Artificial Intelligence-Based Counting Algorithm Enables Accurate and Detailed Analysis of the Broad Spectrum of Spot Morphologies Observed in Antigen-Specific B-Cell ELISPOT and FluoroSpot Assays.

Abstract

Antigen-specific B-cell ELISPOT and multicolor FluoroSpot assays, in which the membrane-bound antigen itself serves as the capture reagent for the antibodies that B cells secrete, inherently result in a broad range of spot sizes and intensities. The diversity of secretory footprint morphologies reflects the polyclonal nature of the antigen-specific B cell repertoire, with individual antibody-secreting B cells in the test sample differing in their affinity for the antigen, fine epitope specificity, and activation/secretion kinetics. To account for these heterogeneous spot morphologies, and to eliminate the need for setting up subjective counting parameters well-by-well, CTL introduces here its cutting-edge deep learning-based IntelliCount™ algorithm within the ImmunoSpot® Studio Software Suite, which integrates CTL's proprietary deep neural network. Here, we report detailed analyses of spots with a broad range of morphologies that were challenging to analyze using standard parameter-based counting approaches. IntelliCount™, especially in conjunction with high dynamic range (HDR) imaging, permits the extraction of accurate, high-content information of such spots, as required for assessing the affinity distribution of an antigen-specific memory B-cell repertoire ex vivo. IntelliCount™ also extends the range in which the number of antibody-secreting B cells plated and spots detected follow a linear function; that is, in which the frequencies of antigen-specific B cells can be accurately established. Introducing high-content analysis of secretory footprints in B-cell ELISPOT/FluoroSpot assays, therefore, fundamentally enhances the depth in which an antigen-specific B-cell repertoire can be studied using freshlyisolated or cryopreserved primary cell material, such as peripheral blood mononuclear cells.