High-density polymorphisms analysis of 23 candidate genes for association with bone mineral density

Bone 47 (2010) 975–981 Contents lists available at ScienceDirect Bone j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / b o n e High-density polymorphisms analysis of 23 candidate genes for association with bone mineral density Sylvie Giroux a, Latifa Elfassihi a,b,c, Valérie Clément a, Johanne Bussières a, Alexandre Bureau b,d, David E.C. Cole e,f, François Rousseau a,b,g,⁎ a Centre de Recherche de l'Hôpital St-François d'Assise du Centre hospitalier universitaire de Québec, Québec, Canada G1L 3L5 Faculté de Médecine, Université Laval, Québec, Canada Direction de la surveillance de l'état de santé de la population, Direction générale de la santé publique, Ministère de la santé et des services sociaux, Québec, Canada d Centre de recherche, Université Laval Robert-Giffard, Québec, Canada e Department of Clinical Pathology, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, Ontario, Canada f Department of Laboratory Medicine & Pathobiology, University of Toronto, Toronto, Ontario, Canada g Centre for the Development, Evaluation and Rational Implementation of New Diagnostic Tools in Medicine (CEDERINDT)/The CanGèneTest Research Consortium on Genetic Laboratory Services, Québec, Canada b c a r t i c l e i n f o Article history: Received 19 April 2010 Revised 9 June 2010 Accepted 12 June 2010 Available online 30 July 2010 Edited by: Stuart Ralston Keywords: Bone mineral density Osteoporosis CSF3R Association TagSNPs a b s t r a c t Osteoporosis is a bone disease characterized by low bone mineral density (BMD), a highly heritable and polygenic trait. Women are more prone than men to develop osteoporosis due to a lower peak bone mass and accelerated bone loss at menopause. Peak bone mass has been convincingly shown to be due to genetic factors with heritability up to 80%. Menopausal bone loss has been shown to have around 38% to 49% heritability depending on the site studied. To have more statistical power to detect small genetic effects we focused on premenopausal women. We studied 23 candidate genes, some involved in calcium and vitamin-D regulation and others because estrogens strongly induced their gene expression in mice where it was correlated with humerus trabecular bone density. High-density polymorphisms were selected to cover the entire gene variability and 231 polymorphisms were genotyped in a first sample of 709 premenopausal women. Positive associations were retested in a second, independent, sample of 673 premenopausal women. Ten polymorphisms remained associated with BMD in the combined samples and one was further associated in a large sample of postmenopausal women (1401 women). This associated polymorphism was located in the gene CSF3R (granulocyte colony stimulating factor receptor) that had never been associated with BMD before. The results reported in this study suggest a role for CSF3R in the determination of bone density in women. © 2010 Elsevier Inc. All rights reserved. Introduction Osteoporosis is a common disease characterized by a decrease in bone mineral density (BMD) and bone strength leading to an increased risk of fracture. Twin and family studies have shown that genetic factors are important for the development of osteoporosis through their influence on BMD. It has been estimated that up to 60– 80% of the variance in peak bone mass is due to genetic factors [1–3]. ⁎ Corresponding author. Department of Molecular Biology, Medical Biochemistry and Pathology, Faculty of Medicine, Unité de recherche en génétique humaine et moléculaire, Medical Biochemistry Service, Centre hospitalier universitaire de Québec & Université Laval, 10, rue de l'Espinay, Québec, QC, Canada G1L 3L5. Fax: +1 418 525 4195. E-mail addresses: sylvie.giroux@crsfa.ulaval.ca (S. Giroux), latifa.elfassihi@crsfa.ulaval.ca (L. Elfassihi), valerie.clement.2@ulaval.ca (V. Clément), johanne.bussiere@crsfa.ulaval.ca (J. Bussières), Alexandre.Bureau@msp.ulaval.ca (A. Bureau), davidec.cole@utoronto.ca (D.E.C. Cole), Francois.rousseau@mac.com (F. Rousseau). 8756-3282/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.bone.2010.06.030 Genetic effects are thought to be stronger in younger women when the accelerated bone loss observed during menopausal transition has not begun [4,5]. Identification of genes involved in BMD regulation is believed to be important in understanding the disease. It is estimated that multiple gene variants are involved and that each gene has a modest effect on the final phenotype [6]. With the availability of human sequence information and improvements in performance of genotyping methods, hundreds of association studies have been published with phenotypes related to bone density and/or osteoporosis [7]. However, relatively little success has been achieved, and inconsistent results have accumulated [8–10]. The main reasons for the poor success are the lack of power due to small samples but also