Article Text

Download PDFPDF

Original article
Epigenetic state and expression of imprinted genes in umbilical cord correlates with growth parameters in human pregnancy
  1. Ai Lin Lim1,
  2. Shilen Ng1,
  3. Suet Ching Pamela Leow1,
  4. Robin Choo1,
  5. Mitsuteru Ito2,
  6. Yiong Huak Chan3,
  7. Siew Kheng Goh1,
  8. Emilia Tng1,
  9. Kenneth Kwek4,
  10. Yap Seng Chong5,
  11. Peter D Gluckman1,6,
  12. Anne C Ferguson-Smith1,2
  1. 1Department of Growth Development and Metabolism, Singapore Institute for Clinical Sciences, Agency for Science Technology and Research (A-STAR), Singapore, Singapore
  2. 2Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, UK
  3. 3Department of Obstetrics and Gynaecology, Yong Loo Lin School of Medicine, Biostatistics Unit, National University of Singapore, Singapore, Singapore
  4. 4Department of Maternal Fetal Medicine, KKH Women's and Children's Hospital, Singapore, Singapore
  5. 5Yong Loo Lin School of Medicine, Department of Obstetrics and Gynaecology, National University of Singapore, Singapore, Singapore
  6. 6Liggins Institute, University of Auckland, Auckland, New Zealand
  1. Correspondence to Professor Anne C Ferguson-Smith, Department of Physiology, Development and Neuroscience, University of Cambridge, Downing Street, Cambridge CB2 3EG, UK; afsmith{at}


Background Genomic imprinting is a process causing genes to be expressed according to parental origin. Imprinting acts to coordinate fetal and prenatal growth, as well as control postnatal adaptations. Studies on human imprinting are confounded by tissue availability, sampling variability and limitations posed by tissue-specific expression and cellular heterogeneity within tissues. The human umbilical cord is an easily available, embryonic-derived fetal tissue with the potential to overcome many of these limitations.

Methods In a sensitive, gene-specific quantitative expression analysis, we show for the first time robust imprinted gene expression combined with methylation analysis in cords isolated from Asian Chinese full-term births.

Results Linear regression analyses revealed an inverse correlation between expression of pleckstrin homology-like domain, family A, member 2 (PHLDA2) with birth weight (BW). Furthermore, we observed significant down-regulation of the paternally expressed gene 10 (PEG10) in low BW babies compared to optimum BW babies. This change in PEG10 gene expression was accompanied by concomitant methylation alterations at the PEG10 promoter.

Conclusions These data are the first to demonstrate relative expression of an imprinted gene associated with epigenetic changes in non-syndromic fetal growth restriction in babies. They show that perturbed expression in compromised fetal growth may be associated with in utero modulation of the epigenetic state at the imprinting control regions and implicate specific imprinted genes as new biomarkers of fetal growth.

  • Epigenetics
  • Genetics
  • Imprinting
  • Obstetrics and Gynaecology
  • Reproductive medicine
View Full Text

Statistics from


Genomic imprinting is a normal process causing genes to be expressed according to parental origin. Imprints are regulated by epigenetic mechanisms, especially by DNA methylation marks that are differentially acquired on the two parental chromosomes during gametogenesis. These primary differentially methylated regions (DMRs) acquire gamete-specific methylation during either spermatogenesis or oogenesis and maintain these differences during the early preimplantation epigenetic programme and usually through later development. Secondary DMRs are established after fertilisation as a consequence of primary DMRs and contribute to stable monoallelic expression. Most imprinted genes are found in clusters and share a common primary DMR regulating the imprinting of multiple genes in the cluster.1 Mouse models have shown that defective imprinting results in prenatal and postnatal abnormalities including intrauterine growth and placentation.

