Predicting best treatment in rheumatoid arthritis

Abstract

BackgroundAlthough targeted biological treatments have transformed the outlook for patients with rheumatoid arthritis (RA), 40% of patients show poor clinical response, and there is an imperative to unravel the molecular pathways and mechanisms underlying non-response and disease progression. 5–20% of RA individuals do not respond to all current medications including biologic and targeted therapies, which suggests that distinct pathogenic processes underlie multi-drug refractoriness. ObjectivesIn this brief review we discuss advances from recent studies in precision medicine in rheumatoid arthritis. MethodsBulk RNA-Sequencing of synovial biopsies from RA individuals combined with histology and deep clinical phenotyping has revealed substantial insights into divergent pathogenic pathways which lead to disease progression and illuminated mechanisms underlying failure to response to specific treatments. Biopsy-driven randomised controlled trials, such as R4RA and the forthcoming STRAP trial, have enabled the development of machine learning predictive models for predicting response to different therapies. ResultsIn the Pathobiology of Early Arthritis Cohort (PEAC), gene expression analysis showed that individuals could be classified into three gene expression subgroups which correlated with histopathotypes defined by histological markers: pauci-immune fibroid pathotype characterised by fibroblasts and an absence of immune inflammatory cells; diffuse-myeloid pathotype characterised by macrophage influx; and the lympho-myeloid pathotype delineated by the presence of B cells, but typically containing a complex inflammatory infiltrate with ectopic lymphoid structure formation. In the R4RA biopsy-driven randomised controlled trial, patients were randomised to either rituximab or tocilizumab. Comprehensive analysis of synovial biopsies pre/post-treatment identified gene signatures of response associated with pathogenic pathways which could be tracked over time. A group of true refractory patients were identified who had failed anti-TNF prior to the study (it was an entry criterion) and then subsequently failed both trial biologics during the trial. RNA-Seq analysis and digital spatial profiling identified specific cell types including DKK3+ fibroblasts as being associated with the refractory state. We identified machine learning predictive models based on specific gene signatures which were able to predict future response to therapy as well as the refractory state. ConclusionsRNA-sequencing of synovial biopsies has enabled substantial progress in understanding disease endotypes in RA and identifying synovial gene signatures which predict prognosis and future response to treatment.