Alzheimer’s disease (AD) is a progressive neurodegenerative disease that usually has a slow progression and long course. The typical pathological feature of AD is extracellular β-amyloid protein (Aβ) deposition, which starts a decade or more before the onset of illness and appears to be a trigger of the pathological cascade of events leading to AD dementia. Observations suggest that Aβ deposition has reached a peak 10–12 years before the onset of AD symptoms, it is hypothesized that Aβ initiates tangle formation and neuronal cell death (Hardy and Allsop, 1991; Klunk et al., 2006). Recently, biomarkers have been emphasized in the diagnosis of AD. The National Institute on Aging and Alzheimer’s Association (NIA-AA) Research Framework state that Aβ changes, pathologic tau, and neurodegeneration (ATN) comprise the diagnostic standard of AD and highlight the importance of neuroimaging and fluid biomarkers for the accurate diagnosis of AD (Jack et al., 2018).

β-amyloid deposition accumulates early as disease progresses, and varies among brain regions, including deposition in some key regions which mediate cognition (Grimmer et al., 2009). The brain regions susceptible to Aβ accumulation comprise large areas of the medial and lateral association cortex in amyloid-positive individuals without dementia (Palmqvist et al., 2017). The posterior cingulate and the frontal and parietal cortices are most commonly regions affected early in AD and in mild cognitive impairment (MCI) due to AD (Kemppainen et al., 2006, 2007), which is consistent with other studies including post-mortem evaluations (Klunk et al., 2004; Driscoll et al., 2012). The presence of Aβ deposition in different brain regions at different stages may be associated with inconsistencies in the effects of Aβ on spatial areas of the brain, which may reflect regional differences in susceptibility to AD pathology. More recently, Aβ deposition in the cerebral cortex has been shown to have a hierarchical organization in elderly cognitively normal individuals, with four Aβ clusters based on spatial features (Sepulcre et al., 2017). It is uncertain if this hierarchical clustering organization of cognitively normal elderly is present in symptomatic AD patients and whether it reflects the spatial distribution of AD pathological changes.

β-amyloid is a critical hallmark in AD diagnosis whereas ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography (PET) is a strong predictor of progression from MCI to AD dementia (Landau et al., 2010). In AD dementia patients Aβ deposition is widespread but is present in many individual who have not cognitive symptoms and it has a weak association with cognitive decline (Klunk et al., 2004). Reduced glucose metabolism is used as an indicator of synaptic dysfunction and neurodegeneration caused by Aβ. Patients with AD typically show temporal and parietal hypometabolism on FDG PET imaging in patients with AD (Ossenkoppele et al., 2012), where gray matter atrophy is common. Some studies have attempted to correlate metabolic function with the presence of Aβ deposition. However, the spatial distribution of hypometabolism and Aβ deposition is different in both AD patients and normal older adults (La Joie et al., 2012). A few multimodal imaging studies using FDG-PET and amyloid PET approached the question of whether local amyloid plaque deposition is correlated with local levels of glucose metabolism. These studies showed that the correlation was discordant, and changed with disease stages (Li et al., 2008; Cohen et al., 2009; Altmann et al., 2015). The spatial distribution relationship between the Aβ deposition and metabolism in AD, if any, remains uncertain. Studies have shown that Aβ tends to be deposited in core brain regions with higher structural and functional connections (Daianu et al., 2015) which may also be areas with high glucose metabolism.

The purpose of the present study was to examine the spatial distribution and extent of Aβ deposits and glucose metabolism and verify whether regions with high Aβ deposition are regions with high glucose metabolism by using florbetapir (¹⁸F-AV-45) and FDG PET. Additionally, we attempted to characterize a hierarchical structure of amyloid burden and metabolism organization that contains meaningful information about regional covariance patterns in AD patients. We further explored the relationship between regional Aβ deposition and glucose metabolism in AD patients.

Characteristics of the study participants are given in Table 1. At the time of scan, patients with AD were on average 64.94 ± 8.14 years old. Forty-one percent of the patients were male and 88% had a Clinical Dementia Rating (CDR) score greater than one. There were no significant differences in chronic diseases like hypertension, type 2 diabetes mellitus and hyperlipidemia between these two groups.