Knowledge of the function and regulation of imprinting has primarily been extrapolated to humans from studies in mice, but it is becoming increasingly evident that inherent differences in imprinting may exist between humans and mice. Imprinting studies in humans are generally limited due to normal tissue availability and are also confounded by tissue-specific expression and recognised differences in epigenetic control between placenta and embryonic tissues in Eutherian mammals. Furthermore, genes with placenta-specific imprinting in the mouse (Ascl2, Nap1l4, Cd81 and Phemx) are biallelically expressed in human fetal and placental tissues. IGF2R, imprinted in mice, is polymorphically imprinted in humans.2 In contrast, L3MBTL1 is imprinted in humans but not in rodents.3 ,4

Attempts to provide a practical and available resource for epigenetic and imprinting studies in humans have been challenging. Easily accessible tissues such as the adult peripheral blood have been shown to be unsuitable for imprinting studies as the majority of imprinted genes examined show low/no expression in lymphoid or myeloid cells.5 Analyses of term placenta tissue, although extensive, are confounded by issues of maternal contamination, cell and tissue heterogeneity and tissue integrity. Here, we investigate the potential of the human umbilical cord as a suitable and easily available tissue for imprinting studies. Since the cord is developed from cells derived from the inner cell mass, from which the embryo develops, rather than the trophectoderm which contributes to the non-vascular cells of the placenta, we hypothesised that the epigenetic status and imprinted gene expression in the cord is appropriately reflective of fetal development fulfilling the requirements of easy availability and relative tissue homogeneity.

The fetal origin of adult disease hypothesis suggests that developmental plasticity and adaptation in response to compromised environmental and nutritional signals during early life can program ‘the fetus for a spectrum of adverse health outcomes as an adult’.6 ,7 The associations between low birthweight (LBW) and increased incidence of disease in adult life have been extensively supported in epidemiological8–14 and animal studies.15 ,16 LBW has been defined by the WHO and National Institutes of Health as babies with BWs below 2500 g born, irrespective of gestational age. Reports from many countries have also shown LBW to be highly predictive for perinatal mortality and childhood morbidity.17–19 Often, causes of LBW are not well understood. Fetal growth restriction can be a response to disease in mother, to a metabolic or genetic disorder in the fetus, or abnormalities in the placenta.

This fetal programming appears largely independent of genomic DNA (gDNA) sequence, and has been suggested to be mediated by epigenetic mechanisms.20 ,21 Methylation of imprinted genomic regions are of particular interest because of their known influences in regulating fetal and prenatal growth22 ,23 and because of emerging data suggesting a correlation between epigenetic differences at these DMRs in response to early exposure to adverse environments.24–26 Although a limited number of studies demonstrate deregulation of imprinted genes in intrauterine-growth restricted (IUGR) placentas,27 ,28 accompanying epigenetic changes are not usually reported. Similarly, empirical support for perturbation of gene expression dosage is rarely considered alongside epigenetic abnormalities occurring at the imprinted loci in growth restricted individuals.

Here, we test the hypothesis that alterations in epigenetic control mechanisms are associated with dysregulation of growth-related imprinted genes that affect placenta and/or fetal growth and lead to LBW babies. We present a comprehensive analysis of the allelic expression of multiple imprinted genes and methylation patterns at their DMRs in umbilical cords obtained from Chinese pregnancies. Relative expression of 13 imprinted genes was quantified in cords of 75 normal BW and 15 LBW newborns and examined for associations with various fetal growth parameters. Using quantitative pyrosequencing assays, we interrogated changes in methylation status of six germline DMRs (IG-DMR, H19 DMR/imprinting control region (ICR), IGF2R DMR, KvDMR, PEG10 DMR, ZAC1/PLAG1 DMR) and two secondary DMRs (GTL2 DMR, H19 promoter DMR) in low and normal BW babies.

Materials and methods

Subjects and tissue collection

Ninety umbilical cord samples from unrelated newborn Chinese babies were collected using study procedures approved by the Institutional Review Board. Details are provided in online supplemental methods and summarised in table 1.

Table 1

Allelic status of imprinted genes and quantification of DNA methylation at differentially methylated regions (DMRs) in the umbilical cords of newborns

Nucleic acid extraction and cDNA preparation

The cord tissue was homogenised using the Dispomix system (Miltenyi Biotec, Germany), followed by extraction of RNA or gDNA. gDNA (5 µg) was isolated by a phenol-chloroform method using Phase Lock Gel (Eppendorf, Hamburg, Germany). Briefly, a 100 mg sample was homogenised and digested overnight at 56°C with 500 µl Lysis Buffer (50 mM Tris-HCl pH 7.4, 100 mM NaCl, 50 mM EDTA, 1% SDS) containing 4 µl of proteinase K (20 mg/ml). Each sample underwent two phenol/chloroform extractions and one chloroform extraction. gDNA was precipitated by 100% isopropanol, washed with 70% ethanol and resuspended in 50 µl TE-Tris-EDTA buffer. The gDNA was treated with RNase A (Qiagen, GmbH, Germany) to remove contaminating RNA. Total RNA extracted with Trizol based on the manufacturer's instructions (Invitrogen, UK) with further details on cDNA synthesis provided in online supplementary methods.

