Genomic Regions Identified for BMD in a Large Sample IncludingEpistatic Interactions and Gender-Specific Effects

Peng Xiao,1 Hui Shen,2 Yan-Fang Guo,1,4 Dong-Hai Xiong,1 Yao-Zhong Liu,1 Yong-Jun Liu,1 Lan-Juan Zhao,1

Ji-Rong Long,1 Yan Guo,1,3 Robert R Recker,1 and Hong-Wen Deng2,3,4

ABSTRACT: A genome-wide linkage scan was conducted using a large white sample to identify QTLs forBMD. We found QTLs in the total sample and the gender-specific subgroups, as well as significant epistaticinteractions underlying BMD variations.

Introduction: Low BMD is an important risk factor for osteoporosis and under strong genetic control.Materials and Methods: To identify quantitative trait loci (QTLs) for regulation of BMD, we performed alarge-scale whole genome linkage scan (WGS) involving 4126 individuals from 451 families. In addition to theconventional linkage analyses in the total combined sample of males and females, we conducted epistaticinteraction analyses and gender-specific linkage analyses.Results: Significant linkage was detected on 5q23 for wrist BMD (LOD � 3.39) and 15q13 for female spineBMD (LOD � 4.49). For spine BMD, we revealed significant epistatic interactions between 3p25 and 2q32(p � 0.0022) and between 3p25 and 11q23 (p � 0.0007). We replicated several genomic regions that showedlinkage with BMD in previous studies by others and ours, such as 3p21, 1p36, and Xq27.Conclusions: This study highlights the importance of large sample size, incorporation of epistatic interaction,and consideration of gender-specific effects in identifying QTLs for BMD variation. The results of this studyprovide a foundation for the future fine mapping and gene identification in our population.J Bone Miner Res 2006;21:1536–1544. Published online on July 17, 2006; doi: 10.1359/JBMR.060717

Key words: BMD, whole genome linkage, quantitative trait loci, epistatic interaction, gender-specific

INTRODUCTION

OSTEOPOROSIS IS A major public health problem andresults in >1.3 million fractures per year in the United

States, with an estimated direct cost of more than $17 bil-lion.(1,2) BMD is the most powerful, measurable determi-nant of osteoporotic fracture risk.(3–5) Genetic studies haveshown that 50–90% of BMD variation is genetically deter-mined.(6–9)

Several whole genome linkage scans (WGSs) have beenconducted on BMD, searching for genomic regions harbor-ing quantitative trait loci (QTLs) underlying BMD varia-tion.(10–19) However, the results are largely inconsistent,(9)

which may be partially caused by genetic heterogeneity orlimited statistical power.(18) Moreover, as a complex trait,the variability of BMD is also likely to be affected by theinteraction of multiple genes, whereas epistasis analysis hasnot been performed for BMD earlier. In addition, previousstudies have suggested gender-specific effects for humanBMD variations.(18–21)

In this study, we performed a WGS on BMD in a samplecontaining 4126 whites from 451 pedigrees. To identify po-tential gender-specific QTLs and significant gene–gene in-teractions, we also performed gender-specific linkageanalyses and two-locus epistatic interaction analyses.

MATERIALS AND METHODS

Subjects

The study was approved by the Creighton University In-stitutional Review Board, and all the subjects signed in-formed consent documents before entering the project. Allthe study subjects were whites of European origin and re-cruited from the vicinity of Creighton University.

The sample contains a total of 4498 phenotyped subjectsfrom 451 pedigrees, of whom 4126 were genotyped. Amongthe genotyped subjects, 1816 were used in our earlier WGSstudy.(18) The exclusion criteria are the same for all recruit-ment and have been detailed in our earlier study.(22)

Briefly, patients with chronic diseases and conditions thatmight potentially affect bone mass, structure, or metabo-The authors state that they have no conflicts of interest.

1Department of Biomedical Sciences and Osteoporosis Research Center, School of Medicine, Creighton University, Omaha, Nebraska,USA; 2Departments of Orthopedic Surgery and Basic Medical Science, School of Medicine, University of Missouri-Kansas City, KansasCity, Missouri, USA; 3The Key Laboratory of Biomedical Information Engineering of Ministry of Education, and Institute of MolecularGenetics, School of Life Science and Technology, Xi’an Jiaotong University, Xi’an, Shanxi, China; 4Laboratory of Molecular andStatistical Genetics, College of Life Sciences Hunan Normal University, Changsha, Hunan, China.

JOURNAL OF BONE AND MINERAL RESEARCHVolume 21, Number 10, 2006Published online on July 17, 2006; doi: 10.1359/JBMR.060717© 2006 American Society for Bone and Mineral Research

1536

JO602135 1536 1544 October

lism were excluded. For the 451 pedigrees, 87 were ascer-tained through probands with low BMD (Z score � −1.28at the hip or spine, belonging to the bottom 10% in the agematched population), and 27 were recruited through pro-bands with high BMD (Z score � −1.28 at the hip or spine,belonging to the top 10% in the age matched population).The remaining samples were recruited from 337 families,each having two parents and at least two healthy children25–40 years of age without regard to BMD values. As wehave shown previously,(18) such a sampling scheme will gen-erally not change the identity of loci underlying bone sizevariation in the normal population and thus is unlikely tointroduce heterogeneity of loci identified. The 451 familiesvaried in size from 4 to 416 individuals, with a mean of 11.6± 28.5 (SD). The pedigree structure is shown in Table 1.