likely due to genetic heterogeneity, to low linkage disequilibrium between the markers tested and a putative causal variant and the study of phenotypes not highly influenced by genes. In contrast, a few consistent positive associations have been reported between gene variants and bone mineral density. The LRP5 gene was convincingly shown to be important in the full development of peak bone mass and those results have been reproduced in many different samples 976 S. Giroux et al. / Bone 47 (2010) 975–981 [11–22]. Also, two different genome-wide association studies and subsequent meta-analyses have identified TNFRSF11B gene as a potential candidate [22–25]. In this study, we analyzed the association between bone mineral density and a total of 23 genes densely covered with SNPs (single nucleotide polymorphisms) and with a robust study design involving three independent samples of women (two premenopausal and one postmenopausal). Thirteen (13) candidate genes were selected for their involvement in vitamin D and calcium metabolism and ten (10) genes because they had been shown to be induced by estrogen in mouse and their level of expression was correlated with humerus trabecular BMD [26]. These included (A) genes involved in the biosynthesis of vitamin D such as CYP2R1, CYP27A1, CYP27B1 and degradation such as CYP24A1, (B) genes involved in the transport or regulation of vitamin D such as GC a vitamin-D binding protein and KL coding for klotho, a membrane protein acting as a negative regulator of 1-25-(OH)2D synthesis [27], (C) genes involved in the regulation of calcium such as PTHR1, PTH, CASR and its analogous calcium sensor GPRC6A, CALCR and CALCA and (D) ten genes identified by Lindberg et al. [26] including a bone matrix-related gene IBSP, growth factorrelated genes such as TGFB1, IGFBP4 and seven cytokine receptors (IL1RN, IL1R2, LIFR, CSF3R, CSF2RB, IL4R and TNFRSF1A). Each candidate gene was analyzed with a high-density set of SNP selected with tools available with HapMap data to achieve a complete coverage of the common genetic variation and thus to maximise the relative power to detect associations. Furthermore, we used a study design that optimizes power while minimizing the number of subjects genotyped for cost effectiveness. The study design included the analysis of a highly heritable phenotype, namely bone mineral density measured in women before menopause, when bone mineral density is close to the peak bone mass. This increases the likelihood of detecting small genetic effects since it minimizes the variance in bone density due to non-genetic determinants that are present during and after menopause [28]. In addition, we performed the analysis in a homogeneous group of white premenopausal women from the metropolitan region of Québec city (709 women) and all significant associations were replicated in a second group of white premenopausal women from the metropolitan region of Toronto (673 women). The most promising genetic variants remaining associated in the combined samples of premenopausal women (1382 women) were then tested in a sample of postmenopausal women from the metropolitan region of Quebec city (1401 women). This way, we expected to increase the likelihood of identifying variants truly associated with bone density in women. Materials and methods Subjects Quebec sample Recruitment for a study on genetic and environmental factors affecting BMD was achieved, between 1997 and 2001, through volunteers responding to a local newspaper advertisement and a preventive campaign on cardiovascular disease and osteoporosis risk factors in women held at multiple public and work places [15]. After informed consent was obtained, participants answered a detailed questionnaire on osteoporosis risk factors derived from the Mediterranean osteoporosis (MEDOS) study questionnaire [29]. Postmenopausal women were defined as women not having menses in the last 12 months. Women were eligible for the present study if they were French-Canadian for three generations, had their medical history, lifestyle habits, environmental information, anthropometric measures, bone measures (only half of the women came to have a DXA measure) and blood sample collected. Women were excluded from the study if they had a medical condition affecting bone homeostasis (nutritional disorder, alcoholism, phosphocalcic metab- olism disease, degenerative bone disease, hepatic disease or renal insufficiency) or had used medication known to influence bone metabolism (etidronate disodium, alendronate, calcitonine) other than past use of oral contraceptives and hormone therapy. After all exclusions, 2110 women with DXA measures were used in the analyses. Premenopausal women (N = 709) were aged between 18 and 58 years. Postmenopausal women (N = 1401) were aged between 33 and 84 years (Table 1). Toronto sample Between 1995 and 1997, women between 18 and 35 years of age were recruited through advertisements in local newspapers and posted flyers [30,31]. They