Metabolic Asymmetry Relates to Clinical Characteristics and Brain Network Abnormalities in Alzheimer’s Disease

Abstract

Background:

Metabolic asymmetry has been observed in Alzheimer’s disease (AD), but different studies have inconsistent viewpoints.

Objective:

To analyze the asymmetry of cerebral glucose metabolism in AD and investigate its clinical significance and potential metabolic network abnormalities.

Methods:

Standardized uptake value ratios (SUVRs) were obtained from ¹⁸F-FDG positron emission tomography (PET) images of all participants, and the asymmetry indices (AIs) were calculated according to the SUVRs. AD group was divided into left/right-dominant or bilateral symmetric hypometabolism (AD-L/AD-R or AD-BI) when more than half of the AIs of the 20 regions of interest (ROIs) were < –2SD, >2SD, or between±1SD. Differences in clinical features among the three AD groups were compared, and the abnormal network characteristics underlying metabolic asymmetry were explored.

Results:

In AD group, the proportions of AD-L, AD-R, and AD-BI were 28.4%, 17.9%, and 18.5%, respectively. AD-L/AD-R groups had younger age of onset and faster rate of cognitive decline than AD-BI group (p < 0.05). The absolute values of AIs in half of the 20 ROIs became higher at follow-up than at baseline (p < 0.05). Compared with those in AD-BI group, metabolic connection strength of network, global efficiency, cluster coefficient, degree centrality and local efficiency were lower, but shortest path length was longer in AD-L and AD-R groups (p < 0.05).

Conclusion:

Asymmetric and symmetric hypometabolism may represent different clinical subtypes of AD, which may provide a clue for future studies on the heterogeneity of AD and help to optimize the design of clinical trials.

Keywords

Alzheimer’s disease cerebral glucose hypometabolism graph theory metabolic asymmetry metabolic connectivity Mini-Mental State Examination

INTRODUCTION

Asymmetries of brain structure and function are common phenomena in human beings. The morphological asymmetry is generally considered to be associated with normal cognition, including memory, attention, and visuospatial function in normal populations [1 –3]. However, the presence of asymmetry between the cerebral hemispheres may also reflect pathological changes and may serve as a biomarker or a risk factor for neurodegenerative diseases [4, 5]. For example, semantic variant primary progressive aphasia is characterized by asymmetric atrophy in the anterior temporal lobes[6].

Alzheimer’s disease (AD) is the most common form of neurodegenerative disease. There have been several studies demonstrating regional hemispheric asymmetries in brain structure and function in AD, including cortical volume [7 –9], cortical surface area [4, 10], cortical thickness [3], white matter properties [8, 11], and functional connectivity [12]. For example, the volume of the left hemisphere was found to be smaller than the right hemisphere in AD patients, and faster left hemisphere degeneration was associated with poorer performance on neuropsychological tests [4 , 10].

However, previous studies had paid little attention to metabolic asymmetry and had inconsistent findings regarding this feature in AD [13]. Some researchers reported that AD patients showed symmetrical hypometabolism on ¹⁸F-FDG PET images in the temporal lobe, parietal lobe, and posterior cingulate gyrus [14]. In contrast, a longitudinal study based on regional cerebral metabolic rates for glucose (rCMRgl) found bilateral hemispheric asymmetry in AD, and this feature became more significant with disease progression [15]. Moreover, AD showed altered topological properties and lateralized deficits in connectome topologies on magnetic resonance imaging [11], but related research on asymmetric metabolism of AD was scarce [16]. Thus, the first aim of the present work was to investigate the proportion of metabolic asymmetry in AD. To this end, we calculated the asymmetry index in a relatively large cohort. The second aim was to observe whether significant differences in clinical manifestations exist between AD individuals with asymmetric and symmetric hypometabolism. Finally, we investigated the discrepancies in brain metabolic connectivity among different groups to discover the abnormal network underlying the asymmetry in AD.

METHODS

Participants

A total of 389 subjects, including 54 healthy controls (HC) and 335 AD patients, were enrolled in this study, and all of them underwent ¹⁸F-FDG PET imaging at the PET center of Huashan Hospital (Shanghai, China). The clinical diagnosis of AD was made according to the 2011 National Institute on Aging (NIA)-Alzheimer’s Association (AA) criteria [17], and amyloid-β (Aβ) PET imaging showed abnormal deposits of senile plaques in the cerebral cortex in each patient. Healthy controls had no complaint or family history of cognitive impairment and had normal cognitive test results. Cases with any history of cerebrovascular disease, encephalitis, head injury, severe psychiatric illness, alcohol abuse, metabolic disease, hormone use within the previous 6 months, or with abnormal findings on magnetic resonance imaging or computer tomography (CT) were excluded from the study. All of the above subjects are right-handed. Participants in HC and AD groups were matched for age (60.1±7.8 years versus 65.6±9.8 years [mean±SD], p = 0.62) and sex (22:32 versus 147:188 [male: female], p = 0.67). The age of onset and disease duration in the AD group were 59.8±10.2 years and 2.5±1.9 years respectively. In addition, Mini-Mental State Examination (MMSE) scores (19.0±6.7) were obtained at the same time as ¹⁸F-FDG PET scans in 259 ADpatients.

