Morphological and Structural Network Analysis of Sporadic Alzheimer’s Disease Brains Based on the APOE4 Gene

Abstract

Background:

Alzheimer’s disease (AD) is an increasingly common type of dementia. Apolipoprotein E (APOE) gene is a strong risk factor for AD.

Objective:

Here, we explored alterations in grey matter structure (GMV) and networks in AD, as well as the effects of the APOE ɛ4 allele on neuroimaging regions based on structural magnetic resonance imaging (sMRI).

Methods:

All subjects underwent an sMRI scan. GMV and cortical thickness were calculated using voxel-based morphological analysis, and structural networks were constructed based on graph theory analysis to compare differences between AD and normal controls.

Results:

The volumes of grey matter in the bilateral inferior temporal gyrus, right middle temporal gyrus, right inferior parietal lobule, right limbic lobe, right frontal lobe, left anterior cingulate gyrus, and bilateral olfactory cortex of patients with AD were significantly decreased. The cortical thickness in patients with AD was significantly reduced in the left lateral occipital lobe, inferior parietal lobe, orbitofrontal region, precuneus, superior parietal gyrus, right precentral gyrus, middle temporal gyrus, pars opercularis gyrus, insular gyrus, superior marginal gyrus, bilateral fusiform gyrus, and superior frontal gyrus. In terms of local properties, there were significant differences between the AD and control groups in these areas, including the right bank, right temporalis pole, bilateral middle temporal gyrus, right transverse temporal gyrus, left postcentral gyrus, and left parahippocampal gyrus.

Conclusion:

There were significant differences in the morphological and structural covariate networks between AD patients and healthy controls under APOE ɛ4 allele effects.

Keywords

Alzheimer’s disease apolipoprotein E cortical thickness grey matter volume structural covariate network structural magnetic resonance imaging

INTRODUCTION

Alzheimer’s disease (AD) is the most common form of dementia. The clinical manifestations include memory and cognitive decline and personality changes, which seriously affect daily behavior [1]. With the intensification of population aging, the incidence of AD will increase, and the economic and social burden it causes will also increase day by day. According to the report, after 65 years of age, the incidence of AD doubles every five years [2]. The incidence in specific age groups has increased from less than 1% per year before the age of 65 years to 6% per year after the age of 85 years. The prevalence rate has increased from 10% after the age of 65 years to 40% after 85 years of age every year [3].

In AD, pathological changes may occur as early as 20 years before the onset of symptoms [4]. Currently, there is no treatment available to cure or delay the course of AD. Therefore, finding an effective method for the early identification of AD is important for early intervention and improvement of the clinical prognosis of AD. In recent years, more attention has been paid to the application of computer-aided diagnosis technology based on neuroimaging for the early diagnosis and disease classification of AD [5]. Compared with other commonly used imaging modalities, such as PET and functional magnetic resonance imaging (fMRI), structural magnetic resonance imaging (sMRI) can clearly show the anatomical structure of the brain. According to sMRI measurements, grey matter atrophy is the main indicator of neurodegenerative disease [6 –8]. Obvious grey matter atrophy has been observed in the AD brain, particularly in the bilateral hippocampus, medial temporal lobe, cingulate gyrus, and precuneus [9 –11]. Apolipoprotein E (APOE) is undoubtedly a risk factor for AD, but its association with clinical phenotypic and morphological changes in AD is still controversial [12, 13]. In addition, APOE genotype has been proved to effect the brain of AD patients [14 –17]. While less study was performed to explore the role of APOE ɛ4 allele effects on the grey matter atrophy in Chinese AD patients. Moreover, no study examined the role of APOE ɛ4 allele effects on related structure network in Chinese AD patients.

In this study, we grouped patients with AD (PT) and healthy controls (HC) based on APOE4 genotyping, collected sMRI image data of all subjects, and compared grey matter atrophy between the two groups using voxel-based morphological analysis methods. We further constructed a grey matter structure network based on graph theory analysis to explore the influence of morphological changes on the overall brain network, and provide a reference for early clinical diagnosis of AD.

MATERIALS AND METHODS

Subjects

A total of 42 PT and HC were recruited from July 2020 to March 2021, including 16 patients (7 men and 9 women) in the case group and 26 patients (15 men and 11 women) in the control group. The data from one subject was deleted because of excessive head movement (see data preprocessing). All patients with AD enrolled in this study were diagnosed with AD for the first time and had not received any related treatment. A clinical diagnosis of probable Alzheimer’s disease was established according to the National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer’s Disease and Related Disorders Association criteria [18]. The Mini-Mental State Examination (MMSE) was used as a neuropsychological evaluation index, while the Hamilton Depression Scale (HAMD) and Hamilton Anxiety Scale (HAMA) were used to exclude depression. Participants were excluded if they had non-AD dementia, cerebrovascular disease, brain trauma, intracranial infectious diseases, poisoning or metabolic abnormalities, or a history of cardiovascular and cerebrovascular surgery. None of the participants had a history of tobacco or alcohol use. All controls had no above diseases. All samples were recruited from cognitive disability clinic and neurological ward in Nanjing University Affiliated Wuxi No. 2 People’s Hospital.

Our study was performed in accordance with the Declaration of Helsinki. This study was approved by the hospital ethics committee (NO. Y-103). All subjects or their legal guardian were informed about the MRI examination and signed an informed consent form.

