Large-Scale Genome-Wide Linkage Analysis for Loci Linked to BMDat Different Skeletal Sites in Extreme Selected Sibships

Yi-Hsiang Hsu,1 Xin Xu,1 Henry A Terwedow,1 Tianhua Niu,1 Xuimei Hong,2,3 Di Wu,3 Lihua Wang,4

Joseph D Brain,1 Mary L Bouxsein,5 Steve R Cummings,6 Cliff J Rosen,7 and Xiping Xu2,3

ABSTRACT: Few genome-wide linkage studies of osteoporosis have been conducted in the Asian population.We performed a genome-wide scan involving 3093 adult siblings with at least one sib-pair extremely concor-dant or discordant for hip BMD. Our results indicated four genome-wide significant QTLs for BMD. Incomparison with 12 previous reported linkage studies, we reveal novel linkage regions that have reachingglobal significance.

Introduction: The genetic basis for osteoporosis has been firmly established, but efforts to identify genesassociated with this complex trait have been incomplete, especially in Asian populations. The purpose of thisstudy was to identify quantitative trait loci (QTLs) for BMD in a Chinese population.Materials and Methods: We performed a genome-wide scan involving 3093 siblings 25–64 years of age from941 families, with at least one sib-pair extreme concordant or discordant for total hip BMD from a largecommunity-based cohort (n � 23,327) in Anhui, China. Linkage analysis was performed on BMD residualsadjusted for age, height, weight, occupation, cigarette smoking, physical activity, and alcohol consumptionusing the revised Haseman-Elston regression-based linkage model.Results: Our results revealed significant QTLs on chromosome 7p21.2 for femoral neck BMD (LOD � 3.68)and on chromosome 2q24.3 for total hip BMD (LOD � 3.65). Suggestive linkage regions were found tooverlap among different skeletal sites on chromosomes 2q, 7p, and 16q. Sex-specific linkage analysis furtherrevealed a significant QTL for lumbar spine BMD on chromosome 13q21.1 (LOD � 3.62) in women only.When performing multivariate linkage analysis by combining BMDs at four skeletal sites (i.e., whole body,total hip, femoral neck, and lumbar spine BMD), an additional significant QTL was found at chromosome5q21.2 (LOD � 4.56). None of these significant QTLs found in our study overlapped with major QTLsreported by other studies.Conclusions: This study reveals four novel QTLs in a Chinese population and suggests that BMD at differentskeletal sites may also share common genetic determinants.J Bone Miner Res 2007;22:184–194. Published online on October 23, 2006; doi: 10.1359/JBMR.061015

Key words: genome-wide scan, linkage, BMD, quantitative trait loci, siblings

INTRODUCTION

OSTEOPOROSIS IS A systematic skeletal disease character-ized by low bone mass and microarchitectural dete-

rioration of bone tissue, with a consequent increase in fra-gility and susceptibility to fractures. It is estimated that by2020, one half of all American citizens over age 50 will be atrisk for fractures from osteoporosis and low bone mass.(1)

BMD is frequently used as a proxy measure of osteoporosisand is also the most important predictor for osteoporoticfractures.(2) Several lines of evidence strongly suggest that

BMD is heritable. Twin studies estimate that genetic factorsaccount for 75–85% of interindividual variance in BMD.(3–6)

The notion of high heritability in BMD has also been sup-ported by familial aggregation studies that show high cor-relations in BMD among child-parent pairs.(7,8) Segrega-tion analysis in healthy families suggests that bone mass isunder polygenic control.(9) In an attempt to identify genesthat are involved in the regulation of BMD, genetic linkageanalysis and candidate gene association studies have beenused to implicate several loci and candidate genes.(10) How-ever, genes that contribute to genetic susceptibility to os-teoporosis remain to be elucidated.

Linkage analysis has been successfully used to identifyThe authors state that they have no conflicts of interest.

1Program for Population Genetics, Harvard School of Public Health, Boston, Massachusetts, USA; 2Center for Population Genetics,School of Public Health, University of Illinois at Chicago, Chicago, Illinois, USA; 3Anhui Medical University, Institute of Medicine,Anhui, China; 4School of Public Health, Beijing University, Beijing, China; 5Beth Israel Deaconess Medical Center, Boston, Massachu-setts, USA; 6San Francisco Coordinating Center, UCSF, San Francisco, California, USA; 7Maine Center for Osteoporosis Research andEducation, St Joseph Hospital, Bangor, Maine, USA.

JOURNAL OF BONE AND MINERAL RESEARCHVolume 22, Number 2, 2007Published online on October 23, 2006; doi: 10.1359/JBMR.061015© 2007 American Society for Bone and Mineral Research

184

JO605346 184 194 February

chromosomal regions and to map causal genes in classicalMendelian diseases. A few whole genome-scan linkagestudies on BMDs have been reported. Of which, most stud-ies are in populations of European origin,(11–17) and othersare either in Mexican Americans(18) or a mix of white andblacks.(19,20) Among these studies, genome-wide significantquantitative trait loci (QTLs) of BMD with LOD � 3.6(21)

have been mapped at 1q21–23(19) and 14p12(20) for the lum-bar spine, and at 2pter(18) and 10q21(17) for the femoralneck. Suggestive QTLs of BMD were found at 1p36, 1q31.3,2p23–24, 2q33.1, 3p22, 3q24, 4q25, 4q31–32, 4qter, 6q27,7p14, 9q22, 11q23, 12q23-q24, 13q14, 13q33–34, 16p13,16q23, 18p11, 20p12, 20q13, 21q21.2, 21qter, and Xq27 fordifferent skeletal sites. The reported genomic regions areinconsistent, which reflects the polygenic characteristic ofBMD. It may also be caused by the difference of subjectascertainment, sample size, and statistical methods.

In general, 5–15% of genetic variation occurs across ma-jor populations living on different continents, with the re-maining majority of the variation occurring within suchpopulations.(22) However, comparing the polymorphismfrequencies of 63,012 single nucleotide polymorphisms(SNPs) among four populations, the correlation betweenAsian American and others, including blacks, EuropeanAmericans, and Hispanics, are 0.23, 0.54, and 0.6, respec-tively.(23) Differences in allele frequencies certainly contrib-ute to population differences in the incidence of somecommon diseases,(24,25) such as diabetes(26,27) and hyper-tension,(28) as well as low BMD and osteoporosis.(29–31)

Relatively few linkage studies for BMD have been con-ducted in Asian populations. Thus far, only one relativelysmall-scale genome-wide scan for linkage in Chinese se-lected for hypertension has been reported by our group,with modest linkage to chromosomes 2p21 and 13q34 forforearm BMD.(32) Seeking to extend the initial observation,we conducted a large-scale genome-wide scan for BMD atfour different skeletal sites (i.e., whole body, total hip,femoral neck, and lumbar spine) in 3093 Chinese men andwomen from 941 families using the extreme-sib-pairs (ESP)design to identify the loci that regulate BMD.

