Genome-Wide Analysis of DNA Methylation in Human Amnion

The amnion is a specialized tissue in contact with the amniotic fluid, which is in a constantly changing state. To investigate the importance of epigenetic events in this tissue in the physiology and pathophysiology of pregnancy, we performed genome-wide DNA methylation profiling of human amnion from term (with and without labor) and preterm deliveries. Using the Illumina Infinium HumanMethylation27 BeadChip, we identified genes exhibiting differential methylation associated with normal labor and preterm birth. Functional analysis of the differentially methylated genes revealed biologically relevant enriched gene sets. Bisulfite sequencing analysis of the promoter region of the oxytocin receptor (OXTR) gene detected two CpG dinucleotides showing significant methylation differences among the three groups of samples. Hypermethylation of the CpG island of the solute carrier family 30 member 3 (SLC30A3) gene in preterm amnion was confirmed by methylation-specific PCR. This work provides preliminary evidence that DNA methylation changes in the amnion may be at least partially involved in the physiological process of labor and the etiology of preterm birth and suggests that DNA methylation profiles, in combination with other biological data, may provide valuable insight into the mechanisms underlying normal and pathological pregnancies.


Introduction
e human amnion is the inner layer of the fetal membranes composed of a monolayer of epithelial cells attached to a basement membrane overlying a collagen-rich stroma [1,2]. is tissue, which encloses the amniotic �uid, protects the fetus from external mechanical forces and provides an environment that supports fetal movement and growth [3,4]. e amnion is also a metabolically active tissue involved in the synthesis of various substances with important functions during pregnancy, including prostaglandins and cytokines [1,5,6]. It is particularly well known as a major source of prostaglandin E2, a potent molecule mediating cervical ripening and myometrial contraction [7][8][9][10], whose levels dramatically increase before and during labor [11,12].
e amniotic membrane provides most of the tensile strength of the fetal membranes, and alterations in its integrity can lead to undesirable pregnancy outcomes such as preterm premature rupture of membranes (PPROMs) [1,13], which complicates 3% of all pregnancies and is responsible for approximately one-third of all preterm births (PTBs) [14]. Given the important role of the amnion in the maintenance of pregnancy and parturition, investigation into molecular events occurring in this tissue may contribute to a better understanding of physiological and pathological processes involved in pregnancy.
Considering that the amniotic �uid is in a constantly changing state, it may be critical that the amnion properly responds to environmental cues from the amniotic �uid to accommodate the dynamic needs of the fetus, which could be mediated through epigenetic processes. A previous study by Wang et al. [15] has shown that matrix metalloproteinase 1 (MMP1), whose genetic variation is associated with susceptibility to PPROM [16], is regulated at the epigenetic level, e Scienti�c World �ournal speci�cally by DNA methylation, and that MMP1 promoter methylation status correlates with its expression in the amnion and association with PPR�M. is �nding suggests that the amnion represents an intriguing source of tissue for studying epigenetic events of potential physiological and pathological relevance.
In this study, we performed genome-wide methylation pro�ling of human term and preterm amnion in order to explore the possible importance of DNA methylation in physiologic labor as well as the etiology of PTB. In addition, independent of the genome-wide methylation study, we carried out methylation analysis of the promoter region of the oxytocin receptor (OXTR) gene whose role in human parturition is well established [17]. Given that OXTR expression in the amnion increases in association with the onset of labor [18] and that its aberrant methylation in other tissue types has been implicated in autism [19], a disorder that has been associated with PTB [20,21], we sought to investigate if DNA methylation could represent one mechanism regulating OXTR gene function in the contexts of normal parturition and prematurity.

