Genetic factors are important in the pathogenesis of osteoporosis and the estrogen receptor has been suggested as a possible candidate gene for regulation of bone mineral density (BMD). We investigated the relationship between PvuII, XbaI, and dinucleotide (TA)n repeat polymorphisms of the estrogen receptor α (ER-α) gene and BMD in a study of women from northeast Scotland in the United Kingdom. No significant association was observed between BMD values at the lumbar spine (LS) and femoral neck (FN) in relation to PvuII and XbaI polymorphisms individually, but haplotype analysis showed that BMD values (Z score) were significantly lower in those who carried the Px haplotype (n = 36) compared with those who did not (n = 170) at both the LS (mean ± SEM; −0.775 ± 0.125 vs. −0.285 ± 0.082; p = 0.002) and the FN (−0.888 ± 0.130 vs. −0.335 ± 0.083; p = 0.0006). In keeping with this, the Px haplotype also was found to be an independent predictor of LS BMD (p = 0.019) and FN BMD (p = 0.005) in a multiple regression analysis model that included other possible predictors of BMD including age, years since menopause (YSM), hormone-replacement therapy (HRT) use, weight, and height. This model explained 15.7% and 23.4% of the total observed variance in LS and FN BMD, respectively, with the Px haplotype accounting for ∼3% of the variance at both sites. Although the TA repeat polymorphism was in strong linkage disequilibrium (LD) with the PvuII (χ2 = 109.8; p < 0.0001) and XbaI (χ2 = 97.2; p < 0.0001) polymorphisms, there was no overall association between TA repeat number and BMD. We conclude that polymorphisms of the ER-α gene are significantly related to BMD in our population and that this association is dependent on the Px haplotype, suggesting that it is the Px haplotype, or a linked polymorphism, that confers risk.
OSTEOPOROTIC FRACTURES are an important healthcare problem in developed countries(1,2) and several studies have shown that reduced bone mineral density (BMD) is an important predictor of osteoporotic fracture risk.(3–5) BMD is a complex trait that is determined by an interaction of both genetic and environmental factors. Studies in twins have indicated that genetic factors play an important role in regulation of BMD(6–8) whereas segregation analysis in normal families was consistent with polygenic influences on BMD.(9) Genetic linkage studies(10,11) and candidate-gene association studies(12,13) have identified several loci and candidate genes that appear to be involved in the regulation of bone mass and the pathogenesis of osteoporotic fractures. One such candidate is the estrogen receptor α (ER-α) gene. The importance of ER-α in the regulation of bone mass is indicated by the occurrence of osteoporosis in a man with a coding region mutation of the ER-α gene(14) and by the observation that mice lacking a functional ER-α gene have 20–25% lower BMD values than wild-type mice.(15) Several investigators have reported associations between PvuII and XbaI intronic polymorphisms of the ER-α gene and bone mass(16–18) and between a (TA)n repeat polymorphism of the ER-α promoter and bone mass.(19,20) To further define the role of ER-α in the regulation of bone mass and to determine if there is a relationship between the intronic and promoter polymorphisms, we investigated the association between BMD and haplotypes defined by the PvuII, XbaI, and (TA)n repeat polymorphisms at the ER-α locus in a study of women from northeast Scotland (UK).
MATERIALS AND METHODS
The study group was comprised of a white population from northeast Scotland (Aberdeen). A series of 220 community-dwelling women were recruited at random from the local population postal register as part of a European collaborative study investigating the prevalence of osteoporotic fractures.(21) Individuals with diseases known to affect bone metabolism (corticosteroid use, primary hyperparathyroidism, pituitary disease, immobilization, neoplasia, and thyrotoxicosis) were excluded from the study, leaving a final study population of 206 women, of whom 92.7% (n = 191) were postmenopausal, as defined by the absence of menstruation for 6 months. The mean age (±SEM) of the study group was 64.4 ± 0.7 years.
