Abstract
Context: Several genome-wide association studies (GWAS) have been performed to identify genes contributing to bone mineral density (BMD), typically in samples of elderly women and men.
Objective: The objective of the study was to identify genes contributing to BMD in premenopausal women.
Design: GWAS using the Illumina 610Quad array in premenopausal European-American (EA) women and replication of the top 50 single-nucleotide polymorphisms (SNPs) for two BMD measures in African-American (AA) women.
Subjects: Subjects included 1524 premenopausal EA women aged 20–45 yr from 762 sibships and 669 AA premenopausal women aged 20–44 yr from 383 sibships.
Interventions: There were no interventions.
Main Outcome Measures: BMD was measured at the lumbar spine and femoral neck by dual-energy x-ray absorptiometry. Age- and weight-adjusted BMD values were tested for association with each SNP, with P values determined by permutation.
Results: SNPs in CATSPERB on chromosome 14 provided evidence of association with femoral neck BMD (rs1298989, P = 2.7 × 10−5; rs1285635, P = 3.0 × 10−5) in the EA women, and some supporting evidence was also observed with these SNPs in the AA women (rs1285635, P = 0.003). Genes identified in other BMD GWAS studies, including IBSP and ADAMTS18, were also among the most significant findings in our GWAS.
Conclusions: Evidence of association to several novel loci was detected in a GWAS of premenopausal EA women, and SNPs in one of these loci also provided supporting evidence in a sample of AA women.
Evidence of association to several novel loci was detected in genome-wide association studies of premenopausal European-American women, and SNPs in one of these loci also provided supporting evidence in a sample of African-American women.
Bone mineral density (BMD) is a complex quantitative trait that involves the interaction of multiple genetic and environmental components (1). Although environmental factors including nutrition and physical activity influence the peak BMD attained in early adulthood, there is strong evidence for a major genetic contribution to normal variation in BMD (1), with estimates suggesting that genetic factors may account for as much as 80% of the variability in peak BMD (1,2,3).
Studies in both humans and animal models have consistently found that unique quantitative trait loci contribute to BMD at the hip and spine (4,5,6,7,8). Furthermore, segregation studies have yielded evidence of sex-specific genetic regulation of BMD (9), and a number of sex-specific quantitative trait loci have been reported in mice (10,11) and suggested in humans (12,13,14). There are also racial differences in both peak BMD and fracture risk; African-Americans (AAs) on average have higher peak BMD and lower fracture rates than Caucasians (1).
Several genome-wide association studies (GWAS) have identified genes contributing to BMD variation (15,16,17,18,19,20,21). Several studies identified single-nucleotide polymorphisms (SNPs) near the estrogen receptor 1 locus (ESR1) (15,17) with an effect on BMD along with LRP5 and LRP4 from the Wnt signaling pathway (15,16,21). Loci from the RANK/OPG pathway were identified as associated with BMD in one or more studies (16,17,20,21); these included SNPs in or near RANKL (17,20); osteoprotegerin (TNFRSF11B; 16,17), and RANK (17,21). Two different studies found significant association with SNPs in or near distinct disintegrin/metallopeptidase genes, ADAM19 (21) and ADAMTS18 (18). Genes known to be involved in bone were identified by at least one of the published GWAS, including COL1A1 and PPARG (15), SP7/osterix and IBSP (19,21), and CLCN1 (20).
Four of the previous GWAS included cohorts with mean age 50 yr or older (15,17,20,21). The Twins U.K. sample included only women (16), the Icelandic sample was more than 85% women (17), and the Framingham population included men and women in approximately equal proportions (15). The study by Xiong et al. (18) contained equal numbers of men and women, with pre- and postmenopausal women in equal numbers. In this study, we limited our sample to premenopausal European American (EA) women, thereby allowing us to identify genes contributing to peak BMD, a highly heritable phenotype related to osteoporosis. We then genotyped our most significant SNPs in an independent replication sample of AA women to test for race-specific associations.
Subjects and Methods
Subjects
To identify genes contributing to BMD, a study of healthy siblings from Indiana was initiated (4,5). Recruitment focused on families with two or more healthy sisters. Sisters were required to be within 10 yr of one another in age. More recently recruitment was expanded to include healthy women who met all study recruitment criteria but did not have a sibling eligible or desiring to participate.
Exclusion criteria included a history of chronic disease, taking medications known to affect bone mass or metabolism, inability to have BMD measured because of obesity (weight >136 kg), and abnormal blood biochemistry tests. Women who had irregular menses or a history of pregnancy or lactation within 3 months before enrollment were excluded; however, women taking oral contraceptives were not excluded. Health and lifestyle questionnaires were administered and anthropometric variables measured. A blood sample was collected for DNA. BMD was measured by dual-energy x-ray absorptiometry (DPXL and Prodigy; Lunar Corp., Madison, WI) at lumbar spine (vertebrae 2–4) and hip (femoral neck). Image analysis was performed using Lunar software versions 4.6/4.7. Coefficient of variation measured in 230 subjects who had duplicate dual-energy x-ray absorptiometry measures made after they were repositioned on the instrument was 1.0% for femoral neck and 0.5% for lumbar spine.
Studies were performed at the General Clinical Research Center of Indiana University School of Medicine. Studies were approved by the Institutional Review Board of Indiana University-Purdue University Indianapolis. Informed written consent was obtained from all subjects before their participation in the study.