Placental Tissue Collection and Preparation.
Fresh human placentas were collected in 2009 and 2010 at the University of Iowa Hospitals and Clinics in IA, USA and Instituto de Maternidad y Ginecología Nuestra Señora de las Mercedes in Tucumán, Argentina with signed informed consent and an institutional review board approval. We examined 121 placentas from three groups of patients undergoing: term cesarean delivery without labor (term no labor (TNL) group, ), normal term vaginal delivery (term labor (TL) group, ), and spontaneous preterm (<37 weeks of gestation) delivery (preterm labor (PTL) group, ). Gestational age (GA) was determined using the �rst day of the last menstrual period as well as by ultrasound examination and was con�rmed by assessment at birth. Each placenta was dissected into fetal (amnion, chorion) and maternal (decidua basalis) components within an hour of delivery. e amnion and chorion obtained from the extraplacental membranes (re�ected membranes) were separated by blunt dissection under sterile conditions. Decidual tissue samples were macroscopically isolated from the surface of the basal plate of the placenta. Aer being cut into small pieces, the dissected tissues were placed in RNA later solution (Applied Biosystems, Carlsbad, CA, USA) and stored per manufacturer's recommendations until used. A subset of these samples was selected for genome-wide methylation analysis on the basis of their informativity in relation to our previous gene expression pro�ling study (unpublished). Additional samples used for validation experiments were selected primarily based on the quality of DNA or RNA extracted from the tissue samples. . Details of the design and general properties of this platform have been previously described [22]. A total of 24 samples were assayed on two BeadChips (12 samples per chip) following the standard protocol provided by Illumina. e samples examined included 9 individual and 1 pooled amnion samples each from the TNL and TL groups, one pooled amnion sample from the PTL group obtained by combining 6 individual samples, and 3 controls (methylated DNA control treated with M.SssI methyltransferase (New England Biolabs, Ipswich, MA, USA), Universal Methylated Human DNA Standard (Zymo Research), and bisul�te-untreated control). ese samples were selected from among patients who had participated in our previous gene expression pro�ling study (unpublished), performed independently of the current work. Based on this previous study, which showed heterogeneous global gene expression patterns among PTL samples, we only included one pooled PTL sample to assess a group DNA methylation average. e samples were arranged randomly on each chip and were processed in a blinded fashion. Table 1 summarizes the clinical characteristics of the three groups of samples studied.

Quality Control and Statistical
Analysis. Data analysis was conducted on a fee-for-service basis by the W.M. Keck Biostatistics Resource at Yale University with GenomeStudio Methylation Module v1.0 (Illumina). We evaluated the quality of the data based on the signals of assay builtin control probes (staining, hybridization, target removal, extension, bisul�te conversion, methylation signal speci�city, background determination, and overall assay performance) and three experimental controls (two positive methylated controls and one non-bisul�te-converted control), and con-�rmed the reliability of our data. Principal component analysis (PCA) demonstrated that there is no signi�cant batch effect among the three groups of samples examined. e methylation status of each interrogated CpG site was e Scienti�c World �ournal 3 T 1: Clinical characteristics of the three sub�ect groups studied by genome-wide DNA methylation pro�ling.

Parameter
TNL ( ) 1 TL ( ) 1 PTL ( ) 2 Gestational age (weeks) 3 39. Examined both individually and as a pooled sample. 2 Examined as a pooled sample. 3 Data are presented as mean ± standard deviation (SD). Abbreviations: TNL: term no labor; TL: term labor; PTL: preterm labor. determined employing the -value (de�ned as the fraction of methylation, calculation details described in a previous study [23]) method. An average -value (AVG_Beta) for each CpG locus ranging from 0 (unmethylated) to 1 (completely methylated) was extracted utilizing the GenomeStudio soware and used in further analyses. For the determination of differential methylation between two given groups, we used the Illumina custom error model. is model assumes a normal distribution of the methylation value ( ) among replicates corresponding to a set of biological conditions (TNL, TL, and PTL). We prioritized differentially methylated CpG sites by difference score (DiffScore). DiffScore, which takes into account background noise and sample variability [24], was calculated using the following formula: DiffScore 10 sgn( condition − reference )log 10 , where condition TL / PTL , reference TNL or condition TNL / PTL , reference TL . e resulting differentially methylated CpG sites were annotated with respect to their nearest gene based on the information provided by Illumina. A more detailed description of the Illumina custom error model and the DiffScore has been provided previously [25].

