Childhood Bone Mineral Content Is Associated With Methylation Status of the RXRA Promoter at Birth

Authors

  • Nicholas C Harvey,

    1. Medical Research Council (MRC) Lifecourse Epidemiology Unit, University of Southampton, UK
    2. National Institute for Health Research (NIHR) Southampton Biomedical Research Centre, University of Southampton and University Hospital Southampton National Health Service (NHS) Foundation Trust, Southampton General Hospital, Southampton, UK
    Search for more papers by this author
    • NCH and AS are joint first authors.
  • Allan Sheppard,

    1. Liggins Institute, University of Auckland, New Zealand
    2. AgResearch, Ruakura Research Centre, Hamilton, New Zealand
    Search for more papers by this author
    • NCH and AS are joint first authors.
  • Keith M Godfrey,

    1. Medical Research Council (MRC) Lifecourse Epidemiology Unit, University of Southampton, UK
    2. National Institute for Health Research (NIHR) Southampton Biomedical Research Centre, University of Southampton and University Hospital Southampton National Health Service (NHS) Foundation Trust, Southampton General Hospital, Southampton, UK
    Search for more papers by this author
  • Cameron McLean,

    1. Liggins Institute, University of Auckland, New Zealand
    2. AgResearch, Ruakura Research Centre, Hamilton, New Zealand
    Search for more papers by this author
  • Emma Garratt,

    1. Southampton Institute of Developmental Sciences, University of Southampton, Southampton, UK
    Search for more papers by this author
  • Georgia Ntani,

    1. Medical Research Council (MRC) Lifecourse Epidemiology Unit, University of Southampton, UK
    Search for more papers by this author
  • Lucy Davies,

    1. Medical Research Council (MRC) Lifecourse Epidemiology Unit, University of Southampton, UK
    Search for more papers by this author
  • Robert Murray,

    1. Medical Research Council (MRC) Lifecourse Epidemiology Unit, University of Southampton, UK
    Search for more papers by this author
  • Hazel M Inskip,

    1. Medical Research Council (MRC) Lifecourse Epidemiology Unit, University of Southampton, UK
    Search for more papers by this author
  • Peter D Gluckman,

    1. Liggins Institute, University of Auckland, New Zealand
    2. Singapore Institute for Clinical Sciences, Singapore
    Search for more papers by this author
  • Mark A Hanson,

    1. Southampton Institute of Developmental Sciences, University of Southampton, Southampton, UK
    Search for more papers by this author
  • Karen A Lillycrop,

    1. Southampton Institute of Developmental Sciences, University of Southampton, Southampton, UK
    Search for more papers by this author
    • KAL and CC are joint senior authors.
  • Cyrus Cooper

    Corresponding author
    1. Medical Research Council (MRC) Lifecourse Epidemiology Unit, University of Southampton, UK
    2. National Institute for Health Research (NIHR) Southampton Biomedical Research Centre, University of Southampton and University Hospital Southampton National Health Service (NHS) Foundation Trust, Southampton General Hospital, Southampton, UK
    3. NIHR Musculoskeletal Biomedical Research Unit, University of Oxford, Oxford, UK
    • Address correspondence to: Cyrus Cooper, MA, DM, FRCP, FFPH, FMedSci, MRC Lifecourse Epidemiology Unit, Southampton General Hospital, Southampton SO16 6YD, UK. E-mail: cc@mrc.soton.ac.uk

    Search for more papers by this author
    • KAL and CC are joint senior authors.

ABSTRACT

Maternal vitamin D deficiency has been associated with reduced offspring bone mineral accrual. Retinoid-X receptor-alpha (RXRA) is an essential cofactor in the action of 1,25-dihydroxyvitamin D (1,25[OH]2-vitamin D), and RXRA methylation in umbilical cord DNA has been associated with later offspring adiposity. We tested the hypothesis that RXRA methylation in umbilical cord DNA collected at birth is associated with offspring skeletal development, assessed by dual-energy X-ray absorptiometry, in a population-based mother-offspring cohort (Southampton Women's Survey). Relationships between maternal plasma 25-hydroxyvitamin D (25[OH]-vitamin D) concentrations and cord RXRA methylation were also investigated. In 230 children aged 4 years, a higher percent methylation at four of six RXRA CpG sites measured was correlated with lower offspring bone mineral content (BMC) corrected for body size (β = −2.1 to −3.4 g/SD, p = 0.002 to 0.047). In a second independent cohort (n = 64), similar negative associations at two of these CpG sites, but positive associations at the two remaining sites, were observed; however, none of the relationships in this replication cohort achieved statistical significance. The maternal free 25(OH)-vitamin D index was negatively associated with methylation at one of these RXRA CpG sites (β = −3.3 SD/unit, p = 0.03). Thus, perinatal epigenetic marking at the RXRA promoter region in umbilical cord was inversely associated with offspring size–corrected BMC in childhood. The potential mechanistic and functional significance of this finding remains a subject for further investigation. © 2014 American Society for Bone and Mineral Research.

