## Abstract

The apolipoprotein E (APOE) ϵ4 allele is a confirmed genetic risk factor and the APOE ϵ2 allele is a protective factor related to late-onset Alzheimer's disease (AD). Intriguingly, recent studies demonstrated similar brain function alterations between APOE ϵ2 and ϵ4 alleles, despite their opposite susceptibilities to AD. To address this apparent discrepancy, we recruited 129 cognitively normal elderly subjects, including 36 ϵ2 carriers, 44 ϵ3 homozygotes, and 49 ϵ4 carriers. All subjects underwent resting-state functional MRI scans. We hypothesized that aging could influence the APOE ϵ2 and ϵ4 allele effects that contribute to their appropriate AD risks differently. Using the stepwise regression analysis, we demonstrated that although both ϵ2 and ϵ4 carriers showed decreased functional connectivity (FC) compared with ϵ3 homozygotes, they have opposite aging trajectories in the default mode network—primarily in the bilateral anterior cingulate cortex. As age increased, ϵ2 carriers showed elevated FC, whereas ϵ4 carriers exhibited decreased FC. Behaviorally, the altered DMN FC positively correlated with information processing speed in both ϵ2 and ϵ4 carriers. It is suggested that the opposite aging trajectories between APOE ϵ2 and ϵ4 alleles in the DMN may reflect the antagonistic pleiotropic properties and associate with their different AD risks.

## Introduction

Aging remains the primary risk factor for Alzheimer's disease (AD). The apolipoprotein E (APOE) gene serves as the established genetic factor that influences AD onset in an isoform-dependent manner (ϵ4 > ϵ3 > ϵ2) (Corder et al. 1994; Seripa et al. 2009). A more complete understanding of the functional link between the 2 factors could provide insights into the APOE polymorphism's role in insidious AD progression. During the last decade, researchers introduced an imaging genetics strategy to study the APOE polymorphism's influence on human cognition at a neural system level (Bigos and Weinberger 2010). Specifically, most studies are dedicated to uncovering the neural basis of the APOE ϵ4 allele. The effect of APOE ϵ4 allele on brain function is directly linked to heightened AD risk. The resulting pathogenic signs have a potential clinical application in terms of their role as an AD biomarker.

However, several recent studies belie the existence of such a linkage. Trachtenberg, Filippini, Cheeseman, et al. (2012) reported that cognitively normal adult APOE ϵ4 and ϵ2 carriers showed nearly identical brain activation changes during memory and nonmemory tasks, relative to ϵ3 homozygotes . Further study using the resting-state functional connectivity (FC) MRI approach revealed overwhelming intrinsic neural network patterns that were similar between ϵ4 and ϵ2 carriers, compared with ϵ3 homozygotes (Trachtenberg, Filippini, Ebmeier, et al. 2012). These new results are intriguing. It was demonstrated that, in cognitively normal subjects, the ϵ4 allele was linked to accelerated cognitive decline (Caselli et al. 2009), whereas ϵ2 carriers exhibited relatively preserved cognitive function through time (Wilson et al. 2002; Blair et al. 2005). This raises the question of why ϵ4 and ϵ2 carriers exhibit similar brain function alterations relative to ϵ3 homozygotes rather than showing opposite changes in respect to their different AD risks. One answer lies in understanding the APOE neural effects beyond AD (Trachtenberg, Filippini, Cheeseman, et al. 2012). However, such an interpretation seems arbitrary in that it neglected the consistent association between the APOE polymorphism and AD, as well as the controversy about the APOE's role in other neurological disorders (Biffi et al. 2011).

To reconcile these differing views, an alternative interpretation could be that age contributes to the reported inconsistencies in the direction and magnitude of brain function changes (Trachtenberg, Filippini, Mackay, 2012). We hypothesized that the link between the different AD risks and associated brain function changes related to the APOE polymorphism was not merely dependent on the absolute brain activity detected at a specific age range. Rather, the link was associated with the trajectory of brain function changes across the life span. That is, brain function changes related to the APOE polymorphism proved to be age dependent. The ϵ4 allele had a negative trajectory from high to low with aging, whereas the ϵ2 allele had a positive trajectory from low to high. These opposite aging trajectories between the APOE ϵ2 and ϵ4 alleles contributed to their different AD risks. In this study, we tested these hypotheses by addressing the following questions. Did the APOE ϵ4 and ϵ2 alleles have similar effects on brain FC compared with the ϵ3 allele in cognitively normal elderly subjects? Did the APOE ϵ4 and ϵ2 alleles have the same or opposite aging trajectories in FC changes? At what specific ages would these APOE polymorphism trajectories intersect? Did the APOE polymorphism modulate the relationship between FC changes and the education years? If so, how?

We selected the default mode network (DMN) and its association with the APOE polymorphism to address these questions for 3 reasons. (1) It is suggested that the DMN, which consists of a set of brain hubs, including the posterior cingulate cortex (PCC)/precuneus (Pcu), inferior parietal lobule, medial prefrontal cortex, and lateral temporal cortex, is essential for interconnecting distinct networks. The DMN facilitates functional integration of the whole brain (Deshpande et al. 2011; Li et al. 2011). (2) The DMN is implicated in the AD development, as its disruption is consistently identified in AD-spectrum patients and cognitively normal elderly subjects with the APOE ϵ4 allele (Greicius et al. 2004; Sorg et al. 2007; Sheline et al. 2010). (3) The cortical distribution of the β-amyloid (Aβ) deposition in AD patients resembles the DMN pattern (Buckner et al. 2009). Therefore, in this study, we focused on the DMN to detect the APOE polymorphism's neural effects in cognitively normal elderly subjects between the ages of 54 and 80. Then, partial correlation analysis examined the cognitive significance of the altered DMN FC by APOE polymorphism. Finally, stepwise regression analyses examined how the altered DMN FC strengths were differently modulated by aging and the education years among the 3 APOE alleles.

## Materials and Methods

### Subjects

Subjects were recruited using media advertisements and community health screening events. The Affiliated ZhongDa Hospital of Southeast University Research Ethics Committee approved the study. Each participant provided a written informed consent. In this study, we initially enrolled 135 cognitively normal elderly subjects, including 37 APOE ϵ2ϵ3 subjects, 46 ϵ3 homozygotes, and 52 ϵ4 carriers. Specifically, 1 APOE ϵ2 carrier, 2 ϵ3 homozygotes, and 3 ϵ4 carriers were excluded for excessive motion artifacts (i.e., during the resting-state fMRI scan, the head motion exceeded 2 mm of maximum displacement in any direction or 2° of angular motion relative to the first volume) or incomplete image scans. Finally, the remaining 129 subjects, including 36 APOE ϵ2ϵ3 subjects (abbreviated as ϵ2 carriers), 44 ϵ3 homozygotes, and 49 ϵ4 carriers (47 subjects with the APOE ϵ3ϵ4 genotype and 2 ϵ4 homozygote subjects) entered further analysis.