Measurements

BMD (g/cm2) for the lumbar spine (L1–L4), total hip(femoral neck, trochanter, and intertrochanteric region),and wrist (ultradistal region of the forearm) were measuredby Hologic 1000, 2000+, or 4500 DXA scanners (Hologic,Bedford, MA, USA). All scanners were calibrated daily,and long-term precision was monitored with external phan-toms. BMD data obtained from different machines weretransformed into a compatible measurement using the for-mula described(23) and the algorithm that we developedin-house and used extensively.(24–26) In particular, membersof the same pedigree were usually measured on the sametype of machine, which ensured minimum or no effect onour linkage analyses caused by measurements by differentscanners. The measurement precision, as reflected by CVsfor spine, hip, and wrist BMD measured by Hologic 2000+were 0.9%, 1.4%, and 2.3%, respectively. Similar CVs wereobtained on Hologic 1000 and 4500.(13) Weight (kg) andheight (m) were measured at the same visit of the BMDmeasurement.

Genotyping

For each subject, DNA was extracted from peripheralblood using the Puregene DNA isolation kit (Gentra Sys-tems, Minneapolis, MN, USA). All the subjects were geno-typed with 410 microsatellite markers (including 392 mark-ers for 22 autosomes and 18 markers for the Xchromosome) from the Marshfield screening set 14 byMarshfield Center for Medical Genetics (Marshfield, WI,USA). The markers have an average population heterozy-gosity of 0.75 and were spaced on average 8.9 cM apart. Thedetailed genotyping protocol is available at http://research.marshfieldclinic.org/genetics/Lab_Methods/

methods.html. Pedcheck(27) was performed to ensure thatthe genotype data conform to Mendelian inheritance pat-tern at all the marker loci. In addition, we usedMERLIN(28) to detect genotyping errors of unlikely recom-bination (e.g., double recombination) in our sample. Thegenotyping error rate was ∼0.3%.

Statistical analyses

Variance component linkage analyses(29–31) for quantita-tive traits were performed using SOLAR (Sequential Oli-gogenic Linkage Analysis Routines).(31) Multipoint LODscores were calculated for chromosomes 1 through 22, andtwo-point LOD scores were computed for chromosome X.Because the current version of SOLAR cannot handle mul-tipoint linkage analysis on chromosome X and other soft-ware, such as MERLIN,(28) cannot process our large pedi-grees, we used MEGA2 (Manipulation Environment forGenetic Analyses)(32) to break down the large pedigreesinto small ones and reserved as many relative pairs as pos-sible. Hence, 451 pedigrees were broken down into 883nuclear families with family size ranging from 4 to 14. Mul-tipoint linkage analysis on chromosome X was accordinglyconducted in the 883 families using the variance componentmethod implemented in MERLIN.(28)

Applying polygenic screen model in SOLAR, age, sex,height, weight, age2, and the sex-by-age interaction weretested for importance on BMD variations, and significantfactors (p � 0.05) were adjusted as covariates for raw BMDvalues.

Empirically adjusted LOD scores for each trait werecomputed by the procedure “lodadj” enclosed in SOLAR.This procedure sampled the null distribution of LOD scores(the distribution obtained under the hypothesis of “no link-age”). This was achieved by permutating 10,000 replicates,with a fully informative marker unlinked to BMD simu-lated in each replicate. IBDs (identity by descent) and LODscores were computed for linkage to BMD for the simu-lated markers. Adjusted LOD scores resulted from multi-plying the original LOD scores by a correction constant, aninverse slop of the regression line of the observed LODscores on the expected LOD scores under the null distri-bution. The proportion of expected LOD scores greaterthat the observed LOD score is the latter’s empirical pvalue.

The epistatic analyses were performed using the chromo-somal region showing the strongest evidence of linkage(i.e., the locus with the highest LOD score) paired withother peak loci that show LOD score � 1.00 in the single-locus multipoint WGS. Two levels of modeling in additionto single-locus modeling were performed: (1) two-locusmodels with only additive effects for each pair of loci and(2) two-locus models with additive effects and an epistaticterm for interaction between the two loci. The increase ofLOD scores under the epistatic model over that under theadditive model suggested genetic interactions. One-tailed pvalues were generated using �2 statistic with 1 degree offreedom for all hypotheses tested with respect to eachBMD variable.(33) Significance for a single test for interac-tion effect on any one variable was assessed at a type I error

TABLE 1. DISTRIBUTION OF PEDIGREE SIZE

Pedigree size(subjects)

Number offamilies

Number ofsubjects

<10 387 183210–19 27 33020–29 12 30730–39 8 270>39 17 1759Total 451 4498

WHOLE GENOME SCAN FOR BMD 1537

rate of 0.05/N according to Bonferroni adjustment for mul-tiple comparison. N was the number of independent testsconducted for each trait.

