Alzheimer's disease pathogenesis and the role of APOE ε4 in remodeling functional gradients of the default mode network

Abstract

Background

Alzheimer's disease (AD) involves Aβ and tau pathology, initiating years before symptoms. the apolipoprotein E ε4 allele (APOE ε4) is the major genetic risk factor, influencing neurodegeneration and functional network disruption. This study investigates how APOE ε4 modulates the default mode network (DMN)'s hierarchical organization to accelerate cognitive decline.

Objective

This study aimed to elucidate how APOE ε4 accelerates AD pathological progression by altering the functional gradient hierarchy of DMN, and to evaluate the potential of DMN gradient abnormalities as an early diagnostic biomarker for AD.

Methods

We enrolled 271 participants, categorized by diagnosis and APOE ε4 status. All underwent neuropsychological assessment, plasma Aβ_42/40 measurement, and resting-state functional MRI. DMN functional gradients were quantified using the BrainSpace toolbox. Statistical analyses included a 2 × 2 factorial design, mediation analysis, and correlation testing with cognitive scores.

Results

A significant Group × Genotype interaction was identified in the right inferior temporal gyrus (RITG). AD patients with APOE ε4 showed the most severe gradient attenuation, while non-carrier patients exhibited compensatory elevation. The RITG gradient correlated with multiple cognitive domains exclusively in APOE ε4 carriers. Altered connectivity between the left superior frontal gyrus (LSFG) and RITG mediated the effect of Aβ burden on gradient disruption.

Conclusions

APOE ε4 accelerates cognitive decline in AD by specifically disrupting the DMN functional gradient hierarchy, particularly within the right ITG. Amyloid-β deposition contributes to macro-scale DMN topological disorganization by impairing LSFG-RITG functional connectivity. DMN functional gradient mapping provides a sensitive biomarker for early AD diagnosis and a novel target for APOE ε4-targeted interventions.

Keywords

Alzheimer's disease default mode network gradient

Introduction

Alzheimer's disease (AD), the most prevalent form of dementia in the elderly, is characterized by two core molecular pathologies: the accumulation of neurotoxic amyloid-β (Aβ) oligomers and the formation of neurofibrillary tangles composed of hyperphosphorylated tau protein.^1–3 These pathological hallmark are thought to contribute synergistically to synaptic dysfunction and progressive neuronal degeneration.^4,5 However, it is not known how these pathological hallmarks are mechanistically related to synaptic dysfunction and progressive neuronal degeneration. Synaptic dysfunction may be driven by APOE mechanisms independently of Aβ and tau, with these depositions just being scars of the underlying critical process. Further, the tau changes are associated with the loss of synapses and cognitive dysfunction, while Aβ is not, an important issue that addressed by the findings of this paper.

Of particular clinical significance, the Aβ depositions typically occur 15–20 year prior to clinical symptom onset,⁶ and these depositions appear to occur about 9 years of age earlier according to APOE genotype.⁷ Elucidating the spatiotemporal trajectory of AD pathogenesis therefore represents a critical prerequisite of understanding the early stage changes of AD.

Over 95% of AD cases are categorized as sporadic late-onset AD (LOAD), with the APOE ε4 allele representing the most significant genetic risk factor.^8–10 Accumulating evidence demonstrates that APOE ε4 exerts pleiotropic effects on AD progression through multiple pathophysiological mechanisms: including facilitating Aβ plaque nucleation and fibrillization,^11–13 potentiating tau hyperphosphorylation and subsequent neurofibrillary tangle formation,^14–17 amplifying neuroinflammatory responses via microglia and astrocytes activation,¹⁸ and compromising cerebrovascular integrity through blood-brain barrier dysfunction.^19,20 Neuroimaging studies also reveal that APOE ε4 carriers exhibit accelerated gray matter atrophy and disrupted functional network architecture, which correlate significantly with steeper cognitive decline trajectories.^21–24

The default mode network (DMN), a key neural circuit affected in preclinical AD stages, demonstrates characteristic functional connectivity (FC) abnormalities that may serve as alterations that may serve as prodromal biomarkers. Notably, the topographic distribution of early Aβ deposition foci, particularly within the precuneus and posterior cingulate cortex, exhibits striking spatial convergence with major DMN hubs.^25,26 This neuroanatomical colocalization suggests potential mechanistic links between DMN dysregulation and early AD pathobiology. While recent studies propose APOE ε4-mediated modulation of DMN connectivity patterns,^24,27 conventional region-of-interest analyses lack sensitivity to detect nuanced alterations in the DMN's hierarchical organization.

Functional connectivity gradient mapping, an innovative framework for quantifying cortical hierarchical organization, enables precise characterization of connectivity pattern variations along a sensorimotor-to-transmodal gradient.²⁸ This methodology has identified distinctive neurosignatures in major depressive disorder²⁹ and autism spectrum conditions,³⁰ yet its application in neurodegenerative contexts remains nascent. Critical knowledge gaps persist regarding whether APOE ε4 modulates DMN gradient topography to accelerate AD progression, and whether such connectomic signatures possess predictive validity for clinical outcomes.

