A whole genome linkage scan for QTLs underlying peak bone mineral density
- First Online:
- Cite this article as:
- Zhang, F., Xiao, P., Yang, F. et al. Osteoporos Int (2008) 19: 303. doi:10.1007/s00198-007-0468-z
- 56 Views
We conducted a whole genome linkage scan for quantitative trait loci (QTLs) underlying peak bone mineral density (PBMD). Our efforts identified several potential genomic regions for PBMD and highlighted the importance of epistatic interaction and sex-specific analyses in identifying genetic regions underlying PBMD variation.
Peak bone mineral density (PBMD) is an important clinical risk predictor of osteoporosis and explains a large part of bone mineral density (BMD) variation.
To detect susceptive quantitative trait loci (QTLs) for PBMD variation including consideration of epistatic and sex-specific effects, we conducted a whole genome linkage scan (WGLS) for PBMD using 2,200 Caucasians from 207 pedigrees, aged 20–50 years. All the individuals were genotyped with 410 microsatellite markers. In addition to WGLS in the total combined sample of males and females, we conducted epistatic interaction analyses, and sex-specific subgroup linkage analyses.
We identified several potential genomic regions that met the criteria for suggestive linkage. The most impressing region is 12p12 for hip PBMD (LOD = 2.79) in the total sample. Epistatic interaction analyses found a significant epistatic interaction between 12p12 and 22q13 (p = 0.0021) for hip PBMD. Additionally, we detected suggestive linkage evidence at 15q26 (LOD = 2.93), 2p13 (LOD = 2.64), and Xq27 (LOD = 2.64). Sex-specific analyses suggested the presence of sex-specific QTLs for PBMD variation.
Our efforts identified several potential regions for PBMD and highlighted the importance of epistatic interaction and sex-specific analyses in identifying genetic regions underlying PBMD variation.
KeywordsEpistatic interactionPBMDQTLSex-specificWhole genome linkage scan
Osteoporosis is a major metabolic skeletal disease, characterized by low bone mass and high risk of fragility fracture . In the United States alone, as many as 10 million people suffered from osteoporosis in 2002 , with an estimated healthcare expenditure of over $17 billion on osteoporotic fractures .
The most prominent risk predictor of osteoporosis is bone mass, which is clinically measured by bone mineral density (BMD). BMD varies across an individual’s life, peaking in early adulthood (aged 20–30) and generally remaining relatively stable until the age of ∼50 in both males and females or until the menopause in females, if it occurs before the age of 50; after that BMD begins decreasing due to bone loss . Therefore, for the purposes of this study, we use peak BMD (PBMD) to refer to BMD in men and premenopausal women aged 20–50. Previous studies have suggested that PBMD is the most important factor influencing BMD  and that PBMD variation accounts for a large part of total BMD variation . However, it was also found that the QTLs underlying BMD varied in different age subgroups, including the PBMD subgroup [7–9].
To date, whole genome linkage scan (WGLS) for PBMD in human is few; sex-specific WGLS for PBMD is more rare . Compared with WGLS for BMD, WGLS for PBMD, which diminish the effect of age in the genetic regulation of BMD, has potential advantages in QTLs mapping for osteoporosis. In this study, we conducted a WGLS for PBMD in 2,200 Caucasians from 207 pedigrees. Two-locus epistatic interaction and sex-specific subgroup analyses were also conducted to identify potential gene–gene interactions and sex-specific QTLs.
Materials and methods
Distribution of pedigree size
Pedigree size (participants)
Number of families
Number of participants
Bone mineral density (g/cm2) for the wrist (the distal region of the ulna and radius), total hip (the femoral neck, trochanter, and intertrochanteric region), and the lumbar spine (L1–L4) were measured by dual-energy X-ray absorptiometry (DXA) with a Hologic QDR 1000, 2000+ or 4500 scanner (Hologic, Bedford, MA, USA). All scanners were calibrated daily and long-term precision was monitored with external phantoms. The measurement precision, as reflected by coefficients of variation (CV) for spine, hip, and wrist BMD measured by Hologic 2000+, were 0.9%, 1.4%, and 2.3% respectively. Similar CV were obtained on Hologic 1000 and 4500 . Members of the same pedigree were generally measured on the same scanner to ensure a minimal effect of the difference in scanners on our linkage analyses. In addition, BMD data obtained from different machines were transformed into a compatible measurement using a previously described formula  and an algorithm that we developed in-house and have used extensively [13–15]. Weight (kg) and height (m) were obtained at the same visit when BMD measurement took place.