To identify gender-specific QTLs, we also conductedlinkage analyses in men and women separately in the 451pedigrees. In gender-specific WGS analyses, phenotype val-ues for individuals of the opposite sex were recorded asmissing data. IBD estimation was obtained by the com-bined sample of males and females. Age, height, and weightwere tested, and significant factors (p � 0.05) were adjustedas covariates in the analyses. Furthermore, the Marshfieldgender-specific genetic maps were used in the subgroupanalysis. The adjusted LOD scores and empirical p valueswere also obtained by the procedure “lodadj.” Because weconducted two independent genome-wide searches for thegender-specific analyses, we conducted a Bonferroni adjust-ment approach following Ralston et al.(19) The LODthresholds increased from 1.90 to 2.45 for suggestive linkageand from 3.30 to 3.87 for significant linkage.

RESULTS

Table 2 summarizes the basic phenotypic characteristicsof the study subjects. All three BMD variables are understrong genetic control, with heritability ranging from 0.58 to0.76 (Table 3). In the total of 451 pedigrees, the kurtosisvalues of covariate-adjusted BMD on the spine, hip, andwrist were 0.74, 0.30, and 0.67, respectively. Covariate ofage was not significant for spine BMD; height and age2

were not significant for hip BMD; all the detected six co-variates were significant for wrist BMD. Only the signifi-cant covariates were retained and used in the adjustment ofBMD phenotypes.

Single-locus WGS for the entire sample

Results of the single-locus multipoint WGS on auto-somes in the entire sample are shown in Fig. 1. Detailed

information of suggestive or significant genomic regions(including chromosome X) is presented in Table 4 andFig. 2.

Applying LOD 1.9 and 3.3 for thresholds of “suggestive”and “significant” evidence in pedigree-based linkage analy-sis,(34) we detected a significant linkage at 5q23 (LOD �3.39, p < 0.0001) for wrist BMD (Table 4; Fig. 2C) andsuggestive linkages at 2q, 3p, 3q, 7p (Table 4; Figs. 2A,2B, and 2D), and Xq (Table 4; Figs. 2E and 2F). Particu-larly, 2q32 showed suggestive linkage with both hip andwrist BMD phenotypes (Table 4; Fig. 2A). For chromo-some X, a suggestive two-point LOD score of 2.78 was

TABLE 2. BASIC CHARACTERISTICS OF THE STUDY SUBJECTS

Total(n = 4498)

Female(n = 2682)

Male(n = 1816)

Height (m) 1.695 ± 0.099 1.638 ± 0.068 1.782 ± 0.072Weight (kg) 78.50 ± 18.21 71.33 ± 16.01 89.40 ± 15.79Age (years) 47.7 ± 16.0 47.6 ± 16.0 48.0 ± 16.1BMD (g/cm2)

Spine 1.036 ± 0.162 1.011 ± 0.163 1.072 ± 0.153Hip 0.974 ± 0.161 0.925 ± 0.151 1.045 ± 0.146Wrist 0.467 ± 0.085 0.430 ± 0.070 0.521 ± 0.076

Z scoreSpine 0.37 ± 1.47 0.47 ± 1.45 0.21 ± 1.49Hip 0.28 ± 1.15 0.19 ± 1.38 0.42 ± 1.09Wrist 0.31 ± 1.17 0.36 ± 1.12 0.23 ± 1.23

Values are mean ± SD.The reported BMD are raw phenotype data and normally distributed. Z

scores are computed against the mean and SD as functions of age and sexof a population of healthy, normal people studied around 1990 by theHologic Corporation.

TABLE 3. ESTIMATES OF HERITABILITY FOR BMD PHENOTYPES

IN DIFFERENT GROUPS

BMD Total Female Male

Spine 0.663 ± 0.027 0.743 ± 0.046 0.743 ± 0.060Hip 0.619 ± 0.027 0.668 ± 0.048 0.761 ± 0.062Wrist 0.582 ± 0.030 0.668 ± 0.047 0.619 ± 0.066

Heritabilities are estimated by SOLAR with adjustment for significantcovariates and are presented as mean ± SE.

FIG. 1. Results of single-locus multipoint WGS for autosomes inthe entire sample. The two horizontal lines indicate the suggestivelinkage level (LOD � 1.9) and the significant linkage level (LOD� 3.3).

XIAO ET AL.1538

achieved at Xq27 for wrist BMD (Table 4; Fig. 2E). Furthermultipoint analysis on chromosome X by MERLIN at-tained a LOD score of 1.39 at Xq27 for wrist BMD (Table4; Fig. 2F).

Two-locus epistatic interaction analyses

After applying the Bonferroni correction for multiplecomparisons, the thresholds of significant interaction

were calculated as 0.0042 for spine BMD, 0.0063 for hipBMD, and 0.0042 for wrist BMD. Two significant inter-actions, between 3p25 and 2q32 (epistatic/additive two-locus model, LODE/LODA � 3.78/2.02; p � 0.0022) andbetween 3p25 and 11q23 (epistatic/additive two-locusmodel LODE/LODA � 3.29/1.10; p � 0.0007), were re-vealed on spine BMD (Table 5; Fig. 3). No significant epi-static effects were detected on hip or wrist BMD.

FIG. 2. Suggestive and significant linkageregions detected in the single-locus WGS inthe entire sample. (A–D) Results of auto-somes (wrist BMD, solid line; spine BMD,dotted line; hip BMD, dashed line). (E) Two-point signals on chromosome X for wristBMD. (F) Multipoint signals on chromosomeX for wrist BMD.

