Combinatory strategy using nanoscale proteomics and machine learning for T cell subtyping in peripheral blood of single multiple myeloma patients

Abstract

T cells play crucial roles in our immunity against hematological tumors by inducing sustained immune responses. Flow cytometry-based detection of a limited number of specific protein markers has been routinely applied for basic research and clinical investigation in this area. In this study, we combined flow cytometry with the simple integrated spintip-based proteomics technology (SISPROT) to characterize the proteome of primary T cell subtypes in the peripheral blood (PB) from single multiple myeloma (MM) patients. Taking advantage of the integrated high pH reversed-phase fractionation in the SISPROT device, the global proteomes of CD3+, CD4+ and CD8+ T cells were firstly profiled with a depth of >7 000 protein groups for each cell type. The sensitivity of single-shot proteomic analysis was dramatically improved by optimizing the SISPROT and data-dependent acquisition parameters for nanogram-level samples. Eight subtypes of T cells were sorted from about 4 mL PB of single MM patients, and the individual subtype-specific proteomes with coverage among 1 702 and 3 699 protein groups were obtained from as low as 70 ng and up to 500 ng of cell lysates. In addition, we developed a two-step machine learning-based subtyping strategy for proof-of-concept classifying eight T cell subtypes, independent of their cell numbers and individual differences. Our strategy demonstrates an easy-to-use proteomic analysis on immune cells with the potential to discover novel subtype-specific protein biomarkers from limited clinical samples in future large scale clinical studies.