Microarray genotyping and quality assessment
A sample of 776 sibships of EA sisters (n = 1,552 individuals) were originally selected for genotyping as part of a GWAS. For sibships of size larger than 2, the most informative pair of sisters was selected based on maximal discordance for lumbar spine or femoral neck BMD. In some cases, sample depletion or replacement of samples performing poorly in initial testing resulted in singletons (n = 32) and three-sister sibships (n = 32 families) in the final genotyped sample. Genotyping was performed on the Illumina Human610Quadv1_B BeadChips (Illumina, San Diego, CA) by the Center for Inherited Disease Research using the Illumina Infinium II assay protocol (22). This array contains 592,532 markers with a mean spacing of 5.8 kb. Allele cluster definitions for each SNP were determined using Illumina BeadStudio genotyping module version 3.2.32 and the combined intensity data from 100% of the study samples. The resulting cluster definition file was used on all study samples to determine genotype calls and quality scores. Genotype calls were made when a genotype yielded a quality metric (Gencall score) of 0.15 or higher. The final raw data set released by the Center for Inherited Disease Research to the investigators and to the Genotype and Phenotype database (http://www.ncbi.nlm.nih.gov/gap) contained called genotypes for 589,171 SNPs and 1544 unique DNA samples, with eight samples not released due to inadequate quality of genotyping. Blind duplicate reproducibility was 99.99% based on 36 paired samples.
Additional sample and SNP quality control measures were then applied. Samples having genotypes for at least 98% of the SNPs were considered for inclusion in analyses. These samples were rigorously checked for previously undetected family relationships and population stratification. Samples were removed from further analyses and from the final dbGaP data set due to familial relationships that could not be unambiguously determined (n = 14). Because this sample included sisters, all relationships were confirmed using identity-by-state clustering in PLINK (23). Nineteen pairs of sisters were found not to be full siblings as ascertained but rather were half-siblings; the family structure was altered to reflect the new information. These half-sisters were retained for the association analysis; however, their data are not included in the final data set available at dbGaP (http://www.ncbi.nlm.nih.gov/sites/entrez?db=gap; accession no. phs000138.v1.p1). A principal component-based analysis was performed in PLINK to cluster these samples along with HapMap reference samples (Utah, Ibadan, Beijing, and Tokyo) to verify the samples included for analysis were of European ancestry. Three pairs of sisters (n = 6 samples) were removed from further analysis due to evidence of substantial admixture with other populations. Our final analytic sample included 1524 women from 762 different families.
SNPs with a call rate of 98% or greater (n = 581,255) were included and subjected to further quality control analyses. From these, SNPs were removed if either the minor allele frequency was less than 0.01 in this data set (n = 32,948) or there was significant deviation (P < 0.00001) from Hardy Weinberg equilibrium (n = 1,998). The final data set for analysis consisted of 547,971 SNPs that passed all quality control measures.
Statistical analysis of genome-wide association data
Stepwise regression analysis was performed to test which measured covariates (height, weight, oral contraceptive use in women, pack-years of smoking, and age) were significant covariates accounting for a substantial proportion of the variation in BMD. Only age and weight, which together accounted for 10 and 18% of the variation in spine and neck BMD, respectively, were included in the final analytic model. Regression residuals, representing age- and weight-adjusted BMD values, were used in all analyses.
The initial analysis used a population-based association test using a linear regression framework. SNP genotypes were modeled as taking on three levels (0, 1, and 2) corresponding to the observed genotypes, with measured BMD as the dependent variable. This model was fitted using the qfam-total analysis in PLINK. Correlation between subjects in the same family was corrected by obtaining empiric P values for all GWAS SNPs using a permutation approach (23). Permuted replicates were generated by randomly assigning calculated between- and within-family components of association across sibships, thus preserving the phenotypic correlations and linkage disequilibrium (LD) structure from one replicate to another. For the 50 SNPs with the greatest evidence of association with hip BMD and the 50 SNPs with the greatest evidence of association with spine BMD, we performed 108 permutations to accurately estimate the P value for the BMD-SNP association. P values for those SNPs with nominal P values between 10−3 and 10−5 were empirically estimated using 104 to 106 permutations, respectively, and SNPs with nominal P values greater than 10−2 using 103 permutations; in all cases, a minimum of 10 false-positive results were generated under the null at each level of nominal significance. The GWAS study design had 80% power to detect an effect size of 2.4% of total BMD variation and 50% power for an effect size of 1.8%.
Replication
The 50 most significant SNPs from the association analyses were genotyped in AA women (n = 669 from 383 families). Assays were designed for the MassArray system (Sequenom, San Diego, CA) using MassArray assay design software. Genotyping was performed used iPLEX assays (Sequenom), with alleles discriminated by mass spectrometry as described previously (24). Allele frequencies were estimated using the USERM13 option of MENDEL (http://csg.sph.umich.edu/boehnke/mendel.php). One of the 100 prioritized SNPs did not meet our criteria of greater than 90% genotyping success in the replication samples and was not analyzed. All other replication SNPs met the Hardy-Weinberg and minor allele frequency criteria implemented for the GWAS, as described above.
As described for the analysis of the genome-wide association data, covariates were identified, which contributed to the variation in BMD (P < 0.10). Many of these samples had been previously genotyped (4,5), and the full sibling relationships had already been confirmed in 94% of the families. Because the replication samples included families consisting of sibships as well as singletons, we used a population-based association test using the same permutation framework described for the GWAS analysis. This approach is robust to different family types (i.e. varying number of genotyped individuals within the family) (23). The replication SNP sets typed for both spine and femoral neck BMD each contained information equivalent to 39 uncorrelated SNPs as measured by the simpleM method (25), leading to a significance threshold of P = 0.003 for testing the one-sided replication hypothesis in the AA women. The replication design had 80% power to detect an effect size of 2.3% of total BMD variation and 50% power for an effect size of 1.4%.
