Objective
Cesarean section (CS) has been associated with a greater risk for asthma, diabetes, and cancer later in life. Although elective CS continues to rise, it is unclear whether and how it may contribute to compromised future health. Our aim was to investigate the influence of mode of delivery on the epigenetic state in neonatal hematopoietic stem cells.
Study Design
This was an observational study of 64 healthy, singleton, newborn infants (33 boys) born at term. Cord blood was sampled after elective CS (n = 27) and vaginal delivery. Global deoxyribonucleic acid (DNA) methylation in hematopoietic stem cells (CD34+) was determined by luminometric methylation assay, and genome-wide, locus-specific DNA methylation analysis was performed by Illumina Infinium 450K (Illumina, San Diego, CA), validated by bisulfite-pyrosequencing.
Results
CD34+ cells from infants delivered by CS were globally more DNA methylated (+2%) than DNA from infants delivered vaginally ( P = .02). In relation to mode of delivery, a locus-specific analysis identified 343 loci with a difference in DNA methylation of 10% or greater ( P < .01). A majority of the differentially methylated loci in neonatal CD34+ cells (76%) were found to be hypermethylated after vaginal delivery. In these infants, the degree of DNA methylation in 3 loci correlated to the duration of labor. The functional relevance of differentially methylated loci involved processes such as immunoglobulin biosynthetic process, regulation of glycolysis and ketone metabolism, and regulation of the response to food.
Conclusion
A possible interpretation is that mode of delivery affects the epigenetic state of neonatal hematopoietic stem cells. Given the functional relevance indicated, our findings may have important implications for health and disease in later life.
Cesarean section (CS) rates increase rapidly worldwide. Today CS is the most common surgical procedure performed in women of child-bearing age. In up to 15% of deliveries, CS may be indicated according to recommendations from the World Health Organization. However, most countries exceed this recommendation, suggesting that many women are undergoing CS without a medical indication. Although short-term outcomes after CS are well characterized, the basis for suggested long-term consequences of this global change in childbirth is mostly unknown.
Recent clinical and epidemiological studies have shown that birth by CS is associated with a greater risk of developing diseases later in life, such as asthma, allergies, type 1 diabetes, celiac disease, obesity, and malignancies. Although known confounders have in many of these studies been accounted for, it is still unclear whether and how CS may compromise health in the offspring. The lack of appropriate gut colonization and microbiome exposure, the lack of the immune-activating effects of labor, and epigenetic changes that may modify the immune system have all been suggested as mechanisms for CS-related effects on health and disease risk.
We previously found support for altered epigenetic states in blood cells from newborns delivered by elective CS compared with those vaginally born. Epigenetic states provide mechanisms for the functional genome and mediate adaptations to a dynamic environment. Epigenetic deoxyribonucleic acid (DNA) methylation may retain its stability for the cells’ lifetime, even through divisions. Accordingly, DNA methylation and epigenetic cell memory associated with the mode of delivery could be mechanisms for later differences in disease risk, especially if these occur and are propagated in progenitor cells.
Before a more conclusive answer can be given on an epigenetic memory of birth, several important questions remain to be resolved. We hypothesized that DNA methylation in neonatal stem cells differs in relation to mode of delivery and between different genes/gene regions. Data presented herein suggest that CS is associated with altered epigenetic states of neonatal CD34+ hematopoietic stem cells, involving differential DNA methylation of genes/gene regions relevant for later immune-mediated diseases.
Materials and Methods
Participants
Pregnant women were recruited at the delivery units at Danderyd Hospital in Stockholm, Sweden. Multiple pregnancies, maternal diabetes or gestational diabetes, maternal hypertension, preeclampsia, smoking during the index pregnancy, preterm delivery (gestational age <37 weeks), small-for-gestational-age infants (birthweight ≥2 SD below the mean for a Swedish reference population, neonatal asphyxia (Apgar score <7 at 1 and 5 minutes), malformations, chromosomal disorders, or congenital infection were all exclusion criteria. No pregnancy resulting from assisted reproductive technology was included in the study.
For measurement of global DNA methylation by luminometric methylation assay (LUMA) in cord blood stem cells, we included 40 infants (18 girls) to women delivered by elective CS before the start of labor and under spinal analgesia, and as reference group, 49 infants (22 girls) born after spontaneous, nonassisted vaginal delivery (VD) were included. After cell separation and DNA extraction from stem cells (see the following text), 43 samples (18 CS and 25 VD) contained sufficient DNA (>500 ng DNA) for methylation analyses.
In the VD group, the start of labor was defined as the time point at which the pregnant woman for the first time perceived regular (3-4 per 10 minutes) and painful uterine contractions. When admitted to the hospital, all women were asked about the time point (hours and minutes) for the start of labor. If labor had not started before admission, the start of labor was noted by the attending midwife. Deliveries with induction of labor were not included in this study.
