Fifty preterm (<32 weeks gestation) and 40 term (37–42 weeks gestation) infants were recruited within the first week of life from the Simpson Center for Reproductive Health, Edinburgh, UK, with informed written parental consent. Ethical approval was obtained from the South East Scotland Research Ethics Committee (Reference 11/AL/0329). NHS management approval was obtained (Lothian R&D Project number 2011/R/NE/03). Perinatal samples were collected under the Edinburgh Reproductive Tissue BioBank (ERTBB) (West of Scotland Research Ethics Service Reference 09/S0704/3). All parents gave written informed consent and all studies were performed in accordance with the declaration of Helsinki. Infant samples were collected under the framework of the ERTBB following an amendment to ethical approval (Reference AM07/1). Demographic details were obtained during clinic visits and from hospital records. All of the preterm infants were admitted to the neonatal unit; of these, six infants died. In three of these infants, saliva for buccal cells was not collected after birth due to clinical instability and DNA was of poor quality in a fourth infant. Five of the term infants were also admitted to the neonatal unit for a short period (respiratory distress syndrome, weight loss, and hemolysis from Rhesus isoimmunization) but none required follow-up. Preterm infants were seen at birth, at term-corrected age, and at 3 months and 1-year corrected ages; term infants were seen at birth, 3 months, and 1 year. All visits occurred in the afternoon, supervised by one researcher (Chinthika Piyasena).

Weight, length, and occipitofrontal head circumference (OFC) was measured by one trained researcher. Percentage body fat mass was measured by air displacement plethysmography in preterm infants at term-corrected age and 3 months corrected age, and in term infants at birth and 3 months using the PEAPOD Body Composition System (COSMED, Chicago, IL, USA). Skin fold thickness (subscapular and triceps) was measured at 1 year by the same trained researcher. Term infants were measured at a median of 2 days (range 0–8) after birth, at 3 months (13.3 weeks; range 10.4–16.9), and 1 year (52.3 weeks; range 48.1–57.4). Preterm infants were measured at term-corrected age at a median of 40 (range 35 + 0 to 44 + 1)-corrected weeks, at 3 months corrected (13.4 weeks; range 10.3–18.3) and 1 year corrected (53.9 weeks; range 52.1–68.7).

Saliva was collected using the Oragene DNA (OG-250) kits and saliva sponges CS-1 and extracted using prepIT-2LP (DNA Genotek, Ottawa, ON, Canada). DNA was quantified using the Qubit 2.0 Fluorometer (Life Technologies, Paisley, UK). Five hundred nanograms of DNA were bisulfite converted using the EZ DNA Methylation Gold Kit (Zymo Research Corporation, CA, USA). Pyrosequencing was performed to analyze DNA methylation for DMRs controlling IGF2 expression: IGF2 DMR2 (IGF2DMR2, n = 9 CpGs) and the H19 imprinting control region (H19ICR, n = 8 CpGs) as previously described (22). Primers were purchased from Invitrogen (Life Technologies, Paisley, UK). DNA was amplified using the AmpliTaq Gold 360 kit (Applied Biosystems, Warrington, UK) and pyrosequencing performed using PyroMark Q24Gold reagents on a PyroMark Q24 Pyrosequencer (Qiagen, Crawley, UK). Data were analyzed using PyroMark Q24 1.0.10. Percentage DNA methylation is expressed as the average across all CpGs in each of the two loci in IGF2.

Methylation analysis of three CpGs in FKBP5 intron 7, two of which are located in consensus GRE motif (CpG 2 and 3), was performed in triplicates using a protocol adapted from Klengel et al. (20). One hundred twenty nanograms genomic DNA was bisulfite converted using the EZ DNA Methylation Kit (Zymo Research Corporation, CA, USA). Bisulfite converted DNA was amplified in a 50 µl reaction mix (4–10 µl DNA; each bisulfite specific Primer with a final concentration of 0.2 µM, FKBP5int7_P1_F: GTTGTTTTTGGAATTTAAGGTAATTG, and FKBP5int7_P1_R_biot: biotin-TCTCTTACCTCCAACACTACTACTAAAA) using the Kapa HIFI Uracil + Hot start Ready Mix (Kapa Biosystems Inc., Wilmington, DE, USA). Cycling conditions of the touchdown PCR were 98°C for 5 min, 2× (98°C – 40 s, 62°C – 30 s, 72°C – 60 s), 5× (98°C – 40 s, 60°C – 30 s, 72°C – 60 s), 8× (98°C – 40 s, 58°C – 30 s, 72°C – 60 s), 34× (98°C – 40 s, 56°C – 30 s, 72°C – 60 s), 72°C for 1 min and cooling to 4°C. Pyrosequencing of FKBP5 CpGs was performed on a PyroMark Q96 ID system (QIAGEN GmbH, Hilden) using PyroMark Q96Gold reagents with the following sequencing primer: FKBP5int7_P1_S2: 5′-GTTGATATATAGGAATAAAATAAGA-3′ for CpG1 and CpG2 and FKBP5int7_P1_S3: 5′-TGGAGTTATAGTGTAGGTTTT-3′ for CpG3. PyroMark Q96 ID Software 2.5 (QIAGEN GmbH, Hilden) was used to calculate percentage methylation.

