These authors contributed equally to this study
Large-Scale Population-Based Study Shows No Association Between Common Polymorphisms of the TGFB1 Gene and BMD in Women
Article first published online: 23 OCT 2006
Copyright © 2007 ASBMR
Journal of Bone and Mineral Research
Volume 22, Issue 2, pages 195–202, February 2007
How to Cite
McGuigan, F. E., Macdonald, H. M., Bassiti, A., Farmer, R., Bear, S., Stewart, A., Black, A., Fraser, W. D., Welsh, F., Reid, D. M. and Ralston, S. H. (2007), Large-Scale Population-Based Study Shows No Association Between Common Polymorphisms of the TGFB1 Gene and BMD in Women. J Bone Miner Res, 22: 195–202. doi: 10.1359/jbmr.061016
- Issue published online: 4 DEC 2009
- Article first published online: 23 OCT 2006
- Manuscript Accepted: 18 OCT 2006
- Manuscript Revised: 14 AUG 2006
- Manuscript Received: 3 MAY 2006
The TGFB1 gene is a strong functional candidate for regulating genetic susceptibility to osteoporosis. We studied five common polymorphisms of TGFB1 in relation to osteoporosis-related phenotypes in a population-based cohort of 2975 British women, but found no significant association with bone mass, bone loss, bone markers, or fracture.
Introduction: The gene encoding TGFB1 is a strong functional candidate for genetic susceptibility to osteoporosis. Several polymorphisms have been identified in TGFB1, and previous work has suggested that allelic variants of TGFB1 may regulate BMD and susceptibility to osteoporotic fracture.
Materials and Methods: We studied the relationship between common polymorphisms of TGFB1 and several osteoporosis-related phenotypes including BMD at the lumbar spine and femoral neck, measured by DXA; bone loss over a 6-year period; biochemical markers of bone turnover (urinary free deoxypyridinoline and free pyridinoline/creatinine ratio and serum N-terminal propeptide of type 1 collagen), and fractures in a population-based study of 2975 women from the United Kingdom. Participants were genotyped for single nucleotide polymorphisms (SNPs) in the TGFB1 promoter (G-800A; rs1800468; C-509T; rs1800469), exon 1 (T29C; rs1982073 and G74C; rs1982073); and exon 5 (C788T; rs1800471) on PCR-generated fragments of genomic DNA. Haplotypes were constructed from genotype data using the PHASE software program, and genotypes and haplotypes were related to the phenotypes of interest using general linear model ANOVA, with correction for confounding factors including age, height, weight, menopausal status, hormone replacement therapy (HRT) use, physical activity score, and dietary calcium intake.
Results: The polymorphisms were in strong linkage disequilibrium, and four common haplotypes accounted for >95% of alleles at the locus. There was no association between individual SNPs and BMD, bone loss, or biochemical markers of bone turnover. Haplotype analysis showed a nominally significant association with femoral neck BMD (p = 0.042) and with incident osteoporotic fracture (p = 0.013), but these were not significant after correcting for multiple testing.
Conclusions: Common polymorphic variants of the TGFB1 gene did not influence BMD or bone loss in this population.
Twin and family-based studies have shown that genetic factors play an important role in the regulation of BMD and other phenotypes relevant to the pathogenesis of osteoporotic fractures, such as bone turnover, ultrasound properties of bone, and femoral neck geometry.(1) It is currently believed that these phenotypes are regulated by a complex interplay between polymorphic variations in several candidate genes, each with a relatively modest effect, and environmental factors such as diet and exercise.(1) One of the most extensively studied candidate genes for osteoporosis susceptibility is the TGFB1 gene on chromosome 19q13, which encodes TGFB1. TFGβ1 is abundant in bone matrix, and it has been suggested that release and activation of TGFB1 during osteoclastic bone resorption plays a role in coupling bone resorption to bone formation.(2) The potential importance of TGFB1 as a genetic regulator of bone mass and turnover is shown by the fact that mutations of TGFB1 cause Camurati-Engelmann disease (CED), a condition associated with increased bone turnover and osteosclerosis affecting the diaphysis of the long bones.(3) Functional studies have shown that the mutations that cause CED impair binding of the inhibitory latency-associated peptide (LAP) to mature TGFB1, causing activation of SMAD signaling.(4) Several studies have now been conducted in which common polymorphic variants of TGFB1 have been examined in relation to BMD, osteoporotic fracture, or biochemical markers of bone turnover; however, these studies have yielded inconclusive and sometimes contradictory results possibly explained by differences in the ethnic makeup of the populations studied and small sample size.(5–12) It is now appreciated that many initial studies overestimate the effect size of a genetic variant and subsequent larger studies may provide a more accurate representation of the contribution of a given polymorphism to BMD.(13) Here, focusing on the well-characterized polymorphisms of the TGFB1 gene that have been studied by other groups in relation to osteoporosis, we examined their relationship with bone mass, bone turnover, and fracture in a large, ethnically homogenous cohort of white women from northeast Scotland.