### Clinical Evaluation

All subjects underwent a standardized clinical interview, including demographic inventory, medical history, and neurological and mental status examination. General cognitive function was assessed by the Mini-Mental State Examination (MMSE) and Mattis Dementia Rating Scale-2 (MDRS-2). Moreover, a neuropsychological battery test covering episodic memory, visuospatial function, information processing speed, and executive domains was utilized. It incorporated an Auditory Verbal Learning Test-20-min delayed recall (AVLT-20-min DR), Logical Memory Test-20-min delayed recall (LMT-20-min DR), Rey–Osterrieth Complex Figure Test (CFT) with its 20-min delayed recall (CFT-20-min DR), Clock Drawing Test (CDT), Digital Symbol Substitution Test (DSST), Trail Making Test-A and B (TMT-A and TMT-B), Stroop Color and Word Test A, B, and C, Verbal Fluency Test (VFT), Digital Span Test (DST), and Semantic Similarity Test (Similarity).

### Inclusion and Exclusion Criteria

For all study subjects, the following criteria were met: (1) Age was between 54 and 80; (2) education level was above junior middle school; (3) right-handed; (4) generally in good health with adequate visual and auditory acuity allowing cognitive testing; (5) normal neurological examination and no complaints of cognitive impairment; and (6) MMSE score of >24, MDRS-2 score of >120, and the neuropsychological battery performance was in normal range. The exclusion criteria included: (1) Existence or history of neurological or psychiatric diseases; (2) ferrous or electronic implants; and (3) gross structural abnormalities revealed by MRI images.

### APOE Genotyping

Each subject's genomic DNA was extracted from 250-μL EDTA-anticoagulated blood, using a DNA direct kit (Tiangen, China). A polymerase chain reaction-based restriction fragment length polymorphism (PCR-RFLP) assay detected the alleles of rs7412 and rs429358, respectively. Ultimately, the haplotype of rs7412 and rs429358 determined the APOE genotype. The specific process is outlined in Supplementary Material.

### MRI Data Acquisition

MRI images were acquired in a Siemens Verio 3.0-T scanner (Siemens, Erlangen, Germany) with a 12-channel head coil at the Affiliated ZhongDa Hospital of Southeast University. The subjects were instructed to relax and close their eyes during the scan. Their ears were occluded with earplugs. A pair of stabilizers immobilized their heads. Resting-state functional images, including 240 volumes, were obtained by gradient-recalled echo-planar imaging (GRE-EPI) sequence: repetition time (TR) = 2000 ms; echo time (TE) = 25 ms; flip angle (FA) = 90°; acquisition matrix = 64 × 64; field of view (FOV) = 240 × 240 mm; thickness = 4.0 mm; gap = 0 mm; number of slices = 36. High-resolution T1-weighted anatomical images covering the whole brain were acquired by a 3D-magnetization prepared rapid gradient-echo sequence: TR = 1900 ms; TE = 2.48 ms; FA = 9°; acquisition matrix = 256 × 256; FOV = 250 × 250 mm; thickness = 1.0 mm; gap = 0 mm, number of slices = 176. Additionally, routine axial T2-weighted images were obtained to exclude subjects with major white matter (WM) changes, cerebral infarction, or other lesions.

### Image Preprocessing

The resting-state functional imaging data were preprocessed, using the Analysis of Functional NeuroImages (AFNI) software (http://afni.nimh.nih.gov/afni) and MATLAB programs (The MathWorks, Inc., Natick, MA, USA). The first 5 volumes of the data were discarded to allow for T1 equilibration. Then, the raw data spikes were removed (3dDespike). Correction was performed for the intravolume acquisition time differences among slices and the intervolume motion effects during the scan (3dvolreg). Detrending was conducted to remove Legendre polynomials (3dDetrend). The obtained image was spatially normalized into a standard stereotaxic space with a 12-parameter affine approach and an EPI template image, which was resampled to 4 × 4 × 4 mm3 voxels. Possible confounding signals of WM, cerebrospinal fluid, and six-motion vectors were regressed out from the voxel-wise time series. No significant difference of head motion was observed among the 3 groups. We calculated the Global Negative Index, the ratio of the number of voxels negatively correlated with the global signal to the total number of voxels, for each subject. All were greater than 0.03, suggesting that our data's global signal was irrelevant to nonneural noise and should not be regressed out (Chen et al. 2012). Finally, a band-pass filter was applied to ensure that only low-frequency fluctuations between 0.015 and 0.1 Hz could be analyzed further.

### Seed-Based Functional Connectivity Analysis

The PCC cortical hub [coordinate in the Montreal Neurological Institute (MNI) space: −2, −45, 34] was selected to generate a 6-mm radius spherical seed region (Buckner et al. 2009), followed by coregistration to the functional data (3dfractionize). Individual time courses were extracted based on the coregistered seed region (3dmaskave). Furthermore, voxel-wise cross-correlation (CC) values between the seed region and the whole brain were calculated (3dfim+). A Fisher's Z-transformation was then applied to improve the normality of the CC values [m = 0.5ln(1 + CC)/(1 − CC)] (3dcalc). Finally, the voxel-wise values were smoothed with a 6-mm Gaussian kernel, and the individual DMN patterns were obtained; they were represented by the PCC FC network.

### Voxel-wise-Based Gray Matter Volume Correction

The voxel-wise gray matter (GM) volumes were included as covariates in the FC analysis to avoid the bias of FC strength due to anatomical variation (Bai, Watson, et al. 2011; Bai, Xie, et al. 2011). First, individual GM volume maps were obtained and then normalized to the MNI space, using the toolbox of voxel-based morphometry 8 (VBM8; http://dbm.neuro.uni-jena.de/vbm/). Second, the normalized GM volume maps were resampled to the same voxel size as the functional data, and further subjected to a logit transformation [logit(a) = 0.5ln(a/1 − a)] to improve the normality. Third, the voxel-wise values were smoothed as the functional image. Finally, the resulting GM values were voxelwisely regressed out as the nuisance regressor from the FC values to control the GM volume influence on the FC strength. The voxel-wise GM volume correction was performed for each subject. Additionally, we voxelwisely investigated the APOE genotype and aging effects, and their interaction, on the GM volume. The detailed method and results are provided in Supplementary Material, Supplementary Figures 4 and 5, as well as in Supplementary Tables 3 and 4.

## Statistical Analysis

### Demographic and Neuropsychological Data

One-way analysis of variance (ANOVA) and χ2 tests (only applied in the comparisons of gender and family history, FH) were used to compare the demographic data and neuropsychological performances among the 3 groups. The statistical significance was set at P < 0.05.

### Group-Level Intrinsic Connectivity Analysis

Individual DMN maps were grouped together for each group, respectively. Then, each group's DMN pattern was determined by a random-effects one-sample t-test, according to the AlphaSim program based on the Monte Carlo simulation algorithm (α = 0.05, voxel-wise P < 0.05, cluster sizes >5504 mm3; http://afni.nimh.nih.gov/pub/dist/doc/manual/AlphaSim.pdf).