Introduction

Poor intrauterine growth is a predictor of later osteoporosis and fracture risk,[1, 2] and we have previously demonstrated that maternal diet, lifestyle, body build,[3, 4] and 25-hydroxyvitamin D (25[OH]-vitamin D) status during pregnancy[5, 6] are associated with bone mineral accrual in the offspring. Although it has been suggested that a large proportion of the variance in peak bone mass, achieved in early adulthood, is attributable to heritable fixed genetic factors, several genomewide association studies have failed to identify single-nucleotide polymorphisms (SNPs) which might explain more than a modest proportion of the overall variation in bone mass in the general population.[7, 8] Attention is now turning to epigenetic processes, such as DNA methylation, which might provide potential mechanisms to account for relationships between transgenerational and perinatal environmental influences and offspring bone development.[9] Indeed, epigenetic mechanisms such as methylation may be critical to imprinting, allowing expression of a gene copy dependent on the parent of origin, demonstrated to be relevant to disease outcomes and factors involved in growth such as insulin-like growth factor 2 (IGF2).[10-12] Furthermore, epigenetic effects on gene expression in the offspring may originate from variations in maternal diet during pregnancy[13]; we have previously found evidence, using a candidate approach in two independent cohorts initially informed by an array-based investigation, that methylation at the promoter region of the retinoid-X receptor-alpha (RXRA, also known as NR2B1) in umbilical cord is associated with the mother's diet in early pregnancy and with the offspring's adiposity in later childhood.[14] RXRA is a member of the nuclear hormone superfamily and forms heterodimers with a number of nuclear receptors including the vitamin D receptor, thyroid hormone receptor (TR), glucocorticoid receptor (GR), and peroxisome proliferator-activated receptor (PPAR), which are known to influence bone metabolism.[15, 16] Heterodimerization with RXRA is critical for both DNA binding and transactivation activity of these receptors. RXRA is also known to play a role in fetal development and in the epigenetic regulation of vitamin D activation.[17] Given these critical roles of RXRA in pathways that may influence bone mineral accrual, together with our previously demonstrated associations between perinatal RXRA promoter methylation and offspring body composition, we reasoned that variation in perinatal epigenetic marking of the RXRA promoter region might be associated with differences in offspring bone size and density in childhood. The aim of this study, therefore, was to investigate the relationships between offspring bone mineral accrual and epigenetic marking in the RXRA promoter region in umbilical cord tissue, at sites previously found to be associated with childhood adiposity. We also investigated whether maternal 25(OH)-vitamin D concentrations might relate to RXRA methylation.

Subjects and Methods

Participants

The Southampton Women's Survey (SWS) is a prospective cohort study of 12,583 women aged 20 to 34 years recruited from the general population.[18] Assessments of lifestyle and anthropometry were performed at study entry and then in early (11 weeks) and late (34 weeks) gestation in SWS women who became pregnant. Maternal height was measured with a stadiometer, weight with calibrated digital scales, and skin folds (biceps, triceps, and subscapular and suprailiac regions) with Harpenden calipers (Baty International, Burgess Hill, UK). The research nurses carrying out the measurements underwent regular assessment and retraining during the study to optimize consistency. The women were asked to characterize their current walking speed into one of five groups (very slow, stroll at an easy pace, normal speed, fairly brisk, or fast). The women's own birth weight was recorded (by recall, checked by asking her own parents). Then 25(OH)-vitamin D and vitamin D binding protein (DBP) concentrations were measured in serum at 34 weeks gestation (Diasorin RIA, Stillwater, MN, USA). A subset of 900 offspring aged 4 years was recruited sequentially from the SWS cohort. The mother (or father/guardian) and child were invited to visit the Osteoporosis Centre at Southampton General Hospital for assessment. At this visit written informed consent for a dual-energy X-ray absorptiometry (DXA) scan was obtained from the mother or father/guardian. The child's height (using a Leicester height measurer; Seca Ltd, Birmingham, UK) and weight (in underpants only, using calibrated digital scales; Seca Ltd) were measured. A whole-body DXA scan was obtained, using a Hologic Discovery instrument (Hologic Inc., Bedford, MA, USA) in pediatric scan mode. To encourage compliance, a sheet with appropriate colored cartoons was laid on the couch first; to help reduce movement artifact, the children were shown a suitable DVD cartoon. The total radiation dose for the scans was 4.7 µSv for whole-body measurement (pediatric scan mode). Thirty-two scans were found to have unacceptable movement artifact so were excluded. The manufacturer's coefficient of variation (CV) for the instrument was 0.75% for whole-body bone mineral density (BMD), and the experimental CV when a spine phantom was repeatedly scanned in the same position 16 times was 0.68%. We studied 230 children selected as having both umbilical cord DNA and DXA bone measurements available at age 4 years.

