The role of DNA methylation in human trophoblast differentiation

ABSTRACT

The placenta is a vital fetal exchange organ connecting mother and baby. Specialised placental epithelial cells, called trophoblasts, are essential for adequate placental function. Trophoblasts transform the maternal vasculature to allow efficient blood flow to the placenta and facilitate adequate nutrient uptake. Placental development is in part regulated by epigenetic mechanisms. However, our understanding of how DNA methylation contributes to human trophoblast differentiation is limited. To better understand how genome-wide methylation differences affect trophoblast differentiation, reduced representation bisulfite sequencing (RRBS) was conducted on four matched sets of trophoblasts; side-population trophoblasts (a candidate human trophoblast stem cell population), cytotrophoblasts (an intermediate progenitor population), and extravillous trophoblasts (EVT, a terminally differentiated population) each isolated from the same first trimester placenta. Each trophoblast population had a distinct methylome. In line with their close differentiation relationship, the methylation profile of side-population trophoblasts was most similar to cytotrophoblasts, whilst EVT had the most distinct methylome. In comparison to mature trophoblast populations, side-population trophoblasts exhibited differential methylation of genes and miRNAs involved in cell cycle regulation, differentiation, and regulation of pluripotency. A combined methylomic and transcriptomic approach was taken to better understand cytotrophoblast differentiation to EVT. This revealed methylation of 41 genes involved in epithelial to mesenchymal transition and metastatic cancer pathways, which likely contributes to the acquisition of an invasive EVT phenotype. However, the methylation status of a gene did not always predict gene expression. Therefore, while CpG methylation plays a role in trophoblast differentiation, it is likely not the only regulatory mechanism involved in this process.

KEYWORDS:

• Placenta
• trophoblast
• cytotrophoblasts
• extravillous trophoblasts
• DNA methylation

Introduction

The placenta is the first human organ to develop and forms the vital connection between the mother and fetus in utero. The placenta primarily functions as an exchange organ, allowing for nutrients, gases, and wastes to be transferred between the maternal and fetal circulations. The placenta is often described as the ‘tree of life’, in part due to its branching villous structure. Each placental villus is comprised of an outer bilayer of trophoblasts and an inner mesenchymal core containing fetal blood vessels (Figure 1(a)). Trophoblasts are the functional epithelial cells of the placenta of which there are three mature subtypes; cytotrophoblasts, extravillous trophoblasts (EVTs) and the syncytiotrophoblast. These mature trophoblast populations are thought to originate from a trophoblast stem cell (TSC) population (Figure 1(b)) [1]. The first differentiation step that human TSCs make is traditionally believed to be to villous cytotrophoblasts, a highly proliferative progenitor population that facilitates rapid early placental growth [2]. Cytotrophoblasts can then either 1) fuse to form the outer multinucleated syncytiotrophoblast that transports nutrients and oxygen from the maternal blood, or 2) in the tips of anchoring villi, differentiate into EVTs that migrate away from the placenta and invade the decidualised maternal endometrium (decidua) (Figure 1(a)) [1]. In the decidua, EVTs transform the maternal spiral arteries eroding the smooth muscle of the vessel walls and dilating the arteries to ensure an increased amount of maternal blood can perfuse the placenta in an uninterrupted fashion ensuring maximal exchange.

Figure 1. (a) Schematic diagram of a first trimester human placental villus showing the three major mature trophoblast populations. Extravillous trophoblasts are shown invading into the decidua and remodelling a maternal spiral artery (b) Flow diagram showing the lineage differentiation pathway of trophectoderm derived trophoblast populations.

Impaired trophoblast differentiation is associated with pregnancy disorders such as fetal growth restriction (where babies are born pathologically small) and preeclampsia (when maternal blood pressure becomes dangerously elevated) [3]. However, our understanding of how and why trophoblast differentiation is inadequate in these pregnancy disorders is poor. What we do know is that the pathophysiology of pregnancy disorders is established early in pregnancy [3]. Thus, understanding the regulation of trophoblast differentiation in the first trimester is key to unlocking the underlying mechanisms of these disorders.

Studies in rodent models highlight the importance of DNA methylation in placental development. Inhibiting methylation in pregnant rats results in smaller placentae with anatomical abnormalities and altered ratios of trophoblast populations [4]. Similar outcomes are observed in knockout models of DNA methyltransferases where embryonic lethality results from defective trophoblast differentiation [5]. The impact of DNA methylation on trophoblast development begins at approximately day 5 post fertilization when the morula splits into the hypomethylated trophectoderm (that gives rise to the trophoblast populations of the placenta) and hypermethylated inner cell mass (that gives rise to the embryo proper) [6]. Segregation of murine TSC from embryonic and extra-embryonic populations (e.g., embryonic stem cells (ESC) and extraembryonic endoderm XEN cells) is associated with distinct methylation signatures in key genes including caudal type homeobox 2 (Cdx2), eomesodermin (Eomes), E74 like ETS transcription factor 5 (Elf5), placenta expressed transcript 1 (Plet1) and transcription factor AP-2, gamma (Tcfap2c), and these methylation changes can be correlated with transcriptional changes between these lineages [6–8]. In contrast, the role of DNA methylation in regulating murine trophoblast differentiation has been explored far less [9]. However, it is clear that while some genes (eg. Elf5 promoter hypomethylation) have similar levels of methylation between trophoblast subsets, differential gene methylation (eg. dimethylarginine dimethylaminohydrolase 2 (Ddah2) enhancer region) exists between the different murine trophoblast lineages [8–11]. While some transcriptional and regulatory differences exist between rodent and human trophoblast lineage segregation, significant anatomical differences exist in early murine and human trophoblast differentiation pathways, and our understanding of how methylation networks translate across species is incomplete.