TABLE 4. RESULTS OF SINGLE-LOCUS LINKAGE ANALYSES FOR THE ENTIRE SAMPLE

BMD Location Marker LOD*Empirical

p value

Spine 3p25 (37 cM) GGAA4B09N 2.09 0.0020Hip 2q32 (193cM) GATA65C03M 2.11 0.0010

3p21 (78 cM) ATA10H11 2.29 0.00083q27 (200cM) TTTA040 2.55 0.0004

Wrist 2q32 (186cM) GATA65C03M 2.23 0.00065q23 (128cM) GATA62A04 3.39 <0.00017p15 (42 cM) GGAA3F06 2.15 0.0007

Xq27 (166cM) TATC043 2.78 (1.39)† 0.0002 (0.006)†

* The significant LOD score (�3.3) is bold, and others are suggestive LOD scores (�1.9).† For chromosome X, the multipoint LOD score is in parentheses, and the two-point LOD score is outside; the same for the empirical p values.

WHOLE GENOME SCAN FOR BMD 1539

Gender-specific linkage analyses

Results of the gender-specific linkage analyses are sum-marized in Table 6. For females, significant linkages wereachieved at 15q13 (LOD � 4.49, p < 0.0001) for spine BMD(Table 6; Fig. 4B and 4E). In addition, we also observedsuggestive linkage at four other genomic regions (Table 6;Fig. 4). For males, four suggestive linkages were detected(Table 6; Fig. 5).

DISCUSSION

Compared with our earlier WGSs on BMD,(18) thesample size has been enlarged more than double in thisstudy. Consequently, this WGS possesses higher powerthan our previous linkage studies to detect genomic regionsfor BMD in our population.

In the single-locus WGS, we detected a significant link-age for wrist BMD at 5q23 (LOD � 3.39). This region hasalso been linked to whole body BMD(35) and serum insulin-like growth factor binding protein-5 (IGFBP-5) levels(36) inprevious mice WGS studies. 5q23 anchors two candidategenes, IL-4,(37,38) associated with human bone resorptionand BMD, and LOX (lysyl oxidase).(39) Another interest-ing region is 3p21, which showed suggestive linkage for hipBMD in this study and has been repeatedly linked to BMDin other white studies.(16,40,41) Furthermore, 3p21 harborsthree prominent candidate genes for BMD, includingPTHR1 (PTH receptor 1),(41) CCR2 (CC chemokine recep-tor 2),(42) and CCR3 (CC chemokine receptor 3).(43) Par-ticularly, our recent in vivo gene expression study in hu-mans(43) suggested CCR3 may have a critical role in thepathophysiological mechanism of osteoclastogenesis for os-teoporosis.

In two-point analyses on chromosome X, we detectedsuggestive linkage evidence at Xq27 for wrist BMD, whichagreed with our previous linkage finding.(18) The multipointanalyses with broken-down nuclear families showed consis-tent linkage evidence at Xq27 for wrist BMD but withlower LOD scores. This is not unexpected with regard toconsiderable reduced relative pairs in the broken-downnuclear families compared with the original 451 pedigrees,which would result in a corresponding loss of statisticalpower. For instance, there are only ∼5500 relative pairs inthe 883 broken-down families in comparison with >150,000relative pairs in the original 451 pedigrees. However, Xq27

FIG. 3. Significant two-locus epistatic interactions for spineBMD. Single-locus model, dotted line; additive two-locus model,dashed line; epistatic two-locus model, solid line.

TABLE 5. RESULTS OF TWO-LOCUS EPISTATIC LINKAGE ANALYSES FOR SPINE BMD

MaximumLOD locus in

single-locus WGS Tested locus* Marker LOD 1† LODA‡ LODE

§p

(E vs. A)¶

→ 2q32 (193cM) GATA65C03M 1.97 2.02 3.78 0.00223p25 (37 cM)

→ 11q23 (113cM) GATA23E05L 1.19 1.10 3.29 0.0007

* Twelve loci with peak LOD > 1.0 in one-locus model were tested, but only the two significant results were reported here.† LOD scores of single-locus model linkage analyses.‡ LOD scores of additive two-locus model linkage analyses.§ LOD scores of epistatic two-locus model linkage analyses.¶ The significant threshold was obtained by Bonferroni adjustment (significant p � 0.05/12 � 0.0042).

XIAO ET AL.1540

for wrist BMD still achieved the highest LOD score in themultipoint linkage analysis on chromosome X for the threeBMD traits, despite changing the pedigree structures andthe program for analysis. Two rare skeletal dysplasia syn-dromes have been mapped to Xq27.(44,45)

3p25 and 2q32 are another two interesting regions; bothshowed significant or suggestive linkage evidence withBMD in the entire sample or gender-specific subgroups.Moreover, we detected a strong epistatic interaction be-tween these two regions. However, the biological mecha-nisms of the epistasis between 2q32 and 3p25 are unclear.Both regions harbor several potential candidate genes. Oneprominent candidate gene on 2q32 is GDF8 (growth differ-

entiation factor 8/myostatin), a member of the TGF-� su-perfamily, which was functionally related to femur BMD(46)

and spine BMD.(47) Other candidate genes on 2q32 includeSTAT1 (signal transducer and activator of transcription1)(48) and HOXD (homeobox D) gene families.(49,50) Theregion 3p25 contains the PPAR� (peroxisome proliferative-activated receptor �) gene, which plays an important role inmaintaining bone homeostasis(51–53) and has been associ-ated with BMD variation.(54) Future fine mapping andfunctional genomics and proteomics analyses may beneeded to uncover the exact gene and the nature of theinteraction.