APOE genotyping

Fasting venous blood samples were collected, placed in EDTA anticoagulation tubes, and stored at –20°C. DNA was extracted using a DNA extraction kit (Promega) according to the instructions. The purity and concentration of the DNA were determined by UV spectrophotometry. The APOE genotype was determined by polymerase chain reaction (PCR) using a previously described method [19]. The main steps included: 1) Design primers, upstream primer: CAAATCGGAACTGGAGGAACAACT, downstream primer: CCCGGCCTGGTACACTGC; synthesized by Shanghai Tianhao Biotechnology Co., Ltd.; 2) Design PCR reaction system: 10 x PCR Buffer 2.0μl, genomic DNA 50 ng, Taq DNA polymerase 1U (Takara), primers 5 pmol each, dimethyl sulfoxide 2μl (DMSO, Beijing Sanbo Biological Co., Ltd.), followed by the addition of distilled water to make the total system volume 20μL; 3) Circulation procedure: pre-denaturation at 95°C for 10 min, then denaturation at 95°C for 1 min, annealing at 64°C for 1 min, extension at 72°C for 1 min, cycled 34 times, and finally extension at 72°C for 5 min; 4) The amplified product was digested by Hha I at 37°C for 4 h; 5) The digested product was analyzed by polyacrylamide gel electrophoresis; 6) The digested fragment separated by electrophoresis was stained with ethidium bromide (EB, Sigma) for imaging analysis to determine the genotype. Subjects who were homozygous or heterozygous for the APOE ɛ4 allele were considered to have APOE ɛ4 (+).

MRI image acquisition

All MRI images were collected using a Siemens 3.0T machine, and 176 layers were scanned without gaps in the T1-weighted images (MPRAGE sequence). The scanning parameters were as follows: layer thickness, 1 mm; FOV, 256×248 mm²; TR = 2300 ms; TE = 2.98 ms; flip angle, 9°; voxel resolution, 1*1*1 mm³; and matrix size, 256×248.

Data preprocessing

For T1-weighted images, the original DICOM format was first converted into the NIFTI format using the dcm2niigui software. After checking the image quality, it was found that one subject in the case group had artefacts; therefore, it was excluded.

Voxel-based morphometry (VBM) analysis

The VBM analysis was based on Statistical Parametric Mapping 8 (spm8) software on the MATLAB platform [20]. The specific processing procedures were as follows: 1) The Montreal Neurological Institute (MNI) standard template was used to successively perform linear and non-linear transformations on the collected images, so that all the subjects’ images were standardized to the same three-dimensional space for subsequent statistical analysis; 2) The tissue probability map that comes with the software was used as a reference to segment the white matter, grey matter, and cerebrospinal fluid of the structural images; and 3) To make the data conform to the normal distribution, improve the signal-to-noise ratio of the images, and compensate for the registration error, a Gaussian kernel (8*8*8 mm³) was used for image smoothening.

Surface-based morphometry (SBM) analysis

FreeSurfer software (version 6.0) was used to process and analyze the cortical surface of the T1 structure images [21]. The method comprises the following specific steps: 1) registering T1 structural images to a Talairach template, performing normalization processes on the grey values of the images, and removing non-brain tissues such as the skull; 2) segmenting the grey-white interface according to the signal intensity and carrying out topology correction; and 3) according to the Desikan– Killiany DK40 Atlas, the entire cerebral cortex was divided into 68 brain regions, 34 for each hemisphere. The data were smoothed with Gaussian kernels of 10 mm³, 15 mm³, and 20 mm³, respectively, to calculate the cortical thickness of each brain region.

Structural covariant network

Using the cortical thickness as the morphological measurement index, according to the Desikan– Killiany DK40 Atlas, the Pearson correlation coefficient was calculated to construct the network matrix, and the graph theory attributes were calculated, including the clustering coefficient (Cp), hc (Lp), normalized clustering coefficient (Gamma, γ), normalized shortest path length (Lambda, λ), small-worldness (sigma, σ), Eglobal, Elocal, Betweenness Centrality (BC), Degree Centrality (degree, DC), and efficiency [22, 23].

Statistical analysis

All subjects were grouped as follows: 1) PT versus HC; 2) APOE ɛ4 (+) PT versus APOE ɛ4 (+) HC; 3) APOE ɛ4 (–) PT versus APOE ɛ4 (–) HC; 4) APOE ɛ4 (+) PT versus APOE ɛ4 (–) PT; and 5) APOE ɛ4 (+) HC versus APOE ɛ4 (–) HC.

Statistical analysis of general data

SPSS 25 software was used to perform statistical analyses of the demographic data and clinical indicators of the subjects. Continuous variables are shown as means±standard deviations (X±S), and categorical variables are expressed as percentages. A two-sample T test was used to compare continuous variables, and the chi-square test was used to compare categorical variables. Statistical significance was set at p < 0.05.

Comparison of grey matter volume (GMV) and correlation analysis with MMSE

The general linear model (GLM) in SPM 8 was used to conduct a t-test on the whole brain GMV. Age, sex, and years of education were used as covariates, and the significance level was corrected using Cluster-Level Family-Wise Error. The correction was set as voxel p < 0.001 and cluster p < 0.05. To further explore the association between the difference in GMV and severity of clinical symptoms, the corrected positive results were regarded as the region of interest (ROI). Then, the GMV of the ROI among different groups was extracted, and partial correlation coefficient analysis was performed on the GMV and MMSE scores (p < 0.05).

Comparison of cortical thickness

The t-test was performed on the cortical thickness of the three groups under different smooth kernel conditions with the FreeSurfer plug-in, and the covariates of sex, age, and education year were added. A p < 0.05 indicated that the difference was considered statistically significant.

Structural covariant network analysis

The non-parametric permutation test was used to compare the differences between groups of network attributes of each individual subject. Sex, age, and years of education were used as covariates.

RESULTS

Demographic characteristics

Table 1 summarizes the demographic data and clinical features of the subjects, including age, sex, years of education, MMSE score, HAMD score, HAMA score, and distribution of hypertension, hyperlipidemia, and diabetes. Compared with the HCs, PTs had significantly fewer years of education and lower MMSE scores (p < 0.05). The proportion of patients with AD carrying the APOE ɛ4 (+) gene was also significantly higher than that in the control population (p < 0.05). There was no statistical difference in the other indices between the PT and HC groups.