Functional Enrichment Analysis.
Differentially methylated genes (DMGs) with a DiffScore of >20 (equivalent to -value of <0.01) were evaluated for functional enrichment using prede�ned gene sets from the Molecular Signatures Database (MSigDB) [26]. We searched for signi�cantly enriched gene sets by computing overlaps between the lists of DMGs and the CP collection (canonical pathways, 880 gene sets) or the C5 collection (GO gene sets, 1454 gene sets) in the MSigDB. Gene sets with a -value (based on the hypergeometric distribution) less than 0.05 were considered signi�cant.

2.�. �isul�te �e�uencin� ����.
To validate methylation differences revealed by the genome-wide methylation assay, we performed bisul�te sequencing on urocortin (UCN), a gene identi�ed as differentially methylated between the TL and PTL groups, and OXTR, a gene whose methylation status has recently been shown to be important in the pathogenesis of autism [19]. We investigated the methylation status of OXTR, given its signi�cant role in parturition [17] and its labor-associated expression pattern in the amnion [18], which makes it a potential candidate gene for PTB. ere are two OXTR CpG sites targeted by the Illumina In�nium BeadChip assay, both of which were not identi�ed as being differentially methylated. However, because there is currently no evidence supporting the biological importance of the regions containing the two sites, we focused our BS analysis on CpG sites of known biological signi�cance that are located in a different region of the OXTR gene. Primers for UCN were designed to cover the CpG site identi�ed as being differentially methylated by genome-wide methylation pro�ling, using the default parameters of MethPrimer [27]. PCR ampli�cation using the primer pair results in a 278 bp product that spans part of the promoter, exon 1, and part of intron 1 of UCN (-439 to -162 relative to translation start site (TSS)) containing 16 CpG sites. For OXTR, we used the same primers and PCR conditions as those used in the previous study [19]. PCR ampli�cation using the primer set results in a 358 bp product that spans the OXTR promoter (-1195 to -838 relative to TSS) containing 22 CpG sites that has been associated with tissue-speci�c OXTR expression [28] and the development of autism [19]. e regions examined in both genes were located within CpG islands. We carried out our analysis using the same samples assayed on the BeadChips and eight additional independent PTL samples 4 e Scienti�c World Journal (TNL,  ; TL,  ; PTL,  ). Bisul�te-converted DNA was PCR-ampli�ed using ZymoTaq DNA polymerase (Zymo Research). e resulting PCR products were run on an agarose gel and cloned into the pGEM-T Easy vector (Promega, Madison, WI, USA). Individual clones were isolated, ampli�ed following standard protocols, and puri�ed using the PureLink Quick Plasmid Miniprep Kit (Invitrogen, Carlsbad, CA, USA) per manufacturer's instructions. Ten clones per sample, on average, were isolated and sequenced at the University of Iowa DNA facility. Percentage methylation was determined for each CpG site similarly as done in previous work [19]. Statistical analysis was conducted using SigmaPlot 11.0 (Systat Soware, San Jose, CA, USA). Significance of differential methylation (DM) was assessed using the -test (two-tailed), Mann-Whitney (M-W) rank sum test (two-sided), one-way ANOVA, or Kruskal-Wallis (K-W) one-way ANOVA by ranks, as indicated in the text and/or �gure legends. Post hoc analysis following ANOVA was performed using either the Holm-Sidak or Dunn's Method. A was considered signi�cant.

Encyclopedia of DNA Elements (ENCODE) ChIP-Seq
Data. We examined the potential functional signi�cance of the region of the OXTR gene containing CpG sites with statistically different DNA methylation status (CpGs-959 and -1084) using the ChIP-seq data from the ENCODE project available in the University of California Santa Cruz (UCSC) genome browser [29,30]. We speci�cally used the suppressor of zeste 12 homolog (Drosophila) (SUZ12) and Pol2 ChIPseq data generated by the laboratories of Michael Snyder at Stanford University and Vishy Iyer at the University of Texas Austin. e ChIP-Seq data were obtained using human cells (NT2-D1 for the SUZ12 data; GM18526, 18951, 19099, 19193, and ProgFib for the Pol2 data).

