This is the first study to comprehensively identify functional connectivity alterations in the three RSNs implicated in cognition, in middle-aged adults with genetic risk for AD, preceding cognitive decline and gray-matter atrophy. We found the greatest Fc disruptions in the DMN, followed by the ECN and the least in the SN. We also have preliminary evidence to show that the alterations in the DMN and ECN are associated with memory performance in certain frontal, parietal and temporal regions across all subjects. The R-fMRI technique has great potential as an imaging endophenotype in preclinical AD research.
Our findings of diminished correlated and anticorrelated DMN Fc in APOE-ε4 carriers, especially in the frontal lobe and medial temporal lobe (MTL) structures, have been previously demonstrated in the absence of brain fibrillar β-amyloid plaque deposition 
. Also, diminished DMN Fc was also found in elderly APOE-ε4 carriers 
. In contrast, while predominantly increased DMN connections to the prefrontal cortices and MTL regions were seen in young healthy ε
4 carriers 
, APOE status had no effect on the DMN in another study 
. The inconsistencies may be a result of the differences in the age of enrolled participants 
and the methodology used to identify the RSNs (i.e., the independent component analysis 
versus the seed-based ROI analysis used in our study). Regardless, we found increases in PCC Fc, though this was limited to the basal ganglia structures. We believe that these increased functional connections in the DMN reflect the compensatory recruitment of additional neural resources by which the high-risk subjects retain normal cognitive functions, as previously described in fMRI task-activation and resting-state studies 
We demonstrate for the first time that the core cognitive RSN disruptions extend beyond the DMN to include the ECN and SN in middle-aged individuals at genetic risk for AD. Structures within the task-positive ECN were significantly correlated to each other; the ECN seed was anticorrelated with brain regions in the DMN in our control group. This is consistent with the literature 
. We believe that the intactness of the correlated and anticorrelated functional connectivity patterns in these RSNs is critical for performing episodic and working memory, and executive control functions in healthy individuals. Interestingly, while we did not observe significant differences in the cognitive measures between the APOE-ε4 carriers and noncarriers, decreases in the ECN connectivity were seen in the high-risk group. In AD, increased ECN connectivity previously has been reported relative to cognitively healthy normal persons 
. Moreover, the increased frontal network connectivity observed in AD is more extensive, in comparison to aMCI and controls 
. Patients with mild AD rely on increased ECN connectivity to compensate for a poorly functioning network (e.g., the DMN) as an attempt to improve cognitive performance. Our findings suggest that the increases in the ECN connections in the genetically at-risk individuals only may occur when cognitive decline is imminent and in those who have incipient AD. Future longitudinal studies are essential to elucidate this hypothesis.
The increased connections between the right AI with the dorsal ACC, PCC and precuneus, show for the first time that close functional interactions exist between the DMN and the SN in middle-aged individuals at high risk for AD. AI and ACC, core regions of the SN, are often coactivated during various internal and external salient stimuli. They play an important role in guiding behaviors associated with social, affective and higher order mental processes. AI, a crucial hub of the SN, is thought to mediate the dynamic interactions between other networks implicated in cognition, mainly the DMN and ECN. Increased Fc in the SN and decreased DMN Fc have been previously described in elderly cognitively healthy APOE-ε4 carriers and in patients with mild AD 
. More sophisticated Granger causality analyses are needed to further address the dynamic interactions between these RSNs in individuals at risk for AD 
. It is significant that the ECN and SN are originally defined with the right DLPFC and right AI as the seed regions according to the task-driven fMRI studies 
. In using the left DLPFC and left AI as seeds, different results were found (see Table S5
, Figure S2
and Figure S3
). Further study is needed to address neurobiological significance.
