This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Copyright Permission Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Peacock, M. Articles by Foroud, T. Search for Related Content PubMed PubMed Citation Articles by Peacock, M. Articles by Foroud, T.

Genetics of Osteoporosis

Department of Medicine (M.P., M.J.E.), Department of Orthopaedic Surgery (C.H.T.), and Department of Medical and Molecular Genetics (M.J.E., T.F.), Indiana University School of Medicine, Indianapolis, Indiana 46202; and Department of Biomedical Engineering (C.H.T.), Indiana University Purdue University, Indianapolis, Indiana 46202

Correspondence: Address all correspondence and requests for reprints to: Munro Peacock, M.D., General Clinical Research Center, University Hospital and Outpatient Center, Room 5595, 550 North University Boulevard, Indianapolis, Indiana 46202. E-mail: mpeacock{at}iupui.edu

	Abstract

Top Abstract I. Introduction II. Osteoporosis III. Heritability IV. Locating the Susceptibility... V. Identification of the... VI. Animal Studies VII. Bioethics References

 
Osteoporosis is a common multifactorial disorder of reduced bone mass. The disorder in its most common form is generalized, affecting the elderly, both sexes, and all racial groups. Multiple environmental factors are involved in the pathogenesis. Genes also play a major role as reflected by heritability of many components of bone strength. Quantitative phenotypes in bone strength in the normal population do not conform to a monogenetic mode of inheritance. The common form of osteoporosis is generally considered to be a polygenic disorder arising from the interaction of common polymorphic alleles at quantitative trait loci, with multiple environmental factors. Finding the susceptibility genes underlying osteoporosis requires identifying specific alleles that coinherit with key heritable phenotypes in bone strength. Because of the close correspondence among mammalian genomes, identification of the genes underlying bone strength in mammals such as the mouse is likely to be of major assistance in human studies. Identification of susceptibility genes for osteoporosis is one of several important approaches toward the long-term goal of understanding the molecular biology of the normal variation in bone strength and how it may be modified to prevent osteoporosis. As with all genetic studies in humans, these scientific advances will need to be made in an environment of legal and ethical safeguards that are acceptable to the general public.

I. Introduction

II. Osteoporosis

A. Normal variation in bone mass and structure

B. Definition of osteoporosis

C. Diagnosis of osteoporosis

D. Phenotypes predicting fracture risk

E. Bone strength and physical activity

F. Fragility fractures

III. Heritability

A. Estimation of heritability

B. Heritability of bone mass

C. Heritability of bone size and structure

D. Heritability of bone loss

E. Heritability of bone turnover

F. Heritability of fracture

G. Heritability of falls

IV. Locating the Susceptibility Genes

A. Candidate gene approach

B. Candidate gene association studies with BMD

C. Family-based association approach: the transmission disequilibrium test

D. Linkage approach

E. Family studies

F. Genome scans in sibling pairs

V. Identification of the Susceptibility Genes

VI. Animal Studies

VII. Bioethics

	I. Introduction

Top Abstract I. Introduction II. Osteoporosis III. Heritability IV. Locating the Susceptibility... V. Identification of the... VI. Animal Studies VII. Bioethics References

 
OSTEOPOROSIS IS A common multifactorial disorder of reduced bone mass manifesting clinically as fragility fracture. Fracture arises as a stochastic event from minor trauma acting on a skeleton that has reduced bone strength (1) (Fig. 1). The pathogenesis of fragility fracture almost always involves trauma and is not necessarily associated with reduced bone mass. Thus, fragility fracture should neither be used synonymously nor interchangeably as a phenotype for osteoporosis.

View larger version (20K): [in this window] [in a new window]   Figure 1. Schema of the effect of environmental and genetic risk factors on the interaction between bone strength and trauma that leads to osteoporotic fracture.

	II. Osteoporosis

Top Abstract I. Introduction II. Osteoporosis III. Heritability IV. Locating the Susceptibility... V. Identification of the... VI. Animal Studies VII. Bioethics References

 
A. Normal variation in bone mass and structure
Bone mass and skeletal proportions exhibit a wide range in the normal population (Fig. 2). This variation is further magnified by differences due to age, sex, and race. Skeletal size and mass increase into adolescence. With closure of the epiphyses, the skeleton achieves adult size, although further accumulation of bone mineral continues for several years thereafter. The range in peak bone mineral density and content at the femoral neck in white women measured by dual x-ray densitometry is 50% and 60% of the mean, respectively (Normal White Population Database, DPX-IQ Reference Manual, Documentation Version 5/96, Lunar Corp., Madison, WI). After the third decade, bone mass is steadily lost until the end of life. About 22% of the bone mineral density is lost at the femoral neck in white women from age 30 to age 80 yr (Normal White Population Database, DPX-IQ Reference Manual, Documentation Version 5/96, Lunar Corp., Madison, WI) (17). The main determinants of bone mass in the elderly, who are at greatest risk of osteoporosis, are peak bone mass and the rate of age-related bone loss (18). At all ages, the variance remains relatively stable (19). Furthermore, bone mass among different skeletal sites is highly correlated (20, 21). Age-related bone loss is accompanied by deterioration in bone architecture (22, 23) and an overall expansion of the skeleton (24). Men on average have larger skeletons and have more bone mass at all ages than women (Fig. 2). American blacks have more bone mass than American and European whites (25, 26), who in turn have more bone mass than Asians (27, 28). Common polymorphisms probably underlie much of the normal variation in bone mass and structure. Thus, bone mass and structure phenotypes are key quantitative traits that are used for searching for the susceptibility genes for osteoporosis.

