Abstract
Large-scale brain networks are vulnerable to change with aging and become dysregulated. How these networks are altered at the cellular level remains unclear owing to challenges of bridging data across scales. Here, we integrate in vivo cortical similarity networks with whole brain spatial transcriptomics to characterize the aging brain in a lifespan cohort of macaques (N=64, ages 1-26 years). Deep-layer excitatory neurons and oligodendrocytes emerged as dominant correlates of cortical similarity, linking infragranular cell type composition to macroscopic network structure. Age-related declines in network strength were most pronounced in transmodal networks, including default mode and limbic, and aligned with regions enriched in inhibitory and glial cell types. Parvalbumin-enriched chandelier cells showed the strongest association with regional vulnerability, suggesting a role in network disconnection. Cell-type enrichment was conserved across species, with both human and macaque transcriptomic data aligning with the cortical functional hierarchy. These findings uncover a cellular basis for cortical network aging and highlight the value of imaging-transcriptomic integration across scales.