Besides 3p25 and 2q32, another significant interactionwas observed between 3p25 and 11q23. 11q23 harbors theMMP (matrix metalloproteinase) gene family, which is in-volved in the breakdown of bone extracellular matrix innormal physiological processes.(55) The gene identity andthe molecular mechanisms underlying the interaction be-tween 3p25 and 11q23 on BMD variation await further ex-ploration.

In the gender-specific WGS, we detected one potentialgender-specific QTL, 15q13 (LOD � 4.49), for spine BMDin females. In other suggestive linkage regions, notably,1p36 has been linked to BMD variation in a number ofprevious WGS studies.(10,12,16,17,56) Furthermore, three can-didate genes at 1p36, including TNFR2 (TNF receptor2),(57) MTHFR (methylenetetrahydrofolate reductase),(58)

and PLOD1 (procollagen-lysine 1),(59) have been associ-ated with BMD variation. However, it may be pointed outthat subsample analyses such as sex-specific analyses heremay suffer from the problems of reduced power. In addi-tion, the existence of gender-specific genes should be ulti-

FIG. 4. Female-specific suggestive and sig-nificant linkage regions. Wrist BMD, solidline; spine BMD, dotted line; hip BMD,dashed line.

TABLE 6. RESULTS OF GENDER-SPECIFIC LINKAGE ANALYSES

BMD Location Marker LOD* p

FemaleSpine 3p25 (27 cM) GGAA4B09N 2.61 0.0020

15q13 (19 cM) GTTTT001P 4.49 <0.0001Wrist 3p25 (24 cM) GATA164B08P 3.31 <0.0001

5q23 (158cM) GATA62A04 2.82 0.00016q24 (200cM) GATA184A08 3.05 0.0001

MaleHip 7p12 (74 cM) GAT24D12P 3.01 0.0002Spine 13q12 (6 cM) GATA23C03P 2.50 0.0003

13q33 (67 cM) GATA51B02M 2.96 0.0003Wrist 1p36 (25 cM) GGAA3A07M 2.81 <0.0001

* The significant LOD score (�3.87) is bold, and others are suggestiveLOD scores (�2.45).

WHOLE GENOME SCAN FOR BMD 1541

mately tested not just by our evidence here of detecting aregion in one sex and not being able to detect it in the othergender, but by the evidence that we detect a region in onesex and are also able to exclude this region in the other sex.Because we could not detect a gene in a subgroup (one sex)does not mean the gene is not there. Genes of differentialeffects (but not sex-specific genes) in the two sexes (gene-by-sex interaction) may also lead to differential detection ofthe genes in the subgroup analyses of the two sexes.

In summary, using total sample, gender-stratified sub-groups and epistatic interaction linkage analyses, this largeWGS study for BMD detected two QTLs (5q23 and 15q13),two pairs of interacting genomic regions (3p25–2q32 and3p25–11q23), and several important suggestive linkage re-gions such as 2q32, 3p21, 3p25, 1p36, and Xq27. The link-ages of 3p25 and 11q23 with spine BMD are consistent withsignificant results of a recent meta-analysis for humanBMD.(60) With those prominent results, this study showsthe importance of the large sample size, the incorporationof epistatic interaction, and the consideration of gender-specific effects in linkage analyses for identifying QTLs forcomplex diseases, such as osteoporosis. The findings of thisstudy give important clues to the following: fine mapping,the candidate gene association, and molecular biologicalstudies.

ACKNOWLEDGMENTS

Investigators of this work were partially supported bygrants from NIH (K01 AR021170-01, R01 AR45349-01,and R01 GM60402-01A1) and an LB595 grant from theState of Nebraska. The study was also benefited from

grants from National Science Foundation of China, HuoYing Dong Education Foundation, Hunan Province, Xi’anJiaotong University, and the Ministry of Education ofChina. The genotyping experiment was performed byMarshfield Center for Medical Genetics and supported byNHLBI Mammalian Genotyping Service (ContractHV48141).

REFERENCES

1. Ray NF, Chan JK, Thamer M, Melton LJ III 1997 Medicalexpenditures for the treatment of osteoporotic fractures in theUnited States in 1995: Report from the National OsteoporosisFoundation. J Bone Miner Res 12:24–35.

2. Melton LJ III 2003 Adverse outcomes of osteoporotic fracturesin the general population. J Bone Miner Res 18:1139–1141.

3. Hui SL, Slemenda CW, Johnston CC Jr 1989 Baseline mea-surement of bone mass predicts fracture in white women. AnnIntern Med 111:355–361.

4. Black DM, Cummings SR, Genant HK, Nevitt MC, Palermo L,Browner W 1992 Axial and appendicular bone density predictfractures in older women. J Bone Miner Res 7:633–638.

5. Nevitt MC, Johnell O, Black DM, Ensrud K, Genant HK,Cummings SR 1994 Bone mineral density predicts non-spinefractures in very elderly women. Study of Osteoporotic Frac-tures Research Group. Osteoporos Int 4:325–331.

6. Smith DM, Nance WE, Kang KW, Christian JC, Johnston CCJr 1973 Genetic factors in determining bone mass. J Clin Invest52:2800–2808.