were screened by telephone questionnaire before enrolment. Of the 993 subjects assessed, some were excluded because of age or comorbid conditions known to be associated to bone loss (nutritional disorders, alcoholism, pre-existing disorders of bone and mineral metabolism, degenerative bone disease, hepatic disease or renal insufficiency) or because they had undergone a bilateral oophorectomy. Further sample selection was applied so that only unrelated women with European ancestry were included. In the end, clinical data and DNA samples were available on 673 subjects (Table 1). The Ethics Review Office of the University of Toronto approved the study protocol. After obtaining their written consent, each subject completed a standardized questionnaire about lifestyle factors and the variables were categorized in the same way as the data for the previously reported Quebec sample [21]. DXA measurements BMD was determined at the lumbar spine from levels L2 to L4 inclusively (LS BMD; g/cm2) and at the femoral neck (FN BMD; g/ cm2) by dual-energy X-ray absorptiometry (DXA) (software version Table 1 Characteristics of premenopausal and postmenopausal women. Quebec N = 709 Toronto N = 673 Quebec N = 1401 Initial premenopausal women Replication premenopausal women Post menopausal Continuous variables Mean ± SD Mean ± SD Mean ± SD Age (years) range Weight (kg) Lumbar spine BMD (g/cm2) Lumbar spine Z-score Femoral neck BMD (g/cm2) Femoral neck Z-score 44.5 ± 7.2 18–58 63 ± 11 1.177 ± 0.142 27.5 ± 4.5 18–35 63 ± 12 1.192 ± 0.132 57.1 ± 7.3 33–84 66 ± 12 1.096 ± 0.171 0.094 ± 1.136 0.92 ± 0.13 0.0765 ± 1.036 1.007 ± 0.122 0.203 ± 1.31 0.872 ± 0.136 −0.091 ± 1.000 0.237 ± 0.954 0.077 ± 1.006 Categorical variables Number (%) Number (%) Number (%) Smoking -never -ever -current Physical activity b 1 activity/week 1 to 2 activities/week ≥ 3 activities/week Age at menarche b 12 years 12–13 years ≥ 14 years Hormone therapy use ≥ 5 years 1–4 years never 419 (59) 192 (27) 98 (14) 478 (71) 114 (17) 81 (12) 781 (55.7) 447 (31.9) 173 (12.3) 186 (26.2) 239 (33.7) 284 (40.1) 59 (9) 113 (17) 501 (74) 355 (25.3) 384 (27.4) 662 (47.2) 149 (21) 340 (48) 220 (31) 106 (16) 390 (58) 177 (26) 270 (19.3) 689 (49.2) 442 (31.5) (0) (0) (100) (0) (0) (100) 461 (32.7) 477 (33.8) 463 (32.8) S. Giroux et al. / Bone 47 (2010) 975–981 3.2; DPX-L Lunar radiation Corp., Madison, WI, USA). All BMD measurements were performed by a trained technician from the Service of Nuclear Medicine of the CHUQ, Hôpital St-François d'Assise and interpreted by a nuclear medicine physician. The long-term reproducibility evaluated on a daily basis using a standard bone phantom consistently showed a coefficient of variation (CV) b 1%. In the Toronto sample, BMD was measured at the lumbar spine from levels L2 to L4 inclusively (LS BMD; g/cm2) and at the femoral neck (FN BMD; g/cm2), as in the Quebec sample using the same type of densitometry instrumentation (DPX-L Absorptiometer, Lunar Corporation, Madison WI, USA; software version 3.2). The CV was 1.2% for the spine and 1.3% for the femur. Candidate gene selection and SNP selection Selected genes, gene ID, chromosomal location and sizes are indicated in Table 2. All the genotype data from 30 trios of European ancestry covering the 23 gene sequences plus 5 kilobases in 5' upstream were downloaded from the (data from February 2007) International HapMap project (http://www.hapmap.org/). We used Haploview 4.1 to select SNPs from each chromosomal region [32] with the algorithm Tagger. SNPs were considered if minor allele frequency was N 5%, at least 75% of individuals were genotyped and Hardy– Weinberg equilibrium was respected (p-value N 0.01). Tagger pairwise was run with an r2 N 0.8 to select the Tag SNPs. 231 SNPs were necessary to capture the genetic diversity of the 23 genes (Table 2). 977 This way, 6 SNPs could be recovered with a call rate N 99%. In the end, two SNPs failed and could not be recovered with another SNP, one in IL1RN gene and one in CALCR gene and eight SNPs were not in HWE (p b 0.01). The error rate of this technology is lower than 0.5% according to duplicated and control samples tested at the Sequenom platform at Génome Québec Innovation centre. The sample of premenopausal women from Toronto (673 women) was genotyped for 23 SNPs (associated with bone density in the Québec City premenopausal sample) using the Sequenom Technology at the McGill University and Génome Québec Innovation Centre, Montréal, Canada. Four SNPs failed and call rate was N 98% for the remaining 19 SNPs. These four SNPs were reanalyzed with TaqMan technology at the same centre with success and a call rate N 98%. The sample of postmenopausal women (1401 women) was genotyped for the top two associated SNPs (in both premenopausal samples) using allele-specific PCR assays developed in-house as described in [15]. The primers used to genotype rs1534882 were 5'CAGCCTTCACCTACCCGACTC3'—common, 5'GCCTGCACTGCGTTCCTA3'—minor allele and 5'GCCTGCACTGCGTTCCTG3'—major