Real-time expression analysis

The ABI Prism 7900HT sequence detection system (Applied Biosystems Inc., Foster City, California, USA) was used with inventoried Taqman assays to assess transcript expression levels of 13 genes with details and accession numbers presented in online supplementary methods.

Analysis of expression and imprinting

Single nucleotide polymorphisms (SNP) were identified by interrogating SNP databases and confirmed by DNA sequencing of PCR products after amplification using specific primers (see online supplementary table S2). PCR was performed on cDNAs using the Sequalprep Long PCR kit according to the manufacturer's protocol. All PCR products were sequenced in both the forward and reverse orientations using the ABI Prism 3100 DNA sequencer and analysed for monoallelic/biallelic expression using the informative SNPs.

MicroRNA expression analysis

Procedures for the analysis of the microRNAs are detailed in online supplementary methods. TargetScan 5.1 ( and the miR database miRBase ( were used to identify potential miRNA targets to the PHLDA2 and PEG10 gene. Relative quantification of microRNA expression was carried out using TaqMan MicroRNA Assay kits according to the manufacturer's protocol (Applied Biosystems, Foster City, California, USA). All assays including no template controls were done in triplicates. Relative quantification of miRNA expression was calculated by the 2−ΔΔCT method.29

DNA bisulphite conversion and sequencing

Bisulphite conversion was carried out on total gDNA using EZ Methylation Gold kit (Zymo Research, Orange, California, USA) following the manufacturer's protocol. Subsequently, bisulphite-converted gDNA was used in pyrosequencing analysis or bisulphite sequencing (see online supplementary table S3).The PCR products were cloned into PCR-TOPO vector (Invitrogen. Carlsbad, California, USA) and colonies cultured overnight in LB-ampicillin (100 µg/ml). The plasmids were purified using the PureLink Quick96 Plasmid Kit (Invitrogen) and sequencing carried out on the ABI Prism 3100 DNA sequencer.

Pyrosequencing assay

PCR was performed on bisulphite-converted gDNA using the Sequalprep Long PCR kit (Invitrogen, Carlsbad, California, USA), with each 20 µl PCR reaction consisting of 1× Sequalprep reaction buffer, 1× Sequalprep enhancer and 1.8 U of Sequalprep long polymerase, 0.5 µM of each primer and 2 µl of bisulphite converted DNA. PCR conditions are as follows: denaturation at 94°C for 2 min; 50 cycles at 94°C for 10 s, 55°C for 30 s, 72°C for 40 s and a final extension at 72°C for 5 min (see online supplementary table S2 for primers and PCR conditions). Bisulphite pyrosequencing was carried out on the PyroMark MD pyrosequencer (Biotage AB, Uppsala, Sweden) with the PyroGold SQA reagent kit (Biotage). Output data was analysed using Pyro Q-CpG-1.0.9 software (Biotage). For the pyrosequencing assay, between 3–6 CpG dinucleotides were interrogated for each DMR region.

Statistical analysis

Biostatistical analysis was performed using SPSS V.16.0 ( Methylation Indices (MI) were obtained by averaging the methylation across the CpG dinucleotides and expressing it as MI (mean±SD) and their differences between LBW versus normal BW babies were compared using parametric tests when normality and homogeneity assumptions were satisfied. Otherwise the equivalent non-parametric tests were applied. Box plots and error plots were presented as appropriate. Scatter plots with Pearson coefficient (r) were presented to visualise the relationship between two quantitative variables. Statistical significance was set at p<0.05.