In the Princess Anne Hospital study, white women ≥16 years old with singleton pregnancies of <17 weeks gestation were recruited; diabetics and hormonally-induced conceptions were excluded. When the children approached 9 years, we wrote to 461 still living locally; 216 (47%) attended a clinic and adiposity was measured using DXA (Lunar DPX-L; GE Lunar, Madison, WI, USA); 64 of these had DNA available from an umbilical cord sample collected at birth and stored at −80°C.

Follow-up of the children and sample collection/analysis was carried out under Institutional Review Board approval (Southampton and SW Hampshire Research Ethics Committee) with written informed consent. Investigations were conducted according to the principles expressed in the Declaration of Helsinki.

Quantitative DNA methylation analysis

The methods used to select the region of interest and measure RXRA methylation have been described.[14] In brief, using a commercial tiled oligomer microarray (NimbleGen Systems HG17_min_promoter array, using 50-mer oligonucleotides positioned around the transcription start site of 24,134 human genes), we focused initially on 78 candidate genes, and chose five genes for further study based on individual oligomers showing evidence of correlation with DXA measurements at age 9 years, biological plausibility, and feasibility of designing amplicons suitable for Sequenom analysis. Of the five genes chosen, RXRA showed associations between umbilical cord DNA methylation levels and the child's later adiposity in two independent cohorts.[14] Given the common lineage of adipocytes and osteoblasts, we reasoned that such epigenetic marks might also be related to bone development. Genomic DNA was isolated from frozen archived umbilical cord tissue by classical proteinase K digestion and phenol:chloroform extraction. Quantitative analysis of DNA methylation was carried out using the Sequenom MassARRAY Compact System (http://www.sequenom.com), following bisulphate conversion. These methods have been described in detail.[19] Chromosomal coordinates are based on the University of California, Santa Cruz (UCSC) human genome browser database, March 2006 assembly (hg18).

Electrophoretic mobility shift assays

Nuclear extracts were prepared from the human MCF-7 cell line using the method described by Dignam and colleagues.[20] The protein concentration of nuclear extracts was determined with the bicinchoninic acid (BCA) protein assay kit (Pierce, Thermo Scientific Inc, Rockford, IL, USA) using the manufacturer's instructions. For electrophoretic mobility shift assay (EMSA), double-stranded DNA oligonucleotides (biomers.net) labeled with biotin at the 5′ termini of both strands were used. The sense strand sequence for RXRA CpG8 was 5′-CCTTCTCTCTGCAGGCCGCTGCTCAGCC-3′; single-stranded oligonucleotide probes were annealed by heating equimolar amounts of complementary strands to 95°C for 5 minutes and slowly cooling the reaction mixture to room temperature. EMSAs were performed using LightShift Chemiluminescent EMSA Kit (Thermo Scientific Inc). All binding reactions were carried out in the presence of 1 µg of the nonspecific DNA Poly (dI-dC), 20 fmol of biotin-labeled probe, 1× binding buffer, 2.5% glycerol, 0.05% NP-40, and 5 mM MgClg (20 µL final volume), incubated on ice for 20 minutes. For reactions carried out in the presence of nuclear extract, 5 µg total protein was used. Competition was performed in the presence of 500-fold excess of unlabelled oligonucleotide incubating on ice for 30 minutes prior to addition of labeled probe. Complexes were resolved on pre-run 4% nondenaturing polyacrylamide gel in 0.5× Tris-borate-EDTA for 45 minutes at 100 V followed by semidry transfer (200 mA, 1 hour) to a positive nylon membrane and UV cross-linking. After incubation in blocking buffer for 15 minutes at room temperature the membrane was incubated with streptavidin–horseradish peroxidase (HRP) conjugate for 15 minutes. The membrane was then washed and visualized with a chemiluminescent substrate (Thermo Scientific Inc).

