A recent meta-analysis found that secreted phosphoprotein-1 (SPP1) can predict the risk of both osteoporosis and fracture. No study has explored the association of SPP1 haplotype-tagging single nucleotide polymorphisms (htSNPs) and haplotypes with bone mineral density (BMD).
This is a cross-sectional study. A total of 1,313 healthy Taiwanese women aged 40 to 55 years were recruited from MJ Health Management Institute from 2009 to 2010. BMD was dichotomized into high and low BMD groups. Three common (allele frequency ≥5%) htSNPs were selected to examine the association between sequence variants of SPP1 and BMD.
Homozygosity for the T allele of rs4754 were protective from low BMD [TT vs. CC: adjusted OR (AOR) = 0.58, 95% confidence interval (CI) = 0.83–0.89]. A protective effect was also found for women carrying 2 copies of Hap3 TCT (AOR = 0.57, 95% CI = 0.34–0.95). Menopausal status marginally interacted with SPP1 rs6839524 on BMD (p = 0.049). Postmenopausal women carrying variant rs6839524 (GG+GC vs. CC: AOR = 2.35, 95% CI = 1.06–5.20) or Hap1 TGC (AOR = 2.36, 95% CI = 1.06–5.24) were associated with 2.4-fold risk of low BMD. For women with low BMI (<18.5 kg/m2), variant rs6839524 (AOR = 7.64) and Hap1 (AOR = 6.42) were associated with increased risk of low BMD. These findings did not reach statistical significance after correction for multiple tests.
Citation: Chen J-H, Chen Y-C, Mao C-L, Chiou J-M, Tsao CK, Tsai K-S (2014) Association between Secreted Phosphoprotein-1 (SPP1) Polymorphisms and Low Bone Mineral Density in Women. PLoS ONE 9(5): e97428. https://doi.org/10.1371/journal.pone.0097428
Editor: Maria Eugenia Saez, CAEBi, Spain
Received: January 16, 2014; Accepted: April 21, 2014; Published: May 15, 2014
Copyright: © 2014 Chen et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Funding for the study was provided by National Science Council grants 98-2314-B-002-081, 99-2314-B-002-128, and 101-2118-M-001-011-MY3. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Osteoporosis, characterized by low bone mass and propensity to fracture, has become a global health issue as the rapid growth of aging population , . About 21% of women aged 50 to 84 years has osteoporosis, which is three times higher than that in men . In the US (1988–1994), the prevalence of osteoporosis was highest among white (19%), followed by Mexican American (16%) and black (7%) , which in part correspondence to the levels of bone mineral density (BMD) ranged from lowest in Asian, followed by native Americans, Hispanic, and then African American . This may be explained by different lifestyle in the East and West. Compared with other Asian countries, Japanese and Korean, but not Taiwanese, women showed clear age-dependent loss of BMD . Osteoporosis is a “silent disease” until a sudden strain, twist, fall, or fracture, which is associated with increased mortality in later life . Therefore, it is important to identify osteoporosis risk at an early age to prevent fall and fracture risk in late life.
Secreted phosphoprotein-1 (SPP1) is known as osteopontin. It is a glycoprotein related to bone formation and anchoring of osteoclasts to the bone remodeling matrix via binding with vitronectin receptor , . SPP1 exists in osteoblasts and mineralized bone matrix and intramembranous ossification , which enhances osteoblastic differentiation and proliferation , . SPP1 modulates both bone formation and resorption . As compared to wildtype mice, SPP1 gene knockout mice are resistant to bone resorption . A Chinese study found that the heritability of BMD was quite high (0.6 to 0.9, vary by body site and sex) , therefore, genetic difference may pay a role on BMD level. Because SPP1 is involved in osteogenesis and bone remodeling, SPP1 polymorphisms may play an important role in the pathogenesis of osteoporosis. Variation on functional single nucleotide polymorphisms (SNPs) may affect the production of SPP1 protein and then bone formation; intronic SNPs may affect BMD via regulating the alternative splicing and the subsequent protein production , .
