Introduction

Osteoporosis and obesity are two common complex diseases and major health problems. It is also known that obesity and osteoporosis are closely related diseases1. Human body composition generally changes with aging, mainly including a reduction in lean body mass, an accumulation of fat body mass, and a loss of bone mass2. These changes may correspondingly lead to osteoporosis and obesity. Both bone mass and obesity phenotypes are known to be under strong genetic regulation3, 4. In the past decade, a number of candidate genes have been identified that may contribute to bone mineral density (BMD), but far less is known about genes affecting obesity phenotypes such as lean mass and fat mass5, 6. Recently, Tang et al7 conducted a bivariate whole-genome linkage scan and identified several genomic regions shared by obesity and osteoporosis.

The vitamin D receptor (VDR) is one of the most extensively studied genes in relation to BMD, due to the important role of vitamin D in bone metabolism8. A functional genetic polymorphism in VDR could be involved in other tissues that respond to vitamin D, such as muscle cells and adipocytes. Indeed, it has been demonstrated that vitamin D stimulates the differentiation of preadipocytes to adipocytes in the OB 17 cell line9. Some studies have also suggested that VDR genotype may function as a determinant of body composition3.

Estrogen receptor alpha (ESR1) is also an important potential candidate gene for osteoporosis, and extensive studies have been performed on the relationship between polymorphisms of this gene and BMD variation. The ESR1 gene is known to be involved in metabolic pathways influencing body growth, which may correlate with body mass index (BMI)10, 11. One study demonstrated the association of ESR1 polymorphisms with body fat distribution in Japanese women12. Another study in Caucasians found PvuII polymorphisms within the ESR1 gene to be associated with BMI, with the PvuII genotype giving rise to the highest BMI values in postmenopausal women13.

All of the above data indicate that the VDR and ESR1 genes could be pleiotropic genetic factors influencing both osteoporosis and obesity phenotypes. However, until now, the clear relationship between polymorphisms in the VDR and ESR1 genes with peak BMD and obesity phenotypes have not been elucidated. In this study, we investigated three VDR loci and two ESR1 loci with peak BMD, BMI, total fat mass and total lean mass in Chinese male nuclear families. Dual-energy X-ray absorptiometry (DXA) is a precise, accurate, non-invasive, safe, and convenient technique, found on a three compartment model separating the body into total body mineral mass, fat mass, and lean mass 14, 15. Most association studies use traditional association approaches in random population, and such study designs are prone to population stratification/admixture, which produces false positive/negative results. In addition, the linkage approach often lacks statistical power with the currently used sample. However, the transmission disequilibrium test (TDT), a family-based association approach, is immune to population stratification, much more powerful compared with the traditional linkage approach, and can be used in nuclear families with or without parental phenotypes. Thus, in this study, we used quantitative transmission disequilibrium tests (QTDTs) to determine whether VDR and ESR1 polymorphisms were associated with peak BMD and obesity phenotypes in a relatively large sample of Chinese male nuclear families.

Materials and methods

Subjects

All subjects involved in the study were collected by the Department of Osteoporosis of Shanghai Jiao Tong University Affiliated Sixth People's Hospital from the local population of Shanghai City (located on the mid-east coast of China) and signed informed consent documents before entering the project. Between 2004 and 2007, we recruited 1296 individuals from 427 male nuclear families whose offspring were sons. Of these, samples from 15 individuals could not be amplified and discriminated genotypes due to the poor quality of the DNA, and 12 sons deviated from Mendelian inheritance. Thus, there were a total of 400 male nuclear families composed of both parents and at least one healthy male child (1215 individuals) whose ages were largely between 20 to 40 years old. The average family size was 3.03; 385 families had one child and 15 families had 2. For each study subject, we also collected information on age, sex, medical history, family history, marital status, physical activity, alcohol use, diet habits and smoking history. The recruited sons were healthy. The exclusion criteria for the study subjects were a history of: (1) serious residual effects of cerebral vascular disease; (2) diabetes mellitus, except for easily controlled, non–insulin-dependent diabetes mellitus (defined as adult asymptomatic hyperglycemia controlled by diet or oral agents); (3) chronic renal disease manifested by a serum creatinine level of 11.9 mg/dl; (4) chronic liver disease or alcoholism; (5) chronic lung disease; (6) 12 weeks of corticosteroid therapy at pharmacologic levels; (7) 16 months of treatment with anticonvulsant therapy; (8) evidence of other metabolic or inherited bone diseases (eg, hyper- or hypoparathyroidism, Paget's disease of bone, osteomalacia, or osteogenesis imperfecta); (9) rheumatoid arthritis or collagen disease; (10) major gastrointestinal disease (eg, peptic ulcer, malabsorption, chronic ulcerative colitis, regional enteritis, or any significant chronic diarrhea state); (11) significant disease of any endocrine organ that would affect bone mass (eg, diabetes, hyperthyroidism, etc); (12) any neurologic or musculoskeletal condition that would be a nongenetic cause of low bone mass; and (13) any disease, treatment, or condition that would be a nongenetic cause of low bone mass16.