The institutional review board of Huashan Hospital approved the study (No. KY2018-318), and all participants or their statutory guardians provided written informed consent.

Image acquisition and processing

Patients had ceased anti-AD medications (if used) for at least 24 h prior to scanning. Whole-brain ¹⁸F-FDG PET images of all subjects were acquired using a Siemens mCT Flow PET/CT scanner (Siemens, Erlangen, Germany) in three-dimensional (3D) mode. After fasting for more than 6 h and with normal blood glucose level, all subjects received an intravenous injection of 185 Mbq ¹⁸F-FDG and rested for 1 h in a dimly lit and quiet room. A CT transmission scan was performed first for attenuation correction and the emission scan lasted for 10 min. All participants remained awake during the course of tracer uptake and scanning procedure. Reconstruction was performed using the 3D Ordered Subset Expected Maximizedmethod.

PET images were spatially normalized to the Montreal Neurological Institute (MNI) space and smoothed with a 10-mm full-width at half-maximum (FWHM) Gaussian filter to increase the signal-to-noise ratio using statistical parameter mapping (SPM) software (SPM8, Wellcome Department of Imaging Neuroscience, Institute of Cognitive Neurology, London, UK). Then, based on standardized Automated Anatomic Labeling (AAL) template [18], ¹⁸F-FDG PET images of the left and right cerebral hemispheres were divided into 90 regions of interest (ROIs). The standardized uptake value ratios (SUVRs) of the 90 ROIs were calculated with the cerebellum as the reference region.

Asymmetry index

The calculation of asymmetry indices (AIs) was performed to assess the metabolic differences between brain regions in the left and right hemispheres. In order to eliminate the influence of inherent asymmetry, SUVR value of each ROI in the left or right hemisphere was divided by the corresponding average SUVR value of the HC group to obtain a SUVR ratio (SUVRr) value for each subject. The AI was then computed using the formula: 2*(SUVRr_L-SUVRr_R)/(SUVRr_L+SUVRr_R), ensuring that the average value of the AIs of each ROI in HC group was 0.

To assess the degree of asymmetry of glucose metabolism in AD, we calculated the AIs for key brain regions known to be affected by AD (20 pairs in total): Frontal lobe (Frontal_Sup, Frontal_Sup_Orb, Frontal_Mid, Frontal_Mid_Orb, Frontal_Inf_Oper, Frontal_Inf_Tri, Frontal_Inf_Orb, Frontal_Sup_Medial, Frontal_Mid_Orb), Parietal lobe (Parietal_Sup, Parietal_Inf, Angular, Precuneus), Temporal lobe (Temporal_Sup, Temporal_Pole_Sup, Temporal_Mid, Temporal_Pole_Mid, Temporal_Inf, Hippocampus), and Cingulum_Post [13, 19]. We classified AD patients based on the degree of deviation of their AIs from the corresponding AIs of HC group. Specifically, AD patients with more than half of the AIs of the 20 ROIs < –2SD or >2SD were considered to have left-dominant hypometabolism (AD-L) or right-dominant hypometabolism (AD-R), and those with more than half of the AIs between±1SD were considered to have bilateral symmetric hypometabolism (AD-BI). The remaining patients were considered to be in an intermediate state (AD-IM).

Graph theory analysis

To analyze the brain glucose metabolism networks in the HC group and AD groups, we constructed a collection of 90 brain regions (nodes) and examined the links (edges) between them. To determine the strength of the connection, we calculated the correlation coefficient between each pair of nodes using Pearson’s correlation method in an inter-subjective manner, and then took the absolute value of the correlation coefficient to define the weight of the edge [20].

To characterize the topology of brain network, we used both global and nodal properties. We applied graph theoretical approaches using the GRETNA toolbox [21] to identify metabolic connectivity patterns in different groups. Global properties, including global efficiency (E_global) and clustering coefficient (C), were used to assess the network’s overall characteristics. The global efficiency measures the network’s ability to transfer information efficiently across all nodes, while the clustering coefficient of a given node indicates the probability that its neighboring nodes are connected to each other [21]. These properties were defined as follows: $E_{global} (G) = \frac{1}{N (N - 1)} \sum_{i \neq J \in G} \frac{1}{L_{i, j}}$

and $C = \frac{1}{n} \sum_{i \in N} C_{i} = \frac{1}{n} \sum_{i \in N} \frac{2 t_{i}}{k_{i} (k_{i} - 1)}$

Where N is the set of all nodes in a graph (G), n is the number of nodes, L_i, j is the shortest path length between nodes i and j, C_i is the clustering coefficient of node i, t_i is the actual number of edges between adjacent nodes of node i, and k_i is the number of adjacent nodes of node i. When k_i < 2, C_i = 0 [22].

To further analyze the brain glucose metabolism networks, we also examined the nodal properties, including degree centrality (DC), local efficiency (LE), and shortest path length (SPL) [11 , 23]. DC for a given node reflects its ability to communicate information within the functional network, while the LE measures how efficiently the communication flows among the node’s immediate neighbors when it is removed. SPL quantifies the mean distance or routing efficiency between the node and all other nodes in the network [21].