To address these unresolved questions, we implemented resting-state functional MRI (rs-fMRI) to construct the first comprehensive DMN gradient atlas in AD patients, systematically mapping deviations in hierarchical network architecture. Through multimodal integration of APOE genotyping and neuropsychological assessments, we further investigate ε4 allele dose-dependent effects on DMN gradient configurations and their clinical correlates. Our study provides empirical validation of the hypothesis that APOE ε4 accelerates AD-related cognitive deterioration through functional reorganization of DMN gradient hierarchies.

Methods

Participants

Study population

A cohort of 271 right-handed participants was consecutively recruited from the neurology outpatient and inpatient units at Nanjing University Medical School Affiliated Drum Tower Hospital during September 2022 to January 2025. As illustrated in Figure 1A, the study population comprised four subgroups: 108 APOE ε4 non-carrier healthy controls (HC-ε4-), 25 APOE ε4 carrier healthy controls (HC-ε4+), 81 APOE ε4 non-carrier AD patients (AD-ε4-), and 57 APOE ε4 carrier AD patients (AD-ε4+). All enrolled AD patients were newly diagnosed through standardized clinical assessments. Within a unified one-month protocol, all participants completed tripartite evaluations: hematological analyses, comprehensive neuropsychological testing, and multimodal MRI acquisitions. This study received Institutional Review Board approval from Nanjing University Medical School Gulou Hospital and strictly adhered to Helsinki Declaration guidelines. Written informed consent was obtained from all participants or legally authorized representatives prior to baseline assessments.

Figure 1.

Flowchart of the proposed method in this study A) Selection of subjects and grouping; B) Flowchart for constructing the functional gradient of DMN; C) Flowchart for constructing the FC.

Inclusion criteria

Healthy Control Group: (i) Absence of subjective cognitive complaints; (ii) Age- and education-adjusted scores on the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) within normative ranges; (iii) Clinical Dementia Rating (CDR) global score of 0; (iv) Neurologist-certified cognitive normality through standardized neuropsychological battery assessments, excluding mild cognitive impairment (MCI) or AD diagnoses. AD Patient Group: (i) Documented progressive cognitive decline ≥12 months duration reported through caregiver-assisted clinical history; (ii) Neuropsychological confirmation of dementia syndrome meeting NIA-AA 2023 core clinical criteria; (iii) CSF/blood biomarker profile consistent with AD pathophysiology (Aβ_42/40 ratio <0.08 and p-tau181 > 24.5 pg/mL); (iv) Exclusion of alternative dementia etiologies including frontotemporal lobar degeneration, Lewy body dementia, vascular dementia, and Parkinson's disease dementia through multidisciplinary diagnostic consensus.

Exclusion criteria

(i) Early-onset cognitive impairment (age <50 years) or rapid progression (symptom duration ≤12 months); (ii) Acute cerebrovascular events detected on structural MRI (Fazekas scale ≥2); (iii) Systemic/metabolic contributors to cognitive dysfunction including: Major depressive disorder (HAM-D > 18), Neuropsychiatric comorbidities (DSM-V schizophrenia spectrum disorders), Endocrine disorders (TSH beyond 0.4–4.0 mIU/L; B12 < 200 pg/mL), Active infections (HIV/syphilis seropositivity), Chronic alcohol misuse (AUDIT score ≥8); (iv) MRI contraindications or inability to complete neuropsychological evaluations.

Baseline characteristics and neuropsychological assessment

Demographic and clinical baseline characteristics were systematically recorded, including age, sex, educational attainment (years), comorbidities, and smoking status. All participants completed a standardized neuropsychological battery administered by trained personnel. The assessment protocol encompassed six cognitive domains³¹:

(i) Global Cognition: Evaluated using the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA); (ii) Episodic Memory: Assessed through the Auditory Verbal Learning Test long-delayed recall (AVLT-LR); (iii) Visuospatial Processing (VPF): Derived from the Clock Drawing Test and Visual Reproduction-copy Test; (iv) Memory Span: Quantified using digit span performance (forward and backward trials) from the Digit Span Test (DST); (v) Executive Function (EF): Measured via Trail Making Test Part B (TMT-B) and the interference score of the Stroop Color-Word Test (SCWT-C); (vi) Information Processing Speed (IPS): Assessed using Trail Making Test Part A (TMT-A) and the word/color naming subtests of the Stroop Test (SCWT-A, SCWT-B).

For TMT and SCWT subtests (TMT-A, TMT-B, SCWT-A, SCWT-B), prolonged completion times inversely correlate with performance. To ensure uniform interpretation, raw scores were transformed by calculating the reciprocal of completion time (1/time) multiplied by 100. All neuropsychological measures were standardized into z-scores using cohort-specific means and standard deviations. Composite domain scores were generated by averaging z-scores within each cognitive domain.

Quantification of blood-based AD biomarkers

Peripheral blood samples were collected via venipuncture into EDTA-coated tubes by blinded technicians. Following a 20-min clotting period at room temperature, samples were centrifuged at 2000 g for 10 min at 4°C. Plasma aliquots were stored at −80°C until batch analysis.