allele. The primers used to genotype rs3917989 were 5'CAGCCTTTCTTGATCCTTC3'—major allele, 5'CAGCCTTTCTTGATCCTTT3'—minor allele and 5'CAGGGCTGGAAGTATGGTAGG3'—common. All reactions were performed in a final volume of 15 μl with about 25 ng DNA and HotStart Taq DNA polymerase from Qiagen. Call rate was N 99% and error rate b 2%. Statistical analyses Genotyping The sample of premenopausal women from Quebec (709 women, Table 1) was genotyped for the 231 selected SNPs using the Sequenom Iplex Gold technology at the McGill University and Génome Québec Innovation Centre, Montréal, Canada. Call rate was above 95% for 99% SNPs, the lowest call rate was 92.2%. Eight SNPs were not in Hardy– Weinberg equilibrium (p b 0.01) and 7 failed. These SNPs were tested a second time in a different panel using the same technology and when possible, a synonymous SNP (correlated N 80%) was also tested. Population structure analysis was performed for the three samples (premenopausal Québec, premenopausal Toronto and postmenopausal Québec) using Structure software [33]. For each SNP, Hardy–Weinberg equilibrium (HWE) was tested using a standard Chisquare test comparing the expected and observed allele frequencies. Lumbar spine (LS) BMD and femoral neck (FN) BMD were tested separately for association with each genetic marker in each sample by analysis of covariance (ANCOVA) adjusted for the environmental variables (age, weight, smoking, age at menarche and physical Table 2 List of genes analyzed in this study. Candidate gene Genes induced by estrogen and level of expression correlated with bone density in mouse. Product IL1RN IL1R2 IGFBP4 CSF3R IBSP CSF2RB IL4R TNFRSF1A Genes with a role in Vitamin D and calcium metabolism. LIFR TGFB1 PTHR1 PTH CALCA CALCR CASR KL CYP2R1 CYP3A4 CYP27A1 CYP27B1 CYP24A1 GPRC6A GC Total interleukin 1 receptor antagonist interleukin 1 receptor, type II insulin-like growth factor binding protein 4 colony stimulating factor 3 receptor integrin-binding sialoprotein colony stimulating factor 2 receptor, beta 1, low-affinity interleukin 4 receptor, alpha tumor necrosis factor receptor superfamily, member 1A leukemia inhibitory factor receptor alpha Transforming growth factor, beta-1 parathyroid hormone receptor 1 Parathyroid hormone calcitonin/calcitonin-related polypeptide, alpha Calcitonin receptor calcium-sensing receptor klotho Vitamin D-hydroxylase Cytochrome p450, family 3 Vitamin D- hydroxylase Vitamin D- hydroxylase Vitamin D- hydroxylase G protein-coupled receptor, C6A Vitamin-D binding protein 23 genes Gene ID Chrom. location Size #Tag SNP Failed SNP 3557 7850 3487 1441 3381 1439 2q14 2q12–q22 17q12–q21 1p35 4q21 22q13 21 kb 42 kb 19 kb 22 kb 17 kb 22 kb 11 13 4 8 7 13 1 0 0 0 0 0 3566 7132 16p11 12p13 56 kb 18 kb 21 5 0 0 3977 7040 5745 5741 796 5p13 19q13 3p22 11p15 11p15 125 kb 28 kb 31 kb 9 kb 11 kb 7 6 4 4 1 0 0 0 0 0 799 846 9365 120227 1576 1593 1594 1591 222545 2638 7q21 3q13 13q12 11p15 7q21 2q33 12q13 20q13 6q22 4q11 155 kb 108 kb 55 kb 19 kb 32 kb 38 kb 10 kb 26 kb 42 kb 47 kb 953 kb 28 28 15 8 3 3 3 22 6 11 231 1 0 0 0 0 0 0 0 0 0 2 HWE failed 1 1 1 2 1 2 8 978 S. Giroux et al. / Bone 47 (2010) 975–981 activities). When the two premenopausal samples were combined, a variable for the origin (Québec or Toronto) was included in the model. In the subgroup of post-menopausal women, hormonal therapy (HRT used for less than 5 years, more than 5 years and never) in three categories was added in the model. The analysis with all the women combined included a variable for menopausal status, origin, age, weight, smoking, age at menarche and physical activities. The alphalevel was set at alpha = 0.05 for all associations given the context of this hypothesis-driven study and the sequential confirmation in two independent samples. To control the false discovery rate at 0.05, we applied the procedure described by Benjamini and Hochberg [34] on the results of the combined analysis (two premenopausal samples). To calculate the power in the postmenopausal women group we used the genetic effect size and the genotype frequencies observed in the premenopausal women group which were applied to the postmenopausal sample size expecting an alpha of 0.05 (Zβ = [effect size in premenop / SD in postmenopausal × square root of number carrying genotype 1 × number of carriers genotype 2/1401 ] − Zα/2. The analyses were performed using statistical software packages SAS 9.2 (SAS Institute, Cary, NC, USA) and SPSS version 11.0 for MAC (SPSS an IBM company, Chicago, Illinois, USA). Results ANCOVA statistical analysis was performed with the first sample of 709 premenopausal women from Québec and 221 SNPs in HWE. For each SNP, two analyses were independently performed with LS and FN BMD for a total of 442 tests. Therefore, 22 significant associations could emerge by chance (α = 0.05). We obtained 25 positive associations with one bone measure or the other (Table 3). Two SNPs gave a positive result with both skeletal sites (FN and LS); therefore 23 SNPs were tested in the sample from Toronto. ANCOVA statistical analysis was performed with the combined sample of 1362 premenopausal women from Toronto and Québec and 23 SNPs (Tables 4a and 4b) in HWE. Among the 46 analyses performed, seven were significantly (p b 0.05) associated with femoral neck (FN) BMD and six with lumbar spine (LS) BMD. However, only five with FN BMD and two with LS BMD had a p-value lower than the Table 3 Positive results of ANCOVA analysis with Quebec sample. Table 4a Associations observed with femoral neck BMD in combined sample. p-values smaller than those first observed in the Quebec sample are shaded. * adjusted means. p-value first observed in the Québec sample only (shaded in Tables 4a and 4b) indicating an effect in the same direction in the sample from Toronto. One SNP (rs1534882) was located in CSF2RB gene coding for the beta subunit of granulocyte macrophage colony stimulating factor (GM-CSF) receptor, two SNPs (rs3917989 and rs 3917981) were located in CSF3R gene coding for granulocyte colony stimulating factor (G-CSF) receptor, two SNPs (rs10500804 and rs4674344) were in genes involved in the biosynthesis of vitamin D (CYP2R1 and CYP27A1) and the last one (rs2234898) was located in IL4R gene coding for the interleukin-4 receptor. When controlling the FDR at the 0.05 level, none of these SNPs could be declared significantly associated. In order to have more power to observe an effect in the postmenopausal women group, we combined the genotypes being associated with a similar mean BMD according to a recessive or a dominant model of transmission for the six associated SNPs (Table 5). With the same gene effect size observed in premenopausal women as well as the same genotype frequencies, we calculated the power we had to observe an effect of that size in our sample of 1401 Table 4b Associations observed with lumbar spine BMD in combined sample. Associations with p-value b 0.05 are shaded. p-values smaller than those first observed in the Quebec sample are shaded. * adjusted means. S. Giroux et al. / Bone 47 (2010) 975–981 979 Table 5 Adjusted means observed in premenopausal women according to a recessive or dominant model of transmission and calculated power for α level = 0.05 in the postmenopausal women sample. NA, not applicable given that it was not significantly associated in premenopausal women. Power N 70% is shaded. * adjusted means (age, weight, age at the menarche, smoking habits and level of physical activities). We were aware that this procedure would reduce power to detect small genetic effect but it was cost-effective. These two independent samples of premenopausal women were previously studied with LRP5 variant and the Val667Met was associated with LS BMD exactly in the same manner; same gene effect magnitude, same frequency and same size of p-value [21]. Although the women in the Toronto sample were much younger and much more active than those from Quebec, no statistically significant difference was observed between these two samples for their adjusted BMD at both skeletal sites. Also, the structure analysis did not detect significant stratification in any of the three samples (premenop Québec, premenop Toronto and postmenop Québec). No SNP was significantly associated in the combined sample of premenopausal women (n = 1362) when controlling the FDR at the 0.05 level. The two most promising SNPs were tested in the postmenopausal women sample (n = 1401) and only one in CSFR3 was also associated in that sample. A tetranucleotide repeat in PTHR1 gene promoter had previously been found associated with height and femoral neck BMD in the same sample of young women from Toronto [35]. While our study did not include this type of variation, none of the TagSNPs selected in the PTHR1 gene showed any potential association with either LS or FN BMD in the sample of premenopausal women from Quebec. It is also possible that no SNP could tag the tetranucleotide repeat. Similarly, Vilarino-Güell et al. reported significant associations between the tetranucleotide repeat and LS BMD as well as haplotypes of PTHR1 with BMD in the youngest tertile of their population [36]. TGFB1 gene was also studied in large samples and was not found associated with bone mineral density [37–39] as we observed with the 709 premenopausal women from Québec. Recently, a collaborative metaanalysis reported associations with 150 candidate genes [23]. Nine postmenopausal women with an α level = 0.05 (Table 5). We next analyzed the two most promising SNPs (rs1534882 and rs3917989) in this independent sample of postmenopausal women (n = 1401) in which we had a 70% chance of observing the effect if it is a true association. Only rs3917989 was associated in this third sample and the association was in the same direction as for premenopausal women (Table 6). We observed a significant association with FN BMD (p = 0.023) and a trend with LS BMD (p = 0.094). As hypothesized the gene effect size was larger in premenopausal women (effect/ SD = 0.02/0.13 = 0.154 SD) compared to postmenopausal women (0.015/0.136 = 0.11 SD). After combining the premenopausal and postmenopausal women a highly significant association was observed with FN BMD and to a lesser extent with LS BMD (Table 6). Discussion In this study, we analyzed 23 candidate genes with high-density polymorphisms for association with bone mineral density in women. We focused on bone mineral density, and not fracture risk, because BMD is a well characterized and strongly heritable phenotype. 