Table 1

Demographic characteristics of all subjects

Demographics	PT (n = 15)	HC (n = 26)	p
Age, y	72.80±7.06	72.77±5.81	0.988
Education, y	6.87±5.00	11.38±3.38	0.001*
Male, n (%)	7 (46.7)	15 (57.7)	0.495
MMSE	13.73±5.65	27.15±2.40	<0.001*
APOEɛ4 (+), n (%)	10 (66.7)	9 (34.6)	0.047*
Hypertension, n (%)	5 (33.3)	8 (30.8)	0.865
Hyperlipidemia, n (%)	2 (13.3)	3 (11.5)	0.866
Diabetes, n (%)	7 (46.7)	7 (26.9)	0.199
HAMD	3.47±2.80	3.65±2.53	0.827
HAMA	3.67±2.38	4.04±2.27	0.623

Data are the mean±standard deviation or number (%). *statistically significant. APOE, apolipoprotein E; HC, healthy controls; HAMD, Hamilton Depression Scale; HAMA, Hamilton Anxiety Scale; MMSE, Mini-Mental State Examination; PT, patients with AD.

GMV and correlation analysis with MMSE

We calculated the GMV of 90 cerebral regions of all subjects based on the AAL atlas, and then compared the difference between the PT and HC groups. The results showed that compared with the HC group, the GMV of the bilateral inferior temporal gyrus, right middle temporal gyrus, and right inferior parietal lobule in the PT group was significantly reduced (cluster p < 0.05), and no area with increased volume was found. The results are presented in Supplementary Table 1 and Fig. 1A. After further grouping according to the APOE gene, it was found that there was a reduction in the GMV in the right limbic lobe, right frontal lobe, left anterior cingulate gyrus, and bilateral olfactory cortex in the APOE ɛ4 (–) PT group compared to the APOE ɛ4 (–) HC group (cluster p < 0.05). The results are presented in Supplementary Table 2 and Fig. 1B. No significant GMV changes were noted in APOE ɛ4 (+) PT versus APOE ɛ4 (+) HC, APOE ɛ4 (+) PT versus APOE ɛ4 (–) PT, and APOE ɛ4 (+) HC versus APOE ɛ4 (–) HC.

Fig. 1

A) Brain regions with different gray matter volumes in the PT group and the HC group. The red area was the brain regions in the HC group whose volume was larger than that in the PT group; B) Brain regions with different gray matter volumes in the APOE ɛ4 (–) PT group and the APOE ɛ4 (–) HC group. The red area was the brain regions in the APOE ɛ4 (–) HC group whose volume was larger than that in the APOE ɛ 4 (–) PT group.

The brain regions with GMV differences between the two groups were extracted as ROI and analyzed for correlations with the MMSE score. It was found that in the PT and HC groups, GMV was moderately correlated with MMSE scores to a medium degree (r1 = 0.598, r2 = 0.694, r3 = 0.693, p < 0.001), whereas in the APOE ɛ4 (–) PT and APOE ɛ 4 (–) HC groups, GMV and MMSE scores were moderately correlated (r = 0.530, p < 0.05). Supplementary Table 3 presents the results.

Comparison of cortical thickness

For the PT and HC groups, we performed a two-sample t-test and adopted the Monte Carlo simulation correction based on displacement; the corrected p value was 0.05. The results showed that when the smooth kernels were 10 mm³ or 15 mm, the cortex of the left lateral occipital lobe and the inferior parietal gyrus in the HC group were thicker than those in the PT group (p < .05). No significant differences were observed in the right brain. The results are presented in Supplementary Table 4 and Fig. 2.

Fig. 2

Brain regions with different cortical thickness in the PT group and the HC group. The red area was the brain regions where the cortical thickness in the HC group was greater than that in the PT group. A) Gaussian kernel 10 mm³; B) Gaussian kernel 15 mm³.

For each gene subgroup, since the number of enrolled subjects was small and the corrected p value was large, the p value for the subsequent inter-group comparison was 0.001 without correction.

The thickness of the cerebral cortex in the left transverse temporal gyrus and inferior temporal gyrus was reduced (smooth kernels of 10 mm³ and 15 mm³) in the APOE ɛ4 (+) HC group compared to the APOE ɛ4 (–) HC group. The results are presented in Supplementary Table 5 and Fig. 3, respectively.

Fig. 3

Brain regions with different cortical thickness in the APOE ɛ4 (+) HC group and the APOE ɛ4 (–) HC group. The blue area was the brain regions where the cortical thickness in the APOE ɛ4 (+) HC group was less than that in the APOE ɛ4 (–) HC group. A) Gaussian kernel 10 mm³; B) Gaussian kernel 15 mm³.

When the Gaussian kernel was 10 mm³, for the HC and PT groups carrying the APOE ɛ4 (+) gene, the cortical thickness of the former was larger in the left medial orbital frontal region than in the latter, while it was smaller in the right superior parietal lobe. The results are presented in Supplementary Table 6 and Fig. 4, respectively.

Fig. 4

Brain regions with different cortical thickness in the APOE ɛ4 (+) HC group and the APOE ɛ4 (+) PT group. A) The yellow area was the brain regions where the cortical thickness in the APOE ɛ4 (+) HC group was greater than that in the APOE ɛ4 (+) PT group. B) The blue area was the brain regions where the cortical thickness in the APOE ɛ4 (+) HC group was less than that in the APOE ɛ4 (+) PT group.

Compared with the APOE ɛ4 (–) HC group, in the case of three different smooth kernels, the APOE ɛ4 (–) PT group showed extensive cortical thickness reduction, mainly including the bilateral fusiform gyrus, superior frontal gyrus, left lateral orbital frontal, precuneus, superior parietal gyrus, right precentral gyrus, middle temporal gyrus, parsopercularis gyrus, insular gyrus, and superior marginal gyrus, among others. The results are presented in Supplementary Table 7 and Fig. 5, respectively.