MATERIALS AND METHODS
The participants had been enrolled in the Aberdeen Prospective Osteoporosis Screening Study (APOSS), a population-based screening study for osteoporotic fracture risk, involving 5119 women 45–54 years of age who were initially assessed over a 4-year period between 1990 and 1994. At this assessment, participants were weighed on a set of balanced scales calibrated to 0.05 kg (SECA, Hamburg, Germany), and height was measured with a stadiometer (Holtain, Crymych, UK). Participants also completed a risk factor questionnaire including questions on menopausal status and use of hormone replacement therapy (HRT). Women were classified as premenopausal if they were menstruating regularly, as perimenopausal if their menstruation was irregular and/or up to 6 months had elapsed since their last period, and as postmenopausal if their menstruation had ceased for 6 months or more. On the basis of this information, we categorized subjects into five groups as follows: 1 = premenopausal, no HRT; 2 = perimenopausal, no HRT; 3 = postmenopausal, no HRT; 4 = postmenopausal, previous HRT user; 5 = postmenopausal, current HRT user. After the baseline assessment, BMD results were disclosed to participants through their general practitioners and subjects with BMD values in the lowest quartile of the population were advised to consider taking HRT therapy.(14) A subset of the women were also recruited into a 2-year, randomized controlled trial of population screening for osteoporosis risk. The uptake of HRT assessed after this time was 6% higher compared with nonscreened women.(14)
All participants were invited to undergo further assessment between 1997 and 1999 (mean follow-up time, 6.3 ± 0.86 [SD] years) and a total of 3883 women, of the original 5119, attended at this time (75.8%). The baseline questionnaire was repeated and additional information was collected on diet, by a food frequency questionnaire and on physical activity. Physical activity level (PAL) was calculated from the questionnaire by recording the numbers of hours in a 24-h period spent doing heavy, moderate, or light activities and the numbers of hours in the same period spent sleeping or resting in bed.(15) At this visit, blood samples were collected for DNA extraction and analysis of a bone formation marker (serum N-terminal propeptide of type 1 collagen [P1NP]), and urine samples were collected for analysis of markers of bone resorption (free deoxypyridinoline [fDPD] and free pyridinoline [fPYD]). The participants that were included in this study comprised 2975 women who were successfully genotyped for at least three of the five polymorphisms in the TGFB1 gene under study.
BMD was measured at the femoral neck (FN-BMD) and lumbar spine (L2–L4) (LS-BMD) by DXA using Norland XR26 and XR36 densitometers. The majority of women were scanned using an XR26 densitometer, but 357 women (11.5%) were scanned using an XR36 densitometer. There was a small difference (1.26%) in mean BMD values between the densitometers and a correction factor was used to convert the XR36 values to XR26-equivalent values. This correction factor has previously been validated and has been shown to give similar results in assessing genotype specific differences in BMD in analyses in which scanner type is entered into the statistical model as a covariate.(16,17)
Biochemical markers of bone turnover
Biochemical markers of bone turnover were measured at the follow-up visit. The bone resorption markers free pyridinoline cross-links (fPYD) and free deoxypyridinoline cross-links (fDPD) were measured in urine samples using high-performance liquid chromatography as previously described.(18) Creatinine (Cr) was measured in urine by standard automated techniques (Roche, Lewes, UK), and results were expressed as fPYD/Cr and fDPD/Cr (nmol/mmol). The marker of bone formation, serum N-terminal propeptide of type 1 collagen (P1NP), was measured using an Electro Chemiluminescent Immuno-Assay (ECLIA) supplied by Roche products (Penzberg, Germany). The P1NP assay has a sensitivity of 2 μg/liter and an interassay and intra-assay CV of <4% across the range 5–100 μg/liter.