The legends for Figures 1 and 2 state the number of values/measurements analyzed, excluding samples, which were discarded because of insufficient DNA and/or poor pyrosequencing quality. Additionally, we were unable to collect saliva from several babies for various reasons, although these babies still underwent anthropometry.

FKBP5 rs1360780 genotyping was performed on a Roche LightCycler 480 System using a TaqMan SNP Genotyping Assay (Applied Biosystems). Thermal cycling conditions were 95°C for 10 min, 45× (95°C – 15 s, 60°C – 1 min, 50°C – 30 s). Genotypes were called using endpoint genotyping LightCycler 480 Software version 1.5. Genotypes were in Hardy–Weinberg equilibrium (p = 0.17). Participants were divided into protective–genotype (CC) and risk-allele carriers (CT or TT).

Covariates that could confound the association or be in the causal pathway were added into the model in a hierarchical manner: maternal smoking, male gender, and social deprivation for FKBP5 and IGF2/H19. Additionally, breast milk at 3 months for IGF2/H19 at 1 year and risk-allele carriage for FKBP5 at all time points. Social deprivation was coded as deprivation category (DEPCAT) scores based on the mother’s postcode at booking and obtained from the “Carstairs scores for Scottish postcode sectors from the 2001 Census.” In this system, scoring is based on the material affluence/deprivation of the area in which a person lives. Postcode sectors are allocated a DEPCAT score, derived from four sets of information: overcrowding, male unemployment, car ownership, and the proportion of people in households in social class 4 or 5 and scores range from 1 to 7 where 7 indicates the worst social deprivation. Maternal smoking was categorized as current smoker, never smoked, former (stopped pre-pregnancy), or former (stopped during pregnancy). Breast milk at 3 months indicates whether or not the infant was receiving any breast milk at 3 months corrected age.

Weight, length, OFC, skin fold thickness, and weight gain were adjusted for age and gender by converting to SD scores (SDS) (z-scores) using LMSgrowth, a Microsoft Excel Add-in to access growth references that define the UK-WHO growth charts (23). Demographic, clinical characteristics, and risk-allele carriage between preterm and term infants were compared using independent samples t-testing and chi-square analyses, as appropriate. Multivariate linear regression was used to assess variation in body composition and to test the hypothesis that preterm birth is associated with altered DNA methylation. Outcome variables were percentage DNA methylation at birth, term-corrected age, and 1 year. Unstandardized regression (β) coefficients from these models indicate the change in percentage methylation associated with prematurity and a one-unit change in the other predictors. Paired samples t-testing was used where appropriate. One-way ANOVA with Dunnett’s post hoc t testing was used to test the effect of maternal smoking on DNA methylation at FKBP5. Statistical significance for all analyses was set at p < 0.05 (two-tailed).

Characteristics of the cohort are shown in Table 1. Birth weight SDS was lower in preterm infants p < 0.001, and there were more males in the preterm group, p = 0.006. There was no significant difference between the frequency of FKBP5 risk-allele carriage between the groups (p = 0.43). Maternal age, body mass index (BMI), DEPCAT scores, folic acid in first trimester, and maternal smoking were different between the two groups.

