Background TNF was the first cytokine to be validated as a molecular target of Rheumatoid Arthritis (RA) and still remains at the first-line of therapeutic interventions with biologics. Even though widely used anti-TNF agents, such as infliximab, etanercept, adalimumab and certolizumab pegol all target the same protein, they nonetheless exhibit differences, likely due to their biological origin (humanised or fully human) and biochemical properties (antibody or receptor). In this work, we sought to characterise this variability, at the molecular level, through the creation of robust gene and functional signatures from reproducible gene expression profiles of drug intervention in the huTNFTg model of RA (Tg197).

Materials and methods Global gene expression of aqueous extracts from synovial joints of untreated and treated Tg197 mice were compared with wild-type controls in 10 biological replicates for each condition. Profiles were analysed with ANOVA, coupled with Dunnett’s multiple comparison test using wild-type as control, giving rise to robust lists of differentially expressed genes.

Results Preliminary analysis of untreated samples resulted in the definition of a broad subset of >850 differentially expressed genes, which we used to define a disease signature at both gene and functional levels. Functions linked to this gene subset included both known (TNF and NF-kB signalling, Inflammatory response, RA and Inflammatory Bowel Disease) and less obvious pathways (Fatty Acid Metabolism, Oxidative Phosporylation, Autophagy). Functional dissection also revealed interesting links to various types of cardiomypathy phenotypes, a common comorbidity of RA patients. We found that etanercept, adalimumab and certolizumab pegol shared extensive similarities in the way they restore expression of disease signature genes, while infliximab diversified exhibiting increased efficiency in the modulation of over-expressed genes. The latter appears to predominantly modify genes of inflammatory and cytokine-related functions while the former show particular efficacy for genes associated with various types of cardiomypathy phenotypes.

Conclusions Combined, our analyses reveal modest, yet detailed and significant discrepancies between gene expression of different anti-TNF agents and may provide a valuable framework for the benchmarking of novel biosimilars or the repositioning of existing substances at a preclinical stage.