231 SNPs were selected to densely cover the genetic variation of these genes and 221 of those SNPs were in Hardy–Weinberg equilibrium. A poor man's approach was used for the genotyping procedure. We first divided the sample according to the menopausal status given that the accelerated loss observed during the perimenopausal period is critical in women [4] and that many studies have shown that peak bone mass was highly heritable and more important in younger ages [5]. Therefore, we focused on premenopausal women closer to their peak bone mass to discover new variants. In addition, two independent samples of premenopausal women were analyzed sequentially. Table 6 Ancova analysis in the sample of postmenopausal women and all samples combined with SNP rs 3917989. Premenop Postmenopb Combinedc pre and post a b c Number in a dominant model FN BMD meansa 95% confidence interval p-value LS BMD meansa 95% confidence interval p-value 612 750 625 774 1237 1524 0.962 0.942 0.870 0.855 0.915 0.899 0.952 0.932 0.859 0.846 0.907 0.892 0.001 1.189 1.168 1.095 1.081 1.145 1.128 1.177 1.157 1.081 1.069 1.136 1.119 0.003 to to to to to to 0.973 0.952 0.880 0.865 0.922 0.906 0.023 0.00039 Adjusted means. In the model with postmenop women only, HRT use in 3 categories was included. In the combined sample no HRT adjustment could be made but the means were adjusted for menopausal status. to to to to to to 1.202 1.180 1.109 1.094 1.154 1.137 0.094 0.002 980 S. Giroux et al. / Bone 47 (2010) 975–981 candidate genes included in the present study (IL1RN, TGFB1, PTHR1, PTH, CALCA, CALCR, KL and GC) were also studied in the large metaanalysis and no association was observed [23]. Only two polymorphisms were analysed in postmenopausal women because we expected to have less power in that sample to observe an effect given that women have already started to lose bone mineral density through their transition to menopause and that gene effect might be masked by the use of hormonal therapy. Only the CSF3R variant was associated among postmenopausal women. The associated SNP (rs3917989) was located in the intron 11 of the gene coding for cytokine granulocyte colony-stimulating factor receptor (GCSFR or CSF3R as official nomenclature). CSF3R was studied because it was shown that its expression was induced by estrogen in bone cells and the expression level was correlated (r2 = 0.76) with trabecular bone density in young mice [26]. G-CSF, the ligand of CSF3R, plays a crucial role in the production and function of neutrophilic granulocytes, white blood cells having an essential role against infection [40]. The cytokine is able to mobilize various precursor cells, stimulate the proliferation and differentiation of cells along the neutrophilic lineage and also activate the functions of mature neutrophils. This cytokine is widely used in the treatment of neutropenia due to congenital defect (in severe congenital neutropenia) or to other neutropenic conditions associated with chemotherapy and bone marrow transplantation [40]. The biological effects are mediated through the CSF3 receptor, a specific cell surface receptor, and a member of the hematopoietin receptor superfamily [41]. In addition, it was shown that in vitro, G-CSF (CSF3) and granulocyte macrophage-CSF (GM-CSF), although less effective, could replace macrophage-CSF (M-CSF) in the induction of late monocytic cell along the osteoclast pathway [42]. Monocytes/ macrophages and osteoclasts, but not osteoblasts, express the CSF3 receptor [43]. In humans, it was shown that patients treated over a long period of time with G-CSF develop marked osteopenia due to an increased bone resorption [44] and transgenic mice overexpressing G-CSF have increased numbers of osteoclasts and develop osteoporosis [45]. It was also shown that conditional inactivation of TNF-α converting enzyme (TACE) in mice is associated with disregulated G-CSF expression that is causally related to both osteoporosis-like phenotype and defects in the hematopoietic system [46]. Given all these observations, CSF3R could indeed play a role in the determination of bone density in the general healthy population. It is not clear how the SNP located in intron 11 could impact the function of CSF3R. The SNP is located in a large block of linkage disequilibrium extending some 100 kb in the 3' end of the gene where many other SNPs are correlated with rs3917989. Therefore, any of those could