Background It is well-established that reducing the inflammation of rheumatoid arthritis (RA) early leads to improved patient outcomes. Patients who fail to respond to conventional DMARDs move on to biologic drugs, however 30% will fail to respond to the first drug tried. Biologics are trialled for at least 3 months before their efficacy can be assessed, during which time irreversible joint damage may occur.
Currently there is no way to predict which drug will be effective in individual patients. It is therefore a research priority to develop a method to predict response to each class of biologic drug. I hypothesise that the immune phenotype of a patient will influence treatment response.
Objectives This study aims to immunophenotype T cells using mass cytometry and to test novel unbiased methods of analysing high-dimensional data.
Methods Ten healthy controls (HC) and 10 RA patients with established disease were included. T cells were isolated and stimulated for 4 hours using anti-CD3/anti-CD28 beads in the presence of monensin and brefeldin A. Cells were stained with a 37-channel mass cytometry panel including surface markers, intracellular antigens and transcription factors.
Analysis was performed by conventional biaxial gating alongside tools available on the MRC Cytobank platform, namely Visual t-distributed stochastic neighbour embedding (viSNE), and Cluster identification, characterization, and regression (CITRUS). The CITRUS analysis compared stimulated HC with RA T cells. To compare abundances of cell clusters, the association model Prediction Analysis for Microarrays (PAMR) was applied with a minimum cluster size of 2.2%, 5 cross-validation folds and a false discovery rate of 1%.
Results Conventional biaxial gating showed that there were large differences in the proportions of both IFNγ+ CD4 T cells (1–15%) and IL-17A+ CD4 T cells (0–9%) within and between RA and HC. The proportion of IFNγ+ CD4 T cells (Th1) did not correlate with that of IL-17A + T cells (Th17), leading to the generation of 4 immunophenotypes: “Th1”, “Th17”, “double-hi”, and “low”, determined by cytokine expression.
Using CITRUS we identified 3 clusters of cells which differed significantly in abundance between HC and RA. Cluster 1 was CD4+CD38+, had characteristics of regulatory T cells and was more abundant in HC. Cluster 2 was CD28lowCD8+ expressing perforin and Tbet, and cluster 3 a CD127highCCR6+population, both of which were more abundant in RA.
Conclusions Cytokine expression of ex vivo stimulated T cells from RA and HC is highly variable and can be detected using mass cytometry. Cytokine signatures of this kind may be informative when predicting treatment response.
CITRUS identified 3 cell clusters which may have been missed using conventional methods of analysis. Importantly, CITRUS allowed inspection of the phenotype of each cluster.
Our next step is to compare RA responders to non-responders using the methods described. The function of these clusters will be further investigated by cell isolation with fluorescence-activated cell sorting (FACS) and may go some way in predicting treatment response. Finally we recommend the use of both automated clustering algorithms alongside conventional gating methods when analysing high-dimensional data.
Disclosure of Interest None declared