These nodal properties were defined as follows: ${DC}_{i} = \sum_{j \in N} w_{i, j}$

where w_ij is the weight of the connection between nodes i and j, and N is the number of network nodes. When i= j, w_ij= 0. $LE (G_{i}) = \frac{1}{N_{G} (N_{G} - 1)} \sum_{j, k \in G} \frac{1}{L_{j, k}}$

where N_G is the number of nodes in the subgraph G_i, and the subgraph G_i is the network that contains the neighboring nodes of node i [24]. ${SPL}_{ij} = \sum_{a_{uv} \in g_{i \leftrightarrow j}} f (w_{uv})$

where f is a map from weight to length and g_i↔j is the shortest weighted path between i and j [22].

Statistical analysis

Statistical analysis of the demographic and clinical data was performed using GraphPad Prism 8.0 (GraphPad Software Inc.). Sex differences among groups were compared using the chi-squared test. Other data such as the disease duration, age of onset, MMSE scores, and MMSE decline rate of the three AD groups were analyzed by one-way analysis of variance (ANOVA) with Bonferroni test. Correlations between age of onset or sex and the absolute values of AIs were calculated using Spearman’s correlation analysis. The difference in network connections and properties were assessed using permutation test (family wise errors-corrected, 5000 times) for statistical significance [25]. In all tests, statistical significance was defined as p < 0.05.

RESULTS

Metabolic characteristics in HC and AD group

In 32/45 pairs of ROIs in HC group, we found that glucose metabolism in the left hemisphere was slightly lower than in the right hemisphere (Supplementary Figure 1), but there was no significant difference. Among 335 patients with AD, 155 patients (46.3%) presented with asymmetric hypometabolism including 95 AD-L patients (28.4%) and 60 AD-R patients (17.9%), and 62 patients (18.5%) were assigned to the AD-BI group.

Fifteen AD patients were followed up with ¹⁸F-FDG PET imaging for 1-2 years. Among them, 3 patients converted from the AD-BI group to the AD-IM group and 2 patients converted from the AD-IM group to the AD-L/R group (Supplementary Table 1). The absolute values of AIs in 10 pairs of ROIs (including Frontal_Sup, Frontal_Sup_Orb, Frontal_Mid, Frontal_Inf_Orb, Frontal_Sup_Medial, Frontal_Mid_Orb, Temporal_Mid, Temporal_Pole_Mid, Temporal_Inf and Hippocampus) became higher at follow-up than at baseline (p < 0.05), and did not change significantly in the remaining 10 pairs of ROIs (p > 0.05) (Fig. 1).

Fig. 1

The absolute values of AIs at baseline and at follow-up. The absolute values of AIs in 10 pairs of ROIs (including Frontal_Sup, Frontal_Sup_Orb, Frontal_Mid, Frontal_Inf_Orb, Frontal_Sup_Medial, Frontal_Mid_Orb, Temporal_Mid, Temporal_Pole_Mid, Temporal_Inf, and Hippocampus) became higher at follow-up than at baseline (p < 0.05), but did not change significantly in the remaining 10 pairs of ROIs (p > 0.05). AIs, asymmetry indices.

Clinical differences among three groups of AD

As shown in Table 1, there was no difference in disease duration (p = 0.884, Bonferroni) or sex (p = 0.180, chi-squared test) among the three AD groups. Age at imaging was younger (p < 0.001, Bonferroni), and age of onset was earlier (p < 0.0001, Bonferroni) in the AD-L and AD-R groups than in the AD-BI group. The proportions of early-onset Alzheimer’s disease (EOAD, age of onset <65 years) [26] were higher in AD-L (78.9%) and AD-R groups (81.7%) than in AD-BI group (56.5%) (p = 0.001, chi-squared test). The MMSE scores were lower (p < 0.0001, Bonferroni) and the MMSE decline rate [(30-MMSE)/ disease duration] was faster (p = 0.002, Bonferroni) in AD-L and AD-R groups than in AD-BI group. There was no difference in age of onset, proportion of EOAD, MMSE scores, and the MMSE decline rate between AD-L and AD-R groups (p > 0.05).

Table 1

Demographic and clinical differences among AD-BI, AD-L and AD-R groups

	AD-BI group	AD-L group	AD-R group	p ^a	Post hoc Bonferroni test
age at imaging	68.2±8.0	62.1±9.0	59.0±9.2	<0.001^b	AD-BI versus AD-L^*
					AD-BI versus AD-R^***
disease duration	2.4±1.9	2.5±1.8	2.5±1.9	0.884^b	–
age of onset	63.1±10.4	57.8±8.8	56.4±8.2	<0.0001^b	AD-BI versus AD-L^*
					AD-BI versus AD-R^***
sex (male: female)	30: 32	44: 51	20: 40	0.180^d	–
EOAD^c (%)	56.5%	78.9%	81.7%	0.0001^d	–
MMSE scores	23.1±4.1 (n = 54)	14.9±6.6 (n = 72)	16.7±6.4 (n = 49)	<0.0001^b	AD-BI versus AD-L^****
					AD-BI versus AD-R^****
MMSE decline rate^e	4.9±4.8	9.0±7.1	8.0±6.1	0.002^b	AD-BI versus AD-L^**
					AD-BI versus AD-R^*

