Genome-wide association study of Degenerative Disc Disease (DDD)

Abstract

Low back pain is a common problem, where 70-85% of people experienced back pain in their life. There are a number of causes for LBP and the most important one is degenerative disc disease (DDD). The etiology of DDD is complex. Both environmental and genetic factors are possible contributing factors. Familial aggregation and similar degeneration patterns observed in twins suggested that genetic factors play an important role in DDD. Genetic variants related to DDD were already suggested in several candidate gene studies. To decipher the genetic component of DDD, we have initiated a whole-genome case-control association study for DDD using Affymetrix Mapping 500K microarray. Volunteers of Southern Chinese origin, aged from 18 - 65 years, were recruited and underwent blood sampling and magnetic resonance imaging (MRI) of the spine with consent. DDD status was graded according to Schneiderman’s classification to give a DDD score. MRIs were analyzed by two experienced physicians (K.M.C.C., J.K.) blinded to the results of genetic analysis and clinical history. Scores of each of the five lumbar intervertebral discs were summed. An age-adjustment procedure for the DDD score was carried out using a standardization process to give age-adjusted DDD score (AADS). In this procedure, each sample’s DDD score was standardized with the mean and standard deviation calculated from individuals having ages below and above a specified bandwidth. Cases and controls were selected from samples having the highest and lowest, respectively, AADSs. From these criteria, 200 cases and controls, respectively, were selected. DNA was extracted from whole blood and quality was checked to be conforming to the quality threshold for genotyping with 500K array set. Genotypes were called with BRLMM algorithm. Analysis was carried out using Plink, and Haploview was used for visualization of the data. Quality control was performed with Plink. Samples and markers with bad qualities were excluded. Relationships between individuals were checked by means and variances of the identity-by-state (IBS) to ensure there are no close relatives. The filtered samples and markers were corrected for hidden population structure with Eigenstrat. 318,747 SNPs remained after per-SNP QC filtering. After a further correction of hidden population substructure by Eigenstrat, markers with p-values of order 10-5 could be detected.