Each supported association was reviewed in more detail to determine whether additional genotyping would allow further delineation of the associated region. The extent of LD in the region around the replicated SNPs was initially examined in the EA women using the GWAS data. Additional SNPs were selected to fill gaps between associated and nonassociated SNPs in the region. The genotyping of additional SNPs in the EA and AA women was performed using the Sequenom MassArray system, as described for the initial replication genotyping.
To facilitate comparison of the association results with top SNPs from other BMD GWAS typed on different platforms, we used the IMPUTE software (26) to generate genotype data for the SNPs in HapMap release 27 using the Illumina 610Quad genotypes on our samples and the reference phased haplotypes from HapMap release 27. We limited imputation to only the single chromosomal region in which a SNP associated in a previous study was not included in the Illumina array genotyped in this sample.
Results
The demographics of the EA sister sample used in the genome-wide association analysis are provided in Table 1. An overview of the results of the association analysis is presented in Tables 2 and 3. One SNP each was identified with P < 10−7 for lumbar spine BMD (rs10201742, chromosome 2) and neck BMD (rs11914434, chromosome 3). In addition, three SNPs for spine and two SNPs for neck BMD were identified with P < 10−6. The top two SNPs each for both spine and neck BMD met a false discovery rate threshold of 10% (27). A Q-Q plot for the GWAS is provided in Fig. 1, demonstrating that inflation of the test statistics beyond expected levels was not an issue in this sample; the calculated inflation factors were 1.002 and 1.003 for spine and neck BMD, respectively.
Table 1.
Sample characteristics
| EA women (GWAS) | AA women (replication) | |
|---|---|---|
| Sample size | 1524 | 669 |
| Number of families | 762 | 383 |
| Age (yr)a | 33.15 ± 7.21 | 32.71 ± 6.82 |
| Height (cm)a | 165.32 ± 6.06 | 164.36 ± 6.25 |
| Weight (kg)a | 71.82 ± 16.93 | 82.40 ± 19.87 |
| Spine BMD (g/cm2)a | 1.2801 ± 0.1328 | 1.3401 ± 0.1390 |
| Femoral neck BMD (g/cm2)a | 1.0173 ± 0.1212 | 1.1160 ± 0.1336 |
Mean ± sd.
Table 2.
Results of association analysis with lumbar spine BMD
| SNP | Chr | Position (bp)a | Geneb | Minor allele | EA women
|
AA women
|
|||
|---|---|---|---|---|---|---|---|---|---|
| MAFc | P value | Effectd | MAFc | P value | |||||
| rs10201742 | 2 | 127,045,243 | LOC150554 | C | 0.20 | 8.0E-07 | + | 0.47 | 0.47 |
| rs8104319 | 19 | 61,631,383 | ZNF542/ZNF582/ZNF583/ZNF667 | G | 0.46 | 1.7E-06 | + | 0.48 | 0.71 |
| rs12605652 | 18 | 74,536,154 | LOC731341 /LOC649290 | A | 0.41 | 3.0E-06 | + | 0.08 | 0.31 |
| rs6698119 | 1 | 48,047,574 | LOC388630 | G | 0.35 | 7.0E-06 | − | 0.27 | 0.59 |
| rs10832519 | 11 | 15,768,506 | · | C | 0.03 | 1.5E-05 | + | 0.03 | 0.56 |
| rs17269775 | 15 | 57,830,014 | RPS3AP6 | C | 0.06 | 1.7E-05 | − | 0.004 | 0.53 |
| rs3760849 | 19 | 61,645,397 | ZNF582/ZNF583/ZNF667 | A | 0.46 | 1.7E-05 | + | 0.50 | 0.40 |
| rs2296596 | 10 | 13,775,953 | FRMD4A | G | 0.11 | 2.2E-05 | − | 0.18 | 0.38 |
| rs4801163 | 19 | 61,667,198 | ZNF583/ZNF667/ZNF471 | T | 0.42 | 2.4E-05 | + | 0.40 | 0.34 |
| rs9785225 | 9 | 93,022,313 | AUH | A | 0.12 | 2.7E-05 | + | 0.09 | 0.22 |
| rs16972369 | 16 | 72,310,483 | LOC649800 | G | 0.04 | 2.7E-05 | + | 0.16 | 0.06 |
| rs7788807 | 7 | 4,734,564 | FOXK1/KIAA0415 | C | 0.06 | 2.8E-05 | − | 0.12 | 0.11 |
| rs9944433 | 17 | 68,593,547 | SLC39A11 | A | 0.38 | 3.5E-05 | + | 0.27 | 0.08 |
| rs2309883 | 4 | 28,341,579 | · | T | 0.26 | 3.9E-05 | + | 0.09 | 0.16 |
| rs3774830 | 4 | 5,490,984 | STK32B | A | 0.43 | 4.0E-05 | − | 0.29 | 0.76 |