Phenotype measurements

The BMD (g/cm2) of the lumbar spine and left proximal femur including lumber spine 1–4 (L1-4), femoral neck, total hip, total fat mass(kg), and total lean mass (kg) were measured by a Lunar Prodigy DXA densitometer (Lunar Corp, Madison, WI), and data were analyzed by Prodigy encore software (ver.6.70, standard-array, mode). The percentage of fat mass (PFM) was calculated as the ratio of fat mass to body weight17. The DXA scanner was on fan-beam mode. The machine was calibrated daily, and the coefficient of variability (CV) values of the DXA measurements (which were obtained from 15 individuals repeatedly measured three times) were: for lumbar spine, 1.39%; for femoral neck, 2.22%; for total hip, 0.70%; and for trochanter, 1.41% 18. For body composition, the CVs were 1.18% and 3.72% for total lean mass and fat mass, respectively. The long-term reproducibility of our DXA data during the trial, based on weekly repeated phantom measurements, was 0.45%4. Body mass index (BMI) was defined as the weight/height2 in units of kg/m2.

Genotyping

Genomic DNA was isolated using the phenol-chloroform extraction method. Genotypes were detected using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) and allele-specific multiple PCR (ASM-PCR). The primers are shown in Table 1. A 265 bp fragment containing the FokI (rs10735810) polymorphism in the start codon of the VDR gene was amplified by PCR19. The FokI genotypes were identified by electrophoresis of the DNA samples in 1.5% agarose gels. The FokI genotypes were named as follows: FF (absence of the restriction site); ff (presence of the restriction site); Ff (heterozygous for the restriction site). A 740 bp fragment containing the ApaI (rs7975232) polymorphism near the 3′-end of the VDR gene was amplified by PCR20. The ApaI genotypes were named as follows: AA (absence of the restriction site); aa (presence of the restriction site); Aa (heterozygous for the restriction site). The CDX2 (rs11568820) polymorphism was determined using allele-specific multiple PCR (ASM-PCR) according to the method established by Fang et al21. The position of the CDX2 polymorphism is located in the promoter region of the VDR gene. Two sets of primers were designed for the ASM-PCR tests: These four primers generate three PCR fragments: primer set G-upstream and G-downstream specifically amplifies the G allele with a size of 110 bp; A-upstream and A-downstream specifically amplify the A-allele with a size of 235 bp; and the out-primer pair (G-upstream and A-downstream) amplifies the internal control PCR fragment with a size of 297 bp. Therefore, the GG genotype produced 297 bp and 110 bp fragments, the AA genotype produced 297 bp and 235 bp fragments, and the AG genotype produced 297 bp, 235 bp and 110 bp fragments22. PCR products were size-separated on a 2.5% agarose gel. A 1.3 kb fragment containing the PvuII (rs2234693) and XbaI (rs9340799) polymorphisms in intron 1 of the ESR1 gene was amplified by PCR23. The PCR products were digested with PvuII and XbaI, respectively, and separated by 2% agarose gel electrophoresis with ethidium bromide staining. The genotypes are represented as PP, Pp, and pp for PvuII and XX, Xx, and xx for XbaI24.

