Supervised Machine Learning for Semi-Quantification of Extracellular DNA in Glomerulonephritis

Area of Science:

• Nephrology
• Immunology
• Biotechnology

Background:

• Glomerular cell death is key in myeloperoxidase anti-neutrophil cytoplasmic antibody-associated vasculitis (MPO-AAV).
• Extracellular deoxyribonucleic acid (ecDNA) is released during various cell death pathways, but its measurement in kidney tissue is challenging.
• Current methods for ecDNA analysis are time-consuming and often lack organ-specific insights.

Purpose of the Study:

• To develop and validate a protocol for quantifying ecDNA in formalin-fixed paraffin-embedded (FFPE) kidney tissue.
• To utilize machine learning for objective analysis and differentiation of ecDNA levels in healthy and diseased renal tissue.
• To demonstrate the adaptability of the method for identifying neutrophil extracellular traps (NETs) and extracellular myeloperoxidase (ecMPO) in MPO-AAV models.

Main Methods:

• A protocol for staining ecDNA in FFPE human and murine kidney tissue was established.
• Autofluorescence was quenched, and ecDNA was quantified using the trainable Weka segmentation tool in ImageJ.
• Machine learning was trained to classify ecDNA within glomeruli and applied to experimental murine anti-MPO glomerulonephritis models.

Main Results:

• The developed protocol successfully stained and quantified ecDNA in FFPE kidney tissue.
• Trainable Weka segmentation objectively distinguished ecDNA levels between healthy and diseased kidney tissues.
• The method effectively identified NETs and ecMPO in murine models of anti-MPO glomerulonephritis.

Conclusions:

• This protocol offers an objective and adaptable method for analyzing ecDNA in FFPE kidney tissue.
• The machine learning approach streamlines the quantification of ecDNA, NETs, and ecMPO in renal pathology.
• This technique has potential applications for studying MPO-AAV and other diseases involving ecDNA in various organs.