Indications for CS included maternal request, previous CS, breech position, or pelvic disproportion. In the VD group, the median duration of labor was 14.5 hours (range, 1–53 hours), and the median duration of ruptured membranes was 4 hours (range, 0–17 hours).
Because of exhausted blood samples from the first group, a second group of infants was recruited. Cord blood from 12 infants (6 CS) was used to fill 1 Illumina 450K array (Illumina, San Diego, CA) to measure genome-wide, locus-specific DNA methylation. The DNA content in 9 of these blood samples was sufficient for subsequent validation analysis using bisulfite pyrosequencing. To increase the numbers and power of the validation analysis, we recruited an additional 10 infants (4 CS) to the second study group. There were no differences in maternal characteristics, gestational age (GA), sex distribution, or birthweight between the first and second study groups.
In the whole cohort of 64 mothers, the median maternal age was 35 years (range, 23–43), the body mass index (BMI) was 22.8 kg/m 2 (range, 15.9–38.4), and 17 of 64 mothers were primigravida. The GA was 39.2 (range, 37.6–42) weeks, and all infants had birthweights appropriate for gestational age (3667 g; range, 2820–4915 g).
Mothers in the CS group (n = 27) were older compared with mothers in the VD group (n = 37) (37 [range, 25–43] vs 34 [range, 23–41] years; P = .03), and GA was shorter in the CS group compared with the VD group (38.9 [range, 37.7–39.9] vs 40.3 [range, 37.6–42.0] weeks; P < .001). Maternal and infant characteristics by mode of delivery for LUMA, Illumina, and bisulfite pyrosequencing groups are presented in the Table .
| Characteristic | CS (n = 27) | VD (n = 37) | P value |
|---|---|---|---|
| Maternal age, y a | 37 (25–43) | 34 (23–41) | .03 |
| Prepregnancy BMI, kg/m 2 | 22.9 (18.9–33.4) | 22.7 (15.9–38.4) | .75 |
| Parity, n | 2 (1–4) | 2 (1–4) | .19 |
| Gestational age, wks b | 38.9 (37.7–39.9) | 40.3 (37.6–42.0) | < .001 |
| Birthweight, g | 3625 (2820–4645) | 3675 (2985–4915) | .97 |
| Infant sex, girls/boys | 11/16 | 20/17 | .32 |
a Maternal age higher in the CS group ( P = .03)
The regional ethical review board approved the study protocol and informed consent was obtained from parents before birth (number 2010/440-31/4; 2012/1029/32).
Blood sampling and preparation of hematopoietic stem cells
In all participating infants, 15-20 mL blood was sampled in EDTA tubes from the umbilical cord directly after cord clamping. The cord was clamped after 30 seconds to obtain the targeted volume of cord blood. Blood cells were sorted with commercially available tools (Dynabeads positive isolation kit; Invitrogen by Life Technologies Corp, Carlsbad, CA) to separate CD34+ stem cells from other DNA-containing cells. DNA in stem cells was extracted using Illustra DNA extraction (GE Healthcare Europe GmbH, Freiburg, Germany), and DNA was quantified using the NanoDrop ND-1000 (NanoDrop Technologies Inc/Thermo Fisher Scientific Inc, Wilmington, DE).
Only samples with sufficient DNA for methylation analyses (>500 ng DNA; n = 43, 18 CS and 25 VD) were frozen (–70°C) until DNA methylation analyses. All DNA methylation analyses were performed with the investigators blinded to mode of delivery.
Global DNA methylation analysis in hematopoietic stem cells
LUMA was used for assessing the global methylation (n = 43) and was performed as described.
Genome-wide, locus-specific DN-methylation analysis in hematopoietic stem cells
Quantitative DNA methylation measurements of bisulfite-treated genomic DNA were performed using Illumina Infinium Human Methylation 450K BeadChip (Illumina). One chip was used, allowing for 12 samples (6 CS) from cord blood stem cells to be analyzed. In this chip, more than 450,000 cytosine-phosphate-guanine (CpG) sites are interrogated. For each CpG and sample, the methylation level was estimated as a ratio (β) of the methylated signal to the sum of methylated and unmethylated signals. Before normalization, CpG sites located at known single-nucleotide polymorphisms (list provided by the manufacturer) were discarded to avoid potential confounding by single-nucleotide polymorphisms in addition to CpGs located on the X and Y chromosomes. In addition, we discarded CpG probes with detection values of P > .01.
To estimate the β values for the included CpG probes, a 3-step pipeline, previously described as optimal, was used. Differential DNA methylation was computed by transforming β values into M values and using a software package for the analysis of gene expression microarray data, limma (Linear Models for Microarray Data), to define a linear model. Differentially methylated positions (DMPs) were defined as those that exhibited a 10% or greater difference in the DNA methylation between the VD and CS, at a value of P < .01.