A one-way ANOVA test compared the DMN pattern among the 3 groups (3dANOVA program, corrected by AlphaSim, α = 0.05, voxel-wise P < 0.05, cluster sizes >5504 mm3). Then, post hoc tests by the 3dttest++ program were separately applied in the comparison between groups of ϵ4 carriers and ϵ3 homozygotes, as well as between ϵ2 carriers and ϵ3 homozygotes. We also detected the difference between ϵ2 and ϵ4 carriers. All post hoc tests, controlling the effects of age, gender, FH, and education years, were corrected by the AlphaSim program. The α-value was set at 0.01, as 3 post hoc tests were involved and the significance level was no greater than 0.05/3 = 0.0167, according to the Bonferroni correction principle (α = 0.01, voxel-wise P < 0.05, cluster sizes >7104 mm3).

### Cognitive Significance of the Altered DMN Pattern

The FC strengths of the regions showing altered DMN FC in ϵ2 and ϵ4 carriers, relative to ϵ3 homozygotes, were, respectively, extracted. Partial correlation analyses were applied between the extracted FC strengths and the neuropsychological performances within each group, controlling the effects of age, gender, education years, and FH. To increase statistical power by reducing random variability, as previously described (Xie et al. 2012), we grouped the neuropsychological tests into 4 cognitive domains and transformed the raw scores into 4 composite Z scores. First, for each neuropsychological test, the individual raw scores were transformed to Z scores, according to the mean and standard deviation of the scores for all subjects, as shown in following equation:

(1)
$Zi=ri−r¯S$
where $Zi$ is the Z score of the ith subject; $ri$ is the raw score of the ith subject; $r¯$ is the average raw score of the neuropsychological test for all subjects; S is the standard deviation of the scores. Notably, for tests measured by timing, including TMT-A, TMT-B, Stroop A, Stroop B, and Stroop C, the raw scores were defined as the reciprocal of the time for the tests. Then, each cognitive domain's composite Z score was determined by averaging the Z scores of relevant tests, according to the following divisions: episodic memory (3 tests, including AVLT-20-min DR, LMT-20-min DR, and CFT-20-min DR), visuospatial function (2 tests, including CFT and CDT), information processing speed (4 tests, including DSST, TMT-A, Stroop A, and Stroop B), and executive function (5 tests, including VFT, DST-backward, TMT-B, Stroop C, and Similarity). All statistical procedures utilized the SPSS 20.0 software (SPSS, Inc., Chicago, IL, USA). Bonferroni correction for multiple comparisons was used with the significance level considered at P < 0.0125 (P = 0.05/4 composite scores).

### Stepwise Regression Analyses on the Altered DMN Pattern

To investigate whether aging may differently influence the similar effects of APOE ϵ2 and ϵ4 alleles on the DMN FC, we focused on the intersection region disconnecting to the PCC in both ϵ2 and ϵ4 carriers compared with ϵ3 homozygotes (see the Results section). A stepwise regression analysis, including the APOE status and age interaction, was employed. The initial model is shown below:

(2)
$m=β0+β1×APOE2+β2×APOE4+β3×age+β4×edu+β5×gender+β6×FH+β7×(APOE2×age)+β8×(APOE4×age)+ϵ$
where m is the Fisher's Z-transformed CC value of the intersection region for each subject across the 3 groups; $β0$ is the intercept of the fitting line; $β1andβ2$ are the effects of the APOE ϵ2 and ϵ4 alleles relative to the ϵ3 allele; $β3,β4,β5,andβ6$ are the effects of age, education years, gender, and family history, respectively; $β7andβ8$ represent the interaction between APOE status and age separately for the ϵ2 and ϵ4 alleles, using the APOE ϵ3ϵ3 as the reference group; $ϵ$ denotes the random errors. Finally, the different aging trajectories of the m values among groups were plotted, if the interaction between APOE status and age was found in the final model best predicting the m values.

Furthermore, as the education years yielded a positive correlation with the m values in the final model (see the Results section), we performed a post hoc stepwise regression analysis. This was done to detect if education years differently modulated the FC strength among groups, in addition to the effects of APOE status and age. The initial model is illustrated below:

(3)
$madjusted=β0+β1×edu+β2×(APOE2×edu)+β3×(APOE4×edu)+ϵ$

where $madjusted$ represents the adjusted m values that subtract the contribution of age and APOE status from the original m values in the intersection region, according to the final model stem from equation (2); $β1$ is the effect of education years; $β2andβ3$ represent the interaction between APOE status and education years separately for ϵ2 and ϵ4 alleles, using the APOE ϵ3ϵ3 as the reference group; the implications of $β0andϵ$ refer to those in equation (2). Similarly, the discrepant trajectories of the adjusted m values with education years among groups were depicted, if the interaction between APOE status and education years was observed in the final model.

## Results

### Demographic and Cognitive Data

As Table 1 and Supplementary Table 1 illustrate, no significant difference was observed in the demographic information and neuropsychological performance among the 3 groups, except for education years (F = 3.43, P = 0.04).

Table 1

Demographic and neuropsychological data for all subjects

APOE ϵ2ϵ3 (n = 36) APOE ϵ3ϵ3 (n = 44) APOE ϵ4+ (n = 49) F or χ2 P-value
Age (years) 69.33 ± 6.80 68.77 ± 6.62 67.41 ± 6.46 0.98 0.38a
Gender (M/F) 17/19 23/21 22/27 0.52 0.77b
Education (years) 13.33 ± 3.00 12.42 ± 3.15 11.60 ± 2.91 3.43 0.04a,
Family history (yes/no) 4/32 10/34 13/36 3.11 0.21b
MMSE 28.03 ± 1.52 28.57 ± 1.15 28.06 ± 1.54 1.97 0.14a
MDRS-2 138.17 ± 4.00 137.86 ± 3.73 137.73 ± 3.20 0.15 0.86a
Composite Z scores of each cognitive domain
Episodic memory 0.05 ± 0.67 −0.01 ± 0.78 −0.02 ± 0.67 0.10 0.90a
Visuospatial function 0.13 ± 0.79 −0.15 ± 0.73 0.04 ± 0.72 1.44 0.24a
Information processing speed −0.02 ± 0.78 0.02 ± 0.75 −0.01 ± 0.87 0.03 0.97a
Executive function 0.07 ± 0.60 −0.02 ± 0.63 −0.03 ± 0.80 0.27 0.77a
APOE ϵ2ϵ3 (n = 36) APOE ϵ3ϵ3 (n = 44) APOE ϵ4+ (n = 49) F or χ2 P-value
Age (years) 69.33 ± 6.80 68.77 ± 6.62 67.41 ± 6.46 0.98 0.38a
Gender (M/F) 17/19 23/21 22/27 0.52 0.77b
Education (years) 13.33 ± 3.00 12.42 ± 3.15 11.60 ± 2.91 3.43 0.04a,
Family history (yes/no) 4/32 10/34 13/36 3.11 0.21b
MMSE 28.03 ± 1.52 28.57 ± 1.15 28.06 ± 1.54 1.97 0.14a
MDRS-2 138.17 ± 4.00 137.86 ± 3.73 137.73 ± 3.20 0.15 0.86a
Composite Z scores of each cognitive domain
Episodic memory 0.05 ± 0.67 −0.01 ± 0.78 −0.02 ± 0.67 0.10 0.90a
Visuospatial function 0.13 ± 0.79 −0.15 ± 0.73 0.04 ± 0.72 1.44 0.24a
Information processing speed −0.02 ± 0.78 0.02 ± 0.75 −0.01 ± 0.87 0.03 0.97a
Executive function 0.07 ± 0.60 −0.02 ± 0.63 −0.03 ± 0.80 0.27 0.77a

Note: Data are presented as mean ± stand deviation (SD).

