Amyloid-β (Aβ) accumulation differentiates Alzheimer's disease (AD) from other neurodegenerative diseases (Jack et al., 2018). Additionally, the Aβ deposition has been known to progress non-linearly in decades before the onset of clinical symptoms of AD (Jack et al., 2013). In this regard, the preclinical stage of AD has therefore been defined as the presence of Aβ pathology without signs of significant cognitive impairment due to AD dementia (Knopman et al., 2012). In addition, the period before AD symptoms, which has become apparent, also attracts clinical attention as the right time for primary intervention because the Aβ-dependent pathophysiology is not yet fully advanced. Although this period could be overlooked in the clinical field due to the lack of apparent clinical symptoms, preclinical AD participants have demonstrated an increased risk of transition to MCI and AD (Knopman et al., 2012), as well as the deteriorating effects on the cognitive performance, brain structure, and function (Hedden et al., 2009; Mattsson et al., 2014; Baker et al., 2017).

Amyloid-β deposition is evaluated using Aβ-positron emission tomography (PET) for the early detection of preclinical AD and there has been renewed interest in Aβ accumulation below the threshold for a positive scan (Bischof and Jacobs, 2019). Even among individuals with a negative PET scan, 65% showed an early autopsy stage of Aβ accumulation (Thal phase 2), and 15% displayed an advanced stage (Thal phase 4 or 5) (Salloway et al., 2017). Additionally, specific brain regions, including the orbitofrontal cortex, the anterior cingulate cortex, and the precuneus, are prone to the earliest Aβ accumulation compared with other brain regions (Sojkova et al., 2011; Driscoll et al., 2012; Villeneuve et al., 2015). In addition, the regional Aβ deposits predicted cognitive decline more accurately than global deposition in cognitively normal (CN) older adults with subthreshold Aβ accumulation (Farrell et al., 2018). Moreover, the subthreshold Aβ deposits were associated with functional impairment (Insel et al., 2017), brain atrophy (Mattsson et al., 2014), and dysfunction of the brain functional networks (Palmqvist et al., 2017), and predicted further Aβ and tau deposition (Leal et al., 2018; Farrell et al., 2021).

APOE ε4 allele has been demonstrated to modulate the penetrance and weight of Aβ pathophysiological cascade (Frisoni et al., 2022) and to account for the largest proportion of the genetic risk factors for sporadic AD (Sims et al., 2020). In this regard, the APOE ε4 allele has been known to increase the risk of sporadic AD occurrence in a dose-dependent manner (Corder et al., 1993). The APOE ε4 allele has also been reported to increase Aβ production (DeMattos et al., 2004), reduce the clearance of Aβ (Castellano et al., 2011), and affect tau binding (Small et al., 2009). Lastly, APOE ε4 fragments have been demonstrated to interact synergistically with AD pathology, deteriorating the degree of neurodegeneration (Andrews-Zwilling et al., 2010; Bien-Ly et al., 2011). Therefore, the progression of sporadic AD cannot be fully understood without the consideration of the APOE genotype.

In addition, the changes in brain function have been demonstrated to precede those in the brain structure and track pathophysiological processes in the preclinical phase (Jack et al., 2013). Among the various methodologies employed for evaluating brain function, functional connectivity (FC) is a widely used and reliable method for evaluating functional interactions in the brain connectome (Biswal et al., 2010). In addition, the earliest accumulation of Aβ is known to affect FC (Palmqvist et al., 2017), and existing research has recognized the impact of the APOE ε4 allele on brain function and cognitive performance in the preclinical phase of AD. Previous research has found disrupted FC from the precuneus in the APOE ε4 carrier of the CN without positive Aβ deposition (Sheline et al., 2010). In another study, the authors demonstrated an altered within-network FC in the APOE ε4 carrier of the CN (Wu et al., 2016). Recently, investigators have reported a simultaneous structural and functional disruption that mediates memory impairment in the CN with the APOE ε4 allele (Li et al., 2021). However, the generalizability of this previous research is limited because most of these studies lacked information on Aβ deposition; they only evaluated the remote FC of predefined brain regions and showed results with weak statistical significance.