| rs10852681 | 16 | 7,254,704 | A2BP1 | G | 0.10 | 4.2E-05 | − | 0.20 | 0.40 |
| rs6436459 | 2 | 224,588,580 | MRPL44/SERPINE2 | T | 0.23 | 4.2E-05 | − | 0.21 | 0.35 |
| rs11659609 | 18 | 74,564,026 | LOC649290 | A | 0.27 | 4.4E-05 | + | 0.27 | 0.80 |
| rs7202238 | 16 | 50,947,319 | · | C | 0.40 | 4.6E-05 | − | 0.41 | 0.69 |
| rs3845972 | 3 | 60,002,874 | FHIT | G | 0.37 | 4.6E-05 | + | 0.49 | 0.48 |
| rs4398085 | 15 | 24,924,700 | · | T | 0.26 | 5.1E-05 | − | 0.27 | 0.60 |
| rs4314962 | 10 | 13,777,035 | FRMD4A | C | 0.15 | 5.2E-05 | − | 0.48 | 0.76 |
| rs2062140 | 5 | 44,532,172 | · | T | 0.07 | 6.5E-05 | − | 0.05 | 0.22 |
| rs1919555 | 3 | 122,929,590 | GOLGB1/IQCB1 | T | 0.06 | 6.7E-05 | − | 0.01 | 0.11 |
| rs10778724 | 12 | 79,236,063 | LOC651200/FLJ90579 | C | 0.03 | 6.8E-05 | + | 0.27 | 0.20 |
| rs6436652 | 2 | 157,358,203 | · | C | 0.39 | 7.1E-05 | + | 0.3 | 0.35 |
| rs2354025 | 16 | 85,224,910 | · | T | 0.34 | 7.6E-05 | + | 0.16 | 0.21 |
| rs2291607 | 8 | 62,533,814 | MGC34646/ASPH | G | 0.21 | 7.7E-05 | + | 0.29 | 0.17 |
| rs13169538 | 5 | 44,605,797 | · | A | 0.07 | 7.9E-05 | − | 0.09 | 0.18 |
| rs9288624 | 2 | 157,359,026 | · | T | 0.43 | 7.9E-05 | + | 0.26 | 0.09 |
| rs6827240 | 4 | 5,491,125 | STK32B | T | 0.45 | 7.9E-05 | − | 0.41 | 0.59 |
| rs9859339 | 3 | 142,391,751 | SPSB4/ACPL2 | T | 0.16 | 8.1E-05 | + | 0.11 | 0.89 |
| rs12881798 | 14 | 94,416,311 | · | A | 0.33 | 8.4E-05 | + | 0.45 | 0.36 |
| rs16970118 | 13 | 106,892,968 | LOC728215 | G | 0.15 | 8.6E-05 | − | 0.28 | 0.49 |
| rs1751449 | 9 | 22,186,863 | LOC729983 | G | 0.41 | 9.3E-05 | − | 0.49 | 0.67 |
| rs4606154 | 9 | 13,458,276 | FLJ41200 | A | 0.38 | 9.4E-05 | + | 0.36 | 0.64 |
| rs16901892 | 5 | 44,548,826 | · | C | 0.07 | 9.58E-05 | − | 0.05 | 0.50 |
| rs6857026 | 4 | 80,992,485 | · | T | 0.29 | 1.1E-04 | − | 0.30 | 0.21 |
| rs17537076 | 15 | 83,265,502 | ALPK3/SLC28A1 | A | 0.37 | 1.1E-04 | + | 0.07 | 0.86 |
| rs9511790 | 13 | 24,887,752 | LOC246717 | T | 0.34 | 1.2E-04 | + | 0.12 | 0.16 |
| rs4808629 | 19 | 17,364,996 | PLVAP/BST2/LOC93343 | C | 0.12 | 1.3E-04 | − | 0.47 | 0.16 |
| rs12654032 | 5 | 135,462,508 | TGFBI/SMAD5 | G | 0.07 | 1.4E-04 | − | 0.02 | 0.14 |
| rs2450993 | 12 | 47,015,421 | H1FNT/ZNF641/LOC730678/OR5BK1P/OR5BT1P | T | 0.03 | 2.0E-04 | + | 0.003 | 0.11 |
| rs5929069 | X | 30,081,782 | · | T | 0.02 | NA | NA | 0.01 | 0.13 |
Chr, Chromosome; MAF, minor allele frequency.
Position obtained from build 130.
Defined as any gene within 25 kb of the SNP. If there is no gene within 25 kb of the SNP, a dot (.) is indicated.
Minor allele frequency.
+, Minor allele increases BMD.
Table 3.
Results of association analysis with femoral neck BMD
| SNP | Chr | Position (bp)a | Geneb | Minor allele | EA women
|
AA women
|
|||
|---|---|---|---|---|---|---|---|---|---|
| MAFc | P value | Effectd | MAFc | P value | |||||
| rs11914434 | 3 | 195,157,275 | · | T | 0.08 | 7.0E-07 | − | 0.11 | 0.84 |
| rs3805941 | 6 | 43,165,559 | CUL7/MRPL2/KLC4/PTK7 | C | 0.37 | 2.1E-06 | + | 0.41 | 0.40 |
| rs9827714 | 3 | 128,383,182 | LOC729455/C3orf56 | C | 0.47 | 6.9E-06 | − | 0.43 | 0.65 |
| rs1826601 | 16 | 76,059,687 | ADAMTS18 | T | 0.43 | 1.0E-05 | + | 0.34 | 0.25 |
| rs6459375 | 6 | 14,849,285 | · | G | 0.10 | 1.1E-05 | − | 0.29 | 0.74 |
| rs6667721 | 1 | 76,640,441 | ST6GALNAC3 | A | 0.15 | 1.3E-05 | + | 0.03 | 0.70 |
| rs1215494 | 1 | 89,839,394 | LRRC8B/LRRC8C | T | 0.36 | 1.6E-05 | + | 0.38 | 0.57 |
| rs333616 | 8 | 56,963,551 | LYN | C | 0.43 | 1.7E-05 | − | 0.10 | 0.78 |
| rs2327672 | 6 | 136,269,912 | PDE7B | A | 0.14 | 1.8E-05 | + | 0.15 | 0.82 |
| rs2038016 | 6 | 14,845,488 | · | A | 0.10 | 1.9E-05 | − | 0.28 | 1.00 |
| rs4887348 | 15 | 86,372,538 | NTRK3 | G | 0.29 | 2.2E-05 | + | 0.10 | 0.06 |
