Study participants
Initial participants in the present study consisted of 702 Japanese women aged ≥ 20 years who gave birth to a single child between November 2015 and April 2018. Maternal and fetal information was ascertained from medical and prenatal records. The following data was collected: maternal age, previous pregnancies (i.e., primipara or multipara), medical history (e.g., infections, tumors, hypertension, diabetes, and mental disorders), HDP, GDM, smoking, maternal height, pre-pregnancy body weight, gestational weight gain (GWG), fetal gender, gestational age, and birth weight.
The flow chart of this cohort is shown in figure 1. Of the 702 subjects, considering the effect on fetal DNA methylation, mothers who delivered at preterm, with complications, and with smoking habit in the preconception period and/or during pregnancy were excluded.
After excluding these mothers, the data of 386 mother-fetal pairs were available for DNA methylation analysis. Overall, FT-LBW (defined as a birth weight < 2,500 g, born at 37–41 weeks of gestation) was 25 subjects. Among these subjects, propensity score was matched for infant gender, maternal age, and primipara/multipara status. Subsequently, by randomized selection according to the median birth weight (3,019 g), four FT-LBWs and five FT-NBWs were used as references for analysis.
Genome-wide DNA methylation analysis
Umbilical cord blood samples were collected at birth and frozen at -80 °C until DNA extraction. DNeasy Blood and Tissue Kits (Qiagen) were used to extract DNA from the white cell fraction of cord blood and purified.
For each subject, 865,918 CpG sites across the genome were interrogated using the Illumina Infinium Human Methylation BeadChip (Illumina, USA) [23]. Briefly, 1 µg DNA was converted with sodium bisulfite, amplified, fragmented, and hybridized according to the manufacturer’s instructions. After implementing bias adjustment by the beta-mixture quantile normalization method, 862,260 CpGs were used for analysis. β-values were converted into M-values (log2 ratio of β-value) to account for heteroscedasticity. Differences in methylation in promoter regions relative to UCSC reference sequences were identified between FT-LBW and FT-NBW. A Benjamini and Hochberg false discovery rate control was applied. DNA methylation differences were selected based on log2 ratio of β-value differences ≥ 0.6.
Gene ontology functional enrichment analysis
We performed functional enrichment analysis using the Database for Annotation, Visualization and Integrated Discovery v6.8 (https://david.ncifcrf.gov/) to identify the biological functions of hyper- and hypomethylated DNA [24]. This approach evaluates DNA methylation data in terms of categories of gene function rather than individual genes. Genes were ranked by magnitude of correlation with a class distinction in the GOTERM_BP algorithm, and an enrichment score was calculated. Significance was defined as P < 0.05.
Statistical analysis
Participants’ characteristics and baseline data are presented as n (%) for categorical variables or mean ± standard deviation for continuous variables. The Chi-square test (categorical data) and the student’s t-test (continuous data) in SPSS (Institute, version 23, IBM) was performed to compare characteristics between FT-LBW and FT-NBW groups. Significance was defined as P < 0.05.