Genetics of Fracture: Challenges and Opportunities
Article first published online: 1 JUL 2000
Copyright © 2000 ASBMR
Journal of Bone and Mineral Research
Volume 15, Issue 7, pages 1253–1256, July 2000
How to Cite
Nguyen, T. V. and Eisman, J. A. (2000), Genetics of Fracture: Challenges and Opportunities. J Bone Miner Res, 15: 1253–1256. doi: 10.1359/jbmr.2000.15.7.1253
- Issue published online: 2 DEC 2009
- Article first published online: 1 JUL 2000
Over the past 30 years or so epidemiological and genetic studies have consistently shown that bone mineral density (BMD), a measure of bone strength, is a primary determinant of fracture risk in the general population (1–4); that a large component of the variance of BMD (up to 80%) is determined by genetic factors(5–9); and that a family history of fracture confers significant increase in risk of fracture for a relative.(10) In this issue, Deng et al.,(11) in an elegant study, report that approximately 25% of the liability to Colles' fracture is attributable to genetic factors. How should we interpret this finding in relation to what we already know about the BMD-fracture relationship and the genetics of BMD? Is the observed influence of genetic factors on Colles' fracture mediated through or independent from the genetic influence on BMD? What are the implications of this finding to the search for genes currently underway in many research groups around the world?
Given the extent of genetic contribution to the variation in BMD and that BMD predicts fracture risk, it is possible to estimate the contribution of BMD-related genes to fracture in the form of familial relative risk (RR) of fracture for a given genetic relationship in BMD.(12) Consider a familial correlation in BMD of 0.5 (e.g., between mothers and daughters) and subjects aged 60+ years with a relative fracture risk (RR for each SD decrease in BMD) of 2.4 for hip fracture and 1.8 for upper limb fracture (Nguyen et al, unpublished data). It can be estimated that the familial risk of hip and upper limb fractures is 2 and 1.1, respectively. In the Deng et al. study, the familial RR of Colles' fracture for a woman with an affected mother and an affected sister was 1.3 and 1.9, respectively. Daughters of mothers with hip fracture have lower BMD at the hip and daughters of mothers with spine fractures lower BMD at the spine.(13–14) A similar relationship can be expected at the proximal forearm and humerus sites, because femoral neck BMD predicts fracture risk at these sites.(1) Thus, it could be argued that the observed familial risk of Colles' fracture results from the the genetics correlation in BMD.
However, there is evidence that the influence of family history of fracture is partially independent of BMD. For example, even after adjusting for BMD, the risk of hip fracture remains increased in those with a history of hip fracture in their mother (RR 1.5), sister (RR 1.8), or brother (RR 2.3). Similarly, the risk of wrist fracture is increased by maternal (RR 1.5) and paternal (RR 2.4) history of wrist fracture.(10) Thus, it seems there is a genetic component in fracture risk independent of those that determine BMD.
Factors other than low BMD also contribute to the risk of fracture. Among them, quadriceps weakness and postural instability,(1) aspects of bone quality as assessed by quantitative ultrasound (QUS) measurements(15) have been shown to be predictors of fracture risk. Muscle strength and QUS measurements are modulated largely by genetic factors,(16–17) so that familial aggregation of these parameters also could contribute to risk for fractures. Therefore, the significance of BMD-adjusted family history fracture risk may reflect the genetics of QUS measurements, muscle strength, and other clinical risk factors for osteoporotic fractures. However, BMD, QUS measurements, and muscle mass share some common genetic influences(9, 16, 18); thus, the familial influences of QUS measurements, BMD, and muscle mass may not be independent. This implies that one set of common genes contributes to variation in intermediate phenotypes such as bone mass, quality, size, and density and hence to the liability to fracture. Other genes may specifically contribute to each of these bone and nonbone factors (e.g., quadriceps strength and falls), which also contribute to the liability to fracture.
Apart from individual genes contributing to fracture risk, some bone- and nonbone-related genes may interact such that some alleles only have “deleterious” effects in the presence of specific alleles found at other loci. Apart from the complexity of gene-gene interactions, genes also may interact with nongenetic factors to convert vulnerability into actual fracture outcomes. Nongenetic factors often are equated with “environmental” factors, but there are other types of nongenetic factors that act during development, including epigenetic factors, such as stochastic or chance factors that act during bone development. Indeed, even in monozygotic (MZ) twins, who share 100% of their DNA, a difference in phenotypes such as fingerprints is still observed, likely because of such stochastic factors. Thus, fracture does not necessarily arise from a simple biological “defect” but rather represents an event along a continuum of phenotypes.