MATERIALS AND METHODS

Study population

This study is part of an ongoing community-based osteo-porosis study initiated in 2003 among residents of AnhuiProvince, China. Men and women 25–64 years of age wererecruited.(33) We recruited families when they met the fol-lowing criteria: (1) at least two siblings in the 40–64 agerange and (2) siblings who had the same biological parents.All siblings 25–64 years of age in eligible families were re-cruited. Participants with a history of the following condi-tions were excluded from the study: diabetes mellitus type1, renal failure, chronic infections such as tuberculosis (TB)or other diseases, malignancy, rickets or other metabolicbone diseases, chronic glucocorticoid use, viral cirrhosis,and thyrotoxicosis.

This study was approved by the Human Subjects Com-mittee (the institutional review board) of Harvard School of

Public Health and the Ethics Committee of Anhui MedicalUniversity. Written informed consent was explained to,read, and signed by each participant.

BMC, BMD, and body composition measurement

DXA (GE-lunar Prodigy, Waukesha, WI, USA) wasused to measure soft tissue body composition, BMC (g),and BMD (g/cm2) through whole body, anterior-posteriorlumbar spine and total hip scans. Whole body fat mass andlean mass were expressed in terms of weight (g) and as apercentage of body weight. We randomly selected 71 sub-jects for remeasurements of the whole body and total hip.The maximum interval between the two DXA measure-ments was 30 days. The coefficient of variability (CV%) ofthe reproducibility was 1.34 and 2.05 at whole body andtotal hip, respectively. We defined osteoporosis(34) as a to-tal hip BMD of >2.5 SD below the average peak BMD ofyoung healthy Chinese in the same study area between 25and 30 years of age (T score < −2.5). Osteopenia was de-fined as total hip BMD between 1 and 2.5 SD below thepeak BMD (−2.5 < T score < −1).

Questionnaires

Comprehensive questionnaires were used to collect par-ticipants’ demographic, occupational, and lifestyle informa-tion, reproductive history, disease history, consumption ofalcohol, cigarette smoking, physical activity, history of frac-tures, and daily diet. Drug history, including oral contra-ception, hormone replacement therapy (HRT), and calciumand vitamin D supplements, was also collected. A fracturequestionnaire was applied for those participants who self-reported their fracture history. Fracture sites, treatments,and the age of the participants when they had the fractureswere recorded. We defined menopause status by question-naires.

Definition of ESP

We performed a genome-wide scan on sib-pairs withextreme total hip BMDs. To define ESP, we first groupedsubjects into men, premenopausal women, and postmeno-pausal women, and fitted a group-specific multiple linearregression model for total hip BMD including age, height,weight, physical activity (moderate, heavy), occupation(farmer, non-farmer), cigarette smoking (never, former,current), and alcohol consumption (never, former, current).We defined subjects with >5 h/day carrying (on back orshoulder), raising, lifting, or moving a load �20 kg as hav-ing “heavy” physical activity. The total hip BMD residualfrom this model was calculated for each subject. Our ESPsampling scheme required each sibling’s total hip BMD re-sidual to be either above the 90th percentile or below the10th percentile of the group-specific distribution. Using thiscriterion, we identified a total of 1378 ESPs from 23,327subjects currently enrolled. Because no parents are avail-able, additional siblings are also included in the genome-wide scan to improve the information content. In total, weperformed genotyping on 3897 siblings from 1042 families.

GENOME-WIDE LINKAGE ANALYSIS AND BMD 185

Genotyping

Fasting blood samples were collected and stored in ali-quots at –80°C. Forearm venous blood samples were ob-tained from all study participants by venipuncture, and ge-nomic DNA was extracted from blood lymphocytes. Large-scale genotyping was performed at the Marshfield Centerfor Medical Genetics. Subjects were genotyped at 427 mi-crosatellite markers with an average spacing of ∼10 cM,including 400 from autosomes and 27 from the XY chro-mosomes (Weber marker set v15). One hundred three du-plicated samples were randomly selected and genotyped forreproducibility. Markers with higher genotyping discordantrate, higher missing genotype rate, or deviations fromHardy-Weinberg equilibrium (HWE) at the p < 0.001 levelwere excluded.

Statistical analyses

Kinship relationship: Sib-pairs with mean shared identityby state (IBS) deviating 3 SD from the population meanwere dropped. PEDCHECK(35) and RELPAIR(36) pro-gram were used to check for errors in Mendelian inheri-tance and inconsistencies within pedigrees at all markerloci. The resulting cleaned genotype data were used in sub-sequent analyses.

Phenotypes: Standardized residuals of BMD at differentskeletal sites were used as phenotypes. We first built a pre-dictive model for each BMD trait using all 23,327 recruitedparticipants, stratified by men, premenopausal women, andpostmenopausal women. The predictive model was ad-justed for age, height, weight, occupation, physical activity,cigarette smoking, and alcohol consumption. Residualswere derived from group-specific predictive models at dif-ferent skeletal sites. The trait standardized residuals fromthe predictive model of genotyped subjects were subse-quently used in all of our linkage tests. For sex-specificanalysis, all available genotyping data were used and set thephenotypes for subjects who did not fall into the category ofinteresting as missing values.

Linkage analyses: Univariate linkage analysis was per-formed on the standardized residuals of BMD at wholebody, total hip, femoral neck, and lumbar spine using theregression-based linkage program(37) implemented in theMERLIN package.(38) The pedigree-wide regression link-age program is based on a revised Haseman-Elston (H-E)method that performs regression of estimated identity-by-descent (IBD) sharing between relative pairs on thesquared sums and squared differences of trait values of therelative pairs.

Multivariate analyses of BMDs at different skeletal sitesare highly correlated with each other; therefore, these traitsmay share common linkages. To use such multivariate dataefficiently in the analysis, the XWXW program was used toperform a multivariate linkage analysis that combines indi-vidual test statistics obtained from the trait-specific univari-ate analysis for a global assessment on linkage.(39) Gene-Hunter was used to estimate IBD probabilities (z0, z1, z2)among siblings at any arbitrary chromosomal location.(40)

XWXW takes the IBD dump file from GeneHunter di-rectly for all linkage tests. Multipoint linkage analysis by

using the Unified H-E (UHE) regression-based method,which uses a linear combination of the estimates of theproportion of phenotypic variance explained by the addi-tive effects of the QTL from the squared sums and squareddifferences of traits, was performed in XWXW.(41)

Simulation: The genome-wide significant linkage thresh-olds for LOD score have been shown between 2.8 and 3.6depending on the study design and data informative-ness.(21,42) To adjust for multiple comparisons and estimatethe global genome-wide p value, we conducted a simulationanalysis using the gene-dropping method(43,44) imple-mented in the MERLIN package.(38) In brief, we generated10,000 random datasets under the null hypothesis of nolinkage to observed phenotypes. In each simulation, theoriginal phenotypes were used, and a new data set, with thesame allele frequencies, marker order, genetic distances be-tween markers, and missing-data patterns as the originaldata, was generated. Therefore, “significant signals” ob-tained through simulation are chance findings.