Table 1 Information of the analyzed SNPs in this study.
Full size table

Statistical analyses

To test the population homogeneity of the study subjects, the genotype frequencies for each of the five polymorphic sites were tested against the Hardy-Weinberg equilibrium by the χ2 test. To ensure unrelated individual samples, only genotype data from parents of each nuclear family were used in the statistical analysis25. The heritability estimates were done using the linear regression of parents' mean value and offspring's value for every phenotype (described at www.heritability.com). Statistical power was estimated using Piface Software (version 1.65) (http://www.math.uiowa.edu/˜rlenth/Power/) in our current sample size, according to the MAF(minor allele frequency ) of every genotype and the variation of BMD and obesity genotypes. The QTDT program, using the orthogonal model, was used to test for population stratification, total association, linkage and within-family association between SNPs and haplotypes and BMD phenotypes, BMI, fat mass and lean mass. The QTDT software package is available at: http://www.sph.umich.edu/csg/abecasis/QTDT/. This method, as implemented in the QTDT software4, 26, 27, extends the trio-based TDT to quantitative trait data and uses genotypes data from available sibling and parents. Because in our nuclear families, all of the children were sons and the effects of parents' phenotypes were excluded in the QTDTs, sex was not used as a covariate to adjust the sons' bone phenotype variation4, 25. Of course, raw BMD values were adjusted by age, height and weight as covariates. The BMI, fat mass and lean mass were adjusted by age as a covariate. Because false-positive results can be generated in multiple tests (as in the present study) permutations (1000 simulations) were performed to generate the empirical p values to assess the reliability of the results28, 29, 30, 31. The QTDT program generates P values for various tests via asymptotic χ2 distribution. P<0.05 was considered significant for all analyses.

In unrelated sons, differences in BMD among the genotype and haplotype groups were tested using one-way ANOVA and general linear model-ANOVA (GLM-ANOVA), adjusting for confounding variables such as age, height and weight. Differences in BMI, fat mass, lean mass and PFM among the genotypes and haplotypes groups were determined using GLM-ANOVA, adjusting for age. Statistical analyses were performed using the SPSS package, version 11.5 (SPSS, Chicago, IL, USA).

Results

There were 400 nuclear families with 1215 individuals in this study, including 800 parents and 415 sons. The general characteristics of male nuclear families are shown in Table 2, and the basic characteristics of the sons are shown in Table 3. All of the subjects were genotyped at the ApaI, FokI, CDX2, PvuII, and XbaI polymorphic sites. Pearson correlation analysis confirmed that both fat mass and lean mass were positively correlated with BMD. According to the Pearson's coefficient (r), lean mass was more strongly related to BMD (with correlations range from 0.389 to 0.508), and fat mass was related to BMD (with correlations range from 0.198 to 0.243). To determine the relative contributions of lean mass and fat mass to BMD at various sites, we performed logistic regression analysis. In agreement with the Pearson analysis, both lean mass and fat mass showed positive regression coefficients, but lean mass had a significantly greater effect on BMD than fat mass. The r2 between lean mass and BMD ranged from 0.149 to 0.249 and was from 0.056 to 0.068 between fat mass and BMD. The distribution of all of the genotypes was in Hardy-Weinberg equilibrium. The linkage disequilibrium (LD) between PvuII and XbaI was not very strong (D'=0.684, r2=0.256). For the VDR gene, the LDs between ApaI, FokI, CDX2 were all weak (D'<0.1, r2<0.01). Peak BMD is thought to be under strong genetic control. In our sample, the heritability estimates for peak BMD in the spine, femoral neck, and total hip were 0.565, 0.702, and 0.693, respectively.