Relative gene expression levels of imprinted genes in cords

To determine the suitability of the cords as a tissue for imprinting studies, we first investigated mRNA expression levels of 13 imprinted gene transcripts by performing quantitative real-time PCR on cords obtained from 90 newborn Chinese babies. These genes were normalised to the geometric mean of three housekeeping genes (ACTB, GAPDH and B2M). Some genes examined exist within imprinted clusters (PHLDA2, SLC22A18, KCNQ1, SLC22A2, DLK1, GTL2, L3MBTL, SNRPN, IGF2, GNAS, SNRPN), while others (ZAC1 and PEG10) are unclustered. These imprinted genes are known to be expressed in extraembryonic tissues such as the placenta and altered expression of some has been suggested to play a role in placental/fetal growth in humans.27 ,28 ,30 We observed robust expression of all the 13 gene transcripts investigated (figure 1A), several at notably high levels: IGF2, GNAS, DLK1 and GTL2. Four of the genes (ZAC1, GRB10, PEG10, SNRPN) were expressed between 10–100 fold lower than those of the housekeeping genes, with the remaining five (SLC22A2, PHLDA2, SLC22A18, KCNQ1 and L3MBTL) approximately 100–1000 fold lower. Relative expression levels of several genes (PHLDA2, SLC22A18, KCNQ1, SLC22A2, DLK1, GTL2, L3MBTL, SNRPN, IGF2, GNAS, SNRPN, ZAC1 and PEG10) were assessed against various fetal growth parameters, including BW, birth length, head circumference, abdominal circumference, and placental weight. The analysis was performed using Analysis of variance (ANOVA), including the covariates gestational age, gender, parity, gestational diabetes and maternal weight. We found a significant, inverse association between the relative PHLDA2 expression with BW (see online supplementary table S1, β=−81.039, p=0.039). None of the other 12 genes examined showed significant correlation with fetal growth parameters.

Figure 1

Relative expression of imprinted genes in the umbilical cords of babies and significant correlations with birth parameters. Quantitative expression levels of 13 imprinted genes were examined in 90 babies. Genes names listed on the X axis are plotted against 2^-ΔCton the logarithmic Y axis, where ΔCt is the cycle threshold Ct of each transcript normalised to the geometric mean of three housekeeping genes (B2M, GAPDH and ACTB). Real time expression was obtained using triplicates of cords obtained from 90 individual umbilical cords. Based on this calculation, the housekeeping genes would show 2^-ΔCt=1.

Analysis of allelic expression status of imprinted genes in cords

In order to investigate the imprinting status of genes in umbilical cord, we genotyped unrelated samples for exonic SNPs present in imprinted genes. Informative samples heterozygous for the SNPs were analysed for allelic expression using intron(s)-spanning primers (see online supplementary table S2 for list of primer sequences and PCR conditions). Apart from genes in the KCNQ1 and IGF2R clusters, all other imprinted genes interrogated (DLK1, GTL2, L3MBTL, SNRPN, IGF2, H19, ZAC1, PEG10) showed a clear monoallelic expression pattern (table 1), in agreement with their reported imprinted status in other human fetal and adult tissues. Parental origin of expression was determined for samples with maternal gDNA, and the allelic expression was as expected.

Interestingly, genes within the IGF2R and KCNQ1 cluster showed a combination of polymorphic imprinting, biallelic and monoallelic expression. For the KCNQ1 cluster, nine cord samples showed monoallelic expression of PHLDA2, of which five of them were verified to be maternal in origin. There appears to be a relaxation of imprinting as one proceeds telomeric of the KvDMR, with SLC22A18 and adjacent genes (KCNQ1 and ASCL2) showing biallelic expression in cords. NAP1L4, ASCL2, CD81 and PHEMX are genes whose orthologues exhibit placenta-specific expression from the maternal allele in the mouse under the control of the KvDMR. Both NAP1L4 and ASCL2 showed biallelic expression in human cords consistent with their absence of imprinting in all human tissues. PHEMX, CDKN1C and CD81 were uninformative across the entire exonic region for all individuals. One DMR was found corresponding to the previously described KvDMR.2 ,31 Differential methylation of the KvDMR was maintained in the umbilical cord samples examined despite the full monoallelic expression of only one gene in the cluster.