For each participant, DNA was extracted from peripheral blood using Puregene DNA isolation kit (Gentra Systems, Minneapolis, MN, USA). All the participants were genotyped with 410 microsatellite markers, including 392 markers for 22 autosomes and 18 markers for the X chromosome. These markers are from the Marshfield screening set 14 given by the Marshfield Center for Medical Genetics. Generally, these markers are an average of 8.9 cM apart, with an average population heterozygosity of 0.75. The detailed genotyping protocol can be found at http://research.marshfieldclinic.org/genetics/Lab_Methods/ methods.html. We used PedCheck software  to verify Mendelian inheritance of the alleles at all marker loci within each family. In addition, the program MERLIN  was implemented to detect genotyping errors of unlikely recombination (e.g., double recombination) in our sample. The overall genotyping error rate was ∼0.3%.
Multipoint analyses for chromosomes 1–22 and two-point analyses for chromosome X were performed using the variance component (VC) method implemented in SOLAR (Sequential Oligogenic Linkage Analysis Routines, http://www.sfbr.org/solar) [18–20]. Applying a polygenic screen model in SOLAR, age, sex, height, weight, and age-by-sex interactions were tested for importance on PBMD variation and significant factors (p ≤ 0.05) were used to adjust for raw PBMD values as covariates in linkage analyses. To test the robustness of our results, 10,000 simulations were carried out using the procedure “lodadj” implemented in SOLAR , which is considered to be one of the best ways to deal with normality issues . Estimated correction constants for LOD scores of the three skeletal sites ranged from 0.96 to 1.07. All LOD scores reported in the text were empirically adjusted LOD scores. In addition, empiricalp values for adjusted LOD scores were also calculated by SOLAR. Since the current version of SOLAR cannot handle multipoint linkage analyses on chromosome X, the same approach for chromosome X employed in our previous WGLS for BMD  was used here. In brief, MEGA2 (Manipulation Environment for Genetic Analyses) was utilized to break down the large pedigrees into small ones and reserved as many relative pairs as possible . Multipoint linkage analysis on chromosome X was conducted in MERLIN (Multipoint Engine for Rapid Likelihood Inference) . Because three skeletal sites were analyzed, correcting for multiple analyses was performed as described by Camp and Farnham . The estimated genome-wide thresholds of “suggestive” and “significant” evidence for linkage were 2.40 and 3.80 respectively.
Two-locus analyses were performed to search for epistatic interactions on the PBMD variables between the chromosomal region with the maximum LOD score and other regions with LOD score ≥ 1.0. This test extended from the VC model where interactions between two chromosomal regions are considered simultaneously. Two levels of modeling in addition to single-locus modeling were performed: two-locus models with only additive effects for each pair of loci; and two-locus models with additive effects, as well as an epistatic term for interaction for both loci. The significantly higher LOD score under the epistatic model than that under the additive model suggests the existence of an interaction between these two regions. One-tailedp values were calculated using the χ2 test with 1 degree of freedom (d.f.) for all hypotheses tested . The significance of a single test for interaction effect on any one variable was assessed at a type I error rate of 0.05/N according to the Bonferroni adjustment for multiple comparison.N was the number of independent tests conducted for each trait.
For sex-specific WGLS, we used the same approach as for our previous sex-specific WGLS for BMD . The general methods are essentially the same as those used for the total sample WGLS described above. The major differences are that only single-sex participants and the Marshfield sex-specific genetic maps were used in each subgroup analysis instead of the sex-average map.
Characteristics of the study participants (n = 2,200). PBMD peak bone mineral density, h2 heritability, SD standard deviation
Female (mean ± SD)
Male (mean ± SD)
Total (mean ± SD)
36.9 ± 8.3
37.3 ± 8.7
37.08 ± 8.5
165.5 ± 6.2
179.8 ± 6.9
171.4 ± 9.6
70.6 ± 15.5
89.6 ± 16.4
78.5 ± 18.5
0.96 ± 0.13
1.07 ± 0.14
1.01 ± 0.14
0.45 ± 0.06
0.54 ± 0.07
0.49 ± 0.08
1.06 ± 0.13
1.07 ± 0.13
1.06 ± 0.13
0.88 ± 0.06
0.85 ± 0.09
0.73 ± 0.04
0.82 ± 0.07
0.72 ± 0.09
0.71 ± 0.05
0.81 ± 0.06
0.77 ± 0.09
0.76 ± 0.04
Baseline linkage analyses
Summary of whole genome linkage scan (WGLS) for peak bone mineral density (BMD) in the total sample (LOD scores > 1.9)
Summary of WGLS in females and males separately (LOD scores > 1.9)
Epistatic interaction analyses
We conducted a large-scale WGLS for PBMD using 2,200 Caucasians from 207 pedigrees, one of the largest linkage studies on PBMD to date. The large sample size and high heritability ensure the power of the present WGLS study.