| rs2627690 | 4 | 88,872,640 | · | T | 0.30 | 2.3E-05 | − | 0.25 | 0.44 |
| rs2152319 | 1 | 177,581,980 | SOAT1/C1orf125 | G | 0.42 | 2.3E-05 | + | 0.46 | 0.73 |
| rs1298989 | 14 | 91,156,466 | CATSPERB | T | 0.25 | 2.7E-05 | + | 0.37 | 0.006e |
| rs1285635 | 14 | 91,158,608 | CATSPERB | A | 0.25 | 3.0E-05 | + | 0.38 | 0.003e |
| rs8090593 | 18 | 69,667,938 | · | G | 0.37 | 3.1E-05 | − | 0.10 | 0.40 |
| rs2472793 | 6 | 14,856,654 | · | T | 0.10 | 3.2E-05 | − | 0.40 | 0.93 |
| rs4651024 | 1 | 177,578,725 | SOAT1/C1orf125 | T | 0.49 | 3.3E-05 | + | 0.40 | 0.85 |
| rs4712240 | 6 | 14,860,852 | · | C | 0.10 | 3.6E-05 | − | 0.21 | 0.32 |
| rs2505780 | 10 | 86,450,149 | LOC648354 | G | 0.45 | 3.6E-05 | − | 0.50 | 0.25 |
| rs1591687 | 9 | 105,219,667 | · | A | 0.04 | 3.6E-05 | + | 0.17 | 0.41 |
| rs4240241 | 4 | 37,443,327 | · | G | 0.04 | 3.6E-05 | − | 0.29 | 0.28 |
| rs13336488 | 16 | 76,197,733 | · | G | 0.05 | 3.7E-05 | + | 0.09 | 0.97 |
| rs6479168 | 9 | 105,217,183 | · | T | 0.04 | 3.78E-05 | + | 0.17 | 0.52 |
| rs11722802 | 4 | 28,332,952 | · | A | 0.25 | 3.9E-05 | + | 0.04 | 0.29 |
| rs6443126 | 3 | 7,826,708 | · | T | 0.49 | 4.0E-05 | − | 0.29 | 0.48 |
| rs2637502 | 5 | 124,792,046 | · | A | 0.49 | 4.0E-05 | + | 0.19 | 0.07 |
| rs9563646 | 13 | 58,232,735 | · | C | 0.33 | 4.1E-05 | + | 0.32 | 0.27 |
| rs11651617 | 17 | 14,351,933 | · | G | 0.21 | 4.3E-05 | − | 0.23 | 0.63 |
| rs2323284 | 17 | 14,350,891 | · | T | 0.21 | 4.3E-05 | − | 0.33 | 0.90 |
| rs2080115 | 17 | 14,345,391 | · | G | 0.19 | 5.2E-05 | − | 0.19 | 0.53 |
| rs7783471 | 7 | 51,958,833 | · | A | 0.07 | 5.3E-05 | + | 0.09 | 0.75 |
| rs1360003 | 9 | 77,473,244 | · | C | 0.25 | 5.3E-05 | − | 0.30 | 0.24 |
| rs489659 | 18 | 7,547,909 | PTPRM | G | 0.17 | 6.2E-05 | + | 0.33 | 0.05 |
| rs772492 | 4 | 90,591,589 | · | T | 0.32 | 6.3E-05 | + | 0.28 | 0.85 |
| rs10223929 | 7 | 48,374,319 | ABCA13 | A | 0.10 | 7.2E-05 | − | 0.03 | 0.42 |
| rs8057501 | 16 | 76,116,178 | · | G | 0.39 | 7.5E-05 | + | 0.40 | 0.88 |
| rs2126650 | 4 | 88,929,611 | IBSP/MEPE | A | 0.28 | 7.7E-05 | − | 0.08 | 0.38 |
| rs1431381 | 18 | 6,802,509 | ARHGAP28 | A | 0.12 | 7.8E-05 | + | 0.03 | 0.46 |
| rs4686162 | 3 | 7,833,150 | · | C | 0.21 | 8.8E-05 | + | 0.06 | 0.02 |
| rs1385320 | 16 | 76,099,924 | · | A | 0.41 | 1.1E-04 | + | 0.39 | 0.69 |
| rs1873873 | 11 | 128,984,385 | LOC731658 | C | 0.48 | 1.1E-04 | + | 0.48 | 0.52 |
| rs6654021 | X | 17,665,929 | NHS/SCML1 | G | 0.01 | NA | NA | 0.49 | 0.77 |
| rs12387240 | X | 143,324,556 | · | A | 0.16 | NA | NA | 0.13 | 0.61 |
| rs2006793 | X | 29,394,774 | IL1RAPL1 | G | 0.49 | NA | NA | 0.10 | 0.61 |
| rs1801412 | X | 114,048,960 | HTR2C | G | 0.04 | NA | NA | 0.01 | 0.32 |
Position obtained from build 130;
defined as any gene within 25 kb of the SNP. If there is no gene within 25 kb of the SNP, a dot (.) is indicated;
minor allele frequency;
+, minor allele increases BMD;
direction of effect is the same as in the GWAS cohort; bold P value indicates P value less than adjusted replication threshold.
Figure 1.
Q-Q plot illustrating the relationship between the observed (black symbols) permutation-based association results (y-axis) and results expected for a GWAS study with the same properties (gray symbols) under the null hypothesis (x-axis). Minus log10 P values are plotted on both axes.
To test the hypothesis that genetic effects detected in premenopausal women might have effects in AA women as well, the 50 SNPs with the greatest evidence of association with lumbar spine BMD (Table 2) and the 50 SNPs most associated with femoral neck BMD (Table 3) were selected for replication in a sample of AA women. The demographic characteristics of this replication sample are shown in Table 1.