Fig. 5

Brain regions with different cortical thickness in the APOE ɛ4 (–) HC group and the APOE ɛ4 (–) PT group. A–C) The red area was the brain regions where the cortical thickness in the APOE ɛ4 (–) HC group was greater than that in the APOE ɛ4 (–) PT group in the left brain. Gaussian kernels are 10, 15, and 20 mm³ respectively. D–F) The red area was the brain regions where the cortical thickness in the APOE ɛ4 (–) HC group was greater than that in the APOE ɛ4 (–) PT group in the right brain. Gaussian kernels are 10, 15, and 20 mm, respectively.

In the PT group, the cortical thickness in the bilateral rostral middle frontal gyrus of APOE ɛ4 (–) patients was lower than that of APOE ɛ4 (+) patients when the smooth kernel was 10 mm³. The results are presented in Supplementary Table 8 and Fig. 6, respectively.

Fig. 6

Brain regions with different cortical thickness in the APOE ɛ4 (–) PT group and the APOE ɛ4 (+) PT group. A) The red area was the brain region where the cortical thickness in the APOE ɛ4 (+) PT group was greater than that in the APOE ɛ4 (–) PT group in the left brain. B) The yellow area was the brain region where the cortical thickness in the APOE ɛ4 (+) PT group was greater than that in the APOE ɛ4 (–) PT group in the right brain.

Structural covariant network

It should be noted that in the analysis of the structural covariate network, owing to the small number of subjects and large number of brain nodes, the p value after false discovery rate (FDR) correction was larger; therefore, the results shown here were all without FDR correction.

In this study, to explore the abnormalities of brain structural networks in the pathological state of AD, we performed structural network analysis on the brains of the PT and HC groups from two aspects: the overall characteristics and the node characteristics of the network. We constructed a network matrix for the subjects according to the Desikan– Killiany DK40 Atlas, calculated the graph theory attributes, and then compared the groups. The results demonstrated that in all subgroups, the indicators of all global attributes, including CP, LP, Gamma, Lambda, sigma, Eglobal, and Elocal did not show differences between groups (p > 0.05). The results of the global attribute comparisons between the groups are shown in Supplementary Table 9.

We selected BC, degree, and efficiency as indicators reflecting important cerebral regions significantly related to AD. The local attributes of the network showed statistical differences among all the groups (p < 0.05), and were mainly located in the right gyrus, right temporal pole, bilateral middle temporal gyrus, right transverse temporal gyrus, and left postcentral gyrus. Figure 7 shows the nodes with significant differences between the groups, and Table 2 show the results of comparisons for specific node parameters.

Fig. 7

Brain regions with significant differences in node attributes between groups. Each figure represents BC, DC, and efficiency from top to bottom. A) HC versus PT; B) APOE ɛ4 (+) HC versus APOE ɛ4 (–) HC; C) APOE ɛ4 (+) HC versus APOE ɛ4 (+) PT; D) APOE ɛ4 (–) HC versus APOE ɛ4 (–) PT; E) APOE ɛ4 (+) PT versus APOE ɛ4 (–) PT.

Table 2

The p value results of structural covariant network comparisons between the groups

	BC	DC	efficiency
HC versus PT	L_Parahippoc (0.0337)	R_Bankssts (0.0430)	L_Middletem (0.0307)
HCpos versus HCneg	R_Precuneus (0.0448)	R_Temporalp (0.0154)	R_Temporalp (0.0144)
	R_Temporalp (0.0404)
	R_Transverse (0.0354)
HCpos versus PTpos	L_Postcentral (0.0480)	L_Isthmuscin (0.0392)	R_Bankssts (0.0206)
	R_Temporalp (0.0398)	R_Bankssts (0.0252)
		R_pericalcarin (0.0286)
		R_Transverse (0.0298)
HCneg versus PTneg	R_Parsorbitali (0.0228)	R_Temporalp (0.0358)	L_Lateralorbit (0.0254)
PTpos versus PTneg	L_Postcentral (0.0390)	–	–
	R_Middletem (0.0338)
	R_Superiorte (0.0498)

p values in parentheses. BC, Betweenness Centrality; DC, DegreeCentrality; HC, healthy controls; HCpos, APOE ɛ4 (+) healthy controls; HCneg, APOE ɛ4 (–) healthy controls; L, left; PTpos, APOE ɛ4 (+) patients; PTneg, APOE ɛ4 (–) patients; PT, patients with AD; R, right.

DISCUSSION

At present, many research groups have investigated the effect of APOE gene polymorphism on neuroimaging of AD patients, but this study is the first systematic study shows that APOE gene polymorphism alter GMV, cortical structure and structural brain network. Our results suggest that comprehensive analysis of APOE genotyping, gray matter structure, structural structure, and structural brain network in AD is likely to be a new direction for future research and even diagnosis of AD.

Subsequently, we further verified the effect of APOE gene polymorphism on the brain structure of AD in other public databases. The team of Zhang found APOE ɛ4 (+) prodromal AD participants had a thinner cortical thickness of bilateral entorhinal, smaller subcortical volume of the left amygdala, bilateral hippocampus, and left ventral DC relative to APOE ɛ4 (–) prodromal AD in ADNI database [24]. Our study found the cortical thickness in the bilateral rostral middle frontal gyrus of APOE ɛ4 (–) patients was lower than that of APOE ɛ4 (+) patients. In addition, another group of Gemma Salvadó found carrying at least one ɛ2 allele was associated with larger GM volumes in brain areas typically affected by AD and also in areas associated with cognitive resilience in ADNI database [25]; however, our study showed no significant GMV changes in APOE ɛ4 (+) PT versus APOE ɛ4 (+) HC, APOE ɛ4 (+) PT versus APOE ɛ4 (–) PT, and APOE ɛ4 (+) HC versus APOE ɛ4 (–) HC. The team of Raffaele Cacciaglia discovered structural vulnerability in neuronal networks associated with APOE ɛ4 may be an early event in AD pathogenesis, possibly upstream of amyloid deposition in ALFA (ALzheimer and FAmilies) study [26]. These results all found that APOE gene polymorphism affected the imaging structure of AD, but the specific effects were not completely consistent with our results. The reasons can be attributed to the following reasons: the current studies in the database are all European and American populations, which may have an impact on the results due to genetic influences.

