Abstract: Cell differentiation, a complex process transforming stem/progenitor cells into diverse mature cell types, is driven by intra- and extracellular stimuli and orchestrated by dynamic gene expression changes mediated by key transcription factors. Most biomarkers for identifying differentiation states are transcription factors, rather than surface markers. This limits the ability to sort cells using conventional fluorescence-activated cell sorting (FACS), especially in development and regenerative medicine. Even using intracellular proteins as biomarkers, it is often difficult to identify the cellular state while keeping them alive for further downstream analysis.
Ghost cytometry (GC) is a novel flow cytometry technique that analyzes high-content optical signals that reflect cellular morphology using a static patterned light pattern and a single-pixel detector. By applying machine learning directly to these signals, GC enables label-free cell analysis and sorting based on intrinsic properties.
In this study, we induced human iPSCs(201B7) differentiation towards hepatocytes using a monolayer differentiation protocol and harvested them at Day 3 (definitive endoderm), Day 7 (hepatoblast), and Day 15 (hepatocyte). All samples including undifferentiated iPSCs were mixed and measured on GC, where morphological features were analyzed and visualized via Uniform Manifold Approximation and Projection (UMAP) of the GC data. By applying the UMAP model to individual samples, we identified four cell subtypes that were distributed on different locations on UMAP. Based on their distribution, three subpopulations were defined and gated, presumed to represent the endoderm, hepatoblasts, and hepatocytes. Cells from each gate were sorted and enrichment of each cell type with the expression of marker genes, SOX17 and FOXA2 for endoderm, AFP, HNF4A, ALB and SERPINA1 for hepatoblast and hepatocyte was confirmed. These results demonstrate that GC effectively captures morphological transitions during hepatocyte differentiation from iPSCs, enabling assessment and sorting of transitional states without the need for external labels. This concept can be extended to various other differentiation processes to develop label-free workflows for cell monitoring and sorting in the field of regenerative medicine.
