Low BMD is a major risk factor for spine and proximal femur fractures [13, 14]. In women, BMD in adulthood is largely determined by the amount of bone accumulated at the end of their skeletal growth (peak bone mass), their rate of bone loss after menopause when ovaries cease producing estrogens, and age-related bone loss. It has been well established with the study of twins, that peak bone mass is highly heritable with an estimated heritability between 0.50 and 0.80 . Conversely, published data on the heritability of bone loss at menopause are conflicting [16–18]. Therefore, BMD is a trait that lends itself to studies designed to identify the genes underlying its normal variation [16, 17, 19].
The past decade has seen an important increase in the use of association studies with candidate genes for the genetic analysis of complex traits such as BMD and/or fracture risk. Many genes have been examined for their association with normal BMD variation, which yields an ever-expanding candidate gene list. However, this approach has been largely criticized because of discrepancy in the results [20, 21], often related to the small size of the enrolled cohorts. Moreover, most of the studies focused on postmenopausal female populations. In this view, the main purpose of the present study was to evaluate allelic influence of target genes, such as estrogen receptors, on inherited skeletal traits in a large and homogeneous population-based cohort of premenopausal healthy Caucasian women .
For ESR1 and ESR2, two genes worldwide evaluated by independent research groups, the results obtained even if compelling for their involvement in BMD, osteoporosis, or fracture, are, however, not conclusive . Confounding factors encompassed ethnic-specific distribution of ESR1 and ESR2 polymorphisms [8, 23, 24]. For example, in the SWAN study  which enrolled 693 Caucasian participants (366 premenopausal women), specific associations of BMD with ESR1 and ESR2 genotypes varied according to race/ethnicity. Furthermore, 4 independent studies  concluded that ESR2 locus could be involved in FN-BMD in Caucasians, LS-BMD in Japanese postmenopausal women, and LS- and FN-BMD in Chinese premenopausal women. In addition, most of the human studies of genetic association with BMD have been cross-sectional, and only very few studies examined the association of the genotypes to BMD change within specific age ranges. For all these reasons the present study was aiming to compare the results obtained for ESR1 and ESR2 to other data obtained in age-equivalent studies in Caucasian women, examining the relation of ESR1 and ESR2 genes' polymorphisms.
Differently than for the ESR2 locus , allelic variants of ESR1 gene were proposed to affect skeletal growth, through a genotype-dependent estrogen sensitivity at the growth cartilage, with the ESR1 px haplotype being less sensitive to estrogen effects . The ESR1 haplotype effect was supported by functional studies [27, 28] and by multiple association analysis documented for this gene variant [29–33]. For example, body Ht in pre- and postmenopausal women  and estradiol levels in premenopausal women were lower , with the number of copies of ESR1 px haplotype in their genotype. Lorentzon et al.  found an association between reduced Ht and PvuII T and XbaI A alleles, which corresponded to the ESR1 px haplotype. Although this study was performed in adolescent boys , it is in line with other findings in adult women. In 607 Caucasian women (aged 55–80 yrs) in whom vertebral fractures were excluded, Schuit et al.  observed significant association between Ht and ESR1 PvuII-XbaI haplotypes. In contrast to , a significant allele dose effect was observed for ESR1 px haplotype, corresponding to a 0.9-cm decrease in Ht per allele copy (P for trend = 0.02), extreme genotypes varied 1.8 cm. Boot et al.  partially confirmed this allele dose-effect to some extent, as in girls heterozygous for ESR1 px haplotype the Ht was higher than in those homozygous for the ESR1 px haplotype. In our series, higher Ht was slightly (P > 0.05) correlated with ESR1 CC, GG or CCGG, and ESR2 AA genotypes, while ESR1 CCGG plus ESR2 AA-AG genotype was significantly (2.5-cm) taller than the opposite genotype. As ESR2 modulates ESR1 transcriptional activity , this novel biological interaction between ESR2 and ESR1 genotypes is not surprising.