Morphological analysis

AD is the most common neurodegenerative disease, with one of its notable features being the rapid progression of brain atrophy. The medial temporal lobe, including the entorhinal cortex and hippocampus, is an early region of the brain that is closely associated with AD degeneration. A meta-analysis performed by Barnes et al. showed that the average annual atrophy rate of the hippocampus in a normally aging brain is 1.4%, whereas that in the population with AD is 4.7% [27]. With the development of neuroimaging in recent years, many researchers have used sMRI to explore the structural changes in the AD brain. The orbitofrontal cortex, posterior cingulate gyrus, inferior parietal lobule, middle frontal gyrus, fusiform gyrus, temporal pole, and other brain regions have also shown a significant correlation with AD cognitive decline [11 , 28–30].

Our purpose was to verify changes in grey matter structure in the AD brain in the local Han population, and to further explore the effects of APOE ɛ4 alleles on brain atrophy and cognitive function in AD. At present, it is generally believed that there is a link between the hippocampal and entorhinal volume and memory tasks, between the left temporal cortex and semantic memory, between the frontal lobe and executive function, and between the parietal and occipital lobes, as well as cognitive speed and visuoconstruction [31, 32]. Analyzing the whole GMV of patients with AD and the normal older adults, we found that the GMV of some brain areas in the AD group had a significant downward trend, and was significantly correlated with the MMSE score. These brain regions mainly include the bilateral inferior temporal gyrus, right middle temporal gyrus, right inferior parietal lobule, right marginal lobe, right frontal lobe, left anterior cingulate gyrus, and the bilateral olfactory cortex. This result is consistent with published literature and brings attention to the fact that brain atrophy is associated with cognitive impairment. The use of sMRI-based whole-brain volume measurement to evaluate the cognitive status or diagnosis of patients with dementia is feasible [33 –35]. In addition to comparing the volume of grey matter in different brain regions, we also performed a computational analysis of cortical thickness, and similar results were obtained. In general, the cortical thicknesses of the left lateral occipital lobe, inferior parietal lobe, orbitofrontal region, precuneus, superior parietal gyrus, right precentral gyrus, middle temporal gyrus, pars opercularis gyrus, insular gyrus, superior marginal gyrus, bilateral fusiform gyrus, and superior frontal gyrus in patients with AD were significantly lower than those in the normal control group.

The APOE ɛ4 allele is a strong risk factor for late-onset AD [36]. To clarify the importance of APOE ɛ4 in the clinical and biological heterogeneity of AD, Zhang et al. conducted a study on patients with prodromal AD [24]. The results of this study confirmed that, compared with those with APOE ɛ4 (–) prodromal AD, APOE ɛ4 (+) participants had thinner entorhinal cortices in both hemispheres, smaller subcortical volume in the bilateral hippocampus, left amygdala, and left ventral diencephalon, and presented with faster decline of memory and overall cognitive clinical manifestations. In our comparison of gene subgroups, it was discovered that in the normal older population, carriers of APOE ɛ4 (+) also had more significant temporal lobe atrophy than non-carriers, which further illustrates the important influence of APOE ɛ4 status on brain atrophy and cognitive function.

Structural covariant network analysis

A growing body of evidence suggests that changes that lead to cognitive impairment are not limited to specific brain regions, such as the medial temporal lobe, but manifest as broad changes in network topology properties [37]. A large multicenter cohort study by Pereira et al. showed that changes in grey matter network parameters are associated with gradual progression of dementia in patients with mild cognitive impairment [38]. In this study, we used graph analysis to construct structural grey matter networks in patients with AD and healthy controls. The basic elements of the network are the nodes and edges. Brain network analysis based on graph theory mainly analyses the different attributes of nodes and edges, including global and local attributes. The global attributes include CP, LP, gamma, lambda, sigma, Eglobal, and Elocal. CP represents the possibility of interconnection between nodes adjacent to a node, while LP represents the mean of the shortest paths of any two nodes in the network. These two indices are usually combined to define a small-world property; when a network has a large normalized clustering coefficient and an approximate normalized path length relative to a random network, that is, γ>1 and λ ≈ 1 or σ = γ/λ > 1, the network is considered to have a small-world property [22]. Eglobal and Elocal reflect the ability to transmit information [39]. There were no significant differences in global attributes between the AD and HC groups in our comparison, which may be attributed to several reasons. First, there were relatively few enrolled cases and controls, and there were differences in brain structure among individuals. Second, age-related structural network studies have shown that reduced structural connectivity is associated with AD in young adults, whereas cognitive impairment is relatively less affected in older adults [40]. In addition, AD is considered a disconnection syndrome in which connections between brain regions are lost, but abnormalities in this structural network have been proven to have a compensatory mechanism, thus resulting in improved network integrity. This compensation also starts in the medial temporal lobe, and gradually extends to the wider cortex as the disease progresses [41, 42]. However, we still observe a trend that patients with AD have lower small-world properties and lower network efficiency than normal controls. Information interaction forms the basis of cognitive processes. A smaller CP and a larger LP in the AD group indicate that the connection between adjacent brain regions is reduced, and that the efficiency of information transmission and integration is decreased. Moreover, similar to the results of the GMV analysis in patients with AD, the APOE ɛ4 (+) group exhibited a more random topology than the APOE ɛ4 (–) group, thus re-emphasizing the pathogenic risk of APOE ɛ4.

