Computational Analysis of Multiparametric Flow Cytometric Data to Dissect B Cell Subsets in Vaccine Studies.

The generation of the B cell response upon vaccination is characterized by the induction of different functional and phenotypic subpopulations and is strongly dependent on the vaccine formulation, including the adjuvant used. Here, we have profiled the different B cell subsets elicited upon vaccination, using machine learning methods for interpreting high-dimensional flow cytometry data sets. The B cell response elicited by an adjuvanted vaccine formulation, compared to the antigen alone, was characterized using two automated methods based on clustering (FlowSOM) and dimensional reduction (t-SNE) approaches. The clustering method identified, based on multiple marker expression, different B cell populations, including plasmablasts, plasma cells, germinal center B cells and their subsets, while this profiling was more difficult with t-SNE analysis. When undefined phenotypes were detected, their characterization could be improved by integrating the t-SNE spatial visualization of cells with the FlowSOM clusters. The frequency of some cellular subsets, in particular plasma cells, was significantly higher in lymph nodes of mice primed with the adjuvanted formulation compared to antigen alone. Thanks to this automatic data analysis it was possible to identify, in an unbiased way, different B cell populations and also intermediate stages of cell differentiation elicited by immunization, thus providing a signature of B cell recall response that can be hardly obtained with the classical bidimensional gating analysis.