IGF2R and SLC22A2 are polymorphically imprinted in humans, with the majority of individuals showing biallelic expression in fetal placentae2 ,32 and fetal lung.32 In our Asian samples, we found monoallelic expression of IGF2R in only 1 out of 29 informative individuals. However, 6 of 18 informative individuals showed monoallelic expression of the nearby SLC22A2 gene. Of these, four showed discordant imprinting between IGF2R (biallelic) and SLC22A2 (monoallelic) genes, with another four having concordant biallelic expression for both genes. We selected three concordant samples and three discordant samples to examine DNA methylation at the IGF2R DMR (figure 2A). This ICR lies within intron2 of IGF2R and is normally methylated on the maternally inherited chromosome. Differential methylation of maternal and paternal chromosomes at this locus is conserved between human and mouse, even in situations where human IGF2R gene is transcribed from both alleles.33 ,34 Surprisingly, hypermethylation of the IGF2R DMR was observed in the three individuals showing biallelic SLC22A2 expression, polymorphism in two of them confirmed that the gain of methylation occurred at the presumably paternal allele. Importantly, canonical differential methylation was maintained in the three individuals showing monoallelic SLC22A2. This indicates that in contrast to imprinting at IGF2R, imprinting at SLC22A2 is correlated with the methylation status of the IGF2R DMR.

Figure 2

Analysis of the allelic expression of SLC22A2 and IGF2R at the IGF2R differentially methylated region (DMR) locus. (A) Bisulphite sequencing analysis of IGF2R DMR in cord samples showing concordant or discordant expression of the IGF2R and SLC22A2 gene. The region physical map of the human IGF2R cluster on human chromosome 6 (not drawn to scale). Genes known to be maternally expressed are indicated in red; paternally expressed putative human AIRN transcript is indicated in blue. Differential methylation at the IGF2R DMR was maintained in the cord samples (cord 1–3) showing monoallelic SLC22A2 and biallelic IGF2R expression. Hypermethylation of the IGF2R DMR was seen in the cord samples (cord 4–6) showing concordant biallelic SLC22A2 and IGF2R expression. The T/G single nucleotide polymorphisms located between the 5th and 6th CpG dinucleotide in the region examined denotes parental allele. Each line indicates a single clone, and each circle denotes CpG dinucleotide; filled and open circles represent methylated and unmethylated cytosines, respectively. (B) Scatterplot showing relative SLC22A2 mRNA transcript levels with different SLC22A2 allelic expression. Each circle represents the mRNA level in one individual cord sample. Population exhibiting monoallelic (n=6) and biallelic (n=12) SLC22A2 expression showed different, but statistically insignificant, transcript level changes (fold change=1.76, Student's t test, p=0.061).

Furthermore, the allelic expression status of SLC22A2 appears to regulate transcript levels, with loss-of-imprinting resulting in increase in relative gene expression (figure 2B). Of the 18 cord samples showing informative heterozygous SNPs, 6 showed monoallelic expression and the other 12 showed biallelic expression (table 1). We observed a 1.76-fold increase in the relative SLC22A2 mRNA in the population showing biallelic SLC22A2 expression compared with the population showing monoallelic expression; this trend did not reach statistical significance (Student's t test , p=0.061). However, the data does suggest that high expression of SLC22A2 is associated with loss-of-imprinting.

Analysis of DMRs at imprinting control regions

In addition to IGF2R, we quantitatively examined the methylation status of the other DMRs (table 1). Germline DMRs exhibited methylation ranging between 45–65%, consistent with the maintenance of differential methylation at these ICRs. Methylation was expressed as MI (MI=mean methylation±SD). Bisulphite sequencing allowed examination of more CpGs, with amplicon sizes ranging from 250 bp to 600 bp. Bisulphite sequencing data were consistent with the pyrosequencing assay (see online supplementary figure S1), showing mean 50% methylation for the germline DMRs. Each clone showed complete methylation or was fully unmethylated indicative of DMR status. Using polymorphisms we confirmed that all the methylated CpGs arose from one allele and all unmethylated from another.

