Introduction

Genetic risk is a risk factor of AD, but how to identify the differences of neurodegenerative changes between normal and early Alzheimer’s disease (AD) is still a question to be answered.¹ As a high risk genetic factor, Apolipoprotein E4 (ApoE4) carriers accounted for over 65% of AD.² Many clinical studies found that ApoE4, as the most neurotoxic isoform, could induce neuropathology through couples of cellular pathways.³ The aim of our study is to investigate the neuronal and biochemical mechanisms underlying this genetic risk. Resting state functional MRI (rs-fMRI) depicts blood oxygenation level dependent (BOLD) neuronal activities in the resting state. Graph theory is a method which detects the differences of the properties of small-world network of neuronal connections, which could be modulated by glutamatergic system in the hippocampus.

Methods

In total 91 healthy, cognitively normal subjects (age: from 20 to 84, average:51±16.6 years old, gender:58F/33M) were studied. Seven subjects were excluded due to the excessive head motion during scanning. All subjects underwent an MRI examination with a Phlips-3T MR scanner using a standard 8 channel head coil. Structural images were acquired with 3D fast field echo sequence (3D-T1-FFE sagittal, TR=7ms, TE=3.2ms, Flip angle=8, voxel size=111). Functional images were collected by using a gradient-echo echo-planar sequence (parameters: TR=2000ms, TE=30ms, flip angle=90, voxel size=3×3×4) sensitive to blood-oxygen-level-dependent (BOLD) contrast. During the functional scanning, participants were instructed to open their eyes to watch the cross in the mirror and not to think of anything. Single Voxel Spectroscopy (SVS) was performed on left and right hippocampus (Fig. 1) with the following parameters (TR/TE=2,000/39 ms, number of signals averaged=128, phase cycles=16, spectral width=2,000 Hz with spectral resolution of 1.95 Hz per point, and free induction decay=1024). Absolute concentration of Glx ([Glx]_abs) (summation of glutamate and glutamine) was measured and quantified using internal water as reference by QUEST in jMRUI (4.0) with cerebrospinal fluid, grey matter and white matter water content corrected.
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 Automated Anatomical Labeling(AAL) template⁴, the preprocessed fMRI data were segmented into 90 regions (each hemisphere has 45 regions). For every individual subject, the regional time series was obtained by averaging the time series over all voxels of this region. We applied the Fisher Z-Transformation to covert the correlation coefficients so that the data became normally distributed. Then the transformed FC matrix was used to perform the small-world network. We respectively calculated the small-world properties of the ApoE4 carriers and non-ApoE4 carriers with connection densities from 37% to 50%⁵. In the analysis, the cost of the network below 37% would start to fragment, and more than 50% the network becomes more random. The 90×90 FC matrix consisted of a collection of undirected vertices(nodes) and edges(links) between pairs of vertices. The properties considered were computed using brain connectivity toolbox (BCT, http://www.brain-connectivity-toolbox.net/).
Hub is the centrality identified region with high degree of connectivity to other parts of the brain. The local betweenness centrality and local degree were applied together to pick out the hubs of a network.⁶ The values of the two indexes were averaged between two hemispheres and then were ranked in descending order respectively. The top 20% of sum of the ranking score were selected as the network hubs.
The relationships between topological measurements (clustering coefficients and characteristic path length of hubs) and [Glx]_abs in the left and right hippocampus were estimated based on the Pearson correlation method in SPSS package (SPSS Inc., Chicage, USA).

Results

The significant differences in function were found in cluster coefficients in hubs of right inferior frontal gyrus, bilateral precentral gyrus, right inferior temporal gyrus, bilateral precuneus, occipital region, right parahippocampus and bilateral posterior cingulate cortex (PCC). No significant difference was detected in the characteristic path length, but an increased trend of path length could be detected in the carriers (Figure 2). ApoE4 carriers had worse network segregation and possibly integration. In addition, there is significant relationship between [Glx]_abs in left hippocampus and topological metrics in high-risk but not low risk group (Figure 3)., i.e. between the [Glx]_abs in left hippocampus and clustering coefficient of MPFC (ApoE carrier: r=-0.68, p=0.001*; ApoE noncarrier: r=-0.06, p=0.696), characteristic path length of MPFC (ApoE carrier: r=0.55, p=0.012*; ApoE noncarrier: r=0.13, p=0.368), clustering coefficient of left parahippocampus (ApoE carrier: r=-0.56, p=0.012*; ApoE noncarrier: r=-0.06, p=0.674) and clustering coefficient of PCC(ApoE carrier: r=-0.58, p=0.009*; ApoE noncarrier: r=-0.10, p=0.507).

Discussion and conclusion

Small-world network represented the balance of local segregation and global integration in a real network compared to a random network⁷. The results suggested that ApoE ε4 carriers exhibited poorer local interconnectivity. It is tempting to speculate that the function of the brain networks could be affected by the genetic variations. Moreover, the close relationship between glutamate and topological properties in ApoE ε4 carriers might reflect the compensation of the impaired cortical communication efficiency.

Acknowledgements

This work was supported by Research Grants Council of Hong Kong (GRF grant number: 17108315) and the State Key Laboratory of Brain and Cognitive Sciences, the University of Hong Kong.