View larger version (22K): [in this window] [in a new window]   Figure 2. Normal variation (mean and 2 SD) and change in BMD with age in healthy men (black circle) and women (open circle) (Normal Population Database, DPX-IQ Reference Manual, Documentation Version 5/96, Lunar Corp., Madison, WI). Peak bone mass at hip and spine for measurement on Lunar machines is taken as the mean BMD between age 20 and 40 yr, but this age range varies with DXA machine manufacturer.

D. Phenotypes predicting fracture risk
Measurement of BMD by DXA predicts fracture risk (50), particularly when it is made at the skeletal site of future fracture (51). Although there are inherent inaccuracies in the technique (52), it is widely used as a key phenotype in searching for susceptibility genes for osteoporosis. The hip and spine are commonly measured sites because of their high incidence of osteoporotic fracture. For each SD decrease in T score, the lifetime risk of fragility fracture about doubles (50). However, skeletal structure also contributes independently to fracture risk and can be obtained from radiographs (20), QCT (53), and DXA images (54). Although phenotypes based on direct measurements of biomechanical strength cannot be made in humans, a variety of parameters related to bone strength can be derived from structural variables (55, 56). Deterioration in bone quality also leads to fracture. By definition, this is not measurable except by destructive biomechnical tests. New techniques using ultrasound (57, 58) and magnetic resonance (59) may capture some quantitative components of bone quality. Although not all studies agree, fracture risk in elderly women may also be predicted from bone turnover as assessed by biochemical markers (60). Thus, key bone phenotypes involved in fracture risk relate not only to bone mass but also to bone structure, bone loss, and possibly to bone turnover. Because of the wide variety of key phenotypes and because it is not known how the susceptibility genes for osteoporosis affect the skeleton, measurement of multiple skeletal phenotypes is essential. However, it should also be appreciated that in addition to these skeletal risk factors, the frequency of falls (61, 62, 63), the direction of falling (63, 64), and the occurrence of previous fracture (61) (65) are also risk factors for osteoporotic fractures.

E. Bone strength and physical activity
Bone strength cannot be directly measured in vivo in humans. However, it may be assessed indirectly from measuring components of mass and the distribution of structure. Such measures can be used as quantitative traits in searching for the susceptibility genes for osteoporosis (66) and are of particular interest at skeletal sites such as the hip and spine where fragility fractures are common. The strength of bone is normally maintained in balance with the amount of physical activity the skeleton is subject to through mechanisms collectively known as the mechanostat (67). However, the effectiveness of the mechanostat to achieve this balance may also be under genetic influences. Muscle mass, an important covariate of bone strength and an integral component of the mechanostat, is a key phenotype and can be measured simultaneously with BMD by DXA and QCT.

F. Fragility fractures
Fragility fractures may affect any bone. However, they are common at the vertebra (65, 68) and the upper end of the femur (61, 69) (Fig. 3). The incidence of fracture rises steeply with age after the age of 50, and hip fracture is higher in women than men and lower in black than white Americans (70, 71). Thus, fragility fracture incidence inversely tracks bone mass. However, although bone mass predicts fracture risk within discrete populations, it does not identify individuals who will fracture (50). This is explained in part by the fact that other factors such as bone structure predict fracture independent of bone mass (37). However, it is mainly due to the fact that fracture itself is a complex disorder with multiple underlying risk factors (72), many of which are unrelated to bone strength. Fragility fracture, therefore, is a highly complex phenotype in its own right, which should not be used by itself to diagnose or to evaluate risk factors for osteoporosis. As such, it is unlikely to be a useful phenotype in searching for the genes underlying osteoporosis.

View larger version (20K): [in this window] [in a new window]   Figure 3. Incidence of hip fracture in white (open square), Asian (x), and black (black square) men (panel A) and white (open circle), Asian (x), and black (black circle) women (panel B) (71 339 ).

	III. Heritability

Top Abstract I. Introduction II. Osteoporosis III. Heritability IV. Locating the Susceptibility... V. Identification of the... VI. Animal Studies VII. Bioethics References

B. Heritability of bone mass
Studies over the last 30 yr in healthy twins (74, 75, 78, 79, 80) consistently demonstrate a large genetic contribution to bone mass even into old age (Fig. 4). A number of family studies using healthy parent-children pairs (81, 82, 83, 84, 85, 86), healthy sister pairs (87), and parent-children pairs in whom the parent had osteoporosis (88, 89, 90) have corroborated the major role of genes in determining bone mass. Furthermore, heritability of bone mass can be detected during childhood even though the skeleton is undergoing major changes in both skeletal size and mass (78, 80, 91, 92).

View larger version (109K): [in this window] [in a new window]   Figure 4. Pair of elderly Mz twins showing striking similarity both in height, weight, and body dimensions and in vertebral shape, structure, and mineral density. Sister A presented with backache, whereas sister B was asymptomatic.