Human trophoblast differentiation is associated with considerable gene expression changes [12]. However, there is little understanding of how DNA methylation influences these transcriptional pathways as studies often focus on only a handful of genes or provide an overall picture of methylation data [13–16]. Until very recently, the lack of a human TSC population has limited our ability to understand the influences of DNA methylation on early trophoblast differentiation events. Recent advances employing a novel human TSC population, as well as other candidate trophoblast progenitor and stem cell populations may help shed light on the methylation events involved in early trophoblast differentiation [16–19]. Our laboratory works with a candidate TSC population termed side-population trophoblasts, which have some characteristic features of stem cells, both in their ability to rapidly efflux the DNA binding dye Hoechst 33,342 and at the level of the transcriptome [19]. DNA methylation is known to play key roles in stem cell maintenance and promotion of lineage commitment. In particular DNA methylation helps regulate stem cell senescence, renewal, and differentiation [20]. However, much of our understanding of this comes from stem cells that derive from the relatively hypermethylated inner cell mass of the embryo (eg. human embryonic stem cells, neuronal stem cells, mesenchymal stem cells), and it is unclear how they relate to TSC differentiation, which stems from the hypomethylated trophectoderm [20]. In murine TSC hypomethylation of Elf5 and Plet1 promoters contribute to establishing the TSC state and allowing TSC self-renewal [6]. Whether similar DNA methylation events influence human TSC function and differentiation is not yet known however it has been previously reported that that there is a degree of differential methylation between promoter regions of ELF5 between human cytotrophoblasts (3.4% methylated) and EVTs (9.1% methylated) [13].

It is clear that DNA methylation differences exist between primary human trophoblast populations [21]. These methylation changes are known to influence expression of genes involved in regulating; cell proliferation, the immune system, and cell invasion [14,15,22]. Studies using trophoblast-like cell lines (JAr, Jeg-3, TEV-1, BeWo and HTR8/SVneo cells) also suggest that DNA hypermethylation downregulates tumor suppressor genes, whilst DNA hypomethylation allows upregulation of epithelial to mesenchymal transition (EMT) genes aiding acquisition of an invasive and migratory phenotype [23]. However, DNA methylation has not been examined directly in uncultured trophoblasts and the impact of in vitro culture conditions on trophoblast DNA methylation are unclear. Therefore in this work we aimed to gain insights into the epigenetic landscape during early pregnancy by studying genome-wide methylation profiles in freshly isolated, first trimester trophoblast populations.

Methods

Isolation of trophoblast populations

Use of first trimester placental tissue in this study was approved by the Northern Regional Ethics Committee (NTX/12/06/057/AM04). Using a Becton Dickinson Aria II SORP, side-population trophoblasts ($\bar{x}$ = 97,899 cells/sample, n = 13), cytotrophoblasts ($\bar{x}$ = 201,029 cells/sample, n = 13) and EVTs ($\bar{x}$ = 110,427 cells/sample, n = 13) were isolated from human first trimester placentae of 5.6–12.2 weeks of gestation as previously described (Supplementary Table 1) [19]. Live cell populations were sorted directly into 200 µL of sterile PBS in 2 mL RNAse-free Eppendorf tubes for DNA extraction (n = 4 placentae), or into 200 µL of RNA Later® (Cat no. AM7021, Ambion) for RNA (n = 5 placentae) or miRNA (n = 4 placentae) extraction.

Table 1. Table of sequenom massARRAY primers.

Primer Name	Primer Sequence
LOC100289580 Forward	aggaagagagGTATATTTGTGAGTAGATTTGGGGG
LOC100289580 Reverse	cagtaatacgactcactatagggagaaggctAAATAATTCCCATAAACCCTAAAAA
GALNT6 Forward	aggaagagagGTTTGGTGGAGAGGAAGTTTGTAT
GALNT6 Reverse	cagtaatacgactcactatagggagaaggctCCTCTACCCTTTAACCCTCAAAAAT
CPZ Forward	aggaagagagGGTTGGTTGTATTGTTTTTTATATGTTT
CPZ Reverse	cagtaatacgactcactatagggagaaggctAAAAAAAACCTCACCTAATAACCCA
C22orf31 Forward	aggaagagagTTTGAGGTAGAGTTATTGGGTTTTTT
C22orf31 Reverse	cagtaatacgactcactatagggagaaggctCCAATACCACTATTCCTCCAATAAC
GATA2 Forward	aggaagagagTTTTTATTTTTTTTGGTGAGGATTG
GATA2 Reverse	cagtaatacgactcactatagggagaaggctAACTCCTACCCAACAACCCC
CARD9 Forward	aggaagagagAGTTGTATAAGAAGGTTATAGGTAAGGA
CARD9 Reverse	cagtaatacgactcactatagggagaaggctACAACTAAATCAAACCTAACTCCCC

DNA extraction

Sorted cells were centrifuged at 450 g for 30 minutes at 4ºC. The supernatant was removed, and cells were resuspended in 200 µL of sterile PBS to which 20 µL of Qiagen Proteinase K was added. A 200 µL volume of AL buffer from a Qiagen QIAAMP DNA Mini Kit (Cat no. 51304, Qiagen) was added, and samples were incubated overnight at 56ºC. The next day, DNA was extracted using a Qiagen QIAAMP DNA Mini Kit as per the manufacturer’s instructions. The DNA was eluted into 200 µL of Qiagen DNA Mini Kit AE buffer and stored at −20ºC. At the time of DNA extraction, DNA quantity was assessed using the Qubit® dsDNA HS Assay Kit (Cat no. Q32851, Molecular Probes) or Qubit® dsDNA BR Assay Kit (Cat no. Q32850, Molecular Probes).

Reduced representation bisulfite sequencing preparation and analysis

DNA extracted from matched sets (from the same placenta) of side-population trophoblasts, cytotrophoblasts and EVTs were sent to New Zealand Genomics Limited to undergo reduced representation bisulfite sequencing (RRBS). All samples were prepared and sequenced at the same time in order to mitigate batch effects and RRBS libraries were prepared following previously published protocols [24]. Briefly, 500 ng of DNA was digested overnight using Msp1 enzyme (Cat no. R0106S, New England Biolabs). DNA fragments then underwent end-repair and ‘A’-tailing at the 3ʹ ends. Methylated adaptors were ligated and DNA was purified using MinElute PCR purification kit (Cat no. 28006, Qiagen). Samples were eluted into 20 µL TE. Samples then underwent bisulfite conversion using the EZ DNA Methylation Kit (Cat no. D5001, Zymo Research) following the manufacturer’s protocol, followed by PCR amplification with Pfu Turbo Cx Hotstart DNA polymerase (Cat no. 600410, Agilent Technologies). The products were run on a 3% (w/v) NuSieve agarose gel and appropriate sized (150–325 bp) fragments were selected. DNA libraries were purified using the MinElute Gel Extraction Kit (Cat no. 28006, Quiagen) according to the manufacturer’s instructions. DNA quantity was assessed using a Qubit fluorimeter (Life Technologies). High sensitivity DNA chips run on a 2100 Bioanalyser (Agilent Technologies) were used to determine average fragment size. The RRBS libraries underwent single-end (100 bp) sequencing using the Illumina HiSeq2000 platform.

