Systematic Identification of Placental Epigenetic Signatures for the Noninvasive Prenatal Detection of Edwards Syndrome

Background Noninvasive prenatal diagnosis of fetal aneuploidy by maternal plasma analysis is challenging owing to the low fractional and absolute concentrations of fetal DNA in maternal plasma. Previously, we demonstrated for the first time that fetal DNA in maternal plasma could be specifically targeted by epigenetic (DNA methylation) signatures in the placenta. By comparing one such methylated fetal epigenetic marker located on chromosome 21 with another fetal genetic marker located on a reference chromosome in maternal plasma, we could infer the relative dosage of fetal chromosome 21 and noninvasively detect fetal trisomy 21. Here we apply this epigenetic-genetic (EGG) chromosome dosage approach to detect Edwards syndrome (trisomy 18) in the fetus noninvasively. Principal Findings We have systematically identified methylated fetal epigenetic markers on chromosome 18 by methylated DNA immunoprecipitation (MeDIP) and tiling array analysis with confirmation using quantitative DNA methylation assays. Methylated DNA sequences from an intergenic region between the VAPA and APCDD1 genes (the VAPA-APCDD1 DNA) were detected in pre-delivery, but not post-delivery, maternal plasma samples. The concentrations correlated positively with those of an established fetal genetic marker, ZFY, in pre-delivery maternal plasma. The ratios of methylated VAPA-APCDD1(chr18) to ZFY(chrY) were higher in maternal plasma samples of 9 male trisomy 18 fetuses than those of 27 male euploid fetuses (Mann-Whitney test, P = 0.029). We defined the cutoff value for detecting trisomy 18 fetuses as mean+1.96 SD of the EGG ratios of the euploid cases. Eight of 9 trisomy 18 and 1 of 27 euploid cases showed EGG ratios higher than the cutoff value, giving a sensitivity of 88.9% and a specificity of 96.3%. Conclusions Our data have shown that the methylated VAPA-APCDD1 DNA in maternal plasma is predominantly derived from the fetus. We have demonstrated that this novel fetal epigenetic marker in maternal plasma is useful for the noninvasive detection of fetal trisomy 18.