Statistical analysis

The bone outcomes at 4 years were whole body minus head (for simplicity, denoted whole-body or WB) bone area (BA), bone mineral content (BMC), areal bone mineral density (aBMD), and size-corrected bone mineral content (scBMC). scBMC is BMC adjusted for BA, height, and weight (to minimize the effect of body size).[21] Furthermore, we explored the effect of body composition by adjusting BMC for total lean, total fat, and height in an additional model. In the absence of maternal albumin measurements, we modified the method of Bikle and colleagues[22] to estimate free 25(OH)-vitamin D levels by calculating the ratio of serum 25(OH)-vitamin D concentration to that of vitamin D binding protein. Continuous maternal and child characteristics were summarized by mean (SD) or median (interquartile range [IQR]) depending on normality. Categorical variables were summarized by percentages. Differences in the continuous variables between boys and girls were tested using t tests and Wilcoxon rank-sum tests where appropriate. Linear regression analysis was performed to explore associations between bone outcomes and the epigenetic measurements. For the purpose of this analysis bone outcomes at 4 years were adjusted for sex and the epigenetic variables were transformed to normality using a Fisher-Yates transformation. Analyses were performed separately in the discovery and replication cohorts, and then combined using meta-analysis. All statistical analysis was carried out using Stata 12.1 (Statacorp, College Station, TX, USA).

Results

Characteristics of the subjects

There were 230 SWS mother-baby pairs (124 boys) with Sequenom and DXA data. Mean ± SD mother's age at delivery was 30.4 ± 3.6 years; 50% were in their first pregnancy and 11% smoked in late pregnancy. Table 1 summarizes the characteristics of the mothers, and Table 2 summarizes those of the children. The girls had greater bone area and BMC than the boys, thus the outcomes were adjusted for sex of the child. Compared with mothers of children born to the SWS during the same timeframe, but who did not have DXA scans at 4 years, the mothers of children who did have DXA assessments were, on average, slightly older at the birth of their child (mean age 31.2 versus 30.6 years, respectively; p = 0.007), better educated (24.8% versus 20.6% with higher degree, respectively; p = 0.002), and smoked less (8.5% versus 17.4% smoked before pregnancy, respectively; p < 0.001).

Table 1. Characteristics of the SWS Mothers
 n (%)Mean/medianSD/IQR
  1. Values are n (%), median (IQR), or mean (SD).
  2. SWS = Southampton Women's Survey; IQR = interquartile range; 25(OH)-vitamin D = 25-hydroxyvitamin D; CSE = certificate of secondary education; O levels = ordinary level; A levels = advanced level; HND = higher national diploma.
Woman's age at child's birth (years) 30.43.6
Height (cm) 164.26.2
Mid-upper arm circumference at 34 weeks (cm) 29.83.5
Prepregnancy weight (kg) 65.459.4–72.4
Triceps skin-fold at 34 weeks (mm) 19.916.6–26.1
Serum 25(OH)-vitamin D at 34 weeks (nmol/L) 58.539.0–92.0
Free 25(OH)-vitamin D index at 34 weeks (units) 0.10.1–0.2
Smoked at 34 weeks
No197 (88.7)  
Yes25 (11.3)  
Parity
0116 (50.4)  
1+114 (49.6)  
Educational attainment
None1 (0.4)  
CSE18 (7.8)  
O levels68 (29.6)  
A levels58 (25.2)  
HND21 (9.1)  
Degree64 (27.8)  
Walking speed at 34 weeks
Very slow36 (16.2)  
Stroll at an easy pace115 (51.8)  
Normal speed60 (27)  
Fairly brisk10 (4.5)  
Fast1 (0.5)  
Table 2. Characteristics of SWS Children
 Boys (n = 124)Girls (n = 106)p diff
Median/meanIQR/SDMedian/meanIQR/SD
  1. Values are median (IQR) or mean (SD); DXA measures at whole body minus head site.
  2. SWS = Southampton Women's Survey; IQR = interquartile range; RXRA = retinoid-X receptor-alpha; BA = bone area; BMC = bone mineral content; aBMD = areal bone mineral density; scBMC = size-corrected bone mineral content; DXA = dual-energy X-ray absorptiometry.
At birth
Gestational age (weeks)39.938.9–40.740.439.4–41.00.02
RXRA CpG4/5 (chr9:136355593,600+)42.028.5–59.547.036.0–60.00.11
RXRA CpG6 (chr9:136355688+)65.047.5–80.563.551.0–77.00.86
RXRA CpG7 (chr9:136355836+)56.044.0–77.057.546.0–72.00.75
RXRA CpG8 (chr9:136355885+)65.051.0–82.066.055.0–82.00.98
At 4 years old
Age (years)4.14.1–4.24.14.1–4.20.44
BA (cm2)752.442.9774.946.5<0.001
BMC (g)375.641.0388.141.90.02
aBMD (g/cm2)0.50.030.50.030.71
scBMC (g)375.817.3374.717.70.62

