Maternal serum retinol, 25(OH)D and 1,25(OH)2D concentrations during pregnancy and peak bone mass and trabecular bone score in adult offspring at 26-year follow-up

Background Vitamin A and D deficiency is prevalent in pregnant women worldwide. Both vitamins are involved in fetal skeletal development. A positive association between maternal vitamin D levels and offspring bone mineral density (BMD) at adulthood has been observed. The impact of maternal vitamin A status in pregnancy on offspring peak bone mass remains unclear. Method and findings Forty-one mother-child pairs were recruited from a population-based prospective cohort study in Trondheim, Norway, where pregnant women were followed from gestational week 17. Their term-born infants were followed from birth (1986–88). Regression analyses were performed for vitamin A (retinol), 25-hydroxyvitamin D [25(OH)D] and 1,25-dihydroxyvitamin D [1,25(OH)2D] in maternal serum (gestational weeks 17, 33, 37) and cord blood. Offspring BMD and spine trabecular bone score (TBS), a measure of bone quality, were analyzed by dual x-ray absorptiometry at 26 years. Average levels during pregnancy of retinol, 25(OH)D and 1,25(OH)2D were 1.66 (0.32) μmol/L, 59.0 (20.6) nmol/L, and 251.3 (62.4) pmol/L, respectively. 1,25(OH)2D levels were similar in those with 25(OH)D levels <30 and >75 nmol/L. After adjustment for maternal age, BMI, smoking, and education, and offspring birth weight, maternal serum retinol was positively associated with offspring spine BMD [mean change 30.8 (CI 7.6, 54.0) mg/cm2 per 0.2 μmol/L retinol], and with offspring TBS, although non-significant (p = 0.08). No associations were found between maternal 25(OH)D and 1,25(OH)2D levels and offspring bone parameters. Vitamin levels in cord blood were not associated with offspring BMD or TBS. Conclusions This is the first study to show an association between maternal vitamin A status and offspring peak bone mass. Our findings may imply increase future risk for osteoporotic fracture in offspring of mothers with suboptimal vitamin A level. No associations were observed between 25(OH)D and 1,25(OH)2D and offspring BMD.


Introduction
Increasing evidence suggests that the in-utero environment plays an important role in the development of future osteoporosis [1,2]. Vitamin A and D are both important for bone health, and antenatal levels of these vitamins may exert critical influence on this process. Globally, about 20 million pregnant women are considered to be vitamin A deficient [3,4]. Maternal vitamin D deficiency during pregnancy is a worldwide epidemic with a reported prevalence from 18-84% [5][6][7].
In the diet, vitamin A is obtained as retinyl ester or β-carotene, which are metabolized to retinol in the intestine [8]. In the circulation, retinol is incorporated in chylomicrons or bound to retinol binding protein and transthyretin. The liver is the main storage site for vitamin A, but a substantial amount is also transported to other tissues by chylomicrons, bone being the second most important organ for clearance of chylomicron remnants [8,9]. All-trans retinoic acid, the biologically active form, binds to retinoic acid receptors (RARs) which heterodimerize with retinoid X receptors (RXRs) [8,10,11]. Vitamin A has been shown to promote bone formation and to inhibit generation of osteoclast progenitors [11]. The effects are, however, dependent on dose, and both hypo-and hypervitaminosis A may be harmful for the skeleton. Studies in rats show that vitamin A is essential for growth and normal skeletal development in the fetus, whereas excess retinol may have negative skeletal effects [12]. An inverse association between maternal serum retinol in late pregnancy and offspring total bone mineral content (BMC) was reported in neonates within 2 weeks after birth, whereas β-carotene was positively associated [13]. The association between maternal serum retinol during pregnancy and offspring bone health in adulthood has not been addressed.