Introduction
Fetal chromosomal aneuploidies are the main reasons for pregnant women to seek prenatal diagnosis [1]. Definitive diagnosis of fetal aneuploidy often requires obtaining fetal genetic materials by invasive procedures, which carry a risk of procedureassociated fetal loss [2]. The presence of fetal DNA in maternal plasma has opened up new opportunities for noninvasive prenatal diagnosis [3]. However, the noninvasive detection of fetal aneuploidy in maternal plasma is complicated by the low fractional concentration (3-10%) of fetal DNA, which co-exists with a large background of maternal DNA [4,5].
To detect fetal Down syndrome noninvasively, we and other researchers have applied massively parallel genomic sequencing to determine the proportional amounts of chromosome 21 DNA molecules in maternal plasma [6][7][8]. This approach entails the analysis of millions of DNA molecules, derived from both the mother and the fetus, in maternal plasma. As an alternative to the still technically and bioinformatically complex massively parallel sequencing-based approach, we have developed other approaches by targeting fetal-specific DNA or RNA molecules in maternal plasma [9][10][11][12].
Previous findings suggest that the placenta is the major source of cell-free fetal nucleic acids in maternal plasma [13][14][15], whereas maternal blood cells are the major source of maternal nucleic acids in maternal plasma [16]. Thus, we have used epigenetic signatures specific to the placenta but not maternal blood cells to target fetal DNA in maternal plasma [14,[17][18][19]. We have developed a placental epigenetic signature, namely the unmethylated promoter of the serpin peptidase inhibitor, clade B (ovalbumin), member 5 (SERPINB5, NM_002639) gene, into a fetal epigenetic (DNA methylation) marker. Since this fetal epigenetic marker is located on chromosome 18, it is feasible to detect fetal trisomy 18 by assessing the ratio of the alleles in fetuses who are heterozygous for this marker [20]. However, this approach is only applicable to fetuses polymorphic for the marker. Thus, we developed an alternative approach, the epigenetic-genetic (EGG) chromosome dosage approach, with potentially wider population coverage.
We have recently applied the EGG approach for the noninvasive detection of fetal trisomy 21 [11]. This EGG analysis involves a fetal epigenetic marker, holocarboxylase synthetase (HLCS, NM_002639) gene, which is found to be hypermethylated in the placenta in comparison to maternal blood cells. Comparing this fetal epigenetic marker, which is located on the chromosome 21, with a fetal genetic marker on a reference chromosome unaffected by trisomy 21, we could infer the relative dosage of fetal chromosome 21 by analyzing maternal plasma DNA [11]. Since any paternally-inherited fetal sequences that are not found in the pregnant woman (e.g. Y-chromosomal sequences for male fetuses or polymorphic sequences inherited only from the father for both male and female fetuses) and are located on a reference chromosome can serve as the fetal genetic marker for the EGG analysis, potentially all fetuses in the general population may be covered.
Yet, if we use the EGG approach to detect the small increase (,1.5-fold) in the dosage of fetal chromosome 21 relative to any other unaffected fetal chromosome in maternal plasma samples collected from pregnancies involving a Down syndrome fetus, an analytical platform of high precision is required. The bisulfitebased detection method of fetal epigenetic markers can only offer limited precision, because bisulfite has been reported to degrade .90% of the input DNA [21]. A bisulfite-independent method using methylation-sensitive restriction enzyme that digests only unmethylated, but not methylated DNA has been developed to detect methylated fetal epigenetic marker [17]. Adopting this method to detect the methylated HLCS DNA molecules in maternal plasma, we have achieved an analytical precision that was high enough to distinguish between trisomy 21 and euploid fetuses noninvasively [11]. Therefore, the fetal epigenetic marker suitable for the EGG approach has to be resistant to digestion by methylation-sensitive restriction enzymes in maternal plasma. In other words, we need a fetal epigenetic marker that is hypermethylated in the placenta, the predominant source of fetal DNA in maternal plasma [14], and is located on the potentially aneuploid chromosome. Most of the earlier studies searched for fetal epigenetic markers only within selected genomic loci on chromosome 21 [11,22,23]. Using methylated DNA immunoprecipitation (MeDIP) and tiling array analysis [24], other investigators have expanded the search to chromosome 18 and reported numerous loci as potential fetal epigenetic markers [25]. However, no studies have validated whether these MeDIP-identified markers are detectable in maternal plasma, let alone their fetal-specificity in maternal plasma and clinical application.
In this study, we aimed to identify fetal methylated markers on chromosome 18 by MeDIP in a systematic way, and to confirm its detectability and fetal-specificity in maternal plasma. Furthermore, we applied this marker to develop an EGG test for the noninvasive detection of fetal trisomy 18.

Objectives
To systematically test if there are any genomic loci on chromosome 18 that are hypermethylated in the placenta, compared with maternal blood cells, we profiled the DNA methylation levels in the two tissue types by MeDIP and tiling array (MeDIP-chip). To test if the methylated DNA of such a locus is fetal specific in maternal plasma, we measured its plasma concentration before and after delivery of the fetus. To test if this methylated fetal epigenetic marker in maternal plasma can be used to detect fetuses with trisomy 18 using the EGG approach, we measured its concentrations relative to a fetal genetic marker by digital PCR.

Ethics approval
This study was conducted according to the principles expressed in the Declaration of Helsinki. Ethics approval from the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee and the respective institutional review broads was obtained. All patients provided written informed consent for the collection of samples and subsequent analysis.

