Flowct: A Semi-Automated Workflow for Deconvolution of Immunophenotypic Data and Objective Reporting on Large Datasets

Background: The advent of immunotherapy renewed the interest in immune monitoring to identify determinants of treatment response. Flow cytometry is widely adopted in immunotherapy-based clinical trials, but manual analysis of multiparameter files poses a challenge to capture full cellular diversity and to provide unbiased reporting in large datasets.

Methods: Here, we developed a semi-automated pipeline named "FlowCT" which, starting from compensated data obtained with standardized protocols, allows simultaneous analyses of multiple files and automated cell clustering. FlowCT starts with quality control and data normalization followed by an analytical stage with clustering algorithms, dimensional reduction techniques and cluster identification based on antigen expression. Statistical tools are included for immediate analysis of results.

Results: As proof-of-concept, we used FlowCT in three different datasets. First, we applied FlowCT to bone marrow (BM) samples from three multiple myeloma (MM) patients stained with 17-color flow cytometry, to determine the increment in the complexity of analyzing 8 and 17 markers, chosen to characterize T cells. Of note, a single combination of CD3, CD4, CD8, CD45RA, CD56, CCR7, PD1 and TIGIT, allowed the identification of 31 lymphocyte subsets using FlowCT, which increased to 39 different clusters with 17 markers and unveiled a novel population of CD3^- CD56^- CD8⁺ CD16⁺ lymphoid cells in the MM immune microenvironment. Secondly, we applied FlowCT to matched peripheral blood (PB) and BM samples from 10 patients with smoldering MM, to objectively assess if PB represents a good surrogate of T-cell distribution in the BM. Using an 8-color combination to characterize CD4 T cells, up to 26 different subsets were identified, including several CD4 T helper (Th) type subsets. Of note, their distribution within PB CD4 T cells was similar to that found in BM, except for CD4 T CXCR3⁺CCR4⁺ effector memory and Th17 central memory subsets that decreased in the BM tumor immune microenvironment.

Thirdly, we analyzed 30 BM samples from 10 MM patients studied every year during maintenance therapy, monitored with CD4, CD8, CD25, CD45RA, CD127, CCR7, PD1, and TCRγδ to characterize T cells. FlowCT identified 29 different T-cell populations, including 9 CD4 subsets, 14 CD8 subsets, 4 Tγδ cell subsets and 2 distinct Treg subsets. Longitudinal, semi-automated and unbiased analysis unveiled a significant fluctuation of CD4 naïve and transitional memory cells during maintenance, as well as a significant decrease of CD8 CD127^- effector memory and transitional effectors cells after 2 years of maintenance.

Conclusions: Here, we presented FlowCT, a pipeline optimized for the analysis of large flow cytometry datasets that could be easily implemented by research laboratories to unveil full cellular diversity, singular patterns of antigen expression, and to provide unbiased reporting in large studies, like clinical trials.