In addition to germline DMRs, we examined two somatic or secondary DMRs. These are established during embryonic development. The GTL2 secondary DMR, found within the DLK1-GTL2 imprinted domain, showed greater variability between cord samples, with methylation levels ranging from 10–50%. Despite the hypomethylation at the GTL2 promoter, the non-coding GTL2/MEG3 transcript maintained monoallelic expression in 10 informative samples (table 1). Additionally the neighbouring DLK1 was also monoallelically expressed in eight informative samples. We were unable to establish the imprinting status of the Type III iodothyronine deiodinase gene (DIO3), another protein coding gene in the domain known to be expressed from the paternal chromosome, due to an absence of informative SNPs in its exon. Taken together, results suggest that differential methylation at the ICR may be sufficient to maintain monoallelic expression of the imprinted genes in this cluster and that somatic DNA methylation, such as at GTL2 DMR, may be dispensable for conserved imprinting at this region at least in cords. Interestingly, the hypomethylation observed in the GTL2 promoter was also seen for the H19 promoter DMR, the other secondary DMR examined. Like GTL2, despite the hypomethylation at the H19 promoter, H19 showed imprinted expression in 15 cord samples, as did the neighbouring IGF2 gene. This indicates that differential methylation at the germline H19 DMR is sufficient to maintain imprinted expression in cords.

Methylation of DMRs in low and normal BW babies—an effect with PEG10

We examined the methylation pattern of the seven DMRs in cord gDNA obtained from normal BW (n=33) and small BW babies (n=15, table 2). These LBW babies are categorised as full term babies with gestation >37 weeks and BW <2.5 kg. They include constitutionally small babies that may/may not fall within the IUGR category. In each cord sample, we saw little variation in methylation between each CpG. This is characteristic of DMRs. The consistency meant that we could obtain an average methylation value across the CpGs instead of looking at each individual CpG site. We observed no significant differences between low and normal BW in the methylation of most DMRs (KvDMR, H19 DMR, IG-DMR, GTL2 DMR, ZAC1 DMR) (figure 3A). Interestingly, two individuals in the LBW category showed hypermethylation (94–95%) at the IGF2R DMR. However, we lacked informative material to determine allelic expression.

Table 2

Clinical characteristics of the mothers and newborn babies participating in the study.

Figure 3

Significant associations of methylation at differentially methylated regions (DMRs) and expression of imprinted genes with normal and low birth weight (LBW) babies. (A) Comparison of mean methylation at DMRs in cords from normal (n=33) and LBW (n=15) babies. The absolute methylation values were obtained by taking the average methylation percentage across the interrogated CpGs in each DMR pyrosequencing assay. All the DMRs maintained similar methylation levels between the two groups. Only the PEG10 DMR exhibited significant hypermethylation in the LBW group (p=0.014). (B) Linear regression showing significant inverse relationship (p=0.009) between PEG10 DMR methylation with BW of the babies. Babies used (n=48) were all full term (>37 gestational weeks). (C) PEG10 relative expression was significantly downregulated in LBW (n=15) in comparison to control (n=33) babies by Mann-Whitney U test (p=0.036). (D) Linear regression showing significant inverse correlation between PEG10 DMR methylation and PEG10 relative gene expression (n=48).

Investigation of the PEG10 DMR showed a significant hypermethylation in the LBW compared to the normal BW group (figure 3A, Mann-Whitney U test, p=0.014). PEG10 DMR methylation values also showed greater variation amongst the LBW individuals ranging from 49–63%, while the normal BW individuals had little inter-individual variation, with mean methylation ranging from 45–52%. PEG10 DMR methylation was found to have a strong inverse correlation with BW (figure 3B, r=−0.386, p=0.009). We can assume that acquisition of methylation is occurring on the canonically unmethylated allele.

Altered imprinted gene expression levels in LBW babies

Due to significant PEG10 DMR hypermethylation found in LBW infants, we investigated the possibility that this epigenetic change predicted difference in expression of the PEG10 transcript. Student's t test revealed a significant down-regulation (0.68-fold change) of PEG10 expression in the LBW compared to control cords (figure 3C, p=0.010). There was no difference in PEG10 expression between sexes, parity or gestational week. We examined the association between methylation at the PEG10 DMR with log-transformed PEG10 relative gene expression and found a significant correlation (figure 3D, r=−0.382, p=0.013), suggesting that PEG10 transcript levels are inversely associated with methylation at the PEG10 DMR. We hypothesise that hypermethylation at the PEG10 DMR in LBW babies is acting on the paternal allele, and is associated with reduced activity of the usually active PEG10 transcript. We also examined the relative expression of the other imprinted genes: PHLDA2, SLC22A18, KCNQ1, SLC22A2, IGF2R, DLK1, GTL2, L3MBTL, SNRPN, IGF2, GNAS and GRB10. We found no significant difference in expression between the cords of the LBW and normal BW babies. Apart from PEG10, there was no correlation between imprinted gene transcript levels and DMR methylation at the respective imprinted loci.