Data on self-reported fracture history since the age of 19 years was gathered at the baseline visit (1990–1994), but these fractures were not confirmed by X-rays or review of medical records. Fracture history was also recorded, by questionnaire, at the follow-up visit (1997–1999) and by a further postal questionnaire sent out in 2002. These incident fractures, which were reported as occurring between the baseline visit and the questionnaire in 2002, were validated by scrutiny of medical records, radiology reports, or general practitioner records.
We studied five single nucleotide polymorphisms (SNPs). Two were situated in the TGFB1 promoter; 800 bp upstream from the transcription start site (G-800A; rs1800468) and 509 bp upstream from the start site (C-509T; rs1800469). Two were situated in exon 1; a T/C change at position 29 in the cDNA (T29C; rs1982073) resulting in a leucine to proline amino acid change at codon 10 (Leu10Pro); and a G/C change at position 74 (G74C; rs1800471) resulting in an arginine to proline amino acid change at codon 25 (Arg25Pro). The final SNP was situated in exon 5, a C/T change at position 788 in the cDNA (C788T; rs1800472), which results in a threonine to isoleucine amino acid change at codon 263 (Thre263Ile).
The polymorphisms were genotyped by restriction fragment length polymorphism (RFLP) analysis of PCR products using the primers shown in Table 1. For the C-509T; T29C and C788T polymorphisms the PCR products were generated using Qiagen Taq DNA polymerase and Q buffer; for the G-800A polymorphism, PCR products were generated using Promega Taq DNA polymerase; and for the G74C polymorphism PCR products were generated using Invitrogen Platinum Pfx DNA polymerase. Magnesium concentration in the PCR reactions was 1.5 mM except for G74C, where it was 0.75 mM. To detect the polymorphism under study, between 10 and 15 μl of the PCR reaction product was digested with the appropriate restriction enzyme in a 20 μl volume for 16 h according to the manufacturer's instructions and visualized on 3% agarose gels stained with ethidium bromide. Details of PCR primers, RFLP product sizes, and restriction enzymes used in the analysis are included in Table 1. Two observers, blinded to the clinical data, scored the genotypes independently. Genotyping was repeated in the event of discordant results. Repeat genotyping of each SNP for a random 5% of all samples was also performed. In 2809 cases, the G74C polymorphism was also genotyped using allelic discrimination (Applied Biosystems TaqMan), by Dr Andre Uitterlinden at Erasmus Medical Center, Rotterdam. Comparison of genotypes for 50 randomly selected control samples using the RFLP and Taqman methods showed 100% concordance.
Statistical analyses were performed using SPSS version 12.0 (SPSS, Chicago, IL, USA). Hardy-Weinberg equilibrium (HWE) was calculated by the χ2 test. Pairwise linkage disequilibrium (LD) between SNP was calculated by the EH and 2BY2 programs. Genotype- and haplotype-specific differences in BMD were studied using general linear model ANOVA, entering age, weight, height, menopausal status, HRT use, and PAL into the model. Comparison between groups was by Scheffe's posthoc test. Predictors of incident fracture were identified by logistic regression, and we allowed for three possible genetic models: co-dominant, dominant, and recessive. Haplotypes were constructed, using the PHASE program version 2.02,(19) from individuals who had genotype data at three or more SNP sites. Only haplotype data with a >95% probability of correct assignment were used in the statistical analysis.
Study power and significance thresholds
The study had excellent power to detect allele-specific effects on BMD for all of the common haplotypes and genotypes studied. For G-800A and C-509T, we had 99% and 91% power, respectively, to detect a 0.20 SD difference in BMD between alleles, whereas previous studies of the −509T polymorphism have reported a 0.35 SD difference in BMD between extreme genotypes.(20) For T29C, we had 99% power to detect a 0.20 SD difference in BMD between alleles, whereas previous workers have reported differences of between 0.3(8) and 0.5 SD(11) for this polymorphism. For the rare polymorphism at C788T we only had 25% power to detect a 0.20 SD differences in BMD, but we had 90% power to detect a 0.5 SD difference in BMD. The power to detect allele specific differences in BMD of 0.20 SD in relation to common haplotypes was good and ranged from 78% for haplotype 4–99% for haplotype 1.