RNA Extraction and Real-Time qRT-PCR.
Total RNA was extracted from amnion (TNL, ; TL, ; PTL, ) and decidua (TNL, ; TL, ; PTL, ) tissues using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. e quality of extracted RNA was checked using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Reverse transcription was carried out with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems), using random hexamers as primers following the manufacturer's instructions. Realtime qRT-PCR was performed using synthesized cDNA as a template, gene-speci�c primers (UCN and OXTR) and Power SYBR Green PCR Master Mix (Applied Biosystems). e reactions (including no-template controls) were run in triplicate on the 7900HT Fast Real-Time PCR System (Applied Biosystems) using ACTB (beta actin) [31] as an endogenous reference. Data were analyzed with the SDS 2.4 soware (Applied Biosystems), employing the comparative CT method [32]. Absence of nonspeci�c ampli�cation was con�rmed by dissociation curve analysis. Samples with a value that falls outside ±2 standard deviations of the group mean were de�ned as outliers and removed from the study. Statistical analysis was performed similarly as described above in the bisul�te sequencing section. Data were presented as mean ± standard error of the mean (SEM).

Genome-Wide Patterns of DNA Methylation and Differentially Methylated CpG Loci between Term (Non-Labored and Labored) and Preterm Amnion Tissues.
To investigate the possible involvement of epigenetic mechanisms in the physiology of normal labor and the pathogenesis of PTB, we examined the genome-wide methylation pro�les of the amnion obtained following term (TNL and TL, for each) and preterm (PTL, ) deliveries using the Illumina In�nium BeadChip platform. e overall levels of DNA methylation in the experimental samples were low with third quartile AVG_Beta values between 0.4 and 0.55. Principal component analysis (PCA) placed the pooled TL and PTL samples close to each other and very distant from the pooled TNL sample (Figure 1), which indicates that the genome-wide methylation patterns in amnion tissues from the two spontaneous labor groups (regardless of gestational age (GA) at delivery) are more similar to each other than to those observed in non-labor tissues.
We also performed gene/locus level analysis of differential methylation (DM), searching for methylation changes associated with labor and/or PTB at speci�c CpG sites. Using the Illumina custom error model algorithm, we identi�ed 65 CpG sites in 64 and 61 autosomal genes each that are differentially methylated between the TNL and TL groups and the TL and PTL groups, respectively with a DiffScore of >30 (equivalent to -value of 0.001). Listed in Table  2 are the 15 most highly differentially methylated genes (DMGs). It was noted that among the genes with differentially methylated sites, although very few, were those belonging to special classes of genes, including noncoding RNAs and imprinted genes (such as Down syndrome critical region gene 10 (DSCR10), FBXL19 antisense RNA 1 (FBXL19-AS1), and paternally expressed 10 (PEG10) as shown in Table 2), many of which have regulatory functions in diverse biological processes.

Functional Enrichment Analysis.
To determine the biological signi�cance of DMGs, functional annotation analysis was performed. Our approach involved examining the extent of overlap between our lists of DMGs and prede�ned Statistical signi�cance was determined based on -values calculated from DiffScores. All genes listed here have a DiffScore >40 (corresponding to -value of <0.0001). 2 Genes most highly methylated in the TL group compared to the TNL group. 3 Genes most highly methylated in the PTL group compared to the TL group. 4 De�ned by the CpG island track in the �CSC Genome Browser. 5 An imprinted gene. 6 Non-protein coding genes. 7 A gene identi�ed as having three non-island CpG sites with a DiffScore >40. Abbreviations: TNL: term no labor; TL: term labor; PTL: preterm labor. Note that the TNL sample is placed distantly from the TL or PTL samples, indicating that the TNL group displays distinctly different methylation patterns compared to the other two groups. annotated gene sets from the MSigDB [26] (see Section 2 for further details). For this analysis, we used gene lists with a less stringent -value cutoff of <0.01 (corresponding to a DiffScore of >20), given the small number of DMGs ( ) with a -value below 0.001. We found that 7 gene sets were signi�cantly overrepresented ( < ) in the list of 110 DMGs between the TNL and TL groups. e seven enriched gene sets included cation transport, ion channel activity, and those shown in Table 3, most of which are highly relevant to molecular processes involved in physiologic labor. Among the 186 DMGs between the TL and PTL groups, 17 gene sets were overrepresented. Many of the enriched gene sets were found to be associated with the regulation of cell behavior and extracellular matrix-cell interactions, including focal adhesion, cell junction, cell-substrate adherens junction, and integrin binding (Table 3).