Data are presented as mean±1 SD. ^aComparison among AD-BI, AD-L, and AD-R groups; ^bone-way ANOVA with differences between groups examined post hoc with Bonferroni test; ^cearly-onset Alzheimer’s disease; ^dchi-squared test; ^e[(30-MMSE)/ disease duration]; MMSE, Mini-Mental State Examination; AD-BI, Alzheimer’s disease with bilateral symmetric hypometabolism; AD-L, Alzheimer’s disease with left-dominant hypometabolism; AD-R, Alzheimer’s disease with right-dominant hypometabolism; ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001.

There was negative correlation between age of onset and the absolute values of AIs in most ROIs in the Frontal lobe (Frontal_Sup, Frontal_Sup_Orb, Frontal_Mid, Frontal_Mid_Orb, Frontal_Inf_Oper, Frontal_Inf_Tri, Frontal_Inf_Orb, Frontal_Sup_Medial), Parietal lobe (Parietal_Sup, Parietal_Inf, Angular, Precuneus) and Temporal lobe (Temporal_Sup, Temporal_Mid, Temporal_Inf) (p: 0.000∼0.042, R: –0.287∼–0.138), but not in Frontal_Med_Orb, Temporal_Pole_Sup, Temporal_Pole_Mid, Hippocampus, or Cingulum_Post in AD patients (p > 0.05) (supplementary Figure 2). In addition, there was no correlation between sex and the absolute values of AIs in all the 20 ROIs (p > 0.05, chi-squared test).

Brain metabolic network differences among three groups of AD

We found no significant difference in the metabolic connection strength, as measured by adjacency matrix, global and nodal properties between HC group and AD-BI group (p > 0.05). Compared with AD-BI group, both AD-L and AD-R groups showed lower metabolic connectivity mainly associated with 11 pairs of ROIs (including Cingulum_Post, Hippocampus, Parietal_Sup, Parietal_Inf, Angular, Precuneus, Temporal_Sup, Temporal_Pole_Sup, Temporal_Mid, Temporal_Pole_Mid, Temporal_Inf) in the bilateral parietal, temporal lobe and posterior cingulate. Additionally, it was found that in AD-L and AD-R groups but not in HC and AD-BI groups, the metabolic connection strength between hemispheres was stronger than that within the hemisphere with more severe hypometabolism (i.e., the left hemisphere of AD-L group and the right hemisphere of AD-R group), but was weaker than that within the hemisphere with less severe hypometabolism (Fig. 2, Supplementary Figure 3).

Fig. 2

The metabolic connectivity strength presented by adjacency matrix in HC and AD-BI, AD-L and AD-R groups. There were no significant differences in the metabolic connectivity strength between HC group and AD-BI group (p > 0.05). Compared with AD-BI group, the matrices of correlation coefficients in AD-L and AD-R groups showed lower metabolic connectivity mainly associated with 11 pairs of ROIs (including Cingulum_Post, Hippocampus, Parietal_Sup, Parietal_Inf, Angular, Precuneus, Temporal_Sup, Temporal_Pole_Sup, Temporal_Mid, Temporal_Pole_Mid, Temporal_Inf) in the bilateral parietal, temporal lobe, and posterior cingulate. HC, healthy control; AD-BI, Alzheimer’s disease with bilateral symmetric hypometabolism; AD-L, Alzheimer’s disease with left-dominant hypometabolism; AD-R, Alzheimer’s disease with right-dominant hypometabolism.

Compared with HC group and AD-BI group, the global properties of metabolic networks (global efficiency and cluster coefficient) were significantly lower in AD-L and AD-R groups (p < 0.05) (Fig. 3). As for the nodal properties, the degree centrality and local efficiency of frontal lobe, temporal lobe and parietal lobe were lower, but the shortest path length was longer in AD-L and AD-R groups than those in HC and AD-BI group. There was no difference between AD-L and AD-R groups (Fig. 4).

Fig. 3

Differences of average connection and global properties (global efficiency and cluster coefficient) among HC, AD-BI, AD-L, and AD-R groups. There were no significant differences in average connection strength and global properties between HC group and AD-BI group or between AD-L and AD-R groups, but AD-L and AD-R groups had lower average connection strength and global properties than AD-BI group (A-C). HC, healthy control; AD-BI, Alzheimer’s disease with bilateral symmetric hypometabolism; AD-L, Alzheimer’s disease with left-dominant hypometabolism; AD-R, Alzheimer’s disease with right-dominant hypometabolism; ^**p<0.01, ^***p<0.001, ns, no significance.

Fig. 4

Differences in degree centrality (DC), local efficiency (LE) and shortest path length (SPL) among HC, AD-BI, AD-L, and AD-R groups. The DC and LE of the frontal lobe, temporal lobe and parietal lobe were lower, and the SPL was longer in AD-L and AD-R groups than in HC group and AD-BI group, but there was no difference between HC group and AD-BI group or between AD-L and AD-R groups. HC, healthy control; AD-BI, Alzheimer’s disease with bilateral symmetric hypometabolism; AD-L, Alzheimer’s disease with left-dominant hypometabolism; AD-R, Alzheimer’s disease with right-dominant hypometabolism.