Among the 50 GWAS SNPs with the greatest evidence of association with femoral neck BMD, a SNP in CATSPERB on chromosome 14 (rs1298989, P = 2.7 × 10−5; rs1285635, P = 3.0 × 10−5) also provided some evidence of association with femoral neck BMD in the sample of AA women (Table 3: rs1285635, P = 0.003), with further support from another nearby SNP (rs1298989, P = 0.006). Association results for these SNPs were similar under a standard ANOVA model when ancestry-informative SNPs typed previously on the AA subjects were included as covariates. These two SNPs are in very high LD with each other (r2 = 1.0 in the EA sample and 0.95 in the AA sample). No other SNPs genotyped in the AA women attained a P value that exceeded our threshold for evidence of replication (P < 0.003). Based on the supporting AA data, 47 additional SNPs in CATSPERB were genotyped in the EA and AA women. As shown in Fig. 2, this increased SNP coverage of the region detects maximal association with femoral neck BMD in both EA and AA women near rs1285635, with supporting evidence from nearby SNPs becoming less significant both 5′ and 3′ of this position.
Figure 2.
Association results with lumbar spine (plus symbol) and femoral neck BMD (diamond symbol) in the AA women for additional SNPs in the 3′ region of the CATSPERB gene on chromosome 14. SNP rs1285635, identified in the GWAS and also demonstrating evidence of association in AA women, is indicated with the small circle.
We also reviewed our GWAS results to evaluate whether we could replicate previously reported associations. In two regions, we found evidence of association with a SNP reported in a previous GWAS with the same phenotype, although in both cases with a different SNP providing the peak association evidence in the gene or region. Association of femoral neck BMD with rs11864477 and other SNPs in ADAMTS18 on chromosome 16 were reported in a Caucasian sample and then replicated by the same authors in other cohorts (18). The SNP rs11864477 was not included on our genotyping platform, but its association P value based on imputed genotypes in our sample was 0.006. Imputation accuracy of this SNP from our data was excellent, with variance ratio of 0.96 based on a SNP of very similar frequency 1 kb distal (rs16945612). We detected evidence of association between femoral neck BMD and rs1826601 (Table 3: P = 1.1 × 10−5; Fig. 3) near the 5′ terminus of ADAMTS18, approximately 80 kb 3′ from rs1826601. Association with femoral neck BMD and SNPs in the IBSP/MEPE/DMP1 region on chromosome 4 was detected in a GWAS of a predominantly female Icelandic sample (21). The Icelandic association was with SNP rs1054627, which was also typed in our study and provided evidence of association between femoral neck BMD and this region (P = 1.5 × 10−4) in our sample. An additional SNP in this region, rs2126650, was among our top findings for femoral neck BMD and is located 85 kb 3′ from rs1054627 (Table 3; P = 7.7 × 10−5).
Figure 3.
Femoral neck BMD association results with SNPs in and near the ADAMTS18 locus on chromosome. Position and orientation of the gene are indicated by the black arrow.
Discussion
We performed a genome-wide association study to detect common genetic variants influencing variability in peak BMD in premenopausal EA women. This phenotype represents a highly heritable risk factor for low BMD and osteoporotic fracture later in a woman’s life. By studying younger women, the genetic factors influencing peak BMD may be detected without the confounding effects of variants and environmental factors that act on BMD later in life, for example, by modulating the rate of bone loss with age (28). Thus, our study is complementary to the previous GWAS that focused primarily on BMD in older subjects (15,17), osteoporotic fracture in the elderly (20), or a combination of these phenotypes (18).
Whereas genome-wide significance (P < 5 × 10−8) for individual SNP associations were not obtained in our study, several loci with point-wise empiric P values in the 10−6 to 10−7 range were identified for both femoral neck and lumbar spine BMD. Among these findings, support for association of SNPs in CATSPERB across racial groups was also demonstrated in our cohort of premenopausal AA women. Studies to date of this gene (151 kb in size, with 27 exons) have been somewhat limited; the gene is expressed in testis, and its product appears to be a subunit of a cation-specific membrane channel complex essential for male fertility (29). Expression has also been detected in ovary and liver [National Center for Biotechnology Information (Bethesda, MD) expressed sequence tag profile Hs.131755; (30)]. The role of CATSPERB in these tissues is poorly understood, but its expression in ovary suggests a possible role in hormonal regulation of bone metabolism.
Top findings unique to our study might have their primary effect on premenopausal BMD. These include IQCB1 and SSPH, demonstrated to be involved in calmodulin binding and calcium homeostasis, respectively. Also among our most significant findings was a SNP near TGFB/SMAD5, a regulator of the bone morphogenic protein family members. All three of these findings were in the 3′ (for IQCB1 and ASPH) or 5′ (for SMAD5) regions of the respective genes, suggesting possible regulatory roles in gene expression.
The association findings in common with other bone GWAS, typically conducted in older populations, might alternatively suggest that some genes with an effect on peak BMD show an effect later in life as well. Our GWAS for peak BMD in a sample of premenopausal EA women also found evidence for several genes associated with BMD in previous GWAS. These include ADAMTS18, which was associated with hip BMD in females in the study of Xiong et al. (18). The IBSP gene, identified by Styrkarsdottir et al. (21) for hip BMD, was one of the top 50 findings for femoral neck BMD in our study as well. We observed evidence of association with femoral neck BMD and, to a lesser extent, spine BMD across a broad region that also encompasses the genes DMP1 and MEPE, both with well-established roles in bone mineralization.
Through the careful evaluation of association evidence, we have identified genes that may contribute to peak BMD rather than bone loss or other factors affecting changes in BMD later in life, although this hypothesis will need to be tested in premenopausal cohorts with sufficient power to confirm or exclude effects much smaller than is possible in our study. Replication of the novel associations identified here in other, larger cohorts (in both EAs and other ethnicities) is critical for confirmation and in determining the magnitude of their effect in the populations from which the cohorts were drawn.