Based on the WGLS in the entire sample, the most impressing two genetic regions for PBMD are located on 12p12 (LOD = 2.79) and 22q13 (LOD = 2.16). 12p12 anchors a candidate gene,MGP (Matrix Gla protein), which was believed to be involved in bone growth and metabolism [27, 28]. As for 22q13, although no candidate gene is reported in humans, a study in mice has found that theMCHR1 (melanin-concentrating hormone receptor 1) gene, mapping to 22q13 in humans, was significantly correlated with the risk of high bone turnover osteoporosis . Given the linkage evidence in this study, further investigation may be necessary to detect the potential effect of theMCHR1 gene on human PBMD variation. Furthermore, we detected a significant epistatic interaction effect influencing hip PBMD between 12p12 and 22q13 (p = 0.0021), which highlights the importance of these two regions to PBMD variation. As shown by Fig. 4, the linkage signal on 22q13 under two-locus models with only additive effects (LOD = 0.91) is significantly lower than that under two-locus models with additive effects as well as epistatic interaction with 12p13 (LOD = 2.69), suggesting that the genes located in 12p13 influence PBMD variation not only on their own, but also through the interactions with genes on 22q13.
We repeatedly found linkage evidence in Xq27 (LOD = 2.64) with our two WGLS studies for BMD [23, 30], which demonstrates the importance of Xq27 in the genetic control of BMD variation. Although no candidate gene has been reported for Xq27 to date, theBGN (biglycan) gene contained in Xq28, which is very close to Xq27, has been proven to play a important role in osteoblast differentiation and matrix mineralization [31, 32]. Furthermore, two rare skeletal dysplasia syndromes have been mapped to Xq27 [33, 34]. Two other important regions detected in the entire sample were 10p14 and 14q23 for wrist PBMD. 14q23 harbors two prominent candidate genes, namelyER-β (encodes estrogen receptor beta) [35, 36] andBMP-4 (bone morphogenetic protein 4) [37, 38].ER-β has been associated with BMD in several groups [35, 36], andBMP-4 was reported to affect bone formation and osteoblast differentiation [37, 38].
In the sex-specific WGLS for PBMD, two suggestive genetic regions were identified, including 15q26 (LOD = 2.93) for male hip PBMD, and 2p13 (LOD = 2.64) for female wrist PBMD. Although no candidate genes have been reported for 15q26 so far, several Mendelian diseases involved in bone abnormal phenotype have been mapped to 15q26 [39–41], which suggests the effects of genes contained in 15q26 on PBMD variation. As for 2p13, we observed linkage signals in both the total sample and the female subgroup, with LOD scores increasing from 2.04 to 2.64. The potential candidate genes in 2p13 still await further studies. Additionally, we identified several potential sex-specific regions with weaker linkage signals, including 6q24, 7p21, and 11q13. Some candidate genes with bone-related functions have been presented in the above regions. They areER-α (estrogen receptor α) in 6q24 [42, 43]; IL-6 (interleukin 6) in 7p21 [44, 45]; FRA-1 (fos-related Antigen-1),LRP5 (low density lipoprotein receptor-related protein 5), andTCIRG1 (T-cell immune regulator 1) in 11q13 [46–48].
We found that few regions identified in the entire sample overlapped with those in the sex-specific analyses. Similar results have also been reported by previous studies [7–9]. This may partially be explained by the reduced power of sex-specific analyses using smaller sample sizes. In addition, the lower power of linkage analyses involving gender as a covariate in the entire sample than that of sex-specific analyses in identifying sex-specific regions can be a reason . Another important finding of our study is the genetic heterogeneity of PBMD across different skeletal sites. As shown by Tables 3 and 4, the identified regions appeared to affect primarily the wrist or hip, not both, which coincides with previous studies [49, 50]. Site-specific regions support the hypothesis that different genomic regions are responsible for PBMD variation at different skeletal sites.
We also compared our results with those of recent meta-analysis of WGLS for BMD [50, 51]. We found that 10p14, 11q13, 14q23, and 22q13 presented linkage evidence for BMD in the meta-analysis [50, 51]. Given the significant epistatic interaction between 12p12 and 22q13 detected in this study, further fine mapping and functional genomics and proteomics studies may be needed to identify the exact gene and characterize the nature of the interaction. The remaining regions identified in this study did not show a linkage signal in the meta-analysis, which may be explained by the age-specific effects of QTLs on BMD variation. We just included individuals aged 20–50, while the meta-analysis used the sample with ages varying from 18 to 90 years [50, 51].
In conclusion, our efforts identified several suggestive genetic regions underlying PBMD, some of which are epistatic and sex-specific. Further fine mapping, candidate gene association, and molecular biological studies are needed to identify genes contributing to PBMD.
Investigators of this work were partially supported by grants from NIH (K01 AR021170-01, R01 AR45349-01, and R01 GM60402-01A1) and an LB595 grant from the State of Nebraska. The study was also benefitted from grants from National Science Foundation of China, Huo Ying Dong Education Foundation, Hunan Province, Xi’an Jiaotong University, and the Ministry of Education of China. The genotyping experiment was performed by Marshfield Center for Medical Genetics and supported by NHLBI Mammalian Genotyping Service (Contract Number HV48141).