Subjects and sample collection
Women with singleton pregnancies attending the respective hospitals in Hong Kong and the UK were recruited (File S1). Placental tissues were collected from first-and third-trimester pregnant women undergoing pregnancy termination and elective cesarean section, respectively. Maternal peripheral blood samples were collected just before and 24 hours after the obstetrics procedures. Additionally, blood samples were collected from firstand second-trimester pregnant women attending antenatal care. The trisomy 18 status of the fetus was confirmed by full karyotyping of chorionic villus samples. Blood samples were also collected from non-pregnant individuals as negative controls for validating the fetal-specificity of the candidate marker in plasma.

Sample processing
Maternal plasma was harvested from EDTA-blood by our previously established double-centrifugation protocol [26]. The maternal blood cell sample portion was recentrifuged at 2,5006g for removal of any residual plasma. Placenta was rinsed thoroughly in phosphate buffered saline to remove blood. DNA was extracted from plasma, blood cells and the placenta with the methods described in File S1.

MeDIP and tiling array (MeDIP-chip) analysis
The DNA sample was sonicated and subjected to MeDIP by antibody specific for methylated cytosine [24]. The immunoprecipitated product was amplified, labeled and hybridized on the GeneChip Human Tiling 2.0R Arrays (Affymetrix). Genomic loci with significantly higher DNA methylation in the placenta, relative to maternal blood cells, were identified by the Tiling Array Software (TAS) version 1.1 and the Model-based Analysis of Tiling array (MAT) method with the parameters described in File S1 [27,28].

Rapid and quantitative DNA methylation analysis by the Epityper
MeDIP-identified locus was analyzed by a quantitative DNA methylation assay, the Epityper (Sequenom) [29](File S1). Briefly, the genomic locus of interest in a bisulfite-converted DNA sample was subjected to PCR amplification by primers listed in Table S1, in vitro transcription into RNA, and uracil-specific cleavage on the complementary strand. The product derived from PCR amplicons would therefore be fragmented. Fragments derived from the methylated and unmethylated DNA molecules would have different masses due to the difference in nucleotide sequence at the CpG site caused by bisulfite conversion. The mass differences were readily resolved and quantified as distinct peaks by matrixassisted laser desorption and ionization time-of-flight (MALDI-TOF) mass spectrometry. In the Epityper, the DNA methylation level of one or more CpG sites in any one cleaved fragment was reported as an integral unit, namely a CpG unit. For each CpG unit, a methylation index (MI) was calculated as the ratio of the methylated peak height to the sum of the methylated and unmethylated peak heights.
Priority for Epityper analysis was given for MeDIP-identified loci that would (i) allow appropriate PCR design, including a high annealing temperature of the primers; (ii) result in Epityper assays with the maximum number of detectable fragments (CpG units), the masses of which were within the detection range of the mass spectrometer; and (iii) allow specific PCR amplification from bisulfite-converted DNA. (i)-(ii) were mainly calculated by the Epidesigner 2.0 program (Sequenom); and (iii) was checked by the BiSearch search tool [30]. Occurrence of genomic variations in the analyzed regions were checked using the Database of Genomic Variants (http://projects.tcag.ca/variation/) [31]. We made sure that there was no reported variant of higher than 1% frequency in the regions we analyzed.
Single-base single-molecule DNA methylation analysis by bisulfite genomic sequencing To quantify DNA methylation at the resolution of single CpG site in a single molecule, bisulfite genomic sequencing was performed [32]. Briefly, a genomic locus in the bisulfite-converted DNA was amplified by PCR primers (Table S1). The PCR products were cloned and segregated as colonies, each representing a single molecule for sequencing (File S1). For each CpG site in a DNA sample, a MI was calculated as the ratio of the number of methylated clones to the total number of clones sequenced.
Conventional quantitative PCR assays for VAPA-APCDD1, ZFY and b-actin DNA DNA samples, which had been subjected to digestion by methylation-sensitive restriction enzyme (File S1) or mock digestion by 50% glycerol instead of an enzyme, were then analyzed by 3 quantitative polymerase chain reaction (qPCR) assays. One qPCR assay was designed to target an intergenic region between the VAMP (vesicle-associated membrane protein)associated protein A, 33 kDa (VAPA, NM_003574) gene and the adenomatosis polyposis coli down-regulated 1 (APCDD1, NM_153000) gene. Another assay, targeting the zinc finger protein, Y-linked (ZFY, NM_003411), an established genetic marker for detecting fetal DNA in maternal plasma of pregnancies bearing male fetuses, was adopted from our previous study [5]. The third assay was designed to target a completely unmethylated region (positive control for methylation-sensitive restriction enzyme digestion) of the b-actin (ACTB, NM_001101) gene.
All 3 assays involved hydrolysis probes for detection. The primer and probe sequences and reaction conditions are listed in Tables S2 and S3. Concentrations of each target were quantified by a standard calibration curve constructed with known concentrations of a male genomic DNA sample. Any signals detected below the limit of detection (3 copies/PCR for all 3 qPCR assays), as determined by the method in File S1, were considered undetectable. Eight no template controls (water only) were included in every PCR run.