For each simulation, we performed the same set of analy-ses (including univariate linkage analyses for BMD atwhole body, total hip, femoral neck, and lumbar spine; mul-tivariate linkage analysis; and the sex-specific linkage analy-ses) and recorded the maximal LOD score. The distributionof the maximal LOD scores obtained from 10,000 simulateddatasets (LODsimulated) was used to derive the empiricalgenome-wide p value of the observed test statistic for as-sessment of global significance. The adjusted genome-widep values were determined as padjusted � p(LODobserved �LODsimulated).

RESULTS

Characteristics of study subjects

After excluding non–full-sibling individuals, a total of3093 individuals from 941 families were included in the finalanalysis. A majority of families had three siblings geno-typed (47%). Among these 941 families, the numbers offamilies with two, three, four, five, six, and seven siblingsgenotyped were 181, 446, 220, 75, 16, and 6 families, respec-tively. The distribution of ESPs is shown in Table 1. Thereare 332 male ESPs, 246 female ESPs, and 378 male-female

TABLE 1. ESP OF THE STUDY SUBJECTS

ESP M-M F-F M-F Total

EDSP 32 27 38 97EHiCSP 149 102 177 428ELoCSP 151 117 163 431Total 332 246 378 956

Extreme discordant sib-pair (EDSP): one sibling is above the 90% per-centile and the other is below the 10th percentile of the total hip residualdistribution.

Extreme high concordant sib-pair (EHiCSP): both siblings are above the90th percentile of total hip residual distribution.

Extreme low concordant sib-pair (ELoCSP): both siblings are below the10th percentile of total hip residual distribution.

M-M, male-male sib-pair; F-F, female-female sib-pair; M-F, male-femalesib-pair.

HSU ET AL.186

ESPs. Most of the sib-pairs are concordant sib-pairs (90.4%for male sib-pairs and 89% for female sib-pairs).

The principal characteristics of the study subjects areshown in Table 2. In total, 1604 men, 1062 premenopausalwomen, and 427 postmenopausal women were included inthe final analysis. Men had higher weight and height butlower percentage body fat (% Fat) than women. Postmeno-pausal women had lower BMDs at whole body, total hip,femoral neck, and lumbar spine than men and premeno-pausal women. However, no significant differences forBMDs between men and premenopausal women werefound. Sex- and skeletal site–specific T score was calcu-lated. The average T scores of BMD at whole body, femoralneck, and lumbar spine are < −1 for postmenopausalwomen. According to the WHO definition, one half of post-menopausal women can be defined as having osteopenia (Tscore < −1).

The majority of the women used an intrauterine device(IUD) for contraception in our study subjects. Only 2% ofwomen currently used oral contraception. There was nosignificant association between oral contraception use andBMD. No significant association was found between IUDuse and BMD. No woman was currently using HRT. Only22 (0.7%) and 8 (0.3%) subjects regularly took calcium andmultiple vitamins, respectively. No association was foundwith BMD.

Univariate linkage analysis

We genotyped 400 microsatellite markers across auto-somes. Two of the markers were excluded from analysisbecause of a higher drop-out rate (>10%). The averagedrop-out of the remaining 398 markers was 2.3%. The av-erage heterozygosity was 72.2%. The multipoint linkageresults for autosomes are summarized in Fig. 1 along withthe information content at each position. As shown in

Table 3, our linkage analysis suggested several QTLs forBMD.

For BMD at the total hip (TH-BMD), we identified threeQTLs that showed significant or suggestive evidence oflinkage (Fig. 1; Table 3). The highest LOD score was 3.65(pointwise p � 0.00002, padjusted � 0.036) on chromosome2q24.3 at 173 cM between markers D2S1776 and D2S1391and achieved global genome-wide significance level. Theothers were on chromosome 7p21.2 at 22.5 cM betweenmarkers D7S2200 and D7S3051 (LOD � 2.93, p � 0.00012,padjusted�0.128), and on chromosome 16q12.1 at 82.3 cMbetween markers AAT107 and D16S2624 (LOD � 3.14,p � 0.00007, padjusted � 0.085). For BMD at the femoralneck (FN-BMD), we identified three genomic regions thatwere close to the regions identified as QTLs for TH-BMD.The highest LOD score was 3.68 (pointwise p � 0.00002,padjusted � 0.034) on chromosome 7p21.2 at 22.5 cM andachieved the global genome-wide significance level. Theother two QTLs for FN-BMD were on chromosome 2q24.3at 173 cM (LOD � 2.31, p � 0.0005, padjusted � 0.358) andon chromosome 16q22.1 at 94.8 cM between markersD16S2624 and MFD466-TTA001 (LOD � 2.9, p �0.00013, padjusted � 0.134). As for lumbar spine BMD (LS-BMD), we only identified one suggestive QTL on chromo-some 5q21.1 at 107.5 cM between markers D5S1462 andD5S2501 (LOD � 2.71, p � 0.0002, padjusted � 0.181). Fewstudies have reported the QTLs for whole body BMD(WB-BMD); three QTLs were identified as having sugges-tive linkage evidence. They are on chromosome 2q31.1 at180 cM (LOD � 2.71, p � 0.0002, padjusted � 0.181), onchromosome 7p21.3 at 17.5 cM (LOD � 2.47, p � 0.0004,padjusted � 0.268), and on chromosome 16q22.1 at 97.3 cM(LOD � 2.52, p � 0.0003, padjusted � 0.256), respectively.

Sex-specific linkage analysis

Sex-specific linkage analysis was also performed. In ad-dition to the genomic regions on chromosome 7p and 16q,we identified three suggestive QTLs for BMD in women.The highest LOD score was 3.62 (pointwise p � 0.00002,padjusted � 0.038) for LS-BMD on chromosome 13q21.1 at44 cM between markers D13S894 and D13S1807 andachieved global genome-wide significance level. The othersare on chromosome 5q33.1 at 155 cM for TH-BMD be-tween markers D5S1480 and D5S820 (LOD � 2.47, p �0.0004, padjusted � 0.278) and on chromosome 13q13.3 at 34cM for WB-BMD (LOD � 2.5, p � 0.0005, padjusted �0.265). As for men, two suggestive linkage evidences werefound. These are on chromosome 9p21.1 at 50 cM for WB-BMD between markers D9S1121 and D9S304 (LOD �2.65, p � 0.0002, padjusted � 0.202) and on chromosome22q11.23 at 22 cM for LS-BMD between markersATTT019M and D22S689. We also plotted the results,showing the detailed genomic regions on chromosome 2q,5q, 7p, and 16q for all subjects in Fig. 2 and on chromosome13q for women in Fig. 3, respectively.

