Karumanchi , S.A. and Moore , M.J. ( 2017 ) FLT1 and transcriptome-wide polyadenylation site (PAS) analysis in preeclampsia . Sci. Rep. , 7 , 12139 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Løset , M. , Mundal , S.B. , Johnson , M.P. , Fenstad , M.H. , Freed , K.A. , Lian , I.A. , Eide , I.P. , Bjørge , L. , Blangero , J. , Moses , E.K. et al. ( 2011 ) A transcriptional profile of the decidua in preeclampsia . Am. J. Obstet. Gynecol. , 204 , 84 , e1-27 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Smyth , G.K. ( 2004 ) Linear models and empirical bayes methods for assessing differential expression in microarray experiments . Stat. Appl. Genet. Mol. Biol. , 3 , 1 – 25 . Google Scholar Crossref Search ADS WorldCat 31. Benjamini , Y. and Hochberg , Y. ( 1995 ) Controlling the false discovery rate: a practical and powerful approach to multiple testing . Journal of the Royal Statistical Society. Series B (Methodological) , 57 , 289 – 300 . Google Scholar Crossref Search ADS WorldCat 32. Yeung , K.R. , Chiu , C.L. , Pidsley , R. , Makris , A. , Hennessy , A. and Lind , J.M. ( 2016 ) DNA methylation profiles in preeclampsia and healthy control placentas . Am. J. Physiol. Heart. Circ. Physiol. , 310 , H1295 – H1303 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Michels , K.B. , Binder , A.M. , Dedeurwaerder , S. , Epstein , C.B. , Greally , J.M. , Gut , I. , Houseman , E.A. , Izzi , B. , Kelsey , K.T. , Meissner , A. et al. ( 2013 ) Recommendations for the design and analysis of epigenome-wide association studies . Nat. Methods. , 10 , 949 – 955 . Google Scholar Crossref Search ADS PubMed WorldCat 34. Anton , L. , Brown , A.G. , Bartolomei , M.S. and Elovitz , M.A. ( 2014 ) Differential methylation of genes associated with cell adhesion in preeclamptic placentas . PLoS One , 9 , e100148 . Google Scholar Crossref Search ADS PubMed WorldCat 35. Blair , J.D. , Yuen , R.K. , Lim , B.K. , McFadden , D.E. , von Dadelszen , P. and Robinson , W.P. ( 2013 ) Widespread DNA hypomethylation at gene enhancer regions in placentas associated with early-onset pre-eclampsia . Mol. Hum. Reprod. , 19 , 697 – 708 . Google Scholar Crossref Search ADS PubMed WorldCat 36. Anderson , C.M. , Ralph , J.L. , Wright , M.L. , Linggi , B. and Ohm , J.E. ( 2014 ) DNA methylation as a biomarker for preeclampsia . Biol. Res. Nurs. , 16 , 409 – 420 . Google Scholar Crossref Search ADS PubMed WorldCat 37. Herzog , E.M. , Eggink , A.J. , Willemsen , S.P. , Slieker , R.C. , Wijnands , K.P.J. , Felix , J.F. , Chen , J. , Stubbs , A. , van der , P.J. , van Meurs , J.B. et al. ( 2017 ) Early- and late-onset preeclampsia and the tissue-specific epigenome of the placenta and newborn . Placenta , 58 , 122 – 132 . Google Scholar Crossref Search ADS PubMed WorldCat 38. Yuen , R.K. , Peñaherrera , M.S. , von Dadelszen , P. , McFadden , D.E. and Robinson , W.P. ( 2010 ) DNA methylation profiling of human placentas reveals promoter hypomethylation of multiple genes in early-onset preeclampsia . Eur. J. Hum. Genet. , 18 , 1006 – 1012 . Google Scholar Crossref Search ADS PubMed WorldCat 39. Leslie , K. , Whitley , G.S. , Herse , F. , Dechend , R. , Ashton , S.V. , Laing , K. , Thilaganathan , B. and Cartwright , J.E. ( 2015 ) Increased apoptosis, altered oxygen signaling, and antioxidant defenses in first-trimester pregnancies with high-resistance uterine artery blood flow . Am. J. Pathol. , 185 , 2731 – 2741 . Google Scholar Crossref Search ADS PubMed WorldCat 40. Duarte-Pereira , S. , Pereira-Castro , I. , Silva , S.S. , Correia , M.G. , Neto , C. , da Costa , L.T. , Amorim , A. and Silva , R.M. ( 2016 ) Extensive regulation of nicotinate phosphoribosyltransferase (NAPRT) expression in human tissues and tumors . Oncotarget , 7 , 1973 – 1983 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Philips , M.A. , Vikeså , J. , Luuk , H. , Jønson , L. , Lilleväli , K. , Rehfeld , J.F. , Vasar , E. , Kõks , S. and