be the functional SNP that could impact the binding of some regulatory proteins. Only well-designed assays testing each SNP in linkage disequilibrium would allow identification of the functional one that would lead to the underlying mechanism. Replication in independent premenopausal and postmenopausal samples should be undertaken to confirm the associations observed. Conflict of interest statement All authors have no conflicts of interest. Acknowledgments Financial support was provided by the Canadian Institutes for Health Research and the Canadian Genetic Diseases Network, Networks of Centres of Excellence program. Infrastructure support to the various research centers was provided by the Fonds de Recherche en Santé du Quebec. François Rousseau holds a Fonds de la Recherche en Santé du Québec / MSSS Research Chair in Health Technology Assessment and Evidence Based Laboratory Medicine, Alexandre Bureau a scientist award, and Latifa Elfassihi a doctoral award. References [1] Ferrari S, Rizzoli R, Slosman D, Bonjour JP. Familial resemblance for bone mineral mass is expressed before puberty. J Clin Endocrinol Metab 1998;83:358–61. [2] Pocock NA, Eisman JA, Hopper JL, Yeates MG, Sambrook PN, Eberl S. Genetic determinants of bone mass in adults. A twin study. J Clin Invest 1987;80:706–10. [3] Smith DM, Nance WE, Kang KW, Christian JC, Johnston Jr CC. Genetic factors in determining bone mass. J Clin Invest 1973;52:2800–8. [4] Berger C, Langsetmo L, Joseph L, Hanley DA, Davison KS, Josse R, et al. Change in bone mineral density as a function of age in women and men and association with the use of antiresorptive agents. CMAJ 2008;178:1660–8. [5] Brown LB, Streeten EA, Shapiro JR, McBride D, Shuldiner AR, Peyser PA, et al. Genetic and environmental influences on bone mineral density in pre- and postmenopausal women. Osteoporos Int 2005;16:1849–56. [6] Ioannidis JP, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG. Replication validity of genetic association studies. Nat Genet 2001;29:306–9. [7] Liu YZ, Liu YJ, Recker RR, Deng HW. Molecular studies of identification of genes for osteoporosis: the 2002 update. J Endocrinol 2003;177:147–96. [8] Ioannidis JPA. Why most published research findings are false. PLoS Med 2005;2. [9] Shen H, Liu Y, Liu P, Recker RR, Deng HW. Nonreplication in genetic studies of complex diseases—lessons learned from studies of osteoporosis and tentative remedies. J Bone Miner Res 2005;20:365–76. [10] Giguere Y, Rousseau F. The genetics of osteoporosis: “complexities and difficulties”. Clin Genet 2000;57:161–9. [11] Koller DL, Ichikawa S, Johnson ML, Lai D, Xuei X, Edenberg HJ, et al. Contribution of the LRP5 gene to normal variation in peak BMD in women. J Bone Miner Res 2005;20:75–80. [12] Koay MA, Woon PY, Zhang Y, Miles LJ, Duncan EL, Ralston SH, et al. Influence of LRP5 polymorphisms on normal variation in BMD. J Bone Miner Res 2004;19: 1619–27. [13] Ferrari SL, Deutsch S, Choudhury U, Chevalley T, Bonjour JP, Dermitzakis ET, et al. Polymorphisms in the low-density lipoprotein receptor-related protein 5 (LRP5) gene are associated with variation in vertebral bone mass, vertebral bone size, and stature in whites. Am J Hum Genet 2004;74:866–75. [14] van Meurs JB, Rivadeneira F, Jhamai M, Hugens W, Hofman A, van Leeuwen JP, et al. Common genetic variation of the low-density lipoprotein receptor-related protein 5 and 6 genes determines fracture risk in elderly white men. J Bone Miner Res 2006;21:141–50. [15] Giroux S, Elfassihi L, Cardinal G, Laflamme N, Rousseau F. LRP5 coding polymorphisms influence the variation of peak bone mass in a normal population of French-Canadian women. Bone 2007;40:1299–307. [16] Ezura Y, Nakajima T, Urano T, Sudo Y, Kajita M, Yoshida H, et al. Association of a single-nucleotide variation (A1330V) in the low-density lipoprotein receptorrelated protein 5 gene (LRP5) with bone mineral density in adult Japanese women. Bone 2007;40:997–1005. [17] Saarinen A, Valimaki VV, Valimaki MJ, Loyttyniemi E, Auro K, Uusen P, et al. The A1330V polymorphism of the low-density lipoprotein receptor-related protein 5 gene (LRP5) associates with low peak bone mass in young healthy men. Bone 2007;40:1006–12. [18] Bollerslev J, Wilson SG, Dick IM, Islam FM, Ueland T, Palmer L, et al. LRP5 gene polymorphisms predict bone mass and incident fractures in elderly Australian women. Bone 2005;36:599–606. [19] Koh JM, Jung MH, Hong JS, Park HJ, Chang JS, Shin HD, et al. Association between bone mineral density and LDL receptor-related protein 5 gene polymorphisms in young Korean men. J Korean Med Sci 2004;19:407–12. [20] van Meurs JB, Trikalinos TA, Ralston SH, Balcells S, Brandi ML, Brixen K, et al. Largescale analysis of association between LRP5 and LRP6 variants and osteoporosis. JAMA 2008;299:1277–90. [21] Giroux S, Elfassihi L, Cole DE, Rousseau F. Replication of associations between LRP5 and ESRRA variants and bone density in premenopausal women. Osteoporos Int 2008;19:1769–75. [22] Richards