APOE Polymorphism Affects Brain Default Mode Network in Healthy Young Adults

Supplemental Digital Content is available in the text


INTRODUCTION
A lzheimer disease (AD) is a progressive neurodegenerative disease. Apolipoprotein E (APOE) gene (e2, e3, and e4 alleles) is highly associated with AD. 1 APOE e4 allele is a well established genetic risk for decline of memory, e3 allele plays a neutral role, 1,2 whereas e2 allele enhances neuroprotection against AD. 3,4 Recent neuroimaging techniques, including functional magnetic resonance imaging (fMRI), allow for the investigation of genetic influence on the brain. fMRI studies showed decreased 5,6 or increased 7 default mode network (DMN) functional connectivities (FCs) and decreased hippocampal volumes 8 in healthy middle-aged or older e4 carriers compared with noncarriers. Healthy older e2 carriers exhibited greater cortical thickness. 4 All these findings support the fact that the effect of e4 and e2 allele on brain function and structure begins early in life.
Exploration of APOE polymorphism on brain function and structure in young cognitively intact individuals may help identify the early alterations before ageing leaves an imprint on middle-aged or older individuals, 9,10 and better understand the pathophysiology of APOE on AD. Of limited fMRI studies in young e4 carriers, the structural and functional alterations varied widely: e4 carriers exhibited either no difference 11 or alterations of hippocampal volume, 12 and elevated 13,14 or reduced activity 15 during task-related fMRI compared with noncarriers. Majority of these studies recruited only e4 and e4 noncarriers 5,11,14 and depicted no clear profiles of APOE polymorphism on brain cognition, structure, and function. Thus, we used multimodality MRI in cognitively intact e2/e3, e4/e3, and e4/e3 carriers younger than 35 years to investigate the effect of APOE polymorphism on the resting-state brain function, structure, and blood flow.

Participants, Standard Protocol Approvals, and Participant Consents
The local institutional review board of the Jinling Hospital, Nanjing University, approved this study. Written informed consent was obtained from all participants. A total of 215 cognitively intact adults were recruited from the local community between April and December 2013. Inclusion criteria were as follows: age 18 to 35 years; right-handedness; without any clinical diseases; and 12 years or more educational level. Exclusion criteria were as follows: any history of psychiatric or neurological illness; head trauma; drug or alcohol abuse; chronic corticosteroid therapy; diabetes; hypertension; magnetic resonance (MR) contraindications; and excessive head movement (more than 1.0 mm in translation or 1.08 in rotation).

Neuropsychological Assessments
All participants completed a battery of neuropsychological tests, 5,11,12,14  ). The MoCA tests are associated with the domains of cognitive dysfunctions including executive function, naming, memory, language, attention, ability of abstraction, and orientation and delayed recall. Of MoCA tests, language, attention, ability of abstraction, and orientation were similar among the 3 groups; thus only executive functioning, naming, memory, and delayed recall were involved in further analysis. All the neuropsychological assessments were carried out by a trained doctor (XL) with 4 years of experience in neuropsychological tests, according to standard measurements as previously described. 5,11,12,14 Genotyping The analysis of APOE genotypes was performed by an independent laboratory (Shanghai Tianhao Biological Technology Co. Ltd., Shanghai, China). Participants with homozygous genotype of e4/e4 (0.9%, 2/215) were extremely rare and were excluded from the study. No participant with e2/e2 genotype was found in the study population. Since e4 and e2 have been reported having different roles in AD processing, 16 participants with genotype of e4/e2 were excluded from this study. Of the remaining participants, 14 e2/e3 carriers in e2 group, 31 e3/e3 carriers in e3 group, and 31 e4/e3 carriers in e4 group, matched for age, sex, and education level, were included for further analysis.

Data Acquisition
Images were acquired on a 3-Tesla MR instrument (TIM Trio, Siemens Medical Solutions, Erlangen, Germany).

Data Processing
Resting state fMRI data were preprocessed by using SPM8 (statistical parametric mapping, http://www.fil.ion.ucl.ac.uk/ spm/) implemented in Matlab, Version 7.9 (MathWorks, Natick, MA). The first 10 volumes of the functional images were discarded to increase stability of the initial MRI signal and the adaptation of the participants to the environment. Images were corrected for acquisition delays (slice timing) before motion correction. The functional images of the remaining participants were realigned to the standard Montreal Neurological Institute (MNI) template (3 Â 3 Â 3 mm 3 ) and smoothed by convolution with an isotropic Gaussian kernel of 8 mm Full Width at Half Maximum (FWHM) to decrease spatial noise.