Multivariate linkage analysis

As shown in Table 4, the highest correlation coefficientwas 0.93 between FN-BMD and TH-BMD, and the lowest

TABLE 2. MEAN (SD) OF AGE, HEIGHT, WEIGHT, BMI, AND

BMDS IN STUDY SUBJECTS STRATIFIED BY SEX AND

MENOPAUSAL STATUS

VariableMen

(n = 1604)

Premenopausalwomen

(n = 1062)

Postmenopausalwomen

(n = 427)

Age (years) 46.4 (7.5) 41.3 (5.2) 52.3 (4.6)Height (cm) 163.5 (5.7) 153.9 (5.1) 151.9 (4.7)Weight (kg) 57.2 (7.8) 52.8 (7.2) 51.1 (8.0)BMI (kg/m2) 21.4 (2.5) 22.3 (2.7) 22.1 (3.0)Percent fat 12.5 (6.7) 27.7 (6.3) 27.5 (7.0)BMDs (g/cm2)Whole body 1.14 (0.10) 1.12 (0.09) 1.03 (0.11)Total hip 0.99 (0.19) 0.99 (0.17) 0.88 (0.18)Femoral neck 0.94 (0.17) 0.94 (0.15) 0.83 (0.16)Lumbar spine 1.07 (0.17) 1.11 (0.15) 0.94 (0.16)T score (BMDs)

Whole body −0.16 (1.28) 0.02 (1.19) −1.18 (1.44)Total hip −0.40 (1.43) −0.01 (1.41) −0.94 (1.48)Femoral neck −0.63 (1.33) −0.09 (1.28) −1.03 (1.22)Lumbar spine −0.44 (1.22) −0.12 (1.14) −1.44 (1.23)

Sex- and skeletal site–specific T score was defined as the SDs away fromthe mean peak BMD of young healthy Chinese in same study area between25–30 years of age (WHO 1994).

Percent fat, percentage body fat measured by DXA.

GENOME-WIDE LINKAGE ANALYSIS AND BMD 187

correlation coefficient was 0.69 between FN-BMD and LS-BMD. BMD at different skeletal sites are highly correlatedwith each other; therefore, these traits may share commonlinkages. We performed a multivariate linkage analysis bycombining the individual test statistics of these four traitstogether. The results are shown in Fig. 1 and Table 3. Thedetailed genomic regions on chromosomes 2q, 5q, 7p, and16q for all subjects and on chromosome 13q for women areshown in Figs. 2 and 3. Genomic regions on chromosomes2q, 5q, 7p, and 16q still show linkage evidences, with peakLODs on chromosome 2q11.2 at 112 cM (LOD � 2.98, p �0.0001, padjusted � 0.117), on chromosome 5q at 112.5 cM(LOD � 4.56, p � 0.000002, padjusted � 0.011), on chro-mosome 7p at 20 cM (LOD � 2.86, p � 0.0001, padjusted �0.142), and on chromosome 16q at 97.3 cM (LOD � 2.49,p � 0.0004, padjusted � 0.268), respectively. In addition tothese genomic regions that have been found to be QTLs fora single trait, we identified two suggestive QTLs for com-

bined BMD traits. They are on chromosome 3q27.1 at 198cM between markers D3S2427 and TTTA040 (LOD �2.61, p � 0.0003, padjusted � 0.218) and on chromosome13q13.3 at 36.5 cM between markers D13S894 andD 1 3 S 1 8 0 7 ( L O D � 2 . 1 6 ,p � 0.0008, padjusted � 0.451), respectively.

DISCUSSION

The primary objective of this study was to investigateQTLs for BMD in a Chinese population. By using largesamples of selected extreme siblings, we found significantQTLs for TH-BMD on chromosome 2q24.3 at 173 cM(LOD � 3.65, p � 0.00002, padjusted � 0.036) and for FN-BMD on chromosome 7p21.2 at 22.5 cM (LOD � 3.68, p �0.00002, padjusted � 0.034). When stratified by sex, we alsofound significant QTLs for LS BMD on chromosome13q21.1 at 44 cM (LOD � 3.62, p � 0.00002, padjusted �0.038). In our study, the linkage evidence was mostly over-

FIG. 1. QTL plots for total hip (TH), femo-ral neck (FN), whole body (WB), lumbarspine (LS) BMD and multivariate analysis(ALL) for all subjects.

HSU ET AL.188

lapped among QTLs for BMD at the total hip, femoralneck, and whole body, but not for the lumbar spine, and themagnitude of significance was different, which suggests thatgenes may regulate BMD differently at different skeletalsites and that gene–environment interactions may also playan important role at different skeletal sites. However, byusing multivariate linkage analysis, a significant QTL forcombined TH-, FN-, LS- and WB-BMDs was found onchromosome 5q at 112.5 cM (LOD � 4.56, p � 0.000002,padjusted � 0.011), which showed evidence that BMD mea-sured at these four sites might also share common geneticdeterminants.

Compared with 12 previous genome-wide linkage stud-ies, we have unique findings in terms of linkage regions andthe magnitude of the global significance level of linkage. Avariety of study designs have been used in previous re-ported genome-wide linkage studies, including analysis offamilies with probands having lower BMD,(11,13,16,17) fami-lies with a history of osteoporosis,(15) families drawn fromthe normal population,(12,18,45) dizygotic twins,(14) sib-pairs from the normal population,(19,20) and sib-pairsselected for hypertension.(32) We estimated the multiplepoint LOD scores for TH-, FN-, LS- and WB-BMDs atprevious reported QTL regions (LOD � 2.2) in our study.Sex-specific LOD scores were shown if previous studiesreported sex-specific QTLs. As shown in Table 5, nosignificant or suggestive QTL has been found in our linkageanalysis for previously reported genomic regions. Onlytwo previously reported regions have LOD scores >1, butfor different skeletal sites. One is on chromosome 10q21(LOD � 1.4 for WB-BMD in our study), which hasbeen reported as a significant QTL for the femoral neck(LOD � 4.42) in white men.(17) The other is on chromo-some 13q14 (LOD � 1.8 for LS-BMD in our study), whichhas been reported as a suggestive QTL for the femoral neck(LOD � 2.51) and femoral trochanter (LOD � 3.46) inMexican-American men.(18)

