Abstract
Astrocytes are key regulators of sleep and neuroinflammatory responses. However, the relationship between objective sleep parameters and astrocytic fluid biomarkers in cognitively unimpaired individuals remains unclear. We examined how sleep architecture relates to astrocytic, neuroaxonal and Alzheimer's disease-related fluid biomarkers in cognitively unimpaired adults and whether age, sex and APOE ε4 moderate these associations. This cross-sectional study included 51 cognitively unimpaired participants from the Sant Pau Initiative on Neurodegeneration cohort. One-night in-lab polysomnography was used to quantify sleep architecture, fragmentation, slow-wave activity and respiratory parameters. CSF biomarkers included glial fibrillary acidic protein (GFAP), chitinase-like-3 protein 1 (YKL-40), Aβ42, Aβ40, pTau181 and tTau; plasma biomarkers included GFAP and neurofilament light chain (NfL). Associations were analysed using Spearman correlations, multiple linear regression, and moderation models, adjusting for age, sex, body mass index, APOE ε4 status and sleep apnoea. Lighter and more fragmented sleep, characterized by longer N1 duration, increased wake after sleep onset, frequent stage transitions and elevated cortical arousal, was associated with higher CSF YKL-40, Aβ40, pTau181 and tTau (ρ = 0.32-0.62, all P < 0.05). In contrast, deeper, more consolidated sleep, indicated by longer total time of sleep, greater N3 duration and higher slow-wave activity, was associated with lower CSF GFAP and YKL-40 (ρ = -0.35 to -0.44, all P < 0.05). These associations remained significant in adjusted regression models. Plasma GFAP and NfL exhibited an inverse profile, with positive associations with deeper sleep (β: 0.16-0.18, P < 0.05) and negative associations with lighter sleep stages (β: -0.23 to -0.29, P < 0.01). Rapid eye movement (REM) sleep was also associated with astrocytic fluid biomarkers, with negative correlations for CSF and plasma GFAP (ρ = -0.49 and ρ = -0.28, respectively, all P < 0.05), while in regression models, REM duration remained a negative predictor of plasma GFAP (β = -0.23, P = 0.003) and a positive predictor of CSF YKL-40 (β = 0.12, P = 0.037). Notably, APOE ε4 consistently moderated associations between sleep and CSF YKL-40 and GFAP, while age and sex influenced plasma GFAP and CSF YKL-40, respectively (all P < 0.05). In cognitively unimpaired adults, sleep architecture is differentially associated with central and peripheral biomarkers of astrocytic activation, neuroaxonal integrity and Alzheimer's disease-related proteins. These findings support the importance of considering sleep as a key factor in the early pathophysiology of neurodegenerative disease.