P200 Combining protein quantitative trait and genetic risk score analysis to identify biomarkers of treatment response to TNFi in patients with rheumatoid arthritis

Abstract

Abstract Background/Aims Rheumatoid arthritis (RA) is an inflammatory joint disease that can lead to disability if inflammation is not controlled. Biologic disease modifying anti-rheumatic drugs (bDMARDs), including tumour necrosis factor inhibitors (TNFi), are an effective line of therapy for moderate to severe RA. However, treatment response to TNFi is not universal and cannot be predicted from clinical factors. In this study, a protein quantitative trait loci (pQTL) study and genetic risk score (GRS) analysis were combined to identify protein-based biomarkers that are predictive of TNFi treatment response. Methods Protein expression data were generated using Sequential Window Acquisition of All Theoretical Mass Spectra (SWATH-MS) in serum samples taken from 80 RA patients about to commence therapy with etanercept. Protein expression was log2 transformed and the k-nearest neighbour model was used to impute missing values. A pQTL study was performed in patients with available imputed genome-wide genetic variation data to detect cis-acting genetic markers (p < 1E-05). GRS for pQTLs were subsequently tested in 1,430 RA patients with available genome-wide genetic and TNFi treatment response data (improvement in DAS28 between pre-treatment and 3/6 months on drug). The GRS analysis was adjusted for baseline disease activity, sex, conventional synthetic DMARD co-therapy and the first 10 principal components, calculated from the genetic datasets. Results Following imputation, expression levels for 271 proteins were analysed in 69 RA patients with available genotype data. 514 cis-pQTLs were found for 16 proteins. GRS for the proteins Apolipoprotein(a) (UniProt ID P08519, p = 0.017) and Carboxypeptidase N subunit 2 (P22792, p = 0.027) were found to be modestly associated with treatment response; however, scores for both proteins explained less than 1% of the variance in DAS28 difference between time-points. Conclusion This study identified two protein-based genetic biomarkers of treatment response to TNFi. However, genetic scores based on these proteins are unlikely to be useful predictors, explaining little variance in on-treatment disease activity. Disclosure Z. Li: None. S.F. Ling: None. N. Nair: None. J.D. Isaacs: None. K.L. Hyrich: Honoraria; Abbvie. A.W. Morgan: None. A.G. Wilson: None. A. Barton: None. D. Plant: None.