POS0555 THE NON-APOPTOTIC PROGRAMMED-CELL-DEATH-RELATED SIGNATURE PREDICTS ANTI-TNF THERAPY NONRESPONSE IN PATIENTS WITH RHEUMATOID ARTHRITIS

Abstract

BackgroundEULAR guideline in rheumatoid arthritis (RA) recommended the primary failure of the first-line conventional synthetic modifying anti-rheumatic drugs (csDMARDs) patients switching to tumor necrosis factor alpha inhibitors (TNFi) [1]. Nevertheless, approximately 30-40% csDMARDs-IR patients also experience inefficacy of TNFi [2]. There is still no index to predict whether TNFi would be responded or not. Moreover, only few studies had focused on the relationship between TNFi nonresponse and other cell programmed deaths except apoptosis.ObjectivesTo predict the possibility of TNFi response prior to prescript in RA patients with the biomarkers of non-apoptotic programmed cell death in synovial cells.MethodsThe datasets of 22 TNFi treated RA synovial samples were enrolled from the Gene Expression Omnibus (GEO) database (GSE140036 and GSE15602). And the differentially expressed genes (DEGs) and modules related to TNFi treatment through weight gene correlation network analysis (WGCNA) were identified with the R packages “limma” and “WGCNA”. Then the enrichment analysis among the shared genes was performed through the R.4.1.2, Metascape website, and WebGestalt website. Following with the confirmation of the non-apoptotic programmed cell death (NAPCD) genes in the shared genes with Student’s T-Test. Furthermore, the TNFi treatment cohort was clustered based on the hub genes, making the receiver operating characteristic (ROC) curve analysis. Moreover, the least absolute shrinkage and selection operator (LASSO) model was constructed to identify the predictive genes.Results2624 DEGs were identified significantly, including 161 upregulated genes and 2463 downregulated genes. One module with TNFi treatment was constructed in WGCNA, significant in both response and nonresponse. Then the gene signatures for TNFi nonresponse were collected from overlaps 2260 genes in above. And we found 38 NAPCD genes might play role in TNFi nonresponse, but reserved 33 genes which statistically significant with T-Test. 22 TNFi treated synovial samples in GEO database could be classified into response or nonresponse subgroups. The ROC curve showed that the AUC for 32 genes in these samples ranged from 0.7 to 0.9, expected for CD46. At last, the LASSO model indicated that CASP5, CAPN10, ITGB4, NLRP2, and SLC2A8 could predict the TNFi nonresponse, as the risk score = CASP5 × 0.028 + CAPN10 × 0.064 + ITGB4× 0.080+ NLRP2 × 0.317+ SLC2A8 × 0.090 (Figure 1).Figure 1.Predictive model of TNFi nonresponse based on NAPCD genes. (A) Volcano map of differential expressed genes; (B) Correlation heat map of gene modules and phenotypes, the red is positively correlated with the phenotype, blue is negatively correlated with the phenotype; (C) The shared 38 genes of TNFi response & nonresponse DEGs, among the WGCNA turquoise module and cell programmed death genes; (D) Consensus clustering matrix for k = 2; (E) The ROC curve of 33 genes; (F) LASSO regression of the 32 genes,except for CD46; (G) Nomogram for predicting TNFi nonresponse in TNFi treatment RA cohort, indicated five possible indicators (CASP5, CAPN10, ITGB4, NLRP2, and SLC2A8) were closely related to TNFi nonresponse.ConclusionOur study firstly screened out the 38 NAPCD candidate genes signatures in RA synovial tissues which took part in TNFi nonresponse through WGCNA and DEGs. Further analysis confirmed that five possible indicators (CASP5, CAPN10, ITGB4, NLRP2, and SLC2A8) were closely related to TNFi nonresponse.