In total 12 libraries were prepared. Following ‘Processing RRBS samples: a User Guide’ by Peter Stockwell [25], data files were trimmed and cleaned. FastQC reports of each cleaned data file determined they were acceptable for downstream applications. Bismark was then used to map the files to the Homo sapiens GRCh37 reference genome. On average across the 12 samples investigated by RRBS, 17,850,057 reads (±948,819 reads SEM) were obtained, 10,688,624 reads (±563413SEM) were aligned, the depth of coverage was 6.23 reads (±0.41 reads SEM), the proportion of C methylated in a CpG context was 30.62% (±0.605% SEM), and the samples had a bisulfite non-conversion rate of 1.28% (±0.05% SEM). The data was analysed using the R package ‘methylKit’ to determine the methylation at individual CpG sites (Supplementary tables 2–4) [26]. P-values were generated using logistic regression and a q-value cut-off of <0.01, and percentage methylation difference of >25% was employed by methylKit to identify differential methylation events. Gene enrichment analysis was undertaken using the GATHER pathway analysis online tool. These data have been deposited in the Gene Expression Omnibus (GEO) database with accession number GSE109682.

Table 2. Table of microRNA primers.

miRNA	Primer Sequence
Universal Primer	GCATAGACCTGAATGGCGGTA
hsa-let-7b-5p	CGTTCTGAGGTAGTAGGTTGTGTG
hsa-miR-637	CTTTCGGGCTCTGCGTAAA
hsa-miR-193b-3p	GCCCTCAAAGTCCCGCTAA
hsa-miR-525-3p	GAAGGCGCTTCCCTTTAGA
hsa-miR-103a-3p	CAGCATTGTACAGGGCTATGAA
hsa-miR-21-5p	GCTAGCTTATCAGACTGATGTTGAAA

Table 3. Summary table of differential methylation between trophoblast populations.

	SP v CTB	SP v EVT	CTB v EVT
Total differentially methlated CpG sites	158	2991	7061
Hypermethylated CpG sites	11.4% (18)	95% (2842)	96.6% (6819)
Hypomethylated CpG sites	88.6% (140)	5% (149)	3.4% (242)
Total genes impacted	32	1487	3185
Hypermethylated genes	10	1420	3059
Hypomethylated genes	18	45	74
Genes containing both hypo and hypermethylated CpG sites	4	22	52

Table 4. Table of differentially methylated genes regulating a stem cell phenotype and/or cell cycle regulation and proliferation in side-population compared to cytotrophoblasts or extravillous trophoblasts.

Gene	Location	Methylation difference in side-population trophoblasts compared to cytotrophoblasts	q. value	Methylation difference in side-population trophoblasts compared to extravillous trophoblasts	q. value	Function/ Biological Relevance
PRDM2	Chr1: 14109294	3.8	0.9690	−32.6	0.0004	Maintains muscle stem cells in a quiescent state. Aids stem cell self-renewal [56].
	Chr1: 14109300	4.2	0.9690	−28.9	0.0041
LRP5	Chr11: 68139919	0.5	0.9690	5.0	0.0006	Canonical Wnt signalling through LRP5 permits the regenerative capacity of mammary epithelial stem cells [57].
	Chr11: 68141102	0.1	0.9690	32.3	0.0094
	Chr11: 68141123	4.0	0.9690	36.7	0.0005
	Chr11: 68097308	6.2	0.9363	−33.4	0.0026
	Chr11: 68097314	3.9	0.9690	−31.6	0.0029
	Chr11: 68153864	20.8	0.7617	−37.6	0.0002
TPT1	Chr13: 45914874	34.1	<0.0001	35.4	<0.0001	Enhances Oct4 and Nanog expression in embryonic stem cells [42].
	Chr13: 45914880	34.7	<0.0001	35.5	<0.0001
RCC1	Chr1: 28845226	14.1	0.0039	14.1	0.0002	Critical cell cycle regulator involved in the S-phase of mitosis [58].
	Chr1: 28845238	14.1	0.0042	13.2	0.0028
PMF1	Chr1: 156186626	10.6	0.0003	8.3	0.0021	Regulates transcription of genes involved in cell growth and cell death [59].
	Chr1: 156186325	−5.0	<0.0001	8.3	0.0021
	Chr1: 156186433	−2.8	0.0004	−1.2	0.3362
	Chr1: 156186383	−0.1	0.9690	1.6	0.0070
	Chr1: 156186392	−0.7	0.9193	2.3	<0.0001
	Chr1: 156186393	−0.2	0.9690	1.6	0.0008
	Chr1: 156186496	−1.1	0.9193	3.1	0.00123
B4GALT6	Chr18: 29265352	11.3	0.0088	10.3	0.0220	Required for normal proliferation and differentiation of epithelial cells. Murine knockout models show growth retardation and premature fetal death [60].
MAPK11	Chr22: 50717438	−34.5	0.0086	9.7	0.7256	Required for cytotrophoblast differentiation to syncytiotrophoblast. Inhibition enhances cytotrophoblast proliferation [61].
DHX34	Chr19: 47852601	17.6	<0.0001	17.6	<0.0001	Involved in nonsense-mediated mRNA decay, a process important for stem cell differentiation [62].
	Chr19: 47852604	17.2	<0.0001	17.2	<0.0001
	Chr19: 47852629	18.2	<0.0001	18.2	<0.0001
MDM4	Chr1: 204537885	17.0	0.9193	−41.1	0.0010	Promotes cell proliferation though inhibition of the p53 pathway [63].
	Chr1: 204537886	0.7	0.9680	−57.2	<0.0001
	Chr1: 204537940	−1.9	0.9690	−30.5	0.0022
HOXB1	Chr17: 46570458	−0.6	0.9690	−28.6	0.0031	Mediates neural stem and progenitor proliferation and differentiation [64].
DCAF12L1	ChrX:125715240	−3.6	<0.0001	−1.2	0.3769	Expressed in murine TSC and is dcaf12L is down-regulated upon Elf5 depletion [34].
	ChrX:125715250	−3.7	<0.0001	−1.5	0.1722
	ChrX:125715261	−3.0	0.0002	−2.6	0.0003
	ChrX:125715276	−3.0	0.0002	−2.5	0.0007
	ChrX:125715286	−3.0	0.0004	−2.1	0.0011
	ChrX: 125715243	−2.1	0.0272	−2.7	<0.0001
	ChrX: 125715281	−1.0	0.9193	−2.1	0.0011
STAB1	Chr3: 52520822	4.9	0.9690	−35.1	0.0006	Required for rat TSC self-renewal and maintenance [35].