While a static FC is based on the assumed temporal stationarity of functional networks, the dynamic FC considers the dynamic nature of brain activity in faster timescales by selecting a time window that is shifted in time by a fixed number of data points (Hutchison et al., 2013). This dynamic FC has shown a significant association with the subthreshold Aβ accumulation in the preclinical phase; however, the interaction with the APOE ε4 allele was not evaluated in this prior study (Hahn et al., 2019). Moreover, results from dynamic FC must be interpreted with caution due to the arbitrariness of dynamic FC parameters and its vulnerability to physiological noise that could drive dynamic states (Laumann et al., 2017). These factors contribute to weaker reliability and reproducibility of dynamic FC than a static FC (Abrol et al., 2017).

Together, these studies provide important insights into the subthreshold Aβ, the pivotal role of the APOE ε4 allele, and the FC biomarker in the trajectory of AD. However, there remains a paucity of comprehensive evidence on the impact of APOE genotype on associations between subthreshold, positive Aβ accumulation, brain function, and cognitive performance in the preclinical phase of AD.

In this regard, the current study classified the CN group according to the presence of the APOE ε4 allele in each CN group with subthreshold Aβ deposition (CN sub-Aβ) and those with positive deposition (CN Aβ+). We aimed to evaluate the differential impact of the APOE ε4 carrier status on the association between the subthreshold, positive Aβ accumulation, FC, and cognitive performance in the preclinical phase. Additionally, local and remote FC evaluation of functional synchronization at different spatial scales in the brain connectome (Sepulcre et al., 2010) has been suggested to enhance network information capacity and discrimination accuracy of the global network (Deco et al., 2014). In this regard, the local-to-remote FC has been suggested to provide a comprehensive understanding of AD (Liu et al., 2014; Li et al., 2021). Therefore, the present study assessed the effect of the APOE ε4 carrier status by measuring the local and remote FC in an integrative manner. Furthermore, we included information on both global and regional Aβ accumulation and evaluated both memory performance and executive functions being affected at an initial stage for detecting subtle differences in Aβ burden and cognitive function in the earliest course of AD (Buckner, 2004; Farrell et al., 2018). In addition, we hypothesized that the APOE ε4 carrier would show a distinct pathway from Aβ deposition to cognitive function via differences in the local and remote FC in each CN sub-Aβ and Aβ+ group.

In the current study, we found a distinctive association between the regional, global Aβ accumulation, and cognitive performance scores depending on the APOE ε4 carrier status in each CN sub-Aβ and Aβ+ group. In the CN sub-Aβ group, the executive function was higher in the APOE ε4 carrier that showed higher global and regional Aβ accumulation. However, lower memory and global performances were displayed in the APOE ε4 carriers with higher regional Aβ deposition. Taken together, these results show that the area of cognitive function and the slope of plots between the Aβ pathology and cognitive function differed according to the degree of Aβ deposits in the APOE ε4 carrier in the preclinical phase. This result may also reflect the compensatory reaction in the executive function against the subthreshold Aβ accumulation in the CN. Additionally, we found a significant relationship between cognitive function and Aβ deposits in the posterior cortical regions. This finding broadly supports the work of a previous study demonstrating a more accurate prediction of prospective cognitive decline by Aβ accumulation in the posterior brain areas than by global deposits (Farrell et al., 2018). However, this previous study could not confirm the significant impact of the APOE ε4 allele on the results. In another study, a change in the subthreshold Aβ deposition was associated with memory decline but not executive function (Landau et al., 2018). In this prior study, the average age of the participants was older than that in the current study, and the presence of the APOE ε4 allele was not adjusted for the analysis. Moreover, the proportion of APOE ε4 carriers was only 15%. Therefore, such differences may have caused the discrepancy between the studies. Finally, the current study evaluated the association between Aβ accumulation and subtle cognitive decline within the normal range, which might have contributed to the relative lack of statistical robustness for the interaction.