Amyloid-β isoforms (Aβ₄₀, Aβ₄₂), Aβ_42/40 ratio, and phosphorylated tau-181 (p-tau181) were quantified using the Single Molecule Array (Simoa^®) platform (Neurology 3-Plex A Advantage Kit, Quanterix Corporation, Lexington, MA). To ensure analytical precision, triplicate measurements were performed for each sample, with intra-assay coefficients of variation (CV) exceeding 20% triggering repeat analyses. Calibration curves were generated using manufacturer-provided standards, and sample values were interpolated against these curves.

Neuroimaging acquisition and preprocessing

Structural and functional magnetic resonance imaging (MRI) was performed using a 3.0 Tesla Philips Achieva scanner (Philips Healthcare, Best, The Netherlands) with an 8-channel phased-array head coil. To minimize motion artifacts, foam padding was applied around participants’ heads, and noise-reducing headphones were utilized. Participants were instructed to maintain supine positioning with eyes closed while remaining awake throughout the scanning session.

The multimodal imaging protocol comprised: (i) Anatomical Imaging: High-resolution 3D T1-weighted magnetization-prepared rapid gradient-echo (MPRAGE) sequence (TR/TE = 9.8/4.6 ms; flip angle = 8°; matrix = 256 × 256; slice thickness = 1.0 mm; isotropic voxel size = 1 × 1 × 1 mm³); (ii) Fluid-Attenuated Inversion Recovery (FLAIR): 3D volumetric acquisition (TR/TE = 4500/333 ms; FOV = 258 × 247 mm²; 200 slices; slice thickness = 1.0 mm); (iii) Resting-State fMRI: Gradient-echo planar imaging (EPI) sequence (TR/TE = 2000/30 ms; flip angle = 90°; FOV = 192 × 192 mm²; matrix = 64 × 64; 35 axial slices; slice thickness = 4.0 mm; 240 volumes).

Resting-state fMRI preprocessing was implemented in DPABI v8.0 ((V8.0, http://rfmri.org/dpabi/) and SPM12 (www.fil.ion.ucl.ac.uk/spm) under MATLAB R2020b platform, following standardized pipelines: (i) DICOM-to-NIFTI format conversion (ii) Elimination of the initial 10 time points to ensure signal stabilization; (iii) Slice timing correction (interleaved ascending acquisition); (iv) Realignment with motion censoring (exclusion criteria: more than 3 mm of translation or more than 3 degrees of rotation); (v)T1-based segmentation of brain tissues (gray matter, white matter, and cerebrospinal fluid) along with the calculation of total intracranial volume (TIV); (vi) Co-registration of functional images to T1 images and normalization to Montreal Neurological Institute (MNI) space (3 × 3 × 3 mm³ isotropic voxels); (vii) Spatial smoothing using a 6 × 6 × 6 mm³ Gaussian kernel; (viii) Removal of linear trends; (ix) Regression of nuisance signals (white matter, CSF, and Friston 24 head motion parameters); (x) Detrending and bandpass filtering (0.01 Hz < f < 0.1 Hz).

DMN functional gradient mapping

Cortical gradient analysis was implemented using BrainSpace v0.1.³² The DMN was delineated based on the 400-region Schaefer atlas,³³ with 8962 DMN-specific voxels selected for blood-oxygen-level-dependent (BOLD) time-series extraction. Subject-level functional connectivity matrices (8962 × 400) were constructed by computing Pearson correlations between each DMN voxel's BOLD signal and Schaefer parcel-averaged signals. Connectivity matrices underwent Fisher z-transformation for variance stabilization, followed by row-wise sparsification retaining the top 10% edge weights.

A cosine similarity kernel (8962 × 8962) was embedded into low-dimensional manifolds via diffusion mapping.³⁴ Group-level gradient templates were derived from averaged connectivity matrices, with individual gradients aligned through Procrustes rotation. The first three principal gradients (G1-G3) were retained for analysis, with G1 demonstrating maximal variance explained (Figure 1C).

Voxel-wise gradient deviations in G1 were subjected to false discovery rate (FDR) correction (q < 0.05) to identify DMN subregions exhibiting significant associations with cognitive deterioration. These thresholded regions served as seeds for subsequent functional connectivity (FC) analysis. Seed-to-voxel FC strength was computed using Fisher z-transformed Pearson correlation coefficients between seed time courses and whole-brain voxels, yielding individual-level connectivity spatial maps for group comparisons.

Statistical analysis

Demographic and clinical characteristics were processed using SPSS Statistics (v25.0, IBM Corp., Armonk) with a two-tailed statistical significance threshold of p < 0.05. All variables underwent normality assessment through skewness (<|3|) and kurtosis (<|10|) evaluations, with non-parametric alternatives (Kolmogorov-Smirnov/Mann-Whitney U tests) applied to non-normally distributed datasets.

The cohort (N = 271) was stratified into four clinically defined groups: APOE ε4-negative healthy controls (HC-ε4-, n = 108); APOE ε4-positive healthy controls (HC-ε4+, n = 25); APOE ε4-negative AD patients (AD-ε4-, n = 81); APOE ε4-positive AD patients (AD-ε4+, n = 57).