RXRA methylation and offspring bone size and density

The region of interest contained 13 CpG dinucleotides; however, five of these could not be assayed due to silent signals or high fragment mass, leaving eight CpG sites included in this analysis. For ease of interpretation, these have been labeled sequentially CpG1 to CpG8 and their locations are shown in Fig. 1. Four sites were clustered in pairs (CpG1/2 and CpG4/5), and were treated as single sites in the analysis, yielding a total of six CpG-site variables. The included RXRA CpG sites demonstrated a wide range of methylation and four were associated with childhood size-corrected bone mineral content (scBMC) at 4 years (Tables 2 and 3). These sites were denoted RXRA CpG4/5 (chr9:136355593,600 +), RXRA CpG6 chr9:136355688 + , RXRA CpG7 (chr9:136355836 +), and RXRA CpG8 (chr9:136355885 +). scBMC at the WB site was negatively and statistically significantly associated with percent methylation at each of these four RXRA sites; these are summarized graphically in Fig. 2. The relationships between RXRA methylation and BMC adjusted for total fat, total lean, and height were of a similar effect-size to those observed with scBMC, but failed to achieve statistical significance, apart from RXRA CpG7. There were no statistically significant relationships between RXRA methylation at these sites and WB BA, BMC, or aBMD, and no associations were observed between any bone outcome and methylation at the other two RXRA sites [CpG1/2 (chr9:136355556+,560) and CpG3 (chr9:136355569+)]. Table 3 summarizes these associations.

Figure 1.

Location of RXRA CpG sites and transcription start site. DMR = differentially methylated region.

Table 3. Relationships Between Methylation at Sites Within RXRA Promoter (Fisher-Yates Z-scores) in Umbilical Cord and Sex-Adjusted Whole-Body DXA Indices in Offspring at 4 Years Old: Whole-Body DXA Measures
 BA, β (cm2/SD)BMC, β (g/SD)aBMD, β (g/cm2 per SD)scBMC, β (g/SD)BMC adjusted for lean, fat, and height, β (g/SD)
  • Values are regression coefficient (95% CI).
  • RXRA = retinoid-X receptor-alpha; DXA = dual-energy X-ray absorptiometry; BA = bone area; BMC = bone mineral content; aBMD = areal bone mineral density; scBMC = size-corrected bone mineral content; RXRA CpG4/5 = (chr9:136355593,600 +); RXRA CpG6 = (chr9:136355688 +); RXRA CpG7 = (chr9:136355836 +); RXRA CpG8 = (chr9:136355885 +); CI = confidence interval.
  • *p < 0.05.
  • **p < 0.01.
RXRA CpG4/5 (SD)2.37 (−3.38, 8.11)0.26 (−5.09, 5.61)−0.0016 (−0.0059, 0.0027)−2.59* (−4.81, −0.36)−2.46 (−5.47, 0.55)
RXRA CpG6 (SD)0.41 (−4.84, 5.66)−0.98 (−5.86, 3.90)−0.0020 (−0.0059, 0.0020)−2.07* (−4.10, −0.031)−2.34 (−5.07, 0.39)
RXRA CpG7 (SD)0.58 (−4.93, 6.08)−1.41 (−6.53, 3.70)−0.0026 (−0.0067, 0.0015)−3.43** (−5.54, −1.33)−3.90** (−6.75, −1.05)
RXRA CpG8 (SD)0.76 (−4.60, 6.12)−0.27 (−5.26, 4.71)−0.0013 (−0.0053, 0.0027)−2.38* (−4.45, −0.31)−2.38 (−5.20, 0.43)
Figure 2.

Percent methylation at umbilical cord RXRA promoter CpG sites and offspring size-corrected BMC (g) at 4 years old. RXRA CpG4/5 (chr9:136355593,600 +); RXRA CpG6: (chr9:136355688 +); RXRA CpG7 (chr9:136355836 +); RXRA CpG8 (chr9:136355885 +).