We used BC, DC, and efficiency to describe node attributes. The BC measures the connection ability between different nodes connected to a certain node. Node degree is defined as the number of nodes connected to a node. The larger the node degree, the higher the DC of the node, and the more important the node. Efficiency refers to the information transfer capability of a node in a network [39]. We found that the node attributes of the grey matter network were significantly different between the case and control groups, and that these differences were mainly located in the right bank, right temporal pole, bilateral middle temporal gyrus, right transverse temporal gyrus, left postcentral gyrus, and left parahippocampal gyrus. Regions such as the hippocampus, parahippocampal gyrus, and entorhinal cortex form a feedback loop by interacting with neocortical regions, which become the basis for memory processing [43, 44]. Patients with AD had lower BC in the left parahippocampal gyrus, left postcentral gyrus, and right temporal pole; lower DC in the left isthmus cingulate, right bankssts, and right transverse temporal gyrus; and lower efficiencies in the left middle temporal gyrus and the lateral orbitofrontal gyrus than controls. Decreased node attributes imply fewer nerve fibers connecting to these brain regions, disrupting local functional and structural connections and potentially affecting patients’ memory and cognitive abilities. The AD group exhibited higher node attributes in some brain regions. BC was higher than that of the controls in the right pars orbitalis and left postcentral gyrus; DC was higher than that of the control in the right bank, right pericalcarine, and right temporal pole; and efficiency was higher in the right bankssts. This may be related to the compensation mechanism mentioned above, with similar studies previously reporting this[39, 45].

As in the previous analysis, APOE ɛ4 gene subgroup comparisons were performed. In the AD group, there were significant differences in BC values in the left postcentral gyrus, right middle temporal gyrus, and right superior temporal gyrus between APOE ɛ4 (+) carriers and non-carriers. This difference was more pronounced in the control group. All APOE ɛ4 (+) subjects had lower node attributes than APOE ɛ4 (–) subjects, and the main brain regions involved included the right precuneus, temporal pole, and transverse temporal gyrus. APOE ɛ4 not only increases the risk of AD but has also been proposed to be associated with memory impairment and hippocampal atrophy by many studies. The existence of the APOE ɛ4 allele accelerates cognitive decline in AD, mild cognitive impairment, and healthy subjects [46 –48]. Therefore, among the subjects we included, it is reasonable to speculate that healthy people who carry the APOE ɛ4 allele will experience a greater rate of cognitive decline in the future than non-carriers will. However, our study included a relatively small number of samples, which is the shortcoming of this paper. Hence, more studies are needed to perform to verify this relation.

Application prospect

The underlying mechanism of APOE ɛ4 affecting AD progress has not been fully understood. To date, many studies proved that APOE genotyping influences on MRI results in AD patients. However, our study first proved that APOE genotyping influences altered the GMV in the right limbic lobe, right frontal lobe, left anterior cingulate gyrus, and bilateral olfactory cortex, the thickness of the cerebral cortex in the left transverse temporal gyrus and inferior temporal gyrus. Moreover, APOE genotyping also changed the structural covariate network. We hypothesized that APOE genotyping in AD population combined with GMV, cortical thickness, and structural covariant network would be one of the diagnostic methods for AD. As we all known, the use of functional networks in AD patients is of great significance, and relevant teams have conducted relevant studies [49, 50]. And the brain’s structural and functional network are closely shared [51]. We hope to verify its importance in the diagnosis of AD in our population and to explore the characterization of functional networks in AD and their relationship with structural networks in the future studies.

Conclusion

In conclusion, our study revealed the differences in grey matter structure and network between patients with AD and normal controls, and further explored the effect of the APOE ɛ4 allele on clinical manifestations and neuroimaging regions. Our findings provide new evidence of structural changes in grey matter in patients and provide a reference for understanding the AD brain and applying neuroimaging to diagnose and monitor disease progression.

Footnotes

ACKNOWLEDGMENTS

I would like to all authors for their guidance through each stage of the process.

This work was supported by grants from the National Natural Science Foundation of China (No. 81801054), the National Natural Science Foundation of Jiangsu Province (No. BK20221201), the Wuxi Municipal Health and Family Planning Commission Fund (No. M202137), Wuxi Top Talent Support Program for Young and Middle-aged People of Wuxi Health Committee (No. HB2020023) and China Postdoctoral Funding.

The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Authors’ disclosures available online ().

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

References

Sheppard

, Coleman

(2020) Alzheimer’s disease: Etiology, neuropathology and pathogenesis. In Alzheimer’s Disease: Drug Discovery, Huang

, ed., Brisbane (AU).

Scheltens

, De Strooper

, Kivipelto

, Holstege

, Chetelat

, Teunissen

, Cummings

, van der Flier

(2021) Alzheimer’s disease. Lancet 397, 1577–1590.

Kumar

, Sidhu

, Goyal

, Tsao

(2021) Alzheimer disease. In StatPearls, Treasure Island (FL).

Villeneuve

, Rabinovici

, Cohn-Sheehy

, Madison

, Ayakta

, Ghosh

, La Joie

, Arthur-Bentil

, Vogel

, Marks

, Lehmann

, Rosen

, Reed

, Olichney

, Boxer

, Miller

, Borys

, Jin

, Huang

, Grinberg

, DeCarli

, Seeley

, Jagust

(2015) Existing Pittsburgh Compound-B positron emission tomography thresholds are too high: Statistical and pathological evaluation. Brain 138, 2020–2033.

Rathore

, Habes

, Iftikhar

, Shacklett

, Davatzikos

(2017) A review on neuroimaging-based classification studies and associated feature extraction methods for Alzheimer’s disease and its prodromal stages. Neuroimage 155, 530–548.

Frisoni

, Fox

, Jack

Jr. , Scheltens

, Thompson

(2010) The clinical use of structural MRI in Alzheimer disease. Nat Rev Neurol 6, 67–77.