This study examined the spatial distribution and hierarchical structure of amyloid burden and metabolism organization in AD patients. Several clear findings about the relationship between Aβ and energy metabolism are presented here. First of all, from the spatial distribution view, cortical regions with highest florbetapir binding like cingulate gyrus, precuneus, lingual gyrus, frontal and parietal areas, were also the regions with high glucose metabolism. This is consistent with several previous studies which described the topographic patterns of AD, MCI and HC groups (Klunk et al., 2004; Kemppainen et al., 2007; Rowe et al., 2007). Previous studies have shown that the cingulate gyrus and precuneus are the hub regions for structural and functional brain networks, which are central in brain communication and neural integration (van den Heuvel and Sporns, 2013). Frequent and massive information operations require high energy consumption. The posterior cingulate, precuneus and retrosplenial cortices together show the highest level of glucose use of any area of the cerebral cortex in humans (Gusnard and Raichle, 2001). These hub regions carry a large burden in everyday cognitive activities, making themselves prime targets for toxic metabolites accumulation like Aβ.

In terms of the magnitude of the change in Aβ deposition and glucose hypometabolism, some very interesting phenomena were found. Briefly speaking, the brain regions that changed the most in florbetapir binding and FDG uptake were not those regions with the highest absolute levels in those indicators. The percentage changes of amyloid deposition were greatest in the gyrus rectus and middle orbitofrontal gyrus, and the hypometabolism was greatest in the angular gyrus. The indicators of cingulate gyrus and precuneus ranked top by absolute magnitude, but they did not change the most from HC to AD. This is not surprising because these brain regions have already existed high deposition in the early stage of the disease (shown in Figure 1), and as the deposition of Aβ has a platform, it will not continue to quickly accumulate after a certain amount of accumulation. As a result, the final variability of these regions between AD and HC might probably not be as big as we thought. In contrast, the major indicators for determining the progression from mild to severe disease phase will be frontal and other temporal regions.

From the perspective of regional clustering, although the number of clustering of the two indicators was different, with three categories in florbetapir binding and five categories in FDG uptake, they were not that distinct. For example, Aβ cluster 1 and FDG cluster 1 are basically overlapped, mainly including the temporal lobe region such as hippocampus and parahippocampal, Pearson correlation coefficients between amyloid deposition and glucose metabolism of these two clusters was significant (r = −0.35). Another example, FDG cluster 5 is totally part of Aβ cluster 2 in anatomical location (r = −0.36). The similarity of location distribution and subsequent correlation analysis of different data clustering implies the internal relationship between these two indicators. Further, we also found Aβ deposit exerts a negative influence on energy metabolism not only in local areas, but also contralaterally distant brain areas. The underlying reason may include the following aspects. Aβ deposition appears to follow distinct pathways, spreading progressively through interconnected brain regions, rather than emerging from stochastic aggregation of Aβ in different brain areas over time (Heilbronner et al., 2013; Eisele and Duyckaerts, 2016; Condello and Stöhr, 2017). The distant brain areas with hypometabolism may be affected by propagated Aβ, the majority of which may be monomers or oligomers of Aβ. Oligmeric Aβ is reported to exert more toxic effects on neurons than fibrillar Aβ (Sun et al., 2015).

Decades before the onset of AD dementia, abnormal accumulation of insoluble amyloid proteins are detectable in the temporal lobe and association cortex (Villain et al., 2012; Grothe et al., 2017). It has been shown in vivo that Aβ deposits follow some degree of spatial specificity. In our study, we tried to describe the hierarchical spatial organization of Aβ pathology. With our hierarchical clustering analysis of amyloid PET data, we identified that the pattern of distribution of Aβ deposition in AD patients resembled the proposed Braak stages (Braak and Braak, 1991). The brain regions in the first cluster was closely linked with the amyloid pathology at the early stage, with areas of major changes being the fusiform, hippocampus, parahippocampal, rectus, lateral orbitofrontal, inferior temporal, and inferior occipital areas. The second cluster had almost closed spatial distribution with the amyloid pathology at the mid-stage, including cingulate gyrus, insular, lingual gyrus, and precuneus. The third cluster contained the majority of cortical areas, which correspond with amyloid pathology at late stage. The clustering results may suggest that some brain areas share similar pathological mechanisms, so that these areas are threatened by disease at the same stage.