Table 2 Basic characteristics of the subjects (mean±SD).
Full size table
Table 3 The bacic characteristics of the sons (n=415).
Full size table

We investigated the association between every genotype and BMD, BMI, fat mass and lean mass in 400 unrelated sons (randomly selected from 415 sons) using ANOVA. There was a significant association between CDX2 genotypes and BMI (P=0.006), and adjusted BMI values were 5.4% higher in the AA genotype compared to the AG genotype (P=0.002). A significant association between CDX2 genotypes and fat mass (P=0.004) was found, and adjusted fat mass values were 18.8% higher in the AA genotype compared to the AG genotype (P=0.002). There was a significant association between CDX2 genotypes and FPM (P=0.002), and adjusted PFM values were 14.8% higher in the AA genotype compared to the AG genotype (P=0.001). However, no significant association was found between CDX2 genotypes and peak BMD. In addition, marginally significant association was found between the PvuII genotypes and BMD at the total hip (P=0.049). Moreover, a significant association between PvuII genotypes and fat mass was observed (P=0.022); adjusted fat mass values were 14.3% higher in the pp genotype compared to the Pp genotype (P=0.006). A significant association between the PvuII genotypes and PFM was also observed (P=0.020), and adjusted PFM values were 10.0% higher in the pp genotype compared to the Pp genotype (P=0.006). (Tables 4 and 5) Analysis of these SNPs showed that genetic variations in CDX2 explained 2.40%, 3.08%, and 3.40% of the variation in BMI, fat mass, and PFM, respectively; while PvuII explained 1.86%, 2.18%, and 2.12% of population variance in BMI, fat mass, and PFM, respectively.

Table 4 Association of the three SNPs of VDR gene with phenotypic values in sons from one offsping (mean±SD, n=400). Values for BMD are raw values (means±SD); P values are results of analysis of variance (ANOVA) test for the least square mean of BMD among various genotypes after adjusting for significant covariates of age, weight, and height. P values are results of ANOVA test for the lease square mean of BMI, fat mass, lean mass among various genotypes after adjusting for significant covariates of age. Bold indicates significant P values (P<0.05). cP<0.01 vs AG genotype. fP<0.01 vs GG genotype..
Full size table
Table 5 Association of the two SNPs of ESR1 gene with phenotypic values in sons from one offsping (mean±SD, n=400). Values for BMD are raw values (means±SD); P values are results of analysis of variance (ANOVA) test for the least square mean of BMD among various genotypes after adjusting for significant covariates of age, weight, and height. P values are results of ANOVA test for the lease square mean of BMI, fat mass, lean mass among various genotypes after adjusting for significant covariates of age. Bold indicates significant P values (P<0.05). bP<0.05 vs pp genotype, cP<0.01 vs pp genotype. eP<0.05, fP<0.01 vs 2.
Full size table

Association between VDR and ESR1 genotypes and peak BMD were performed using the QTDT program for nuclear families. The MAF of five SNPs was >30%, and they had high heterozygosis in our population, so we obtained 290, 309, 296, 283, and 246 informative nuclear families for the QTDT analysis for the ApaI, FokI, CDX2, PvuII, and XbaI genotypes, respectively. In this study, using 400 nuclear families has more than 80% of the power to test a candidate gene as a QTL, which can explain about 10% of the BMD or obesity phenotypes' variation. The results of the QTDT analysis are presented in Table 6. There was no population stratification for the genotypes at any sites of BMD. For the total association, PvuII was associated with the variation of the lumbar spine, femoral neck and total hip BMD (P=0.008, P=0.006, and P=0.016, respectively). XbaI genotypes were associated with the variation of the lumbar spine BMD (P=0.040). For the within-family association, we did not obtain significant association evidence at any bone sites between VDR or ESR1 genotypes and peak BMD at any site.

Table 6 P value of tests for population stratification and within-family association between VDR, ESR1 genotypes and peak BMD. BMD values were adjusted by significant covariate effects of age, height and weight. Bold indicates significant P values (P<0.05).
Full size table

The results of the associations between the five genotypes and obesity phenotypes in the QTDT analyses are presented in Table 7. There was population stratification between XbaI and fat mass (P=0.008). For the total associations, CDX2 and PvuII were associated with fat mass (P=0.001 and P=0.009, respectively) and PFM (P=0.003 and P=0.047, respectively). For the within-family associations, CDX2 genotypes were associated with BMI (P=0.046), fat mass (P=0.004), and PFM (P=0.020), and PvuII genotypes were associated with the variations of fat mass and PFM (P=0.002 and P=0.039, respectively). Considering multiple parameters were tested, 1000 permutation tests were performed, and subsequent permutations were in agreement with these significant within-family association results.