Acknowledgments
We thank the individuals who participated in this study as well as the study coordinators, without whom this work would not have been possible. Replication in independent samples was completed by MALDI-TOF mass spectrometry at the facilities of the Center for Medical Genomics at Indiana University School of Medicine, which is supported in part by a grant from the Indiana Genomics Initiative (INGEN). INGEN is supported in part by the Lilly Endowment, Inc.
Footnotes
This work was supported by National Institutes of Health (NIH) Grants P01 AG-18397 and M01 RR-00750. Genotyping services were provided by Center for Inherited Disease Research, which is fully funded through a federal contract from the NIH to the Johns Hopkins University (Contract HHSN268200782096C). This research was supported in part by the Intramural Research Program of the NIH, National Library of Medicine.
Disclosure Summary: D.L.K., S.I., D.L., L.R.P., K.F.D., E.P., J.P., S.L.H., H.J.E., X.X., M.P., M.J.E., and T.F. have nothing to declare.
First Published Online February 17, 2010
Abbreviations: AA, African-American; BMD, bone mineral density; EA, European-American; GWAS, genome-wide association studies; LD, linkage disequilibrium; SNP, single-nucleotide polymorphism.
References
- Peacock M, Turner CH, Econs MJ, Foroud T 2002 Genetics of osteoporosis. Endocr Rev 23:303–326 [DOI] [PubMed] [Google Scholar]
- Slemenda CW, Christian JC, Williams CJ, Norton JA, Johnston Jr CC 1991 Genetic determinants of bone mass in adult women: a reevaluation of the twin model and the potential importance of gene interaction on heritability estimates. J Bone Miner Res 6:561–567 [DOI] [PubMed] [Google Scholar]
- Hustmyer FG, Peacock M, Hui S, Johnston CC, Christian J 1994 Bone mineral density in relation to polymorphism at the vitamin D receptor gene locus. J Clin Invest 94:2130–2134 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Econs MJ, Koller DL, Hui SL, Fishburn T, Conneally PM, Johnston Jr CC, Peacock M, Foroud TM 2004 Confirmation of linkage to chromosome 1q for peak vertebral bone mineral density in premenopausal white women. Am J Hum Genet 74:223–228 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Peacock M, Koller DL, Hui S, Johnston CC, Foroud T, Econs MJ 2004 Peak bone mineral density at the hip is linked to chromosomes 14q and 15q. Osteoporos Int 15:489–496 [DOI] [PubMed] [Google Scholar]
- Ioannidis JP, Ng MY, Sham PC, Zintzaras E, Lewis CM, Deng HW, Econs MJ, Karasik D, Devoto M, Kammerer CM, Spector T, Andrew T, Cupples LA, Duncan EL, Foroud T, Kiel DP, Koller D, Langdahl B, Mitchell BD, Peacock M, Recker R, Shen H, Sol-Church K, Spotila LD, Uitterlinden AG, Wilson SG, Kung AW, Ralston SH 2007 Meta-analysis of genome-wide scans provides evidence for sex- and site-specific regulation of bone mass. J Bone Miner Res 22:173–183 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Beamer WG, Shultz KL, Donahue LR, Churchill GA, Sen S, Wergedal JR, Baylink DJ, Rosen CJ 2001 Quantitative trait loci for femoral and lumbar vertebral bone mineral density in C57BL/6J and C3H/HeJ inbred strains of mice. J Bone Miner Res 16:1195–1206 [DOI] [PubMed] [Google Scholar]
- Edderkaoui B, Baylink DJ, Beamer WG, Wergedal JE, Dunn NR, Shultz KL, Mohan S 2006 Multiple genetic loci from CAST/EiJ chromosome 1 affect vBMD either positively or negatively in a C57BL/6J background. J Bone Miner Res 21:97–104 [DOI] [PubMed] [Google Scholar]
- Duncan EL, Cardon LR, Sinsheimer JS, Wass JA, Brown MA 2003 Site and gender specificity of inheritance of bone mineral density. J Bone Miner Res 18:1531–1538 [DOI] [PubMed] [Google Scholar]
- Hsu YH, Xu X, Terwedow HA, Niu T, Hong X, Wu D, Wang L, Brain JD, Bouxsein ML, Cummings SR, Rosen CJ, Xu X 2007 Large-scale genome-wide linkage analysis for loci linked to BMD at different skeletal sites in extreme selected sibships. J Bone Miner Res 22:184–194 [DOI] [PubMed] [Google Scholar]
- Xiong DH, Wang JT, Wang W, Guo YF, Xiao P, Shen H, Jiang H, Chen Y, Deng H, Drees B, Recker RR, Deng HW 2007 Genetic determination of osteoporosis: lessons learned from a large genome-wide linkage study. Hum Biol 79:593–608 [DOI] [PubMed] [Google Scholar]
- Huang QY, Ng MY, Cheung CL, Chan V, Sham PC, Kung AW 2006 Identification of two sex-specific quantitative trait loci in chromosome 11q for hip bone mineral density in Chinese. Hum Hered 61:237–243 [DOI] [PubMed] [Google Scholar]
- Turner CH, Sun Q, Schriefer J, Pitner N, Price R, Bouxsein ML, Rosen CJ, Donahue LR, Shultz KL, Beamer WG 2003 Congenic mice reveal sex-specific genetic regulation of femoral structure and strength. Calcif Tissue Int 73:297–303 [DOI] [PubMed] [Google Scholar]
- Orwoll ES, Belknap JK, Klein RF 2001 Gender specificity in the genetic determinants of peak bone mass. J Bone Miner Res 16:1962–1971 [DOI] [PubMed] [Google Scholar]