DISCUSSION

In the present study, we analyzed the asymmetric hypometabolism in AD and its clinical significance, as well as the potential network abnormalities by graph theory among AD-BI, AD-L, and AD-R groups. The main findings of this study were as follows. First, compared with HC group, 46.3% of AD patients showed asymmetric hypometabolism, and left-dominant hypometabolism was more common. The metabolic asymmetry became more significant with disease progression. Second, compared with AD-BI group, the age of onset was earlier and the rate of decline of MMSE scores was faster in AD-L and AD-R groups. Moreover, graph theory analysis showed that the metabolic connection strength of network was weaker, the global efficiency and the clustering coefficient were lower, the shortest path length was longer and the degree centrality, local efficiency of the frontal lobe, temporal lobe, and parietal lobe were lower in AD-L and AD-R groups compared with AD-BI group. In a word, about half of AD patients showed metabolic asymmetry and presented with younger age of onset, faster cognitive decline, and worse metabolic connectivity in bilateral cerebral hemispheres.

In this study, we found that glucose metabolism in the left hemisphere was slightly lower than that in the right hemisphere in HC group. This was consistent with the characteristic of brain structure in HC subjects reported in previous studies, which showed that the regional volume of the left hemisphere (especially the temporal lobe including the hippocampus) was smaller on average than that of the right hemisphere [1 , 28]. It is worth noting that the current study used the SUVRr value, which was calculated by dividing the SUVR value of each ROI in the left or right hemisphere of each subject by the corresponding average SUVR value derived from the HC group, to calculate the AIs. As a result, the AIs of the HC group were all 0. It was helpful to eliminate the influence of the congenital brain asymmetry and to accurately identify the true state of abnormal metabolic asymmetry in AD patients.

Previous studies had found that, in terms of metabolism, patients with AD had asymmetric rCMRgl decreases with prominent left-sided metabolic reduction [13 , 30]. Consistently, in terms of structure, the degree and rate of cortical atrophy in the left and right hemispheres of AD patients were different, with the faster rate of gray matter atrophy in the left hemisphere [10]. Therefore, the finding that the left hemisphere presented more severe hypometabolism in AD patients may be due to the fact that AD patients had more pronounced neuronal damage in the left brain. Furthermore, a previous study suggested a prominent asymmetric pattern of cerebral Aβ deposition by ¹⁸F-florbetaben PET scans in patients with mild cognitive impairment (MCI) [31]. Another study using tau protein imaging observed that Aβ+ patients in preclinical AD had an asymmetric cortical tau deposition pattern associated with younger age and decreased cortical thickness [32]. In addition, AD patients with apolipoprotein E type 4 allele (APOE ɛ4) presented increased left-right asymmetry in the parietal lobe compared to those without APOE ɛ4 [33], this may reflect that APOE ɛ4 plays a role in the clinically heterogeneity of AD. Nevertheless, further studies are needed to elucidate the interaction between asymmetric metabolism, structural atrophy, abnormal protein deposition and risk genes.

Our study showed that the proportion of EOAD was higher in the AD-L and AD-R groups than that in the AD-BI group, and the absolute values of AIs in most ROIs in the frontal lobe, temporal lobe and parietal lobe negatively correlated with age of onset in AD. Consistent with these findings, a previous study showed patients with EOAD had lower metabolism in the ROI of left posterior cingulate compared with late-onset Alzheimer’s disease (LOAD) at a similar disease stage, and EOAD showed greater cortical atrophy, more cognitive deficits (as tested by MMSE), and more rapid progression than LOAD [26]. Although previous studies did not investigate the asymmetry of EOAD or LOAD, taking their and our findings together, we can speculate that AD with asymmetric hypometabolism are more likely to be EOAD with more cognitive impairments. In addition, our finding that the AD-L/AD-R groups had worse cognitive performance than the AD-BI group was further supported by previous findings that AD patients with asymmetric rCMRgl decline, especially the left-dominant decline, were closely associated with cognitive decline [13, 15]. In general, EOAD and cognitive deterioration related to asymmetric features of glucose metabolism, and metabolic lateralization may represent worse cognitive function[34 –37].

Regarding the influence of sex, a previous study suggested that the glucose metabolism rate of the left cerebral hemisphere was lower than that of the right hemisphere in male AD patients, but not in the females [38]. However, consistent with the finding in our study, the other study did not report the sex difference in cortical metabolic asymmetry [39]. Therefore, further studies are needed to determine the relationship between sex and metabolic asymmetry.

The longitudinal results of fifteen AD patients in the present study showed more pronounced asymmetry in 10 pairs of ROIs at follow-up than at baseline. This finding was in line with a similar longitudinal study of metabolic asymmetry in MCI patients [15]. In contrast, in another follow-up study with ¹⁸F-FDG imaging, asymmetric patterns of rCMRgl decline were observed in MCI patients, particularly pronounced in the left hemisphere, but diminished with disease progression [29]. Therefore, further research with a larger scale is needed to validate the variable metabolic model in AD.