The mechanisms that contribute to the Fc differences in the cognition-related neuronal circuits between carriers and noncarriers are not fully understood. One explanation is that the APOE-ε
4 isoform increases β-amyloid aggregation in brain regions involved in these RSNs, thus leading to the observed functional disconnections. Aβ plaque deposition can precede neurodegeneration and clinical symptom onset in middle-aged adults at risk for AD. Interestingly, the MTL and neocortical areas that are functionally altered in the three RSNs also are early targets of β-amyloid plaque deposition 
. In contrast, occipital lobe structures are only affected in the advanced phases of AD 
. Moreover, alterations in noncognitive-associated RSNs are found in healthy ε
, and diminished DMN Fc is seen in the absence of cerebral amyloid pathology 
. These findings suggest that the effects of APOE genetic status on the functional connections may not be fully mediated by amyloid pathology.
APOE also plays a critical role in the neuronal development, and the presence of ε
4-allele is found to decrease neuronal growth and synaptic plasticity 
, thereby detrimentally affecting functional integration in cognitive networks by nonβ-amyloid-dependent mechanisms. If so, it is plausible that the loss of functional integration, as evidenced by R-fMRI alterations in ε
4-carriers, may precede commonly used AD biomarkers, including cerebral β-amyloid deposition, structural atrophy and neuropsychological deficits. APOE-ε
4 also stimulates abnormal tau hyperphosphorylation and formation of neurofibrillary tangles (NFTs) 
. Finally, APOE-ε
4 impairs cholesterol delivery from astrocytes to neurons, and can accentuate neurodegeneration 
Our exploratory voxel-based whole-brain analyses showed a weak but statistically significant relationship between episodic memory performance and the DMN and ECN Fc in several cortical areas across the whole sample. Interestingly, the brain areas within these cognitive circuits that are recruited to perform episodic memory functions across all participants, after controlling for several factors, including the group status, do not overlap with the areas differentially affected in carriers. These results indicate that distinct brain structures are often involved in performing episodic memory-related functions in cognitively healthy normal individuals and APOE-ε4 carriers, further providing evidence of the brain’s ability to balance and reorganize in normal and at-risk disease states.
Our study is not without limitations. First, this is a cross-sectional study. Future longitudinal studies are needed to disentangle the mechanisms linking APOE status, baseline cognition and brain function with the future incidence of AD. Second, while the lack of knowledge regarding the cerebral amyloid burden status is a limitation, DMN connectivity differences observed in this study are consistent with the findings reported in a sample free of cortical β-amyloid 
. Third, in the ROI-based DMN analysis, different seeds have been used to study Fc, and it is possible that distinct locations may produce divergent results. However, our findings are consistent with recent observations that have used different DMN and SN seed regions 
. Because our analysis is hypothesis driven, we limited our analysis to three RSNs; we plan to perform a data-driven, large-scale functional connectivity analysis in this cohort 
. Fourth, cerebrovascular risk factors and disease are related to memory decline in ε
4-allelic carriers 
. Although our study did not enroll individuals with significant vascular disease, future studies need to examine the relationship of varying degrees of white-matter disease and altered structural connectivity with Fc differences in cognitively healthy ε
4-allelic carriers. Fifth, because the primary objective of this study was to determine the effects of APOE on Fc in the cognitive RSNs, we did not examine the influence of family history of AD on the functional architecture of these networks. Finally, a gene dose-dependent relationship on RSN Fc may exist in APOE carriers 
. We were unable to explore this further because of small samples in ε
2) and ε
In summary, we have shown for the first time that intrinsic functional connectivity alterations is present in the core RSNs implicated in cognition beyond the DMN, even before gray-matter volume loss and the onset of cognitive decline, in middle-aged, ε4 carriers. Functional connectivity endophenotypes that are identified, using R-fMRI, can unravel the influence of APOE in middle-aged, cognitively healthy individuals with a genetic risk for AD. We also show the preliminary evidence of the effects of episodic memory performance on the altered Fc in the DMN and ECN, in cognitively normal individuals. Prospective studies that address the use of R-fMRI technique as a potential biomarker for predicting conversion from normal to early AD is needed, in ε4 carriers.