BMD at the lumbar spine (LS; L2-L4) and femoral neck (FN) was measured in each participant by dual-energy X-ray absorptiometry (DXA) using a Norland XR-26 densitometer (Norland Corp., Fort Atkinson, WI, USA). The precision of the technique in our hands is 0.9% for the LS and 2.3% for the FN. Z score results were calculated by relating the measured BMD values to a locally acquired normal reference range.(22)
Analysis of variance (ANOVA) was used to evaluate the potential association between phenotypic characteristics and genotype groups defined by the PvuII and XbaI restriction fragment length polymorphisms (RFLPs) and Student's t-test was used when two groups were compared. Analysis of BMD and other phenotypic characteristics in relation to the (TA)n repeat polymorphism was performed using ANOVA after grouping by the number of repeats. Individuals were then coded according to whether they had two copies [1,1], one copy [1,0], or no copies [0,0] of a particular TA allele group. Stepwise multiple linear regression was used to test for independent predictors for BMD entering genotypes, age, years since menopause (YSM), weight, duration of hormone-replacement therapy (HRT) use, and height into the model. Linkage disequilibrium (LD) between the polymorphisms was analyzed using the estimate of haplotype (EH) algorithm.(23) Direct haplotyping was possible for the PvuII and XbaI polymorphisms, and here LD was assessed using standard χ2 test comparing the expected and actual haplotype frequencies. All statistical analyses were performed using Minitab version 12.1 (Minitab, Inc., Coventry, UK). Analysis of potential regulatory elements surrounding the PvuII and XbaI polymorphisms was performed using the Signal Scan program(24) (Human Genome Mapping Project Resource Center, Hinxton, UK).
Genomic DNA analysis
DNA was extracted from peripheral blood leukocytes using a kit according to the manufacturer's instructions (Nucleon II DNA extraction kit; Scotlab, Coatbridge, UK). A 1.3-kilobase (kb) fragment of genomic DNA containing the PvuII and XbaI polymorphisms in intron 1 of the ER-α gene was amplified by polymerase chain reaction (PCR) using previously described oligonucleotide primers and amplification conditions.(16) After amplification, the PCR products were digested with PvuII and XbaI restriction endonucleases (Promega, Southampton, UK) and separated by 1.5% agarose gel electrophoresis. The genotypes were represented as Pp (PvuII) and Xx (XbaI), with uppercase letters signifying the absence of and lowercase letters the presence of the restriction site. Haplotypes for the individuals carrying the PpXx genotype were determined by double digestion with PvuII and XbaI. The dinucleotide repeat polymorphism of the ER-α gene was investigated by PCR using primers designed to amplify the (TA)n repeat, which is situated approximately 1.2 kb upstream of exon 1 in the human ER-α gene.(19) The length of TA repeats was determined by analysis of PCR products on an ABI377 DNA sequencer using Genescan and Genotyper software (Applied Biosystems, Perkin Elmer, Warrington, UK). All PCR reactions were performed in a 25-μl volume containing 100 ng genomic DNA using an Omnigene thermal cycler (Hybaid Ltd., Teddington, UK).
Data on frequency of the PvuII and XbaI polymorphisms in relation to BMD and other relevant clinical and anthropometric variables are shown in Table 1. Genotype distributions were in Hardy-Weinberg equilibrium and similar to those previously reported in white women.(18,25,26) There was no significant difference in any variable between the different genotype groups, although there were trends for individuals with the xx genotype to have lower LS and FN BMD values than women with the Xx and XX genotypes and for individuals with the PP genotype to have an earlier menopause. Moreover, individual PvuII and XbaI genotypes did not predict BMD at both LS and FN in a multiple linear regression model including age, height, weight, YSM, and HRT use (data not shown). We then analyzed the relationship between genotypes defined by the combination of PvuII and XbaI polymorphisms in relation to BMD and other variables. The frequencies of the genotype combinations were PpXx, 76 (36.9%); ppxx, 62 (30.1%); Ppxx, 26 (12.6%); PPXX, 27 (13.1%); PPXx, 8 (3.9%); ppXx, 5 (2.4%); and PPxx, 2 (1.0%). Analysis of BMD data in these subgroups showed lower BMD values (Z score) in the combined PPxx and Ppxx groups (LS = −0.8901 ± 0.1286; FN = −1.0243 ± 0.1535; n = 28) compared with the other groups (LS = −0.2890 ± 0.0793; p = 0.0002; FN = −0.3390 ± 0.0795; p = 0.0003; n = 178). In view of this, we analyzed the relationship between haplotypes defined by the PvuII and XbaI polymorphisms in relation to BMD. The frequencies of haplotypes were px, 231 (56.1%); PX, 138 (33.5%); Px, 38 (9.2%); and pX, 5 (1.2%), indicating that the PvuII and XbaI polymorphisms are in strong LD (χ2 = 195.5; p < 0.0001, EH algorithm). The PX haplotype (138/412) is present two times more frequently than expected (61/412) and the least common haplotype pX (5/412) is present 16 times less frequently than expected (82/412). The predicted coefficient of gametic LD between the P allele and X allele (DPX) as calculated by the EH algorithm was 0.1854, which is very similar to the result obtained when DPX was calculated using actual haplotype frequencies (DPX = 0.1868).