This definition was chosen to disclose the most important set of DMPs while considering the following: (1) most differences in methylation are mild; (2) we investigated a large number of probes that required multiple testing; and (3) we had only a small number of samples. We considered a 10% methylation difference as a conservative threshold for a greater difference in DNA methylation after observing the density of the median differences between CS and VD ( Appendix ; Supplementary Figure 1 ). By using the annotation provided by Illumina, we summarized the DMP information over genes by computing the number of probes for each gene, the number of DMPs, and how many DMPs were hypermethylated or hypomethylated.
Due to a misclassification error of gestational age in the clinical records, 1 preterm infant (GA 33 weeks, CS group) was inadvertently included in the genome-wide, locus-specific DNA methylation analysis. The results from this infant were excluded from all statistical calculations comparing CS with VD; hence, 11 samples instead of 12 were analyzed from the Illumina 450K BeadChip.
Functional enrichment
To investigate the functional relevance of the selected DMPs, the Genomic Regions Enrichment of Annotations Tool (GREAT) analysis was applied. GREAT helps in identifying gene set enrichments (GREAT version 2.0.2; http://bejerano.stanford.edu/great ).
Validation by bisulfite pyrosequencing
A subset of CpGs, a priori clinically relevant, or CpGs exhibiting the largest DNA methylation differences in relation to the mode of delivery, was validated ( Supplementary Table 1 ). The selected CpGs are associated with the genes COLEC11 , PCK2 , PGBD5 , and HLA-F . Genomic DNA (500 ng) was treated with sodium bisulfite (EpiTect bisulfite kit; QIAGEN, Valencia CA). One microliter of converted DNA (∼10 ng) was applied as a template in the PCRs performed with the PyroMark PCR kit (QIAGEN). The entire PCR product and 4 pmol of the respective sequencing primer, and streptavidin sepharose high-performance beads (GE Healthcare), were used for pyrosequencing performed with the PSQ 96 system and the PyroMark Gold Q96 reagent kit (QIAGEN). The PyroMark CpG software 1.0.11 served for data analysis.
Statistical analyses
The power calculation for the sample size was based on findings in our previous publication in which DNA methylation ( Hpa II/ Msp I ratio) in the VD and CS groups had the following mean values (SD): 0.30 (0.046) and 0.25 (0.022). Assuming normal distribution and equal variance, the power calculation estimated 10 samples in each of the 2 groups would be needed to detect a difference in the DNA methylation of the same magnitude (2-sided test, 5% significance level, 80% power). The power calculation was performed in STPLAN version 4.5 (Obtained from MD Anderson Cancer Center. Available at: https://biostatistics.mdanderson.org/SoftwareDownload/SingleSoftware.aspx?Software_Id=41 . Accessed June 12, 2014).
Data are expressed as median values and range. Statistical analyses were performed using rank sum tests (Mann-Whitney U test and Wilcoxon signed-rank test), and associations were tested for by calculating the Spearman’s correlation coefficients. Tests for DNA methylation differences between CS and VD included GA as an independent variable. All analyses regarding the locus specific methylation were performed using R.
Correlation between DNA methylation and duration of labor
Spearman correlation was used to compute the correlation between duration of labor and β value of each probe. Supplementary Figure 2 shows the distribution of the correlations; for all nondiscarded probes, the distribution is centered around 0 (mean, –0.05; median, –0.04), whereas for DMP, the correlation is larger (mean, 0.1; median, 0.17). We used a Kruskal Wallis test to compute the significance of the difference between DMPs vs non-DMPs, obtaining a value of P < 10 –10 . Considering the all-probe correlations as a null distribution, we used as thresholds the 0.01 and 0.99 quantiles to identify significant correlations in the DMP set. All statistical analyses were conducted using R.
Results
Global DNA methylation
In this study, isolated CD34+ hematopoietic stem cells displayed significantly more methylated DNA in cells from CS infants compared with cells from VD infants ( P < .02, Figure 1 ). Global DNA methylation in the neonatal stem cells did not correlate with maternal characteristics (age, prepregnancy BMI, parity, duration of delivery, and duration of ruptured membranes) or infant risk factors (GA, sex, birthweight) (values of P = .13 to P = .91).
Locus/gene-specific DNA methylation
To obtain better knowledge about specific genes differentially modified in relation to mode of delivery, we assessed locus-specific DNA methylation. This resulted in the identification of 343 DMPs exhibiting a difference in DNA methylation of 10% or more ( P < .01; Supplementary Table 2 and Figure 2 ). The maximal locus-specific difference in DNA methylation in the stem cells from CS and VD infants was 40%. Among the 343 DMPs, 179 (52%) were associated with known genes ( Supplementary Table 3) , and a majority of the DMPs were in gene bodies.