Family history is a major risk factor for osteoporotic fractures . In white postmenopausal women, increased BMD-independent risk for vertebral (but not non-vertebral) fractures was found in ESR1 px haplotype carriers . Moreover, the GENOMOS Consortium found a BMD-independent protective effect against vertebral fractures in ESR1 XX homozygous individuals, while no effects on fracture risk were seen for ESR1 PvuII polymorphism . Similarly to the InCHIANTI study , we could not demonstrate any strong association between FHF and ESR1 rs2234693 and rs9340799 genotypes. However, our study might have had not enough power to detect any differences.
Variants of ESR2 gene, alone and in interaction with ESR1 genotypes influenced the fracture risk in postmenopausal women. Moron et al.  suggested that ESR2 rs4986938 (but not ESR1 rs2234693) could have a role (P = 0.04) in osteoporosis in Spanish postmenopausal women. Furthermore, they detected a joint effect of ESR1 gene in osteoporosis modulating the penetrance of ESR2 rs4986938 genotype . Rivadeneira et al.  showed for the first time that white postmenopausal women (≥ 55 yrs of age) who are homozygous for a common intron 2–3'UTR ESR2 haplotype allele have 40–80% increased risk of fragility and vertebral fracture. Interestingly, we also observed ESR2 rs4986938 genotypes significantly correlated with FHF risk but not with FHO, suggesting ESR2 variants may affect bone strength independently of BMD.
According to our findings, McGuigan et al.  observed a modest association between ESR1 PvuII genotypes and BMD at the hip (P = 0.034) but not at the spine in 216 young Irish women (mean age 22.6 ± 1.6 yrs), with no differences regarding the ESR1 XbaI locus . On the other hand, Valero et al.  found no significant relations between FN- or LS-BMD with both ESR1 PvuII and XbaI loci in 194 older Caucasian women aged 22–45 yrs. Furthermore, a cross sectional study of XbaI and BMD in women who were premenopausal and perimenopausal, did not confirm this association . Finally, in perimenopausal Caucasian women (older than 48.5 yrs) enrolled in the GENOMOS consortium, none of two ESR1 intron 1 polymorphisms (i.e. PvuII and XbaI loci) or derived haplotypes had any statistically significant effect on BMD, with estimated differences between genetic contrasts being 0.01 g/cm2 or less . Collectively, our findings and the published studies [22, 39–41] make possible to support a significant effects of the ESR1 rs2234693 (but not rs9340799) locus on the BMD mainly in the young adult next to her achievement of bone peak mass.
Previous approaches have also suggested the role of ESR2 in BMD within different ethnic backgrounds . No association between ESR2 rs4986938 with LS- or FN-BMD were detected in 1291 Caucasian women (from 192 families) aged 33.2 ± 7.1 yrs (range 20–50 yrs) . Similarly, no associations between ESR2 rs4986938 genotypes and Ht, LS-BMD and serum OC levels were detected in 147 healthy peri and postmenopausal Greek women (mean age 54 ± 7.9 yrs) . On the other hand, we detected significant BMD variations of the ESR2 rs4986938 genotypes only in the later age group (i.e. 41–50 yrs old women). Together with ESR1 rs2234693 data, this reinforces the hypothesis that ESR1 and ESR2 genes affect bone metabolism in precise and distinct age-sequential windows. Larger pre-planned analysis will be necessary to confirm our interpretation.
In conclusion, taken together, our findings indicated that, although the effect size may be small, allelic variations in ESR1 and ESR2 genes are associated with various and different bone traits (e.g. Ht, BMD and FHF risk) in normal premenopausal Caucasian subject. Furthermore, multiple genotype interactions were detected that reinforced the polygenic and complex character of skeletal system. In some cases however, the mean pattern of bone trait values for a gene polymorphism with evidence of association was not in agreement with previously published studies. Therefore, even though family history of fragility fractures is one of the risk factors , we cannot recommend genetic testing for clinical use in humans to better identify population at risk for pathologic bone traits such as fragility fractures. However, as it has been shown for other diseases , extended panels of several polymorphic markers could be used in the future, in addition to traditional risk factors, to evaluate the skeletal disorder risk in humans.