Maternal influences

Maternal height and weight prepregnancy, and mid-upper arm circumference, smoking and strenuous exercise in late pregnancy, which are factors previously associated with neonatal bone mass, did not predict RXRA methylation at any site (all p > 0.05). In contrast there was a negative, statistically significant relationship between maternal free vitamin D index (ratio of serum 25[OH]-vitamin D to DBP concentrations) measured at 34 weeks gestation and percent methylation at RXRA CpG4/5 (β = −3.3 SD/unit, p = 0.03), shown in Fig. 3. However, serum concentrations of 25(OH)-vitamin D or DBP were not predictive of methylation status at any site.

Figure 3.

Maternal free 25(OH)-vitamin D index (units), and % methylation at RXRA CpG4/5 (chr9:136355593,600 +).

RXRA methylation and birth weight

To investigate whether the associations between RXRA methylation and offspring bone indices might be mediated through an effect on overall size at birth, we examined the relationships between RXRA methylation and birth weight. There was a trend toward a positive association for RXRA CpG4/5 methylation and birth weight (β = 46.7g/SD, p = 0.11), but no relationships were observed for RXRA CpG6, RXRA CpG7, and RXRA CpG8 (p = 0.5, p = 0.7 and p = 0.8 respectively).

Replication study and functional investigation

We examined these relationships in the Princess Anne Hospital cohort,[14] in which we had previously found associations between RXRA methylation and offspring fat mass. In these 64 subjects, the effect sizes and direction of association for two sites (RXRA CpG7 and RXRA CpG8) were similar to those observed in SWS; the associations for RXRA CpG4/5 and RXRA CpG6 were positive, but none of the relationships achieved statistical significance. However, when combined with the SWS results using meta-analysis, consistent, statistically significant, negative associations were observed at RXRA CpG4/5, CpG7, and CpG8. These findings are summarized in Table 4.

Table 4. Results of Replication Study and Meta-analysis: Whole-Body DXA
 scBMC, β (g/SD)
SWSPAHCombined
  • Values are regression coefficient (95% CI).
  • DXA = dual-energy X-ray absorptiometry; scBMC = size-corrected bone mineral content; SWS = Southampton Women's Survey; PAH = Princess Anne Hospital Cohort; RXRA CpG4/5 = (chr9:136355593,600 +); RXRA CpG6 = (chr9:136355688 +); RXRA CpG7 = (chr9:136355836 +); RXRA CpG8 = (chr9:136355885 +); CI = confidence interval.
  • *p < 0.05.
  • **p < 0.01.
RXRA CpG4/5 (SD)−2.59* (−4.81, −0.36)3.55 (−6.24, 13.34)−2.29* (−4.46, −0.12)
RXRA CpG6 (SD)−2.07* (−4.10, −0.031)4.06 (−5.93, 14.05)−1.82 (−3.82, 0.17)
RXRA CpG7 (SD)−3.43** (−5.54, −1.33)−1.90 (−11.72, 7.93)−3.37** (−5.42, −1.31)
RXRA CpG8 (SD)−2.38* (−4.45, −0.31)−1.80 (−12.02, 8.43)−2.36* (−4.39, −0.32)

Finally, we used an EMSA to explore the effect of methylation of the CpG sites within the RXRA promoter on transcription factor binding. A biotin-labeled double-stranded oligonucleotide probe corresponding to a region of the RXRA promoter from −2322 to −2295 (which contains RXRA CpG8) was incubated with nuclear extracts from the MCF-7 cell line. One retarded complex was seen binding to the unmethylated probe, the binding of which was reduced markedly by co-incubation with 500-fold excess of the unlabelled specific competitor. Binding to the methylated probe was also observed, but at a much reduced level.

Discussion

We have demonstrated, to our knowledge for the first time, that alteration of epigenetic marking of specific regions of the RXRA promoter in umbilical cord is associated with childhood scBMC in the offspring. These associations were present for only four of the six CpG sites measured, suggesting possible site specificity of methylation; furthermore, these findings were supported by results from a second independent cohort. Additionally, methylation at one of these sites was negatively associated with the maternal free 25(OH)-vitamin D index measured at 34 weeks of gestation.

