Breaking Down Bone Strength: A Perspective on the Future of Skeletal Genetics

WITH INCREASING INTENSITYover the past decade, researchers have pursued genetic approaches to understanding the basis of skeletal fragility. This month's issue of the JBMR includes a report by Beamer and associates1 that identifies quantitative trait loci (QTLs) for femoral and vertebral volumetric bone mineral density (vBMD) and presents us with an opportunity to reflect on recent progress and future directions in bone genetics. BMD is only one of several parameters related to biomechanical performance. Fracture, not BMD, is the clinically relevant outcome; ideally, scientists would perform genetic analyses with fracture as an endpoint rather than on surrogate traits that correlate imperfectly with this abnormality. Deng and colleagues2 estimated that the narrow-sense heritability of Colles' fracture was approximately 0.25 in a cohort of white American women, thus accounting for approximately one-quarter of total Colles' fracture risk variation observed. This indicates that the magnitude of the genetic contribution to fracture warrants further study, but important caveats exist. Fractures are relatively rare and tend to occur late in life. Should the genetic basis of human fracture prove to be quantitative, then the contributions of individual loci to the total fracture risk are likely too small to allow successful mapping of individual QTLs with realistic sample sizes. These considerations impose important and possibly insurmountable obstacles to recruitment of an adequate population in which to perform mapping studies of Colles' fracture in humans. In addition, because Colles' fractures are frequent relative to hip fractures, collecting an adequate sample to perform equivalent analyses of hip fracture may prove a daunting challenge. Further, vertebral fractures, although relatively common, do not always come to clinical attention3, 4 and their diagnosis may prove uncertain,5-9 posing potential problems in the study of vertebral fracture risk. Investigators have adopted alternative strategies using surrogate endpoints to overcome the practical limitations of studying human fracture directly. One approach uses biomechanical performance in a contrived experimental scheme as an analog for natural fracture; this allows functional assessment of the bone, but the mechanisms in biomechanical testing differ from those of naturally occurring fractures. In addition, cadaver specimens and specimens obtained from model organisms, both used frequently in these investigations, impose additional differences from in vivo human fractures. A second approach, based on epidemiological data, focuses on predictors of fracture rather than analogs of fracture and is reflected by the vast literature on densitometric assessment of bone. The success of these methods is incontrovertible; indeed the World Health Organization (WHO) diagnostic criteria for osteoporosis rely primarily on densitometric assessment.10 The same logic underlies investigations based on instruments that assess lifestyle factors, comorbidities, and mobility, and there is broad agreement in the community regarding the risk factors for fracture.11, 12 However, markers of fracture risk are variously defined and although related are nevertheless distinct. This is readily apparent when comparing densitometric parameters; cross-calibrating fracture risk based on T scores obtained by different techniques at different sites13-19 poses an important patient care problem. Similarly, considerable overlap exists among the data included in assessments of clinical risk factors for fracture, but cross-calibration among instruments has not yet been achieved. Optimizing predictive ability does not necessarily advance mechanistic understanding of fracture biology. Areal BMD (aBMD) measured by dual-energy X-ray absorptiometry (DXA) is the single best predictor of fracture risk at the studied site (using lateral view at the spine). Scientists have established that large bones have higher aBMDs than small bones having equal vBMD20 because of projection artifact. We also know that larger bones are stronger than smaller bones having equivalent tissue strength. By combining both size and mineral density into a single measure, aBMD achieves better fracture prediction than vBMD (in which the contribution of bone size is explicitly removed from consideration). Yet, to understand the underlying bone biology, we must treat bone size and BMD as distinct properties.21-23 In contrast, biomechanical studies24, 25 divide performance into components that promote investigation of mechanisms. The most basic distinction is between structural and material properties. Structural properties are possessed by an entire bone and are determined jointly by its anatomy and its tissue strength. Material properties are those of the bone tissue per se, after correction for differences in anatomy. Distinct components of biomechanical performance also can be identified in interpreting the load-deformation curve obtained in testing a specimen, and each of these properties has a corresponding material parameter obtained by normalizing for anatomy. Different testing schemes yield information on other mechanical properties; for example, repetitive, low-intensity loading can be used to assess fatigue. Similarly, the anatomic component of structural strength can be divided further; the distinction between cortical and trabecular bone reflects this division, as do ongoing efforts to quantify various architectural bone properties. Biomechanical studies naturally lead to efforts to analyze bone strength into ever more refined components, but such partition is not always achieved in practice. Bone is an anisotropic tissue, and its biomechanical performance varies according to the orientation of the tissue relative to the applied load26, 27; this is particularly evident in trabecular bone.28-30 Despite these limitations, a reductionistic program does advance our understanding of the biomechanical properties of bone. The reductionist approach embodied by biomechanics does not preclude epidemiological assessment. Both epidemiological and interventional investigations have identified risk factors for fracture other than bone mass. Advancing age,31, 32 past personal atraumatic fracture history,11, 33-35 and maternal hip fracture history12 increase fracture risk. In trials of alendronate and