APOE ϵ2ϵ3, subjects with APOE ϵ2ϵ3 genotype; APOE ϵ3ϵ3, subjects with APOE ϵ3ϵ3 genotype; APOE ϵ4+, subjects carrying APOE ϵ4 allele; M/F, male/female; MMSE, Mini-Mental State Examination; MDRS-2, Mattis Dementia Rating Scale-2.

aP-values were obtained by one-way ANOVA analysis.

bP-values were obtained by χ2 tests.

*Significant difference was found in education years among the 3 groups. The performances of MMSE and MDRS-2 are presented as raw scores. The level of each cognitive domain is denoted by the composite Z scores. Raw scores and corresponding Z scores of individual neuropsychological tests are shown in Supplementary Table 1.

### Altered DMN Patterns Among Groups

The DMN FC patterns of the 3 APOE genotype groups were obtained by using PCC seed-based FC analysis (Supplementary Fig. 1). Voxel-wise one-way ANOVA detected the APOE polymorphism effects on the DMN FC strength among groups. As shown in Supplementary Figure 2, the APOE polymorphism had a significant effect on the DMN FC in the bilateral Pcu and anterior cingulate cortex (ACC) regions.

Further post hoc analysis between the ϵ2 carrier and ϵ3 homozygote groups revealed that the DMN FC reduction occurred in the bilateral Pcu and ACC regions for ϵ2 carriers compared with ϵ3 homozygotes (Fig. 1A and Supplementary Table 2). For numeric presentation, we averaged the DMN FC values within the 6-mm radius spherical regions centered at the peak coordinate in the ACC and Pcu areas, respectively. As shown in Figure 1B, the DMN FC in ϵ2 carriers was reduced 69.4% in the ACC region and 57.6% in the Pcu region, compared with ϵ3 homozygotes. With partial correlation analysis, the averaged DMN FC strength of these 2 regions was significantly correlated with the information processing speed in APOE ϵ2 carriers (indexed as the Zspeed score) (P = 0.002, R2 = 0.26). The lower the FC was, the worse the information processing speed, as shown in the left panel of Figure 1C. However, such a relationship was not found for ϵ3 homozygotes (right panel of Fig. 1C).

Figure 1.

Decreased DMN FC in APOE ϵ2 carriers with its cognitive significance in information processing speed. (A) Brain regions showing significant decreased DMN FC in the APOE ϵ2ϵ3 group compared with the APOE ϵ3ϵ3 group (P < 0.05, AlphaSim correction). Color bar is presented with Z scores. (B) Numerical representation of the regions showing significant decreased DMN FC in (A) (m value is the Fisher's Z-transformed CC coefficient, same below). Error bars represent standard error of the mean. (C) APOE ϵ2ϵ3 subjects showed a significant positive correlation between the decreased DMN FC and the information processing speed scores (left panel), while such a correlation was not observed in the APOE ϵ3ϵ3 group (right panel). B Pcu, bilateral precuneus; B ACC, bilateral anterior cingulate cortex; Zspeed, composite Z scores of information processing speed.

Figure 1.

Decreased DMN FC in APOE ϵ2 carriers with its cognitive significance in information processing speed. (A) Brain regions showing significant decreased DMN FC in the APOE ϵ2ϵ3 group compared with the APOE ϵ3ϵ3 group (P < 0.05, AlphaSim correction). Color bar is presented with Z scores. (B) Numerical representation of the regions showing significant decreased DMN FC in (A) (m value is the Fisher's Z-transformed CC coefficient, same below). Error bars represent standard error of the mean. (C) APOE ϵ2ϵ3 subjects showed a significant positive correlation between the decreased DMN FC and the information processing speed scores (left panel), while such a correlation was not observed in the APOE ϵ3ϵ3 group (right panel). B Pcu, bilateral precuneus; B ACC, bilateral anterior cingulate cortex; Zspeed, composite Z scores of information processing speed.

Using a similar post hoc analysis, as described above, it was found that the ϵ4 carriers also had reduced DMN FC in the bilateral ACC/right dorsolateral prefrontal cortex (DLPFC) region. The ϵ4 carriers' mean FC value of the 6-mm radius spherical region centered at the peak coordinate was 62.7% below than ϵ3 homozygotes (Fig. 2A,B and Supplementary Table 2). The DMN FC of individual ϵ4 carriers had a strong tendency to positively correlate with the information processing speed in ϵ4 carriers (P = 0.02, R2 = 0.11, left panel of Fig. 2C), as it did not reach the statistical significance level after performing the Bonferroni correction. For ϵ3 homozygotes, no such a relationship was observed (right panel of Fig. 2C). In addition, we compared the DMN FC between ϵ2 and ϵ4 carrier groups, and found that ϵ2 carriers exhibited decreased DMN FC in the left Pcu relative to ϵ4 carriers, as illustrated in Supplementary Figure 3.

Figure 2.

Decreased DMN FC in APOE ϵ4 carriers with its cognitive significance in information processing speed. The descriptions for (AC) are the same as provided in Figure 1, except that the comparison was made between the APOE ϵ4+ group and the APOE ϵ3ϵ3 group. B ACC, bilateral anterior cingulate cortex; R DLPFC, right dorsal lateral prefrontal cortex.

Figure 2.

Decreased DMN FC in APOE ϵ4 carriers with its cognitive significance in information processing speed. The descriptions for (AC) are the same as provided in Figure 1, except that the comparison was made between the APOE ϵ4+ group and the APOE ϵ3ϵ3 group. B ACC, bilateral anterior cingulate cortex; R DLPFC, right dorsal lateral prefrontal cortex.

