Background and objectives Flow cytometry offers quantification of multidimensional characteristics at single cell level for millions of cells. Multiplex flow cytometry or mass cytometry enable to screen for dozens of antigens on a single cell. Conventional analysis of such data requires user defined gating and is time consuming. Using the new bioinformatics tools immunoClust for automated and user-independent analysis, we investigated the complexity of phenotypic changes of immune cells upon migration from peripheral blood (PB) to synovial fluid (SF) in rheumatoid arthritis (RA).
Materials and methods Seven paired samples of PB and SF from RA patients were stained in 10 different antibody cocktails and data investigated by the newly developed immunoClust pipeline for sample specific populations and differences between PB and SF. Population clustering and comparative meta-clustering assume finite mixture models and use Expectation Maximisation (EM)-iterations with integrated classification likelihood (ICL) criterion to stabilise the number of reasonable clusters. For meta-clustering, a probability measure on Gaussian distributions was invented, which is based on the Bhattacharyya Coefficients. Meta-clusters were manually annotated and classified. The clustering tools of immunoClust are available as open source R-package in Bioconductor.
Results Automated clustering with 46 different surface markers detected all major leukocyte subsets and several activation markers in PB and SF samples including neutrophils, eosinophils, T-cells and sub-populations, monocytes, B-cells, NK-cells and dendritic cells. The comparison revealed about 10 highly significant changes per staining cocktail. For example the percentage of monocytes/macrophages was doubled in SF and dominated by CD16+ cells, the frequencies of effector/memory subpopulations of lymphocytes were increased and naïve T-cells and B-cells were almost completely absent. In addition several unexpected populations like CCR7+ monocytes were found in SF only.
Conclusion In conclusion, the results give a reasonable starting point to face the next field of research for marker detection and prediction analysis. The data will be further exploited for changes in cell activation and differentiation in SF in order to screen for these populations also in PB. This approach is not only applicable to fluorescence-based flow data but could be also used for multi-parametric data sets generated by mass spectrometry-based cytometry (CyTOF).
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.