BoneKEy-Osteovision | Perspective
Genomics and osteoporosis: What are the implications?
Serge L Ferrari
DOI:10.1138/20040121
IntroductionNearly 30 years have passed since the first published evidence of high heritability for bone mineral density (BMD) in twins (), and it has been nearly 10 years since vitamin D receptor (VDR)-3′UTR alleles were the first described gene variants associated with BMD in humans (). In April 2003, the Human Genome Project — the sequencing of our genome — was completed, and a nonredundant set of variations of nearly six million single nucleotide polymorphisms (SNPs) was identified therein (i.e., there is a mean spacing of one SNP every 556 base pairs). More than two million of the SNPs have been experimentally validated and 34 x 103 of them represent nonsynonymous changes that alter the amino acid sequence of proteins (see http://www.ncbi.nlm.nih.gov/SNP/snp_summary.cgi). Consequently, several “great challenges” in genomics have been defined (). For example, within a few years, the HapMap project should be able to characterize the patterns of linkage disequilibrium and haplotypes across the human genome and identify subsets of SNPs that capture most of the information about these patterns of genetic variation to enable large-scale genetic association studies. The Genome to Life project aims to identify the protein machines that carry out critical life functions, characterize the gene regulatory networks that control these machines, and develop the computational capabilities to integrate and understand these data and begin to model complex biological systems. Yet, most of the currently available information waits to be translated into clinically oriented research and furthermore into clinical practice. For this purpose, we first need to recognize what we have learned from osteoporosis genetic studies and define what are the implications for targeted research in this field.
What have we learned?Osteoporosis genetics has expanded along three axes of research. First, quantitative trait loci (QTLs) for BMD, ultrasound properties of bone (), and femur geometry () have been identified by linkage analysis in the general human population (, for review). Five of the six published genome-wide screening studies for BMD were conducted in whites (), and one of the studies was conducted in Asians (). In whites, about 20 QTLs with a LOD score above 1.8 (LOD > 3, evidence of linkage; LOD 2-3, suggestive linkage) have been mapped on most autosomes (Table 1). However, only one-third of the QTLs were identified in more than one independent population (), whereas many other human QTLs for bone mass likely remain unidentified, pertaining to both the structure (such as healthy sibs and small pedigrees) and limited sample size of most studies (). Moreover, because of the rather low density of microsatellite markers used, typically 300 to 400 across the entire genome, each thus far identified QTL spans millions (on average, 10 to 30 million) of base pairs and contains therefore hundreds of genes and thousands of allelic variants. Although most of these QTLs harbor osteoporosis candidate genes (Table 1), secondary identification of gene variants associated with bone mass within the QTLs has been very limited (). Investigators from deCode Genetics (Reykjavik, Iceland) have only recently reported the comprehensive approach that allowed them to identify a new gene variant for osteoporosis (). First, they mapped QTLs for osteoporosis in extended Icelandic osteoporotic families with a framework scan of 1000 microsatellite markers and subsequently added more markers at one locus with the highest LOD score (20p12) to narrow down the chromosomal region of interest to only 2.2 megabases. The investigators subsequently performed expression analysis of all genes in this region to identify those expressed in bone. Eventually, they used a very dense set of polymorphic markers (both microsatellites and SNPs) in and around those genes and thereby identified the gene coding for bone morphogenetic protein 2 (BMP-2) and a specific missense substitution therein, which were associated with a very significantly increased risk of osteoporosis and fractures. Identification of gene variants for bone density in the general population may also be prompted by the discovery of novel gene mutations responsible for rare Mendelian disorders affecting bone mass. This has recently been illustrated by low-density lipoprotein receptor-related protein 5 (LRP5) mutations responsible for osteoporosis-pseudoglioma (OPPG) and high bone mass (HBM) syndromes, with or without sclerosing bone dysplasias (), with the finding that the gene maps to a previously identified locus (11q12-13) for BMD in the population ().
In addition to the crucial impact of transgenic and knockout experiments in mice, the second axis of research concerns the identification of QTLs for bone mass and structure using crosses of various inbred strains of mice (). A major advantage of this approach is that mice allow investigation of discrete skeletal traits pertaining to bone “quality” and BMD. Moreover, sample size is fairly unlimited, the genetic background homogeneous, and the environmental variance low. Mice also allow direct probing of the causal linkage between a given QTL and skeletal phenotype(s) by generating congenic animals (). Yet, the mouse approach also has its limitations, including the fact that the skeletal phenotype observed in adult congenic mice might reflect adaptation to genetic influences expressed earlier during bone mass development, rather than direct QTL effects on the phenotype. Moreover, just as in humans, the loci identified thus far in mice are large, and strategies to map precise genes within the loci are just now emerging. A beautiful illustration of these new strategies has recently been provided by a joint group of academic and industry investigators, which used microarrays for gene expression profiling to identify two new genes — Frzb1, which codes for an inhibitor of Wnt signaling (the ligand of LRP5), and Alox15, which codes for a lipid-peroxidizing enzyme — implicated in the regulation of peak bone mass in congenic mice ().