Serum 25-hydroxyvitamin D [25(OH)D] is a measure of vitamin D status, and a precursor of the active form 1,25-dihydroxyvitamin D 3 [1,25(OH) 2 D] which binds to the vitamin D receptor (VDR). Like RAR, VDR is dependent on heterodimerizing with RXR [14]. The primary action of vitamin D is to facilitate intestinal calcium absorption [15]. Severe vitamin D deficiency leads to rickets in children and osteomalacia in adults [15]. During pregnancy, a 2-3 time rise in 1,25(OH) 2 D concentration occurs to optimize calcium absorption and mineralization of the fetal skeleton [16,17].
Data on the association between maternal 25(OH)D concentrations during pregnancy and offspring bone health are conflicting. Zhu et al. reported a positive association between maternal serum vitamin D level and offspring peak bone mass [18]. This was supported by a study showing that maternal 25(OH)D associated positively to offspring BMC at 9 years of age [19]. In contrast, two larger studies found no relationship between maternal serum 25(OH)D and offspring bone health at 6 and 9 years [20,21]. No studies have evaluated the relation between maternal 1,25(OH) 2 D concentrations and adult offspring bone health.
Antagonizing effects of vitamin A and D have been reported in several studies [11,22]. High levels of vitamin D seem to protect against vitamin A toxicity, whereas high vitamin A levels reduce the adverse effects of hypervitaminosis D [23]. High intake of vitamin A and concomitant low intake of vitamin D may enhance bone fragility [23][24][25]. It has been postulated that these opposing effects may be attributed to that both RAR and VDR heterodimerize with the RXR receptor [11,26]. Thus, high levels of retinol could attenuate binding of the heterodimerized receptors VDR RXR to responsive elements of DNA, thereby impairing the actions of vitamin D.
We aimed to examine the association between maternal serum levels of retinol, 25(OH)D, and 1,25(OH) 2 D during second and third trimester and offspring BMD and trabecular bone score (TBS) at the age of peak bone mass (26 years).

Study design and participants
Mother-child pairs were recruited from a Caucasian-population-based prospective cohort study in Trondheim, Norway [27]. Pregnant women were followed from gestational week 17. Their term-born infants were followed up from birth. Mother-child pairs were included based on the access to maternal serum samples and offspring dual x-ray absorptiometry (DXA) scans. At inclusion, offspring age was 26 years. Exclusion criteria were cerebral palsy and pregnancy. The study was approved by the Regional Committee for Medical and Health Research Ethics in Central Norway (No: 2013/636/REK Mid-Norway). We obtained informed written consent from all participants.

Statistical analyses
Continuous variables are presented as mean and standard deviation (SD) or median and interquartile range (IQR) dependent on the data distribution. The student's t-test or nonparametric test were applied. Differences in frequencies of categorical variables were analyzed by the Pearson chi-square test. Multivariate linear regression was used to examine associations between maternal serum retinol, 25(OH)D and 1,25(OH) 2 D, and offspring bone parameters. Retinol, 25(OH)D, and 1,25(OH) 2 D were treated as continuous exposure variables. A composite exposure variable was generated for each maternal serum parameter by averaging the concentration across three gestational time-points (weeks 17, 33 and 37). Adjustments were made for maternal age, preconception BMI, smoking and education, and offspring birth weight. These potential confounders were chosen based on previous studies [18][19][20], and as they were assumed to be associated both with exposure and outcome. Additional adjustment for offspring covariates (sex, BMI, smoking, physical activity and retinol and 25(OH)D levels) did not affect the results and were therefore not included in the analyses. Statistical analyses were conducted using SPSS statistics version 22�0 (IBM, Chicago, IL).

Results
Forty-one mother-child pairs were included. Maternal data are presented in Tables 1 and 2.
Mean age at delivery was 29.6 (4.3) years and mean pre-conception BMI 21.3 (3.1) kg/m 2 . Seventy-three % of the women smoked at conception, and 51% smoked during pregnancy.  Offspring (n = 41) data are given in Table 3. Among the offspring, 16 had birth weight below the 10 th percentile, and 25 above the 10 th percentile. Mean birth weight was 3410 (542) g. The age at inclusion was 26.1 (0.6) years; 61% (n = 25) were males. Mean BMI was 25.2 (5.0) kg/m 2 . None had vitamin A insufficiency, three had levels above the upper reference level (>2.8 μmol/L). Vitamin D insufficiency (25(OH)D <50 nmol/L) was observed in 51% (n = 21). The data collected on vitamin D intake were incomplete and did not allow estimation of daily intake. The median calcium intake from milk was 85 (52.5-184.0) mg/day.