This study helps us to comprehensively examine the pathological mechanism of AD from A multi-dimensional perspective, and researches about the pathological mechanism of Aβ from the perspective of energy metabolism are still not sufficient. Only a few multimodal imaging studies using FDG-PET and amyloid-PET approached the question of whether local amyloid plaque deposition is correlated with local levels of glucose metabolism. These studies showed that the correlation could be complex and changed with disease stages (Landau et al., 2012; Altmann et al., 2015). Some suggested that the amyloid deposition in MCI patients is associated with higher metabolism as a compensatory response (Cohen et al., 2009; Oh et al., 2014). However, negative correlations were observed between amyloid deposition and metabolism in AD patients (Landau et al., 2012; Grothe and Teipel, 2016), which is consistent with our observations. A mechanistic view linking accumulation of Aβ to the hypometabolism, however, has been lacking so far. The possible explanation underlying the association between Aβ and glucose metabolism may include insulin resistance (Neth and Craft, 2017; Kellar and Craft, 2020), mitochondrial dysfunction (involving TCA cycle and oxidative phosphorylation system), reactive oxygen species, apoptosis, inflammatory factors, excitotoxicity, glycation end products, hyper-activation of some protein kinases and so on (Devi et al., 2006; Chen and Zhong, 2013). Accumulating evidence suggests that mitochondrial dysfunction may play a fundamental role among these above pathways. Several in vitro studies posit that neurodegenerative disorders are associated with changes in mitochondrial dynamics and can be induced by Aβ that progressively accumulates within this organelle, acting as a direct toxin (Ferreira et al., 2010). Accumulation of the Aβ precursor protein, at mitochondrial membrane can cause mitochondrial dysfunction by blocking the translocation of other intra-mitochondrial molecules/proteins and disrupting the electron-transport chain (Sun et al., 2015). The Aβ localized in mitochondria can bind to two pro-apoptotic factors including Aβ-binding alcohol dehydrogenase and cyclophilin D, consequently increasing neurodegenerative cell death (Lustbader et al., 2004; Moura et al., 2010). Aβ induces activation of glutamate N-methyl-D-aspartate receptors and/or excessive release of calcium from endoplasmic reticulum that may underlie mitochondrial calcium dyshomeostasis thereby disturbing organelle functioning like energy conversion, and ultimately, damaging neurons (Ferreira et al., 2010).

There are limitations of our study. First, it is very important in the future to validate continuity and change in the AD progression by longitudinal studies in cohorts including MCI. Interrogation of a longitudinal dataset is also warranted to verify the hierarchical clustering results from our cross-sectional analyses. AD in this sample has likely been present longer than recorded, given the difficulty of identifying and documenting early cognitive changes. It remains unclear if apolipoprotein E gene is implicated in the AD-related effects of Aβ load patterns and this should be addressed in future studies.

In summary, we demonstrated that cortical regions with more Aβ accumulation were the regions with high glucose metabolism. The hierarchical clustering provides evidence that Aβ accumulation and glucose metabolism are region-specific and regions in the same cluster may be specifically affected in AD. Amyloid in each hierarchical category is significantly negatively correlated with metabolism in multiple categories supporting the hypothesis that Aβ deposition is an early event of the pathological process and relates to neurodegenerative changes of multiple brain regions.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

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

YC had full access to all of the data in the study and took responsibility for the integrity of the data and accuracy of the data analysis. YC and LA conceived the original idea for the study, supervised the conception, and revised and drafted the manuscript. YC, D-AZ, KX, XZ, QC, FS, DF, ZZ, and LA recruited the study population and conducted the neuropsychological tests. YC, FS, and LS analyzed the data. All authors read and approved the final 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.