We estimated adjusted levels of significance for the number of independent tests performed in relation to the primary outcomes of bone mass at the spine and hip, bone loss at the spine and hip, and fracture, in relation to the four common haplotypes of TGFB1 that were studied. Excluding the biochemical markers that were analyzed as explanatory variables, nine different tests were conducted, although these were not independent, because BMD at the spine and femoral neck were significantly correlated (r = 0.668) as were rates of bone loss at the spine and femoral neck (r = 0.49). Accordingly, we conducted 7.83 independent tests, which gives an adjusted level of significance of p = 0.006 (equivalent to p = 0.05), taking into account multiple testing.
The genotype frequencies for the SNP studied were in keeping with those reported in other white populations. All frequencies were in HWE except for C788T; however, this could be accounted for by the low frequency of the rare allele, such that the TT genotype was observed in two individuals compared with the expected single incidence (Table 2).
There was strong and highly significant linkage disequilibrium (LD) between most of the polymorphisms studied (p < 0.0000), with the exception of G74C and C788T where LD was moderate (D′ = 0.58) and nonsignificant (p = 0.069). Reflecting the strong LD, PHASE analysis showed that four common haplotypes accounted for >95% of alleles at the TGFB1 locus (Fig. 1). These were as follows (in order: G-800A/C-509T/T29C/G74C/C788T): haplotype 1, GCTGC (53%); haplotype 2, GTCGC (25%); haplotype 3, ACTGC (9.5%); haplotype 4, GCTCC (7.5%). In view of this, association analysis with clinical variables of interest was performed on the basis of haplotype rather than genotype and for these analyses, participants were coded according to haplotype copy number (i.e., whether they carried zero, one, or two copies of the haplotype under study.
Relevant clinical and demographic data for these analyses in each of the four haplotypes (under a co-dominant model) are shown in Tables 3–6. Homozygotes for haplotype 2 were slightly but significantly younger than the other haplotype groups (p = 0.013) using this model; however, there was no significant difference in BMD, bone loss, or biochemical markers of bone turnover in relation to carriage of haplotypes 2, 3, or 4 in the study population. For haplotype 1, we observed a trend with FN-BMD, with lower values in homozygotes (p = 0.042), which increased to p = 0.01 under a recessive model, although this is nonsignificant when multiple testing is taken into consideration. Assuming a dominant model, haplotype 3 showed a trend with change in FN-BMD (p = 0.035), with those individuals carrying one or more copies of the haplotype loosing the most bone. When we repeated the above analyses using individual genotypes rather than haplotypes, no significant associations were observed with BMD, bone loss, or biochemical markers of bone turnover (data not shown).
Between the baseline visit and 2002 (a duration of 9.7 ± 1.1 [SD] years), incident fractures occurred in 255 (8.6%) women. Of these, 81 (31.7%) were fractures of the hip, wrist, or spine. Logistic regression analysis showed that regardless of model (co-dominant, dominant, or recessive) occurrence of incident fracture at any site was predicted independently by LS-BMD and FN-BMD (β = −3.07, p < 0.001 and β = −3.88, p < 0.001, respectively). Menopausal status was predictive of fracture (β = −0.94, p = 0.033) only in models containing FN-BMD. TGFB1 haplotype was not predictive of incident fracture under any model. When we considered only osteoporotic incident fractures (of the hip, wrist, and spine combined), spine BMD and femoral neck BMD were predictive of fracture independently (β = −4.25, p < 0.001 and β = −4.60, p = 0.001, respectively). Only when a recessive model was assumed was TGFB1 haplotype 1 found to be a nominal predictor of fracture incidence (β = −0.967, p = 0.013), with individuals carrying two copies of the haplotype having a lower incidence of fracture compared with those with one or zero copies, respectively (0.4%; 1.0%; 1.3%; χ2 = 7.1; df = 2; p = 0.029).
We observed no association between TGFB1 haplotype 2 and self-reported adult fractures or an association between adult fracture and any other TGFB1 haplotype or genotype.