Few epidemiologic studies have explored the association between SPP1 polymorphisms and osteoporosis or BMD. One study found that increased plasma SPP1 level was associated with low BMD . The other candidate-gene study including white and African Americans reported no difference between average hip or spine BMD level by genotypes of SPP1 rs11730582 or rs4754 . However, the selection of these two SNPs only captured limited genetic information in SPP1 gene (r2 = 0.62). In addition, these studies did not assess the association between SPP1 genetic polymorphisms and low BMD or osteoporosis. A meta-analysis study, included 5 genome-wide association studies [GWASs, using data mining approach to explore a massive number of SNPs for specific outcome(s)] on BMD and fracture, found that polymorphisms of 9 genes (ESR1, LRP4, ITGA1, LRP5, SOST, SPP1, TNFRSF11A, TNFRSF11B, and TNFSF11) were significantly associated with BMD level; and SPP1, SOST, LRP5, and TNFRSF11A were also related to elevated risk of fracture . Another recent meta-analysis, included 17 GWASs on BMD, found that polymorphisms of FAM210A, SLC25A13, LRP5, MEPE/SPP1, SPTBN1, and DKK1 were significantly associated with both BMD and fracture risk . Among the genes that can predict both BMD and fracture risk, SPP1 is the only one that modulates both osteoblast and bone resorption , and associated with the risk of both vertebral and non-vertebral fracture . In addition, no studies have explored how SPP1 genetic polymorphisms affect the risk of low BMD by using representative haplotype-tagging single nucleotide polymorphisms (htSNPs) and haplotypes. Data in Asian is also lacking.
Because of the above research gap, this study was aimed to explore the association between SPP1 polymorphisms and the risk of low BMD in middle-aged women. A systematic approach was used to select representative htSNPs in SPP1 to capture sufficient genetic information and to identify SNPs representative for Asian population. We also evaluated the interactions between SPP1 polymorphisms and menopausal status or body mass index (BMI) on BMD, which has not been explored previously.
Materials and Methods
This is a cross-sectional study. A total of 1,567 healthy Taiwanese (Chinese ethnicity) women, aged 40 to 55 years, were recruited from MJ Health Management Institution from October 2009 to August 2010. The outcome of this study is spinal BMD (g/cm2). Spinal BMD is the major site measured at the MJ Health Management Institute. Participants with the following conditions or diseases were excluded (n = 254): (1) diseases known to affect BMD levels (e.g., hyperparathyroidism, hyperthyroidism, type 1 diabetes, inflammatory bowel disease, chronic active hepatitis, liver cirrhosis, chronic cholestatic diseases, and multiple myeloma), (2) took medications for osteoporosis (e.g., raloxifene), (3) received hormone replacement therapy or other medications (e.g., steroid, oral contraceptive agents) that may affect BMD, (4) lack of BMD at lumbar spine, (5) lack of blood samples or genotyping data. A total of 1,313 women were included for data analyses.
A questionnaire was administered to collect information on demography, lifestyle (e.g., smoking status, alcohol consumption, and exercise), and disease history, etc. A blood sample was collected in an 8 ml EDTA tube from each participant. Genomic DNA was extracted by using QuickGene-Mini80 kit (Fujifilm, Tokyo, Japan). Participants with the following conditions were excluded: BMD was measured at sites other than spine (n = 85), lack of blood sample (n = 113), or genotyping data (n = 2), and had steroid or hormone therapy (n = 70). Some participants may lack of two or more information above.
The study protocol has been approved by the Institutional Review Boards of National Taiwan University and MJ Health Management Institution. Written informed consent was obtained from each study participant. The consent from the legal guardian/next of kin was obtained when patients had serious cognitive impairment. This research carried out with human subjects complies with the World Medical Association Declaration of Helsinki - Ethical Principles for Medical Research Involving Human Subjects.
Measurement of bone mineral density
The BMD (g/cm2) of the lumbar spine (L1-L4) was measured by a dual-energy X-ray absorptiometry densitometer (DXA, General Electric Lunar Health Care, DPX-L, USA). Calibration of BMD measurement was performed daily. The long-term coefficient of variation in BMD was around 1%. This healthy population included few participants with osteoporosis (<1%). Instead of using osteoporosis as the outcome variable, BMD was tertiled (i.e. T1, T2, and T3) on the basis of the data of the whole population in order to identify the subgroup with an elevated risk of low BMD. Previous studies ,  and our recent study  have used similar approaches that involve BMD tertiles. The high BMD group comprised participants in T2 plus T3 (i.e., the reference group) and the low BMD group comprised participants in T1 (i.e., the comparison group) with a BMD cut-off point of 1.27 g/cm2.