Digital PCR assays for EGG dosage analysis
Since high analytical precision is required for chromosome dosage analysis, this part of the study was performed by digital PCR [33]. We performed digital PCR through the dilution of the DNA sample to an average concentration of one template molecule in every two reaction wells [34]. This diluted DNA sample was then distributed to hundreds of reaction wells on a 384-well plate for PCR amplification of VAPA-APCDD1, ZFY and b-actin DNA. Primer and probe sequences and reaction conditions are listed in Tables S2 and S3. The actual number of template molecules was calculated by the direct counting of the number of positive wells followed by correction for the Poisson distribution (File S1).

Statistical analysis
Statistical analyses were performed with the Sigma Stat v3.5 (Systat).

Systematic identification of chromosome 18 loci hypermethylated in the placenta by MeDIP and the Epityper
We have embarked on studying DNA methylation levels of the entire chromosome 18, which contains about 351,500 CpG sites in 74.7 million bases of non-repetitive DNA sequence [35]. Ten DNA samples from 5 first-trimester euploid placentas and 5 firsttrimester maternal blood cell samples were subjected to MeDIP and tiling array analysis. The array interrogated DNA methylation level every 35 bases (average inter-probe distance) on essentially all the non-repetitive DNA sequences of the entire chromosome.
Genomic loci with a signal log ratio between the placenta and maternal blood cells of .0.4 in the TAS analysis or with a P,10 25 in the MAT analysis were considered as hypermethylated in the placenta. We identified 3,043 CpG sites located in 178 loci (68 and 110 loci from the TAS and MAT algorithms, respectively) with higher DNA methylation levels in the placenta, relative to maternal blood cells. We found that 140 loci (79%) were located within genes and the remaining 38 loci (21%) were located in the intergenic regions. The results from the TAS and MAT algorithms are listed in Table S4 and Table S5, respectively. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus [36], and are accessible through GEO Series accession number GSE22837.
Selecting for CpG sites with high DNA methylation levels in the placenta. Quantitative DNA methylation levels at the resolution of CpG sites are required for developing fetal epigenetic markers in maternal plasma. For this purpose, we analyzed the MeDIP-identified loci by quantitative DNA methylation assays, the Epityper, which were less labor-intensive and time-consuming than bisulfite sequencing, because it did not involve cloning and sequencing [29,37]. Priority was given for 370 CpG units (located in 26 loci) which were most efficiently analyzed by the Epityper platform ( Figure 1A). For each CpG unit, we compared the DNA methylation levels of two first-trimester placentas, with two firsttrimester maternal blood cell samples ( Figure 1B and Table S6).
To select CpG units potentially useful for developing fetal epigenetic markers in maternal plasma, we identified CpG units with a methylation index (MI) #0.20 in the maternal blood cells (criterion #1), and with a difference in MI of $0.50 between the placenta and maternal blood cells (criterion #2). These criteria were adopted with slight modification from our previous study in developing fetal markers in maternal plasma [22]. Among the 370 CpG units analyzed by the Epityper, 40 CpG units, located in 8 MeDIP-identified loci, fulfilled both criteria #1 and #2 (Table 1; Figure 1B, highlighted as green bars). Three or more potentially useful CpG units were found in each of 5 MeDIP-identified loci ( Table 1, top 5 loci). Thus, we further confirmed these observations in more cases. The majority of the analyzed CpG units in these 5 loci were confirmed to possess significantly higher DNA methylation levels in 5 first-trimester placentas, relative to 5 first-trimester maternal blood cell samples ( Figure 2; Table S7, Mann-Whitney test, p,0.05).
Selecting for CpG sites with low inter-individual variation in placental DNA methylation levels. We then assessed the inter-individual variations of the MI of the 5 selected loci in 10 first-trimester placental tissue samples. These 5 loci were interrogated by 6 Epityper assays ( Figure 3). For each CpG unit, a coefficient of variation (CV = standard deviation (SD)/mean) of the MI in 10 placentas was calculated ( Figure 3 and Table S8). For each Epityper assay, a median CV of all the CpG units within the assay was also calculated. The 4 Epityper assays with the lowest median CV were selected for further investigation (Figure 3, first 4 assays).
These 4 Epityper assays were performed on 6 euploid and 6 trisomy 18 placental tissue samples. No significant difference between the two groups were found in the methylation of any CpG unit, suggesting that placental DNA methylation was not altered by trisomy 18 (Mann-Whitney test, P.0.05; Table S9).
Selecting for CpG sites overlapping with methylationsensitive restriction enzyme sites. To select specific CpG sites that would be most promising as a noninvasive fetal DNA marker, we performed bisulfite sequencing on the 4 selected regions. Consistent with the MeDIP and the Epityper data, most of the sequenced CpG sites were predominately methylated in the placenta and almost completely unmethylated in maternal blood cells ( Figures 4A and S1). We examined these CpG sites for any overlapping sites recognized by two commonly used methylationsensitive restriction enzymes, HinP1I and HpaII. There were 5, 3, 1 and 1 such sites within 100 bp from the following regions: MAT. 18 This MeDIP-identified region, MAT.18.0094.1, is an intergenic region located 73 kb downstream of the VAPA gene and 421 kb upstream of the APCDD1 gene ( Figure 4A). Hence, we refer to this locus as VAPA-APCDD1.

Detection and characterization of digestion-resistant VAPA-APCDD1 DNA in maternal plasma
We designed a qPCR assay to target region 1 of VAPA-APCDD1. Since the PCR primers flanked 5 methylation-sensitive HinP1I and HpaII sites, only VAPA-APCDD1 molecules methylated in all 5 CpG sites would result in amplifiable qPCR signal ( Figure 4B). Thus, the concentration of digestion-resistant VAPA-APCDD1 DNA detected by this qPCR assay is reflective of the level of the methylated DNA of this potential marker.