We compared our reported QTL regions to previousstudies. As shown in Table 6, we searched LOD within40 cM of our significant or suggestive QTL regions in otherstudies. None of the major QTLs found in our study over-lapped with major QTLs reported by other studies. Welisted the highest LOD score reported by other studies, ofwhich the peak genome regions were within 40 cM of ourreported QTL regions. Most of our reported peak LODregions on chromosomes 2q, 7p, and 16q are ∼20 cM dis-tance away from previous reported regions. Kammerer etal.(18) found a suggestive linkage at 13q14 in male MexicanAmericans (LOD � 2.51), which is the closest to our QTLregion on chromosome 13q. However, we found significantQTLs for LS-BMD only in women (LOD � 3.62, p �0.00002), but not in men. The inconsistency of reportedQTLs may be caused by the differences of subject ascer-tainment, sample size, statistical methods, and ethnicity.The ethnicity/geographic location might influence bonedensity through different environmental factors, differentfrequencies of genetic variants, as well as their interactionsthat underlie risk of osteoporosis. For example, the Sp1binding site polymorphism in the collagen, type I, �1(COL1A1) gene has been shown to be associated withlower spine and femoral neck BMD and higher risk of frac-tures(46,47) in the white population, but didn’t find polymor-phism in Asian populations.(30,48) Ferrari et al.(49) reportedthat healthy white males with the V667M polymorphism inthe low-density lipoprotein receptor-related protein 5(LRP5) gene had significantly lower BMC, lower bonearea, and lower stature at the lumbar spine, but we didn’tfind this polymorphism in our study population.(31) How-ever, the Q89R polymorphism in the LRP5 gene (no poly-morphism in white population) is associated with a lowerBMD in the Chinese population.(31)

Mouse models offer a useful tool for causal gene identi-fication and functional analysis.(50) By using F2 intercross ofhigh BMD and low BMD inbred progenitor strains of mice,

TABLE 3. QTLS (LOD � 2.2) FOR BMD AT DIFFERENT SKELETAL SITES FOR ALL SUBJECTS

Location Phenotype* cM† Nearby markers LOD‡ p§ padjusted¶ Ic**

Chr2 2q24.3 TH 173 D2S1776, D2S1391 3.65 0.00002 0.036 0.552q24.3 FN 173 D2S1776, D2S1391 2.31 0.0005 0.358 0.552q31.1 WB 180 D2S1776, D2S1391 2.71 0.0002 0.181 0.402q11.2 ALL 112 D2S1790, D2S2972 2.98 0.0001 0.117 0.52

Chr3 3q27.1 ALL 198 D3S2427, TTTA040 2.61 0.0003 0.218 0.49Chr5 5q21.1 LS 107.5 D5S1462, D5S2501 2.71 0.0002 0.181 0.53

5q21.2 ALL 112.5 D5S1462, D5S2501 4.56 0.000002 0.011 0.50Chr7 7p21.2 TH 22.5 D7S2200, D7S3051 2.93 0.0001 0.128 0.52

7p21.2 FN 22.5 D7S2200, D7S3051 3.68 0.00002 0.034 0.527p21.3 WB 17.5 D7S2200, D7S3051 2.47 0.0004 0.278 0.527p21.2 ALL 20.0 D7S2200, D7S3051 2.86 0.0001 0.142 0.50

Chr16 16q21 TH 82.3 AAT107, D16S2624 3.14 0.00007 0.085 0.5416q22.1 FN 94.8 D16S2624, MFD466-TTA001 2.90 0.00013 0.134 0.4916q22.1 WB 97.3 D16S2624, MFD466-TTA001 2.52 0.0003 0.256 0.5616q22.1 ALL 97.3 D16S2624, MFD466-TTA001 2.49 0.0004 0.268 0.56

* TH, total hip BMD; FN, femoral neck BMD; LS, lumbar spine BMD; WB, whole body BMD; ALL, combined all four skeletal sites.† Centimorgans (cM) based on Marshfield map.‡ Multipoint LOD.§ Pointwise p value without adjustment for multiple comparisons.¶ Empirical genome-wide p value derived from simulation adjusted for multiple comparisons.** Information content.

GENOME-WIDE LINKAGE ANALYSIS AND BMD 189

studies have found several QTLs for bone mass in F2 malemice,(51) whole body BMD,(52) femur length and width,(53)

femur and vertebral BMD,(54) and trabecular BMD volumein F2 female mice.(55) Based on the human-mouse homol-ogy map, several of our reported QTLs have been foundwith observed regions of the mouse QTL including (1) hu-man chromosome 5q with mouse chromosome 18 at markerD18Mit36 with LOD � 13.7 for femoral BMD, LOD � 8.4

for vertebral BMD,(54) and LOD � 4.3 for trabecularthickness(55); (2) human chromosome 13q with mousechromosome 14 at marker D14Mit35 with LOD � 3.4for femur length,(53) and at marker D14Mit160 withLOD � 8.3 for trabecular thickness(55); and (3) humanchromosome 16q with mouse chromosome 8 at markerD8Mit75 with LOD � 3.4 for bone volume of the vertebraltrabecular region.(55) A QTL with LOD � 4 for ultimateforce at lumbar vertebra (L5) in F2 female rats also matchedup with our human chromosome 16q QTL region.(56) Thisevidence suggests that these may be relevant QTLs. How-ever, we should notice that the mouse QTLs were for thetraits that differed from those in our reported QTLs.

To avoid selection bias, it is desirable that the studypopulation is genetically and ethnically homogeneous. Ourstudy population comes from a rural area, with limited ac-cess to public transportation and seems to be relativelystable and fairly homogeneous with respect to ethnicity,lifestyle variables, and social and cultural norms, which sig-nificantly minimized nongenetic confounding factors. Bycomparing the correlation coefficient of the proximal radialBMD among 290 dizygotic (DZ), 170 monozygotic (MZ)twins, and 387 sib-pairs from the same area, the estimatedheritability for proximal radial BMD is 0.70, again indicat-ing a strong genetic component in our study population(unpublished data).

The information content varies according to marker het-erozygosity, marker density, and family structure. Therange of information content in our study is from 0.25 to0.70, with an average of 0.51. Because we did not have theparents’ genotypes, and the majority of families only geno-typed three to four siblings, compared with other linkagestudies with both parents genotyped, the information con-tent in our study is low; therefore, we may not have enough

FIG. 2. LOD score for chromosomes 2, 5, 7, and 16. LOD scoreplot for HIP-BMD is in black solid line, for FN-BMP is in graysolid line, for WB-BMD is in black dot line, and for combined allBMDs is in gray dot line. The flanking markers closest to the peakLOD score are also indicated.

FIG. 3. LOD score for chromosome 13 in women. LOD scorefor HIP-BMD is in black solid line, for FN-BMP is in gray solidline, for WB-BMD is in black dot line, and for LS-BMD is in graydot line. The flanking markers closest to the peak LOD score arealso indicated.

