Deep Learning-Guided Holotomography Reveals Early Structural Remodelling During Pluripotency Exit.

Area of Science:

• Biotechnology
• Stem Cell Biology
• Regenerative Medicine

Background:

• Assessing human pluripotent stem cell (hPSC) quality is crucial for safe and reproducible regenerative medicine.
• Current methods for hPSC quality assessment are often invasive, labor-intensive, and operator-dependent.

Purpose of the Study:

• To develop a non-invasive, automated, and statistics-driven platform for real-time hPSC quality assessment.
• To integrate 3D refractive index imaging with deep learning for pluripotency evaluation.

Main Methods:

• Developed DeepHOPE (Deep-learning-guided Holotomography for Pluripotency Evaluation).
• Utilized 3D refractive index imaging and deep learning algorithms.
• Validated performance across various differentiation contexts (germ-layer, retinoic acid, reprogramming).

Main Results:

• DeepHOPE demonstrated robust performance in diverse hPSC culture conditions.
• The platform enabled streamlined cell production and improved differentiation efficiency via informed colony selection.
• Mechanistically, DeepHOPE detected subtle topological changes preceding pluripotency loss, linked to F-actin remodeling and cytoskeletal dynamics.

Conclusions:

• DeepHOPE offers a non-invasive, efficient method for real-time hPSC quality control.
• Cytoskeletal dynamics, particularly actomyosin tension, are identified as upstream regulators of early pluripotency exit.
• This technology can enhance reproducibility and safety in regenerative medicine applications.