C. Heritability of bone size and structure
Height and other anthropometric variables related to skeletal size have been known for many years to be highly heritable (94, 95, 96 ; see Table 2). In the early reports on the heritability of bone mass, variables related to bone size and structure such as forearm width (78) and phalangeal cortical and medullary width (79) were shown to be as highly heritable as bone mass. More recent reports examining structural variables that are important in maintaining bone strength at skeletal sites where osteoporotic fracture is common, highlight the strong heritability of bone structure at both the hip (97, 98) and at the spine (99). Hip axis length measured from DXA images (97) and femur axis length measured from radiographs (98) are both highly heritable and predict risk for fracture at the hip.

E. Heritability of bone turnover
Bone formation and resorption can be assessed from a number of biochemical markers in blood and urine (102). These vary with age (103), race (104), and disease (102) and its treatment (105). Formation and resorption markers show a strong correlation among themselves (103). These markers do not measure bone mass or structure but in some circumstances may reflect bone gain (103) or bone loss (106, 107). In elderly white women, they may also predict hip fracture (60). Their relationship to bone strength is tenuous. Nevertheless, a number of studies have reported that there is a heritable component to bone turnover markers although with little consistency in the turnover markers across studies and little corroboration among markers within studies (108, 109).

F. Heritability of fracture
The heritability of fracture, as expected with such a complex phenotype, is not strong. In a large prospective study of white American women 65 yr of age or older, a history of hip fracture in their mother doubled the risk of hip fracture (110). The increased risk remained after adjusting for BMD, indicating that factors other than bone mass are involved. In a study of 2308 Mz and 5241 Dz male and female twins followed prospectively in Finland, although the concordance in the rate of fracture (111) was greater in Mz twins than Dz twins, the magnitude of difference was small and osteoporotic fracture was not strongly influenced by genes. When these data were reanalyzed using a variance components approach (112), genetic factors were estimated to account for, at most, only about one third of the variance in the liability to fracture. In a questionnaire study of white American women older than 40 yr identified because they had participated in any type of bone study, there was a significant history, recalled by the proband, of a forearm fracture in their mother and their sister. The heritability of forearm fracture was calculated to be less than one third (113). Although there appears to be a genetic component to fragility fracture, it seems equally likely from these studies that the underlying genes may not relate to bone strength but to the risk factors for falls, which are also highly heritable (114, 115).

G. Heritability of falls
Trauma is an essential component in most fractures (Fig. 1). The most common source of trauma in the elderly is falls. Most age-related osteoporotic fractures, particularly at the hip, result from simple falls from a standing height (61, 62, 63, 116, 117). Falls increase with age (118) and are more common in women than men (116). Although the minority of falls result in hip fracture, the majority of hip fractures result from a fall. The incidence of hip fracture is greatly reduced if the energy of the fall is attenuated by padding around the hip (119, 120). The number of falls along with decreased bone strength, age, and previous fracture are major predictors of hip fracture (61, 63). Falls are a highly complex phenotype with multiple environmental risk factors (62, 72, 116, 118, 121, 122, 123). Although falls also have a heritable component (114, 115), the susceptibility genes for fracture resulting from falls are unlikely to relate to the genes underlying bone strength.

	IV. Locating the Susceptibility Genes

Top Abstract I. Introduction II. Osteoporosis III. Heritability IV. Locating the Susceptibility... V. Identification of the... VI. Animal Studies VII. Bioethics References

 
A. Candidate gene approach
For phenotypes with a substantial genetic contribution to phenotypic variability, studies to identify the genetic loci influencing the phenotypes are more likely to be eventually successful. However, high heritability does not ensure large individual gene effects. Thus, the power to detect a QTL contributing to a multifactorial phenotype is directly proportional to the magnitude of the specific effects of the QTL. For example, regardless of the average heritability of the phenotype, there is greater power to detect a QTL accounting for 50% of the phenotype variance as compared with a QTL explaining only 5% of the overall variation of the phenotype. There are several experimental approaches that can be employed to identify genetic loci contributing to the risk for osteoporosis. One of the most commonly employed experimental designs is that of candidate gene analysis, which seeks to test the association between a particular genetic variant (i.e., allele) and a specific phenotype. Many of these candidate gene studies use population-based association methods. As applied to the study of osteoporosis, two samples are collected: a group of osteoporotic patients and a control group of nonosteoporotic subjects. The allele frequencies at a polymorphism within or near the candidate gene are then compared in the two groups. Ideally, the two groups should be matched so that they differ only in their disease status. In theory, evidence of differences in allele frequencies within the two populations (association) may be the result of linkage disequilibrium with the candidate gene or possibly with another gene in close proximity; however, in practice, the candidate gene is thought likely to be causative for disease.

