Autofluorescence lifetime imaging to identify cancer stem cells from bulk cancer cells

Cancer Stem Cells (CSCs) can drive drug resistance, tumor recurrence, and metastasis. CSCs can be difficult to identify due to their small proportion of tumor and need for antibody-based labeling techniques. Here, autofluorescence lifetime imaging of the metabolic coenzymes NADH and FAD was combined with multivariate analysis and machine learning classification models to develop a label-free imaging method to accurately identify CSCs and bulk tumor cells. The HCT116 colorectal cancer cell line was segregated into 4 distinct populations: Bulk tumor cells, Cancer stem cells (CSCs) identified by LGR5-positive expression, and two variable groups (V1 and V2) representing mixtures of the former two cell types. NADH and FAD fluorescence lifetime images were acquired for each group using a multiphoton fluorescence lifetime microscope. The FLIM features including the short and long fluorescence lifetimes, amplitude ratios, weighted average fluorescence lifetime and optical redox ratio were extracted to characterize the metabolic signatures of CSCs and bulk cancer cells. Random forest models identified cells as CSCs or bulk cancer cells from the autofluorescence features with high accuracy.