Identifying reliable tissue-level markers of biological aging has long been a bottleneck for both clinical assessment and longevity research. A structural feature visible in routine muscle biopsies — nuclear size — may now serve as a measurable, reproducible index of how much a muscle has biologically aged, with implications for diagnosing premature aging in inflammatory conditions and benchmarking anti-aging interventions.
Analyzing 974 hematoxylin-eosin whole-slide images from a population-level biobank, researchers deployed a dual-attention convolutional neural network alongside an independent Mask R-CNN segmentation model to quantify skeletal muscle nuclear morphology at scale. The classifier distinguished young from aged muscle tissue with an AUC of 0.91 and 86.2% accuracy. Nuclear diameter correlated with donor age at a Spearman's ρ of 0.71, a notably strong association for a purely structural measurement. Attention maps consistently flagged nuclear enlargement and spatial disorganization as the most discriminative features. Transcriptomic data linked to the same donors deepened the picture: muscles with larger nuclei were enriched for chromatin remodeling, proteostasis failure, cellular senescence signaling, and telomere dysregulation — all canonical aging hallmarks — while smaller nuclei associated with active DNA repair and anti-inflammatory programs. Critically, pediatric patients with inflammatory myopathies showed nuclear enlargement comparable to elderly muscle, suggesting inflammatory stress can accelerate histological aging independently of chronological age.
This work is methodologically significant for several reasons. Prior aging biomarkers in muscle have leaned heavily on functional measures (grip strength, fiber type ratios) or molecular assays requiring specialized tissue processing. Nuclear morphometry from standard H&E stains is vastly more accessible. The convergence of structural and transcriptomic signals around nuclear size adds biological credibility beyond a purely correlative finding. Key limitations remain: the analysis is cross-sectional, causality cannot be established, and generalizability across ethnic and disease populations requires validation. Nonetheless, the identification of a deep learning-quantifiable, histologically accessible marker that mirrors molecular aging programs represents a meaningful step toward scalable muscle-aging assessment in both research and clinical pathology contexts.