7. Dequeker J, Nijs J, Verstraeten A, Geusens P, Gevers G 1987Genetic determinants of bone mineral content at the spine andradius: A twin study. Bone 8:207–209.

8. Pocock NA, Eisman JA, Hopper JL, Yeates MG, SambrookPN, Eberl S 1987 Genetic determinants of bone mass in adults.A twin study. J Clin Invest 80:706–710.

9. Liu YJ, Shen H, Xiao P, Xiong DH, Li LH, Recker RR, Deng

FIG. 5. Male-specific suggestive linkage re-gions. Hip BMD, solid line; spine BMD, dot-ted line; wrist BMD, dashed line.

XIAO ET AL.1542

HW 2006 Molecular Genetic Studies of Gene Indentificationfor Osteoporosis: The 2004 update. J Bone Miner Res (inpress).

10. Devoto M, Shimoya K, Caminis J, Ott J, Tenenhouse A,Whyte MP, Sereda L, Hall S, Considine E, Williams CJ, TrompG, Kuivaniemi H, Ala-Kokko L, Prockop DJ, Spotila LD 1998First-stage autosomal genome screen in extended pedigreessuggests genes predisposing to low bone mineral density onchromosomes 1p, 2p and 4q. Eur J Hum Genet 6:151–157.

11. Niu T, Chen C, Cordell H, Yang J, Wang B, Wang Z, Fang Z,Schork NJ, Rosen CJ, Xu X 1999 A genome-wide scan for locilinked to forearm bone mineral density. Hum Genet 104:226–233.

12. Devoto M, Specchia C, Li HH, Caminis J, Tenenhouse A,Rodriguez H, Spotila LD 2001 Variance component linkageanalysis indicates a QTL for femoral neck bone mineral densityon chromosome 1p36. Hum Mol Genet 10:2447–2452.

13. Deng HW, Xu FH, Huang QY, Shen H, Deng H, Conway T,Liu YJ, Liu YZ, Li JL, Zhang HT, Davies KM, Recker RR2002 A whole-genome linkage scan suggests several genomicregions potentially containing quantitative trait Loci for osteo-porosis. J Clin Endocrinol Metab 87:5151–5159.

14. Kammerer CM, Schneider JL, Cole SA, Hixson JE, SamollowPB, O’Connell JR, Perez R, Dyer TD, Almasy L, Blangero J,Bauer RL, Mitchell BD 2003 Quantitative trait loci on chro-mosomes 2p, 4p, and 13q influence bone mineral density of theforearm and hip in Mexican Americans. J Bone Miner Res18:2245–2252.

15. Styrkarsdottir U, Cazier JB, Kong A, Rolfsson O, Larsen H,Bjarnadottir E, Johannsdottir VD, Sigurdardottir MS, BaggerY, Christiansen C, Reynisdottir I, Grant SF, Jonasson K,Frigge ML, Gulcher JR, Sigurdsson G, Stefansson K 2003Linkage of osteoporosis to chromosome 20p12 and associationto BMP2. PLoS Biol 1:E69.

16. Wilson SG, Reed PW, Bansal A, Chiano M, Lindersson M,Langdown M, Prince RL, Thompson D, Thompson E, BaileyM, Kleyn PW, Sambrook P, Shi MM, Spector TD 2003 Com-parison of genome screens for two independent cohorts pro-vides replication of suggestive linkage of bone mineral densityto 3p21 and 1p36. Am J Hum Genet 72:144–155.

17. Karasik D, Cupples LA, Hannan MT, Kiel DP 2004 Genomescreen for a combined bone phenotype using principal compo-nent analysis: The Framingham study. Bone 34:547–556.

18. Shen H, Zhang YY, Long JR, Xu FH, Liu YZ, Xiao P, ZhaoLJ, Xiong DH, Liu YJ, Dvornyk V, Rocha-Sanchez S, Liu PY,Li JL, Conway T, Davies KM, Recker RR, Deng HW 2004 Agenome-wide linkage scan for bone mineral density in an ex-tended sample: Evidence for linkage on 11q23 and Xq27. JMed Genet 41:743–751.

19. Ralston SH, Galwey N, MacKay I, Albagha OM, Cardon L,Compston JE, Cooper C, Duncan E, Keen R, Langdahl B,McLellan A, O’Riordan J, Pols HA, Reid DM, UitterlindenAG, Wass J, Bennett ST 2005 Loci for regulation of bonemineral density in men and women identified by genome widelinkage scan: The FAMOS study. Hum Mol Genet 14:943–951.

20. Jones G, Nguyen TV 2000 Associations between maternalpeak bone mass and bone mass in prepubertal male and femalechildren. J Bone Miner Res 15:1998–2004.

21. Duncan EL, Cardon LR, Sinsheimer JS, Wass JA, Brown MA2003 Site and gender specificity of inheritance of bone mineraldensity. J Bone Miner Res 18:1531–1538.

22. Deng HW, Deng H, Liu YJ, Liu YZ, Xu FH, Shen H, ConwayT, Li JL, Huang QY, Davies KM, Recker RR 2002 A genome-wide linkage scan for quantitative-trait loci for obesity pheno-types. Am J Hum Genet 70:1138–1151.