Mean BMD Z-scores were -0.505 (1.02), -0.071 (0.98), 0.066 (0.87), and -0.420 (0.94), at the lumbar spine, femoral neck, total hip and whole body, respectively. Spine TBS was 1.42 (0.10). BMD and TBS did not differ significantly between participants with birth weight below and above the 10 th percentile (spine BMD: 1100 (70) and 1130 (90) mg/cm 2 , respectively, p = 0.28; TBS 1.40 (0.11) vs 1.43 (0.10), respectively, p = 0.40). Offspring spine BMD was significantly higher in those with maternal retinol levels above 1.54 μmol/L (mean level in week 37) compared to those with levels below [1040 (120) vs 960 (80) mg/cm 2 , p = 0.023] Five subjects reported allergy, and three of them used antihistamine drugs; one had asthma and used glucocorticoid inhalation. Three had depression, of whom two were treated with antidepressants (escitalopram and sertraline). One participant had hypothyroidism and was adequately treated with levothyroxine. One subject with a history of eating disorder had normal BMI and BMD at inclusion. Another subject successfully treated for lymphoma also displayed satisfactory BMD. Additionally, four reported to have migraine and one chronic pain. Nine were currently using hormonal contraceptives and six were former users.

Associations between maternal vitamin A and D levels and offspring bone parameters
After adjustment for maternal confounders and birth weight, offspring spine BMD and Zscore were positively associated with average retinol level across three gestational time-points, and increased by 30.8 (CI 7.6, 54.0) mg/cm 2 and 0.24 (CI 0.03, 0.46) SD, respectively per 0.2 μmol/L increment in maternal retinol (Table 4). A positive, but non-significant (p = 0.08) association was also seen between maternal retinol levels and offspring TBS after adjustment (Table 4). Maternal retinol concentration at week 17 was positively associated with spine and whole-body BMD (S1 Table). Positive associations were also observed between retinol concentration at week 37 and offspring spine and total hip BMD (S3 Table). No associations were observed between average maternal 25(OH)D and 1,25(OH) 2 D levels and offspring BMD or TBS (Table 4). Retinol in cord blood was not associated with offspring BMD or TBS (spine, p = 0.77; femoral neck, p = 0.23; total hip, p = 0.

Discussion
This is the first study to demonstrate a relationship between vitamin A status in pregnant women and offspring peak bone mass. After adjustment for maternal confounders and birth weight, we observed a significant positive association between average maternal retinol levels during second and third trimester and offspring spine BMD at 26 years of age. Offspring spine BMD Z-score increased by 0.24 SD per 0.2 μmol/L increment in maternal retinol. This is of clinical significance, as a BMD increase of 1 SD translates to a 2-3 times fracture reduction [28]. Retinol concentration at week 17 was also associated with offspring spine and whole body BMD and at week 37 with spine and total hip BMD. No associations were found between maternal 25(OH)D and 1,25(OH) 2 D levels and offspring bone parameters. TBS is a measure of microarchitecture and bone quality, which has been shown to predict occurrence of major fractures independent of BMD [29]. We observed a positive association between average maternal retinol concentration and offspring TBS, albeit, not significant (p = 0.08). Impairment of bone quality in addition to lower peak bone mass in offspring of mothers with insufficient vitamin A status may enhance future fracture risk. In contrast to a study by Hyde et al. [30], no relation between maternal 25(OH)D concentration and TBS was found.