It must, therefore, be concluded that the liability to fracture is genetically complex. The exact nature of the genetic component in fracture may be a combination of polygenic effects and gene-gene and genetic-environmental interactions. Under this supposition, no single genetic locus will by itself cause the event or even, for that matter, determine the major share of risk of fracture or any osteoporosis phenotype. It is likely that multiple alleles at multiple loci within the genome interact to produce vulnerability to fracture. Also, the genes may act at different times in the development of bone and perhaps even in different bone envelopes.
From the functional point of view, there are two kinds of genes, namely, “osteoporosis” genes that actually determine an abnormality in bone metabolism and hence contribute to the risk of the fracture event per se and “innocent bystander” genes that are completely unrelated to these phenotypes but lie so close to and/or in linkage disequilibrium with the osteoporosis genes that they can be used to track them. There are several good reasons to search for both kinds of genes. In particular the definite causes of osteoporosis are not completely known and known environmental factors such as lifestyle and dietary habit account for only a modest proportion of variance of the liability to fracture. Thus, genetics may prove to be the single most powerful tool for gaining mechanistic insights into the pathophysiology of osteoporosis, as has been the case for genes in familial aggregations of breast cancer. Once an osteoporosis gene has been identified, researchers can elucidate its functions, study its interactions with developmental and environmental factors, examine its ability to predict fracture in an individual, and use this understanding to devise novel treatments. For example, most drugs work by binding to a specific protein, altering its activity to achieve a therapeutic outcome; thus, identification of a gene and hence its protein product can lead to new targets for new therapeutic agents. In the future, such knowledge of gene action may even open an opportunity for gene therapy. In terms of public health, identification of genes can lead to new diagnostics for the early detection of individuals with high risk of fracture and also help identify responder/nonresponder groups for particular therapies.
It has been estimated that there are between 60,000 and 70,000 genes located within 3 billion base pairs of DNA in the human genome.(19) However, the number of genes thought to be involved in osteoporosis (either directly or indirectly) is much smaller, probably less than 100 and only some of these will have major effects. Finding these genes in such a large pool is undoubtedly a difficult task, and one of the major challenges at present is how to design powerful genetic studies to search for these genes. Some relevant and important questions in study design include the following: What are appropriate sampling plans? What analytical strategies maximize the statistical power to detect genes? What phenotypes should be considered?
Currently, there are essentially two strategies for searching genes that influence a complex trait such as osteoporosis: candidate genes and genome screening. In the candidate-gene approach, individual genes can be tested directly for their association with or linkage to a phenotype. In a genome search approach, all genes are systematically screened using panels of microsatellite DNA markers “uniformly” distributed throughout the entire genome. The genomewide search approach can only identify region(s) where genes are located, not the genes themselves. Any region identified by genome scanning typically spans 5–10 centimorgans (cM) and localization has to be progressively refined until a gene actually can be identified. Actually, there is another strategy of “positional candidates” for the search of genes. In this approach, any identified region of several centimorgans is searched intensively for candidate genes using established databases and those coming from the Human Genome Project.
In each of the candidate gene and genome search approaches, genes can be identified based on a demonstration of a significant association and/or linkage relationship. Association analysis compares the frequency of alleles between affected (e.g., fracture or low BMD) and unaffected individuals; a given allele is considered to be associated with the disease if that allele occurs at a significantly higher frequency among affected individuals. On the other hand, linkage analysis tests whether the genes show correlated transmission within pedigrees.
The majority of genetic studies in osteoporosis in the past have been based on association analysis of candidate genes, and results of these studies often have been inconsistent and sometimes difficult to interpret. A major disadvantage of the candidate gene approach is that if multiple genes are involved (as is expected in the case of osteoporosis), analysis of each candidate gene in isolation of the others may amount to testing every gene on the human genome, an endeavor fraught with statistical problems relating to false-positive results. It is, of course, possible to correct for these associations by statistical adjustment, but by definition, candidate genes have a reasonable prior probability of being involved in disease susceptibility; therefore, correction for the number of genes tested should be conservative or it could be counterproductive. It is important to recognize that a significant association between a marker and a trait could arise from three major possibilities: the “marker” locus is the disease locus; the marker locus has no direct effect on fracture but it is in linkage disequilibrium with the disease locus; and there is an artifactual association caused by population admixture or other population phenomenon. Population admixture can occur, for example, in a case-control study conducted in a mixture of two subpopulations, one of which has both a higher fracture prevalence and a higher marker frequency. Moreover, even with close linkage, low frequency and modest effect of a locus may lead to lack of association, that is, the absence of an association does not exclude linkage.