Regarding the local FC, the APOE ε4 carriers exhibited significantly higher ReHo, mainly in the brain region of the DMN in the CN sub-Aβ group. However, there is little published research on the impact of the APOE ε4 allele on local FC in the preclinical phase of AD. Given the positive relationship between ReHo and regional glucose metabolism (Nugent et al., 2015), APOE ε4 carriers display increased glucose metabolism in the medial, frontal, and anterior temporal regions of the CN (Yi et al., 2014). However, this previous study did not evaluate Aβ deposition, and it assessed the impact of the APOE ε4 allele by including this high-risk genotype as the covariate in the analysis with small sample size. Another study with subjects complaining of subjective cognitive decline (SCD) showed a significant difference in the ReHo according to the Aβ status. However, this previous study did not explore the interaction with the APOE ε4 allele; it included subjects who were older than those in the present study, and also performed the analysis in the restricted sample size (Li et al., 2021). These previous results, therefore, need to be interpreted with caution regarding possible bias.

Among the ROIs showing a significantly higher ReHo in the APOE ε4 carrier of the CN sub-Aβ group, the precuneus is the known hub region of the DMN, and the ReHo in this region reflects the progression of AD (He et al., 2007; Kang et al., 2017). Although prior research has documented the negative relationship between Aβ deposition and ReHo of the precuneus in the CN with a positive Aβ PET scan, another previous study has demonstrated that SCD subjects with Aβ deposits show a higher ReHo in the right precuneus than those without Aβ accumulation, suggesting the compensatory role of the increased local FC (Li et al., 2021). However, given the Aβ-dependent neuronal hyperactivation (Zott et al., 2019), we could not rule out the possibility that early Aβ deposition induced a higher ReHo in the AD vulnerable brain region.

However, despite the importance of the APOE ε4 allele, there remains a paucity of evidence on the effect of the APOE ε4 allele on the local FC change in the earliest phase of AD. Concerning the middle occipital gyrus, which is another ROI in the CN sub-Aβ group, the FC between this ROI and precuneus has been shown to be higher in SCD subjects with Aβ burden than in those without Aβ deposits (Li et al., 2021). Additionally, we found a higher ReHo in several posterior regions of the cerebellum in the APOE ε4 carriers in the CN sub-Aβ group. The posterior cerebellum has been reported to be functionally mapped to the DMN brain regions (Buckner et al., 2011; Buckner, 2013) that are prone to Aβ accumulation (Mormino et al., 2011). In this regard, the Aβ-associated neuronal hyperactivation might affect the high activity of the DMN brain regions and functionally related posterior cerebellum (Pasquini et al., 2017). Additionally, neuronal hyperactivity has also been reported to induce further Aβ deposition (Bero et al., 2012). This positive correspondence between Aβ and intrinsic FC has been demonstrated to start in the preclinical stage (Pasquini et al., 2017). Additionally, a previous study has also shown the impact of the APOE ε4 allele on neuronal hyperactivity in a mouse model without definite Aβ pathology (Nuriel et al., 2017). Therefore, a synergistic effect of the APOE ε4 allele with the Aβ burden could affect the high activity of the ROIs in the current study. However, the posterior cerebellum did not show a significant difference depending on the APOE ε4 carrier status in the CN Aβ+ group of the present study. The detrimental effect of Aβ on FC has been reported to appear in brain networks where both the brain activity and Aβ deposits are high (Pasquini et al., 2017). In addition, this deteriorating effect has been reported to initiate already in the preclinical phase and reach a peak in the MCI stage (Pasquini et al., 2017). In this regard, a higher Aβ burden might weaken the brain activity of the ROIs in the CN Aβ+ group. However, a further longitudinal study is required to determine exactly how Aβ influences the brain functional change, conjoining with APOE ε4 allele in the preclinical phase.