23. Genant HK, Grampp S, Gluer CC, Faulkner KG, Jergas M,Engelke K, Hagiwara S, Van Kuijk C 1994 Universal standard-ization for dual x-ray absorptiometry: Patient and phantomcross-calibration results. J Bone Miner Res 9:1503–1514.

24. Recker R, Lappe J, Davies K, Heaney R 2000 Characterizationof perimenopausal bone loss: A prospective study. J BoneMiner Res 15:1965–1973.

25. Deng HW, Mahaney MC, Williams JT, Li J, Conway T, Davies

KM, Li JL, Deng H, Recker RR 2002 Relevance of the genesfor bone mass variation to susceptibility to osteoporotic frac-tures and its implications to gene search for complex humandiseases. Genet Epidemiol 22:12–25.

26. Huang QY, Xu FH, Shen H, Deng HY, Conway T, Liu YJ, LiuYZ, Li JL, Li MX, Davies KM, Recker RR, Deng HW 2004Genome scan for QTLs underlying bone size variation at 10refined skeletal sites: Genetic heterogeneity and the signifi-cance of phenotype refinement. Physiol Genomics 17:326–331.

27. O’Connell JR, Weeks DE 1998 PedCheck: A program foridentification of genotype incompatibilities in linkage analysis.Am J Hum Genet 63:259–266.

28. Abecasis GR, Cherny SS, Cookson WO, Cardon LR 2002 Mer-lin–rapid analysis of dense genetic maps using sparse gene flowtrees. Nat Genet 30:97–101.

29. Amos CI 1994 Robust variance-components approach for as-sessing genetic linkage in pedigrees. Am J Hum Genet 54:535–543.

30. Amos CI, Zhu DK, Boerwinkle E 1996 Assessing genetic link-age and association with robust components of variance ap-proaches. Ann Hum Genet 60:143–160.

31. Almasy L, Blangero J 1998 Multipoint quantitative-trait link-age analysis in general pedigrees. Am J Hum Genet 62:1198–1211.

32. Mukhopadhyay N, Almasy L, Schroeder M, Mulvihill WP,Weeks DE 2005 Mega2: Data-handling for facilitating geneticlinkage and association analyses. Bioinformatics 21:2556–2557.

33. Mathias RA, Freidhoff LR, Blumenthal MN, Meyers DA,Lester L, King R, Xu JF, Solway J, Barnes KC, Pierce J, StineOC, Togias A, Oetting W, Marshik PL, Hetmanski JB, HuangSK, Ehrlich E, Dunston GM, Malveaux F, Banks-Schlegel S,Cox NJ, Bleecker E, Ober C, Beaty TH, Rich SS 2001 Ge-nome-wide linkage analyses of total serum IgE using variancecomponents analysis in asthmatic families. Genet Epidemiol20:340–355.

34. Lander E, Kruglyak L 1995 Genetic dissection of complextraits: Guidelines for interpreting and reporting linkage results.Nat Genet 11:241–247.

35. Klein RF, Mitchell SR, Phillips TJ, Belknap JK, Orwoll ES1998 Quantitative trait loci affecting peak bone mineral densityin mice. J Bone Miner Res 13:1648–1656.

36. Mohan S, Masinde G, Li X, Baylink DJ 2003 Mapping quan-titative trait loci that influence serum insulin-like growth factorbinding protein-5 levels in F2 mice (MRL/MpJ X SJL/J). En-docrinology 144:3491–3496.

37. Duncan EL, Brown MA, Sinsheimer J, Bell J, Carr AJ, Words-worth BP, Wass JA 1999 Suggestive linkage of the parathyroidreceptor type 1 to osteoporosis. J Bone Miner Res 14:1993–1999.

38. Saidenberg-Kermanac’h N, Bessis N, Lemeiter D, de Verne-joul MC, Boissier MC, Cohen-Solal M 2004 Interleukin-4 cel-lular gene therapy and osteoprotegerin decrease inflammation-associated bone resorption in collagen-induced arthritis. J ClinImmunol 24:370–378.

39. Khakoo A, Thomas R, Trompeter R, Duffy P, Price R, PopeFM 1997 Congenital cutis laxa and lysyl oxidase deficiency.Clin Genet 51:109–114.

40. Wilson SG, Reed PW, Andrew T, Langdown M, Prince RL,Spector TD 2004 Fine mapping provides further evidence oflinkage for bone mineral density to 3p21. 25th Annual Meetingof the American Society of Bone and Mineral Research, Min-neapolis, MN, USA, September 19–25, 2003.

41. Duncan EL, Brown MA, Sinsheimer J, Bell J, Carr AJ, Words-worth BP, Wass JA 1999 Suggestive linkage of the parathyroidreceptor type 1 to osteoporosis. J Bone Miner Res 14:1993–1999.

42. Yamada Y, Ando F, Niino N, Shimokata H 2002 Associationof a polymorphism of the CC chemokine receptor-2 gene withbone mineral density. Genomics 80:8–12.

43. Liu YZ, Dvornyk V, Lu Y, Shen H, Lappe JM, Recker RR,Deng HW 2005 A novel pathophysiological mechanism forosteoporosis suggested by an in vivo gene expression study ofcirculating monocytes. J Biol Chem 280:29011–29016.

