Osteoporosis is a common aging-related disease characterized by progressive thinning and porosity of bone, resulting in increased fracture risk. Affecting more than 9 million individuals per year worldwide, osteoporotic fractures represent a significant healthcare burden. Clinical risk factors—with bone mineral density (BMD) being the most relevant—are measured to diagnose osteoporosis and to predict fracture risk, but high-risk individuals are often only diagnosed after suffering a low trauma fracture. Taken together, this means that we urgently need to better understand the causes of osteoporosis and to find novel drug targets. Therefore, my PhD research focused on elucidating the determinants of this disease.
1. Genetic determinants. Genome-wide association studies (GWAS) can identify biological mechanisms for osteoporosis pathophysiology and drug therapies. By 2013, GWAS of hip and spine BMD (N=84,000) had identified 56 loci, including four regions containing genes (or their pathways) targeted by existing drugs. Despite their success, these GWAS only accounted for ~5.8% of the estimated heritability of BMD. This suggested that more genes (and possible drug targets) remained to be discovered. To address this, we used the UK Biobank, a large cohort with data for 500,000 participants. I co-led analyses of the UK Biobank’s interim (N=150,000) and full data releases, identifying a total of 518 loci (454 novel), explaining up to 20% of its estimated heritability, using data from 426,824 participants. We published both interim and full data analyses in Nature Genetics, the top-ranked genetics research journal, highlighting GPC6 and DAAM2 as novel candidates for drug targets. These studies were of interest for 1) bone researchers, as they provided hundreds of new genes for experimental follow-up, and 2) for geneticists, as a guide on how to study common diseases and complex traits. Furthermore, our studies enable personalized genomics research for disease prediction, given the high amount of estimated heritability. An example of this is recent work by Forgetta et al. (bioRxiv 2018) which used genomics-predicted heel bone traits to identify UK Biobank participants at high-risk for fracture.
2. Epigenetic determinants. As the role of epigenetics in common disease is poorly defined, many international cohorts have measured DNA methylation in whole blood to study its relation to various traits or diseases. To study osteoporosis, I assembled an international working group (five cohorts from four countries) and led an epigenome-wide association study (EWAS) testing the association between DNA methylation levels and BMD. This was one of the largest EWAS of its time (N=5,515) and remains the largest for any musculoskeletal trait. Researchers had hypothesized that findings from EWAS would explain the “missing heritability” of many traits or diseases. We found that this was not the case for osteoporosis. Our study was statistically well-powered but failed to identify any strong epigenetic signals for BMD. We published our findings in the Journal for Bone and Mineral Research, the top-ranked bone research journal, as a cautionary tale for other researchers. We described that whole blood is not a suitable tissue in which to study osteoporosis epigenetics, but that directly studying more difficult to obtain bone cells (e.g. osteoblasts) may be worth pursuing.
3. Causal risk factors. Clinical risk factors are measured to predict fracture risk, but whether they were causal for fracture was unknown. I co-led the largest genetic study of fracture of its time with 562,258 participants (185,057 cases) to estimate the causal effect of 15 clinical risk factors for fracture. Our work was published in the British Medical Journal, one of the leading general medical journals. We found that low BMD was the only risk factor with strong causal evidence for fracture, indicating that increasing BMD is the most effective way to protect against fracture. Importantly, intervention treatments targeting other risk factors may not be useful (e.g. vitamin D supplementation in the general population), therefore, our findings have implications for medical guidelines on preventing fracture. Furthermore, this study reinforces Contribution 1 on studying the genetic determinants of BMD, and leveraging genomics-prediction methods, for fracture prevention.