Because of its simplicity, the population-based association study of candidate genes has been widely used and perhaps abused. There are a number of major problems with this approach. The first is the choice of candidate gene. A large number and variety of proteins are involved in skeletal biology, and their genes are all potential candidates. Thus, the number of candidates is large (124) and multiplies daily with the application of new technologies to bone such as gene expression microarray analysis. Second, analysis of each candidate in isolation of the others is difficult to interpret statistically (10, 125, 126). Third, association studies often use simple polymorphisms in introns with doubtful biological effect. Fourth, studies in multiple populations are required. Fifth, there is no chance of finding genes outside those hypothesized. Sixth, spurious associations are common due to racial admixture and to the failure to use polymorphisms in genes known not to be involved in bone biology as a negative control (127, 128). Finally, disequilibrium appears to exist in blocks separated by regions of excess recombination (129, 130, 131), suggesting that complete characterization of the boundaries of linkage disequilibrium is essential for the accurate interpretation of results that could lead to erroneous interpretation of apparent linkage disequilibrium with candidate genes. About 20 candidate genes have been shown to be associated with BMD (Table 3). None, however, have been replicated over all populations. Unfortunately, meta-analysis (132) does not overcome this lack of consistency because it fails to take into account the many unpublished negative studies and also results in the problem of racial admixture. Another common disturbing feature of the candidate gene approach is that in studies that fail to detect the original association such as with BMD, other associations, some quite unrelated to BMD, emerge and when the polymorphism is examined in other disorders, new associations emerge. The vitamin D receptor (VDR), a candidate gene that has been extensively studied in relation to BMD, has been reported to be associated with phenotypes as diverse as body size (133), height (134), infant growth (135), age at menarche (136), muscle strength (137), calcium absorption (138), calcium intake (139), urinary calcium excretion (140), blood lead levels (141), blood pressure (142), primary hyperparathyroidism (143), type 1 diabetes mellitus (144), severity of hyperparathyroidism in renal failure (145), calcium rickets (146), calcified aortic valve stenosis (147), multiple sclerosis (148), thyrotoxicosis (149), intervertebral disc degeneration (150), osteoarthritis (151), biliary cirrhosis (152), breast cancer (153), rheumatoid arthritis (154), benign prostatic hyperplasia and prostatic cancer (155), and tuberculosis (156).

₂HS-glycoprotein is present in bone matrix and was the first candidate gene shown to be associated with BMD (157, 158). It is preferentially concentrated in bone matrix (159) and functions as an immunoregulator (160).

The VDR is largely responsible for the broad range of actions of 1,25-(OH)₂ vitamin D including its effect on calcium transport and homeostasis and bone resorption (161). Mutations in the gene result in vitamin D-resistant rickets (162). Polymorphisms in the introns of VDR were initially said to account for about 80% of the variability in bone mass in twins (163). Subsequent studies, however, were unable to confirm linkage in twin samples (76, 164), and the original observation of linkage was retracted because of genotyping errors (165). Despite these negative linkage studies, VDR polymorphisms have been extensively studied in association studies. More than 50 association studies have been published, none of which show a consistent association with BMD and many showing associations with various nonskeletal phenotypes. In two large sib pair studies, no evidence of linkage with the VDR locus at 12q12–14 was found (87, 166).

The estrogen receptors (ERs) are responsible, in large part, for the broad range of actions of estrogenic steroids on target tissues including those in skeletal tissues. There are two ER genes, ER1 and ER2. ER (167) and ERß (168) have a wide tissue distribution. Estrogen is essential for closure of the bone epiphyses in adolescent girls (169) and for maintaining bone mass in women (170). In men, mutations in the ER result in prolonged skeletal growth and osteoporosis (7). Polymorphisms in the ER1 (171, 172, 173, 174) have been most extensively studied, but more recently ER2 (175) has also been examined. There is no consistent association between ER1 or ER2 polymorphisms and phenotypes of bone mass across studies. In a large sib pair study, no evidence of linkage with the ER1 locus and ER2 locus was evident (87).

Collagen type 1 I (COL1A1) along with collagen type 1 2 (COL1A2) make up bone collagen, which is the main structural protein in the skeleton. Mutations in COL1A1 and COL1A2 cause a dominant monogenetic osteoporotic disease, osteogenesis imperfecta (4). Polymorphisms outside the coding region but in an SP1 binding site in COL1A1 were reported to account for part of the variance in BMD in the normal population and to be associated with fragility fracture (176, 177). Subsequent studies have not replicated an association between COL1A1 polymorphisms with phenotypes of bone mass (178, 179). No association was found with polymorphisms in COL1A2 (179). In a large sib pair study, no evidence of linkage of BMD with the COL1A1 locus was evident (87).

At least another 16 candidate genes have been examined for association with BMD (Tables 3 and 4). They have been selected because their products are known either to be involved in some aspect of the metabolism or structure of bone. In addition, association studies have also been performed with bone phenotypes whose heritability has not been clearly established such as bone loss and vertebral fracture. "Noncandidate" genes have also been studied including HLA DRB1 (180) and apolipoprotein E (181, 182, 183) because of their known associations with other diseases.

C. Family-based association approach: the transmission disequilibrium test
To avoid the pitfalls of population-based association studies, a family-based association test, the transmission disequilibrium test or TDT (126, 187, 188) was developed. The primary advantage of the TDT is that it avoids the necessity of collecting a matched control sample. As originally proposed, the TDT analyzes a nuclear trio consisting of an affected individual and his/her parents (Fig. 5). These three individuals are genotyped at the polymorphism in or near the candidate gene. The alleles transmitted by the genotyped parents to the affected offspring are the "affected" sample, and the alleles not transmitted from these two parents are then used as "control" alleles. Through the use of such a within-family design, the control sample of alleles is perfectly matched to the affected sample of alleles, because they are transmitted from the same two parents. Thus, spurious association results due to population stratification are avoided. When the TDT is performed with one affected offspring from each family, it is a valid test of linkage and association (linkage disequilibrium), because affected individuals are unrelated and provide independent meiotic information toward the test of association. Application of the TDT with multiple affected siblings remains a test of linkage, but due to the lack of independent meiotic data from siblings, it is no longer a valid test of association (126).