### Opposite Aging Effects Between APOE ϵ2 and ϵ4 Carriers

To specifically investigate if aging influences similar effects differently between the APOE ϵ2 and ϵ4 alleles, the intersection region in the bilateral ACC (between Figs 1A and 2A) was obtained (Fig. 3A). In this region, both ϵ2 and ϵ4 carriers showed comparably reduced DMN FC relative to ϵ3 homozygotes, and the mean FC values were 39.0 and 37.8% lower than ϵ3 homozygotes for ϵ2 and ϵ4 carriers, respectively, as shown in Figure 3B. With the stepwise regression analysis using APOE ϵ3ϵ3 as the reference group, the model with APOE2, APOE4, APOE2 × age, APOE4 × age, and edu variables provided the best prediction of the DMN FC value in this region (adjusted R2 = 0.185, P < 0.0001, Table 2). Of note, the ϵ2 and ϵ4 carriers showed opposite neural trajectories with aging (Fig. 3C). In the case of ϵ2 carriers, the aging trajectory of DMN FC changes was positive. The older the ϵ2 carriers were, the stronger their FC values. In the case of ϵ4 carriers, however, the aging trajectory was negative. That is, the older the ϵ4 carriers were, the weaker their FC values. Interestingly, the modeled trajectories of ϵ3 homozygotes and ϵ4 carriers intersected at around age 55. After this age, the ϵ4 carriers started to show lower FC values than the ϵ3 homozygotes, despite the fact we estimated that they would exhibit higher FC at a younger age. Additionally, the modeled trajectories of ϵ2 and ϵ4 carriers intersected at around age 70. The FC of ϵ2 carriers was always lower than that of ϵ4 carriers. However, it gradually increased with age, and was estimated to surpass the FC of ϵ4 carriers at approximately age 70.

Table 2

Effects of APOE status, age, and education years on the altered DMN FC with the stepwise regression analysis

Model summary Adjusted R2 = 0.185 P < 0.0001
Predictors* Regression coefficients (95% CI) P-value
APOE2 −0.213 (−0.424 to −0.002) 0.05
APOE4 0.227 (0.029 to 0.424) 0.03
APOE2 × age 0.003 (0.000 to 0.006) 0.08
APOE4 × age −0.004 (−0.006 to −0.001) 0.02
Edu 0.006 (0.001 to 0.010) 0.02
Model summary Adjusted R2 = 0.185 P < 0.0001
Predictors* Regression coefficients (95% CI) P-value
APOE2 −0.213 (−0.424 to −0.002) 0.05
APOE4 0.227 (0.029 to 0.424) 0.03
APOE2 × age 0.003 (0.000 to 0.006) 0.08
APOE4 × age −0.004 (−0.006 to −0.001) 0.02
Edu 0.006 (0.001 to 0.010) 0.02

DMN, default mode network; APOE, apolipoprotein E; CI, confidence interval; edu, education years.

*Predictors of APOE genotype treated APOE ϵ3ϵ3 as the reference group.

Figure 3.

Opposite trajectories of DMN FC with aging between APOE ϵ2 and ϵ4 carriers. (A) The bilateral anterior cingulate cortex region showed decreased DMN FC in both APOE ϵ2ϵ3 subjects and APOE ϵ4 carriers, compared with APOE ϵ3ϵ3 subjects. (B) Numerical representation of the region in (A) shows significant decreased DMN FC for APOE ϵ2 and ϵ4 carriers. Error bars represent the standard error of the mean. (C) The interaction between APOE status and age on the DMN FC at the bilateral anterior cingulate cortex region. The trajectory of the DMN FC with aging is positive for APOE ϵ2ϵ3 group (red line) and negative for APOE ϵ4+ group (black line). B ACC, bilateral anterior cingulate cortex.

Figure 3.

Opposite trajectories of DMN FC with aging between APOE ϵ2 and ϵ4 carriers. (A) The bilateral anterior cingulate cortex region showed decreased DMN FC in both APOE ϵ2ϵ3 subjects and APOE ϵ4 carriers, compared with APOE ϵ3ϵ3 subjects. (B) Numerical representation of the region in (A) shows significant decreased DMN FC for APOE ϵ2 and ϵ4 carriers. Error bars represent the standard error of the mean. (C) The interaction between APOE status and age on the DMN FC at the bilateral anterior cingulate cortex region. The trajectory of the DMN FC with aging is positive for APOE ϵ2ϵ3 group (red line) and negative for APOE ϵ4+ group (black line). B ACC, bilateral anterior cingulate cortex.

### Discrepant Effects of Education Years Among the 3 APOE Alleles

Also, we observed that the factor of education years yielded a regression coefficient of 0.006 in the final model (Table 2). This indicates that the greater the education years, the stronger the DMN FC values were for the whole cohort. Moreover, the post hoc stepwise regression analysis, including the interaction between the APOE status and education years, revealed the degree to which the effect of education years on the DMN FC values changed as a function of the APOE status (adjusted R2 = 0.084, P = 0.003). Particularly, relative to the ϵ3 homozygotes exhibiting the regression coefficient of 0.006, ϵ4 carriers showed a significantly increased slope, whereas ϵ2 carriers showed a decreased slope. The absolute slopes for the ϵ2, ϵ3, and ϵ4 groups were 0.004, 0.006, and 0.008, respectively, with the highest slope for the ϵ4 carriers and the lowest for the ϵ2 carries, as shown in Figure 4 and Table 3.

Table 3

Effects of APOE status and education years on the altered DMN FC with the post hoc stepwise regression analysis

Model summary Adjusted R2 = 0.084 P = 0.003
Predictors* Regression coefficients (95% CI) P-value
Edu 0.006 (0.001 to 0.010) 0.01
APOE2 × edu −0.002 (−0.003 to 0.000) 0.02
APOE4 × edu 0.002 (0.001 to 0.004) 0.001
Model summary Adjusted R2 = 0.084 P = 0.003
Predictors* Regression coefficients (95% CI) P-value
Edu 0.006 (0.001 to 0.010) 0.01
APOE2 × edu −0.002 (−0.003 to 0.000) 0.02
APOE4 × edu 0.002 (0.001 to 0.004) 0.001

DMN, default mode network; APOE, apolipoprotein E; CI, confidence interval; edu, education years.

*Predictors of APOE genotype treated APOE ϵ3ϵ3 as the reference group.

Figure 4.

Discrepant functional trajectories with education years among the 3 APOE alleles. The education years positively correlated with the DMN FC in the bilateral anterior cingulate cortex region across the 3 APOE genotype groups. Notably, education years showed the strongest contribution for ϵ4 carriers (black line), the lowest for ϵ2 carriers (red line), and in between for ϵ3 homozygotes (green line).

Figure 4.

Discrepant functional trajectories with education years among the 3 APOE alleles. The education years positively correlated with the DMN FC in the bilateral anterior cingulate cortex region across the 3 APOE genotype groups. Notably, education years showed the strongest contribution for ϵ4 carriers (black line), the lowest for ϵ2 carriers (red line), and in between for ϵ3 homozygotes (green line).