In summary, genome-wide screens in humans and mice have taught us where to look for specific genes, and recent developments in molecular technology now make it possible to identify these genes and their variants. Most importantly, these examples of newly identified genes for bone mass in both humans and mice (LRP5, BMP-2, Alox15, and Frzb1)provide proof of the theory that osteoporosis genetics has the ability to improve our understanding of bone physiology.
The third major and most popular approach to osteoporosis genetics has been through association studies with allelic variation in known candidate genes. Doubtlessly, this approach has led to fewer brilliant discoveries than have those mentioned above and has certainly been the most criticized. Nonetheless, the modern era of osteoporosis genetics began with this approach and will end with it, because the ultimate demonstration for an implication of any gene variant to osteoporosis risk will require large-scale association studies in humans (). In this case, we should be able to learn from both our errors and successes. Thus, we know now the limitations of small sample sizes, as seen in many individual studies, but also the value of metaanalyses that have shown the actually small, but significant, effect size of some gene variations, including in VDR (), estrogen receptor alpha (ESR1) (), and Col1A1 genes (). We also acknowledge the risks of false positive associations due to stratification effects in populations with high genetic heterogeneity and should consider the use of “control” gene variants in this situation. Moreover, we became aware of the small contribution to the population-based variance for the trait(s) of isolated and synonymous SNPs and have learned the utility of multiple SNPs and their related haplotypes, as well as of functional SNPs, such as missense substitutions and variations in gene regulatory (promoter) regions. Eventually, we now understand the importance of gene interactions with environmental influences on the trait, as shown in early studies on calcium intake interaction with VDR alleles on bone mass acquisition and maintenance (). Furthermore, association studies with polymorphisms in the methylenetetrahydrofolate reductase gene (MTHFR) () and their possible interaction with folate levels in humans have brought our attention to a dietary factor (folate) and a metabolic pathway (homocysteine) possibly implicated in bone health.
What are the implications?Considering the evidence summarized above, it is now time to delineate some potential lines of research in the field of osteoporosis genetics. It is unlikely that many other genome-wide screening studies will be funded to identify additional QTLs for areal bone mineral density (aBMD), at least in whites. In contrast, near-future availability of novel devices to assess bone structure in humans noninvasively (such as MRI and microcomputed tomography) will likely allow linkage studies to be performed for bone “quality” traits, including microarchitecture of cancellous bone. In the meantime, we could theoretically gain power and confidence from the existing linkage studies by attempting to perform a metaanalysis and/or by pooling the multiple datasets available. There would obviously be personal and technical difficulties in this task, mostly because of the heterogeneity of the study samples (sibling pairs vs. multigeneration families; see Table 1) and the heterogeneity of the microsatellite markers used. Synteny between humans and mice should also allow prioritizing candidate QTLs and genes for dense SNP mapping in both species. In addition, we shall develop new strategies to move beyond QTLs. Thus, a recent study with “only” 119 markers restricted to one contig under the linkage peaks for rheumatoid arthritis (RA) at the 1p36 locus — also a strong QTL for aBMD in humans () — revealed a clear picture of strong association to only one gene (). Other technological improvements, such as high-throughput multiplex PCR-Invader® assays, have allowed the concomitant genotyping of more than 100 x 103 SNPs, mapping within 16,000 genes in hundreds of RA cases and controls, thereby identifying at once more than 1000 SNPs, and meaningfully, 25 haplotype blocks associated with the disease ().
More simply, we should begin directly assessing whether the candidate gene polymorphisms thus far identified in association studies could be translated into clinical practice. Guidelines for population screening, as applied to genetic susceptibility to disease, have recently been published (). A most important goal will be to evaluate the positive and negative predictive values of gene markers with respect to the disease in the targeted population. For instance, association of functional variants in the Col1A1 gene with collagen composition and fracture (), as well as association of functional interleukin 6 (IL-6) promoter alleles with bone resorption markers (), suggest that genetic variation could influence bone “quality” above and beyond BMD. Thus, Col1A1 and IL-6 alleles could be tested with regard to fracture prediction, in addition to dual-energy x-ray absorptiometry (DXA) measurements. Furthermore, suggestions about distinct gene polymorphisms interacting with environment and/or lifestyle factors should prompt nutrigenetic studies in the osteoporosis field. Thus, we could prospectively test whether improving bone mass changes with calcium supplements in childhood and in the elderly population would require different doses according to VDR alleles, as suggested by some experimentally derived schemes about interactions between these factors (). Furthermore, we could prospectively investigate whether preventing bone loss with estrogens or selective estrogen receptor modulators in postmenopausal women would be influenced by gene variants along the biological pathway regulating bone resorption, including ESR1 and IL-6 alleles. Insights from such pharmacogenetic studies have indeed been advocated to develop targeted strategies with improved cost effectiveness for the prevention and treatment of osteoporosis ().
In conclusion, tremendous developments have been made in genomics and in osteoporosis genetics in the past 10 years, and both the quantity and quality of this information will continue to rise steeply. Rather than looking disdainfully at the “genetic revolution of medicine” () as applied to our field, we should expend effort to bring this new knowledge from the bench to the bedside.
This work is licensed under a
Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.