Correlation between digestion-resistant VAPA-APCDD1
DNA and an established fetal genetic marker in maternal plasma. To determine if there was any correlation between the concentrations of digestion-resistant VAPA-APCDD1 DNA and those of the ZFY DNA, an established genetic marker for detecting male fetal DNA in maternal plasma, we analyzed the maternal plasma samples from 13 pregnant women bearing male fetuses in the previous experiment. Digestion-resistant VAPA-APCDD1 and ZFY concentrations were positively correlated (r = 0.91; P,0.00001; Spearman correlation) ( Figure 5B). Both assays were optimized to run under identical PCR thermal profiles (Table S3) and showed similar efficiencies as reflected by the slopes (23.77 and 23.53 for the VAPA-APCDD1 and the ZFY assays, respectively) and y-intercepts (39.7 Cq and 38.9 Cq for the VAPA-APCDD1 and the ZFY assays, respectively) of the two calibration curves.
Postpartum clearance of digestion-resistant VAPA-APCDD1 DNA in maternal plasma. To further investigate if digestion-resistant VAPA-APCDD1 DNA in maternal plasma would be cleared upon delivery of the fetus, we collected pre-delivery and 24-hour post-delivery maternal plasma samples from 10 other pregnant women. In all 10 cases, the digestion-resistant VAPA-APCDD1 DNA was rapidly cleared from maternal plasma to almost undetectable levels ( Figure 5C), demonstrating that its existence in maternal plasma was fetal-specific. The plasma concentrations of digestion-resistant VAPA-APCDD1 DNA before and after delivery of the fetuses were statistically significantly   Table S7. doi:10.1371/journal.pone.0015069.g002 different (Wilcoxon signed-rank test, P = 0.002). As a control for successful DNA extraction, we quantified the total DNA amount by the VAPA-APCDD1 DNA qPCR assay, and detected DNA in all pre-delivery and post-delivery maternal plasma samples before enzyme digestion. As a control for complete enzyme digestion, a qPCR assay was designed to target a region on the b-actin gene, which was known to be completely unmethylated (MI = 0.00) in both the placenta and maternal blood cells ( Figures 4B and S2), and no signals were detected in any of the HinP1Iand HpaIIdigested maternal plasma samples. The lower limits of detection of the b-actin and VAPA-APCDD1 qPCR assays were both 3 copies per reaction, as determined by the methods in File S1.

Epigenetic-genetic (EGG) dosage analysis of fetal chromosome 18 in maternal plasma by digestionresistant VAPA-APCDD1 DNA
Digital PCR assays for the VAPA-APCDD1, ZFY and bactin DNA. To detect the relatively small (,1.5-fold) increase in the dosage of chromosome 18 in trisomy 18, the digital PCR platform, which featured a higher precision than conventional qPCR, was used. These digital PCR assays were first validated in HinP1I and HpaII-digested plasma samples collected from 5 pregnant women before and after delivery, and 4 non-pregnant females. By these digital PCR assays, digestion-resistant VAPA-APCDD1 DNA and ZFY DNA were detected in pre-delivery maternal plasma samples, but were almost undetectable in postdelivery maternal plasma and non-pregnant female plasma samples (Table S10). b-actin DNA was not detected in all 3 groups of digested samples (Table S11).
EGG dosage analysis of chromosome 18 in the placenta. Since the placenta is the main source of fetal DNA in maternal plasma, we tested EGG analysis in the placenta before attempting it in maternal plasma. Digital PCR assays for VAPA-APCDD1 and ZFY DNA were performed on HinP1Iand HpaIIdigested DNA samples extracted from placental tissues of five trisomy 18 and five euploid male fetuses. The ratio of digestionresistant VAPA-APCDD1 to ZFY was calculated for each sample. These ratios were significantly higher in the trisomy 18 placentas than the euploid placentas (Mann-Whitney test, P = 0.029; Figure  S3). A reference interval, defined as the mean ratio digestionresistant VAPA-APCDD1 to ZFY 61.96 SD, was calculated from the euploid placentas as 1.20-1.66. The ratios in all of the trisomy 18 placenta were above the upper reference limit ( Figure S3).