Table 7 P value of tests for population stratification and within-family association between VDR, ESR1 genotypes and obesity phenotypes; BMI, total fat mass, total lean mass and PFM values were adjusted by age. Bold indicates significant P values (P<0.05).
Full size table

There were eight different haplotypes of the VDR gene and four haplotypes of the ESR1 gene. The frequencies of the haplotypes for all parents are shown in Table 8. We further observed the associations between VDR and ESR1 haplotypes and peak BMD using QTDTs. There was no significant population stratification, total association, or within-family association between the VDR or ESR1 haplotypes and BMD. We also observed associations between VDR haplotypes and BMI, fat mass, lean mass, and PFM using QTDTs. For haplotype 1 (aFG) and fat mass, there was population stratification (P=0.008), significant total association (P=0.022), and within-family association (P=0.001 and P=0.010 of 1000 permutation tests). For haplotype 1 (aFG) and PFM, there was significant total association (P=0.012) and within-family association (P=0.010 and P=0.008 of 1000 permutation tests). For haplotype 6 (AFA) and fat mass, there was significant total association (P=0.018) and within-family association (P=0.012 and P=0.041 of 1000 permutation tests). Also, there was significant total association (P=0.017) and within-family association (P=0.040 and P=0.041 of 1000 permutation tests) for haplotype 6 (AFA) and PFM. Moreover, we investigated the association between ESR1 haplotypes and BMI, fat mass, lean mass, and PFM using QTDTs. For haplotype 1 (px) and fat mass, there was population association (P=0.011) and significant within-family association (P=0.002 and P=0.009 of 1000 permutation tests) (data not shown).

Table 8 Frequencies of VDR and ESR1 haplotypes for all parents in the study.
Full size table

In addition, we also investigated associations between the most common haplotypes of the two genes (aFG for VDR and px for ESR1) and BMD, BMI, fat mass, lean mass and PFM in 400 unrelated sons using GLM-ANOVA. There was no significant association between the aFG haplotype and BMD, BMI, fat mass, lean mass or PFM (data no shown). There was, however, significant association between the px haplotype and BMD at the lumbar spine (P=0.011) and total hip (P=0.048). Subjects carrying two copies of the px haplotype had higher BMD at the lumbar spine and total hip compared with those carrying no copy of the px haplotype (P=0.004 and P=0.032, respectively) (Table 5). No relationship between such combinations of genotypes and peak BMD or obesity phenotypes in 400 unrelated sons was found using two-way analysis of variance.

Discussion

Extensive population-based association studies have been performed in different ethnic groups to test the relationships between VDR and ESR1 genotypes and BMD variation32, 33. However, the results have been inconsistent or even contradictory. Our previous study did not find any relationship between ApaI genotypes and BMD in Chinese nuclear families34, and this did not agree with Dundar's findings which indicated that the VDR ApaI polymorphism may be responsible for an important part of the heritable component of lumbar spine BMD in postmenopausal women35. Lau et al36 reported that the VDR FokI polymorphism may have a weak effect on the BMD of elderly Chinese women, but Remes et al37 failed to find any association between VDR FokI and ApaI polymorphisms on BMD in middle-aged men. We found that ESR1 PvuII and XbaI polymorphisms may have minor effects on peak BMD in Chinese women24, 25, and another study38 also found that the ESR1 PvuII and XbaI genotypes may modulate the relationship between BMD in men. However, Finnish researchers failed to support the opinion that the ESR1 XbaI and PvuII polymorphisms have a substantial impact on the development of peak bone mass in young men39. This may be due to the ethnic difference of the participants, to some specific effects of the gene mutation, and/or different genetic architecture and allele frequencies. In this study, we collected 400 Chinese male nuclear families, and used TDTs to simultaneously test linkage and/or association of the VDR and ESR1 gene polymorphisms with peak BMD. We found that VDR and ESR1 genotypes were not associated with lumbar spine and hip BMD in young Chinese males. Moreover, no significant within-family association was found between the VDR and ESR1 haplotypes and peak BMD. Long et al's study40 suggested that the ESR1 haplotypes, not single markers, may be associated with BMD variation at some skeletal sites in Chinese male samples. This result is in partial agreement with our findings. Therefore, further studies in other ethnic male populations are needed to better define the relationship between VDR and ESR1 genotypes and BMD.