SNP selection and genotyping assays
Common (frequency ≥0.05) SNPs in SPP1 were identified from genotyping data of Han Chinese in Beijing, China (CHB) of the International HapMap Project (http://hapmap.ncbi.nlm.nih.gov). Haplotype block was defined by Haploview program (http://www.broadinstitute.org/haploview/haploview) using modified Gabriel algorithm , . Three representative htSNPs [rs11730582 in 5′untranslated region (UTR), rs6839524 in intron, and rs4754 in exon] were selected from 12 common SNPs using tagSNP program  based on the common disease/common variant hypothesis . TaqMan Assay was used to determine genotypes using HT7900 (Applied Biosystems Inc., CA, USA). Genotyping success rate was greater than 95% for each SNP. Quality control samples were replicates of 5% study participants and the concordance rate was 100%.
Hardy-Weinberg equilibrium (HWE) test was performed for each SNP to check genotyping error. The expectation-maximization algorithm was used to estimate haplotype frequencies in the haplotype block using tagSNP program . High and low BMD was defined as above. Power analysis was performed by using QUANTO program (http://hydra.usc.edu/GxE/).
Logistic regression model was used to estimate the adjusted odds ratio (AOR) and 95% confidence interval (CI) for the risk of low BMD in participants carrying either 1 or 2 versus 0 copies of minor allele of each SNP and each multilocus haplotype. Haplotype trend regression  was used to test the global association between SPP1 haplotypes and low BMD. Given a significant global test, haplotype-specific tests can provide some guidance as to which variant(s) contributes to the significant global test. The association between SPP1 genetic polymorphisms and continuous BMD were also assessed by using general linear model (GLM). After stepwise model selection and the inclusion of variables with biological importance, age, menopausal status (yes/no), BMI (kg/m2), alkaline phosphatase (ALP, IU/L), uric acid (UA, mg/dL), low-density lipoprotein (LDL, mg/dL), and exercise (frequency × duration × intensity) were adjusted in the models. All participants had normal creatinine level (<1.3 mg/dL) and thus this variable was not explored in this study.
A likelihood ratio test was used to evaluate how menopausal status (pre- and post-menopause) and BMI groups (<18.5, 18.5 to <24, ≥24 kg/m2) modified the association between SPP1 polymorphisms and risk of low BMD. Stratified analyses were performed by menopausal status and BMI groups. Correction for multiple tests was performed by false discovery rate (FDR) using method of Benjamini and Hochberg (1995) . Statistical analyses were performed by using SAS 9.2 (SAS Institute, Cary, NC) and all statistical tests were two-sided.
Characteristics of the study population
This study included 1,313 participants. The differences between participants with low and high BMD are summarized in Table 1
SPP1 polymorphisms and BMD
Three SPP1 htSNPs [rs11730582 (5′ UTR), rs6829524 (intron), and rs4754 (exon)] were genotyped. The minor allele frequencies (MAFs) of these SNPs ranged from 31% to 42%, which were similar to the MAFs of CHB data from International HapMap Project (29 to 38%). All SPP1 SNPs were in HWE among participants with low BMD, high BMD, or the whole population. Power analysis showed that given 881 and 432 participants with low and high BMD, respectively, rs4754 (MAF = 0.31) has over 0.99 of power to detect an OR at 0.58. Because of the modest effect and high MAF, the power is low (<0.7) for rs11730582 (MAF = 0.33) and rs6829524 (MAF = 0.42) to detect an OR at 1.14 and 0.91, respectively.
Homozygosity for the T allele of rs4754 were protective from low BMD (TT vs. CC: AOR = 0.58, 95% CI = 0.83–0.89, p = 0.005, Table 2). This association did not reach statistical significance after correction for multiple tests. The other two htSNPs did not show significant relationship with the outcome.
Three common htSNPs (rs11730582, rs6839524, and rs4754) spanning SPP1 gene formed one block using the modified Gabriel algorithm , . Four common (frequency ≥0.05) haplotypes were identified (cumulative frequency, 97.9%); the global test for the association between SPP1 haplotypes and low BMD was significant (p<0.0001, Table 2). Two copies of Hap3 TCT were protective from low BMD (AOR = 0.57, 95% CI = 0.34–0.95, p = 0.03, Table 2). Other haplotypes were not associated with the outcome. The conditional haplotype analysis was performed conditioning on other haplotypes and the results did not reach statistical significance (Hap1: ref; Hap2: AOR = 1.14, 95% CI = 0.90–1.44; Hap3: AOR = 0.93, 95% CI = 0.74–1.17; Hap4: AOR = 0.85, 95% CI = 0.59–1.24). These findings did not remain significant after correction for multiple tests.