EGG dosage analysis of fetal chromosome 18 in maternal
plasma. Since the digestion-resistant VAPA-APCDD1 DNA and ZFY DNA are fetal-specific in maternal plasma, we reasoned that EGG dosage analysis of the fetus or the placenta could also be performed in maternal plasma. Maternal plasma samples were collected from 27 women with euploid male fetuses, and 9 women with trisomy 18 male fetuses. The median gestational ages at sample collection were 14.1 weeks (IQR, 12.9-16.6) and 13.3 weeks (IQR 12.9-14.5) among the euploid and trisomy 18 fetuses, respectively. The HinP1Iand HpaII-digested plasma samples were subjected to digital PCR assays for VAPA-APCDD1 and ZFY DNA ( Figure S4). To facilitate a fair comparison, we diluted each plasma sample to a comparable average template concentration (m) of ZFY (reference) molecules per reaction well. The median m values were 0.08 and 0.05 per reaction well for the euploid and trisomy samples, respectively. The ratio of digestion-resistant VAPA-APCDD1 to ZFY was calculated for each sample. The ratios of digestion-resistant VAPA-APCDD1 to ZFY were significantly higher in maternal plasma samples of trisomy 18 fetuses than those of euploid fetuses (Mann-Whitney test, P,0.001) ( Figure 6). A reference interval of 0.34-3.04 was calculated for the 27 maternal plasma samples from euploid pregnancies. The ratio of one euploid sample fell outside the reference interval. The ratios of 8 out of 9 trisomy 18 samples were above the upper reference limit. b-actin DNA was essentially undetectable by the digital PCR assay in all of the 36 digested maternal plasma DNA samples, implying that complete digestion had occurred. Ten cases were selected for comparing the detectable concentrations of b-actin DNA before and after digestion. b-actin DNA was detectable in all the maternal plasma samples before, but not after, digestion ( Figure S5).

