Single-Cell Behavioromics Enables Fate Prediction of Individual Hematopoietic Stem Cells

Abstract

Achieving near-single-cell precision in the prospective identification of long-term engrafting hematopoietic stem cells (LT-HSCs) remains challenging, likely because current static, point-in-time single-cell phenotyping technologies do not capture the dynamic nature of cell fate decisions. Here, we overcome this barrier by longitudinally tracking individual hematopoietic stem and progenitor cells (HSPCs) in ex vivo niche cultures using live imaging, coupled with downstream immunophenotyping, single-cell transcriptomics, and transplant assays. Deep learning analysis of live imaging data achieved 98.5% average precision in predicting serially-engrafting LT-HSC (serial-LT-HSC) identity from the behaviorome (the totality of observed behaviors) of each cell, outperforming currently available technologies. Serial-LT-HSCs exhibited lower motility among HSPC populations, which was associated with reduced Activin signaling activity and elevated expression of cell adhesion genes. Correspondingly, antagonizing Activin signaling by Inhibin-A treatment in ex vivo niche cultures suppressed HSPC motility while increasing the frequency of post-expansion serial-LT-HSCs, revealing a novel role for Activin signaling in HSPC behavior and fate determination. Collectively, this study establishes a broadly applicable framework that connects behavioromic analysis and single-cell transcriptomics to enable near-single-cell precision in predicting stem cell fate and uncover molecular mechanisms governing behavior-fate relationships.