Biomarkers

Altered resting-state functional connectivity (RSFC) has been reported in early Alzheimer's disease (AD). Graph metrics derived from RSFC networks provide valuable insights into brain organization. However, their potential in characterizing early network dysfunction and their relationship with AD biomarkers and cognitive performance remains understudied.

Using RSFC data from 326 cognitively unimpaired (CU) individuals in the ALFA cohort (mean age=60.8, SD=4.74), we analyzed graph metrics in relation to cerebrospinal fluid (CSF) biomarkers. CSF Aβ42 and Aβ40 were assessed with the NeuroToolKit, a panel of exploratory robust prototype assays, while p-tau181 was measured with the Elecsys® Phospho-Tau (181P) CSF immunoassay (both Roche Diagnostics International Ltd). Interactions with age and sex were further inspected. RSFC networks were computed from 218 regions (Brainnetome atlas) using CONN, and thresholded at a density of 35%. The following graph metrics were extracted: average path length (APL), Local Efficiency (LE), Betweenness Centrality (BC), Closeness Centrality (CC), and Strength. Linear regression models assessed associations between CSF biomarkers and graph metrics, adjusting for age, sex, years of education, and APOE-ε4. Interactions between CSF biomarkers and graph metrics were analyzed for their impact on longitudinal cognitive measures (Preclinical Alzheimer Cognitive Composite [PACC]; mean follow-up=3.35 y, SD=0.53).

We found a positive main effect of CSF Aβ42/40 (p = 0.011) and p-tau181 (p = 0.012) on BC of the Dorsal Attention Network (DAN) (Figure 1). Significant interactions between Aβ42/40 and sex were observed on the APL, CC and Strength of the Default Mode Network (DMN) (Figure 2). Finally, significant interactions were observed between Aβ42/40 and graph metrics of the DMN- APL (p = 0.005), LE (p = 0.009), and CC (p = 0.008)- on PACC changes (Figure 3).

Our data suggest that, in CU individuals, soluble Ab and p-tau exert opposite effects in the DAN information flow. Moreover, our interaction models suggest that a lower integration between DMN and the rest of the brain, as well as a lower centrality of DMN regions might be beneficial to preserve cognitive performance in the presence of AD. This study highlights the role of network topology in early AD and its potential to support cognitive resilience, providing potential targets for intervention.