Genetic Trail of Osteoporosis: Candidate Genes Versus Genome Screens? Small Families Versus Large Families?

Defining the genetic mutations that predispose to common diseases is obviously a major challenge for the next decade or so of medical research. A prime example is osteoporosis. The disease is clearly multifactorial, but there is an increasing body of evidence to support earlier suggestions that genetic mutations predispose to osteoporosis. The paper in the current issue by Duncan et al.1 presents a careful analysis to test for linkage between alleles of candidate genes and bone mineral density (BMD). The authors assayed BMD and markers for 23 different candidate genes in 115 probands with low BMD and at least one first- or second-degree relative with normal BMD. They found "suggestive" evidence for linkage of BMD with markers in the region of the parathyroid receptor type (PTHR1) gene. They also found "some evidence" of linkage to regions containing seven other candidate genes. None of the data, however, reached the threshold for "significant linkage"2 that is generally regarded as the minimum necessary to justify extensive analyses of a gene or locus for mutations. The results may encourage further attempts to search for linkage with the same genes in larger populations. Also, the results may encourage a few brave investigators, at considerable risk of time and effort, to search for meaningful mutations in the genes. A major value of the paper, however, is the sophisticated and clear analyses of the data that bring into focus several questions about different strategies for defining the genetic causes of osteoporosis: What are the relative merits of searching for linkage with candidate genes versus carrying out genome-wide screens? What are the relative merits of using data from small family groups versus large families? As Duncan et al.1 point out, the candidate gene approach capitalizes on the large body of knowledge that has already been developed about osteoporosis and that can be used to develop specific hypotheses. The search by Duncan et al.1 for genetic linkage of BMD with candidate genes in probands and relatives is a far more powerful strategy than a search for gene associations among large numbers of unrelated individuals. The initial report3 of an association of low BMD with an allele of the vitamin D receptor in an Australian population was not confirmed by most subsequent reports in other populations.4 Duncan et al.1 found only "some evidence" of linkage to the locus for the vitamin D receptor. Also, they found only "some evidence" of linkage to the COL1A1 gene which was reported to be associated with low BMD through a rare Sp1 polymorphism in the gene.5, 6 The strategy employed by Duncan et al.1 provides an important means of verifying such gene associations among unrelated individuals, particularly if it is applied to larger numbers of probands and their relatives. At the same time, Duncan et al.1 point out several of the limitations of their strategy. One is that their analysis does not distinguish between the genetic influences of a given candidate gene and nearby genes. The problem can be overcome by using genome-wide screens as several groups are now doing,7 but genome-wide screens require large numbers of probands and families. A second problem with the strategy employed by Duncan et al.1 is that there is still an uncertainty as to statistical thresholds to be used to establish linkage with candidate genes when testing multiple loci. Also, the statistical programs for the analyses are not entirely adequate. A third limitation is that data from a large number of unrelated probands and small families must be pooled. As a result, important genetic components can be obscured by differences in genetic background and environment. A less problematic approach to any genetic analysis is to recruit large families from demographically restricted and rapidly growing populations. Data from such families are ideal for genome-wide screens, and they can be analyzed by both classical parametric analysis and nonparametric analysis. In addition, the power of recombination events between genetic markers at a specific locus can be invoked. Data on recombinants cannot be used with the currently available methods of nonparametric analyses for sibling pairs and small families. Therefore, the ability to define the size of a locus is severely limited. Without the power of recombinants, a locus cannot be defined to a region of <20–30 cM and, therefore, a region that on average can contain up to a thousand different genes.8, 9 With the identification of recombinants in large families, however, a locus can frequently be defined within <1 cM and, therefore, to a region that on average would contain 50 genes.8, 9 In addition, genetic studies have repeatedly demonstrated the great value of carrying out analyses on demographically restricted populations that are rapidly growing.10 Hence, there are continuing efforts to carry out genetic studies in populations such as French Canadians and Icelanders.10, 11 A large effort is obviously required to recruit large families with idiopathic low BMD from demographically restricted populations. There are now multiple examples from other diseases, however, that demonstrate such data are well worth the effort.8, 9 In the end, linkage data cannot provide more than statistical evidence that a chromosomal locus may contain a gene predisposing to low BMD. The data simply provide the basis for convincing investigators to make a major commitment to the essential next step: analysis of the locus that may contain 50 or more genes for mutations that change the properties of bone. Completion of the Human Genome Project will certainly facilitate the work. Also, clues as to the genes at fault may come from current efforts to define genes that influence BMD in inbred strains of mice.12 However, developing definitive data for mutations that predispose to a disease as complex as osteoporosis is still a formidable undertaking. Therefore, the genetic trail to osteoporosis is likely to be a long one. The paper by Duncan et al.1 emphasizes that for the moment it is probably wise to pursue a number of different strategies simultaneously.