Although the importance of the VDR and ESR1 genes to bone biology is widely acknowledged, their importance to obesity is seldom reported. Obesity is an excess proportion of total body fat. Several studies have shown that body weight is the most common index of obesity, but it cannot be used to distinguish body fat from lean mass. More defined phenotypes have been proposed for studying obesity, such as fat mass, lean mass, and percentage fat mass. Genetic variation in VDR is associated with muscle strength, fat mass, and body weight in Swedish women3. Recently, however, Moreno et al41 failed to find an association between VDR genotypes and haplotypes with fat-free mass in postmenopausal Brazilian women. Roht's research42 showed that VDR genotype is associated with fat-free mass in elderly Caucasian men. In the present study, we measured whole body fat mass, lean mass, and FPM as indices of the degree of obesity, using DXA. Using ANOVA and QTDT, we all found that CDX2 was significantly associated with BMI, fat mass, and PFM, and PvuII was significantly associated with fat mass and PFM. Haplotype analysis also supported the above findings. These results are inconsistent with findings in Caucasian populations; evidence of a possible link between genetic variation in ESR1 and obesity is provided by the observation that the PvuII polymorphism is associated with fat mass in women but not men43. However, Grundberg et al 44 found that a TA-repeat polymorphism in the ESR1 gene did not correlate with muscle strength or body composition in young adult Swedish women. This may be due to the difference of the participants, especially sex and menopause-associated changes.

The present study differs from most other studies in the following three characteristics. First, most of the studies mentioned above were based on the traditional population association approach, which is susceptible to population structure and with which it is easy to generate spurious results. In this study, we applied a more robust method, QTDT, to estimate the relationship between polymorphisms in the VDR and ESR1 genes with spine and hip BMD. Peak BMD is thought to be under strong genetic control. In our sample, the heritability estimates for peak BMD in the spine, femoral neck, and hip are 0.565, 0.702, and 0.693, respectively. Second, we have now learned that lean mass and fat mass may be important determinants of the BMD45, 46. Many previous studies used BMD as a surrogate marker of bone strength, which did not take into account determinants of bone strength (eg, fat mass, lean mass and PFM) other than bone mass. Additionally, few studies have focused on the association of genetic polymorphisms with fat mass and lean mass. To our knowledge, this is the first study to investigate the possible influence of VDR and ESR1 genotypes and haplotypes to BMD, BMI, fat mass, lean mass, and FPM variation in Chinese males. Third, we used a relatively large sample of male offspring nuclear families.

Of course, our study has several limitations. We tested only five common loci in the two genes, so we cannot rule out the possibility that association may exist between other polymorphisms in the genes with bone and obesity phenotypes. Therefore, further studies using denser markers are needed to test the effects of the VDR and ESR1 genes (and other candidate genes) in the Chinese or other populations. In addition, because all nuclear families were represented by only two generations in this study, no sibling pairs were informative for the linkage analyses. As a result, no linkage for the VDR or ESR1 genotypes and haplotypes with BMD, BMI, fat mass, lean mass, and PFM were detected.

In conclusion, our results showed that the VDR and ESR1 polymorphisms were associated with fat mass in young Chinese men, but we failed to find a significant association between these polymorphisms and peak BMD. These findings suggested that the VDR and ESR1 genes are the QTL underlying fat mass variation in young Chinese men. Confirmation of our results is needed in other populations and with more functional markers of the two genes.

Author contribution

Zhen-lin ZHANG designed research; Jie-mei GU, Wen-jin XIAO, Jin-wei HE, and Wen-zhen FU performed research; Hao ZHANG, Wei-wei HU, Yun-qiu HU, Miao LI, Yu-juan LIU, Jin-bo YU, Gao GAO, Hua YUE, and Yao-hua KE recruited subjects; jie-mei GU and Zhen-lin ZHANG wrote the paper.