We also kept only participants with the lowest and the highest BMD tertile and compared them for the same analyses above. Because of removing one-third of the study population (the 2nd tertile), the statistical power decreased and the protective effect for rs4754 and Hap3 no longer reached statistical significance.
Interactions between menopausal status and SPP1 polymorphisms
Menopausal status has known as an important modifier for BMD. Interaction between menopausal status and SPP1 htSNPs or haplotypes for low BMD did not reach statistical significance. After stratification by menopausal status, postmenopausal women carrying variant rs6839524 were associated with low BMD (GG+GC vs. CC: AOR = 2.35, 95% CI = 1.06–5.20, p = 0.03). Postmenopausal women carrying Hap1 TGC was associated with low BMD (AOR = 2.36, 95% CI = 1.06–5.24, p = 0.03). These findings did not reach statistical significance after correction for multiple tests. No significant association was found in other subgroups or for other SPP1 htSNPs/haplotypes after stratification by menopausal status.
Interaction between BMI and SPP1 polymorphisms
No significant interaction was observed between SPP1 SNPs or haplotypes and low BMD. After stratification by BMI groups (<18.5, 18.5 to <24, ≥24 kg/m2), women with low BMI (<18.5 kg/m2) carrying rs6839524 variant were associated with low BMD (GG+GC vs. CC: AOR = 7.64, 95% CI = 1.42–40.97, p = 0.02). Women with low BMI (<18.5 kg/m2) carrying Hap1 TGC were associated with low BMD (AOR = 6.42, 95% CI = 1.23–33.60, p = 0.03). These findings did not reach statistical significance after correction for multiple tests.
The power for assessing interaction between SPP1 SNPs and menopausal status or BMI is low because of the smaller sample size in subgroup analysis and the results should be interpreted with caution.
To the best of our knowledge, this is the first study exploring the association between SPP1 polymorphisms and low BMD using htSNPs in Asian population. We found that homozygosity for the T allele of SPP1 rs4754 (TT) and Hap3 TCT were associated with low BMD; the former result remained significant after correction for multiple tests for SNP analysis but lost significance after correction for multiple tests for SNP and haplotype analysis. The only candidate-gene study , which included white and black populations, only compared mean BMD by SPP1 genotypes of 2 SNPs (i.e., no estimation of multivariable OR and 95% CI) and no significant difference was observed. Previous GWASs and meta-analysis , , – for BMD or fracture risk mainly focused on white population. In addition, haplotype analysis, which offers more information than single-locus SNPs, has not been performed previously. Therefore, our results provide important information because of the estimation of outcome risk by using multivariable OR, large sample size (n>1,300), selection of representative htSNP, and performing haplotype analysis.
Among 3 htSNPs genotyped in this study, rs4754 is a synonymous SNP. That is, rs4754 does not lead to the change of amino acid but may affect BMD level via its influence on translational efficiency. Interestingly, C allele is the major allele of rs4754 in Chinese (MAF: C = 0.66, T = 0.34) but the minor allele in white (MAF: C = 0.23, T = 0.77, http://hapmap.ncbi.nlm.nih.gov). Therefore, the inconsistent findings were observed between this Asian study and previous studies focused on whites , , –. Hap3 TCT also showed significant association with high BMD, which may be attributable to the only SNP rs4754 with the minor allele in Hap3. Three htSNPs were in one haplotype block and strong linkage disequilibrium (LD) were observed between rs11730582 and rs6839524 (|D′| = 0.98) as well as between rs6839524 and rs4754 (|D′| = 0.96). However, the pairwise r2 for any two SNPs were low (0.02 to 0.34). Especially, both D′ (0.30) and r2 (0.02) were low between rs11730582 and rs4754, this may explain the non-significant association between SPP1 rs11730582 and BMD.
SPP1 plays a role in a wide spectrum of physiologic and pathologic processes –. First, it mediates the attachment of osteoclasts to bone matrix and then regulates bone resorption and normal bone development . The polymorphisms of SPP1 gene may regulate SPP1 structure, decrease serum SPP1 level, change or reduce SPP1 protein, which may affect bone formation, resorption and the osteoclastic process. The downgrading of osteoclastic process may slow BMD decline and thus prevent osteoporosis. Second, SPP1 plays an important role in regulating immune response . Therefore, polymorphisms of SPP1 may block or reduce the inflammation responses and thus showed increased BMD. In addition, SPP1 polymorphisms may also interact with two important modifiers, menopausal status or BMI, on BMD as detailed below.