Sequenom MassARRAY EpiTYPER analysis

In order to interrogate CpGs of interest, primers specific to corresponding genomic regions were designed using the EpiDesigner (Agena Bioscience) online tool. The primers were then validated using the ‘MassArray’ R package (Thompson and Greally 2009) and NCBI BLAT tool. The selected primers (Table 1) (HPLC grade, Integrated DNA technologies, Inc.) were then used to amplify the regions of interest from the bisulfite converted DNA using the EpiTYPER® Reagent and Chip Set reagents. The bisulfite conversion was performed with an EZ DNA methylation kit using 200 ng of DNA from each sample (12 trophoblast samples, as well as 0% and 100% methylation controls (Cat no. D5014, Zymo Research)). Water was included as a negative control. 1 µL of each PCR product was run on a 1.5% Agarose gel to ensure successful amplification of the target regions. The successful PCR reactions were sent to Grafton Clinical Genomics (GCG), University of Auckland, for processing and analysis. At GCG, the PCR products were subjected to SAP treatment and T7 conversion, followed by spot firing and Quantitative DNA Methylation Analysis using the EpiTYPER software (Agena Bioscience). Mean methylation was calculated for the each CpG of interest identified in the RRBS data set. Statistical analysis was conducted using GraphPad Prism v7.03 using a one-way ANOVA and Bonferroni post-test correction.

miRNA extraction and quantification

RNA was extracted using a mirVana™ miRNA Isolation Kit (Cat no. 4464066, Ambion) and eluted into 100 µL of nuclease-free water. cDNA was generated using a qScript™ microRNA cDNA Synthesis Kit (Cat no. 95107–100, Quanta Biosciences™) and real-time PCR was conducted using a Quanta PerfeCTa SYBR Fastmix Low ROX mix (Cat no. 95074–250, Quanta Biosciences™) and run on Applied Biosystems Quant Studio K12 Flex as per the manufacturer’s instructions using the primers listed in Table 2. Gene expression was normalised to the geometric mean of three stably expressed miRNAs (miR525-3p, miR103a-3p and miR21-5p) using the ΔΔC_T method [27].

Transcriptomic comparison of cytotrophoblasts and EVT gene expression

RNA was extracted using an Invitrogen PureLink Mini-Kit (Cat no. 12183018A, Life Technologies) according to the manufacturer’s instructions and evaluated for quality and quantity using Experion RNA HighSens chips (Bio-Rad) as previously described [19]. cDNA was generated and amplified using a NuGEN Ovation Pico WTA system V2 kit in accordance with the manufacturer’s instructions (Cat no. NUG-3302–12, NuGEN). Purified cDNA was biotin labelled and hybridised to Affymetrix PrimeView microarrays (Cat no. 901837, Affymetrix).

Microarray data were quality checked and lacked obvious outliers. Data were normalised using the ‘limma’ R package as previously described [19]. Genes that were significantly different between cytotrophoblasts and EVTs were identified after Benjamini–Hochberg false-discovery rate (FDR) correction, and a cut-off coefficient of 1.2 (corresponding to a 2.29-fold change) with corrected p-values <0.05 were used. These data can be accessed using the GEO database with accession number GSE57834.

Results

Trophoblast populations have distinct methylomes

Hierarchical clustering identified that the greatest methylation differences resulted from variation between placentae, rather than by cell type (Figure 2(a)).Despite these placenta specific differences, the three trophoblast populations had distinct methylomes. When the methylome of side-population trophoblasts were compared to cytotrophoblasts 158 CpG sites were differentially methylated between the groups, corresponding to 32 genes (Figure 2(b), Table 3, Supplementary Table 2). 2991 CpG sites, corresponding to 1487 genes, were differentially methylated between side-population trophoblasts and EVTs (Figure 2(b), Table 4, Supplementary Table 3). Finally, 7061 CpG sites, corresponding to 3185 genes, were differentially methylated between cytotrophoblasts and EVTs (Figure 2(b), Table 5, Supplementary Table 4). Taken together, these data indicate that the methylome of side-population trophoblasts and cytotrophoblasts are more closely related, while EVT have a more distinctive methylome.