Discussion
Using MeDIP-chip analysis, the Epityper and bisulfite sequencing, we have systematically identified methylated fetal epigenetic markers on chromosome 18. We have also demonstrated that one such marker, namely the digestion-resistant VAPA-APCDD1 DNA, was readily detectable in maternal plasma during pregnancy, but rapidly cleared to almost undetectable levels upon delivery of the fetus. Further characterization has shown that the concentrations of this novel marker in maternal plasma were positively correlated with those of an established fetal genetic marker.
Since these data have suggested that this digestion-resistant VAPA-APCDD1 DNA was predominantly derived from the fetus in maternal plasma, we further applied it for the EGG dosage analysis of fetal chromosome 18 in maternal plasma. We have determined the ratio of digestion-resistant VAPA-APCDD1 to ZFY in maternal plasma samples involving 9 trisomy 18 male fetuses, and 27 euploid male fetuses. A reference interval of this ratio calculated from the euploid samples was calculated. We then observed the ratios of 8 trisomy 18 fetuses, and only 1 euploid fetus was higher than the upper reference limit. Hence using the upper reference limit as a threshold, all except one trisomy 18 fetuses was detected (i.e. 1 false negative), and only one euploid was also detected (i.e. 1 false positive). The sensitivity and specificity were 88.9% and 96.3%, respectively, for this EGG analysis in maternal plasma.
Of note, in this EGG analysis, we have quantified the reference chromosome by a fetal genetic marker, i.e. the ZFY DNA, instead of a fetal epigenetic marker, e.g. the hypermethylated RASSF1A [17]. This is because we have previously shown that the dosage analysis had greater power in distinguishing the trisomic from the   cases, if the reference chromosome was quantified by a fetal genetic marker instead of a fetal epigenetic marker [11]. For fetal epigenetic markers, there are small degrees of heterogeneity in the DNA methylation levels between samples. If we determined the relative ratio between an epigenetic marker on the aneuploid chromosome and another epigenetic marker on a reference chromosome, the combined variance caused by the heterogeneous DNA methylation levels in both epigenetic markers would be large. The resultant inter-individual variation in the chromosome dosage ratio would be too large to discriminate trisomy from euploid cases. Thus, it is preferable to quantify the reference chromosome by a fetal genetic marker.
Of a similar concern, to minimize the inter-individual variability in the DNA methylation level of the fetal epigenetic marker used for quantifying the potentially aneuploid fetal chromosome 18, we have studied how the DNA methylation levels of each CpG unit varied across different individuals by the Epityper. Among the six potential markers, the median of the inter-individual CVs was observed to vary from 8% to 38% (Fig. 3). The data suggested that the development of fetal epigenetic markers by MeDIP-based study should be followed up by quantitative DNA methylation studies at single CpG resolution. Otherwise, a low degree of interindividual variation, which is important for developing quantitative applications such as the assays for the EGG analysis, could not be guaranteed by the MeDIP-based study alone.