It has been known that sex hormone plays an important role in maintaining bone strength . For most of women, BMD decreases rapidly during the first few years after menopause  as a result of excessive osteoclastic activities via unopposed osteoclastic activation after rapid declination of estrogen. Because SPP1 modulates osteoclast and thus sequence variants of SPP1 may affect bone resorption. This may explain our finding that postmenopausal women carrying 1 or 2 copies of variant rs6839524 were associated with low BMD (AOR = 2.35), which did not reach statistical significance after correction of multiple tests. An association was also observed for Hap1, which rs6839524 is the only SNP with minor allele. It is possible that variant rs6839524 affects bone formation and this effect becomes more evident after menopause. In addition, rs6839524 is an intronic SNP and its variation may affect the alternative splicing, e.g., altering mRNA folding or the stability of mRNA structure, and then the subsequent protein production , . All these may explain the associations, which did not reach statistical significance after correction of multiple tests, between rs6839524 or Hap1 and low BMD in postmenopausal women.
BMI has been related to BMD previously. Low BMI has been associated with increased risk of osteoporosis and fracture  and the association varied by ethnic groups, e.g., positive association between one unit increase of BMI and BMD in white women but negative association was observed in African American women . Because of different diet and lifestyle, body shape can be quite different between people in Western and Eastern countries. However, relevant data and research in Asian population are sparse. Our research, for the first time, explored that among women with low BMI (<18.5 kg/m2), variant carriers of rs6839524 (AOR = 7.64) or Hap1 TGC (AOR = 6.42) were associated with low BMD, which did not reach statistical significance after correction for multiple tests. The application of these polymorphisms will help us to identify women with low BMD.
This study has several strengths. First, the selections of a set of representative htSNPs for Asian captured the majority of genetic information of SPP1 (r2 = 0.82, estimated by tagSNP program) as compared with that (r2 = 0.65) of the only candidate-gene study . Second, haplotypes capture unknown variants via LD between these SNPs and thus provide more information than single SNP. In addition, unlike most previous studies, this study included premenopausal women (n∼1000) that allowed us to assess how menopausal status interacted with SPP1 polymorphisms on BMD and, importantly, to predict outcome risk at an early age.
This study had some limitations. First, this is a cross-sectional study, which causal inference is usually not available. Functional analysis will be needed to unravel the underlying mechanism between SPP1 polymorphisms and BMD. In addition, the original questionnaire did not collect fracture information. Because this population is healthy, fracture frequency is low. We also assessed the association between SPP1 genetic polymorphisms and continuous BMD by using GLM and no significant findings were observed. Because the aim of this study is to identify a high-risk population of low BMD, no further analyses were performed by using continuous BMD as outcome.
SPP1 plays a role in bone formation and resorption. This study has some first findings. Homozygosity for the T allele of rs4754 and Hap3 TCT in SPP1 were significantly associated with low BMD in this Asian population. rs6839524 and haplotypes in SPP1 have not been explored before. Variant carriers of rs6839524 and Hap 1 TGC were associated with low BMD in menopausal women or women with low BMI. These findings did not reach statistical significance after correction for multiple tests. Because of the complex role of SPP1 in bone physiology, functional and larger studies are warranted to confirm our findings.
The authors gratefully acknowledge Dr. Wen-Chung Lee for epidemiological consultation.
Conceived and designed the experiments: KST YCC JHC. Performed the experiments: CLM. Analyzed the data: CLM JMC YCC JHC. Contributed reagents/materials/analysis tools: JHC YCC KST. Wrote the paper: CLM JHC JMC YCC CKT KST. Approval of the final version of the manuscript: CLM JHC JMC YCC CKT KST.
- 1. Sànchez-Riera L, Wilson N, Kamalaraj N, Nolla JM, Kok C, et al. (2010) Osteoporosis and fragility fractures. Best Pract Res Clin Rheumatol 24: 793–810.
- 2. Vestergaard P, Rejnmark L, Mosekilde L (2007) Increased mortality in patients with a hip fracture-effect of pre-morbid conditions and post-fracture complications. Osteoporos Int 18: 1583–1593.
- 3. Kanis JA, Johnell O, Oden A, Jonsson B, De Laet C, et al. (2000) Risk of hip fracture according to the World Health Organization criteria for osteopenia and osteoporosis. Bone 27: 585–590.