In the CN Aβ+ group in the present study, the APOE ε4 carriers showed significantly lower ReHo in the insula, which is a known hub region of the SN. The insula has been demonstrated to have a modulatory role in the association between the DMN and CEN (Sridharan et al., 2008). Additionally, the SN has been documented to display distinctive FC activation from normal aging to AD progression (He et al., 2014). Contrary to the current findings, the APOE ε4 carrier in the preclinical phase has demonstrated increased FC related to the insula, which is attributed to the reduction of the inhibitory control of the DMN (Machulda et al., 2011). However, in the present study, there were no significant differences in the intra-network FC of the DMN and the inter-network FC between the DMN and SN depending on the APOE ε4 carrier status in the CN Aβ+ group. Nevertheless, caution should be exercised here because there was a lack of information on Aβ deposition in the above-mentioned previous studies, and the difference in the ReHo has been shown depending on Aβ positivity during the preclinical phase (Kang et al., 2017).

In the present study, the difference in ReHo in the APOE ε4 carriers was related to the higher executive and memory function in each CN sub-Aβ and CN Aβ+ group, respectively. Additionally, this finding broadly supports the work of other studies linking the functional activation of the posterior cerebellum with executive function (Baumann et al., 2015; Castellazzi et al., 2018). In this regard, this combination of findings provides some support for the compensatory role of the difference in the ReHo found in the APOE ε4 carriers in the preclinical phase. However, these results differ from some published studies reporting a non-significant association between FC and cognitive function in APOE ε4 carriers with intact cognition (Chen et al., 2015). These differences can be explained in part by not considering the effect of the Aβ burden in the previous study.

Another interesting finding was the differential association between regional Aβ burden and local connectivity according to the APOE ε4 carrier status in each CN sub-Aβ and Aβ+ group. In the APOE ε4 carriers in the CN sub-Aβ group, the ReHo was higher in ROI, including the posterior cerebellum, as the Aβ accumulation in the temporal lobe was higher, but in the CN Aβ+ group, the ReHo was lower in the right cerebellum crus 1 as the Aβ burden in the PCC/precuneus was higher. Although the subjects of this study were not in the age group in which Aβ deposition was actively increasing (Rodrigue et al., 2012), the temporal lobe and PCC/precuneus have been reported as early Aβ deposition regions (Sojkova et al., 2011). These regions have been demonstrated to be vulnerable to the APOE ε4 allele (Sheline et al., 2010) and to show an overlap between the topological distribution of Aβ deposits and the DMN (Kang et al., 2017). Additionally, hyperactivity in these regions has been assumed to induce further Aβ accumulation, which could cause neuronal dysfunction and reduced FC in the CN (Bero et al., 2011, 2012). In addition, among the ROIs in the current findings, the posterior cerebellum has been demonstrated to be functionally connected to the temporal gyrus and precuneus, which are the vulnerable regions in AD (Guo et al., 2016). Significantly, the cerebellum crus 1 and cerebellum 7 have also been mapped to the DMN brain regions (Buckner et al., 2011), including the association cortex where Aβ accumulation is prone to occur (Mormino et al., 2011). However, few previous studies on the distinctive effect of the APOE ε4 allele on the function of the posterior cerebellum have been performed. Therefore, starting from this study, additional reproducible studies are needed to verify the differential effects of the APOE ε4 allele. Furthermore, the temporal gyrus has also been known to be correlated with tau pathology (Insel et al., 2020) and a synergistic effect between Aβ and tauopathy for the changes in the FC has been reported (Schultz et al., 2017). Therefore, additional studies evaluating tauopathy are necessary to understand the earliest pathophysiology of AD more accurately.