Gonuguntla

, Yang

, Guan

, Koo

, Kim

(2022) Brain signatures based on structural MRI: Classification for MCI, PMCI, and AD. Hum Brain Mapp 43, 2845–2860.

Shaikh

, Ali

(2019) Automated atrophy assessment for Alzheimer’s disease diagnosis from brain MRI images. Magn Reson Imaging 62, 167–173.

Weintraub

, Dietz

, Duda

, Wolk

, Doshi

, Xie

, Davatzikos

, Clark

, Siderowf

(2012) Alzheimer’s disease pattern of brain atrophy predicts cognitive decline in Parkinson’s disease. Brain 135, 170–180.

10.

Bakkour

, Morris

, Wolk

, Dickerson

(2013) The effects of aging and Alzheimer’s disease on cerebral cortical anatomy: Specificity and differential relationships with cognition. Neuroimage 76, 332–344.

11.

Tondelli

, Wilcock

, Nichelli

, De Jager

, Jenkinson

, Zamboni

(2012) Structural MRI changes detectable up to ten years before clinical Alzheimer’s disease, Neurobiol Aging 33, 825e825–836.

12.

Mattsson

, Eriksson

, Lindberg

, Scholl

, Lampinen

, Nilsson

, Insel

, Lautner

, Strandberg

, van Westen

, Zetterberg

, Blennow

, Palmqvist

, Stomrud

, Hansson

(2018) Effects of APOE epsilon4 on neuroimaging, cerebrospinal fluid biomarkers, and cognition in prodromal Alzheimer’s disease. Neurobiol Aging 71, 81–90.

13.

Mattsson

, Ossenkoppele

, Smith

, Strandberg

, Ohlsson

, Jogi

, Palmqvist

, Stomrud

, Hansson

(2018) Greater tau load and reduced cortical thickness in APOE epsilon4-negative Alzheimer’s disease: A cohort study. Alzheimers Res Ther 10, 77.

14.

Lee

, Ha

, Park

, Lee

, Moon

, Chung

, Kim

, Mun

, Kim

(2016) Impact of Apolipoprotein E4 polymorphism on the gray matter volume and the white matter integrity in subjective memory impairment without white matter hyperintensities: Voxel-based morphometry and tract-based spatial statistics study under 3-Tesla MRI. J Neuroimaging 26, 144–149.

15.

Susanto

, Pua

, Zhou

, Alzheimer’s Disease Neuroimaging Initiative (2015) Cognition, brain atrophy, and cerebrospinal fluid biomarkers changes from preclinical to dementia stage of Alzheimer’s disease and the influence of apolipoprotein E. J Alzheimers Dis 45, 253–268.

16.

Moller

, Vrenken

, Jiskoot

, Versteeg

, Barkhof

, Scheltens

, van der Flier

(2013) Different patterns of gray matter atrophy in early- and late-onset Alzheimer’s disease. Neurobiol Aging 34, 2014–2022.

17.

, Yang

, Lei

, Lin

, Li

, Jiang

, Wei

, Xu

(2019) The association between the ApoE polymorphisms and the MRI-defined intracranial lesions in a cohort of southern China population. J Clin Lab Anal 33, e22950.

18.

Tamaoka

(2011) [Alzheimer’s disease: Definition and National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA)]. Nihon Rinsho 69 Suppl 10 Pt 2, 240–245.

19.

Wenham

, Price

, Blandell

(1991) Apolipoprotein E genotyping by one-stage PCR. Lancet 337, 1158–1159.

20.

Ashburner

, Friston

(2000) Voxel-based morphometry–the methods. Neuroimage 11, 805–821.

21.

Fischl

, Dale

(2000) Measuring the thickness of the human cerebral cortex from magnetic resonance images. Proc Natl Acad Sci U S A 97, 11050–11055.

22.

Watts

, Strogatz

(1998) Collective dynamics of ‘small-world’ networks. Nature 393, 440–442.

23.

Bullmore

, Bassett

(2011) Brain graphs: Graphical models of the human brain connectome. Annu Rev Clin Psychol 7, 113–140.

24.

Zhang

, Kong

, Wei

, Zhang

, Ma

, Ba

, Alzheimer’s Disease Neuroimaging Initiative (2020) The effect of ApoE epsilon 4 on clinical and structural MRI markers in prodromal Alzheimer’s disease. Quant Imaging Med Surg 10, 464–474.

25.

Salvado

, Ferreira

, Operto

, Cumplido-Mayoral

, Arenaza-Urquijo

, Cacciaglia

, Falcon

, Vilor-Tejedor

, Minguillon

, Groot

, van der Flier

, Barkhof

, Scheltens

, Ossenkoppele

, Kern

, Zettergren

, Skoog

, Hort

, Stomrud

, van Westen

, Hansson

, Molinuevo

, Wahlund

, Westman

, Gispert

, ALFA study, BioFINDER, ADNI (2022) The protective gene dose effect of the APOE epsilon2 allele on gray matter volume in cognitively unimpaired individuals. Alzheimers Dement 18, 1383–1395.

26.

Cacciaglia

, Molinuevo

, Falcon

, Arenaza-Urquijo

, Sanchez-Benavides

, Brugulat-Serrat

, Blennow

, Zetterberg

, Gispert

, study

(2020) APOE-epsilon4 shapes the cerebral organization in cognitively intact individuals as reflected by structural gray matter networks. Cereb Cortex 30, 4110–4120.

27.

Barnes

, Bartlett

, van de Pol

, Loy

, Scahill

, Frost

, Thompson

, Fox

(2009) A meta-analysis of hippocampal atrophy rates in Alzheimer’s disease. Neurobiol Aging 30, 1711–1723.

28.