Functional Activity and Connectivity Analysis
For amplitude of low-frequency fluctuation (ALFF) analysis, the smoothed data were detrended and temporally filtered to extract frequencies of 0.01 Hz to 0.08 Hz, then the ALFF map of each participant was extracted according to a previous fMRI report, 18 using REST 1.8 software (http://resting-fmri.sourceforge.net). For DMN analysis, the group spatially independent component analysis (ICA) was performed to extract DMN on the smoothed fMRI data by using GIFT software (Vision2.0;http://icatb.sourceforge.net/). ICA separated linear mixed data into spatially independent components. To determine the number of independent components, dimension estimation of the smoothed data of the 3 groups was conducted by using the minimum description the length criterion, and the result was an average of 20 independent components. Then, fMRI data from all participants in each group were concatenated, and the temporal dimension of the aggregated data set was reduced by means of principal component analysis, followed by an independent component (with time courses and spatial maps) estimation by using the infomax algorithm. To each ICA, the time courses corresponding to the waveform of a specific pattern of coherent brain activity and the intensity of this pattern of brain activity across the voxels were expressed by the associated spatial map. Then, using the GIFT software (Vision 2.0d; http://icatb.sourceforge.net/), the DMN component of each participant was selected based on the largest spatial correlation with DMN templates provided by Dr Wei Liao (Center for Cognition and Brain Disorders and the Affiliated Hospital, Hangzhou Normal University, Hangzhou, China). 19

Voxel-based Morphometry (VBM) Data Processing
Voxel-based morphometry (VBM) data processing was performed in VBM8 toolbox (http://dbm.neuro.uni-jena.de/ vbm). First, high-resolution structural data were bias-corrected, tissue classified, and normalized to the MNI. Second, the standardized images were segmented as gray matter (GM), white matter (WM), and cerebrospinal fluid. Subsequently, analysis was performed on GM and WM segments separately, which was multiplied by the nonlinear components derived from the normalization matrix to preserve actual GM and WM values locally (modulated GM volume). Finally, all GM and WM images were spatially smoothed with an 8-mm FWHM Gaussian kernel.

Voxel-based DTI Data Processing
For DTI images, the 20 diffusion-weighted images were registered to b 0 image using SPM8, and then corrected for difference in spatial distortion due to eddy currents using FMRIB Diffusion Toolbox (FDTv2.0) as implemented in FMRIB Software Library (FSLv4.1;www.fmrib.ox.ac.uk/fsl). Fractional anisotropy (FA) and mean diffusivity (MD) maps were then computed by using Diffusion Toolkit software based on our previous study. 20 The b 0 maps were standardized into a standard MNI space by using the echo-planar imaging template supplied with SPM8, which generated the deformation parameters that were applied to the normalization for FA and MD images. The normalized FA maps and MD maps were smoothed with an 8-mm FWHM Gaussian kernel.

ASL Data Processing
The cerebral blood flow (CBF) values in whole brain and regional brain were calculated within the ASL Toolkit (ASLtbx, http://www.cfn.upenn.edu) based on SPM8 software according to our previous study. 17 Ninety ASL images were acquired from each participant, and the first one was taken as the equilibrium magnetization of brain (M0) image for CBF calculation. The first 2 pairs of ASL images were discarded during establishment of equilibrium of spin dynamics. After realigning, motion correction, and spatial normalization to MNI space, the 43 remaining of control/tag CBF image pairs were converted to mL/100 g/min using a single-compartment model. To eliminate the individual differences, the CBF images were normalized by using the equation CBF corrected ¼ CBF uncorrected /(GM þ 0.4WM). Since the hippocampus is a key area associated with AD pathology, CBF values in bilateral hippocampus were extracted. Apart from the whole brain maps and 90 regional CBF maps, the bilateral hippocampus CBF maps, and the maps of the brain regions showing ALFF, DMN, FA, or MD differences among groups were extracted. The corrected CBF images were then spatially normalized.