Additionally, our data have suggested that not every CpG site in the MeDIP-identified regions could fulfill the stringent criteria for developing fetal epigenetic marker in maternal plasma. Some CpG sites might possess undesirably high levels of DNA methylation in the maternal blood cells ( Figure 1B, not fulfilling criterion #1), thus preventing the removal of the maternal DNA sequence in maternal plasma. On the other hand, some CpG sites might not have high enough DNA methylation levels in the placenta relative to maternal blood cells, thus preventing the specific detection of the fetal DNA sequence in maternal plasma ( Figure 1B, not fulfilling criterion #2). Thus, the quantitative DNA methylation data provided by the Epityper for each CpG unit within the MeDIP-identified regions were indispensable for developing fetal epigenetic markers (Figure 3). The Epityper data have facilitated us to select for a suitable epigenetic signature, with marked difference between the placenta and maternal blood cells and with the least inter-individual variation, to target the fetal chromosome 18 in maternal plasma for the EGG analysis.
In this study, we performed the EGG analysis only in pregnancies involving male fetuses because a precise digital PCR assay targeting the Y-chromosome, namely the ZFY DNA assay, is well established in our laboratory [5]. However, we envision that this type of EGG analysis can be adopted for pregnancies involving female fetuses, because any fetal DNA sequence, including an autosomal sequence, that is inherited only from the father, and is located on a chromosome unaffected by the concerned trisomy, can serve as a fetal genetic marker for quantifying the reference chromosome. Through the latter approach, EGG analysis can be performed for both female and male fetuses.
In this study, we demonstrated the feasibility of applying the EGG approach for the noninvasive prenatal detection of fetal trisomy 18. The diagnostic accuracy of the test requires further evaluation in a larger cohort. Nevertheless, we are the first to use a bisulfite-independent approach to detect a methylated epigenetic marker for fetal chromosome 18 in first-trimester maternal plasma, and to use this approach to achieve the noninvasive detection of Wallis ANOVA test showed that there are significant differences among the three trimesters (P,0.001). The P-values of the pairwise comparisons with significant differences are shown (Dunn's test). (B) Concentrations of the epigenetic (Y-axis) and genetic (X-axis) markers in 13 maternal plasma samples (Spearman rank order correlation test, P,0.00001, R = 0.91). (C) The concentrations of this marker in 10 pairs of pre-delivery and post-delivery maternal plasma (Wilcoxon Signed-rank test, P = 0.002). Corresponding plasma samples from the same pregnant women are connected by a line. doi:10.1371/journal.pone.0015069.g005 fetal trisomy 18 in early gestation. Compared with tests based on massively parallel genomic sequencing, the approach described here might represent a lower 'barrier for entry' by many laboratories interested in noninvasive prenatal diagnosis, as the equipment and bioinformatics support requirements are much lower [38]. Figure S1 DNA methylation levels by bisulfite sequencing in two promising loci for developing fetal epigenetic markers. Data on each locus are shown in two panels, each of which contains 3 sub-panels. Top sub-panel. The genomic location of the promising locus in relation to the associated gene. Middle sub-panel. The locus, which was identified by MeDIP as possessing higher methylation in the placenta relative to maternal blood cells, was analyzed by bisulfite sequencing. See Figure 2 for the legend on the bar graph for MeDIP. Bottom sub-panel. Single-base DNA methylation levels by bisulfite sequencing. For each sample, 8 randomly-picked clones (rows) were scored for each CpG site (column). Filled circles, methylated CpG sites. Empty circles, unmethylated CpG sites. Upward arrows, cutting sites of the methylation-sensitive restriction enzymes HpaII and HinP1I.