Bi��l�t� S����n�in� �BS� Anal��i� of Di��r�ntial M�th�lation.
To validate DM detected by genome-wide methylation pro�ling, we performed BS analysis on UCN, a gene identi�ed as being overmethylated in the PTL group compared with the TL group with a DiffScore >50 (Table 2). We performed the same analysis on one additional gene named oxytocin receptor (OXTR) whose mRNA and protein expression has been shown to be markedly upregulated in association with labor in primary human amnion epithelial cells [18]. Previous 6 e Scienti�c World Journal T 3: Gene sets overrepresented among differentially methylated genes in amnion tissues from term (TNL, TL) and preterm (PTL) deliveries 1  studies have demonstrated that the methylation status of the promoter region of this gene is associated with tissue-speci�c OXTR expression [28] and the development of autism [19], a disorder linked to PTB [20,21,33,34]. ese �ndings intrigued us to investigate whether DNA methylation could represent one mechanism regulating the labor-associated activity of OXTR in the amnion. We selected the two genes (UCN and OXTR), given their crucial role in normal labor and parturition, which makes them potential candidate genes for PTB. Details on the regions ampli�ed, samples used in the BS experiments, and statistical tests performed for the analysis of the sequencing results are given in Section 2 and Figure 2.
All 16 CpG dinucleotides interrogated in the UCN gene showed some degree of methylation with the ones at positions -361, -335, and -319, being more highly methylated (22.9-55.7%, Table 4) compared with those at other positions (1.1-17.1%). All except two CpG sites were overmethylated in the PTL samples compared to the TL samples, showing the expected direction of DM. However, the differences were not statistically signi�cant.
For OXTR, since we had no priori data on the methylation status of the 22 CpG sites in amnion tissue, all three groups of samples (TNL, TL, and PTL) were examined. Consistent with the �nding of Gregory et al. [19], 5 CpG sites at positions -959, -934, -924, -901, and -860 showed the highest levels (22.2-68.6%, Table 4) and variation in methylation, whereas very little or no methylation (0-5.7%) was observed at the other sites. We found that one (CpG-959) of the �ve sites was signi�cantly differentially methylated among the three groups tested (one-way ANOVA, , Table 4). Pairwise comparisons (Holm-Sidak test) revealed signi�cant differences between the TNL and TL groups ( 7) and the TNL and PTL groups ( ) and borderline signi�cant difference between the TL and PTL groups ( ), demonstrating more distinct differences in methylation at this site between non-labor and labor tissues than between term and preterm tissues.
To determine if the observed DM also occurs in other parts of the placenta where the genes are known to be expressed [17,35], we extended our study to decidua tissues from the same groups of individuals. e decidua, which is of maternal origin, unlike the amnion of fetal origin [4], was selected, given that the function of OXTR in parturition has been well demonstrated in maternal tissue [17], and therefore, the examination of the decidua, along with the amnion, may allow us to compare the methylation state of e Scienti�c World �ournal 7 T 4: UCN and OXTR promoter methylation status in the amnion and decidua from term (TNL, TL) and preterm (PTL) deliveries 1 . Presents average % methylation at each CpG site. 2 Nucleotide positions relative to translation start site. 3 CpG sites in UCN with the lowest -value in each tissue type. 4 CpG sites methylated at higher levels in both tissues than the average methylation level of all sites examined. 5 CpG sites with statistically signi�cant ( ) differential methylation in either tissue. Details of statistical tests used are described in Section 2. Abbreviations: TNL: term no labor; TL: term labor; PTL: preterm labor.

