Genome-wide placental DNA methylations in fetal overgrowth and associations with leptin, adiponectin and fetal growth factors

Characteristics of study subjects

Table 1 presents maternal and neonatal characteristics of the 30 pairs of study subjects. Comparing LGA versus birth weight optimal-for-gestational-age (OGA) control subjects, there were no significant differences in maternal age, ethnicity, parity, education, smoking, gestational hypertension, family history of hypertension, family history of diabetes, maternal blood glycated hemoglobin (HbA1c) levels during the 2nd and 3rd trimesters of pregnancy, and gestational age at delivery. Women bearing a LGA fetus had higher pre-pregnancy body mass index (BMI mean: 24.0 vs. 22.0 kg/m2) and were more likely to have a cesarean section delivery (73.3% vs. 26.7%). As expected, average birth weight and birth length were substantially higher in LGA vs. OGA newborns.

Table 1 Characteristics of study subjects in a matched (1:1) study of 30 pairs of term placentas in LGA and OGA newborns in the Shanghai birth cohort

LGA newborns had significantly higher cord blood IGF-I concentrations (mean: 88.8 vs.68.6 ng/mL, P = 0.006), and lower HMW adiponectin concentrations (14.6 vs. 20.5 μg/mL, P = 0.014) (Table 1). There were no significant differences in cord blood insulin, C-peptide, proinsulin, leptin, IGF-II and total adiponectin concentrations.

Differentially methylated positions (DMPs)

Adjusting for maternal age, pre-pregnancy BMI, whole blood HbA1c levels at the second and third trimesters of pregnancy and the identified four principal components (from principal component analysis) representing placental cell types (other co-variables were excluded since they were similar and did not affect the comparisons), a total of 543 CpG sites were differentially methylated positions (DMPs) comparing LGA and OGA groups accounting for multiple tests with false discovery rate (FDR) < 5% and absolute methylation difference (delta beta) > 0.05, including 397 hypermethylated and 146 hypomethylated DMPs (Fig. 1, Additional file 2: Table S1). These loci were distributed over 316 genes (232 hypermethylated genes, 84 hypomethylated genes).

Fig. 1figure 1

Volcano plot of differentially methylated positions (DMPs) in placental gene DNAs comparing large-for-gestational-age (LGA, birth weight > 90th percentile) versus optimal-for-gestational-age (OGA, 25–75th percentiles, control) newborns. DMPs in the upper left and right quadrants (colored) are differentially methylated at false discovery rate (FDR) < 5%

The top 50 DMPs (25 hypermethylated, 25 hypomethylated CpG sites) are presented in Tables 2 and 3. The hypermethylated loci were annotated to 14 genes (EN2, LOC283999, CADM2, ADGB, KRTAP13-4, CRMP1, GFRA1, NRXN1, VSX1, PPFIA2, PLXNC1, DNAJB5, DAOA and ZPLD1). The hypomethylated sites were annotated to 13 genes (FAM155A, C21orf34, WNT5B, DTNA, OPRM1, SORCS3, KIF26B, SLCO3A1, KIF26B, LOC284930, SLIT3, NXPH1 and HLA-L).

Table 2 Top 25 hypermethylated sites in placental DNAs in LGA versus OGA newbornsTable 3 Top 25 hypomethylated sites in placental DNAs in LGA vs. OGA newborns

Our placental epigenome data validated that the cadherin 13(CDH13) gene was hypermethylated in LGA as reported in a previous epigenome-wide association study [10] (our study: CDH13 methylation increased by 0.05 in LGA; the previous study: CDH13 methylation increased by 0.21 in LGA, according to the publication and communications with the corresponding author).

Pyrosequencing validation study

We sought to validate a few differentially methylated CpG sites with relatively large methylation differences between LGA and OGA groups in the epigenome-wide association analysis. The study subjects were an independent random sample of 47 pairs of LGA and OGA newborns matched by sex and gestational age from the Shanghai birth cohort (Additional file 3: Table S2). Women bearing a LGA fetus had higher pre-pregnancy BMI and were more likely to be have a cesarean section delivery, while other maternal characteristics were similar in LGA and OGA groups. As expected, average birth weight and birth length were substantially higher in LGA versus OGA newborns.

Three CpG sites were selected among the top 25 DMPs in the pyrosequencing validation study. The CpG site (cg11700298) annotated to cell adhesion molecule 2 (CADM2)] was selected because polymorphism in this gene has been associated with obesity and type 2 diabetes [11, 12]. We randomly selected 2 more CpG sites among the top 25 DMPs in the validation study. They were cg17650274 [annotated to gene visual system homeobox 1 (VSX1)] and cg17512353 [annotated to gene major histocompatibility complex, class I, L (HLA-L)]. The pyrosequencing validation vs. epigenome-wide study results for the 3 DMPs are shown in Additional file 3: Table S3; only one DMP (VSX1 gene) was validated in the pyrosequencing study.

