Determinants of radiographic severity in ankylosing spondylitis

Abstract

Ankylosing Spondylitis (AS) is a highly heritable condition characterised by excessive bone formation and stiffness of the spine and pelvis causing functional impairment. Functional impairment is strongly correlated with radiographic signs of bone formation. Clinicians can accurately measure bone formation but have a limited ability to predict or prevent it. A small number of non-genetic factors have been identified that explain approximately one third of the variation in radiographic severity. No genetic predictors of radiographic severity have consistently been identified. This work investigates the contribution of non-genetic and genetic factors to radiographic severity. Knowledge of causal genetic variants could be used to predict prognosis or to design therapeutics. Spinal radiographs were scored using the modified Stoke Ankylosing Spondylitis Severity Score (or a modified version of this score), mostly by a single calibrated reader. Observed modified Stoke Ankylosing Spondylitis Severity Scores were compared between males and females. Cross-sectional modelling was performed on 1392 subjects by comparing previously recorded clinical covariates and demographics with the latest radiographic score for each patient, using a negative binomial model. Residuals from this model were considered to be the phenotype corrected for known covariates. A quantitative Genome Wide Association Study (GWAS) was performed on 1253 cases, comparing the corrected phenotype with millions of genetic markers per subject. Genomewide Complex Trait Analysis (GCTA) was used to estimate the contribution of common variants with small effect sizes to the phenotype. An allelic risk score approach was used to examine the hypothesis that known bone mineral density, height, AS susceptibility and smoking dependence related loci also affect AS radiographic severity. ii Known contributors to radiographic severity - gender, smoking and symptom duration were found to explain approximately one-third of phenotypic variation. GCTA estimated that common variants contribute at least 33% of phenotypic variation. No signal was identified to link known contributors to AS susceptibility, including the MHC region, with radiographic severity. Known AS susceptibility, BMD, height and smoking dependence loci were not associated with severity upon allelic risk score analysis. GWAS suggested that loci involved in bone metabolism and innate inflammation may influence radiographic severity. Some genes in the same pathways as AS susceptibility loci, including the TNF pathway, had a suggestive association with the phenotype. Additionally suggestive associations between loci related to bone formation and immune-microbiota interactions in the gut were present, which, if confirmed, may provide a mechanism for observed patterns of microbiota in human AS cases. This study did not definitively associate any specific loci with the phenotype, probably due to its low power and limitations of the phenotype modelling. Modelling was limited to a cross-sectional analysis, slightly underestimated the frequency of ankylosis and did not correct for some known influences of radiographic severity such as medication use and occupation. Robustness of the model to the effects of confounders was assessed by local sensitivity analysis. As the first comprehensive study of the relationship between genotype and radiographic severity, this study confirms that genetic factors make a significant contribution. Involvement of loci highlighted by this study should be confirmed by larger genetic association studies and functional profiling of the new candidate genes in AS cases. Once identified, genetic determinants of radiographic severity would make attractive targets for therapeutic intervention and prediction of prognosis.