So far, the potential brain network abnormalities of bilateral hemispheric metabolic asymmetry in AD have not been fully explored. To the best of our knowledge, this was the first study to reveal the metabolic connectivity network underlying asymmetric hypometabolism in AD by graph theory. Compared with AD-BI group, both AD-L and AD-R groups presented significantly weaker metabolic connection strength showed by the adjacency matrix, in which one salient feature was that the metabolic connection abnormalities were mainly driven by 11 pairs of ROIs in the bilateral parietal, temporal lobe and posterior cingulate. Considering the short duration of AD patients in this study, this may reflect the higher vulnerability of parietal, temporal lobe and posterior cingulate relative to frontal lobe in the early stage of AD [13, 19]. The other feature showed by the adjacency matrix was that the metabolic connection strength was weaker within the more affected hemisphere than that between hemispheres and that within the less affected hemisphere. A previous small-sample study showed that 42.9% patients with AD presented asymmetric tau deposition in bilateral cortices [40], which suggested tau deposition might be a pathological process detrimental to connectivity function and should be validated longitudinally in a research with a bigger sample size.

Compared with the AD-BI group, both AD-L and AD-R groups presented significantly worse global properties of the metabolic network, which suggested that the corticocortical communication function was more severely destroyed in AD patients with asymmetric hypometabolism [35 , 41], and may explain our finding of worse performance on MMSE in AD-L and AD-R groups. Our study showed that asymmetric hypometabolic patients are more likely to be EOAD, and a previous study found the decline of metabolic connectivity including global efficient and clustering coefficient in patients with EOAD but not in LOAD, which further verified that asymmetric hypometabolic patients may have worse connectivity function [42]. In addition, the worse nodal properties of the frontal, temporal, and parietal lobes in AD-L and AD-R groups than in AD-BI group suggested that the link of the brain nodes most associated with AD became weaker and the metabolic connectivity function became worse in the asymmetric hypometabolic patients [22, 41]. Previous studies had found that AD patients showed decreased network connectivity compared to HC subjects [43, 44]. However, our study showed that the AD-BI group had consistent metabolic connectivity with the HC group. We speculate that the findings of decreased network connectivity in AD patients may result from the fact that about half of the AD patients presented with asymmetric hypometabolism. Our findings suggest that AD patients with symmetric and asymmetric hypometabolism should be analyzed separately in future studies.

There are some limitations in the current study. Firstly, this cross-sectional study had limited information of MMSE scores or other more detailed cognitive measures. Secondly, the analysis of ¹⁸F-FDG PET imaging did not combine with the magnetic resonance imaging so that the partial volume effect may impact the results of our study to some extent. Finally, potential class analysis, multiple comparison and other methods are needed to control variation between individuals for further confirming the findings of this study.

In conclusion, about half of AD patients showed asymmetric hypometabolism, with younger age of onset, faster cognitive deterioration, and worse metabolic connectivity in comparison with patients with symmetric hypometabolism. Moreover, the asymmetry of metabolism became more prominent as the progression of disease. Asymmetric and symmetric hypometabolism may represent different clinical subtypes of AD and it may play an important role in reflecting the heterogeneity of AD and help to optimize the design of clinical trials.

Footnotes

ACKNOWLEDGMENTS

The authors have no acknowledgments to report.

FUNDING

This work was supported by the National Natural Science Foundation of China (Nos.81771483/81902282/82021002/81671239/81702229/81971641/82272039), Youth Medical Talents - Medical Imaging Practitioner Program by Shanghai Municipal Health Commission and Shanghai Medical and Health Development Foundation (SHWRS(2020)_087), Medical Innovation Research Project of Shanghai Science and Technology Commission (No.21Y11903300), Research Project of Shanghai Health Commission (No.2020YJZX0111), Clinical Research Plan of SHDC (No.SHDC2020CR1038B), Science and Technology Innovation 2030 Major Projects (No.2022ZD0211600).

CONFLICT OF INTEREST

The authors have no conflict of interest to declare.

DATA AVAILABILITY

The data supporting the findings of this study are available within the article and its supplementary material.

The supplementary material is available in the electronic version of this article: .

References

Dennis

, Thompson

(2013) Mapping connectivity in the developing brain. Int J Dev Neurosci 31, 525–542.

Goldberg

, Roediger

, Kucukboyaci

, Carlson

, Devinsky

, Kuzniecky

, Halgren

, Thesen

(2013) Hemispheric asymmetries of cortical volume in the human brain. Cortex 49, 200–210.

Kim

, Lee

, Kim

, Roh

, Kim

, Seo

, Kim

, Jeon

, Lee

, Heilman

, Na

(2012) Cortical asymmetries in normal, mild cognitive impairment, and Alzheimer’s disease. Neurobiol Aging 33, 1959–1966.

Thompson

, Hayashi

, De Zubicaray

, Janke

, Rose

, Semple

, Herman

, Hong

, Dittmer

, Doddrell

, Toga

(2003) Dynamics of gray matter loss in Alzheimer’s disease. J Neurosci 23, 994–1005.

