Abstract
The process of reprogramming patient samples to human-induced pluripotent stem cells (iPSCs) is stochastic, asynchronous, and inefficient, leading to a heterogeneous population of cells. In this study, we track the reprogramming status of patient-derived erythroid progenitor cells (EPCs) at the single-cell level during reprogramming with label-free live-cell imaging of cellular metabolism and nuclear morphometry to identify high-quality iPSCs. EPCs isolated from human peripheral blood of three donors were used for our proof-of-principle study. We found distinct patterns of autofluorescence lifetime for the reduced form of nicotinamide adenine dinucleotide (phosphate) and flavin adenine dinucleotide during reprogramming. Random forest models classified iPSCs with ∼95% accuracy, which enabled the successful isolation of iPSC lines from reprogramming cultures. Reprogramming trajectories resolved at the single-cell level indicated significant reprogramming heterogeneity along different branches of cell states. This combination of micropatterning, autofluorescence imaging, and machine learning provides a unique, real-time, and nondestructive method to assess the quality of iPSCs in a biomanufacturing process, which could have downstream impacts in regenerative medicine, cell/gene therapy, and disease modeling.