Abstract 2530: Advancing ICK assay: Real-time, label-free imaging of lymphocyte subsets

Abstract

Abstract Background: Immune Cell Killing (ICK) Assay is pivotal in immunology and clinical research, uncovering the intricate connections between the immune system and diseases. Traditional ICK assays utilizing antibody staining for endpoint data generation pose challenges in capturing dynamic interactions. Moreover, antibody labeling for data quantification may compromise sensitivity, especially with low-affinity antibodies or a limited number of cells. The anticipation of label-free lymphocyte subset classification and real-time effector/target cell interaction analysis in one field of view is expected to provide more valuable insights than traditional ICK assays. Previous studies achieved 70-80% accuracy in image-based lymphocyte subset classification using digital holographic microscopy, light scattering, etc. but the cells’ condition during the cell sorting or sample preparation process may affect the lymphocyte subset status, resulting in real morphological discrepancies. Our approach focuses on directly sorting human PBMCs to enable label-free, dynamic interaction analysis of natural immune system responses in one field of view. Method: In this study, PBMC samples from 6 individuals were separated using standard-density gradient centrifugation and labeled with antibodies. Data were acquired using the Holotomographic Microscope HT-X1 (Tomocube Inc.). We employed Densenet 121 as the deep learning model architecture, considering the trade-off between memory consumption and performance. The cross-entropy loss function was utilized, and the AdamW optimizer was chosen for model parameter optimization. Input data consisted of 3D refractive index (RI) data for individual cells, each of sized (20, 54, 54). Individual cell labeling was performed using staining information. Results: An overall accuracy of 93.75% in Human PBMC subtype cell classification was achieved. In stage 1, an accuracy of 97.56% was demonstrated in classifying [CD14, CD15, Others]. Among the cells classified as "Others" in stage 2, [CD3, CD19, CD16&CD56] displayed an accuracy of 93.5%. In stage 3, classifying [CD4, CD8] within the CD3 subtype achieved an accuracy of 90.2%. This classification is presumably based on pattern recognition of the chromosomal landscape in holotomography images. Conclusion: Our research has successfully demonstrated the feasibility of real-time label-free lymphocyte subset classification. This approach enabled the classification of effector cells as live cells in ICK assay and the analysis of effector-target cell dynamic interactions. Further training in the classification of more detailed subtype cells and various cell death types will provide additional insights. The advancement of this technology is expected to play a pivotal role in disease prevention, treatment, vaccine development, medical research, and academic studies. Citation Format: Sanggeun Oh, Jaephil Do, Hyun-Seok Min, Dongmin Ryu, Wei sun Park. Advancing ICK assay: Real-time, label-free imaging of lymphocyte subsets [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2530.