UCN
the OXTR gene and possibly its importance in both fetal and maternal tissues.
e overall methylation patterns observed in the decidua were similar to those identi�ed in the amnion. However, unlike in the amnion tissues, the methylation levels not at CpG-959, but at different sites (CpGs-934 and -1084), were found to be statistically signi�cantly different ( , 0.008, resp., K-W one-way ANOVA by ranks) among the three groups of the decidua tissues (Table 4). e CpG-1084 site, interestingly, was completely unmethylated in the TL group, whereas it was methylated to some small degree in the other two groups (TNL, 10%; PTL, 8.7%) ( Table 4). Signi�cant differences between the TL and TNL or PTL groups were con�rmed by Dunn's post hoc test ( ). �n the case of CpG-934, the difference was signi�cant only between the TL and PTL groups. Taken together, it appears that there exist compartment-speci�c OXTR methylation patterns in the placenta.

Analysis of UCN and OXTR Gene Expression in the Amnion and Decidua.
To evaluate the functional signi�cance of the methylation status of the two genes, we performed gene expression analysis using qRT-PCR on an extended set of amnion and decidua tissues ( , 59, resp.) from the three groups. Although the DM of UCN was not validated by BS, we observed a statistically signi�cant 2.3-fold increase in its transcript levels in the PTL amnion samples compared to the TNL and TL samples ( , K-W one-way ANOVA by ranks, Figure 3). ere was also a statistically signi�cant, but less than twofold increase in OXTR mRNA levels in the PTL amnion samples compared with the TL samples ( , K-W one-way ANOVA by ranks, Dunn's post hoc test, Figure 3). e results were not replicated in the decidua samples for either gene. ese �ndings suggest that the upregulation of UCN is speci�c to the amnion from spontaneous preterm deliveries, and that the DM observed in OXTR may not correlate with OXTR expression given that methylation generally plays a role in gene silencing.

Methylation�Speci�c PCR �MSP� Analysis of Di�eren� tial Methylation.
As an alternative approach to validate DNA methylation differences captured by our genome-wide methylation study, we carried out methylation-speci�c PCR (MSP) for selected 3 DMGs between the TNL and TL groups (PEG10) and between the TL and PTL groups (LPAR5 and SLC30A3). Analysis of the same set of amnion samples used in BS revealed no intergroup differences in PEG10 and LPAR5 methylation (data not shown). However, the methylation status of SLC30A3 was in good agreement with our genome-wide methylation data with methylated MSP products present in 10 out of 14 (71%) PTL samples and none of the TL samples ( Figure 4).

