Genetics implicates overactive osteogenesis in the development of diffuse idiopathic skeletal hyperostosis

Abstract

Diffuse idiopathic skeletal hyperostosis (DISH) is a condition where adjacent vertebrae become fused through formation of osteophytes. The genetic and epidemiological etiology of this condition is not well understood. Here, we implemented a machine learning algorithm to assess the prevalence and severity of the pathology in ~40,000 lateral DXA scans in the UK Biobank Imaging cohort. We find that DISH is highly prevalent, above the age of 45, ~20% of men and ~8% of women having multiple osteophytes. Surprisingly, we find strong phenotypic and genetic association of DISH with increased bone mineral density and content throughout the entire skeletal system. Genetic association analysis identified ten loci associated with DISH, including multiple genes involved in bone remodeling (RUNX2, IL11, GDF5, CCDC91, NOG, and ROR2). Overall, this study describes genetics of DISH and implicates the role of overactive osteogenesis as a key driver of the pathology.

Fig. 1. Overview of the machine learning results and validation.

A Representative images of individual osteophyte flow scores and calculation of a total DISH score across all vertebrae. B Comparison of machine learning derived DISH scores with median scores from three annotators in the validation dataset (n = 300 images). The centerline of the box represents the median ML DISH score while the whiskers indicate the 25th and 75th percentile of the ML DISH score within that category. Individual comparison between annotators and ML scores can be found in Supplementary Fig. 3. C Comparison of ML scores with assessment by two expert radiologists based on Resnick diagnostic criteria for DISH. Majority of individuals above a score of 12 have DISH based on the Resnick criteria.

Fig. 2. DISH Prevalence.

A The distribution of DISH scores in the UK Biobank Imaging cohort (~40 K participants) stratified by age and gender. Men are more likely to present with multiple osteophytes than women. There is a strong age dependent increase in DISH scores. B Association of DISH with self reported pain questionnaires in multivariate adjusted (age, sex, age*sex, age^2, township deprivation index, smoking, ethnicity) logistic model. DISH is linked to increased use of pain medications and Neck and shoulder pain. The number of participants who replied to these questions in the questionnaire during the imaging visit varied widely from 2360 for stomach pains to 41,233 participants for taking pain medication and the logistic regression is performed with information from all participants who answered the questionnaire. C Association of DISH with pre-existing comorbidities in EHR records in multivariate logistic model (same adjustments as above). DISH is significantly linked to metabolic diseases, renal disorders and other forms of arthritis (n = 41,233). The blue dots represent the mean effect size, the error bars represent 95% confidence intervals.

Fig. 3. DISH Bone Density Associations.

A Multivariate LASSO regression identifies independent physiological predictors of DISH (n = 41,233). Only features where 95% credible intervals do not overlap zero effect size are shown. Among the strongest risk factors are age, sex, and various BMC measures. The blue dots represent the mean effect size, the error bars represent 95% confidence intervals in the regression model. B Increased bone mineral density (BMD) across the entire skeletal system is associated with increased DISH score. Plot shows age and sex adjusted spline fits (GAM model) across multiple distal sites. Association of DISH with BMD in the Head, Femur Shaft, and Femur Ward’s skeletal sites (areas not known to form osteophytes) suggest that BMD is an independent risk factor for osteophyte formation. The shaded regions represent the 95% confidence interval of the spline fit.

Fig. 4. Genetic associations and genetic correlation.

A Manhattan Plot of DISH GWAS. Associations reaching genome-wide significance below p value < 5e⁻⁸ were labeled. The p value is the two-tailed test for association of trait with variant. B Genetic correlation between DISH and other phenotypes derived from DXA scans, estimated using LD score regression from UK Biobank summary statistics. C GDF5 locus showing a clear overlap of genome-wide significant signals with DISH, Osteoarthritis (OA), and lung Forced Expiratory Volume in one second (FEV1). The p value is the two-tailed test for association of trait with variant.

Fig. 5. Genetic Colocalization Analysis.

The colocalization of genetic signals of DISH with different related traits as well as gene expression of neighboring genes from GTEx in any tissue. A colocalized signal for a particular trait is represented by a circle on the corresponding locus. The size of the circle represents the strength of the colocalization signal. PP4 is the posterior probability that both traits are caused by the same variant within this locus, while PP3 represents the posterior probability that both traits are caused by different variants within the same locus. The color of the circle represents whether the risk allele changes DISH as well as the corresponding trait in the same direction (correlated/blue) or opposite directions (anti-correlated/red). L1-L4—lumbar vertebral spine, BMA—DXA based bone mineral area, BMC—DXA based bone mineral content, FEV1—lung Forced Expiratory Volume in one second. ALPL—Alkaline phosphatase.

Fig. 6. Overview of genetic and environmental risk factors associated with development of DISH.

In addition to male sex and age, pre-existing conditions such T2D, obesity, and osteoarthritis are risk factors associated with DISH. Genetic analysis (GWAS, colocalization, and Mendelian randomization) points to genes involved in overactive osteogenesis as drivers of the pathology (highlighted in orange). The increase in osteogenesis, and consequence in increases in BMD and BMC measures is observed throughout the entire body. Molecular mechanisms likely involve gain of function in multiple signaling pathways such Wnt signaling, IL11 signaling, and BMP-signaling. Conversely, loss of inhibitory BMP proteins such as Noggin and CRDL2 likely increases BMP-signaling. In prognostic outcomes, DISH is associated with increased risk of metabolic and sleep disorder diagnoses.