Intergroup comparisons of continuous variables (age, education years, AD biomarker profiles, MMSE scores, neuropsychological battery metrics) were conducted using one-way ANOVA (with Tukey post-hoc) or Kruskal-Wallis H-test, as appropriate. Categorical variables (sex, hypertension, diabetes, hyperlipidemia prevalence) were analyzed through Pearson χ² tests with Yates’ continuity correction.

For neuroimaging analyses, SPM12 (Wellcome Centre for Human Neuroimaging) was implemented to execute a 2 × 2 factorial design (Group [HC/AD] × APOE ε4 status [carrier/non-carrier]) with age, sex, and education as nuisance covariates. Whole-brain functional gradient alterations in the DMN were identified via cluster-level inference using AlphaSim correction (initial voxel-wise p < 0.01, minimum cluster extent = 38 voxels, Gaussian kernel FWHM = 6 mm, 10,000 Monte Carlo simulations).

Significant interaction clusters (p < 0.05 FDR-corrected) were extracted as regions of interest (ROIs) using MarsBar v0.44. Post-hoc seed-based functional connectivity (FC) analyses were conducted in DPABI v8.0,³⁵ employing Gaussian Random Field (GRF) correction at voxel-level p < 0.001 and cluster-level p < 0.05. Partial correlation analyses controlling for demographic confounders were performed to interrogate DMN gradient-cognition relationships in ε4 carriers.

Mediation modeling (5000 bootstrap resamples) examined tripartite relationships between: Independent variable: Plasma Aβ_42/40 ratio; Mediator: resting-state functional connectivity; Dependent variable: DMN functional gradient. Path significance was determined through 95% bias-corrected confidence intervals.

Results

Demographic and clinical characteristics

Demographic distributions and comorbidity profiles across four groups are comprehensively detailed in Table 1. No significant intergroup differences emerged for age (p = 0.11), sex distribution (p = 0.17), educational attainment (p = 0.16), or prevalence of hypertension (p = 0.14), diabetes (p = 0.30), and hyperlipidemia (p = 0.85).

Table 1.

Comparison of general demographic and clinical characteristics of the participants.

Characteristic	HCAPOE ε4 (-)	HCAPOE ε4 (+)	ADAPOE ε4 (-)	ADAPOE ε4 (+)	F/x²	p
Characteristic	(n = 108)	(n = 25)	(n = 81)	(n = 57)	F/x²	p
Age (y)	67.84 ± 8.26	66.00 ± 10.79	69.70 ± 9.26	70.39 ± 9.83	2.01	0.11
Male, n (%)	49 (45.37%)	16 (64.00%)	47 (58.02%)	33 (57.89%)	5.09	0.17
Education (y)	12 (6)	9 (11)	9 (3)	12 (7)	5.21	0.16
Hypertension, n (%)	56 (51.85%)	10 (40.00%)	31 (38.27%)	18 (31.58%)	5.41	0.14
Diabetes, n (%)	20 (18.51%)	5 (20.00%)	13 (16.05%)	4 (7.02%)	3.70	0.30
Hyperlipidemia, n (%)	29 (26.85%)	5 (20.00%)	17 (20.99%)	12 (21.05%)	0.81	0.85
Plasma Aβ₄₂ (pg/mL)	7.55 ± 1.44^a	7.34 ± 1.36^a	5.37 ± 1.43^b	5.40 ± 1.45^b	47.80	<0.001
Plasma Aβ₄₂/40 ratio	0.07 ± 0.01^a	0.07 ± 0.01^a	0.05 ± 0.01^b	0.05 ± 0.01^b	47.51	<0.001
Plasma P-tau181 (pg/mL)	1.84 (1.23)^b	1.78 (1.12)^b	3.00 (2.11)^a	3.07 (2.65)^a	50.50	<0.001

Continuous data are presented as mean ± SD or median (IQR); categorical data as n (%).

Superscript letters (a, b) indicate statistically significant pairwise differences (post-hoc tests, *p* < 0.05).

Aβ: amyloid-β; APOE: apolipoprotein E; P-tau: phosphorylated tau; AD: Alzheimer's disease; HC: healthy controls.

Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001.

Plasma biomarker analyses revealed distinct pathophysiological signatures: Relative to healthy controls, AD patients demonstrated a significant 28.6–28.9% decrease in plasma Aβ₄₂ concentrations (p < 0.001), independent of APOE ε4 status (p > 0.05 for ε4 + versus ε4- comparisons within the AD cohort). Aβ_42/40 ratio showed a significant reduction of approximately 25–26.4% (p < 0.001). P-tau181 levels were markedly elevated by 62.6–72.4% (p < 0.001), with APOE ε4 carriers demonstrating greater increases (potentially reflecting accelerated tau pathology).

Neuropsychological tests

As delineated in Table 2, this study demonstrates a significant association between APOE ε4 carrier status and cognitive impairment in AD. Key findings from neuropsychological assessments across HC and AD groups (stratified by APOE ε4) include:

Table 2.

Comparison of cognitive test results across four groups.

