Purpose

Alzheimer’s Disease (AD) is the most common form of dementia, characterized by memory loss, changes in behavior, and cognitive difficulties¹. The 3xTgAD mouse model produces Tau proteins in addition to amyloid precursor protein (APP) and presenilin-1 (PS1). These mice also begin showing memory deficiencies around 4 months (2 months before plaques and 8 months before Tau); therefore, structural alterations are potentially instituted prior to hallmark deposition². Clinically, MRI is used to diagnose AD by means of volumetrics, mainly focusing on hippocampal atrophy³.
In this study, Diffusion Tensor Imaging (DTI) and graph theory were implemented longitudinally in order to explore structural changes in connectivity across hemispheres as a function of age. Herein, 3xTgAD alterations were compared to data acquired previously using 5xFAD mouse models to verify the robustness of the technique.

Materials & Methods

Preserved ex vivo female mouse brains (4% PFA) expressing the 3xTgAD phenotype and wild type (WT) brains that span a corresponding range of ages were studied using DTI. Brains were harvested between 1-10 months and washed in PBS for 24 h prior to immersion in Fluorinert (3M, Corp) within a 10-mm glass NMR tube for scanning. Using an 11.75-T (500-MHz) magnet, DTI was implemented with a multi-slice, diffusion-weighted 2D spin-echo sequence that utilized 18 diffusion encoding directions (nominal b = 400 s/mm²) and four unweighted acquisitions for which Δ = 11 ms and δ = 3 ms with 15 averages. Brains were scanned individually with a TE/TR = 30/2000ms with a resolution of 100x100x500 µm and acquisition matrix size of 95x95 over a scan time of ~17 h. DTI datasets were analyzed using DSI Studio to reconstruct tracts⁴ using the following parameters: FA ≥ 0.1, Angular threshold ≤ 60^◦, Seeds ≤ 10⁶, min tract length = 1 mm, max tract length = 25 mm. Regions of interest (ROI) were selected as 14 equally spaced ellipsoids in the cortical regions and two ROI for each right and left side of the hippocampus. The segmented nodes were categorized into four neural regions: Piriform, Temporal, Parietal and Hippocampus and were split into left and right hemispheric regions. Tract counts were imported into Gephi for extracting graph properties including: Degree & Weighted degree, Clustering Coefficient, Closeness (Betweenness, Closeness, Eigenvector & Harmonic), and Eccentricity. One-way ANOVA and least significant difference post-hoc tests were used to determine statistical significance among samples (p<0.05).

Results & Discussion

Hemispheric and phenotypic differences are observed in the Piriform region for Betweenness Centrality (Figure 1), which quantifies the number of times a node acts as a bridge along the shortest path between two other nodes. In the 3xTgAD model, the right hemispheric Piriform region increases with age to a much greater extent than the left hemispheric Piriform region, which stays relatively constant. In WT, the opposite pattern is observed, the left hemispheric Piriform region decreases with age to a greater extent than the right hemispheric Piriform region. Notably, the scale differences across phenotype are drastically different, potentially due to synaptic degeneration.Temporal hemispheric and phenotypic differences are not observed for closeness centrality; however, a general decrease for the highest age is observed in both 3xTgAD and WT samples with no hemispheric differences with age.
Because Closeness Centrality is the sum of the shortest path lengths from one node to all other nodes, its decreasing trend with age may reflect the paring of neural connections during maturation; however, Closeness Centrality also may highlight differential time courses or connection remodeling for the AD versus WT specimens.
Clustering coefficient in the Temporal region and Closeness Centrality in the Piriform region shows a decrease in AD over age while WT samples tend to remain fairly constant, which could indicate a loss of direct nodal connections leading to a lengthening of the shortest path with respect to Closeness or neural pairing with respect to Clustering (Figures 2&3).
Hemispheric differences manifest in the Temporal and Piriform regions of the AD models when looking at Clustering Coefficient and Closeness Centrality (Figure 4&5). Here the hemispheric regions show large decreases over time when compared over the right hemispheric regions. It is important to note that decreases are occurring after the onset of plaques but before the presence of Tau, indicating that Tau may not be responsible for brain alterations.

Conclusion

Referencing parameters obtained via DTI to graph theory analysis, this method can detect the loss of connectivity and network efficiency (even with hemispheric differences) during progression of AD pathology, with possible extension to other neurodegenerative diseases. Certain metrics are more useful than others in determining phenotype and hemispheric differences. Future work will extend this study by incorporating male data to investigate sex differences. Additionally, this work will help to expand the application of DTI and network theory to identification and progression of other neurodegenerative diseases.

Acknowledgements

This work was supported by the User Collaborations Grant Program at the National High Magnetic Field Laboratory, which is supported by the National Science Foundation (DMR-1644779) and the State of Florida, and by the NIH (R01-NS102395 & R00-AG049090).