Machine learning-guided reconstruction of cytoskeleton network from live-cell AFM images.

Area of Science:

• Cell Biology
• Biophysics
• Microscopy

Background:

• Understanding the dynamic reorganization of actin filaments (F-actins) in motile cells at the individual filament level remains a challenge.
• High-speed atomic force microscopy (HS-AFM) offers potential for live imaging of intracellular F-actin dynamics, but image noise and low resolution hinder detailed analysis.

Purpose of the Study:

• To develop a novel machine learning method for quantitative recognition and analysis of individual F-actin filaments in HS-AFM images.
• To elucidate the structural dynamics and organization of F-actin networks within living cells, specifically in lamellipodia and the cell cortex.

Main Methods:

• Development of a machine learning algorithm to improve resolution and estimate F-actin orientation from HS-AFM images.
• Application of the developed method to analyze F-actin organization in the lamellipodia and cell cortex of motile cells.

Main Results:

• F-actin filaments in lamellipodia were found to be oriented at ±35° toward the membrane, consistent with Arp2/3 complex-induced branching.
• Analysis of the cell cortex revealed non-random F-actin orientation at four specific angles, suggesting a novel organizational mechanism.

Conclusions:

• The developed machine learning method significantly enhances the ability to recognize and analyze individual F-actin filaments, overcoming limitations of HS-AFM.
• The findings provide new insights into F-actin organization mechanisms in different cellular compartments, advancing our understanding of cytoskeletal dynamics.