## Discussion

This study's major novel finding is that APOE ϵ2 and ϵ4 carriers have opposite aging trajectories of DMN FC changes, primarily in the bilateral ACC regions, although they both showed similar decreased FC compared with ϵ3 homozygotes. Our findings corroborate prior results indicating that ϵ4 carriers exhibit significantly lower DMN FC than noncarriers in an elderly cohort (Sheline et al. 2010; Brown et al. 2011). Furthermore, we expand these results in 4 important ways. First, previous imaging genetic studies established a conceptual framework of APOE neuroimaging characteristics. The brain functional differences due to APOE polymorphism are associated with different AD risks in an ϵ4 allele dose-dependent manner for cognitively normal subjects (Bookheimer et al. 2000; Reiman et al. 2005; Sheline et al. 2010; Schraml et al. 2013). However, most of these studies only compared ϵ4 carriers and noncarriers. They were limited in terms of the separation of the ϵ4 noncarriers group into subgroups of ϵ2 carriers and ϵ3 homozygotes (Suri et al. 2013). We expanded previous studies by comparing functional differences among the 3 APOE alleles. We demonstrated that, in cognitively normal elderly subjects, both APOE ϵ2 and ϵ4 alleles similarly decreased DMN FC, rather than showing opposite effects according to their different AD risks. This result, combined with recent findings in a middle-aged cohort (Trachtenberg, Filippini, Ebmeier, et al. 2012), indicates that the absolute functional differences due to APOE polymorphism are not directly linked to their discrepant predispositions to AD.

Fourth, our study investigated the education effects on the DMN among the 3 APOE alleles. Education level is considered the primary proxy of cognitive reserve in that a high education level would increase the reserve capability and allow the brain to resist AD-related pathogenesis (Valenzuela and Sachdev 2006). In this study, we found that the factor of increased education years could enhance the DMN FC in the ACC region across groups, in addition to the effects of age and APOE status. This result corroborates a recent study identifying the ACC region as the neural basis of education-related cognitive reserve in a cognitively normal elderly cohort (Arenaza-Urquijo et al. 2013). Moreover, after regressing out the influences of age and APOE status, we observed that the contribution of education years differed by APOE status. Education years had the strongest effect on ϵ4 carriers, followed by ϵ3 homozygotes and, finally, ϵ2 carriers. This APOE isoform-dependent effect of education years indicates that a high education level would reverse the negative aging trajectory for ϵ4 carriers. The more education obtained, the stronger the FC at older ages. These inspiring findings indicate that there is cognitive reserve modulation on the aging trajectory of FC, showing the potential of cognitive intervention for subjects with a high AD risk.

Additionally, we found that the disrupted DMN was associated with inferior information processing speed in both ϵ2 and ϵ4 carriers. Recently, it was indicated that the DMN and its anticorrelated task control network are mutually influenced, and such a reciprocal relationship may regulate individual performances for goal-directed tasks (Uddin et al. 2009; Wen et al. 2013). Importantly, for this study, the brain region associating with inferior information processing speed for the ϵ2 and ϵ4 carriers was primarily located in the bilateral ACC extending to the right DLPFC. This area is the principal region subserving cognitive control (Dosenbach et al. 2007), and its greater connectivity to the DMN is believed to facilitate better cognitive control performance (Hampson et al. 2006). In contrast, in attention deficit/hyperactivity disorder patients characterized as having a cognitive control deficit, their ACC connectivity to the DMN was diminished (Castellanos et al. 2008). In addition, similar to our findings, it was demonstrated that the anterior DMN, which primarily involves the ACC and prefrontal regions, exhibited reduced integrity that associated with poorer information processing speed in the context of aging (Damoiseaux et al. 2008). These studies convergently indicate the roles of the ACC and prefrontal region in information processing efficiency. Accordingly, our findings suggest that the reduced DMN integrity in the ACC and prefrontal region may contribute to less effective information processing for ϵ2 and ϵ4 carriers.

It should be noted that our study has several limitations. First, the relatively narrow age range between 54 and 80 in our sample challenges our expectation of different functional trajectories across the lifespan among the 3 APOE alleles. Although the modeled trajectories were estimated to intersect at around age 55 between ϵ4 carriers and ϵ3 homozygotes, we were not able to verify the extrapolation of the trajectories because subjects younger than 54 were not available for this study. Second, this study's cross-sectional design limits our understanding regarding the dynamic characteristics of the DMN integrity with aging for different APOE alleles. In view of the fact that there are large variations of brain functions within each APOE genotype group with the cross-sectional data, we are following up with these subjects in order to statistically confirm the modeled trajectories at an individual level in the future. Ultimately, by overcoming these limitations, we will be able to discern accurately the APOE genotype influence on the neural network across different age groups.

In this study, we reported that APOE ϵ2 and ϵ4 carriers have opposite aging trajectories of DMN FC changes, although both of them showed decreased FC compared with ϵ3 homozygotes. These findings revise previous APOE neuroimaging conceptual framework and indicate that the aging trajectories of brain function change, in addition to the absolute change, may influence susceptibility to AD in regard to different APOE alleles. Our data support the antagonistic pleiotropic effects of both ϵ2 and ϵ4 alleles on the neural network activity in an opposite way to link their different AD risks. Furthermore, we proposed that age should be included as an important regulator when investigating the role of APOE polymorphism in AD development at a neural system level.

## Funding

This work was supported by the National Natural Science Foundation of China (81171021, 91132727, and 91232000); the China Scholarship Council (201206090088); the USA National Institutes of Health grant (R01AG020279); and the Key Program for Clinical Medicine and Science and Technology: Jiangsu Province Clinical Medical Research Center (BL2013025).

## Notes

We sincerely thank Ms Carrie M. O'Connor, M.A., for editorial assistance; Dr Piero G. Antuono, M.D., for neurocognitive assessment instruction; Youming Zheng, Haixia Feng, Hong Zhu, and Xiaofa Huang for subject recruitment; Min Wang and Xiaohui Chen for technical support on MRI scan. Conflict of Interest: None declared.