Items	HCAPOE ε4 (-)	HCAPOE ε4 (+)	ADAPOE ε4 (-)	ADAPOE ε4 (+)	F/x²	p
Items	(n = 108)	(n = 25)	(n = 81)	(n = 57)	F/x²	p
MMSE	29.00 (2)^a	29.00 (4)^a	26.00 (6)^b	23.50 (7)^b	123.87	<0.001
MoCA	26 (2)^a	25 (4)^a	19 (6)^b	21 (9)^b	159.99	<0.001
Memory	−0.08 (0.47)	−0.31 (0.72)	−0.55 (0.49)	−0.31 (0.78)	6.94	0.07
EF	0.23 ± 0.59^a	0.07 ± 0.57^a	−0.52 ± 0.41^b	−0.03 ± 1.45^b	10.23	<0.001
IPS	0.44 ± 0.71^a	0.18 ± 0.71^a	−0.59 ± 0.72^b	−0.45 ± 0.84^b	23.19	<0.001
VPF	0.61 (0.78)^a	0.42 (1.55)^a	−0.78 (1.42)^b	−0.54 (1.28)^b	41.03	<0.001

Data are presented as mean ± SD or median (IQR) based on distribution normality.

Superscript letters (a, b) denote significant intergroup differences (post-hoc tests, p < 0.05).

APOE: apolipoprotein E; AD: Alzheimer's disease; HC: healthy controls; MMSE: Mini-Mental State Examination; MoCA: Montreal Cognitive Assessment; EF: executive function; IPS: information processing speed; VPF: visuospatial function

Both MMSE and MoCA showed consistent intergroup differences (all p < 0.001). Healthy controls (regardless of ε4 status) scored markedly higher than AD groups, indicating generalized cognitive decline in AD. Notably, AD ε4 + group had lower median MMSE (23.5) than AD ε4- (26), suggesting ε4 allele may exacerbate cognitive deficits.

Executive function (EF) and information processing speed (IPS) were significantly impaired in AD (both p < 0.001), but unaffected in healthy ε4 carriers versus non-carriers. The visuospatial function (VPF) exhibited the most pronounced intergroup differences (F = 41.03), with the AD group demonstrating negative scores that were significantly lower than those of the healthy control group, indicating a specific deficit in spatial processing capacity. Memory differences were nonsignificant (p = 0.07), though AD groups trended lower, possibly due to sample size or test sensitivity.

Among AD patients, ε4 carriers and non-carriers showed comparable cognitive performance, implying ε4 primarily increases AD risk rather than modulating disease severity. Healthy ε4 carriers, despite preserved global cognition, had marginally lower IPS and EF scores than non-carriers, warranting investigation into preclinical impairment.

Comparison of the DMN gradient

To compare differences in DMN functional gradients across groups, we computed the relationship between functional gradient values of DMN voxels and their anatomical spatial distances, extracting the first gradient. Figure 2 displays the group-averaged first functional gradient maps of the DMN, with warm colors indicating regions of higher principal gradient values and cool colors representing areas with lower principal gradient values.

Figure 2.

The group-averaged first functional gradient maps of the DMN. L: left; R: right.

After controlling for sex, age, and years of education, we performed a two-way ANOVA on the first functional gradient of the DMN to evaluate the interaction between group (healthy controls [HC] versus AD patients) and APOE ε4 carrier status (carriers versus non-carriers). As shown in Figure 3A, significant differences in functional gradients were observed among the four groups in the right inferior temporal gyrus (ITG; MNI coordinates: 63, −12, −30; p < 0.001).³⁶ Multiple comparisons were corrected using the AlphaSim method (threshold: p < 0.01, minimum cluster size > 38 voxels).

Figure 3.

DMN functional gradient analysis. A) The group (HC versus AD) × APOE ε4 status (carriers versus non-carriers) interaction revealed significant differences in the right inferior temporal gyrus (ITG; MNI coordinates: 63, −12, −30). B) Post hoc comparisons among the four groups. Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001. APOE: apolipoprotein E; AD: Alzheimer's disease; HC: healthy controls; L: left; R: right.

The post hoc comparative analysis among the four groups (Figure 3B) revealed the following key findings:

In APOE ε4 non-carriers: The AD group showed significantly elevated DMN functional gradients compared to healthy controls (p < 0.001), suggesting potential compensatory upregulation.

In APOE ε4 carriers: The AD group exhibited significantly reduced DMN functional gradients in the right temporal lobe region compared to healthy controls (p = 0.031), indicating possible decompensation of DMN gradients with AD progression under the influence of the APOE ε4 allele.

Among healthy controls: No significant difference in DMN functional gradients was observed between APOE ε4 carriers and non-carriers (p > 0.05).

In AD patients: APOE ε4 carriers demonstrated significantly lower DMN functional gradients than non-carriers (p < 0.001), suggesting that the APOE ε4 genotype exerts a significant detrimental effect on DMN gradient reduction during AD progression.

Correlation analysis of the DMN gradient with cognitive decline

After controlling for age, sex, and years of education, APOE ε4 carriers exhibited significant positive correlations between the DMN functional gradient in the right ITG and multiple cognitive domains (Table 3). Specifically, higher DMN functional gradients were associated with better performance in:

Global cognition (MoCA score: r = 0.240, p = 0.034)

MoCA attention subscore: r = 0.273, p = 0.025)

MoCA executive subscore: r = 0.361, p = 0.002)

MoCA calculation subscore: r = 0.287, p = 0.017)

Table 3.