Table 5. Table of miRNA involved in signalling pathways regulating stem cells.

miRNA differentially methylated between side-population trophoblasts and extravillous trophoblasts	miRNA differentially methylated between cytotrophoblasts and extravillous trophoblasts
hsa-miR-3194-5p	hsa-miR-3194-5p
hsa-miR-193b-3p	hsa-miR-193b-3p
hsa-miR-657	hsa-miR-657
hsa-miR-2278	hsa-miR-2278
hsa-miR-339-5p	hsa-miR-339-5p
hsa-miR-3177-5p	hsa-miR-3177-5p
hsa-miR-2110	hsa-miR-2110
hsa-miR-339-3p	hsa-miR-339-3p
hsa-miR-193b-5p	hsa-miR-193b-5p
hsa-miR-548au-5p	hsa-miR-548au-5p
hsa-miR-3174	hsa-miR-3174
hsa-miR-548au-3p	hsa-miR-548au-3p
hsa-miR-637	hsa-miR-637
hsa-miR-200a-3p	hsa-miR-130a-3p
hsa-miR-23a-3p	hsa-miR-150-5p
hsa-miR-200a-5p	hsa-miR-3619-5p
hsa-miR-23a-5p	hsa-miR-140-3p
hsa-let-7b-3p	hsa-miR-2467-5p
	hsa-miR-326
	hsa-miR-330-3p
	hsa-miR-330-5p
	hsa-miR-33a-3p
	hsa-miR-365a-3p
	hsa-miR-708-5p
	hsa-miR-5582-3p
	hsa-miR-140-5p
	hsa-miR-33a-5p
	hsa-miR-135b-5p
	hsa-miR-4783-3p
	hsa-miR-150-3p
	hsa-miR-135b-3p
	hsa-miR-548z
	hsa-miR-199a-3p
	hsa-miR-3943
	hsa-miR-4487
	hsa-miR-378i
	hsa-let-7a-2-3p
	hsa-miR-4525
	hsa-miR-199a-5p
	hsa-miR-708-3p
	hsa-miR-587
	hsa-miR-4786-3p
	hsa-miR-138–2-3p
	hsa-miR-944
	hsa-miR-5008-5p
	hsa-miR-130a-5p
	hsa-miR-5008-3p
	hsa-miR-3619-3p

Figure 2. (a) Cluster dendogram showing hierarchical clustering based on global methylation for side-population trophoblasts (SP), cytotrophoblasts (CTB) and extravillous trophoblasts (EVT) from the four placentae used in this work (denoted by numbers 1–4). (b) Venn diagram of differentially methylated CpG sites between SP and CTB, SP and EVT or CTB and EVT.

Using a genome-wide approach to investigate DNA methylation allows the opportunity to determine the location (introns, exons, promoters, intergenic regions) where DNA methylation modifications may be occurring. Methylation events in promoter regions influence gene expression, while methylation events at exons influence repetitive DNA elements and gene splicing. The role of methylation at intra- and intergenic regions remains more unclear, although there is evidence that differential methylation in these regions can fine tune gene expression [28–31]. When the distribution of all CpGs covered by the 12 trophoblast samples was investigated the majority of CpGs identified were located in either introns (34.21%, Figure 3(a)) or intergenic regions (35.51%, Figure 3(a)). Only 17.50% of CpG were located in promoter regions (Figure 3(a)) and even fewer CpGs in exons (12.77%, Figure 3(a)). However, when trophoblast populations are compared different patterns of CpG methylation are present. The same patterns of genome methylation were observed in all three trophoblast comparison groups (side-population trophoblasts vs cytotrophoblasts, side-population trophoblasts vs EVTs, cytotrophoblasts vs EVTs), with the most differential methylation occurring in promoter regions (>54%, Figure 3(b-d)) and the least occurring in exons (<10%, Figure 3(b-d)). The remaining methylation events (>35%) occurred in intra- and intergenic regions (Figure 3(b-d)). The high proportion of differential methylation in promoter and intra- and inter-genic regions indicates a regulatory role of DNA methylation in promoting trophoblast differentiation through control of gene expression at the promoter region or through the use of short and long non-coding RNA respectively.

Figure 3. (a) Pie chart showing the mean distribution of CpGs in promoter, exon, intron and intergenic regions across all 12 trophoblast samples. b-d) Pie charts showing the percentage of differentially methylated CpG located in promoter, exon, intron and intergenic regions when (b) side-population trophoblasts (SP) are compared to cytotrophoblasts (CTB), (c) SP are compared to extravillous trophoblasts (EVT), or (d) CTB are compared to EVT.

RRBS data was validated using Sequenom MassARRAY. In total 16 CpG sites across 6 genes (LOC100289580, CARD9, GALNT6, CPZ, GATA2, C22orf31) that were significantly differentially methylated in RRBS analysis were investigated by Seqenom MassARRAY (Figure 4(a), Supplementary Table 5). In all of the significantly different methylation events examined in the three trophoblast populations (Figure 4(a)) biologically similar changes in the direction of percentage of methylation were observed (i.e., methylation was in the same direction for each trophoblast population) (Figure 4(b-g), Supplementary Figure 1), but not all biological changes were statistically significant (Figure 4(b-g)). In addition, a number of CpGs examined across the Sequenom MassARRAY fragment were also identified as significantly differentially methylated (Figure 4(b-g)).

Figure 4. (a) Bar graph of the percentage methylation at 16 CpG across 6 genes that were significantly differentially methylation when analysed using RRBS and chosen for further validation using Sequenom methylation analysis in side-population trophoblasts, cytotrophoblasts and extravillous trophoblasts. Each number correlates to the CpG in the Sequenom amplicon presented in Figure 4 plots B-G. (b-g) Dot plots showing percentage methylation of CpGs in GATA2 (b), CPZ (c), CARD9 (d), GALNT6 (e), LOC100289580 (f), and C22orf31 (g) in SP, CTB, and EVT when analysed by Sequenom MassARRAY EpiTYPER Analysis. * indicates an adjusted p value ≤0.05 and ** indicates an adjusted p value ≤0.01, *** indicates an adjusted p value ≤0.001, **** indicates an adjusted p value ≤0.0001. Error bars SEM.

Compared to mature trophoblast populations, side-population trophoblasts exhibit differential gene methylation

To understand how gene methylation affects human trophoblast differentiation and to examine the role that side-population trophoblasts may play in this, we further investigated the differentially methylated genes between side-population trophoblasts, and mature trophoblast populations (cytotrophoblasts and EVTs). First, we examined the differential methylation of genes that play key roles in establishing the trophectoderm lineage in the blastocyst, and/or in murine TSC biology, that are known to be influenced by DNA methylation (CDX2, EOMES, HAND1, TEAD4, ELF5, and PLET1) [6,8,32,33], and found that none of these genes were differentially methylated in side-population trophoblasts compared to cytotrophoblasts or EVTs. However, several genes that are highly expressed in murine TSC, (DDB1 and CUL4 associated factor 12 like 1 (DCAF12L1), and stabilin 1 (STAB1)) were hypomethylated in side-population trophoblasts compared to cytotrophoblasts and/or EVTs (Table 6) [34,35]. Interestingly, genes associated with maintaining pluripotency, proliferation and differentiation were differentially methylated in side-population trophoblasts compared to cytotrophoblasts or EVTs (Table 6). Taken together, this data suggests that side-population trophoblasts have a methylomic signature that distinguishes them from cytotrophoblasts.