In this study, we investigate the use of the umbilical cord as a tissue to detect expression and imprinting. Using sensitive Taqman assays rather than the more commonly used Sybr Green incorporation, we find that all imprinted genes are robustly expressed in the cord. We report the first known analysis of methylation and transcript differences at imprinted loci in cord tissues between LBW and normal BW children and the first study of imprinting in pregnancies of Chinese ethnicity. Except for several genes in the IGF2R and KCNQ1 imprinted clusters, the imprinted genes examined showed monoallelic expression.

Our assessment of expression levels and allelic expression of the imprinted genes revealed no correlation between gene expression levels and loss-of-imprinting. Other studies have also reflected this lack of correlation between expression and loss of imprinting in placenta-related pathologies such as IUGR27 and pre-eclampsia.35 Our results also suggest that mechanisms other than loss-of-imprinting may be associated with the deregulated gene expression of PHLDA2 and PEG10 that we observed in LBW babies. Indeed, we see that the PEG10 gene transcript appears to be regulated by PEG10 DMR methylation, with monoallelic expression maintained and no loss-of-imprinting detected. One exception to the observed absence of a relationship between dosage and loss of imprinting is seen at the SLC22A2 gene; mechanisms and implications of this are explored below. Our data also indicate that those genes whose orthologues show placenta-specific imprinting in mouse (ASCL2, NAP1L4), also exhibit biallelic expression in the cord. This is consistent with the lack of imprinting of these genes previously observed in human placenta and other fetal tissues,2 perhaps suggesting a less stringent requirement for dosage control of these genes in humans.

Interestingly, within the KvDMR cluster, we see conserved imprinting only in the maternally expressed PHLDA2 gene, possibly because of its critical role in regulating fetal growth and BW in babies. Importance of regulating PHLDA2 dosage levels and maintaining imprinting in humans is evidenced in the inverse correlation that we observe between cord PHLDA2 expression and fetal BW in the Chinese babies. This finding reinforces and reproduces in a different tissue and population, findings from placentas of newborns of European descent36 and in others showing higher PHLDA2 expression in placentas from IUGR pregnancies.27 ,28 Recently, Ishida et al37 showed that maternal inheritance of a 15 bp repeat sequence variant in the PHLDA2 promoter resulted in a significant increase in BW in European babies. It would be interesting to investigate whether this genetic effect could explain the relationship between PHLDA2 expression and BW in our Chinese cohort.

The IGF2R gene was previously shown to exhibit polymorphic imprinting in five out of eight placentas.2 Our study with cords revealed only one of 28 informative individuals with monoallelic expression. This low frequency of imprinting in our samples could reflect the different ethnicity of the population examined, but alternatively might reflect differences in tissue-specific imprinting. In mouse, the IGF2R imprinting control element, known as Igf2r DMR2, is the primary imprint regulator of the cluster and germline methylation on the maternally inherited chromosome is required for maternal allelic Igf2r expression. This DMR contains the promoter for the paternally expressed Airn non-coding RNA required for repression of Slc22a2 and Slc22a3 genes on the paternal chromosome.38 ,39 There is no evidence for the presence of a human orthologue of the Airn RNA. Our results show a predominant absence of imprinting of IGF2R despite maintenance of differential methylation at the IGF2R DMR. This has also been noted in the human placenta.2 However, in contrast to previous studies we observe discordance in imprinting between IGF2R (biallelic) and SLC22A2 (monoallelic) within the same individuals. Furthermore, we observe a gain in methylation marks at the IGF2R DMR in individuals specifically showing biallelic SLC22A2 expression. Interestingly, two of the 15 LBW babies fell into this category. The rest of the LBW babies were not informative for SLC22A2 status. This, in combination with the contrasting retention of DMR status in individuals with biallelic IGF2R and monoallelic SLC22A2, suggests a previously unidentified mechanistic relationship between the DMR and SLC22A2 imprinting in human, perhaps with functional consequences for growth.