View larger version (10K): [in this window] [in a new window]   Figure 5. Schema of the TDT. A, TDT trio: The individual (II: 1) is affected with osteoporosis. His father (I: 1) is heterozygous at the marker and transmits allele 2 but not allele 1 to his affected son. This result provides the data for the TDT and would be tabulated in a table such as that shown in panel B. The mother (I: 2) is homozygous and can only transmit to her affected son allele 1 and, therefore, does not provide information toward the TDT. B, Hypothetic data from 100 trios. When considering informative transmissions from a parent heterozygous at a marker (genotype 1/2), allele 2 was preferentially transmitted to the affected offspring in 90 of 100 meioses. If no association exists, the expectation is that the two alleles will be equally transmitted. If association exists, there would be an excess of one of the two alleles transmitted to affected offspring.

Further extensions of the TDT methodology have been developed to enable the investigator to utilize family-based disequilibrium methods to analyze quantitative phenotypes (193). A series of quantitative TDT methods have been proposed depending on the type of ascertainment employed in the collection of the proband. Subsequently, additional modifications of the quantitative TDT have been developed providing greater flexibility if parental DNA is not available (194, 195) or if data are available from multiple generations (196).

The testing of candidate genes using the TDT or other association methodologies has not proven, to date, to provide consistent results across populations. However, as demonstrated in Crohn’s disease (129), a multifactorial disorder, the application of the TDT approach in chromosomal regions previously identified to be in linkage to the disease phenotype can be a powerful means to narrow, and potentially identify, disease susceptibility loci.

D. Linkage approach
Few genes influencing complex traits have been identified by the study of candidate genes alone. Therefore, researchers in the field of osteoporosis have used other experimental designs to identify genes contributing to the risk of osteoporosis. To improve the likelihood that a gene influencing osteoporosis might be identified, investigators search the genome, testing polymorphic markers evenly spaced on all chromosomes. A strength of the genome-wide approach is that it may allow susceptibility genes to be identified that are not candidates based on the current understanding of the pathophysiology of osteoporosis.

Identification of the genes contributing to polygenic traits can be extremely complex, even for a phenotype such as BMD with substantial heritability. Therefore, several types of genetic studies have been employed to dissect the genetic contribution to BMD. One technique has been the identification of families with extreme BMD phenotypes. The rationale for such studies is that genes with a substantial contribution to BMD are more likely to be segregating in families with extreme BMD measurements. This strategy has been employed to identify families with either abnormally high or low BMD. An advantage of this approach is that statistical tests of linkage can be employed that model the genetic contribution to BMD as a single gene effect. Such studies typically employ parametric linkage analyses [i.e., computer package: FASTLINK (197); VITESEE (198)], the most powerful study design for identification of genetic loci contributing to extreme BMD phenotypes. Unfortunately, there are several limitations to this particular experimental design. First, and perhaps most importantly, the genes found to contribute to the extreme BMD phenotypes observed in these unusual pedigrees may not contribute substantially to the normal variation in BMD phenotype observed in the general population. A second limitation of the identification of extreme pedigrees is their rarity in the population, which makes the identification of such families very expensive. A third limitation is the likely faulty assumption that families with low bone mass, but within the normal range, are segregating as a single-gene disorder, whereas the phenotype is due to more than one gene as would occur in a multifactorial disorder.

An alternative to the identification of pedigrees with extreme phenotype is the ascertainment of families with members having BMD within the normal range. In such pedigrees, BMD is inherited in a complex, non-Mendelian fashion, with multiple genes and environmental factors contributing to the phenotype. As a result, a particular model for BMD inheritance may be difficult to specify. In addition, the time and effort required to correctly specify more complex penetrance-based linkage models may outweigh the slight advantage those approaches may have over some penetrance-free linkage approaches (199). Such model-free nonparametric linkage analyses typically involve studying a large number of related subjects thought to be segregating for genes that influence BMD.

All statistical tests of linkage using nonparametric methods are based on "identity by descent" (IBD) marker allele sharing (Fig. 6). Alleles are IBD if siblings inherit the same marker allele from the same parent. If the marker being tested is in close physical proximity to a gene influencing the phenotype, then siblings with similar phenotypic values would be expected to share marker alleles IBD. Conversely, siblings with dissimilar phenotypes would be expected to share fewer marker alleles IBD near the gene influencing the phenotype. An advantage of quantitative linkage methods as employed here is that no arbitrary cutoff for "high" or "low" phenotypic values is necessary; therefore, all sibling pairs are included in the analysis [i.e., computer package: Mapmaker/Sibs (200)].

View larger version (15K): [in this window] [in a new window]   Figure 6. Schema of IBD. In each nuclear family, the genotype for a marker with four alleles is shown. In the left panel, the two siblings have both inherited allele 1. However, the brother inherited this allele 1 from his mother while his sister inherited this allele 1 from her father. Therefore, they have no alleles IBD (IBD = 0). In the middle panel, both siblings inherited the 1 allele from the father (IBD = 1). In the right panel, both siblings inherited the 2 allele from their father and the 3 allele from their mother (IBD = 2).