- Kiel DP, Demissie S, Dupuis J, Lunetta KL, Murabito JM, Karasik D 2007 Genome-wide association with bone mass and geometry in the Framingham Heart Study. BMC Med Genet 8(Suppl 1):S14 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Richards JB, Rivadeneira F, Inouye M, Pastinen TM, Soranzo N, Wilson SG, Andrew T, Falchi M, Gwilliam R, Ahmadi KR, Valdes AM, Arp P, Whittaker P, Verlaan DJ, Jhamai M, Kumanduri V, Moorhouse M, van Meurs JB, Hofman A, Pols HA, Hart D, Zhai G, Kato BS, Mullin BH, Zhang F, Deloukas P, Uitterlinden AG, Spector TD 2008 Bone mineral density, osteoporosis, and osteoporotic fractures: a genome-wide association study. Lancet 371:1505–1512 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, Gudbjartsson DF, Walters GB, Ingvarsson T, Jonsdottir T, Saemundsdottir J, Center JR, Nguyen TV, Bagger Y, Gulcher JR, Eisman JA, Christiansen C, Sigurdsson G, Kong A, Thorsteinsdottir U, Stefansson K 2008 Multiple genetic loci for bone mineral density and fractures. N Engl J Med 358:2355–2365 [DOI] [PubMed] [Google Scholar]
- Xiong DH, Liu XG, Guo YF, Tan LJ, Wang L, Sha BY, Tang ZH, Pan F, Yang TL, Chen XD, Lei SF, Yerges LM, Zhu XZ, Wheeler VW, Patrick AL, Bunker CH, Guo Y, Yan H, Pei YF, Zhang YP, Levy S, Papasian CJ, Xiao P, Lundberg YW, Recker RR, Liu YZ, Liu YJ, Zmuda JM, Deng HW 2009 Genome-wide association and follow-up replication studies identified ADAMTS18 and TGFBR3 as bone mass candidate genes in different ethnic groups. Am J Hum Genet 84:388–398 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Timpson NJ, Tobias JH, Richards JB, Soranzo N, Duncan EL, Sims AM, Whittaker P, Kumanduri V, Zhai G, Glaser B, Eisman J, Jones G, Nicholson G, Prince R, Seeman E, Spector T, Brown MA, Peltonen L, Smith GD, Deloukas P, Evans DM 2009 Common variants in the region around Osterix are associated with bone mineral density and growth in childhood. Hum Mol Genet 18:1510–1517 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cho YS, Go MJ, Kim YJ, Heo JY, Oh JH, Ban HJ, Yoon D, Lee MH, Kim DJ, Park M, Cha SH, Kim JW, Han BG, Min H, Ahn Y, Park MS, Han HR, Jang HY, Cho EY, Lee JE, Cho NH, Shin C, Park T, Park JW, Lee JK, Cardon L, Clarke G, McCarthy MI, Lee JY, Lee JK, Oh B, Kim HL 2009 A large-scale genome-wide association study of Asian populations uncovers genetic factors influencing eight quantitative traits. Nat Genet 41:527–534 [DOI] [PubMed] [Google Scholar]
- Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, Gudbjartsson DF, Walters GB, Ingvarsson T, Jonsdottir T, Saemundsdottir J, Snorradóttir S, Center JR, Nguyen TV, Alexandersen P, Gulcher JR, Eisman JA, Christiansen C, Sigurdsson G, Kong A, Thorsteinsdottir U, Stefansson K 2009 New sequence variants associated with bone mineral density. Nat Genet 41:15–17 [DOI] [PubMed] [Google Scholar]
- Gunderson KL, Steemers FJ, Ren H, Ng P, Zhou L, Tsan C, Chang W, Bullis D, Musmacker J, King C, Lebruska LL, Barker D, Oliphant A, Kuhn KM, Shen R 2006 Whole-genome genotyping. Methods Enzymol 410:359–376 [DOI] [PubMed] [Google Scholar]
- Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, Maller J, Sklar P, de Bakker PI, Daly MJ, Sham PC 2007 PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 81:559–575 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ichikawa S, Koller DL, Johnson ML, Lai D, Xuei X, Edenberg HJ, Klein RF, Orwoll ES, Hui SL, Foroud TM, Peacock M, Econs MJ 2006 Human ALOX12, but not ALOX15, is associated with BMD in white men and women. J Bone Miner Res 21:556–564 [DOI] [PubMed] [Google Scholar]
- Gao X, Starmer J, Martin ER 2008 A multiple testing correction method for genetic association studies using correlated single nucleotide polymorphisms. Genet Epidemiol 32:361–369 [DOI] [PubMed] [Google Scholar]
- Marchini J, Howie B, Myers S, McVean G, Donnelly P 2007 A new multipoint method for genome-wide association studies by imputation of genotypes. Nat Genet 39:906–913 [DOI] [PubMed] [Google Scholar]
- Hochberg Y, Benjamini Y 1990 More powerful procedures for multiple significance testing. Stat Med 9:811–818 [DOI] [PubMed] [Google Scholar]
- U.S. Department of Health and Human Services 2004 Bone health and osteoporosis: a report of the Surgeon General. Rockville, MD: U.S. Department of Health and Human Services, Office of the Surgeon General [Google Scholar]
- Liu J, Xia J, Cho KH, Clapham DE, Ren D 2007 CatSperbeta, a novel transmembrane protein in the CatSper channel complex. J Biol Chem 282:18945–18952 [DOI] [PubMed] [Google Scholar]
- Wang H, Liu J, Cho KH, Ren D 2009 A novel, single, transmembrane protein CATSPERG is associated with CATSPER1 channel protein. Biol Reprod 81:539–544 [DOI] [PMC free article] [PubMed] [Google Scholar]