transdermal estrogen, the fracture benefit exceeded that predicted from the observed BMD increase.36, 37 In contrast, sodium fluoride treatment increases BMD but does not improve and may even worsen fracture outcomes.38, 39 Taken together, these data show that non-BMD factors, often referred to collectively as bone quality, contribute to the material strength of bone, thereby linking disparate research areas. Researchers use several different strategies to study skeletal genetics. Each provides a potentially valuable approach and addresses somewhat different questions.40 The linkage studies considered here address the locations of genes (loci) in which segregation contributes to differences in the trait of interest. In human samples, this question is distinct from that of identifying which allele of that gene is responsible for different values of the trait. In contrast, the existence of inbred mouse lines allows investigation of both loci and alleles to be accomplished simultaneously.41 In human samples scientists have successfully mapped a locus conferring high BMD as a simple Mendelian trait.42 Koller et al. implicate the same genomic region as a QTL contributing to less dramatic variation in BMD.43, 44 Two other human genome scans reported evidence of possible linkage for hip and spine45 and forearm.46 These investigations found evidence for a QTL on chromosome 2p in spite of the different sites and populations studied. Each of these studies used aBMD as the phenotypic endpoint, based on the ease, economy, safety, and acceptance of aBMD as a surrogate measure of skeletal fragility. In mice, researchers have studied a broader range of phenotypes. Klein et al.47 studied whole body aBMD in the B × D recombinant inbred lines. Using DXA scanning, Benes and colleagues48 examined vertebral aBMD in a pair of crosses between SAMP6 and either SAMR2 or AKR/J. Shimizu et al.49 studied femoral cortical thickness index, the fraction of diaphyseal cross-section occupied by cortical bone on plain radiographs, in a cross between SAMP6 and SAMP2. Beamer and colleagues have reported two related studies: a cross between C57BL/6J and Cast/EiJ in which femoral vBMD was mapped50 and a study of insulin-like growth factor 1 (IGF-1) levels in a cross between C57BL/6J and C3H/HeJ.51 Humeral biomechanical performance and dimensions, ash percentage, and body mass in the HcB/Dem recombinant congenic strains were assessed by Yershov et al.52 In this issue of the JBMR, Beamer et al. present the results of mapping femoral and vertebral vBMD in a cross between C57BL/6J and C3H/HeJ.1 Their results are notable for several reasons. First, a subset of the QTLs contributes to vBMD at both sites studied, while others contribute to vBMD at only one site, providing a biological context for the clinical finding of disparate BMDs at different sites in the same individual. Second, in several cases the same QTLs have been mapped in different crosses, using different but related phenotypes. This provides compelling evidence that a bone mass gene in fact has been identified. Third, the number of animals studied in this cross enabled testing for pairwise epistatic interactions, which were not found. It is too soon to say that epistasis does not contribute significantly to vBMD variation in general, but these efforts remind us that we must investigate gene-gene and gene-environment interactions. Fourth, the authors are unequivocal in stating their ambition to help clarify our understanding of the mechanisms underlying bone phenotypes. Successfully mapping QTLs represents an early step in understanding the genetic basis of skeletal function. With the approximate locations of some vBMD QTLs identified, we must face the problem of determining which of the positional candidates thus defined are the relevant gene(s). Similar research efforts in other diseases, notably type I diabetes53-55 and other autoimmune diseases,56-58 suggest that successfully subdividing the overall phenotype into pathophysiologically relevant intermediate phenotypes is one of the major tasks that lies ahead. Subdivision of traits may lead to phenotypes, the genetics of which are Mendelian rather than quantitative, greatly simplifying their genetic analysis, as illustrated by the high bone mass locus.42 Even if we do not succeed in reducing traits to oligogenic phenotypes, the effort undoubtedly will stimulate new ideas about how whole bone anatomy is established, how bone responds to mechanical loading, how matrix proteins and minerals interact, and how bone metabolism changes during development and aging. Successful integration of all these research areas into genetic investigations of fracture risk should enable us to choose among the myriad of potential intermediate phenotypes. In addition, new instruments and new analytical methods provide opportunities to define additional phenotypes.59-62 Work to date has already used a variety of different, albeit related, phenotypes. Intermediate phenotypes will add more traits into the mix and present us with the additional challenge of relating the phenotypes to each other. If we design future genetic experiments to include a variety of traits, we may discover evidence of pleiotropy, that is, a single gene controlling multiple phenotypes.52, 63, 64 The data presented by Beamer and colleagues1 contribute to this effort by distinguishing between QTLs acting at both femur and spine and QTLs acting in a site-specific manner. Finally, progress in genomics favors investigations based on interesting and insightful biology. Recent completion of the human genome sequence65 and the pending completion of the mouse and rat sequences will devalue some of the skills geneticists have mastered. High-density maps and microarray technologies will allow adoption of multigenerational association analyses having greater statistical power than traditional linkage methods.66 These advances demand that the bone geneticists of the future have command of bone biology. The article by Beamer et al.1 provides a first step in achieving this goal. The author gratefully acknowledges support from the Department of the Army award DAMD17-00-1-0071. The U.S. Army Medical Research Acquisition Activity, Ft. Detrick, MD, is the awarding and administering acquisition office. The views expressed herein do not necessarily reflect the position or policy of the U.S. government, and no official endorsement should be inferred.