Linkage analyses for complex diseases are commonly performed using affected sibling pairs or other types of affected relative pairs. In the case of osteoporosis, these might be relatives diagnosed with osteoporosis. Rather than employing a quantitative phenotype such as BMD, these studies classify individuals as "affected" or "unaffected." Tests of linkage are still based on the sharing of marker alleles identical by descent; however, in this type of analysis, because both relatives are affected (i.e., not discordant), linkage tests only for increased sharing of marker alleles. The use of a dichotomous rather than quantitative phenotype is a less powerful approach for gene mapping.

E. Family studies
A family with high bone mass has been reported in which high BMD segregates as an autosomal dominant phenotype (202). The proband was identified on radiographs taken after a car accident. Affected individuals have spine BMD greater than 3 SD above the mean and are perfectly healthy (203) with no evidence of the sequelae of a sclerosing bone dysplasia (204). Using a genome screen approach, linkage to chromosome (Chr) 11q12-l3 was identified, and further fine mapping and sequencing identified the responsible gene as low-density lipoprotein receptor-related protein 5 (LRP5) (205) This is the same gene responsible for the Mendelian disorder autosomal recessive osteoporosis-pseudoglioma syndrome (9). The high-bone mass syndrome results from a mutation causing gain of function whereas the osteoporosis-pseudoglioma syndrome results from a mutation causing loss of function. There is a dosage effect for Lrp5 function because heterozygous carriers of osteoporosis-pseudoglioma syndrome have reduced bone mass (9). Lrp5 is involved in Wnt signal transduction (206, 207) and, as such, Lrp5 represents a new regulatory pathway in osteoblast function and bone mass regulation.

Low-bone mass families ascertained through a proband with low BMD (Z score < -2.0 or radiographic evidence of osteoporotic fracture) consisted of seven families comprising 149 members who had parametric linkage analysis performed to evaluate three candidate genes: COL1A1, COL1A2, and the VDR. Linkage to all three candidate genes was excluded with LOD scores below -2.0 (208). Subsequently, parametric and nonparametric linkage analysis was performed in these seven families using data from a genome screen (209). The maximum parametric LOD score was obtained on Chr 11q. Nonparametric linkage analysis using 74 independent sibling pairs derived from these same families supported linkage to Chr lp36 (LOD = 2.29), Chr 2p23–24 (LOD = 2.25), and Chr 4q (LOD = 2.28).

An independent sample of eight families has been ascertained through a proband under the age of 35 yr with a history of two or more crush fractures and a spinal BMD at least 2.5 SD below the mean for age and sex, i.e., severe early-onset osteoporosis (210). Segregation analyses performed in this sample suggest a major gene of codominant inheritance for spinal BMD. Linkage studies have not been reported in these families. Whether the gene in these rare families will be relevant to the common form of osteoporosis or BMD in the general population is uncertain. In contrast, segregation analyses performed in healthy nuclear families rejected the hypothesis of a single major gene and, instead, supported a polygenic model underlying BMD (86).

F. Genome scans in sibling pairs
BMD.
A study in 374 healthy premenopausal white and black sister pairs reported linkage of femoral neck BMD to Chr 11q12–13 (211). This region harbors the LRP5 gene responsible for autosomal high bone mass trait (205) and the autosomal recessive osteoporosis-pseudoganglioma syndrome (9) and the TCIRG1 subunit of the vacuolar proton pump responsible for a subset of autosomal recessive osteopetrosis (212), suggesting that the same locus may also regulate BMD in the normal population. However, a subsequent analysis of this region in 464 white and 131 black sister pairs weakened the evidence of linkage (87), and analysis in a sample ascertained through a low bone mass proband did not support linkage to 11q12–13 (213).

A 10-cM autosomal genome screen identified six possible QTL (LOD >1.85) in 429 white premenopausal sister pairs (Table 5 and Ref. 87). The linkages on Chr 1, 5, 6, and 22 were at or near a marker locus and were reexamined in an expanded sample of 595 sister pairs. The initial genome screen in 429 sibling pairs had a LOD score of 3.11 at Chr 1q, which increased to 3.86 when the 595 sister pairs were included. This is not the same region of linkage reported on Chr 1 in a genome screen employing pedigrees ascertained on the basis of an osteoporotic proband (209). With the addition of sibling pairs, linkage to Chr 5p also increased from 1.9 to 2.2, linkage to Chr 6p was not substantially increased, and linkage to Chr 22 decreased. These results provide substantial evidence that genetic loci influencing BMD can be detected.

Structure.
In the sister pair sample used to detect linkage of BMD to Chr 1, 5, 6, and 11 reported for linkage to BMD, seven QTL were found for various measures of structure at the proximal femur (Table 6 and Ref. 98). Two chromosomal regions were identified with significant LOD scores (>3.6) for at least one femoral structure phenotype. The maximum LOD score of 4.3 was obtained for femoral neck axis length on Chr 5q. Evidence of linkage to Chr 4q was found with both femoral neck axis length (LOD = 3.9) and midshaft width (LOD = 3.5). Significant linkage was found to Chr 17q with femoral head width (LOD = 3.6). Chromosome 3q showed linkage with pelvic axis length (LOD = 3.1), midshaft femoral width (LOD = 2.8), and femoral head width (LOD = 2.3). Chromosome 19p showed linkage with femoral neck axis length (LOD = 2.8) and femoral head width (LOD = 2.8).