Table 6. Table of differentially methylated genes related to EMT, and/or metastatic cancers were up-regulated in extravillous trophoblasts compared to cytotrophoblasts.

Gene	Methylation status at CpGi in promoter region in EVT compared to CTB	Averaged fold change in mRNA expression in EVT compared to CTB	Function/Biological Relevance	Reference
ABR	hypomethylated	2.8*	BCR related protein with functional activity that may regulate Rho family members involved in cytoskeletal reorganisation.	[65]
ADAM19	hypomethylated	9.1^‡	Involved in the EMT in glioblastoma.	[66]
AGRN	hypomethylated	4.9^‡	AGRN knockdowns reduce oral squamous cell carcinoma migration	[67]
AQPEP	hypomethylated	7.9^‡	An EVT specific protein involved in invasion and migration.	[68]
B3GNT7	hypomethylated	6.8^†	Is upregulated in metastatic breast cancer and enhances cell migration and invasion in vitro.	[69]
CAPN2	hypomethylated	2.7*	Promotes EMT, migration and invasion in metastatic cancer.	[70]
CARD9	hypomethylated	4.7*	CARD9 plays a role in manipulating the tumour microenvironment to enhance colon cancer metastasis.	[71]
CDR2	hypomethylated	3.1^†	Upregulated expression of CDR2 is associated with poor prognosis in renal cell carcinoma and may contribute to cancer metastasis.	[72]
DLX4	hypomethylated	2.5*	Involved in EMT, invasion and migration during embryogenesis and placentation.	[73]
FAM163A	hypomethylated	4.6^†	Associated with high-risk neuroblastoma and metastasis to bone marrow.	[74]
FERMT3	hypomethylated	3.1*	Encourages EMT,invasion, metastasis and angiogenesis in breast cancer.	[75]
FLT4	hypermetylated	5.8‡	Associated with metastatic epithelial malignancies and EMT.	[76]
HCK	hypomethylated	3.2*	Contributes to cell migration and actin remodelling.	[77]
HMHA1	hypermethylated	3.5‡	Regulates reorganisation of the cytoskeleton and enhancing cell motility	[78]
IL17RA	hypomethylated	2.9^†	Involved in gastric cancer migration and osetosarcoma metastasis.	[79]
IRF7	hypomethylated	2.4^†	Enables glioma cell invasion.	[80]
KCNF1	hypomethylated	2.0^†	Ion channel upregulated in invasive cancer.	[81]
KIF3C	hypomethylated	2.7^†	Associated with the EMT and breast cancer progression and metastasis.	[82]
KLF13	hypomethylated	2.1^†	Oral cancer cells overexpress KLF13 and contributes to malignancy in other tissues.	[83]
KLF6	hypomethylated	2.5^†	KLF6 is involved in the EMT and present in HER2-ERBB2 positive ductal breast cancer cells.	[84]
LHX2	hypomethylated	4.4^†	Involved in EMT during normal and cancerous breast epithelium development.	[85]
LIMS2	hypomethylated	3.4*	Involved in interactions which help cell morphology and migration.	[86]
LOXL1	hypomethylated	7.9^‡	Known to drive EMT in cancers.	[87]
LPCAT1	hypomethylated	4.3^†	Associated with increased breast cancer grade and metastasis.	[88]
LRFN4	hypomethylated	2.1^†	Expressed in both solid tumours and blood cancers. Involved in cytoskeletal reorganisation which allows cell motility.	[89]
MICALL2	hypomethylated	11.6^‡	Likely an important regulator of EMT in ovarian cancer. Silencing MICALL2 in ovarian cancers inhibits proliferative, invasive and migratory capacity of tumours.	[90]
MT1E	hypomethylated	4.9^‡	Involved in migration and invasion of bladder cancer and glioma.	[91]
NCOR2	hypomethylated	2.9^‡	Associated with distant metastasis of breast cancer cells	[92]
NFATC1	hypomethylated	2.8^†	Associated with EMT.	[93]
NRN1	hypomethylated	5.3^‡	Plays a role in the migration of bone marrow derived mesenchymal stem cells.	[94]
NUMB	hypermethylated	3.2†	NUMB isoforms are involved enhancing the EMT and invasion and migration of breast cancer cells.	[95]
PIK3CD	hypomethylated	3.5*	Downregulation of PIK3CD inhibits invasion and migration of colorectal cancer cells	[96]
PIM3	hypomethylated	2.0^†	Overexpression of PIM3 enhances the invasive and migratory capacity of ovarian and prostate cancer cells. Regulates genes that enhance the EMT.	[97]
PRDM16	hypomethylated	3.3*	Interacts with SKI to enhance gastric cancer growth.	[98]
RUSC2	hypomethylated	3.7^†	Important migration of non-small cell lung cancer.	[99]
SKI	hypomethylated	2.1^†	Overexpression in melanoma cells enhances migration in vitro.	[100]
SLC16A3	hypomethylated	13.7^‡	Associated with hepatocellular carcinoma invasion and migration. Contributes to EMT in breast cancer.	[101]
SOD2	hypomethylated	3.7*	Mitochondrial enzyme which promotes EMT.	[102]
TGFB1	hypomethylated	6.6^‡	In later stages of tumorigenesis in vivo TGFB1 promotes the EMT and cancer invasion and metastasis.	[103]
TPM2	hypomethylated	5.6^‡	Associated with EMT in lens epithelium	[104]
WNT1	hypomethylated	3.0*	Associated with highly metastatic advanced cancers. Cancer cell migration and invasion can be inhibited though blockade of the Wnt-1 pathway and inhibition of EMT.	[105]
ZYX	hypomethylated	2.8^†	Regulates lung cancer cell motility. Involved in reorganisation of cytoskeletal proteins in EMT.	[106]

Statistically significant values are labelled either *adjusted p value<0.05, † adjusted p value <0.01 or ‡ adjusted p value <0.001. EVT, extravillous trophoblasts; CTB, cytotrophoblast.