- 4. Dawson-Hughes B, Looker AC, Tosteson ANA, Johansson H, Kanis JA, et al.. (2011) The potential impact of the National Osteoporosis Foundation guidance on treatment eligibility in the USA: an update in NHANES 2005–2008. Osteoporos Int.
- 5. Barrett-Connor E, Siris ES, Wehren LE, Miller PD, Abbott TA, et al. (2005) Osteoporosis and fracture risk in women of different ethnic groups. J Bone Miner Res 20: 185–194.
- 6. Sugimoto T, Tsutsumi M, Fujii Y, Kawakatsu M, Negishi H, et al. (1992) Comparison of bone mineral content among Japanese, Koreans, and Taiwanese assessed by dual-photon absorptiometry. Journal of Bone and Mineral Research 7: 153–159.
- 7. Denhardt D, Noda M (1998) Osteopontin expression and function: Role in bone remodeling. J Cell Biochem: 92–102.
- 8. Choi ST, Kim JH, Kang E-J, Lee S-W, Park M-C, et al. (2008) Osteopontin might be involved in bone remodelling rather than in inflammation in ankylosing spondylitis. Rheumatology (Oxford) 47: 1775–1779.
- 9. Moore M, Gotoh Y, Rafidi K (1991) Characterization of a cDNA for chicken osteopontin: expression during bone development, osteoblast differentiation, and tissue distribution. Biochemistry.
- 10. Zohar R, Cheifetz S, McCulloch C (1998) Analysis of intracellular osteopontin as a marker of osteoblastic cell differentiation and mesenchymal cell migration. European Journal of Oral Sciences 106: 401–407.
- 11. Standal T, Borset M, Sundan A (2004) Role of osteopontin in adhesion, migration, cell survival and bone remodeling. Exp Oncol 26: 179–184.
- 12. Rittling SR, Matsumoto HN, Mckee MD, Nanci A, An X-R, et al. (1998) Mice Lacking Osteopontin Show Normal Development and Bone Structure but Display Altered Osteoclast Formation In Vitro. J Bone Miner Res 13: 1101–1111.
- 13. Feng Y, Hsu Y, Terwedow H, Chen C, Xu X, et al. (2005) Familial aggregation of bone mineral density and bone mineral content in a Chinese population. Osteoporos Int 16: 1917–1923.
- 14. Moyer RA, Wang D, Papp AC, Smith RM, Duque L, et al. (2011) Intronic polymorphisms affecting alternative splicing of human dopamine D2 receptor are associated with cocaine abuse. Neuropsychopharmacology 36: 753–762.
- 15. Kawase T, Akatsuka Y, Torikai H, Morishima S, Oka A, et al. (2007) Alternative splicing due to an intronic SNP in HMSD generates a novel minor histocompatibility antigen. Blood 110: 1055–1063.
- 16. Chang IC, Chiang TI, Yeh KT, Lee H, Cheng YW (2010) Increased serum osteopontin is a risk factor for osteoporosis in menopausal women. Osteoporos Int 21: 1401–1409.
- 17. Taylor BC, Schreiner PJ, Doherty TM, Fornage M, Carr JJ, et al. (2005) Matrix Gla protein and osteopontin genetic associations with coronary artery calcification and bone density: the CARDIA study. Hum Genet 116: 525–528.
- 18. Richards JB, Kavvoura FK, Rivadeneira F, Styrkarsdottir U, Estrada K, et al. (2009) Collaborative meta-analysis: associations of 150 candidate genes with osteoporosis and osteoporotic fracture. Ann Intern Med 151: 528–537.
- 19. Estrada K, Styrkarsdottir U, Evangelou E, Hsu YH, Duncan EL, et al. (2012) Genome-wide meta-analysis identifies 56 bone mineral density loci and reveals 14 loci associated with risk of fracture. Nat Genet 44: 491–501.
- 20. Bidoli E, Schinella D, Franceschi S (1998) Physical activity and bone mineral density in Italian middle-aged women. Eur J Epidemiol 14: 153–157.
- 21. Nock NL, Patrick-Melin A, Cook M, Thompson C, Kirwan JP, et al. (2011) Higher bone mineral density is associated with a decreased risk of colorectal adenomas. Int J Cancer 129: 956–964.