TABLE 4. PEARSON’S CORRELATION COEFFICIENTS BETWEEN

BMD PHENOTYPES

TH FN LS WB

TH 1.00 0.93 0.73 0.83FN 1.00 0.69 0.79LS 1.00 0.81WB 1.00

TH, total hip BMD; FN, femoral neck BMD; LS, lumbar spine BMD;WB, whole body BMD.

HSU ET AL.190

power to detect modest QTLs.(39) A denser set ofmarkers in those significant and suggestive QTL regions isnecessary to increase the resolution of gene mapping.

Each of our QTL regions contains 50–200 known genes.Several genes have been found functionally relevant to

bone metabolism. Interesting potential candidate genes inchromosome 7p QTL region include sclerostin domain-containing protein 1 (SOSTDC1) and interleukin-6 (IL6).SOSTDC1 belongs to a class of bone morphogenetic pro-tein (BMP) antagonists that bind BMPs and regulate their

TABLE 5. LOD OF THIS STUDY AT PREVIOUSLY REPORTED QTL POSITIONS WITH LOD � 2.2

BMDs Location LOD* Ethnicity Reference

LOD* in this study†

WB TH FN LS

Total hip 1p36 2.38§ White 14 0.1 0.01 0.0 0.021qter 3.14 White 12 0.0 0.0 0.0 0.0Xq27 2.57 White 16 — — — —

Femoral neck 1p36 2.29 White 11 0.1 0.01 0.0 0.02pter 3.98‡ Mexican Americans 18 0.0 0.2 0.2 0.02q33.1 2.99‡ White + black 20 0.3 003 0.01 0.34q25 2.55§ White 17 0.0 0.0 0.0 0.07p14 2.28‡ White 17 0.7 0.01 0.3 0.019q22 2.71 White 45 0.0 0.0 0.0 0.010q21 4.42‡ White 17 1.4 0.0 0.0 0.013q14 2.51‡ Mexican Americans 18 0.0 0.2 0.4 1.816p13 2.52‡ White 17 0.2 0.2 0.04 0.0120p12 3.18 White 15 0.0 0.4 0.3 0.1

Femoral Ward’s area 21q21.2 2.40 White + black 20 0.2 0.2 0.5 1.0Femoral trochanter 6q27 2.27 Mexican Americans 18 0.0 0.02 0.3 0.0

12q23-q24 3.00§ White 45 0.6 0.03 0.06 0.613q14 3.46‡ Mexican Americans 18 0.0 0.2 0.4 1.8

Lumbar spine 1q21–23 3.86§ White + black 19 0.0 0.1 0.0 0.01q31.3 2.57 White + black 20 0.6 0.9 0.7 0.02p23–24 2.25 White 11 0.0 0.0 0.04 0.03p22 2.72§ White 14 0.0 0.0 0.0 0.03q25 2.43‡ White 17 0.1 0.01 1.0 0.04q31–32 3.08 White 13 0.1 0.1 0.1 0.24qter 2.28 White 11 0.0 0.1 0.3 0.07p14 2.64 White 16 0.7 0.01 0.3 0.0111q23 3.13 White 16 0.2 0.3 0.07 0.113q33–34 2.43 White 13 0.0 0.1 0.01 0.014p12 3.66 White + black 20 0.0 0.0 0.0 0.016q23 2.28§ White 17 0.3 0.1 0.1 0.118p11 2.83§ White 17 0.1 0.2 0.01 0.020p12 2.89 White 15 0.0 0.4 0.3 0.120q13 3.20§ White 17 0.0 0.04 0.2 0.0

* Multipoint LOD.† The sex-specific LODs were reported if previously conducted studies showed the LOD for men or women only.‡ Men only.§ Women only.—, not available.

TABLE 6. COMPARISON OF QTL REGIONS FOR CHROMOSOME REGIONS WITHIN 40 cM BETWEEN THIS STUDY AND PREVIOUSLY

CONDUCTED GENOME-WIDE LINKAGE STUDIES

This study Previously conducted genome-wide linkage studies

Chromosome* cM Chromosome Phenotype LOD† Ethnicity Reference2q (180 cM) 204 2q33.1 FN 2.16 White + black 207p (17.5 cM) 0 7pter LS 1.93 White 1313q (44 cM)§ 46 13q14 FN 2.51‡ Mexican Americans 1816q (82.3 cM) 67 16q12 LS 2.11 White 1416q21 (104.8 cM)§ 125 16q23 FN 1.99 White 15

* The chromosome regions of significant or suggestive QTLs in our study.† Multipoint LOD.‡ For men only.§ For women only.

GENOME-WIDE LINKAGE ANALYSIS AND BMD 191

signaling during cellular proliferation and differentia-tion.(57) IL6 is located on chromosome 7p15, which is 13 cMaway from our peak LOD region. The LOD score near theIL6 region is 2.29 for FN-BMD. Serum IL6 level is associ-ated with femoral BMD loss in postmenopausal women andaccounts for up to 34% of the total variability of change infemoral BMD.(58) A potential candidate gene in chromo-some 5q QTL region is peroxisome proliferative activatedreceptor �, co-activator 1 (PPARGC1B). PPARGC1B is amajor activator to the PPAR� gene, which has been foundto be associated with BMD(59) and has been found to playa major role of the central pathway of bone for congenicmice (C Ackert-Bicknell and C Rosen, unpublished data,2005). Two potential candidate genes in chromosome 2qQTL region are lipoprotein receptor-related protein-2(LRP2) and secreted frizzled-related protein 3 (SFRP3).LRP2 gene is a member of a family of receptors with struc-tural similarities to the low-density lipoprotein receptor(LDLR). LRP2 knockout animals showed severe vitaminD deficiency and bone disease with abnormal urinary ex-cretion of 25(OH)vitamin D3. LRP2 is essential to preservevitamin D metabolites and to deliver the precursor for gen-eration of 1,25(OH)2 vitamin D3.(60) The protein encodedby SFRP3 is a soluble antagonist of WNT signaling. Defi-cient Wnt signaling has been implicated in disorders of re-duced bone mass. A functional polymorphism (Arg324Gly)in the SFRP3 gene was associated with a higher risk of hiposteoarthritis in female.(61) It is necessary to further exploreif the variants in these potential candidate genes are asso-ciated with BMD.

In summary, we identified three significant QTLs at chro-mosomes 2q24.3, 7p21.2, and 13q21.1 in a Chinese popula-tion. When performing multivariate linkage analysis bycombining BMDs at four skeletal sites (i.e., whole body,total hip, femoral neck, and lumbar spine), an additionalsignificant QTL was found at chromosome 5q21.2, whichsuggests that BMD at different skeletal sites may also sharecommon genetic determinants. The different findings be-tween our study in Chinese and previous studies in whiteimply that genetic variation underlying BMD may dependon ethnicity/geographic location. However, we can not ruleout the facts that it could be caused by the differences instudy design and within ethnic group heterogeneity. Fur-ther studies are currently being undertaken to identify thegenes responsible for BMD regulation in these regions.