We show that low maternal vitamin A status may affect bone adversely and influence achievement of an optimal peak bone mass. This is of concern, as peak bone mass is regarded as the most important determinant of future fracture risk [31]. Our findings comply with studies in rodents demonstrating that retinol is mandatory for normal development of fetal bone [12]. Retinol plays a crucial role at an early stage in embryogenesis in patterning the entire axial skeleton and for development of other skeletal structures at a later stage [32]. The second trimester is a critical period for long bone growth [1], where vitamin A is an important actor, both deficiency and excess vitamin A causing inhibition of longitudinal growth [33]. In line with this, we observed a significant association between maternal retinol levels in week 17 and offspring BMD. Vitamin A deficiency in late embryogenic stage has been shown to affect development of the axial and appendicular skeleton in rats [12]. Accordingly, the significant association seen in week 37 in the present study, may indicate a role for vitamin A in skeletal development during the third trimester. As reviewed by Delgado-Calle et al., epidemiological and experimental studies suggest that epigenetic mechanisms influence skeletal development and the risk of osteoporosis [34]. Vitamin A is a potent epigenetic modulator that may produce long-term effects on the phenotype [35]. So far, there are no studies addressing epigenetic effects of vitamin A on the skeleton. Vitamin A is also involved in post-natal maintenance of bone [36]. In vitro studies suggest that retinol stimulates osteoblast differentiation and transition of osteoblasts to osteocytes and inhibits osteoclastogenesis, thus promoting bone formation [36][37][38].
Maternal vitamin D inadequacy may impair mineralization and thus affect the fetal skeleton directly [16,17]. Moreover, skeletal effects may be attributed to epigenetic mechanisms, as vitamin D interacts with the epigenome at multiple levels [39,40]. In contrast to Zhu et al.
[18], we did not observe any association between maternal 25(OH)D and offspring BMD. There are no reports on the relation between maternal 1,25(OH) 2 D level and offspring BMD. In the present study, no association was found. In line with previous studies, maternal 1,25 (OH) 2 D levels increased from second to third trimester. This rise occurs to facilitate calcium absorption and ensure mineralization of the fetal skeleton [16, 17]. We found that circulating 1,25(OH) 2 D was similar across different levels of 25(OH)D, illustrating that compensatory mechanisms may uphold serum 1,25(OH) 2 D levels in a state of maternal hypovitaminosis D [41]. Since vitamin D exerts its actions through 1,25(OH) 2 D, this could explain the lack of association between maternal 25(OH)D and offspring BMD. In a recent study including 855 pregnant women, we observed that 1,25(OH) 2 D level was significantly lower in women with 25(OH)D level < 30 nmol/L compared to those with level above 75 nmol/L [7]. Hollis et al. suggested that 25(OH)D levels of 100 nmol/L are necessary to obtain maximal 1,25(OH) 2 D concentrations during pregnancy [17]. The fact that none of our participants reached this level combined with the small sample size, may have attenuated the possibility to show associations.
In line with previous studies [13,42], we observed a modest decline in retinol levels throughout pregnancy, which may be attributed to hemodilution, and depletion of maternal stores due to fetal demands. The fact that the majority of women were smokers could contribute to the inferior vitamin A status, as smoking seems to attenuate serum levels of beta-carotene [43]. Unfortunately, we do not have data on alterations in smoking status during pregnancy.
The optimal retinol concentration during pregnancy is not settled. Concentrations below 1.05 μmol/L are regarded as inadequate in the non-pregnant state. According to this criterion, five (12.5%) women displayed inadequacy in last trimester. In a meta-analysis addressing vitamin A and fracture risk, retinol levels between 1.99 and 2.31 μmol/L were suggested to be optimal [44]. A U-shaped relationship between serum retinol levels and hip fracture risk was observed, indicating that both high and low levels may be harmful [44]. Several studies have reported negative skeletal effects, however, mainly in those with retinol levels > 2.64 μmol/L [45]. In contrast, in a Norwegian study by Holvik et al. [46], hip fracture risk tended to be increased at low retinol concentrations, whereas high levels (up to 3 μmol/L) showed no adverse effect on hip fracture risk. In the present study, most of the women had retinol levels below 1.99 μmol/L, and about 40% had levels below the reference level. When stratifying offspring BMD according to mean maternal retinol level in week 37 (1.54 μmol/L), we observed a significantly higher spine BMD in offspring whose mothers had concentrations above this level.