Toga

, Thompson

(2003) Mapping brain asymmetry. Nat Rev Neurosci 4, 37–48.

Acosta-Cabronero

, Patterson

, Fryer

, Hodges

, Pengas

, Williams

, Nestor

(2011) Atrophy, hypometabolism and white matter abnormalities in semantic dementia tell a coherent story. Brain 134, 2025–2035.

Sarica

, Vasta

, Novellino

, Vaccaro

, Cerasa

, Quattrone

(2018) MRI asymmetry index of hippocampal subfields increases through the continuum from the mild cognitive impairment to the Alzheimer’s disease. Front Neurosci 12, 576.

Muller

, Greverus

, Dellani

, Weibrich

, Wille

, Scheurich

, Stoeter

, Fellgiebel

(2005) Functional implications of hippocampal volume and diffusivity in mild cognitive impairment. Neuroimage 28, 1033–1042.

Dhikav

, Duraisamy

, Anand

, Garga

(2016) Hippocampal volumes among older Indian adults: Comparison with Alzheimer’s disease and mild cognitive impairment. Ann Indian Acad Neurol 19, 195–200.

10.

Thompson

, Hayashi

, Dutton

, Chiang

, Leow

, Sowell

, De Zubicaray

, Becker

, Lopez

, Aizenstein

, Toga

(2007) Tracking Alzheimer’s disease. Ann N Y Acad Sci 1097, 183–214.

11.

Yang

, Zhong

, Zhou

, Wei

, Wang

, Nie

(2017) The abnormality of topological asymmetry between hemispheric brain white matter networks in Alzheimer’s disease and mild cognitive impairment. Front Aging Neurosci 9, 261.

12.

Wang

, Wang

, Zhang

, McHugh

, Sun

, Li

, Yang

(2015) Interhemispheric functional and structural disconnection in Alzheimer’s disease: A combined resting-state fMRI and DTI Study. PLoS One 10, e0126310.

13.

Murayama

, Ota

, Kasanuki

, Kondo

, Fujishiro

, Fukase

, Tagaya

, Sato

(2016) Cognitive dysfunction in patients with very mild Alzheimer’s disease and amnestic mild cognitive impairment showing hemispheric asymmetries of hypometabolism on ¹⁸F-FDG PET. Int J Geriatr Psychiatry 31, 41–48.

14.

Edison

, Archer

, Hinz

, Hammers

, Pavese

, Tai

, Hotton

, Cutler

, Fox

, Kennedy

, Rossor

, Brooks

(2007) Amyloid, hypometabolism, and cognition in Alzheimer disease: An [¹¹C]PIB and [¹⁸F]FDG PET study. Neurology 68, 501–508.

15.

Haxby

, Grady

, Koss

, Horwitz

, Heston

, Schapiro

, Friedland

, Rapoport

(1990) Longitudinal study of cerebralmetabolic asymmetries and associated neuropsychological patterns inearly dementia of the Alzheimer type. Arch Neurol 47, 753–760.

16.

Long

, Zhang

, Liao

, Jiang

, Qiu

(2013) Distinct laterality alterations distinguish mild cognitive impairment and Alzheimer’s disease from healthy aging: Statistical parametric mapping with high resolution MRI. Hum Brain Mapp 34, 3400–3410.

17.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

Jr , Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

18.

Rolls

, Joliot

, Tzourio-Mazoyer

(2015) Implementation of a new parcellation of the orbitofrontal cortex in the automated anatomical labeling atlas. Neuroimage 122, 1–5.

19.

Ishii

(2002) Clinical application of positron emission tomography for diagnosis of dementia. Ann Nucl Med 16, 515–525.

20.

Zhang

, Huang

, Jiao

, Liu

, Nie

, Liang

, Li

, Sun

, Feng

, Xu

, Shan

(2019) Modular architecture of metabolic brain network and its effects on the spread of perturbation impact. Neuroimage 186, 146–154.

21.

Wang

, Wang

, Xia

, Liao

, Evans

, He

(2015) GRETNA: A graph theoretical network analysis toolbox for imaging connectomics. Front Hum Neurosci 9, 386.

22.

Rubinov

, Sporns

(2010) Complex network measures of brain connectivity: Uses and interpretations. Neuroimage 52, 1059–1069.

23.

Wang

, Zuo

, Gohel

, Milham

, Biswal

, He

(2011) Graph theoretical analysis of functional brain networks: Test-retest evaluation on short- and long-term resting-state functional MRI data. PLoS One 6, e21976.

24.

Liang

, Jiang

, Zhang

, Duan

, Zhang

, Huang

, Sun

, Liu

, Dong

, Liu

, Tao

, Zhao

, Nie

, Chen

, Shan

(2018) Abnormal metabolic connectivity in rats at the acute stage of ischemic stroke. Neurosci Bull 34, 715–724.

25.

Pan

, Liu

, Li

, Zhang

, Zhao

, Zhou

, Nie

, Zhu

, Xu

, Liu

(2022) Retrosplenial cortex effects contextual fear formation relying on dysgranular constituent in rats. Front Neurosci 16, 886858.

26.