GO and KEGG pathways

The GO analysis showed that DMPs were mostly enriched in axon development (biological processes), motile cilium (cellular components), metal-ion transmembrane transporter activity (molecular functions) (Additional file 2: Table S4). KEGG pathway analysis showed that DMPs might mainly be involved in 9 pathways including GnRH secretion, protein digestion and absorption (Additional file 2: Table S5). However, after correction for multiple tests, the results were not statistically significant for all pathways.

Ingenuity pathway analysis (IPA)

IPA identified 16 canonical pathways (Fisher’s exact crude P < 0.05, Additional file 2: Table S6). None of the pathways achieved statistical significance after Benjamini–Hochberg correction for multiple tests. Differentially methylated genes were most enriched in the G-Protein Coupled Receptor Signaling (9 genes, crude P < 0.05). There was also some evidence of enrichment (crude P < 0.05) in the Pentose Phosphate (Oxidative Branch) and Estrogen Biosynthesis pathways.

Differentially methylated regions (DMRs)

To reduce data dimensionality and identify differential methylations over gene areas, we analyzed the methylation data for differentially methylated regions (DMRs) by DMRcate and comb-p. The DMRcate program identified 135 DMRs at FDR < 5% between LGA and OGA groups (Additional file 2: Table S7), and these DMRs were annotated to 94 genes, while the other DMRs were located in the OpenSea areas. The comb-p program identified 31 DMRs with Sidak corrected P < 0.05 (Additional file 2: Table S8), 15 DMRs were annotated to genes, while the other DMRs were located in the OpenSea areas. Three DMRs were identified by both DMRcate and comb-p, and these loci were annotated to 3 genes (PRMT2, BACE1 and TRAK2; Table 4).

Table 4 Placental DNA differentially methylated regions (DMRs) in LGA vs. OGA newborns identified in both comb-p and DMRcate package analysesCorrelations of DMPs with birth weight and cord blood biomarkers

Among the 543 DMPs, 494 CpG sites were correlated with birth weight z score (crude P < 0.05), and the correlations for 486 DMPs remained statistically significant after correction for multiple tests (Additional file 2: Table S9). These 486 DMPs were annotated to 286 genes. In LASSO regression identifying the most important DMPs, 25 DMPs were selected and could explain 80.2% of the variations in birth weight (z) (Additional file 2: Table S10). These 25 DMPs were annotated to 20 genes (QSOX1, FCHSD2, LOC101928162, ADGRB3, GCNT1, TAP1, MYO16, NAV1, ATP8A2, LBXCOR1, EN2, INCA1, CAMTA2, SORCS2, SLC4A4, RPA3, UMAD1,USP53, OR2L13 and NR3C2) (Additional file 2: Table S10).

Additional file 2: Table S11 presents the correlations of DMPs with cord blood insulin, proinsulin, C-peptide, IGF-II, IGF-I, leptin, total or HMW adiponectin. All these correlations did not reach statistical significance after Benjamini and Hochberg correction for multiple tests.

Gene-specific correlations of placental DNA methylations and cord blood biomarkers

Additional file 3: Table S12 presents the gene-specific correlations of placental gene CpG sites and cord blood biomarkers. For INS-IGF2/IGF2AS/IGF2, methylation levels in four CpG sites (cg08014499, cg21728792, cg05203776 and cg22225943) were positively correlated with cord blood insulin (r = 0.26 to 0.28, crude P = 0.03 to 0.05), and in one CpG site positively correlated with cord blood C-peptide (cg17434309, r = 0.28, crude P = 0.03), proinsulin (cg17434309, r = 0.31, crude P = 0.02) and IGF-II (cg23889607, r = 0.30, crude P = 0.02). Methylation levels in four CpG sites (cg13928782, cg21574853, cg07096953 and cg20088847) were negatively correlated with cord blood IGF-II (r = − 0.30 to − 0.35, crude P = 0.007 to 0.024), and in one CpG site negatively correlated with cord blood proinsulin (cg13670288, r = − 0.26, crude P = 0.049). All these correlations, however, did not reach statistical significance after correction for multiple tests.

Several correlations between ADIPOQ gene methylation levels and cord blood total or HMW adiponectin remained statistically significant after correction for multiple tests. Total adiponectin was correlated with cg16126291 (r = − 0.35, adjusted P = 0.03), cg02235049 (r = − 0.32, adjusted P = 0.047), cg10681525 (r = − 0.40, adjusted P = 0.01) and cg18537894 (r = 0.40, adjusted P = 0.01). HMW adiponectin was correlated with cg16126291 (r = − 0.42, adjusted P = 0.01) and cg18537894 (r = 0.36, adjusted P = 0.036).

留言 (0)

沒有登入
gif