Since vitamin A has to be obtained from the diet, it may be challenging to reach satisfactory levels [35]. Data from clinical trials indicate that a vegetarian diet alone is not sufficient to achieve an adequate vitamin A status [47]. A Cochrane review concluded that taking vitamin A supplements during pregnancy reduced the risk of anemia, infection and night blindness in the mother [47]. However, no reduction in deaths of mothers or newborns were seen [48]. The recommended intake during pregnancy differs between countries [49], and the adherence to these recommendations is low [50]. WHO advocates vitamin A supplementation only to pregnant women in regions with severe deficiency to prevent night blindness [3]. According to McCauley et al., the basal requirement during pregnancy is 370 μg/day and the recommended daily allowance 770 μg/dayhttps://www.ncbi.nlm.nih.gov/pubmed/26027509 [48]. It is not specified whether vitamin A supplements should be given as retinol ester or beta-carotene. In a study from rural Nepal, a positive effect on pulmonary function was observed in the offspring at the age of 9-13 years, in those whose mothers received supplements with preformed vitamin A during pregnancy. In children whose mothers were supplemented with beta-carotene, no beneficial effects on pulmonary function were observed [51,52].
In the present study, almost 50% of the women exhibited vitamin D insufficiency at both second and third trimester. The occurrence of deficiency increased from 11.1 to 17.5% from week 17 to week 37. This is in accordance with previous studies showing a high prevalence of hypovitaminosis D during pregnancy [16,17]. Infants whose mothers are vitamin D deficient may be born with hypocalcemia and some develop rickets and craniotabes [53]. The prenatal recommendations vary between countries, and the adherence seems to be poor [54].
Interference between vitamin A and D at receptor level could also affect fetal bone negatively. High levels of retinol combined with vitamin D deficiency could hinder binding of VDR to the heterodimer RAR-RXR and thus block the effects of vitamin D [55]. In the present study, none of the mothers with vitamin D deficiency had hypervitaminosis A, thus interaction between these vitamins is not likely to explain the association between maternal levels and offspring bone health.
The main limitation to the study is the small sample size, which makes it vulnerable to type 2 error, and to a lesser extent to type 1 error. The offspring comprised individuals born at term with low or normal birth weight. At the age of 26 years, they did, however, not differ significantly in BMI, BMD and TBS. We had access to maternal data that allowed adjustment for several confounding factors. However, information on potential confounding factors such as parity and dietary intake of vitamin A and D were lacking. Moreover, we did not have data on vitamin A and D intake in the offspring, and the calcium intake was underestimated as only milk consumption was recorded. The study participants were white, and the findings may thus not be applicable to other ethnic groups.
The study has several notable strengths, including the long-term follow-up. This enabled the assessment of offspring bone health at the age of peak bone mass, thus reducing the influence of factors as growth rate and developmental differences. Application of TBS for assessment of bone quality gave additional information concerning fracture risk. In contrast to most previous studies, serum levels of vitamin A and D were analyzed at several time-points across the last two trimesters of pregnancy and in cord blood at delivery, thus allowing assessment of at which stage of fetal skeletal development these vitamins may have the highest impact. Vitamin A levels preconception and in first trimester could have added further insight.

Conclusion
Maternal retinol concentration during mid and late pregnancy was positively associated with offspring peak bone mass and bone quality. This may imply increase future fracture risk in offspring of mothers with inadequate vitamin A status. Our study contributes to novel knowledge on developmental origins of osteoporosis. Given the high prevalence of hypovitaminosis A worldwide, there is a need for increased attention to ensure sufficient intake during pregnancy. No associations between maternal 25(OH)D and 1,25(OH) 2 D concentrations and offspring bone health were shown. Studies with larger study populations are warranted to confirm our data.
Supporting information S1 Table. Associations