Dickerson

, Stoub

, Shah

, Sperling

, Killiany

, Albert

, Hyman

, Blacker

, Detoledo-Morrell

(2011) Alzheimer-signature MRI biomarker predicts AD dementia in cognitively normal adults. Neurology 76, 1395–1402.

29.

Hall

, Moore

, Lopez

, Kuller

, Becker

(2008) Basal forebrain atrophy is a presymptomatic marker for Alzheimer’s disease. Alzheimers Dement 4, 271–279.

30.

, Coyle

, Maguire

, Watson

, McGinnity

(2011) Gray matter concentration and effective connectivity changes in Alzheimer’s disease: A longitudinal structural MRI study. Neuroradiology 53, 733–748.

31.

Arlt

, Buchert

, Spies

, Eichenlaub

, Lehmbeck

, Jahn

(2013) Association between fully automated MRI-based volumetry of different brain regions and neuropsychological test performance in patients with amnestic mild cognitive impairment and Alzheimer’s disease. Eur Arch Psychiatry Clin Neurosci 263, 335–344.

32.

Cardenas

, Chao

, Studholme

, Yaffe

, Miller

, Madison

, Buckley

, Mungas

, Schuff

, Weiner

(2011) Brain atrophy associated with baseline and longitudinal measures of cognition. Neurobiol Aging 32, 572–580.

33.

Zhang

, Liao

, Zhang

, Yan

, Ngetich

, Zhang

, Jin

, Li

(2021) Predicting MCI to AD conversation using integrated sMRI and rs-fMRI: Machine learning and graph theory approach. Front Aging Neurosci 13, 688926.

34.

Zhang

, Teng

, Liu

, He

(2022) Diagnosis of Alzheimer’s disease based on regional attention with sMRI gray matter slices. J Neurosci Methods 365, 109376.

35.

Feng

, Zhang

, Chen

, Zuo

, Alzheimer’s Disease Neuroimaging Initiative (2022) Detection of Alzheimer’s disease using features of brain region-of-interest-based individual network constructed with the sMRI image. Comput Med Imaging Graph 98, 102057.

36.

Bertram

, Lill

, Tanzi

(2010) The genetics of Alzheimer disease: Back to the future. Neuron 68, 270–281.

37.

Verfaillie

SCJ

, Slot

RER

, Dicks

, Prins

, Overbeek

, Teunissen

, Scheltens

, Barkhof

, van der Flier

, Tijms

(2018) A more randomly organized grey matter network is associated with deteriorating language and global cognition in individuals with subjective cognitive decline. Hum Brain Mapp 39, 3143–3151.

38.

Pereira

, Mijalkov

, Kakaei

, Mecocci

, Vellas

, Tsolaki

, Kloszewska

, Soininen

, Spenger

, Lovestone

, Simmons

, Wahlund

, Volpe

, Westman

(2016) Disrupted network topology in patients with stable and progressive mild cognitive impairment and Alzheimer’s disease. Cereb Cortex 26, 3476–3493.

39.

Sheng

, Chen

, Shao

, Qin

, Zhao

, Xu

, Bai

(2021) Brain structural network compensation is associated with cognitive impairment and Alzheimer’s disease pathology. Front Neurosci 15, 630278.

40.

, Gao

, Potter

, Benoit

, Shen

, Schulz

, Zhang

(2021) Interactions between aging and Alzheimer’s disease on structural brain networks. Front Aging Neurosci 13, 639795.

41.

Delbeuck

, Van der Linden

, Collette

(2003) Alzheimer’s disease as a disconnection syndrome? Neuropsychol Rev 13, 79–92.

42.

Liu

, Hu

, Wang

, Han

, Li

, Qi

, Wang

, Zhang

, He

, Zhao

(2020) A brain network constructed on an L1-norm regression model is more sensitive in detecting small world network changes in early AD. Neural Plast 2020, 9436406.

43.

Pasquini

, Benson

, Grothe

, Utz

, Myers

, Yakushev

, Grimmer

, Scherr

, Sorg

(2017) Individual correspondence of amyloid-beta and intrinsic connectivity in the posterior default mode network across stages of Alzheimer’s disease. J Alzheimers Dis 58, 763–773.

44.

Pasquini

, Rahmani

, Maleki-Balajoo

, La Joie

, Zarei

, Sorg

, Drzezga

, Tahmasian

(2019) Medial temporal lobe disconnection and hyperexcitability across Alzheimer’s disease stages. J Alzheimers Dis Rep 3, 103–112.

45.

Chen

, Sheng

, Luo

, Qin

, Ye

, Zhao

, Xu

, Bai

(2020) The compensatory phenomenon of the functional connectome related to pathological biomarkers in individuals with subjective cognitive decline. Transl Neurodegener 9, 21.

46.

Farlow

, He

, Tekin

, Xu

, Lane

, Charles

(2004) Impact of APOE in mild cognitive impairment. Neurology 63, 1898–1901.

47.

O’Donoghue

, Murphy

, Zamboni

, Nobre

, Mackay

(2018) APOE genotype and cognition in healthy individuals at risk of Alzheimer’s disease: A review. Cortex 104, 103–123.

48.

El Haj

, Antoine

, Amouyel

, Lambert

, Pasquier

, Kapogiannis

(2016) Apolipoprotein E (APOE) epsilon4 and episodic memory decline in Alzheimer’s disease: A review. Ageing Res Rev 27, 15–22.

49.

Zhang

, Suo

, Yang

, Lai

, Pan

, He

, Li

, Kuang

, Wang

, Gong

(2022) Structural and functional deficits and couplings in the cortico-striato-thalamo-cerebellar circuitry in social anxiety disorder. Transl Psychiatry 12, 26.

50.

Suo

, Zuo

, Lan

, Pan

, Zhang

, Kemp

, Wang

, Gong

(2022) COVID-19 vicarious traumatization links functional connectome to general distress. Neuroimage 255, 119185.

51.

Bullmore

, Sporns

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