- 22. You YS, Lin CY, Liang HJ, Lee S, Tsai KS, et al. (2013( Epub ahead of Print)) Association between Metabolome and Low Bone Mineral Density in Taiwanese Women Determined by 1H NMR Spectroscopy. Journal of Bone and Mineral Research.
- 23. Chen YC, Giovannucci E, Lazarus R, Kraft P, Ketkar S, et al. (2005) Sequence variants of Toll-like receptor 4 and susceptibility to prostate cancer. Cancer Res 65: 11771–11778.
- 24. Gabriel SB, Schaffner SF, Nguyen H, Moore JM, Roy J, et al. (2002) The structure of haplotype blocks in the human genome. Science 296: 2225–2229.
- 25. Stram DO, Leigh Pearce C, Bretsky P, Freedman M, Hirschhorn JN, et al. (2003) Modeling and E-M Estimation of Haplotype-Specific Relative Risks from Genotype Data for a Case-Control Study of Unrelated Individuals. Hum Hered 55: 179–190.
- 26. Challenges for the 21st century. Nat Genet 29: 353–354.
- 27. Zaykin DV, Westfall PH, Young SS, Karnoub MA, Wagner MJ, et al. (2002) Testing association of statistically inferred haplotypes with discrete and continuous traits in samples of unrelated individuals. Hum Hered 53: 79–91.
- 28. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Statist Soc B 57: 289–300.
- 29. Hofman A, Breteler MMB, van Duijn CM, Krestin GP, Pols HA, et al. (2007) The Rotterdam Study: objectives and design update. Eur J Epidemiol 22: 819–829.
- 30. Sayed-Tabatabaei FA, van Rijn MJE, Schut AFC, Aulchenko YS, Croes EA, et al. (2005) Heritability of the function and structure of the arterial wall: findings of the Erasmus Rucphen Family (ERF) study. Stroke 36: 2351–2356.
- 31. Richards JB, Rivadeneira F, Inouye M, Pastinen TM, Soranzo N, et al. (2008) Bone mineral density, osteoporosis, and osteoporotic fractures: a genome-wide association study. Lancet 371: 1505–1512.
- 32. Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, Gudbjartsson DF, Walters GB, et al. (2009) New sequence variants associated with bone mineral density. Nat Genet 41: 15–17.
- 33. Kiel DP, Demissie S, Dupuis J, Lunetta KL, Murabito JM, et al. (2007) Genome-wide association with bone mass and geometry in the Framingham Heart Study. BMC Med Genet 8 Suppl 1S14.
- 34. Gravallese EM (2003) Osteopontin: a bridge between bone and the immune system. Journal of Clinical Investigation 112: 147–149.
- 35. Reinholt FP, Hultenby K, Oldberg A, Heinegard D (1990) Osteopontin—a possible anchor of osteoclasts to bone. Proceedings of the National Academy of Sciences of the United States of America 87: 4473–4475.
- 36. Denhardt D (1993) Osteopontin: a protein with diverse functions. The FASEB journal.
- 37. O'Regan AW, Chupp GL, Lowry JA, Goetschkes M, Mulligan N, et al. (1999) Osteopontin is associated with T cells in sarcoid granulomas and has T cell adhesive and cytokine-like properties in vitro. Journal of Immunology 162: 1024–1031.
- 38. Denhardt DT, Noda M, O'Regan AW, Pavlin D, Berman JS (2001) Osteopontin as a means to cope with environmental insults: regulation of inflammation, tissue remodeling, and cell survival. Journal of Clinical Investigation 107: 1055–1061.
- 39. Satoh Y, Soeda Y, Dokou S (1995) Analysis of Relationships Between Sex-Hormone Dynamics and Bone Metabolism and Changes in Bone Mass in Surgically Induced Menopause. Calcif Tissue Int 57: 258–266.
- 40. Saarelainen J, Kiviniemi V, Kröger H, Tuppurainen M, Niskanen L, et al.. (2011) Body mass index and bone loss among postmenopausal women: the 10-year follow-up of the OSTPRE cohort. J Bone Miner Metab.
- 41. Morin S, Tsang JF, Leslie WD (2009) Weight and body mass index predict bone mineral density and fractures in women aged 40 to 59 years. Osteoporosis International 20: 363–370.
- 42. Castro JP, Joseph LA, Shin JJ, Arora SK, Nicasio J, et al. (2005) Differential effect of obesity on bone mineral density in White, Hispanic and African American women: a cross sectional study. Nutr Metab (Lond) 2: 9.