Characterizing liver sinusoidal endothelial cell fenestrae on soft substrates upon AFM imaging and deep learning

Area of Science:

• Cell Biology
• Biophysics
• Medical Imaging

Background:

• Liver sinusoidal endothelial cells (LSECs) possess unique fenestrated structures crucial for liver function.
• Dysregulation of LSEC fenestrae is implicated in various liver diseases.
• Existing methods for fenestrae visualization and quantification are limited.

Purpose of the Study:

• To develop and validate an in situ imaging and analysis method for quantifying LSEC fenestrae.
• To assess the impact of different substrate stiffnesses on LSEC fenestration.
• To provide a quantitative tool for studying LSEC morphology in liver health and disease.

Main Methods:

• Primary mouse LSECs were cultured on collagen-I, PDMS, or PA hydrogel substrates.
• Atomic force microscopy (AFM) in contact mode was used for high-resolution imaging of LSEC fenestrae.
• A custom image recognition program based on fully convolutional networks (FCN) was developed for automated fenestra analysis.

Main Results:

• Optimized AFM scanning parameters enabled visualization of fenestrae across different substrates.
• The FCN-based program achieved 81.6% accuracy in recognizing LSEC fenestrae on soft substrates.
• Quantitative analysis of fenestrae number, size distribution, and cell porosity was achieved.

Conclusions:

• Combining AFM imaging with FCN analysis provides a robust method for quantifying LSEC fenestrae morphology.
• This technique offers global insights into cell surface structures, essential for understanding LSEC regulation and disease relevance.
• The developed method is applicable to LSECs cultured on various stiffness-varied substrates.