Statistical Analysis
The SPSS16.0 software (SPSS, Chicago, IL) was used for statistical analysis. Among demographic characteristics, neuropsychological scales, and the extracted MR imaging statistics, the qualitative data were described by relative ratio or percentage and tested by chi-square test. The normality of the remaining quantitative data, including age, education, and all the neurological assessments, were tested by Kolmogorov-Smirnov analysis. The data variance homogeneity of the 3 groups was analyzed by analysis of variance (ANOVA). In case of significant differences in the variance analysis, post-hoc analysis was used for intergroup comparison. In case of non-Gaussian distribution, the median and interquartile range [M (QU-QL)] represented, and k-independent samples nonparametric test was performed among the 3 groups. P value less than 0.05 was regarded as significant difference.
Amplitude of low-frequency fluctuation and DMN maps were analyzed for each group by 1-sample t test and corrected by false discovery rate (FDR) criterion with a threshold of P <0.05. ANOVA with Alphasim correction (http://afni.nih.gov/ afni/docpdf/AlphaSim.pdf) was conducted among the 3 groups at the threshold of P <0.05, using SPM8 software. ANOVA was performed on normalized ALFF maps and DMN maps among the 3 groups by using SPM8 software in process Alphasim correction (http://afni.nih.gov/afni/docpdf/AlphaSim.pdf) at the threshold of P <0.05 with number of clusters of 79 and 121, respectively. If statistical difference was present, post hoc analysis was performed to detect the intergroup differences.
The ANOVA was performed on normalized GM maps and WM maps, FA and MD maps, and CBF maps (the whole brain, and regional CBF maps including bilateral hippocampus and the brain regions showing functional or structural differences) among the 3 groups, corrected by FDR criterion with a threshold of P <0.05. If the difference among the 3 groups were statistically significant, post hoc analysis was performed for intergroup differences. Bivariate analysis was conducted to analyze the correlation between brain regions showing functional or structural differences and neuropsychological scales. The data with normal distribution were analyzed by Pearson correlation analysis, whereas data with non-normal distribution was evaluated by Spearman correlation analysis with the threshold of P <0.05. Table 1 shows the demographical and neuropsychological results. Age, sex, and educational level showed no difference among groups with APOE e4, e3, and e2 (P ¼ 0.681, 0.988, 0.232, respectively). None of the neuropsychological scales showed any difference (all P > 0.05).
Differences among the groups were mainly located in bilateral MPFC, PCC, and right superior and middle temporal gyrus (STG) and middle temporal gyrus (MTG). Post hoc analysis demonstrated that group e4 showed significantly increased FCs in the left MPFC and bilateral PCC/PCu compared with the group e3, and increased FCs in the left MPFC and right STG and right MTG compared with group e2 (Figure 2 and Table 2). There was no difference between groups e2 and e3.

ALFF maps
Results of 1-sample t test showed that ALFF maps were mainly distributed in the bilateral frontal lobes, temporal lobes, occipital cortices, cingulate cortices, PCu, inferior parietal lobules, thalamus, and midbrain among the groups. ALFF maps showed no significant differences among the 3 groups.

Brain Volume Alteration
The volumes of GM or WM masks were no different among the 3 groups (Supplementary Table 1

FA and MD Maps
No differences in FA and MD values were found among the 3 groups.

CBF Maps
Whole brain and each regional CBF showed no significant differences among the 3 groups, although the average CBF values in these brain regions showed a gradually increased trend of groups e4 < e3 < e2 (Table 3).