Correlation analysis between DMN gradients and cognitive performance.

	p	R
MMSE	0.404	0.095
MoCA (total)	0.034*	0.240
MoCA-Executive	0.002**	0.361
MoCA-Attention	0.025*	0.273
MoCA-Language	0.059	0.232
MoCA-Calculation	0.017*	0.287
VPF	0.320	0.203
Memory span	0.031*	0.323
EF	0.012*	0.434
IPS	0.004**	0.431
AVLT-Long Delay	0.039*	0.316

MMSE: Mini-Mental State Examination; MoCA: Montreal Cognitive Assessment; VPF: visuospatial function; EF: executive function; IPS: information processing speed; AVLT: Auditory Verbal Learning Test.

Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001.

Furthermore, as illustrated in Figure 4, the right temporal DMN functional gradient in APOE ε4 carriers also correlated positively with:

Information processing speed (IPS): r = 0.431, p = 0.004

Executive function (EF): r = 0.434, p = 0.012

Memory span: r = 0.323, p = 0.031

Long-delay recall (RAVLT): r = 0.316, p = 0.039

Figure 4.

Correlation analysis between DMN functional gradients and cognitive functions. (A) DMN gradients showed a positive correlation with MoCA total scores (Montreal Cognitive Assessment). (B) DMN gradients were positively correlated with information processing speed (IPS). (C) DMN gradients exhibited a positive association with executive function (EF). (D) DMN gradients demonstrated a significant positive correlation with memory span.

In contrast, no significant correlations were observed between DMN gradients and cognitive performance in APOE ε4 non-carriers. These findings suggest that disease progression in APOE ε4 carriers is associated with progressive decline in both DMN functional gradients and cognitive performance, highlighting the gene's role in modulating neural network efficiency and cognitive resilience.

Mediation effect of FC on the relationship between blood AD markers and the DMN gradient

Using brain regions exhibiting significant DMN gradient changes correlated with cognitive impairment as seed points, we performed static functional connectivity (FC) analysis. Seed-based FC was calculated, and correlation analyses between gradient values and FC were conducted using DPABI software, with a significance threshold of p < 0.001 (Gaussian Random Field [GRF] correction: voxel p < 0.01, cluster p < 0.05). Based on the Automated Anatomical Labeling 90 (AAL90) atlas,³⁶ the left superior frontal gyrus (SFG) emerged as a significantly correlated region (Figure 5A).

Figure 5.

Functional connectivity (FC) and mediation analysis.

In APOE ε4 carriers, we examined whether resting-state FC between the left SFG (significant cluster) and right inferior temporal gyrus (ITG; seed region) mediated the relationship between DMN gradients and AD blood biomarkers. Mediation analysis (Figure 5B) revealed that: Left SFG-right ITG FC significantly mediated the association between Aβ₄₂/Aβ₄₀ ratio and DMN functional gradients (p < 0.05). Reduced Aβ₄₂/Aβ₄₀ levels → Weaker left SFG-right ITG FC → Lower DMN gradients. The Aβ₄₂/Aβ₄₀ ratio exerted a positive influence on DMN functional gradients through left superior frontal gyrus-right inferior temporal gyrus functional connectivity (indirect effect = 1.295, 95% CI [0.070, 3.670], p < 0.05). Notably, this mediation effect was not observed in APOE ε4 non-carriers.

These results suggest that in APOE ε4 carriers, AD blood biomarkers are associated with disruption of the DMN functional organization and impaired FC between prefrontal and temporal hubs, potentially accelerating cognitive decline. The left SFG's role as a mediator highlights its vulnerability in ε4-related AD pathogenesis

Discussion

This study pioneers the systematic characterization of dynamic reorganization in DMN functional connectivity gradients in AD through resting-state fMRI gradient analysis. By integrating multimodal biomarkers, we delineate mechanistic pathways through which the APOE ε4 allele may modulate neurodegenerative cascades. Critically, we identified a significant group × genotype interaction in the right inferior temporal gyrus (RITG), where gradient compression—operationalized as reduced functional hierarchy dimensionality at this DMN integration hub—appears to reflect Aβ-induced transmodal integration failure. Notably, APOE ε4 carriers with AD demonstrated pronounced DMN gradient attenuation relative to non-carriers, and the gradient reductions correlate strongly with deficits of several multiple cognitive domains. These findings posit that APOE ε4-mediated acceleration of DMN hierarchical disintegration constitutes a novel pathway for cognitive deterioration, offering clinically actionable biomarkers for gene-endophenotype trajectory mapping. Our structural equation modeling revealed a cascading pathway from molecular dysregulation to network failure: diminished serum Aβ₄₂/Aβ₄₀ ratios predicted DMN gradient abnormalities through disrupted RITG-left superior frontal gyrus (LSFG) connectivity. This discovery has established for the first time a quantitative pathway linking amyloid dynamics to macroscopic brain network topological disorder, providing translational medical evidence for the integration of biomarkers in AD.