In the CN sub-Aβ group of the present study, there was a significant interaction between the APOE ε4 carrier status and local FC in the right cerebellum crus 2 for the remote inter-network FC between the DMN and CEN, showing robustness to the decline in this inter-network FC in the APOE ε4 carriers. This is consistent with previous findings showing a significant association between the posterior cerebellum and resting-state networks, including the DMN and CEN (Habas et al., 2009; Buckner et al., 2011). The inter-network FC between the DMN and CEN has been reported to display a U-shaped progression pattern with increasing age, and to show a declining pattern in the age group of the subjects in the current study (Ng et al., 2016). In this regard, the APOE ε4 carrier status could be suggested to have an alleviating role for this declining pattern in the CN before the Aβ positivity. Furthermore, we found that the stronger inter-network FC of DMN-CEN was associated with better cognitive function in the APOE ε4 carrier in both CN sub-Aβ and Aβ+ groups. The function of the CEN has been demonstrated to be supported by that of the DMN in maintaining a normal cognition (Turner and Spreng, 2015). In this respect, the present results support previous observational studies that have shown a positive association between the cognitive function and the inter-network FC of DMN-CEN in the cognitively normal APOE ε4 carrier of similar age to the present study (Ng et al., 2018).

In the current study, the participants with varying risk for AD have not yet been confirmed to have developed further Aβ accumulation or converted to symptomatic disease. Moreover, considering the results of previous studies showing that Aβ changes rather than that baseline accumulation predicts prospective cognitive change (Farrell et al., 2018); accordingly, it is necessary to conduct a longitudinal study for a clear interpretation of the results of the current study. Given that prior study has demonstrated an interaction between Aβ and tau for changes in the FC of the CN (Hasani et al., 2021), another issue with this study is the lack of information on the tau deposits. Additionally, it has been reported that the FMM has a relatively low sensitivity to diffuse plaques that are usually detected in the earliest phase of Aβ accumulation (Salloway et al., 2017). In this respect, the current findings must be interpreted with caution when comparing them with the results of previous studies using other types of Aβ ligands. In addition, a selection bias is another potential limitation. The APOE ε4 carriers whose cognitive function remained intact despite the Aβ burden could have other characteristics, including protective genes (Seto et al., 2021) and higher cognitive reserve (Stern, 2009). The characteristics of these participants might have influenced the non-significant difference in cognitive function according to APOE ε4 carrier status. Finally, in the present study, multiple analyzes were performed after dichotomization according to Aβ positivity due to many analysis variables. Additionally, conventional data-fusion methods have difficulty in capturing complex relationships between multiple data, including brain imaging, genetic, demographic, and clinical data (Hu et al., 2019). In this regard, we need to apply deep collaborative learning, which can link the whole data simultaneously to develop a complete understanding of the Aβ pathophysiological cascade during the preclinical phase (Song and Chai, 2018; Hu et al., 2019). This novel method has been demonstrated to improve the generalization and robustness of brain research (Song and Chai, 2018).

The present study was designed to explore the distinctive effect of the APOE ε4 carrier status on the associations between subthreshold, positive Aβ deposition, FC, and cognitive performance during the preclinical phase. We have depicted the impact of the APOE ε4 carrier status on the pathway from the subthreshold and positive Aβ deposits to cognitive performance via the local and remote FC, as illustrated in the schematic in Figure 5. Finally, this new understanding based on the APOE ε4 carrier status should help to improve predictions of the impact of the earliest Aβ accumulation on the progression of AD with a compliment to the aforementioned limitations.

The datasets generated or analyzed during the current study are not publicly available because of the Patient Data Management Protocol of Yeouido Saint Mary's Hospital, but are available from the corresponding author upon reasonable request.

The studies involving human participants were reviewed and approved by Institutional Review Board of the Catholic University of Korea. The patients/participants provided their written informed consent to participate in this study.

DK contributed to conceptualization, methodology, data curation, writing the original draft, visualization, formal analysis, and funding acquisition. S-MW contributed to methodology, data curation, and writing, reviewing, and editing the manuscript. YU contributed to software and investigation. N-YK contributed to methodology and data curation. CL contributed to conceptualization and supervision. HL contributed to conceptualization, methodology, supervision, project administration, funding acquisition, and writing, reviewing, and editing the manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by the National Research Foundation of Korea grants funded by the Korean government (Ministry of Science and ICT) (Nos. 2019R1A2C2009100 and 2019R1C1C1007608). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.