Super-resolution imaging with deep learning-based segmentation for detailed characterization of mitochondrial arrangement in Pompe disease skeletal muscle

Abstract

Pompe disease (glycogen storage disease type II) is an autosomal recessive lysosomal storage disorder caused by acid -glucosidase deficiency, leading to lysosomal glycogen accumulation, autophagosome buildup, and defective autophagic flux. Mitochondrial abnormalities, traditionally described by electron microscopy as paracrystalline inclusions, cristae loss, and swollen structures, have long been considered secondary consequences of lysosomal dysfunction. However, the spatial organization, organelle crosstalk, and fiber-type-specific progression of these alterations in skeletal muscle remain poorly understood. We combined super-resolution image scanning microscopy with deep learning-based segmentation to quantitatively assess mitochondrial and lysosomal remodeling, and their direct physical interactions, in Pompe disease (Gaa-/-) mice. Organelles were analyzed at pre-symptomatic (1 month) and symptomatic (4 months) stages across different muscle types (Soleus, Gastrocnemius), fiber types, and subcellular regions (subsarcolemmal, intermyofibrillar). Mitochondrial network structure was altered as early as 1 month of age, whereas density changes became widespread at 4 months in a fiber-type- and region-dependent manner. Alongside early lysosomal enlargement and late-stage spatial clustering, we observed a progressive, region-specific increase in mitochondria-lysosome contacts, most pronounced in the intermyofibrillar region. This quantitative framework provides a powerful tool for monitoring pathophysiology and evaluating therapeutic interventions in Pompe disease.