To minimize those artifacts in population-based association studies, family based association studies on related individuals have been proposed, such as the transmission linkage disequilibrium test (TDT).(20) The sampling unit in these methods consists of two parents with an affected child; parental alleles not transmitted to affected children are used as controls. Thus, the TDT considers affected children of heterozygous parents at a marker locus and simply tests whether these children have received this locus with a probability different from 0.5, the value expected under random segregation. The TDT test can be powerful. However, in the more common situation in which the marker locus is different from the disease locus, the power is highly dependent on the disease-marker allele frequencies and the strength of their linkage disequilibrium.
Linkage studies are aimed at tracing cosegregration and recombination phenomena between observed marker alleles and unobserved putative-trait influencing alleles among members of families or pedigrees. Although the association analysis of candidate genes has been productive in osteoporosis, it can only lead to identification of allelic associations for genes known to be involved in bone biology. A genomewide search using linkage analysis offers the potential to identify a range of known (and unknown) genes involved in determination of osteoporosis. The genome screening approach has several advantages, but perhaps the most important is that it requires no assumptions about possible pathophysiological mechanisms. One recent analysis of “osteopenia” in 37 members of 7 families suggested monogenetic inheritance on the basis of a possibly bimodal inheritance.(21) This group analyzed 143 members and concluded that loci other than those associated with the collagen Iα1 and Iα2 and vitamin D receptor (VDR) genes are responsible for the low bone density observed in affectedmembers. By contrast, Johnson and colleagues have mapped an apparently autosomal dominant gene associated with very high bone mass in just 28 members of an extended family pedigree.(22) This mapping was facilitated by the extreme degree of difference in bone density but it is not clear to what extent this infrequent trait contributes to bone mass or density in the overall population. These limitations notwithstanding, the successes of these studies support the power of linkage studies in extended family pedigrees over association studies.
One of the key issues in any study aiming at identify genes is the problem of power. If the power of a study is low, there is a good chance that the study's results will be inconclusive. Among strategies for increasing power, such as the selection of sampling unit, other strategies relating to data analysis also are worth consideration. Past linkage studies largely have been based on the sib-pair design, in which a set of two related individuals is analyzed. However, it is intuitive that larger sibships can provide more information than two-individual sib pairs. For a sibship with m siblings, it is possible to evaluate m (m − 1)/2 sib pairs. However, the fact that only m − 1 is independent can lead to a falsely inflated estimate of significance level (type I error). Nevertheless, with recent advances in statistical genetics, particularly the variance component approach,(23) this problem (of nonindependence) can be resolved by maximizing the likelihood of a sibship jointly on all members in the sibship. In fact, with the variance component method, it is possible to increase the power of a study by including all individuals in any extended pedigree in the analysis.(24) Todorov et al. (1997)(25) have further shown that large sibships can be a cost-effective alternative to the use of sibling pairs.
The ultimate and clinically relevant consequence of osteoporosis is fracture. Thus, it could be argued that fracture is the most relevant trait for a study of osteoporosis. For a discrete categorical trait such as fracture, Risch (1990))(26) developed the use of the ratio λ of the risk to relatives of a proband to the population prevalence of the trait. In this analysis, the power for detecting linkage essentially depends on the risk ratio λ and the recombination fraction θ; with power decreasing rapidly as recombination fraction increases. When θ = 0 and λ = 2, a sample size of only 200 sib pairs would be required to have an 80% power to detect linkage at a LOD score > 3. Considering this approach in relation to the data of Deng et al.,(11) with RR for female first-degree relatives of 1.3–1.9, the search for osteoporosis genes based on the dichotomous phenotype of fracture could be useful. However, the potential for using fracture as the phenotype for genetic studies is limited by the phenomenon of phenocopies, that is, environmentally produced phenotypes that mimic the genetically produced phenotypes. An example of this would be fracture of “nonfragile” bones caused by environment-related falls. In fact, any fracture event is the result of a number of factors including reduced BMD, bone quality, and nonbone factors such as quadriceps weakness and postural instability. Hence, it seems inherently more reliable and informative to use these traits as surrogate phenotypes (rather than the global phenotype of fracture) for genetic studies of osteoporosis.
Despite some advantages of economies of scale and model independence in studying multiple fracture-related traits in genetic studies, it would be suboptimal to examine each phenotype separately, because of the increase in type I error rate. Although procedures for correction such as the Bonferroni adjustment can be applied, the results may be too conservative, because the phenotypes are likely correlated. Thus, instead of adjusting the significance level of univariate analyses of each phenotype separately, a multivariate analysis of multiple phenotypes would be preferable. It has been estimated that linear combination of many traits into factor scores is more powerful than the use of univariate or mean phenotypic data.(27–29) Therefore, some progress in the genetic dissection of osteoporosis will come from the integrated analysis of these surrogate phenotypes alone or even perhaps in combination with fracture as a phenotype.