DISCUSSION
Our study showed that APOE e4 carriers exhibited significantly increased DMN FCs when compared with e3 and e2 carriers. The e4 affects DMN FCs before brain structure and blood flow in cognitively intact young individuals. To our knowledge, this is the first study using multimodality MRI to investigate the influence of the APOE polymorphism (e4, e3, and e2 alleles) on brain structure, function, and blood flow in cognitively intact young adults.
In our study, e4 carriers exhibited significantly increased DMN FCs when compared with e3 and e2 carriers, which was consistent with prior studies 14 : e4 influences functional connectivity within the DMN before the onset of clinical disease in the youth. However, older cognitively intact e4 carriers manifested decreased DMN FCs. 21,22 It is unclear whether these differentiations embodied in age-dependent alterations or potential pathological progress. The involved brain regions such as bilateral MPFC, PCC, and temporal gyrus were consistent with the preferential areas in AD. 22,23 MPFC had been found to be linked with other components of the limbic system anatomically and functionally, and carries out self-referential mentations, executive function, and working memory. 24,25 Both PCC/PCu and STG play leading roles in episodic memory. 26,27 Positive correlation between the right STG FCs and vocabulary learning, delayed recall, and graph recall, which represent processes of productive and receptive memory, and consolidation 23,28 was also reported. Our findings regarding the increased DMN FCs may be a compensatory for slight alteration of cognitive function in a very young age. e4 has a potential to disrupt the dynamic balance of excitatory and inhibitory neurotransmitters, which is crucial for learning, memory, and for DMN alteration. 29,30 The increased FCs in DMN observed in this study might be the consequence of high glutamatergic concentration and/or low GABA compensation in the cognitively intact young e4 carriers. This assumption is supported by a recent study in which high regional inhibitory neurotransmitter levels were associated with enhanced deactivation, whereas high excitatory glutamate concentration was associated with reduced deactivation induced by the task. 31 Thus, DMN might be an indirect biomarker for alterations of neurotransmitters in neurons and/or interneurons. In addition, the pathology of amyloid b disposition (e4 > e3 > e2) and clearance (e2 > e3 > e4), including synaptic deficits, mitochondrial dysfunction, and neuroinflammation in the brain 1,11 in e4 carriers, might contribute to DMN alterations. The DMN describes the spatial linkages between neuronal activities in different brain regions, whereas the ALFF measures the amplitude of the regional spontaneous neuronal activity. 13,32 Both of these brain functional measurements showed abnormalities in MCI and AD. [33][34][35] In this study, the DMN but not the ALFF showed abnormalities among the groups. The inconsistence between DMN and ALFF has been previously described in MCI and AD patients. 36 One possible explanation was that alterations of spatial connectivity might precede alterations of neuron connectivity, that is, APOE-induced interneuron alteration precedes neuron alteration.
Voxel-based morphometry reflects brain macrostructure changes of gray matter and white matter. DTI reflects microstructure alterations of the integrity of white matter. 11 Our study did not find any VBM and DTI differences among the 3 groups, consistent with a previously published study. 11 Otherwise, 1 study found lower gray matter density in e4 carriers relative to e4 noncarriers via VBM in a wide range aged 19 to 80 years. 37 FIGURE 2. Spatial maps of the DMN by ANOVA among the 3 groups (e4, e3, and e2). Spatial maps of the DMN shows that significant differences among the groups located in bilateral medial prefrontal cortices, posterior cingulate cortices, and right superior temporal gyrus. Post hoc analysis shows significantly increased functional connectivity in the left medial prefrontal cortices, and bilateral posterior cingulate cortices/precuneus in group e4 compared with group e3, and increased functional connectivity in left medial prefrontal cortices and right superior and middle temporal gyrus compared to group e2 (Alphasim-corrected, P < 0.05). ANOVA ¼ analysis of variance, DMN ¼ default mode network. However, it is unclear whether ageing or APOE e4 led to decreased gray matter density. ASL is able to noninvasively quantify CBF. In our study, ASL images showed no abnormalities in cognitively intact young e4 carriers inconsistent with previously reported results, 9 in which young e4 carriers (23.6 AE 3.1 y) had increased CBF value in ACC. In this study, we observed the alteration of DMN before the alteration of brain macrostructure and microstructure, and CBF. The pathological mechanism of AD might help to explain the reason behind that functional alteration precedes structural alteration. White matter integrity has been reported to be positively associated with cerebrospinal fluid markers of amyloid-beta (amyloid b [42] and amyloid b [42]/p-Tau [181]) in AD. 38 In a recent study, Tau protein has been reported to induce pathological changes in the brain including gray matter atrophy, increased white matter radial diffusivity, decreased amide proton transfer, and hyperperfusion in the rTg4510 mouse model of tauopathy. 39 Whereas no differences in the concentration of plasma amyloid b peptides have been reported in young (28 AE 7.6 y) e4, e3, and e2 carriers without any memory deficits, 40 suggesting that the alterations of brain macrostructure and microstructure, and CBF may not occur at an early age.
Even though e4 and e2 alleles may have different contributions to AD, e2 carriers manifested no significant alterations   41 and 60 to 80 years. 16 The clear role that e2 plays needs to be further evaluated. This study has some limitations. First, the frequency of APOE gene (e2, e3, and e4 alleles), especially that of e2 allele, is extremely rare. Additional studies with expanded sample size are needed to further investigate the genetic influence on alterations in brain macrostructure and microstructure, as well as CBF. Secondly, whereas the influence of e4, e3, and e2 alleles on brain function, structure, and blood flow was evaluated in this cross-sectional study, further follow-up studies are necessary to investigate whether e4 and e2 alleles have dynamic pathophysiologic changes along aging trajectories. Third, this research was focused only on sporadic e4 carriers, and patients with AD-related family history need to be included in the future studies. Fourth, apart from DMN, other spatially distributed networks including executive-control network and salience network are needed to deepen the understanding of dynamic integration of early pathological manifestations and alterations of brain networks.
In conclusion, this study shows that the effect of APOE alleles on brain function precedes alterations in brain structure and blood flow in healthy young adults. DMN abnormalities may serve as potential biomarkers for detecting brain alterations in healthy young adults with APOE alleles.