We used a prospective cohort with detailed characterization of mothers and children, using the gold standard DXA technique to assess bone mass. There are, however, several limitations to our study. First, methylation analysis was carried out on samples that had been stored for several years, but our local data suggest that DNA methylation is likely to be stable in tissue stored at −80°C, consistent with findings from another study.[23] Moreover, there is no reason to suppose that, if present, sample degradation would have occurred in any but a random distribution across the cohort and therefore should not have led to erroneous associations. Second, the methylation sites studied were upstream from the proximal promoter region, but they were located in a region that has been demonstrated to contain positive regulatory elements of transcription[15] and there are several studies reporting promoter regulation by sites at this distance.[24, 25] We excluded the presence of an SNP at the CpG sites of interest by sequencing, but without genomewide analysis it is not possible to exclude a genetic effect of distant SNPs that could influence both DNA methylation of a particular sequence and a child's phenotype. Third, we analyzed methylation in cells from whole umbilical cord; whereas it is possible that the differential methylation we observed arose from variation in the proportions of different component cells (eg, fibroblasts, epithelial cells) in individual samples, our studies show similar methylation in different umbilical cord cell types (unpublished data). Fourth, measurement of bone mineral in children is hampered by their low absolute BMC. However, we used specific pediatric software, and studies of DXA indices compared to ashed mineral content in piglets have confirmed the accuracy of the technique.[26] Fifth, the study cohort was a subset of the SWS, but mothers whose children underwent DXA scanning and those whose children did not were broadly similar: the former were on average slightly older and smoked slightly less. There is no reason to suppose, however, that relationships between RXRA promoter methylation in umbilical cord and childhood bone mineral accrual would differ between these two groups. Finally, the use of DXA does not allow measurement of true volumetric bone density, thus making it difficult to be certain about differential determinants of skeletal size and volumetric density.

We are not aware of existing data showing that methylation of these CpGs affects RXRA expression. However, the promoter region of RXRA is rich in CpG dinucleotides, and our CpG loci are located in a region that has been shown to be important for RXRA expression: Using an RXRA promoter reporter construct, Li and colleagues[15] demonstrated that, in two different cell lines, the promoter region of RXRA from −2494 to −1500 bp with respect to the transcription start site was important for promoter activity, because deletion of this region led to a significant decrease in luciferase activity. Moreover, they also showed that this same region of RXRA (from −2494 to −1500 bp with respect to the transcription start site) was able to drive heterologous promoter activity. In their detailed description of the RXRA promoter, the same group reported, using the MatInspector program,[27] potential transcription factor binding sites at distances up to 3922 bp upstream from the transcription start site (for example, AP-1 at 196, 2038, and 2241 bp upstream and SRY at 982 and 1763 bp upstream).[15] Finally, we used EMSAs to demonstrate that methylation of CpG8 leads to a reduction in transcription factor binding, suggesting that methylation may have functional consequences for RXRA expression (unpublished data). The 2% to 3% change in methylation we observed across the relationship with bone outcomes is consistent with other studies which have shown that early life environment can induce methylation changes that persist throughout the life course: an example is that of periconceptional famine exposure leading to a 5.2% decrease in methylation of the IGF2 differentially methylated region (DMR)[28] in the offspring. Such changes in DNA methylation are likely to reflect a change in the proportion of cells containing a methylated gene versus an unmethylated gene, with downstream consequences potentially in terms of gene expression and signalling.[29, 30]

Despite its critical role in the nuclear action of a variety of hormones known to influence bone cells, there are scant data linking directly RXRA with bone mineral accrual. We have previously demonstrated an inverse relationship between maternal 25(OH)-vitamin D concentration at 34 weeks gestation and offspring bone mass both at birth[6] and in later childhood.[5] These associations have been confirmed in another mother-offspring cohort.[31] Furthermore, in one study in which umbilical cord calcium concentrations were available, the maternal 25(OH)-vitamin D–childhood bone relationships appeared to be, in part, mediated via placental calcium transfer.[5] Thus maternal 25(OH)-vitamin D concentration might be an important determinant of neonatal bone mass and of the postnatal skeletal growth trajectory.