Nielsen , F.C. ( 2009 ) Characterization of MYG1 gene and protein: subcellular distribution and function . Biol. Cell. , 101 , 361 – 373 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Grover , R. , Burse , S.A. , Shankrit , S. , Aggarwal , A. , Kirty , K. , Narta , K. , Srivastav , R. , Ray , A.K. , Malik , G. , Vats , A. et al. ( 2019 ) Myg1 exonuclease couples the nuclear and mitochondrial translational programs through RNA processing . Nucleic. Acids. Res. , 47 , 5852 – 5866 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Philips , M.A. , Kingo , K. , Karelson , M. , Rätsep , R. , Aunin , E. , Reimann , E. , Reemann , P. , Porosaar , O. , Vikeså , J. , Nielsen , F.C. et al. ( 2010 ) Promoter polymorphism -119C/G in MYG1 (C12orf10) gene is related to vitiligo susceptibility and Arg4Gln affects mitochondrial entrance of Myg1 . BMC. Med. Genet. , 11 , 56 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Roifman , M. , Choufani , S. , Turinsky , A.L. , Drewlo , S. , Keating , S. , Brudno , M. , Kingdom , J. and Weksberg , R. ( 2016 ) Genome-wide placental DNA methylation analysis of severely growth-discordant monochorionic twins reveals novel epigenetic targets for intrauterine growth restriction . Clin Epigenetics , 8 , 70 . Google Scholar Crossref Search ADS PubMed WorldCat 45. Yankovskaya , V. , Horsefield , R. , Törnroth , S. , Luna-Chavez , C. , Miyoshi , H. , Léger , C. , Byrne , B. , Cecchini , G. and Iwata , S. ( 2003 ) Architecture of succinate dehydrogenase and reactive oxygen species generation . Science , 299 , 700 – 704 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Ji , Z. , Song , R. , Regev , A. and Struhl , K. ( 2015 ) Many lncRNAs, 5′UTRs, and pseudogenes are translated and some are likely to express functional proteins . Elife , 4 , e08890 . Google Scholar Crossref Search ADS PubMed WorldCat 47. Chatterton , Z. , Hartley , B.J. , Seok , M.H. , Mendelev , N. , Chen , S. , Milekic , M. , Rosoklija , G. , Stankov , A. , Trencevsja-Ivanovska , I. , Brennand , K. et al. ( 2017 ) In utero exposure to maternal smoking is associated with DNA methylation alterations and reduced neuronal content in the developing fetal brain . Epigenetics Chromatin , 10 , 4 . Google Scholar Crossref Search ADS PubMed WorldCat 48. Tan , Q. , Li , S. , Frost , M. , Nygaard , M. , Soerensen , M. , Larsen , M. , Christensen , K. and Christiansen , L. ( 2018 ) Epigenetic signature of preterm birth in adult twins . Clin Epigenetics , 10 , 87 . Google Scholar Crossref Search ADS PubMed WorldCat 49. Bliddal , M. , Broe , A. , Pottegård , A. , Olsen , J. and Langhoff-Roos , J. ( 2018 ) The Danish medical birth register . Eur. J. Epidemiol. , 33 , 27 – 36 . Google Scholar Crossref Search ADS PubMed WorldCat 50. Sen , A. , Ren , S. , Lerchenmüller , C. , Sun , J. , Weiss , N. , Most , P. and Peppel , K. ( 2013 ) MicroRNA-138 regulates hypoxia-induced endothelial cell dysfunction by targeting S100A1 . PLoS One , 8 , e78684 . Google Scholar Crossref Search ADS PubMed WorldCat 51. Sen , A. , Most , P. and Peppel , K. ( 2014 ) Induction of microRNA-138 by pro-inflammatory cytokines causes endothelial cell dysfunction . FEBS Lett. , 588 , 906 – 914 . Google Scholar Crossref Search ADS PubMed WorldCat 52. Li , J. , Chen , Y. , Qin , X. , Wen , J. , Ding , H. , Xia , W. , Li , S. , Su , X. , Wang , W. , Li , H. et al. ( 2014 ) MiR-138 downregulates miRNA processing in HeLa cells by targeting RMND5A and decreasing Exportin-5 stability . Nucleic. Acids. Res. , 42 , 458 – 474 . Google Scholar Crossref Search ADS PubMed WorldCat 53. Jin , Y. , Chen , D. , Cabay , R.J. , Wang , A. , Crowe , D.L. and Zhou , X. ( 2013 ) Role of microRNA-138 as a potential tumor suppressor in head and neck squamous cell carcinoma . Int. Rev. Cell. Mol. Biol. , 303 , 357 – 385 . Google Scholar Crossref Search ADS PubMed WorldCat 54. Avila , L. , Yuen , R.K. , Diego-Alvarez , D. , Peñaherrera , M.S. , Jiang , R. and Robinson , W.P. ( 2010 ) Evaluating DNA methylation and gene expression variability in the human term placenta . Placenta , 31 , 1070 – 1077 . Google Scholar Crossref Search ADS PubMed WorldCat 55. Non , A.L. , Binder , A.M. , Barault , L. , Rancourt , R.C. , Kubzansky , L.D. and Michels , K.B. ( 2012 ) DNA methylation of stress-related genes and LINE-1 repetitive elements across the healthy human placenta . Placenta , 33 , 183 – 187 . Google Scholar Crossref Search ADS PubMed WorldCat 56. Janssen , B.G. , Byun , H.M. , Cox , B. , Gyselaers , W. , Izzi , B. , Baccarelli , A.A. and Nawrot , T.S. ( 2014 ) Variation of DNA methylation in candidate age-related targets on the mitochondrial-telomere axis in cord blood and placenta . Placenta , 35 , 665 – 672 . Google Scholar Crossref Search ADS PubMed WorldCat 57. Holbrook , J.D. , Huang , R.C. , Barton , S.J. , Saffery , R. and Lillycrop , K.A. ( 2017 ) Is cellular heterogeneity merely a confounder to be removed from epigenome-wide association studies? Epigenomics , 9 , 1143 – 1150 . Google Scholar Crossref Search ADS PubMed WorldCat 58. Tranquilli , A.L. , Dekker , G. , Magee , L. , Roberts , J. , Sibai , B.M. , Steyn , W. , Zeeman , G.G. and Brown , M.A. ( 2014 ) The classification, diagnosis and management of the hypertensive disorders of pregnancy: a revised statement from the ISSHP . Pregnancy Hypertens , 4 , 97 – 104 . Google Scholar Crossref Search ADS PubMed WorldCat 59. Irizarry , R.A. , Hobbs , B. , Collin , F. , Beazer-Barclay , Y.D. , Antonellis , K.J. , Scherf , U. and Speed , T.P. ( 2003 ) Exploration, normalization, and summaries of high density oligonucleotide array probe level data . Biostatistics , 4 , 249 – 264 . Google Scholar Crossref Search ADS PubMed WorldCat 60. Bolstad , B.M. , Irizarry , R.A. , Astrand , M. and Speed , T.P. ( 2003 ) A comparison of normalization methods for high density oligonucleotide array data based on variance and bias . Bioinformatics , 19 , 185 – 193 . Google Scholar Crossref Search ADS PubMed WorldCat 61. Aryee , M.J. , Jaffe , A.E. , Corrada-Bravo , H. , Ladd-Acosta , C. , Feinberg , A.P. , Hansen , K.D. and I.R.A. ( 2014 ) Minfi: a flexible and comprehensive bioconductor package for the analysis of Infinium DNA methylation microarrays . Bioinformatics , 30 , 1363 – 1369 . Google Scholar Crossref Search ADS PubMed WorldCat 62. Triche , T.J. , Weisenberger , D.J. , Van Den Berg , D. , Laird , P.W. and Siegmund , K.D. ( 2013 ) Low-level processing of Illumina Infinium DNA methylation BeadArrays . Nucleic. Acids. Res. , 41 , e90 . Google Scholar Crossref Search ADS PubMed WorldCat 63. Fortin , J.P. , Labbe , A. , Lemire , M. , Zanke , B.W. , Hudson , T.J. , Fertig , E.J. , Greenwood , C.M. and Hansen , K.D. ( 2014 ) Functional normalization of 450K methylation array data improves replication in large cancer studies . Genome. Biol. , 15 , 503 . Google Scholar Crossref Search ADS PubMed WorldCat 64. Teschendorff , A.E. , Zhuang , J. and Widschwendter , M. ( 2011 ) Independent surrogate variable analysis to deconvolve confounding factors in large-scale microarray profiling studies . Bioinformatics , 27 , 1496 – 1505 . Google Scholar Crossref Search ADS PubMed WorldCat 65. Jaffe , A.E. , Murakami , P. , Lee , H. , Leek , J.T. , Fallin , M.D. , Feinberg , A.P. and Irizarry , R.A. ( 2012 ) Bump hunting to identify differentially methylated regions in epigenetic epidemiology studies . Int. J. Epidemiol. , 41 , 200 – 209 . Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2019. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Differential gene expression and limited epigenetic dysregulation at the materno-fetal interface in preeclampsia
JO - Human Molecular Genetics
DO - 10.1093/hmg/ddz287
DA - 2020-01-15
UR - https://www.deepdyve.com/lp/oxford-university-press/differential-gene-expression-and-limited-epigenetic-dysregulation-at-DH6aNMFTq9
SP - 335
VL - 29
IS - 2
DP - DeepDyve
ER -