JB, Rivadeneira F, Inouye M, Pastinen TM, Soranzo N, Wilson SG, et al. Bone mineral density, osteoporosis, and osteoporotic fractures: a genome-wide association study. Lancet 2008;371:1505–12. [23] Richards JB, Kavvoura FK, Rivadeneira F, Styrkarsdottir U, Estrada K, Halldorsson BV, et al. Collaborative meta-analysis: associations of 150 candidate genes with osteoporosis and osteoporotic fracture. Ann Intern Med 2009;151:528–37. [24] Rivadeneira F, Styrkarsdottir U, Estrada K, Halldorsson BV, Hsu YH, Richards JB, et al. Twenty bone-mineral-density loci identified by large-scale meta-analysis of genome-wide association studies. Nat Genet 2009;41:1199–206. [25] Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, Gudbjartsson DF, Walters GB, Ingvarsson T, et al. Multiple genetic loci for bone mineral density and fractures. N Engl J Med 2008;358:2355–65. [26] Lindberg MK, Moverare S, Eriksson AL, Skrtic S, Gao H, Dahlman-Wright K, et al. Identification of estrogen-regulated genes of potential importance for the regulation of trabecular bone mineral density. J Bone Miner Res 2002;17:2183–95. [27] Tsujikawa H, Kurotaki Y, Fujimori T, Fukuda K, Nabeshima Y. Klotho, a gene related to a syndrome resembling human premature aging, functions in a negative regulatory circuit of vitamin D endocrine system. Mol Endocrinol 2003;17:2393–403. [28] Giroux S, Rousseau F. Genes and osteoporosis: time for a change in strategy. Int J Clin Rheumatol 2009;4:221–33. [29] Dequeker J, Tobing L, Rutten V, Geusens P. Relative risk factors for osteoporotic fracture: a pilot study of the MEDOS questionnaire. Clin Rheumatol 1991;10:49–53. [30] Jamal SA, Ridout R, Chase C, Fielding L, Rubin LA, Hawker GA. Bone mineral density testing and osteoporosis education improve lifestyle behaviors in premenopausal women: a prospective study. J Bone Miner Res 1999;14:2143–9. S. Giroux et al. / Bone 47 (2010) 975–981 [31] Rubin LA, Hawker GA, Peltekova VD, Fielding LJ, Ridout R, Cole DE. Determinants of peak bone mass: clinical and genetic analyses in a young female Canadian cohort. J Bone Miner Res 1999;14:633–43. [32] Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005;21:263–5. [33] Pritchard JK, Stephens M, Donnelly P. Inference of population structure using multilocus genotype data. Genetics 2000;155:945–59. [34] Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc 1995;B57:289–300. [35] Scillitani A, Jang C, Wong BY, Hendy GN, Cole DE. A functional polymorphism in the PTHR1 promoter region is associated with adult height and BMD measured at the femoral neck in a large cohort of young Caucasian women. Hum Genet 2006;119:416–21. [36] Vilarino-Guell C, Miles LJ, Duncan EL, Ralston SH, Compston JE, Cooper C, et al. PTHR1 polymorphisms influence BMD variation through effects on the growing skeleton. Calcif Tissue Int 2007;81:270–8. [37] McGuigan FE, Macdonald HM, Bassiti A, Farmer R, Bear S, Stewart A, et al. Largescale population-based study shows no association between common polymorphisms of the TGFB1 gene and BMD in women. J Bone Miner Res 2007;22: 195–202. [38] Langdahl BL, Uitterlinden AG, Ralston SH, Trikalinos TA, Balcells S, Brandi ML, et al. Large-scale analysis of association between polymorphisms in the transforming growth factor beta 1 gene (TGFB1) and osteoporosis: the GENOMOS study. Bone 2008;42:969–81. 981 [39] Hubacek JA, Weichetova M, Bohuslavova R, Skodova Z, Stepan JJ, Adamkova V. No associations between genetic polymorphisms of TGF-beta, PAI-1, and COL1A1, and bone mineral density in Caucasian females. Endocr Regul 2006;40:107–12. [40] Ward AC. The role of the granulocyte colony-stimulating factor receptor (G-CSF-R) in disease. Front Biosci 2007;12:608–18. [41] van de Geijn GJ, Aarts LH, Erkeland SJ, Prasher JM, Touw IP. Granulocyte colonystimulating factor and its receptor in normal hematopoietic cell development and myeloid disease. Rev Physiol Biochem Pharmacol 2003;149:53–71. [42] Takahashi N, Udagawa N, Akatsu T, Tanaka H, Shionome M, Suda T. Role of colonystimulating factors in osteoclast development. J Bone Miner Res 1991;6:977–85. [43] Hirbe AC, Uluckan O, Morgan EA, Eagleton MC, Prior JL, Piwnica-Worms D, et al. Granulocyte colony-stimulating factor enhances bone tumor growth in mice in an osteoclast-dependent manner. Blood 2007;109:3424–31. [44] Sekhar RV, Culbert S, Hoots WK, Klein MJ, Zietz H, Vassilopoulou-Sellin R. Severe osteopenia in a young boy with Kostmann's congenital neutropenia treated with granulocyte colony-stimulating factor: suggested therapeutic approach. Pediatrics 2001;108:E54. [45] Takahashi T, Wada T, Mori M, Kokai Y, Ishii S. Overexpression of the granulocyte colony-stimulating factor gene leads to osteoporosis in mice. Lab Invest 1996;74: 827–34. [46] Horiuchi K, Kimura T, Miyamoto T, Miyamoto K, Akiyama H, Takaishi H, et al. Conditional inactivation of TACE by a Sox9 promoter leads to osteoporosis and increased granulopoiesis via dysregulation of IL-17 and G-CSF. J Immunol 2009;182: 2093–101.