	V. Identification of the Susceptibility Genes

Top Abstract I. Introduction II. Osteoporosis III. Heritability IV. Locating the Susceptibility... V. Identification of the... VI. Animal Studies VII. Bioethics References

Normal genetic variation in complex traits, such as peak BMD, is generally not due to deleterious mutations but to common polymorphisms resulting in more subtle changes in gene function or expression. The large number of genes and the intensity with which each gene must be examined for sequence variation mandates that a logical strategy for ranking candidates is pursued rather than examining in sequence every gene that lies within the region of interest. However, there are pitfalls in ranking candidate genes. First, ranking genes is based largely on current models of the pathophysiology of osteoporosis, which are incomplete. Second, rankings are based on knowledge of the function of the genes, which is also incomplete. An example of the former is the PHEX gene, which is a member of the neutral endopeptidase family and is responsible for X-linked hypophosphatemic rickets (216). Before demonstrating that PHEX mutations were responsible for XLH, investigators had never considered that an enzyme defect could be responsible for the disease. Therefore, the goal of ranking genes should be to analyze genes in a systematic fashion from the most likely to least likely, rather than exclude genes based on current notions of pathophysiology. Indeed, one of the strengths of positional cloning studies is the potential to dramatically alter the field by identifying genes that were not previously known to be involved in the pathophysiology of osteoporosis. Thus, it is reasonable to initially study genes that are expressed in bone and genes that by virtue of homology have a high likelihood of being involved in the pathophysiology of osteoporosis. However, subsequent studies may need to examine candidate genes that are not obviously related to the pathophysiology of osteoporosis.

Once the candidate genes in a fine mapped region are identified and ranked, the next task is to identify polymorphisms in these genes. This can be done by searching the SNP databases, such as the NCBI SNP database (http://www.ncbi.nlm.nih.gov/SNP) and the HGBASE (Human Genic Bi-Allelic Sequences, http://hgbase.interactiva.de/) for known polymorphisms, focusing on polymorphisms that are likely to have functional significance such as those that result in amino acid changes. Although these databases currently have limitations, they are expanding rapidly and are already very useful. Finally, once the polymorphisms are identified, DNA from the subjects can be genotyped using a variety of different methods and the results analyzed (217, 218).

	VI. Animal Studies

Top Abstract I. Introduction II. Osteoporosis III. Heritability IV. Locating the Susceptibility... V. Identification of the... VI. Animal Studies VII. Bioethics References

 
In searching for the susceptibility genes for osteoporosis, complementary studies in animals are essential. Not only do they allow breeding strategies that cannot be performed in humans, but they also provide important bone strength phenotypes that cannot be measured in vivo in humans. Two animal models, the mouse (219) and the baboon (220), have been used for identifying genes underlying bone strength. More recently, the rat has been used (221).

The most intensively studied animal model is the mouse. It is ideally suited for genetic analysis because of its short generation time and its ability to produce large litters in the laboratory (222). Its contribution to the genetics of osteoporosis and skeletal biology is already substantial. A variety of inbred mouse strains have been used in genetic studies. A mouse strain is considered inbred when virtually every genetic locus in its genome is homozygous. Typically, this has been produced from 20 or more consecutive generations of brother-sister mating. As a result, all animals within the inbred mouse strain are genetically identical. This situation is analogous to twin studies in humans. Also, founder effects in genetically isolated populations can be amenable to similar approaches to those employed in mouse studies, again emphasizing the similarity between human and mouse genetics studies.

Many of the genetic mapping studies in mice designed to identify chromosomal regions contributing to osteoporosis or BMD were initially performed in recombinant inbred (RI) strains. RIs are created from an F₂ (second-generation offspring) sample by completing multiple generations of brother-sister mating. As a result, each RI strain is not only inbred but also unique in its genetic composition from each of the inbred founders. The power of the RI methodology to identify genes underlying phenotypic variability lies in the vast amount of genotyping already completed in the various RI lines. However, the limited number of available RI lines compromises the power of these lines to localize and identify genetic loci. As a result, whereas RI studies can detect regions of possible linkage, most researchers have pursued additional confirmation studies in backcross or F₂ progeny derived from inbred animal lines.