Table Table 1.. Characteristics of the PvuII and XbaI Genotypes
We next investigated, in a multiple linear regression, the relation between each of the three frequent haplotypes (px, PX, and Px) and BMD. For each haplotype, subjects were coded according to whether they had two copies [1,1], one copy [1,0], or no copies [0,0] of the haplotype under investigation. The haplotype copy number was then entered in a multiple linear regression model including age, weight, height, YSM, and duration of HRT use. From all haplotypes tested, only the Px haplotype predicted BMD at both LS and FN ( Table 2, data for px and PX are not shown). The model shown in Table 2 explained 15.7% and 23.4% of the total observed variance in LS and FN BMD, respectively. However, age, weight, and Px haplotype copy number were the most significant independent predictors of both LS and FN BMD. Duration of HRT use was of borderline significance for LS (p = 0.053) but not FN (p = 0.099) BMD. Age accounted for the majority of the total observed variance in LS (∼10%) and FN (∼16%) BMD. The Px haplotype accounted for approximately 3% at both LS and FN.
Table Table 2.. Multiple Linear Regression Analysis of BMD
Analysis of the relationship between Px haplotype and bone density by ANOVA showed that BMD values were significantly lower in those with one [1,0] or two [1,1] copies of the Px haplotype, as compared with those who did not carry the Px haplotype [0,0] (Table 3). There was a trend toward an earlier menopause and fewer HRT users among those with the Px haplotype but this was not statistically significant and no other variables differed between the groups.
Table Table 3.. Characteristics of Each Genotype Classified According to the Number of Copies of the Px Haplotype
Next, we investigated the association between the dinucleotide (TA)n repeat polymorphism lying upstream of the human ER-α gene and BMD at both LS and FN in the same study subjects. The most common number of TA repeats was 15 with a range of values between 12 and 30 (Fig. 1). There were 64 different genotypes in the population studied and the most common one was the genotype (15/15; n = 32). The allele frequencies showed a bimodal distribution with a breakpoint occurring at the median number of 20 TA repeats (Fig. 1). In view of the large number of individual genotypes, subjects were classified into three broad groups: those who carried two alleles with >20 TA repeats (designated [1,1]), those who carried one allele with >20 TA repeats (designated [1,0]), and those who carried two alleles with ≤20 TA repeats (designated [0,0]). There was no significant difference in BMD values or other variables among the three different TA genotype groups using ANOVA nor was TA genotype related to LS or FN BMD in a multiple linear regression analysis (data not shown). Further analysis showed that a small group of subjects with the highest number of TA repeats (i.e., having at least one allele TA ≥ 26; n = 15) had lower BMD values at the spine than those with fewer TA repeats (i.e., having both alleles TA < 26; n = 191) (Z score = −0.743 ± 0.149 vs. −0.341 ± 0.077, respectively; p = 0.025) and similar findings were observed at the FN (−0.887 ± 0.185 vs. −0.397 ± 0.078; p = 0.024). No other variables differed between the low (<26) and high (≥26) TA repeat groups (data not shown), but there was significant overrepresentation of the Px haplotype in individuals who had high numbers of TA repeats (χ2 = 9.6; p = 0.002). In view of this, we evaluated the possibility of LD between PvuII, XbaI, and (TA)n repeat polymorphisms in the ER-α gene. The PvuII and XbaI polymorphisms are 45 base pairs (bp) apart (our own sequence data) and located in intron 1, approximately 400 bp upstream from exon 2 whereas the (TA)n repeats are located 1.2 kb upstream of exon 1, approximately 21 kb upstream from the PvuII and XbaI polymorphisms. The average number of TA repeats (average of allele 1 and 2 in each subject) was significantly related to PvuII and XbaI genotype and to Px haplotype (Fig. 2). Strong LD also was observed when data were analyzed using EH algorithm (for TA-PvuII, χ2 = 109.8 and p < 0.0001; TA-XbaI, χ2 = 97.2 and p < 0.0001; TA-PvuII-XbaI, χ2 = 356.1 and p < 0.0001). Taken together, these data show that the PvuII, XbaI, and (TA)n repeat polymorphisms are in LD and that individuals with low bone density carry the Px haplotype in association with having high numbers of TA repeats. To determine whether the association with BMD was driven primarily by the Px haplotype or the TA repeat alleles, we compared BMD values in individuals who were subgrouped based on TA repeat alleles in relation to Px status (Fig. 3A), and based on Px haplotype in relation to TA allele status (Fig. 3B). This showed no significant difference in BMD values when the groups were divided by TA allele status but a highly significant difference when the groups were divided by Px status.