Firstly, we found a significant interaction between group (AD versus HC) and APOE ε4 genotype in the RITG. While APOE ε4-negative AD patients exhibited compensatory RITG gradient elevation, APOE ε4 carriers displayed progressive gradient erosion (p < 0.001 Alpha Sim-corrected), suggesting allele-specific impairment of neuroadaptive capacity. This spatial-topographic dichotomy reveals APOE ε4's dual pathology: suppression of early compensatory mechanisms coupled with potentiation of late-stage network decompensation—a dynamic that may underlie the accelerated cognitive decline observed in ε4 carriers. Our RITG findings converge with emerging evidence of APOE ε4-associated DMN segregation deficits,³⁷ further substantiating its role in network integrity collapse. Methodologically, the gradient analytic framework extends conventional connectivity approaches by quantifying multidimensional hierarchy reorganization.

The RITG engages in DMN-associated higher-order cognitive functions—including memory integration, semantic network regulation, and affective processing—through structural connectivity via white matter tracts (e.g., inferior longitudinal fasciculus, uncinate fasciculus) and functional coupling with DMN hubs (parahippocampal gyrus, posterior cingulate cortex).³⁸ Neuroimaging evidence demonstrates reduced task-based functional activation and coherence in the right ITG of AD patients compared to healthy controls, correlating positively with global cognitive scores.³⁹ Convergent findings from Convit et al.⁴⁰ identify the ITG as an early vulnerable site in AD pathophysiology, with its degeneration strongly associated with episodic memory decline. Pathological insults to the ITG may propagate to DMN core regions via trans-network connections (e.g., inferior longitudinal fasciculus), potentially accelerating whole-brain network disintegration. Our study extends these observations by revealing an APOE ε4 × AD interaction driving abnormal functional gradient reduction in the right ITG (p < 0.001 Alpha Sim-corrected). This gradient attenuation likely reflects diminished functional segregation between RITG and DMN hubs, impairing cross-network information transfer efficiency. The observed gradient decline may signify the ITG's loss of hub dominance within the DMN, disrupting visuo-semantic integration into memory systems—a hypothesis corroborated by significant positive correlations between right temporal DMN gradients and multiple cognitive domains (MoCA: r = 0.240; executive function: r = 0.434; memory span: r = 0.323; Information-processing Speed: r = 0.431; AVLT: r = 0.316) in ε4 carriers after controlling for age, sex, and education (all p < 0.05). Notably, no such associations emerged in non-carriers, suggesting ε4 exerts selective effects on RITG network properties rather than promoting global network degradation. The identified RITG-LSFG disconnection pathway in amyloid-related network failure—a clinically relevant discovery given LSFG's centrality in cognitive integration circuits. These findings align with reports of ε4-associated RITG gradient decline correlating with progressive cognitive deterioration.⁴¹ Mechanistically, ε4 may exacerbate cognitive impairment by reducing RITG segregation from heteromodal cortices, thereby disrupting multimodal integration—a process consistent with the network dedifferentiation hypothesis positing loss of functional specificity in AD. The RITG gradient abnormality may further signal DMN compensatory failure: while prefrontal regions might exhibit early hyperconnectivity to offset DMN damage,⁴² RITG gradient collapse could reflect exhaustion of such adaptive reserves. Longitudinal studies integrating tau-PET and dynamic fMRI are warranted to test this model.

Our mediation analysis delineated a neurobiological pathway underlying gradient abnormalities: in APOE ε4 carriers, functional connectivity between LSFG and RITG significantly mediated the relationship between serum Aβ₄₂/Aβ₄₀ ratio and DMN gradients (95%CI [0.070,3.670]). Given LSFG's role in self-referential cognition and episodic memory integration,⁴³ While plasma Aβ₄₂/Aβ₄₀ effectively reflects cerebral Aβ pathology⁴⁴ and correlates with network dysfunction,⁴⁵ Aβ-induced SFG-ITG decoupling may provoke cortical hierarchy disruption, manifesting as semantic-episodic memory dissociation. Certainly, our biomarker focus on Aβ pathways constitutes a limitation—APOE ε4 may also concurrently exacerbate tau phosphorylation, microglial activation, or cause synaptic pruning dysregulation through complementary mechanisms. Multimodal studies incorporating tau and neuroinflammatory markers are needed to dissect these interactive pathways.

This investigation has several methodological constraints. First, the cross-sectional design precludes causal inferences regarding APOE ε4's temporal effects on functional gradient dynamics in AD. Longitudinal studies tracking gradient trajectories are needed to evaluate their utility as prognostic biomarkers. Second, while APOE ε4 represents the primary genetic risk factor for late-onset AD with dose-dependent effect,⁴⁶ the limited number of ε4/ε4 homozygotes in our cohort (n = 5) restricts analysis of allelic dosage impacts on disease progression. Larger cohorts with balanced genotype distributions are required to validate dose-response relationships. Third, our biomarker focus on amyloid pathways overlooks complementary mechanisms; future studies integrating tau-PET, neuroinflammatory markers (e.g., GFAP, IL-6), and polygenic risk scores could clarify spatiotemporal relationships between proteinopathy propagation and gradient degeneration. Furthermore, the participants in this study were exclusively recruited from a Chinese population, which may limit the generalizability of our findings to other ethnic groups. Future studies incorporating multi-ethnic cohorts are necessary to validate and extend our results.