The most powerful strategy for mapping QTL involves the intercross of two strains discordant for the relevant phenotype of interest. Presumably, these mouse strains are discordant because they have fixed differing alleles at loci relevant to the phenotype. The discordant inbred progenitor strains are mated to produce F₁ hybrid mice. These mice are likely to be obligate heterozygotes at loci contributing to the phenotype. The F₁ mice are then intercrossed (brother-sister mated) to produce an F₂ population. In the F₂ population, the alleles at the loci contributing to the phenotype (the QTL) are segregating, meaning that each F₂ has different combinations of the alleles at the loci contributing to the phenotype. This can be observed in the wide variation in the phenotype which is observed in the F₂ sample with the extreme of the phenotype distribution often exceeding that observed in the progenitor lines. Therefore, the F₂ sample is considered to be segregating for the relevant QTL and is an ideal sample in which to perform QTL mapping. The intercross strategy, as well as the less powerful recombinant inbred strategy, has been used to create extensive data regarding the likely position of QTL contributing to BMD phenotypes (Table 7). Linkage analysis and subsequent fine mapping can provide corroborating information on important QTL syntenic with the human. The effects of individual gene products on skeletal biology can be evaluated using knockout and transgenic technology (Table 8). These techniques can also be used to identify candidate genes for human studies. A number of transgenic or knockout mice have clear skeletal phenotypes. This is an active area of research, and the list will undoubtedly grow in the future.

The first QTL for skeletal BMD in mice were reported (223) using the recombinant inbred approach for the BXD lines (C57BL/6J and DBA/2J cross). Subsequently, femoral BMD QTL were reported using the intercross approach (C57BL/6J and CAST/EiJ cross) (224), as were QTL for femoral cortical thickness index (225) and QTL for spinal BMD in senescence accelerated mice (SAM) using an intercross strategy (SAMP6 and SAMP2 cross) (226). Femoral BMD QTL for both the C57BL/6J and CAST/EiJ cross (225) and the C57BL/6J and C3H/HeJ cross have been reported (227). One QTL was found to be at the same location for both crosses (Chr 1, see Table 7). Interestingly, the two crosses produced several different QTL even though the BMD phenotype is identical. The major QTL for femoral density from the C57BL/6J and CAST/EiJ cross are on Chr 1, 3, 5, 13, and 15 and from the C57BL/6J and C3H/HeJ cross are on Chr 1, 2, 4, 6, 11, 12, 13, 14, 16, and 18. These differences may be due to each strain having fixed differing alleles at the relevant loci or the two progenitor strains in one cross fixing the same QTL allele, resulting in the F₂ sample not segregating for this QTL. This behavior emphasizes the importance of collecting data from several mouse crosses to assure that all QTL contributing to a phenotype are uncovered. As illustrated in Table 7, there are BMD QTL on mouse Chr 1, 2, 13, and 16 that have been uncovered in at least two different linkage studies.

Spinal BMD QTL for the cross of C57BL/6J and C3H/HeJ have also been mapped (228). Interestingly, not all spinal BMD QTL corresponded to femoral BMD QTL. Important spinal BMD QTL on Chr 7 and 9 have no femoral counterparts, suggesting that genetic regulation of BMD is, in part, dependent upon anatomical site. QTL for several vertebral microstructure phenotypes from the C57BL/6J and C3H/HeJ cross are at loci that do not correspond to femoral BMD QTL (227).

One of the clear advantages of using rodents for genetic studies is the availability of the bones for direct measurement of bone biomechanical properties including strength and fragility. Fundamental biomechanical properties include force to failure (a measure of strength), stiffness, and work to failure (a measure of overall fragility). Biomechanical properties can be assessed at several sites including femoral midshaft, femoral neck, and vertebra (229). Preliminary studies of femoral strength in the C57BL/6J and C3H/HeJ cross have identified several important QTL, some of which overlap femoral BMD QTL (Chr 1, 4, 6, and 18) and others which do not (Chr 8, 11, and 13) (230). As with BMD, it is likely that genetic regulation of bone strength is site specific and that femoral and vertebral strength segregate somewhat independently (231).

BMD and skeletal biomechanical properties are polygenic traits. Consequently, there may be numerous interactions among genes contributing to these traits. To isolate gene effects, congenic strains in which a single QTL is moved from the donor strain to a recipient strain can be constructed. This is typically done with selective backcrossing for 6–10 generations. Congenic strains have been created by moving QTL for high bone mass donated by C3H/HeJ onto the low bone mass C57BL/6J line. C3H QTL caused significant differences in B6 BMD and femoral strength. QTL from mouse Chr 1, 4, and 18 increased BMD in recipient mice, whereas the donated QTL at Chr 6 reduced BMD (232, 233) indicating that genetic influences on bone structure can be isolated in congenic mice.

A second animal model that is available is the baboon (220). Colonies represent very large pedigrees that can be used for linkage analysis using many of the markers that are present in the human genome. Importantly, the size and shape of the baboon skeleton at the hip and vertebra approach those of the human much more closely than those of the mouse. Linkage studies performed in a large baboon colony have identified a QTL on baboon Chr 11 that influences BMD (234). Interestingly, this is the same region of Chr 11 identified in the three Mendelian bone-related disorders (9, 205, 212) as well as a large sample of premenopausal sister pairs (211).

More recently, the rat has been developed as a model for studying susceptibility genes of osteoporosis. The advantage of the rat is the extensive information on its physiology and skeletal biology that is available. Variability in femoral, vertebral, and femoral neck fragility among 11 inbred strains of rats has been reported (221). Fischer 344 and Lewis strains show the greatest variance in the vertebral fragility phenotype, and the Copenhagen 2331 and DA strains show the greatest variance in the femoral neck fragility phenotype, indicating that these strains will be useful for QTL analysis using the intercross strategy. As with mice, variation in skeletal fragility phenotypes in rats is dependent upon the anatomical site studied. Therefore, it