High Throughput Identification Of Rare Cell Populations By Functional Phenoyping
Identification of rare cell types can enable early detection of disease and enhance the fundamental understanding of cellular processes. Most current approaches to identify rare cell populations rely on static phenotyping. Although static phenotyping can readily differentiate cells, these methods cannot fully uncover individual cells' dynamic heterogeneity. Cell-to-cell heterogeneity in metabolic profiles has recently been suggested as a critical indicator for their potential cell fate decision. The flexibility of metabolic reprogramming has been raised as a requirement for memory T cell formation, and a similar notion applies to the cancer cell metastasis. It suggests that metabolic profiling, which is part of functional phenotyping, can give a valuable insight into the disease cure and prevention. Such a metabolomics study is currently highly limited due to inadequate research tools. Recently, several approaches have been developed to enable functional phenotyping and now allow some basic single-cell level functional phenotyping. However, these approaches either lack throughput or selective recoverability or analytical and monitoring system. This thesis presents a microfluidic-based device that allows rare cell population identification by on-chip dynamic functional phenotyping and analysis, followed by highly selective recovery of cells for subsequent transcriptomic study. Using this system, we investigate a seemingly homogeneous static phenotypic cell population and de-cluster the subpopulation based on its distinctive dynamic functional phenotype. The high-throughput dynamic monitoring of the individual cell and functional probe's reactions is computationally analyzed using a machine learning technique. The subpopulations are subsequently sorted. Using a novel photoactivated selective recovery (PHASR) membrane, target cells are precisely retrieved, and single-cell level RNA sequencing is conducted. The individual transcriptomic profile is studied in parallel to the observed functional phenotype to investigate the relationship between metabolomics and transcriptomics. Kinetic modeling enables the demultiplexing of simple metabolic profiles into well-defined biological processes. With single-cell level RNA sequencing and high-throughput metabolic profiling, a valuable correlation is established between metabolomics and transcriptomics. Single-cell level heterogeneity in the metabolic profile can be related to the gene expression difference. This thesis shows a successful demonstration of capability for high throughput single-cell level functional phenotyping, computational kinetic modeling system that can decipher the observed functional phenotype to more meaningful biological reaction rates, high-throughput and high-selective recovery microfluidic system that allows single-cell RNA sequencing, and the establishment of the relationship between the transcriptome and functional phenotype. We believe this work can positively impact the fields of single-cell study, microfluidics, and image-based computational analysis.