Differential methylation impacts regulatory miRNAs involved in pluripotency signalling pathways

16%–17% of methylation differences between side-population trophoblasts and cytotrophoblasts, side-population trophoblasts and EVTs, or cytotrophoblasts and EVTs, occurred at intragenic regions that encode gene regulatory molecules, including miRNAs (Figure 3). Differential methylation of eight miRNAs was evident between side-population trophoblasts and cytotrophoblasts, 92 miRNAs between side-population trophoblasts and EVTs, and 212 miRNAs between cytotrophoblasts and EVTs. Eighteen of 92 miRNAs that were differentially methylated between side-population trophoblasts and EVTs, and 48/212 miRNAs that were differentially methylated between cytotrophoblasts and EVTs, were significantly enriched (p < 0.03) for ‘signalling pathways regulating pluripotency of stem cells’ by DIANA-miRPath analysis and of these, 13 were common to both groups (Table 5). This suggests that these miRNAs could play roles in maintaining a stem/progenitor phenotype in side-population trophoblasts and cytotrophoblasts.

To determine whether methylation influenced miRNA expression, the expression of three miRNAs that regulate proliferation and differentiation in stem cell populations and trophoblasts (miR-let-7b, miR-637 and miR-193B) were quantified by real-time PCR in all three trophoblast populations from four placentae. RRBS showed that all three of these miRNAs were hypermethylated in side-population trophoblasts compared to EVTs. Therefore, we would expect to see a corresponding decrease in gene expression of these miRNAs in side-population trophoblasts compared to EVTs. In concordance with expression patterns predicted by RRBS, miR-let-7b expression was significantly lower (>2 fold, p value <0.01) in both side-population and cytotrophoblasts compared to EVT (Figure 5(a)). However, there were no significant differences in mir637 or mir193B expression (Figure 5(b, c)). This demonstrates that methylation alone does not necessarily determine the expression of these miRNAs.

Figure 5. Bar graphs showing the normalised miRNA expression in side-population trophoblasts (SP), cytotrophoblasts (CTB) or extravillous trophoblasts (EVT) for (a) miR-let-7b, (b) miR-637 and (c) miR4487. ** indicates an adjusted p value ≤0.01. Error bars are SEM.

Gene methylation is associated with the acquisition of an invasive phenotype during EVT differentiation

The differentiation of villous cytotrophoblasts to invasive EVTs in many ways mimics an epithelial-mesenchymal transition (EMT) in cell phenotype [36]. This EMT process is also observed in a number of invasive cancers where it is in part regulated by gene methylation. Therefore, in order to understand how methylation may regulate EVT differentiation we compared the transcriptome and methylome of first trimester cytotrophoblasts and EVTs.

When the transcripts of EVTs were compared to the transcripts of cytotrophoblasts 1744 genes were significantly upregulated ≥2 fold, and 2068 genes were significantly downregulated ≥2 fold (Supplementary Table 6). The GATHER gene ontology tool revealed that genes upregulated in EVTs were involved in pathways such as the immune response, defence response, cell motility, and response to external stimulus (Supplementary Table 7). Genes downregulated in EVTs were involved in pathways such as cell proliferation, cell cycle, regulation of protein kinase activity, and cell-cell adhesion (Supplementary Table 8). Markers characteristic of EVTs including; human leukocyte antigen–G (HLA-G) and HLA-E, and pro-migratory and invasive genes including integrin subunit alpha 1 (ITGA1), integrin subunit alpha 5 (ITGA5), and matrix metallopeptidase 2 (MMP2) were all upregulated in EVTs. Genes characteristic of cytotrophoblasts such as integrin subunit beta 4 (ITGB4), integrin subunit alpha 6 (ITGA6) and epidermal growth factor receptor (EGFR) were also all upregulated in cytotrophoblasts compared to EVTs [37,38].

Approximately one-third (2231/7061) of the CpG sites that were differentially methylated between cytotrophoblasts and EVTs were in promoter regions, impacting 633 unique genes. When the methylomes and transcriptomes were compared, 102/633 genes also exhibited transcriptomic changes. Interestingly 41/102 genes were associated with EMT and/or metastatic cancers (Table 6, Figure 6(a)). However, promoter methylation did not always correspond to the expected transcript levels. 97/102 genes were hypermethylated at promoter regions in cytotrophoblasts compared to EVT, yet only 61/97 of these had a corresponding decrease in gene expression (Figure 6(b)).

Figure 6. Bar graphs showing (a) the normalised mRNA fold change in extravillous trophoblasts (EVT) compared to cytotrophoblasts (CTB) of the top 20 significantly different genes associated with EMT or invasive cancer that were hypomethylated at CpG islands in promoters regions of EVT compared to CTB (b) the normalised mRNA fold change in CTB compared to EVT of the top 20 significantly different genes that were hypermethylated at CpG island in promoters regions of CTB compared to EVT.

Discussion

Our genome-wide DNA methylation data generated from freshly isolated human side-population trophoblasts, cytotrophoblasts, and EVTs suggests that trophoblasts have a multi-level system involving both DNA methylation and miRNAs that regulate their differentiation and function. These data support our hypothesis that side-population trophoblasts are a stem-like population in the placenta, and provide insights into the regulatory mechanisms involved in acquisition of a migratory and invasive phenotype during EVT differentiation.