Discussion
e present study investigated if there exist unique genomewide methylation signatures that distinguish among term (non-labored and labored) and preterm amnion tissues. Our methylation pro�ling revealed a higher degree of similarity between the methylation patterns in the TL and PTL pooled samples than those observed in the TNL pooled sample, suggesting the potential role of methylation in the regulation of labor, independent from GA. We identi�ed a relatively small number of DMGs between the TNL and TL groups and the TL and PTL groups (65 genes each) at the 1 signi�cance level. is observation may be attributed to the small sample size and the sample-to-sample variability related to GA. Gene set enrichment analysis of those genes revealed signi�cant overrepresentation of pathways that appear to be functionally relevant (Table 3). e enrichment of pathways related to ion transport, ion channel activity, and cytokine production among the DMGs between the TNL and TL groups re�ects biochemical and molecular events associated with the onset of labor, which, along with e Scienti�c �orld �ournal 9 hormonal factors, help to initiate parturition. ese results are at least partially in line with previous gene expression pro�ling studies reporting labor-associated cytokine-related gene signatures in human amniotic [36] and chorioamniotic [37] membranes. e overrepresentation of heart-(development and contraction) related gene sets may be explained by the presence of myo�broblasts in the connective tissue of the amnion [38], which have contractile ability [39], and hence are involved in heart rhythm regulation [40] and, possibly, prevention of excessive distension of the amniotic membrane [38]. e DMGs between the TL and PTL groups were enriched in gene sets involved in cell adhesion, cell-cell and cell-extracellular matrix interactions, which have crucial roles in the modulation of cellular behavior and tissue maintenance and organization [41]. is observation con�rms the importance of intact fetal membranes as a critical factor in the maintenance of pregnancy. Another overrepresented gene set was the negative regulation of transferase activity. Given the versatile roles of transferases, differential methylation of this group of genes (including HEXIM1, SFN, CBLC, and DUSP2) may in�uence a wide range of cellular processes in a way that interferes with timely onset of labor and parturition. Among these genes, DUSP2 has previously been documented as being signi�cantly upregulated following interleukin-1 (IL-1 ) stimulation in myometrial cells [42], suggesting its potential role in the mediation of uterine contractions. It would be intriguing to examine how the activity of DUSP2 in the amnion may contribute to the process of parturition.
Our study at the individual gene level using BS revealed three CpG sites (CpGs-934, -959, and -1084) in OXTR that exhibit signi�cant DM among the three groups of amnion and decidua tissues, which are of fetal and maternal origin, respectively [4]. Subsequent gene expression analysis demonstrated no correlation between gene expression and methylation and therefore, the functional signi�cance of the observed DM remains undetermined. Previous work showed that site-speci�c methylation can result in transcriptional alterations through its effects on the interaction of transcription factors (TFs) with its cognate DNA sequence [43]. Currently, there are no known TF binding sites around CpG-959, which was previously identi�ed as signi�cantly hypermethylated in peripheral blood mononuclear cells from autistic patients compared with those from control patients [19]. However, Gregory et al. [19] have indicated that CpG-934, whose differential methylation has also been associated with autism, falls within predicted binding domains for vrel reticuloendotheliosis viral oncogene homolog (avian) (c-Rel), zinc �ngers and homeoboxes 2 (�H�2), and lectin, galactoside-binding, soluble, 4 (LGALS4). Using ENCODE ChIP-seq data available in the UCSC genome browser, we also found that CpG-1084 falls within putative binding sites for SUZ12 and Pol2 (see Section 2 for more details), which warrants future studies to dissect the impact of the methylation status at this speci�c dinucleotide on the interactions between these TFs and their binding sites.
Despite the lack of any signi�cant difference in UCN methylation levels between the TL and PTL groups, our observation of a signi�cant, more than 2-fold increase in UCN mRNA levels in the PTL amnion tissues compared with the term tissues suggests a potential role of this gene in the etiology of PTB, which encodes an endogenous ligand for corticotropin releasing hormone receptor (CRHR) that mediates the action of CRH, one of the major endocrine factors in parturition [44]. Given that there are several putative binding sites for TFs (such as C7EBP, GATA, and MyoD) [45] upstream of the region examined in this study, it would be intriguing to investigate whether the methylation status of CpG dinucleotides encompassing those sites correlates with the observed gene expression patterns. It would also be worthwhile to examine if mechanisms other than methylation underlie the transcriptional regulation of UCN in the amnion.
Our MSP analysis identi�ed another gene (SLC30A3) that might play a role in pathogenic processes of PTB. is gene, also known as ZNT3, encodes a zinc transporter responsible for zinc efflux from the cytoplasm to extracellular spaces or intracellular organelles [46]. Given the differential expression of SLC30A3 in relation to dietary zinc and/or glucose supply in mouse placenta [47] and beta cells [48], it is postulated that its dysregulated expression due to aberrant methylation in human amnion may in�uence nutritional homeostasis during pregnancy, ultimately, leading to PTB.
Our work was limited by the small sample size and the lack of control for gender-speci�c methylation differences [49,50]. Another major limitation is that the PTL tissues were examined as a pooled sample, not individually. Previous studies have demonstrated that pooled DNA samples can be used to provide a reliable estimate of average group methylation when analyzed using high-throughput techniques such as MALDI-TOF mass spectrometry [51,52]. erefore, a DNA pooling approach using such systems could be employed in future studies for large-scale assessment of methylation variations in maternal and fetal tissues. Very recently, it has been shown that neonatal DNA exhibits a considerable degree of GA-associated variability in DNA methylation patterns [53]. Given this �nding, a precisely strati�ed analysis based on GA may allow a more accurate characterization of DNA methylation pro�les associated with term and preterm pregnancies.

Conclusion
is work provides preliminary evidence that DNA methylation changes may play at least a partial role in physiologic labor and the etiology of PTB, and suggests that DNA methylation pro�les, together with other types of biological data, hold a promise for the identi�cation of genes involved in normal parturition and preterm birth.

Con�ict o� �nte�ests
e authors declare that there is no con�ict of interests.