In the current study we have demonstrated an inverse association between another measure of maternal vitamin D status: the maternal free 25(OH)-vitamin D index (ratio of 25[OH]-vitamin D to DBP concentrations) and methylation at one of the RXRA sites. However, because relationships with 25(OH)-vitamin D (the measure used in our previous work) or DBP individually did not achieve statistical significance, and we were not able to fully calculate free vitamin D index using the method of Bikle and colleagues[22] (because we did not have albumin concentrations), we view this finding as intriguing, but clearly requiring confirmation through further investigation. Indeed, relationships between concentrations of 25(OH)-vitamin D and 1,25-dihydroxyvitamin D (1,25[OH]2-vitamin D) are not straightforward, with the latter regulated tightly by parathyroid hormone and fibroblast growth factor 23 (FGF-23), which maintains a constant plasma calcium × phosphorus product, appropriate for bone mineralization.[32] The role of 25(OH)-vitamin D and 1,25(OH)2-vitamin D in the human fetus is unclear; animal work has suggested that different species have varying requirements to enable optimal fetal skeletal mineralization.[33] Additionally 1,25(OH)2-vitamin D has been shown to regulate an active placenta plasma membrane calcium ATP-ase (PMCA) in animals.[34] Our previous work has suggested that maternal 25(OH)-vitamin D concentration may be an important determinant of offspring bone development into postnatal life in human pregnancies,[5, 6] but to our knowledge there are no previous data to suggest that either form of vitamin D modulates epigenetic marking at the RXRA gene in this context. However, there is evidence that, in adult life, 1,25(OH)2-vitamin D may influence gene expression through epigenetic mechanisms: in animal studies the 1,25(OH)2-vitamin D–vitamin D receptor (VDR)–RXRA complex, in combination with other factors such as DNA methyltransferase enzymes, is able to negatively modulate transcription of the 1-alpha-(OH)-ase gene via methylation.[17] This provides a negative feedback mechanism, with parathyroid hormone reversing the process and causing demethylation-mediated transcriptional derepression. This mechanism clearly represents a continuous regulatory process but shows that epigenetic mechanisms may influence action of 1,25(OH)2-vitamin D. Whether such mechanisms might lead to methylation change originating during fetal development and persisting into older age remains to be elucidated.

Given that RXRA acts as a necessary cofactor for the action of a range of other nuclear receptors, such as TR, GR, and PPARs,[15, 16] there are several other mechanistic possibilities linking RXRA methylation to offspring bone development. All of these hormone systems have been demonstrated to influence bone density in adults; indeed, circulating cortisol profiles have been linked to growth in early life[35] and maternal dietary manipulation in pregnant rats influences methylation of GR[36] and PPAR[30] in the offspring.

We have previously demonstrated that maternal adiposity, smoking, physical activity, and parity all predict offspring bone size and geometry.[3, 4] However, none of these factors were associated with methylation of the RXRA promoter in the current study; indeed, RXRA methylation appeared to be more strongly related to scBMC than bone size, suggesting that its effect may relate to volumetric bone density rather than development of the overall skeletal envelope. We have previously found associations between RXRA methylation and childhood fat mass.[14] In this study, adjustment of BMC for lean, fat, and height yielded associations with RXRA methylation that were similar in magnitude of effect size to those observed with scBMC but with attenuated statistical significance, except for RXRA CpG7 in which the p value was very similar. These findings are consistent with the associations not being mediated purely through an effect on fat or lean mass and with the possibility of a direct relationship between RXRA methylation and bone mineral accrual. Whatever the underlying mechanism, our results clearly demonstrate that alteration of epigenetic marking in utero is associated with bone outcomes in the offspring, an observation which, if replicated, might lead to a novel biomarker-based prediction of adverse bone outcomes in children.

In conclusion, we have demonstrated that site-specific changes in epigenetic marking within the RXRA promoter in umbilical cord are associated with offspring scBMC in later childhood. The potential mechanistic and functional significance of this finding remains a subject for further investigation.

Disclosures

All authors state that they have no conflicts of interest.

Acknowledgments

This work was supported by grants from the Medical Research Council, British Heart Foundation, Arthritis Research UK, National Osteoporosis Society, International Osteoporosis Foundation, Cohen Trust, the NIHR Southampton Biomedical Research Centre, the National Research Centre for Growth and Development (New Zealand), and NIHR Musculoskeletal Biomedical Research Unit, University of Oxford. We thank the mothers and their children who gave us their time, and a team of dedicated research nurses and ancillary staff for their assistance. Participants were drawn from a cohort study funded by the Medical Research Council and the Dunhill Medical Trust. We thank Mrs G Strange and Mrs Ruth Fifield for helping to prepare the manuscript.

Authors' roles: All authors contributed to manuscript authorship and preparation. NH oversaw the statistical analysis, wrote the draft manuscript, and coordinated manuscript preparation. Laboratory analyses were performed by AS, CM, EG, and RM, and supervised by KL and AS. CC, KG, MH, PG, and KL were responsible for the design and conception of the study. LD and GN undertook the statistical analyses. CC, HMI, and KMG designed and implemented the Southampton Women's Survey. CC oversaw the project and is its guarantor.

Ancillary