Allelic Expression Imbalance Refined Genome-Wide Association Signals In Osteoarthritis

Purpose: A continuum of genome wide association studies (GWAS) on osteoarthritis (OA) with ever increasing sample sizes have resulted in 96 independent loci associating to OA related phenotypes. Consequently, the urge to annotate transcriptional effects of identified risk SNPs in OA relevant tissue is growing. To address functionality of identified risk SNPs we performed allelic expression imbalance (AEI) in whole transcriptomics data of articular cartilage from OA patients. Methods: To perform AEI, we used transcriptome-wide sequencing (RNA-seq) data from articular cartilage of 65 OA patients (56 preserved and 9 lesioned cartilage). First, we calculated the linkage disequilibrium (LD) of each 96 associated loci reported on GWAS catalog database, taking a window of 1 Mb (+/- 500 kb) from the most associated variant reported by each study. Furthermore, the count fraction of the alternative alleles among the alternative and reference alleles together (φ) was determined for each heterozygous individual. Finally, a meta-analysis was performed to generate a meta-φ and P-value for each genetic variant with a false discovery rate (FDR) correction for multiple tests. Additionally, we performed differential expression analysis between paired samples of preserved and lesioned cartilage (N = 35 pairs). Results: We identified 26 transcript SNPs subject to AEI of positional genes (1Mb LD window), in linkage disequilibrium with 20 reported robust OA risk SNPs. Notable among these were AEI SNPs (N=2) located in long intergenic non-coding RNAs (lincRNAs) such as, MALAT1 (AEI meta-φ=0.46, FDR=3.3 x 10-4) and ILF3-DT (meta-φ = 0.4, FDR=4.3 x 10-4) and AEI SNPs (N=16) located in genes not previously highlighted as possible OA genes in the GWAS catalogue, such as ECM1 encoding extracellular matrix protein1 (meta-φ =0.53 , FDR = 3x10-3). To further strengthen these genes as being involved in OA pathophysiology we performed differential expression between preserved and lesioned cartilage and found 9 AEI genes, including TNC (meta-φ=0.47, FDR = 2.08x10-7 , FC= 1.4), COLGALT2 (meta-φ = 0.47, FDR = 0.002, FC=1.42), and lincRNA ILF3-DT (AEI meta-φ = 0.4, FDR=4.3 x 10-4, FC = 0.69) (Table 1).Table 1Allelic Expression Imbalance Genes with differential expression between preserved and lesioned OA cartilage.AEI GeneGWAS GeneGWAS SNPAEI SNPmetaPhiFDRCOLGALT2Not reportedrs11583641rs115836410.462.1x10-2ILF3-DTNot reportedrs1560707rs48045140.44.3X10-4ECM1Not reportedrs12040949rs132940.533.5x10-3MALAT1Not reportedrs10896015rs32004010.463.3 x 10-4TNCTNCrs2480930rs22748360.472.08x10-7 Open table in a new tab Conclusions: Here, we present a robust and comprehensive search for functional AEI SNPs in LD with the OA risk SNPs and identified 16 genes subject to AEI, that were not previously reported by GWAS, including two lincRNAs. Our data provide a roadmap for further investigation in OA genetic variants that act via affecting gene expression in the OA disease-relevant tissue cartilage. In addition, our approach substantially helped to refine and prioritized OA risk genes for further downstream functional (in vitro) analyses.