- The paper demonstrates significant inverted U-shaped non-linear associations between CSF Aβ levels and connectivity in the precuneus and ventral DMN.
- The paper employs 3T resting-state fMRI, quadratic regression modeling, and comprehensive neuropsychological testing to assess network-specific vulnerabilities.
- The paper finds that higher connectivity in Aβ-sensitive networks is robustly associated with superior performance in visual memory, visuospatial ability, and executive function.
Non-linear Associations of Amyloid-β with Resting-State Functional Networks and Cognitive Relevance in Cognitively Normal Older Adults
Introduction
This study investigates the non-linear relationships between cerebrospinal fluid (CSF) amyloid-β (Aβ) levels and resting-state functional connectivity across large-scale brain networks in a substantial, community-based cohort of cognitively normal older adults. The work addresses key limitations in prior research, notably small sample sizes and selection biases inherent to memory clinic-based recruitment, by leveraging a sample of 968 participants from the Emory Healthy Brain Study. The analysis focuses on the early, asymptomatic phase of Alzheimer's disease (AD), aiming to elucidate network-specific vulnerabilities to Aβ pathology and their cognitive correlates.
Methodological Overview
Cohort and Data Acquisition
Participants (N=968, median age 63.8 years, 66.2% female) underwent 3T resting-state fMRI, CSF biomarker quantification (Aβ1-42, total tau, phosphorylated tau 181), and comprehensive neuropsychological testing. Functional networks were delineated using group independent component analysis (ICA) with spatial constraints, yielding 14 canonical networks, including dorsal and ventral DMN, precuneus, salience, executive control, and visual networks.
Statistical Analysis
Associations between CSF biomarkers and network connectivity were modeled using multiple regression with quadratic terms to capture potential biphasic (inverted U-shaped) relationships, controlling for age and sex. Cognitive relevance was assessed via regression of neuropsychological scores on network connectivity, with FDR correction for multiple comparisons.
Key Findings
Non-linear Aβ-Connectivity Associations
The principal finding is the presence of significant inverted U-shaped associations between CSF Aβ and functional connectivity in the precuneus network and ventral DMN. Both linear and quadratic terms for Aβ were significant in the precuneus, while only the quadratic term was significant in the ventral DMN. No such non-linear association was observed in the dorsal DMN or other networks. This pattern indicates network-specific vulnerability, with the precuneus and ventral DMN exhibiting early compensatory hyperconnectivity at intermediate Aβ levels, followed by hypoconnectivity as pathology advances.
Absence of Tau-Connectivity Associations
No significant relationships were detected between CSF tau (total or phosphorylated) and functional connectivity in any network. This supports the hypothesis that tau-related network disruptions manifest later in the disease cascade, or that CSF tau is less sensitive to early network-level changes compared to Aβ.
Cognitive Correlates of Network Connectivity
Higher connectivity in Aβ-sensitive networks (precuneus, dorsal and ventral DMN, posterior salience) was robustly associated with superior performance in visual memory (RCFT immediate and delayed recall), visuospatial ability (JoLO), and executive function (TMTB). Notably, these associations were strongest in the precuneus and dorsal DMN, with additional contributions from the ventral DMN and posterior salience network. These results demonstrate that even subtle, preclinical network alterations have measurable cognitive relevance.
Implications
Theoretical Implications
The findings refine the understanding of the temporal and spatial dynamics of AD pathophysiology. The demonstration of network-specific, non-linear Aβ effects—particularly in the precuneus and ventral DMN—supports models positing early compensatory hyperconnectivity preceding network failure. The lack of dorsal DMN involvement suggests functional and metabolic heterogeneity within the DMN, underscoring the necessity of subsystem-level analyses in future research.
The absence of tau-connectivity associations in this asymptomatic cohort aligns with the view that tau-mediated network disruption is a downstream event, potentially requiring tau PET imaging for greater sensitivity in preclinical populations.
Practical Implications
The study provides evidence that resting-state fMRI-derived network connectivity metrics, particularly in the precuneus and ventral DMN, may serve as sensitive, non-invasive biomarkers for early Aβ pathology and cognitive risk stratification. These imaging markers could be integrated into multi-modal screening pipelines for preclinical AD, enhancing early detection and facilitating targeted intervention trials.
Limitations and Future Directions
The cross-sectional design precludes causal inference regarding the temporal evolution of network changes. The overrepresentation of female participants may limit generalizability. Longitudinal follow-up, currently underway, will be critical to delineate the progression of network alterations and their predictive value for cognitive decline. Incorporation of tau PET and plasma biomarkers may further clarify the dissociation between Aβ and tau effects on network integrity.
Conclusion
This study demonstrates that non-linear, network-specific alterations in functional connectivity are detectable in cognitively normal older adults with early Aβ pathology, particularly in the precuneus and ventral DMN. These changes are cognitively relevant, supporting the utility of functional network metrics as early imaging biomarkers for AD. The results highlight the importance of large, community-based cohorts and advanced network modeling in elucidating the preclinical stages of AD and inform future biomarker development and early intervention strategies.