Introduction

Amyloid-β (Aβ) deposition is of great significance in Alzheimer’s disease (AD) and the further understanding of Aβ burden might lead to the successful therapeutic intervention.¹ Several studies have investigated the effect of amyloid on structural or functional brain changes by magnetic resonance imaging and positron emission tomography in healthy adults and patients with mild cognitive impairment (MCI) or AD.^2,3 However, little is known about how the alterations of structural-functional connectivity coupling relate to Aβ plaques. In this study, graph theory analysis was applied to investigate the small worldness respectively in structural and functional connectivities in MCI-Aβ positive, MCI-Aβ negative and healthy controls. Furthermore, structural and functional connectivities were combined to explore how these two modalities relate to each other in these three groups.

Methods

Thirty-three patients with MCI were recruited from the memory clinic of a university hospital and underwent ¹⁸F-flutametamol PET-CT scans. The dose of 185 MBq ¹⁸F-flutametamol was injected 90 min before image acquisition. Composite Z-score of each subject either above or below the threshold (uptake ratio-0.62) was used as a cut-off to determine the positivity and negativity of the amyloid scans.⁴ There are 21 Aβ negative (Age: 75.9±7.0 years old, Sex: 11F/10M) and 12 Aβ positive (Age: 74.6±7.2 years old, Sex: 7F/5M) subjects. Fourteen healthy controls (HC, Age: 52.3±16.5 years old, Sex: 7F/7M) were recruited from community centers. All subjects also underwent MRI examination using a 3T MR scanner (Philips, Achieva) with a 32-channel head coil. Structural images were acquired with 3D T1-weighted sequence using magnetization prepared rapid gradient-echo imaging (MPRAGE, TR=6.8 ms, TE=3.2 ms, thickness=1.2 mm, Flip angle=8°, FOV=256×240×204 (mm)). The diffusion-weighted images (DWI) were acquired with TR/TE of 3900/81 ms, 2.9× 2.9 × 3 mm³ voxel size, b-values of 0, 1000, 2000 s/mm² and 30 gradient sensitising directions. Resting-state functional images were collected by using a gradient-echo echo-planar sequence (parameters: TR=2000ms, TE=30ms, flip angle=90°, voxel size=1.6×1.6×4 𝑚𝑚³) sensitive to blood-oxygen-level-dependent (BOLD) contrast.
The preprocessing of DWI data were corrected for motion and eddy current geometric distortions and non-brain tissues were removed using fMRI Software Library (FSL, http://fsl.fmrib. ox.ac.uk/fsl). To estimate the metrics fractional anisotropy (FA) and mean kurtosis (MK), the toolbox Diffusional Kurtosis Estimator (DKE, http://academicdepartments.musc.edu/cbi/dki/dke.html) was used. After the diffusion kurtosis images (DKI) approximation of the diffusion orientation distribution function (dODF), the fiber bundle orientations were estimated from DKI with FA>0.1 and angle>35°. Based on the DKI tractography, fiber number (FN), fiber length (FL), FA and MK were then mapped into the Automated Anatomical Labeling (AAL) template to produce a 90×90 structural connectivity (SC) matrix.
The analysis of fMRI data was performed using the Data Processing Assistant for Resting-State fMRI (DPARSF) and Statistical Parametric Mapping (SPM12). Based on the AAL template, the preprocessed fMRI data were segmented into 90 regions. We applied the Fisher Z-Transformation to covert the correlation coefficients so that the data became normally distributed. The functional connectivity (FC) and SC matrices of HC, MCI-Aβ positive and negative subjects were analysed using the small-world network model. The properties (clustering coefficient (γ), characteristic path length (λ) and small worldness (S)) were computed using brain connectivity toolbox (BCT, http://www.brain-connectivity-toolbox.net/) with connection density from 5% to 20%.
We assessed SC-FC coupling as the correlation coefficient between the strengths of structural networks (FN, FL, FA and MK) and functional connectivities. The structural matrices were rescaled into a Gaussian distribution. At last, the SC-FC coupling was quantified by Pearson’s coefficient of correlation. Decoupling is defined as the SC-FC coupling values in subgroups that have significant differences from HC.
One-way ANOVA was used to compare among the three groups and the post-hoc test was applied to compare between each two groups. The relationships between MoCA score and SC-FC coupling were estimated based on the partial Pearson correlation method. (SPSS Inc., Chicage, USA).

Results

The characteristics of patients and HC are summarized in Table 1. No significance was found in FC among groups and between each two groups (Figure 1). Characteristic path length in SC matrices of FA, FL, FA and MK showed significant difference among MCI_postive, MCI_negative and HC. In Figure 2, statistical difference could be seen between HC and MCI_positive and between HC and MCI_negative in characteristic path length of four SC matrices, and in small worldness of MK. In Table 1 and Figure 3, SC-FC decoupling was observed in MCI-positive group from FN, FA and MK matrices. In addition, the decoupling was significantly positively correlated with MoCA score (Figure 4, adjusted with age).

Discussion and conclusion

Structural damage has been shown in MCI patients, as we found an increased characteristic path length in DKI matrices. Small-world network represented the balance to minimize the resource cost and maximize the flow of information among regions.⁵ The increase of path length indicated more cost for global integration. Meanwhile, functional connectivity remained globally normal. This demonstrated structural damage preceded the functional reorganization. Compared with Aβ negative, the MCI positive patients showed significant global decoupling with HC, which suggesting the pathological effects of fibrillar amyloid plaque toxicity in anatomical pathways and functional reorganization happens beyond the confines of structural pathways.

Acknowledgements

This work was supported by the State Key Laboratory of Brain and Cognitive Sciences, the University of Hong Kong.