Rabinovici

, Furst

, Alkalay

, Racine

, O’Neil

, Janabi

, Baker

, Agarwal

, Bonasera

, Mormino

, Weiner

, Gorno-Tempini

, Rosen

, Miller

, Jagust

(2010) Increased metabolic vulnerability in early-onset Alzheimer’s disease is not related to amyloid burden. Brain 133, 512–528.

27.

Good

, Johnsrude

, Ashburner

, Henson

, Friston

, Frackowiak

(2001) A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage 14, 21–36.

28.

Iturria-Medina

, Perez

, Morris

, Canales-Rodriguez

, Haroon

, Garcia Penton

, Augath

, Galan Garcia

, Logothetis

, Parker

, Melie-Garcia

(2011) Brain hemispheric structural efficiency and interconnectivity rightward asymmetry in human and nonhuman primates. Cereb Cortex 21, 56–67.

29.

Weise

, Chen

, Kuang

, Savage

, Reiman

(2018) Left lateralized cerebral glucose metabolism declines in amyloid-beta positive persons with mild cognitive impairment. NeuroimageClin 20, 286–296.

30.

Derflinger

, Sorg

, Gaser

, Myers

, Arsic

, Kurz

, Zimmer

, Wohlschlager

, Muhlau

(2011) Grey-matter atrophy in Alzheimer’s disease is asymmetric but not lateralized. J Alzheimers Dis 25, 347–357.

31.

Frings

, Hellwig

, Spehl

, Bormann

, Buchert

, Vach

, Minkova

, Heimbach

, Kloppel

, Meyer

(2015) Asymmetries of amyloid-beta burden and neuronal dysfunction are positively correlated in Alzheimer’s disease. Brain 138, 3089–3099.

32.

Young

, Winer

, Younes

, Cody

, Betthauser

, Johnson

, Schultz

, Sperling

, Greicius

, Cobos

, Poston

, Mormino

(2022) Divergent cortical tau positron emission tomography patterns among patients with preclinical Alzheimer disease. JAMA Neurol 79, 592–603.

33.

Small

, Mazziotta

, Collins

, Baxter

, Phelps

, Mandelkern

, Kaplan

, La Rue

, Adamson

, Chang

, et al. (1995) Apolipoprotein E type 4 allele and cerebral glucose metabolism in relatives at risk for familial Alzheimer disease. JAMA 273, 942–947.

34.

Tijms

, Wink

, De Haan

, van der Flier

, Stam

, Scheltens

, Barkhof

(2013) Alzheimer’s disease: Connecting findings from graph theoretical studies of brain networks. Neurobiol Aging 34, 2023–2036.

35.

Tetzloff

, Graff-Radford

, Martin

, Tosakulwong

, Machulda

, Duffy

, Clark

, Senjem

, Schwarz

, Spychalla

, Drubach

, Jack

, Lowe

, Josephs

, Whitwell

(2018) Regional distribution, asymmetry, and clinical correlates of tau uptake on [¹⁸F]AV-1451 PET in atypical Alzheimer’s disease. J Alzheimers Dis 62, 1713–1724.

36.

Catani

, Allin

, Husain

, Pugliese

, Mesulam

, Murray

, Jones

(2007) Symmetries in human brain language pathways correlate with verbal recall. Proc Natl Acad Sci U S A 104, 17163–17168.

37.

Yoon

, Kim

, Jeong

, Kim

, Park

, Chun

(2021) Asymmetric amyloid deposition as an early sign of progression in mild cognitive impairment due to Alzheimer disease. Clin Nucl Med 46, 27–531.

38.

Siegel

Jr , Shihabuddin

, Buchsbaum

, Starr

, Haier

, Valladares Neto

(1996) Gender differences in cortical glucose metabolism in Alzheimer’s disease and normal aging. J Neuropsychiatry Clin Neurosci 8, 211–214.

39.

Small

, Kuhl

, Riege

, Fujikawa

, Ashford

, Metter

, Mazziotta

(1989) Cerebral glucose metabolic patterns in Alzheimer’s disease. Effect of gender and age at dementia onset. Arch Gen Psychiatry 46, 527–532.

40.

King

, Bodi

, Nolan

, Troakes

, Al-Sarraj

(2015) Assessment of the degree of asymmetry of pathological features in neurodegenerative diseases. What is the significance for brain banks? J Neural Transm (Vienna) 122, 1499–1508.

41.

Bullmore

, Sporns

(2009) Complex brain networks: Graph theoretical analysis of structural and functional systems. Nat Rev Neurosci 10, 186–198.

42.

Chung

, Yoo

, Kim

, Na

, Jeong

(2016) Glucose metabolic brain networks in early-onset vs. late-onset Alzheimer’s disease. Front Aging Neurosci 8, 159.

43.

Yao

, Zhang

, Lin

, Zhou

, Xu

, Jiang

(2010) Abnormal cortical networks in mild cognitive impairment and Alzheimer’s disease. PLoS Comput Biol 6, e1001006.

44.

Zhou

, Lui

(2013) Small-world properties in mild cognitive impairment and early Alzheimer’s disease: A cortical thickness MRI study. ISRN Geriatr 2013, 542080.