The Human Genome Project will provide crucial opportunities for these studies, although some difficulties will likely remain.(30) With all human genes cloned and the allelic variants identifiable, probably in the form of single nucleotide polymorphisms (SNPs), it will be possible to perform powerful and efficient family based association and linkage studies, for example, using chip technologies.(31) Thus, the ongoing Human Genome Project, the new molecular technologies, and the new statistical methods will collectively enhance the search for osteoporosis genes and hence translate the prediction of genetically complex osteoporosis into the realm of the possible.
- 11993 Prediction of osteoporotic fractures by postural instability and bone density. BMJ 307:1111–1115., , , , , ,
- 21989 Baseline measurement of bone mass predicts fracture in white women. Ann Intern Med 111:355–361., ,
- 31993 Long-term fracture prediction by bone mineral assessed at different sites. J Bone Miner Res 8:1227–1233., , , ,
- 41993 Bone density at various sites for prediction of hip fractures. Lancet 341:72–75., , , , , , , , ,
- 51987 Genetic determinants of bone mass in adults: A twin study. J Clin Invest 80:706–710., , , , ,
- 61995 Determinants of bone mass in 10 to 26 years old females: A twin study. J Bone Miner Res 10:558–567., , . , , , , , ,
- 71995 Bone mineral density determinants in elderly women: A twin study. J Bone Miner Res 10:1607–1613., , , , ,
- 81973 Genetic factors in determining bone mass. J Clin Invest 52:2800–2808., , , ,
- 91998 Bone mass, lean mass, and fat mass: Same genes or same environments? Am J Epidemiol 147:3–16., , ,
- 101998 Family history and risk of osteoporotic fracture. Study of Osteoporotic Fractures Research Group. Osteoporos Int 8:557–562., , ,
- 112000 Genetic determination of Colles fracture and differential bone mass in women with and without Colles fracture. J Bone Miner Res. 15:1243–1252., , , , , , , , ,
- 121991 Modelling the influence of risk factors on familial aggregation of disease. Biometrics 47:933–945.
- 131989 Reduced bone mass in daughters of women with osteoporosis. N Engl J Med 320:554–558., , , , , ,
- 141994 Reduced femoral neck bone density in the daughters of women with hip fractures: The roles of low peak bone density in the pathogenesis of osteoporosis. J Bone Miner Res 9:739–743., , , ,
- 151997 Broadband ultrasound attenuation predicts fractures strongly and independently of densitometry in older women. A prospective study. Study of Osteoporotic Fractures Research Group. Arch Intern Med 157:629–634., , , , , ,
- 161998 Genetic and environmental contributions to the association between quantitative ultrasound and bone mineral density measurements: A twin study. J Bone Miner Res 13:1318–1327., , , ,
- 171997 Genetic influences on muscle strength, lean body mass, and bone mineral density: A twin study. J Bone Miner Res 12:2076–2081.,
- 181996 Do genetic factors explain associations between muscle strength, lean mass, and bone density? A twin study. Am J Physiol 270:E320–E327., , , , ,
- 191994 How many genes in the human genome? Nat Genet 7:345–346., , ,
- 201998 Transmission disequilibrium, family controls and great expectation. Am J Hum Gent 63:935–941.
- 211996 Osteopenia in 37 members of seven families: Analysis based on a model of dominant inheritance. Mol Med 2:313–324., , , , , , , ,
- 221997 Linkage of a gene causing high bone mass to human chromosome 11. Am J Hum Genet 60:1326–1332., , , , ,
- 231998 Multipoint quantitative-trait linkage analysis in general pedigrees. Am J Hum Genet 62:1198–1211.,
- 241999 Comparison of variance components and sibpair-based approaches to quantitative trait linkage analysis in unselected samples. Genet Epidemiol 16:113–134.,
- 251997 Trade-off between sibship size and sampling scheme for detecting quantitative trait loci. Hum Hered 47:1–5., , ,
- 261990 Linkage strategies for genetically complex traits. II. The power of affected relative pairs. Am J Hum Genet 46:229–241.
- 271998 Multiple phenotype modeling in gene-mapping studies of quantitative traits: Power advantages. Am J Hum Genet 63:1190–201., , , , ,
- 281998 Increase in power through multivariate analyses. Behav Genet 28:357–363., ,
- 291998 A comparison of power to detect a QTL in sib-pair data using multivariate phenotypes, mean phenotypes and factor scores. Behav Genet 28:329–340.,
- 301998 It's raining SNPs, hallelujah? Nat Genet 19:216–217.
- 311999 DNA chips: Promising toys have become powerful tools. Trends Biochem Sci 24:168–173., ,