## References

SM
Sanz-Arigita
EJ
Binnewijzend
MA
Ossenkoppele
R
Tolboom
N
van Assema
DM
Wink
AM
Boellaard
R
Yaqub
M
Windhorst
et al
2014
.
Amyloid and its association with default network integrity in Alzheimer's disease
.
Hum Brain Mapp
.
35
:
779
791
.
Arenaza-Urquijo
EM
Landeau
B
La Joie
R
Mevel
K
Mezenge
F
Perrotin
A
Desgranges
B
Bartres-Faz
D
Eustache
F
Chetelat
G
.
2013
.
Relationships between years of education and gray matter volume, metabolism and functional connectivity in healthy elders
.
NeuroImage
.
83
:
450
457
.
Bai
F
Watson
DR
Shi
Y
Wang
Y
Yue
C
Teng
Y
Wu
D
Yuan
Y
Zhang
Z
.
2011
.
Specifically progressive deficits of brain functional marker in amnestic type mild cognitive impairment
.
PLoS ONE
.
6
:
e24271
.
Bai
F
Xie
C
Watson
DR
Shi
Y
Yuan
Y
Wang
Y
Yue
C
Teng
Y
Wu
D
Zhang
Z
.
2011
.
Aberrant hippocampal subregion networks associated with the classifications of aMCI subjects: a longitudinal resting-state study
.
PLoS ONE
.
6
:
e29288
.
Bakker
A
Krauss
GL
Albert
MS
Speck
CL
Jones
LR
Stark
CE
Yassa
MA
Bassett
SS
Shelton
AL
Gallagher
M
.
2012
.
Reduction of hippocampal hyperactivity improves cognition in amnestic mild cognitive impairment
.
Neuron
.
74
:
467
474
.
Bero
AW
Yan
P
Roh
JH
Cirrito
JR
Stewart
FR
Raichle
ME
Lee
JM
Holtzman
DM
.
2011
.
Neuronal activity regulates the regional vulnerability to amyloid-beta deposition
.
Nat Neurosci
.
14
:
750
756
.
Biffi
A
Anderson
CD
Jagiella
JM
Schmidt
H
Kissela
B
Hansen
BM
Jimenez-Conde
J
Pires
CR
Ayres
AM
Schwab
K
et al
2011
.
APOE genotype and extent of bleeding and outcome in lobar intracerebral haemorrhage: a genetic association study
.
Lancet Neurol
.
10
:
702
709
.
Bigos
KL
Weinberger
DR
.
2010
.
Imaging genetics—days of future past
.
NeuroImage
.
53
:
804
809
.
Blair
CK
Folsom
AR
Knopman
DS
Bray
MS
Mosley
TH
Boerwinkle
E
,
Atherosclerosis Risk in Communities Study I
.
2005
.
APOE genotype and cognitive decline in a middle-aged cohort
.
Neurology
.
64
:
268
276
.
Bondi
MW
Houston
WS
Eyler
LT
Brown
GG
.
2005
.
fMRI evidence of compensatory mechanisms in older adults at genetic risk for Alzheimer disease
.
Neurology
.
64
:
501
508
.
Bookheimer
SY
Strojwas
MH
Cohen
MS
Saunders
AM
Pericak-Vance
MA
Mazziotta
JC
Small
GW
.
2000
.
Patterns of brain activation in people at risk for Alzheimer's disease
.
N Engl J Med
.
343
:
450
456
.
Brown
JA
Terashima
KH
Burggren
AC
Ercoli
LM
Miller
KJ
Small
GW
Bookheimer
SY
.
2011
.
Brain network local interconnectivity loss in aging APOE-4 allele carriers
.
Proc Natl Acad Sci USA
.
108
:
20760
20765
.
Buckner
RL
Sepulcre
J
Talukdar
T
Krienen
FM
Liu
H
Hedden
T
Andrews-Hanna
JR
Sperling
RA
Johnson
KA
.
2009
.
Cortical hubs revealed by intrinsic functional connectivity: mapping, assessment of stability, and relation to Alzheimer's disease
.
J Neurosci
.
29
:
1860
1873
.
Caselli
RJ
Dueck
AC
Osborne
D
Sabbagh
MN
Connor
DJ
Ahern
GL
Baxter
LC
Rapcsak
SZ
Shi
J
Woodruff
BK
et al
2009
.
Longitudinal modeling of age-related memory decline and the APOE epsilon4 effect
.
N Engl J Med
.
361
:
255
263
.
Castellanos
FX
Margulies
DS
Kelly
C
Uddin
LQ
Ghaffari
M
Kirsch
A
Shaw
D
Z
Di Martino
A
Biswal
B
et al
2008
.
Cingulate-precuneus interactions: a new locus of dysfunction in adult attention-deficit/hyperactivity disorder
.
Biol Psychiatry
.
63
:
332
337
.
Chen
G
Chen
G
Xie
C
Ward
BD
Li
W
Antuono
P
Li
SJ
.
2012
.
A method to determine the necessity for global signal regression in resting-state fMRI studies
.
Magn Reson Med
.
68
:
1828
1835
.
Corder
EH
Saunders
AM
Risch
NJ
Strittmatter
WJ
Schmechel
DE
PC
Jr
Rimmler
JB
Locke
PA
Conneally
PM
KE
et al
1994
.
Protective effect of apolipoprotein E type 2 allele for late onset Alzheimer disease
.
Nat Genet
.
7
:
180
184
.
Damoiseaux
JS
Beckmann
CF
Arigita
EJ
Barkhof
F
Scheltens
P
Stam
CJ
Smith
SM
Rombouts
SA
.
2008
.
Reduced resting-state brain activity in the “default network” in normal aging
.
Cereb Cortex
.
18
:
1856
1864
.
Dennis
NA
Browndyke
JN
Stokes
J
Need
A
Burke
JR
Welsh-Bohmer
KA
Cabeza
R
.
2010
.
Temporal lobe functional activity and connectivity in young adult APOE varepsilon4 carriers
.
Alzheimers Dement
.
6
:
303
311
.
Deshpande
G
Santhanam
P
Hu
X
.
2011
.
Instantaneous and causal connectivity in resting state brain networks derived from functional MRI data
.
NeuroImage
.
54
:
1043
1052
.
Dosenbach
NU
Fair
DA
Miezin
FM
Cohen
AL
Wenger
KK
Dosenbach
RA
Fox
MD
Snyder
AZ
Vincent
JL
Raichle
ME
et al
2007
.
Distinct brain networks for adaptive and stable task control in humans
.
Proc Natl Acad Sci USA
.
104
:
11073
11078
.
Filippini
N
Ebmeier
KP
MacIntosh
BJ
Trachtenberg
AJ
Frisoni
GB
Wilcock
GK
Beckmann
CF
Smith
SM
Matthews
PM
Mackay
CE
.
2011
.
Differential effects of the APOE genotype on brain function across the lifespan
.
NeuroImage
.
54
:
602
610
.
Filippini
N
MacIntosh
BJ
Hough
MG
Goodwin
GM
Frisoni
GB
Smith
SM
Matthews
PM
Beckmann
CF
Mackay
CE
.
2009
.
Distinct patterns of brain activity in young carriers of the APOE-epsilon4 allele
.
Proc Natl Acad Sci USA
.
106
:
7209
7214
.
Greicius
MD
Srivastava
G
Reiss
AL
Menon
V
.
2004
.
Default-mode network activity distinguishes Alzheimer's disease from healthy aging: evidence from functional MRI
.
Proc Natl Acad Sci USA
.
101
:
4637
4642
.
Hampson
M
Driesen
NR
Skudlarski
P
Gore
JC
Constable
RT
.
2006
.
Brain connectivity related to working memory performance
.
J Neurosci
.
26
:
13338
13343
.
Han
SD