WHOLE GENOME SCAN FOR BMD 1543

44. Dlouhy SR, Christian JC, Haines JL, Conneally PM, HodesME 1987 Localization of the gene for a syndrome of X-linkedskeletal dysplasia and mental retardation to Xq27-qter. HumGenet 75:136–139.I,

45. Zhang W, Amir R, Stockton DW, Van der Veyver IB, BacinoCA, Zoghbi HY 2000 Terminal osseous dysplasia with pigmen-tary defects maps to human chromosome Xq27.3-xqter. Am JHum Genet 66:1461–1464.

46. Hamrick MW 2003 Increased bone mineral density in thefemora of GDF8 knockout mice. Anat Rec A Discov Mol CellEvol Biol 272:388–391.

47. Hamrick MW, Pennington C, Byron CD 2003 Bone architec-ture and disc degeneration in the lumbar spine of mice lackingGDF-8 (myostatin). J Orthop Res 21:1025–1032.

48. Kim S, Koga T, Isobe M, Kern BE, Yokochi T, Chin YE,Karsenty G, Taniguchi T, Takayanagi H 2003 Stat1 functionsas a cytoplasmic attenuator of Runx2 in the transcriptionalprogram of osteoblast differentiation. Genes Dev 17:1979–1991.

49. Dlugaszewska B, Silahtaroglu A, Menzel C, Kubart S, CohenM, Mundlos S, Tumer Z, Kjaer K, Friedrich U, Ropers HH,Tommerup N, Neitzel H, Kalscheuer VM 2006 Breakpointsaround the HOXD cluster result in various limb malforma-tions. J Med Genet 43:111–118.

50. Muragaki Y, Mundlos S, Upton J, Olsen BR 1996 Alteredgrowth and branching patterns in synpolydactyly caused bymutations in HOXD13. Science 272:548–551.

51. Rzonca SO, Suva LJ, Gaddy D, Montague DC, Lecka-CzernikB 2004 Bone is a target for the antidiabetic compound rosig-litazone. Endocrinology 145:401–406.

52. Ali AA, Weinstein RS, Stewart SA, Parfitt AM, ManolagasSC, Jilka RL 2005 Rosiglitazone causes bone loss in mice bysuppressing osteoblast differentiation and bone formation. En-docrinology 146:1226–1235.

53. Akune T, Ohba S, Kamekura S, Yamaguchi M, Chung UI,Kubota N, Terauchi Y, Harada Y, Azuma Y, Nakamura K,Kadowaki T, Kawaguchi H 2004 PPARgamma insufficiencyenhances osteogenesis through osteoblast formation frombone marrow progenitors. J Clin Invest 113:846–855.

54. Ogawa S, Urano T, Hosoi T, Miyao M, Hoshino S, Fujita M,Shiraki M, Orimo H, Ouchi Y, Inoue S 1999 Association ofbone mineral density with a polymorphism of the peroxisomeproliferator-activated receptor gamma gene: PPARgamma ex-pression in osteoblasts. Biochem Biophys Res Commun260:122–126.

55. Shannon WD, Watson MA, Perry A, Rich K 2002 Mantel sta-tistics to correlate gene expression levels from microarrayswith clinical covariates. Genet Epidemiol 23:87–96.

56. Karasik D, Myers RH, Hannan MT, Gagnon D, McLean RR,Cupples LA, Kiel DP 2002 Mapping of quantitative ultrasoundof the calcaneus bone to chromosome 1 by genome-wide link-age analysis. Osteoporos Int 13:796–802.

57. Spotila LD, Rodriguez H, Koch M, Adams K, Caminis J,Tenenhouse HS, Tenenhouse A 2000 Association of a poly-morphism in the TNFR2 gene with low bone mineral density.J Bone Miner Res 15:1376–1383.

58. Miyao M, Morita H, Hosoi T, Kurihara H, Inoue S, Hoshino S,Shiraki M, Yazaki Y, Ouchi Y 2000 Association of methylene-tetrahydrofolate reductase (MTHFR) polymorphism withbone mineral density in postmenopausal Japanese women.Calcif Tissue Int 66:190–194.

59. Spotila LD, Rodriguez H, Koch M, Tenenhouse HS, Tenen-house A, Li H, Devoto M 2003 Association analysis of bonemineral density and single nucleotide polymorphisms in twocandidate genes on chromosome 1p36. Calcif Tissue Int73:140–146.

60. Ioannidis JPA, Ng MY, Sham PC, Zintzaras E, Lewis C, DengHW, Econs MJ, Karasik D, Devoto M, Kammerer CM, Spec-tor TD, Andrew T, Cupples LA, Foroud T, Kiel DP, KollerDL, Langdahl B, Mitchell BD, Peacock M, Recker RR, ShenH, Sol-Church K, Spotila LD, Uitterlinden AG, Wilson SG,Kung AWC, Ralston SH 2006 Collaborative genome searchmeta-analysis for bone mineral density. Proc Natl Acad SciUSA (in press).

Address reprint requests to:Hong-Wen Deng, PhD

Departments of Orthopedic Surgery and BasicMedical Science

School of MedicineUniversity of Missouri–Kansas City2411 Holmes Street, Room M3-C03

Kansas City, MO 64108, USAE-mail: dengh@umkc.edu

Received in original form February 15, 2006; revised form June 24,2006; accepted July 12, 2006.

XIAO ET AL.1544