ACKNOWLEDGMENTS

This study was supported by NIAMS Grant R01AR045651. We thank Melissa Veno for editing this manu-script.

REFERENCES

1. U.S. Department of Health and Human Service 2004 BoneHealth and Osteoporosis: A Report of the Surgeon General.U.S. Department of Health and Human Services, Office of theSurgeon General, Rockville, MD, USA.

2. Cummings SR, Black DM 1995 Bone mass measurements andrisk of fractures in Caucasian women: A review of findingsfrom prospective studies. Am J Med 98:24–28.

3. Pocock NA, Eisman JA, Hopper JL, Yeates MG, Sambrook

NA, Eberl S 1987 Genetic determinants of bone mass in adults:A twin study. J Clin Invest 80:706–710.

4. Christian JC, Yu PL, Slemenda CW, Johnson CC Jr 1989 Heri-tability of bone mass: A longitudinal study in aging male twins.Am J Hum Genet 44:429–433.

5. Slemenda CW, Christian JC, Williams CJ, Norton JA, JohnsonCC Jr 1991 Genetic determinants of bone mass in adultwomen: A reevaluation of the twin model and the potentialimportance of gene interaction on heritability estimates. JBone Miner Res 6:561–567.

6. Harris M, Nguyen TV, Howard GM, Kelly PJ, Eisman JA 1998Genetic and environmental correlations between bone forma-tion and bone mineral density: A twin study. Bone 22:141–145.

7. Jouanny P, Guillemin F, Kuntz C, Jeandel C, Pourel J 1995Environmental and genetic factors affecting bone mass. Arthri-tis Rheum 38:61–67.

8. Barthe N, Basse-Cathalinat B, Meunier PJ, Ribot C, March-andise X, Sabatier JP, Braillon P, Thevenot J, Sutter B, 1998Measurement of bone mineral density in mother-daughterpairs for evaluating the family influence on bone mass acqui-sition: A GRIO survey. Osteoporosis Int 8:379–384.

9. Gueguen R, Jouanny P, Guillemin F, Kuntz C, Pourel J, SiestG 1995 Segregation analysis and variance components analysisof bone mineral density in healthy families. J Bone Miner Res10:2017–2022.

10. Liu YZ, Liu YJ, Recker R, Deng HW 2003 Molecular studiesof identification of genes for osteoporosis: The 2002 update. JEndocrinol 177:147–196.

11. 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.

12. Karasik D, Myers RH, Cupples LA, Hannan MT, Gagnon DR,Herbert A, Kiel DP 2002 Genome screen for quantitative traitloci contributing to normal variation in bone mineral density:The Framingham Study. J Bone Miner Res 17:1718–1727.

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. 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.

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. Shen H, Zhang Y-Y, Long J-R, Xu F-H, Liu Y-Z, Xiao P, ZhaoL-J, Xiong D-H, Liu Y-J, Dvornyk V, Rocha-Sanchez S, LiuP-Y, Li J-L, Conway T, Davies KM, Recker RR, Deng H-W2004 A genome-wide linkage scan for bone mineral density inan extended sample: Evidence for linkage on 11q23 and Xq27.J Med Genet 41:743–751.

17. Ralston SH, Galwey N, MacKay I, Albagha OME, 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.

18. 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.

HSU ET AL.192

19. Koller DL, Econs MJ, Morin PA, Christian JC, Hui SL, ParryP, Curran ME, Rodriguez LA, Conneally PM, Joslyn G, Pea-cock M, Johnston CC, Foroud T 2000 Genome screen forQTLs contributing to normal variation in bone mineral densityand osteoporosis. J Clin Endocrinol Metab 85:3116–3120.

20. Peacock M, Koller DL, Lai D, Hui S, Foroud T, Econs MJ 2005Sex-specific quantitative trait loci contribute to normal varia-tion in bone structure at the proximal femur in men. Bone37:467–473.

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

22. Hinds DA, Stuve LL, Nilsen GB, Halperin E, Eskin E, Ball-inger DG, Frazer KA, Cox DR 2005 Whole-genome patternsof common DNA variation in three human populations. Sci-ence 307:1072–1079.

23. Bamshad M 2005 Genetic influences on health. Does race mat-ter? JAMA 294:937–946.

24. Burchard EG, Ziv E, Coyle N, Gomez SL, Tang H, Karter AJ,Mountain JL, Perez-Stable EJ, Sheppard D, Risch N 2003 Theimportance of race and ethnic background in biomedical re-search and clinical practice. N Engl J Med 348:1170–1175.

25. Race, Ethnicity, and Genetics Working Group 2005 The use ofracial, ethnic, and ancestral categories in human genetics re-search. Am J Hum Genet 77:519–532

26. Fullerton SM, Bartoszewicz A, Ybazeta G, Horikawa Y, BellGI 2002 Geographic and haplotype structure of candidate type2 diabetes susceptibility variants at the calpain-10 locus. Am JHum Genet 70:1096–1106.

27. Carulli L, Rondinella S, Lombardini S, Canedi I, Loria P, Ca-rulli N 2005 Review article: Diabetes, genetics and ethnicity.Aliment Pharmacol Ther 22:16–19.

28. Zhu X, Luke A, Cooper RS, Quertermous T, Hanis C, MosleyT, Gu CC, Tang H, Rao DC, Risch N, Weder A 2005 Admix-ture mapping for hypertension loci with genome-scan markers.Nat Genet 37:177–181.

29. Gong G, Haynatzki G 2003 Association between bone mineraldensity and candidate genes in different ethnic populations andits implications. Calcif Tissue Int 72:113–123.

30. Lei SF, Deng FY, Liu XH, Huang QR, Qin Y, Zhou Q, JiangDK, Li YM, Mo XY, Liu MY, Chen XD, Wu XS, Shen H,Dvornyk V, Zhao L, Recker RR, Deng HW 2003 Polymor-phisms of four bone mineral density candidate genes in Chi-nese populations and comparison with other populations ofdifferent ethnicity. J Bone Miner Metab 21:34–42.

31. Hsu YH, Niu T, Terwedow H, Xu X, Feng Y, Li Z, Brain J,Rosen CJ, Laird N, Xu X 2005 Variation in low-density lipo-protein receptor-related protein 5 (LRP-5) gene is associatedwith bone mineral density in men and women. Annual Meetingof the American Society of Human Genetics, October 26–31,2005, Salt Lake City, UT, USA.

32. 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.

33. Hsu YH, Venners SA, Terwedow HA, Feng Y, Niu T, Li Z,Laird N, Brain JD, Cummings SR, Bouxsein ML, Rosen CJ,Xu X 2006 Relation of body composition, fat mass, and serumlipids to osteoporotic fractures and bone mineral density inChinese men and women. Am J Clin Nutr 83:146–154.