Bondi
MW
.
2008
.
Revision of the apolipoprotein E compensatory mechanism recruitment hypothesis
.
Alzheimers Dement
.
4
:
251
254
.
Jochemsen
HM
Muller
M
van der Graaf
Y
Geerlings
MI
.
2012
.
APOE epsilon4 differentially influences change in memory performance depending on age. The SMART-MR study
.
Neurobiol Aging
.
33
:
832
e815–822
.
Kamenetz
F
Tomita
T
Hsieh
H
Seabrook
G
Borchelt
D
Iwatsubo
T
Sisodia
S
Malinow
R
.
2003
.
APP processing and synaptic function
.
Neuron
.
37
:
925
937
.
Li
R
Chen
K
Fleisher
AS
Reiman
EM
Yao
L
Wu
X
.
2011
.
Large-scale directional connections among multi resting-state neural networks in human brain: a functional MRI and Bayesian network modeling study
.
NeuroImage
.
56
:
1035
1042
.
Liao
D
Miller
EC
Teravskis
PJ
.
2014
.
Tau acts as a mediator for Alzheimer's disease-related synaptic deficits
.
Eur J Neurosci
.
39
:
1202
1213
.
CR
de Quervain
DJ
Buchmann
A
Mustovic
H
Wollmer
MA
Schmidt
CF
Boesiger
P
Hock
C
Nitsch
RM
Papassotiropoulos
A
et al
2007
.
Better memory and neural efficiency in young apolipoprotein E epsilon4 carriers
.
Cereb Cortex
.
17
:
1934
1947
.
Mormino
EC
Smiljic
A
Hayenga
AO
Onami
SH
Greicius
MD
Rabinovici
GD
Janabi
M
Baker
SL
Yen
IV
CM
et al
2011
.
Relationships between beta-amyloid and functional connectivity in different components of the default mode network in aging
.
Cereb Cortex
.
21
:
2399
2407
.
Morris
JC
Roe
CM
Xiong
C
Fagan
AM
Goate
AM
Holtzman
DM
Mintun
MA
.
2010
.
APOE predicts amyloid-beta but not tau Alzheimer pathology in cognitively normal aging
.
Ann Neurol
.
67
:
122
131
.
Mucke
L
Selkoe
DJ
.
2012
.
Neurotoxicity of amyloid beta-protein: synaptic and network dysfunction
.
Cold Spring Harb Perspect Med
.
2
:
a006338
.
Palop
JJ
Mucke
L
.
2010
.
Amyloid-beta-induced neuronal dysfunction in Alzheimer's disease: from synapses toward neural networks
.
Nat Neurosci
.
13
:
812
818
.
Reiman
EM
Chen
K
Alexander
GE
Caselli
RJ
Bandy
D
Osborne
D
Saunders
AM
Hardy
J
.
2005
.
Correlations between apolipoprotein E epsilon4 gene dose and brain-imaging measurements of regional hypometabolism
.
Proc Natl Acad Sci USA
.
102
:
8299
8302
.
Schraml
F
Chen
K
Ayutyanont
N
Auttawut
R
Langbaum
JB
Lee
W
Liu
X
Bandy
D
Reeder
SQ
Alexander
GE
et al
2013
.
Association between an Alzheimer's disease-related index and gene dose
.
PLoS ONE
.
8
:
e67163
.
Seripa
D
Panza
F
Franceschi
M
D'Onofrio
G
Solfrizzi
V
Dallapiccola
B
Pilotto
A
.
2009
.
Non-apolipoprotein E and apolipoprotein E genetics of sporadic Alzheimer's disease
.
Ageing Res Rev
.
8
:
214
236
.
Sheline
YI
Morris
JC
Snyder
AZ
Price
JL
Yan
Z
D'Angelo
G
Liu
C
Dixit
S
Benzinger
T
Fagan
A
et al
2010
.
APOE4 allele disrupts resting state fMRI connectivity in the absence of amyloid plaques or decreased CSF Abeta42
.
J Neurosci
.
30
:
17035
17040
.
Sorg
C
Riedl
V
Muhlau
M
Calhoun
VD
Eichele
T
Laer
L
Drzezga
A
Forstl
H
Kurz
A
Zimmer
C
et al
2007
.
Selective changes of resting-state networks in individuals at risk for Alzheimer's disease
.
Proc Natl Acad Sci USA
.
104
:
18760
18765
.
Suri
S
Heise
V
Trachtenberg
AJ
Mackay
CE
.
2013
.
The forgotten APOE allele: a review of the evidence and suggested mechanisms for the protective effect of APOE varepsilon2
.
Neurosci Biobehav Rev
.
37
:
2878
2886
.
Trachtenberg
AJ
Filippini
N
Cheeseman
J
Duff
EP
Neville
MJ
Ebmeier
KP
Karpe
F
Mackay
CE
.
2012
.
The effects of APOE on brain activity do not simply reflect the risk of Alzheimer's disease
.
Neurobiol Aging
.
33
:
618
.
e1–618.e13
.
Trachtenberg
AJ
Filippini
N
Ebmeier
KP
Smith
SM
Karpe
F
Mackay
CE
.
2012
.
The effects of APOE on the functional architecture of the resting brain
.
NeuroImage
.
59
:
565
572
.
Trachtenberg
AJ
Filippini
N
Mackay
CE
.
2012
.
The effects of APOE-epsilon4 on the BOLD response
.
Neurobiol Aging
.
33
:
323
334
.
Tuminello
ER
Han
SD
.
2011
.
The apolipoprotein e antagonistic pleiotropy hypothesis: review and recommendations
.
Int J Alzheimers Dis
.
2011
:
726197
.
Uddin
LQ
Kelly
AM
Biswal
BB
Castellanos
FX
Milham
MP
.
2009
.
Functional connectivity of default mode network components: correlation, anticorrelation, and causality
.
Hum Brain Mapp
.
30
:
625
637
.
Valenzuela
MJ
Sachdev
P
.
2006
.
Brain reserve and cognitive decline: a non-parametric systematic review
.
Psychol Med
.
36
:
1065
1073
.
Wen
X
Liu
Y
Yao
L
Ding
M
.
2013
.
Top-down regulation of default mode activity in spatial visual attention
.
J Neurosci
.
33
:
6444
6453
.
Wilson
RS
Bienias
JL
Berry-Kravis
E
Evans
DA
Bennett
DA
.
2002
.
The apolipoprotein E epsilon 2 allele and decline in episodic memory
.
J Neurol Neurosurg Psychiatry
.
73
:
672
677
.
Wishart
HA
Saykin
AJ
Rabin
LA
Santulli
RB
Flashman
LA
Guerin
SJ
Mamourian
AC
Belloni
DR
Rhodes
CH
McAllister
TW
.
2006
.
Increased brain activation during working memory in cognitively intact adults with the APOE epsilon4 allele
.
Am J Psychiatry
.
163
:
1603
1610
.
Wright
RO
Hu
H
Silverman
EK
Tsaih
SW
Schwartz
J
Bellinger
D
Palazuelos
E
Weiss
ST
Hernandez-Avila
M
.
2003
.
Apolipoprotein E genotype predicts 24-month bayley scales infant development score
.
Pediatr Res
.
54
:
819
825
.
Xie
C
Bai
F
Yu
H
Shi
Y
Yuan
Y
Chen
G
Li
W
Chen
G
Zhang
Z
Li
SJ
.
2012
.
Abnormal insula functional network is associated with episodic memory decline in amnestic mild cognitive impairment
.
NeuroImage
.
63
:
320
327
.

## Author notes

Hao Shu and Yongmei Shi contributed equally to this work.