We investigated the effect of epigenetic changes associated with growth restriction occurring early in development. Conflicting data exists in the literature for this. H19 DMR/ICR methylation has been shown to be associated with BW in whole blood40 and decreased in normotensive IUGR placentas.41 However, this association was not seen in a separate study using cord blood mononuclear cells, granulocytes, buccal epithelial cells, human umbilical vein endothelial cells and placenta.42 Our data show no effect at IGF2/H19 consistent with the latter study. Indeed, for all but one imprinted gene, there was no association between BW and methylation.

Our quantitative methylation data on the DMRs suggests that LBW is associated with statistically significant increased methylation of the PEG10 DMR. Strikingly, PEG10 mRNA levels correlate inversely with increased DMR methylation likely occurring at the canonically unmethylated paternal allele. Besides showing a lowered PEG10 expression, the LBW babies had a significantly smaller sized placenta, even after accounting for gestational age, parity and gender (data not shown). It has been shown previously in mice that loss of PEG10 expression leads to restricted fetal growth and early embryonic lethality, associated with defective junctional zone in the developing placenta.43 Our results also suggest a role for PEG10 in regulating human fetal growth, perhaps through an effect on placenta development. Furthermore, our data demonstrate that even modest changes in PEG10 expression levels (∼1.5 fold lower) are associated with significant biological changes in human prenatal growth, and this is correlated with epigenetic modulation.

Recently, the microRNA, miR-122, has been reported to regulate the expression of PEG10 in hepatoma cell lines.44 Thus we sought to explore whether microRNAs might potentially contribute to PEG10 regulation in these low and normal BW babies. In silico analysis revealed a list of 18 miRNAs that could potentially target the 3′ untranslated region of the PEG10 transcript. However, we did not see any changes in the relative expression of several miRs (miR-449, miR-182, miR-27a, miR-197 miR-122 and miR-488), between the low and normal BW babies (see online supplementary figure S2). Our results suggest that inappropriate expression of miRs is not responsible for reduced PEG10 in LBW babies. Instead, epigenetic mechanisms such as promoter DMR methylation may contribute to PEG10 transcript expression. Likewise, we examined possible regulation of PHLDA2 expression through microRNA targeting. The relative expression of several miRs predicted to target the PHLDA2 3′UTR were examined: miR-208a, miR-499-5p and miR-1286. We did not find a correlation between microRNA levels with changes in PHLDA2 expression (see online supplementary figure S2). Together this suggests that post-transcriptional modulation of these imprinted genes by miRNAs is not responsible for the  BW-associated changes in transcript levels seen in our cohort.

Other functions have been described for PEG10. These include adipocyte differentiation,45 inhibiting apoptosis46 and mouse liver cell regeneration.47 Many of these processes occur primarily in the intrauterine period, a time of rapid cell division. Our data suggest that the PEG10 locus is subject to epigenetic modulation and dysregulation in utero, perhaps in response to an adverse intrauterine environment. Currently, this birth cohort is part of a longitudinal study to determine later outcomes, therefore, longer term phenotypic consequences of PEG10 perturbation will be interesting to consider. The current study presents the first example of an association between LBW and DNA methylation at a DMR with concomitant expression changes in human. Understanding the dynamic mechanisms regulating this change in epigenetic control in compromised in utero conditions may shed light on the link between increased adult disease risk with adverse conditions during development.


The authors would like to acknowledge Veronica Khee, Tan Li Hua, Rachel Chew, Joanna Fong and Michelle Lee, staff and nurses at KKH Women's and Children's Hospital and National University Hospital for their assistance in data management, sample collection and sample preparation. We would also like to thank Clara Cheong for helpful discussion and contributions.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors The project was managed and executed by ALL, experimental support was conducted by SCPL and SN and contributions to statistical analysis were provided by RC and YHC. Samples and reagents were provided by ET and MI. Placenta and cord and patient procedures were generated and collated by YSC, KK, PDG. The project was designed and supervised by ACF-S.

  • Funding This study was supported by the Translational Clinical Research (TCR) Flagship Research Grant from the National Medical Research Council (NMRCR) and a core grant from the Agency for Science, Technology and Research.

  • Competing interests None.

  • Ethics approval Institutional Review Board (IRB) Singapore.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.