The TGFB1 gene is a strong candidate for susceptibility to osteoporosis. Bone matrix contains large amounts of TGFB1, and it is thought that release of active TGFB1 from bone matrix during osteoclastic bone resorption plays an important role in coupling bone formation to bone resorption.(2) Evidence that genetic variation in TGFB1 can influence bone mass and bone turnover in humans comes from the observation that mutations in TGFB1 cause Camurati-Engelmann disease, a rare autosomal dominant condition resulting in progressive osteosclerosis, predominantly of the diaphysis of the long bones.(3) Many previous association studies of TGFB1 polymorphisms have been conducted in relation to osteoporosis-related phenotypes such as BMD, ultrasound properties of bone, fracture, and bone loss.(5–12,21–25) The inconclusive results reflect the ethnic variation of the populations studied and the often limited sample sizes that are underpowered to detect the modest effects that are expected for candidate gene polymorphisms.(26) For example, one study in Japanese postmenopausal women showed evidence of reduced bone mass and an increased risk of fracture in carriers of the T allele at the T29C polymorphic site encoding proline at codon 10,(11) whereas the opposite association was observed in elderly white postmenopausal women from Western Australia.(5) Another large study of a predominantly white population of postmenopausal women in the United States who took part in the Study of Osteoporotic Fractures (SOF) showed no significant association between the T29C polymorphism and BMD or fracture.(12) In our study, which is the largest and most comprehensive analysis of the TGFB1 gene thus far performed in relation to osteoporosis-related phenotypes, we found no association between the T29C polymorphism or other common SNP in TGFB1 and BMD. Although trends were observed for an association between haplotype 1 and femoral neck BMD, and haplotype 3 and change in femoral neck BMD, the results fell well below the adjusted threshold for significance when the number of tests performed was taken into account The findings reported here differ from those of many previous studies where TGFB1 alleles were found to be associated with BMD.(5–8,11,20) However, most of these reports were based on studies of modest sample size that can lead to false-positive results.(13) We did not have adequate power to detect modest effects of the rare C788T polymorphism on BMD, although this polymorphism has not previously been associated with either BMD or osteoporotic fractures.(8) Similarly, we did not have power to exclude an association between any of the polymorphisms or haplotypes and fractures, and further studies will be needed to investigate this. It is possible that as this population ages and the overall fracture incidence increases, the TGFB1 haplotype may become increasingly important in the determination of fracture occurrence. Furthermore, it is possible that novel polymorphisms, identified in the promoter region of the gene(27) since this study was performed, might be associated with BMD or fracture, either alone or in combination with the polymorphisms studied here. We did not study all known polymorphisms of the TGFB1 gene, such as 713–8delC in intron 4 and T861–20C in intron 5; however, the polymorphisms that were analyzed in this study are likely to have captured much of the genetic variation within the TGFB1 gene. The 713–8delC polymorphism, which has previously been found to be associated with osteoporosis in Danish and Italian populations,(9,28) is in complete LD with C-509T. Likewise, the T861–20C polymorphism, previously found to be associated with femoral neck BMD in a twin study,(7) is in strong linkage disequilibrium with G-800A.(8)
We conclude that, in contrast to previous studies in white populations, which have suggested that polymorphisms of TGFB1 may act as genetic determinants of susceptibility to osteoporosis, this large study suggests that, in this population of early menopausal women, common allelic variants of the TGFB1 gene do not contribute appreciably to the regulation of BMD and fracture.
This study was supported in part by grants from the European Commission (QLRT-2001-02629); the UK Food Standards Agency; the Scottish Executive (Genetic Heath in the 21st Century); and the Arthritis Research Campaign (ICAC grant). AS is supported by a Non-Clinical Career Development Fellowship from the UK Arthritis Research Campaign. Any views expressed are the authors' own. We thank Jay Wallace and Grace Taylor for genotyping, Dr N Hoyle of Roche Germany for the kind donation of reagents to perform the P1NP analysis, and Allan Walker for database support. We also thank the radiographers and research nurses at the Osteoporosis Research Unit and all the women who kindly participated in the study.
- 62001 Association of transforming growth factor-beta1 (TGFbeta1) T29 C gene polymorphism with bone mineral density (BMD), changes in BMD, and serum concentrations of TGF-beta1 in a population-based sample of postmenopausal german women. Calcif Tissue Int 69: 315–320., , , , ,
- 91997 A sequence variation: 713-8delC in the transforming growth factor-beta 1 gene has higher prevalence in osteoporotic women than in normal women and is associated with very low bone mass in osteoporotic women and increased bone turnover in both osteoporotic and normal women. Bone 20: 289–294., , , ,
- 202001 Association of the C-509T polymorphism, alone of in combination with the T869C polymorphism, of the transforming growth factor-beta1 gene with bone mineral density and genetic susceptibility to osteoporosis in Japanese women. J Mol Med 79: 149–156., , , , ,