The ER-α gene is a strong candidate for the regulation of bone density but studies to date have yielded conflicting results. The first reported investigation of ER-α polymorphisms in relation to BMD was that of Sano et al. who found low BMD values and increased bone turnover in a subgroup of Japanese women who had 12 TA repeats.(19) Subsequent work by Kobayashi and colleagues showed no significant difference in BMD values in Japanese women in relation to PvuII and XbaI polymorphisms when they were studied individually but a strong association between the Px haplotype and reduced bone mass.(16) Subsequent studies of ER-α genotypes in relation to BMD have yielded mixed results. Studies on Italian and Korean populations failed to find any significant association between the ER genotypes and LS BMD in postmenopausal women, but these were performed in selected clinic populations rather than on random population surveys.(26,27) Mizunuma et al. reported an association between BMD and the XbaI polymorphism in premenopausal Japanese women but did not observe an association with the PvuII polymorphism.(17) Willing and colleagues also reported a significant association between XbaI and PvuII polymorphism and BMD in a population-based sample of US perimenopausal women from the Michigan Bone Health Study.(18) In a separate study of the same population, Sowers and colleagues found that the (TA)n repeat polymorphism was associated with total body BMD as well as BMD at the spine and hip.(20) Although the association with total body BMD remained after adjusting for confounding factors, the association with hip and spine BMD was attenuated because of earlier menopause as the result of an overrepresentation of hysterectomy and ovariectomy in relation to the number of (TA)n repeats. However, in the study mentioned previously, women with a low average number of TA repeats had reduced BMD, which is the opposite of that reported here. The data on the association between ER genotypes and menopausal age are of interest in relation to the recent report by Weel and colleagues, who found a strong association between carriage of the PvuII “P ” allele and early menopause in the Rotterdam study, particularly when menopause was induced surgically.(25)
The present study of ER-α genotypes in relation to BMD is the first to be performed in UK women and also the first to conduct direct haplotype analysis taking into account information from the PvuII, XbaI, and (TA)n repeat polymorphisms. In agreement with the findings of Kobayashi(16) we found no significant association between BMD values and the ER-α PvuII and XbaI polymorphisms when they were analyzed individually but found a strong association when subjects were categorized by Px haplotype. The differences in BMD were not explained by differences in confounding factors, such as menopausal age, although like Weel et al.,(25) we found that individuals with the PvuII “P” allele tended to have an earlier menopause than those without this allele.
We also found evidence of an association between alleles with ≥26 repeats and BMD but this was in a small subgroup of individuals and was driven primarily by a higher frequency of the Px haplotype in the high (TA)n repeat group. Thus, when individuals were subcategorized separately by TA repeat alleles (≥26 or <26) or by presence or absence of the Px haplotype, the association was significant only with respect to Px haplotype, which was in strong LD with the (TA)n repeat polymorphism. The observation of LD between these polymorphisms, which are approximately 21 kilobases (kb) apart, is in itself of interest, in light of recent theoretical studies that suggested that LD may only extend over short distances (5 kb) in outbred populations.(28)
The mechanisms by which these polymorphisms are associated with BMD remain unclear. Although previous studies have shown that variable number tandem repeat (VNTR) polymorphisms in proximity to some gene promoters can have a significant influence on transcriptional regulation,(29) this seems an unlikely explanation for the findings reported here, because the association was driven by the Px haplotype rather than the TA genotype. Intragenic regulatory elements also may play a significant role in gene regulation.(30) The first intron of the estrogen receptor is not known to play a role in gene regulation, but analysis of potential regulatory elements surrounding the PvuII and XbaI polymorphisms showed that the “P” allele of PvuII polymorphism (—CAGCCG—, polymorphic site underlined) disrupts a potential recognition site for the transcription factor AP4 (recognition sequence = CAGCTG), raising the possibility that this polymorphism could influence gene regulation. Another explanation is that the PvuII and XbaI polymorphisms may be in LD with causal polymorphisms elsewhere in the ER-α gene or in an adjacent gene. Further work using reporter constructs and gel shift assays along with DNA sequence analysis will be required to investigate these possibilities.
In summary, we conclude that polymorphisms of the ER-α gene are significantly related to BMD in our population and that this association is driven primarily by the Px haplotype, which concurs with the findings originally reported by Kobayashi.(16) Although we also found an association between BMD and the TA repeat polymorphism in agreement with the findings of others,(18,19) this was explained by LD with the Px haplotype, suggesting that it is the Px haplotype or a linked polymorphism that confers risk.
We are grateful to Mrs. Rita Smith (Research Nurse) for assistance with the collection of samples for this study. This work was supported financially by The National Osteoporosis Society (UK) in the form of a Ph.D. studentship (O.M.E.A.), by an MRC cooperative group grant, and by an Integrated Clinical Arthritis Centre (ICAC) grant from the Arthritis Research Campaign.