34. World Health Organization 1994 Assessment of Fracture Riskand Its Application to Screening for Postmenopausal Osteo-porosis. World Health Organization, Geneva, Switzerland.

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

36. Duren WL, Epstein M, Li M, Boehnke M RELPAIR: A pro-gram that infers the relationships of pairs of individuals basedon market data. Available at http://www.sph.umich.edu/statqen/boehnke/relDair.html. Accessed June 2004.

37. Sham PC, Purcell S, Cherny SS, Abecasis GR 2002 Powerfulregression-based quantitative-trait linkage analysis of generalpedigrees. Am J Hum Genet 71:238–253.

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

39. Xu X, Palmer LJ, Horvath S, Wei LJ 2001 Combining multiplephenotypic traits optimally for detecting linkage with sib-pairobservations. Genet Epidemiol 21(Suppl 1):S148–S153.

40. Kruglyak L, Daly MJ, Reeve-Daly MP, Lander ES 1996 Para-metric and nonparametric linkage analysis: A unified multi-point approach. Am J Hum Genet 58:1347–1363.

41. Xu X, Weiss S, Xu X, Wei LJ 2000 A unified Haseman-Elstonregression method for testing linkage with quantitative traitAm. J Hum Genet 67:1025–1028.

42. Wiltshire S, Cardon LR, McCarthy MI 2002 Evaluating theresults of genomewide linkage scans of complex traits by locuscounting. Am J Hum Genet 71:1175–1182.

43. Sawcer S, Jones HB, Judge D, Visser F, Compston A, Good-fellow PN, Clayton D 1997 Empirical genomewide significancelevels established by whole genome simulations. Genet Epide-miol 14:223–229.

44. Kruglyak L, Daly MJ 1998 Linkage thresholds for two-stagegenome scans. Am J Hum Genet 62:994–997.

45. Karasik D, Cupples LA, Hannan MT, Kiel DP 2003 Age, gen-der, and body mass effects on quantitative trait loci for bonemineral density: The Framingham Study. Bone 33:308–316.

46. Grant SFA, Reid DM, Blake G, Herd R, Fogelman I, RalstonSH 1996 Reduced bone density and osteoporosis associatedwith a polymorphic Sp1 binding site in the collagen type 1�1gene. Nat Genet 14:203–205.

47. Mann V, Hobson EE, Li B, Stewart TL, Grant SF, Robins SP,Aspden RM, Ralston SH 2001 A COL1A1 Sp1 binding sitepolymorphism predisposes to osteoporotic fracture by affect-ing bone density and quality. J Clin Invest 107:899–907.

48. Beavan S, Prentice A, Dibba B, Yan L, Cooper C, Ralston SH1998 Polymorphism of the collagen type 1a1 gene and ethnicdifferences in Hip-fracture rates. N Engl J Med 339:351–352.

49. Ferrari SL, Deutsch S, Choudhury U, Chevalley T, Bonjour JP,Dermitzakis ET, Rizzoli R, Antonarakis SE 2004 Polymor-phisms in the low-density lipoprotein receptor-related protein5 (LRP5) gene are associated with variation in vertebral bonemass, vertebral bone size, and stature in whites. Am J HumGenet 74:866–875.

50. Rosen CJ, Beamer WG, Donahue LR 2001 Defining the ge-netic of osteoporosis: Using the mouse to understand man.Osteoporos Int 12:803–810.

51. Shimizu M, Higuchi K, Bennett B, Xia C, Tsuboyama T, KasaiS, Chiba T, Fujisawa H, Kogishi K, Kitado H 1999 Identifica-tion of peak bone mass QTL in a spontaneously osteoporoticmouse strain. Mamm Genome 10:81–87.

52. 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.

53. Drake TA, Hannani K, Kabo JM, Villa V, Krass K, Lusis AJ2001 Genetic loci influencing natural variations in femoralbone morphometry in mice. J Orthop Res 19:511–517.

54. Beamer WG, Shultz KL, Donahue LR, Churchill GA, Sen S,Wergedal JR, Baylink DJ, Rosen CJ 2001 Quantitative traitloci for femoral and lumbar vertebral bone mineral density inC57BL/6J and C3H/HeJ inbred strains of mice. J Bone MinerRes 16:1195–1206.

55. Bouxsein ML, Uchiyama T, Rosen CJ, Shultz KL, DonahueLR, Turner CH, Sen S, Churchill GA, Muller R, Beamer WG2004 Mapping quantitative trait loci for vertebral trabecularbone volume fraction and microarchitecture in mice. J BoneMiner Res 19:587–599.

56. Alam I, Sun Q, Liu L, Koller DL, Fishburn T, Carr LG, EconsMJ, Foroud T, Turner CH 2005 Whole-genome scan for link-age to bone strength and structure in inbred Fischer 344 andLewis rats. J Bone Miner Res 20:1589–1596.

57. Yanagita M, Oka M, Watabe T, Iguchi H, Niida A, TakahashiS, Akiyama T, Miyazono K, Yanagisawa M, Sakurai T 2004USAG-1: A bone morphogenetic protein antagonist abun-dantly expressed in the kidney. Biochem Biophys Res Com-mun 316:490–500.

GENOME-WIDE LINKAGE ANALYSIS AND BMD 193

58. Scheidt-Nave C, Bismar H, Leidig-Bruckner G, Woitge H, Sei-bel MJ, Ziegler R, Pfeilschifter J 2001 Serum interleukin 6 is amajor predictor of bone loss in women specific to the firstdecade past menopause. J Clin. Endocrinol Metab 86:2032–2042.

59. Kiel DP, Ferrari S, Cupples LA, Karasik D, Dupuis J, RosenCJ, Imamovic A, Demissie S 2005 Polymorphisms in thePPAR� gene influence bone density in humans. J Bone MinerRes 20:S234.

60. Nykjaer A, Dragun D, Walther D, Vorum H, Jacobsen C, HerzJ, Melsen F, Christensen EI, Willnow TE 1999 An endocyticpathway essential for renal uptake and activation of the steroid25-(OH) vitamin D3. Cell 96:507–515.

61. Loughlin J, Dowling B, Chapman K, Marcelline L, Mustafa Z,Southam L, Ferreira A, Ciesielski C, Carson DA, Corr M 2004Functional variants within the secreted frizzled-related protein

3 gene are associated with hip osteoarthritis in females. ProcNatl Acad Sci USA 101:9757–9762.

Address reprint requests to:Xiping Xu, MD, PhD

Center for Population GeneticsSchool of Public Health

University of Illinois at Chicago1603 West Taylor Street, Room 978A

Chicago, IL 60612, USAE-mail: yhh5402@gmail.com

xipingxu@uic.edu

Received in original form May 26, 2006; revised form September12, 2006; accepted October 18, 2006.

HSU ET AL.194