Weight and OFC was reduced in preterm infants at all time points compared to term infants (Figures 1A–C). Preterm infants were shorter than term infants at term-corrected age and 3 months corrected age, but not at 1-year corrected (Figures 1A–C). Preterm infants had a higher percentage body fat at term-corrected age when compared with term infants at birth (Figure 1D) (mean difference 5.5%, 95% CI [8.0, 3.0], p < 0.001) and this remained significant following adjustment for gender (β = 5.7, 95% CI [3.1, 8.3], p < 0.001). The difference in percentage body fat did not persist at 3 months/3 months corrected, including when adjusted for gender and breast milk intake (β = −0.3, 95% CI [−2.8, 2.2], p = 0.82) (Figure 1D). At 1-year corrected, preterm infants had lower triceps skin fold thickness SDS than term infants, adjusted for breast milk exposure at 3 months (β = −0.4, 95% CI [−1.3, −0.3], p = 0.008), but there were no differences in subscapular skin fold thickness (Figure 1E).

Mean DNA methylation at IGF2DMR2 was lower in preterm infants at birth compared to term infants at birth and this remained significant in adjusted analyses (β = −11.2, 95% CI [−15.2, −7.3], p < 0.001) (Figure 2B). There was no difference in mean DNA methylation at the H19ICR between preterm and term infants at birth after adjusted analysis (β = −1.3, 95% CI [−3.8, 1.2], p = 0.3) (Figure 2A). In preterm infants at term-corrected age, DNA methylation was reduced at IGF2DMR2 and H19ICR compared to infants born at term, and this remained significant in adjusted analyses (β = −2.8, 95% CI [−5.0, −0.6], p = 0.01 and β = −2.3, 95% CI [−4.6, −0.1], p = 0.049, respectively) (Figures 2A,B). However, the significance was attenuated when social deprivation was added into the model for IGF2DMR2 (β = −2.1, 95% CI [−4.3, 0.05], p = 0.055). At 1 year of age, there was no difference in mean DNA methylation in adjusted analyses at either IGF2DMR2 (β = −0.3, 95% CI [−4.0, 3.3], p = 0.86) or H19ICR (β = 0.8, 95% CI [−1.5, 3.0], p = 0.49) (Figures 2A,B).

The presence of the FKBP5 risk allele was associated with higher DNA methylation at FKBP5 CpG3 at birth (β = 4.0, 95% CI [0.3, 7.6], p = 0.03) and term-corrected age (β = 2.5, 95% CI [0.4, 4.5], p = 0.02) across all infants. DNA methylation at CpG1, CpG2, and CpG3 was lower in preterm infants at birth and at term-corrected age when compared to term infants at birth, including after adjusted analysis (Table 2). There was an increase in DNA methylation at all three CpGs in both groups during the first year (Figures 2C–E) and, by 1 year of age, there were no persistent differences in mean methylation at any CpG between preterm and term infants (Table 2). The presence/absence of the FKBP5 risk allele as a covariate did not alter the findings at 1 year and there was no moderation of the relationship between prematurity and DNA methylation by the presence of the risk allele at any time.

IGF2DMR2 methylation was highly significantly related to birthweight SDS in preterm (R = 0.7, p < 0.001), but not term infants. DNA methylation across FKBP5 was also positively associated with birthweight SDS (β = 0.3, 95% CI [0.3, 5.4], p = 0.03) when adjusted for prematurity, but only in the preterm infants when analyzed separately (R = 0.4, p = 0.04). There were no other significant relationships between DNA methylation at IGF2DMR2, H19ICR, or FKBP5 and weight or measures of body composition in either preterm or term infants at any other time point (Table 3).

Social deprivation was an independent contributor toward reducing DNA methylation at IGF2DMR2 in preterm infants at birth (β = −1.6, 95% CI [−2.8, −0.3], p = 0.02) and at term-corrected age (β = −0.9, 95% CI [−1.5, −0.2], p = 0.02) although this did not persist at 1-year corrected age (p = 0.07). Social deprivation was also independently associated with a reduction in DNA methylation at FKBP5 CpG2 in preterm infants at term-corrected age (β = −1.5, 95% CI [−2.9, −0.02], p = 0.047). Finally, maternal smoking was independently associated with a marked reduction in DNA methylation in preterm infants at FKBP5 CpG1 (β = −2.6, 95% CI [−4.4, −0.7], p = 0.008) and CpG2 (β = −2.1, 95% CI [−4.1, −0.03], p = 0.047) at birth, compared to term infants. Additionally, maternal smoking was associated with a reduction in DNA methylation at birth in both groups at these loci (Table 4).