Studies of murine TSC suggest that the DNA methylome is important for trophoblast development and differentiation. Consequently, our current study aimed to determine whether CpG methylation may regulate side-population trophoblast differentiation and function in a similar manner to murine TSC. Our study identified two genes (DCAF12L1 and STAB1) that were hypomethylated in our side-population trophoblasts compared to mature trophoblasts, and these two genes are known to contribute to murine TSC function [34,35]. Our previously published microarray data set (GEO accession number GSE57834) also showed a corresponding significant increase in STAB1 gene expression in both side-population trophoblasts and cytotrophoblasts compared to EVTs. However, no transcriptomic differences in DCAF12L1 between trophoblast populations were observed. In mice, Elf5 contributes to TSC maintenance, and although there is no difference in Elf5 methylation between different murine trophoblast populations, Elf5 gene expression is downregulated in TSC differentiated to mature trophoblast lineages [10,11]. Similarly, in this study there was no differential methylation of ELF5 in the different human trophoblast populations, although the human situation is slightly different as ELF5 is stably expressed in cytotrophoblasts [13,19,39]. We speculate that these differences may result from the fact that the murine placenta is anatomically quite distinct from the human placenta, and that murine and human embryos implant into the maternal tissue in opposite orientations with human embryos implanting via the embryonic pole, while murine embryos implant via the an anembryonic pole. Thus the trophoblasts of the two species are exposed to quite different environmental cues that may influence DNA methylation [38].

In line with their proposed in vivo lineage relationship where side-population trophoblasts differentiate into cytotrophoblasts, this study showed that side-population trophoblasts have a methylome more similar to that of cytotrophoblasts than EVTs [40]. Our differential methylation analyses indicate that side-population trophoblasts may be less differentiated than cytotrophoblasts and EVTs. For example, tumour protein, translationally-controlled 1 (TPT1) is hypermethylated in side-population trophoblasts compared to mature trophoblasts. Hypermethylation of TPT1 inhibits expression of POU class 5 homeobox 1 (OCT4) and Nanog homeobox (NANOG), which are downregulated during trophectoderm lineage segregation during blastocyst formation [41,42]. This suggests that from the perspective of DNA methylation patterns, side-population trophoblasts are more closely related to the trophectoderm (from which all trophoblasts lineages are derived) than mature trophoblast populations. In addition, differential methylation of genes associated with the maintenance of stem cell populations (e.g., DExH-box helicase 34 (DHX34), PR/SET domain 2 (PRDM2), and LDL receptor related protein 5 (LRP5)), was observed in side-population trophoblasts compared to mature trophoblast populations [43,44]. Together this suggests that DNA methylation may contribute to maintaining potency and guiding lineage commitment/differentiation in side-population trophoblasts.

Successful placentation involves the invasion of EVTs through the decidua [22,45]. Similar characteristics are found in cancer cells, and may result from similar molecular circuits involved in EMT and tumour development, which have been suggested to be regulated by epigenetic factors [22,45,46]. Previous data has shown that first trimester placental lysates and cytotrophoblasts are hypomethylated in gene blocks more closely related to cancer phenotypes (some of which were enriched for EMT) compared to third trimester samples [47]. Corresponding gene expression differences in EMT pathways in first versus third trimester placentae have also been reported, and together this data likely represents key gestational differences in placentation, as the majority of EVT differentiation and invasion occurs in the first half of pregnancy [22]. We believe our study is the first to investigate differences in methylation in individual uncultured trophoblast populations from the first trimester placentae, and to our knowledge is the first large scale analysis of matched cytotrophoblasts and EVTs from the same placenta that has connected promoter methylation with gene expression data to understand how methylation influences EVT differentiation. Overall, changes in gene methylation in promotor regions between cytotrophoblasts and EVTs were observed in 19 EMT-associated genes. Of these, 17 genes that were hypomethylated in promoter regions in our methylomic dataset exhibited corresponding changes in gene expression in our transcriptomic dataset, including ADAM metallopeptidase domain 19 (ADAM19), distal-less homeobox 4 (DLX4), LIM homeobox 2 (LHX2), and lysyl oxidase like 1 (LOXL1). This demonstrates that gene methylation likely regulates some (but not all) of the EMT changes in gene expression during EVT differentiation, and supports the emerging concept that EVT differentiation is an EMT-like process [36,48,49].

It has previously been demonstrated that in the placenta methylation differences do not always directly correspond to differences in gene expression [50–52]. Our data shows a similar lack of direct correlation between methylation and gene expression. The majority of placental gene expression regulation is instead likely attributable to changes associated with small RNA, non-coding RNA and histone modifications that in-directly regulate gene expression [52,53]. Indeed our study identified 225 miRNAs that were differentially methylated between trophoblast populations. However, as DNA methylation did not always correlate with miRNA expression, additional regulatory mechanisms must guide miRNA expression in human trophoblasts. Such additional mechanisms may lie in the 3D structure that the genome holds within cells, which means that small changes to the DNA (for example, through the addition or removal of methyl groups) can change the ultrastructure of the DNA in a manner that impacts lineage-specific transcriptional programs [9,54]. For example, DNA methylation mediates chromatin remodelling and contributes to murine trophoblast lineage segregation [9].

The methodology employed in this study only allowed us to investigated CpG methylation in detail. To more comprehensively understand how methylation influences trophoblast differentiation, future studies of human trophoblast populations should investigate non-CpG methylation (CpA, CpT, and CpC). Indeed, non-CpG methylation is a feature of pluripotent cell types and induced pluripotent stem cells [55]. However, as non-CpG methylation accounts for ≤1.7% of methylation in our samples (data not shown) the extent to which this would influence side-population trophoblast function/differentiation in the placenta is unclear.

In conclusion, the methylome of human first trimester side-population trophoblasts, cytotrophoblasts and EVT are distinct from one another. This RRBS analysis of the trophoblast methylome suggests that a multilevel mechanism regulates human trophoblast differentiation. This study provides a scaffold from which individual trophoblast populations can be studied in pregnancy disorders to better understand how DNA methylation maybe impacting the pathogenesis of diseases such as fetal growth restriction and preeclampsia.

Acknowledgments

Thanks are extended to the staff and patients at Auckland Medical Aid Centre for donating the placental tissue used in this work. The authors would like to thank Stephen Edgar for flow cytometry assistance and expertise.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.

Additional information

Funding

This work was supported by funding from Gravida: National Centre for Growth and Development (STF-13-09). TKJB Gamage is the recipient of an Auckland Medical Research Fund Doctoral Scholarship (1214004) and a University of Auckland Doctoral Scholarship. JL James is supported by a Health Research Council Sir Charles Hercus Research Fellowship (16/043). None of these funding bodies had any role in the design of this research or interpretation of the data.