Conclusions

This study applied gradient analysis to the hierarchical dissection of the DMN in AD patients, overcoming the limitations of traditional functional connectivity in sensitivity to network topology. The analysis identified specific regulatory pathways of the APOE ε4 genotype on DMN dynamic reconfiguration, particularly its amplifying effect on the weakening of the RITG gradient, the data also suggest that amyloid deposition is associated with disruption of the LSFG-RITG connection and relates that change to the macroscopic network disorder and excacerbation of cognitive impairment. These findings provide more sensitive neuroimaging biomarkers for the early diagnosis of AD and potential targets for precision medicine prevention and intervention approaches.

Footnotes

Acknowledgements

We are deeply grateful to all the participants in this study, as well as their family members, for their invaluable cooperation and trust. We also extend our sincere appreciation to everyone who contributed to the data collection process. Their dedication and meticulous efforts were fundamental to the completion of this research.

ORCID iDs

WenTing Song

ChengBing Gong

YiFei Li

Dan Yang

HaiFeng Chen

LiMoRang Tang

Yun Xu

Hui Zhao

Ethical considerations

The studies involving human participants were reviewed and approved by the Drum Tower Hospital Research Ethics Committee.

Consent to participate

The participants provided their written informed consent to participate in this study.

Author contribution(s)

WenTing Song: Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Writing – original draft.

ChengBing Gong: Data curation; Methodology; Software.

YiFei Li: Formal analysis.

Dan Yang: Conceptualization; Writing – review & editing.

HaiFeng Chen: Methodology.

LiMoRang Tang: Methodology.

Yun Xu: Writing – review & editing.

Hui Zhao: Conceptualization; Methodology; Writing – original draft; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Nanjing Medical Science and Technology Development Key Projects (ZKX23026) and the National Natural Science Foundation of China (82471454).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding authors.

References

Hardy

Higgins

. Alzheimer’s disease: the amyloid cascade hypothesis. Science 1992; 256: 184–185.

Scheltens

De Strooper

Kivipelto

, et al. Alzheimer’s disease. Lancet 2021; 397: 1577–1590.

Hansson

. Biomarkers for neurodegenerative diseases. Nat Med 2021; 27: 954–963.

Terry

Masliah

Salmon

, et al. Physical basis of cognitive alterations in Alzheimer’s disease: synapse loss is the major correlate of cognitive impairment. Ann Neurol 1991; 30: 572–580.

Ittner

Götz

. Amyloid-β and tau–a toxic pas de deux in Alzheimer’s disease. Nat Rev Neurosci 2011; 12: 65–72.

Jack

Knopman

Jagust

, et al. Hypothetical model of dynamic biomarkers of the Alzheimer’s pathological cascade. Lancet Neurol 2010; 9: 119–128.

Morris

Roe

Xiong

, et al. APOE Predicts amyloid-beta but not tau Alzheimer pathology in cognitively normal aging. Ann Neurol 2010; 67: 122–131.

Corder

Saunders

Strittmatter

, et al. Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer’s disease in late onset families. Science 1993; 261: 921–923.

Farrer

Cupples

Haines

, et al. Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and Alzheimer Disease Meta Analysis Consortium. JAMA 1997; 278: 1349–1356.

10.

Lambert

J-C

Ibrahim-Verbaas

Harold

, et al. Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer’s disease. Nat Genet 2013; 45: 1452–1458.

11.

Rebeck

Reiter

Strickland

, et al. Apolipoprotein E in sporadic Alzheimer’s disease: allelic variation and receptor interactions. Neuron 1993; 11: 575–580.

12.

Castellano

Kim

Stewart

, et al. Human apoE isoforms differentially regulate brain amyloid-β peptide clearance. Sci Transl Med 2011; 3: 89ra57.

13.

Liu

C-C

Zhao

, et al. Apoe4 accelerates early seeding of amyloid pathology. Neuron 2017; 96: 1024–1032.e3.

14.

Zhao

Ren

Yamazaki

, et al. Alzheimer’s risk factors age, APOE genotype, and sex drive distinct molecular pathways. Neuron 2020; 106: 727–742.e6.

15.

Lin

Y-T

Seo

Gao

, et al. APOE4 Causes widespread molecular and cellular alterations associated with Alzheimer’s disease phenotypes in human iPSC-derived brain cell types. Neuron 2018; 98: 1141–1154.e7.

16.

Zhao

Davis

Martens

, et al. APOE Ε4/ε4 diminishes neurotrophic function of human iPSC-derived astrocytes. Hum Mol Genet 2017; 26: 2690–2700.

17.

Chung

W-S

Verghese

Chakraborty

, et al. Novel allele-dependent role for APOE in controlling the rate of synapse pruning by astrocytes. Proc Natl Acad Sci U S A 2016; 113: 10186–10191.

18.

Mian

, et al. Neuroinflammation: a modifiable pathway linking obesity, Alzheimer’s disease, and depression. Am J Geriatr Psychiatry 2023; 31: 853–866.

19.

Shi

Yamada

Liddelow

, et al. Apoe4 markedly exacerbates tau-mediated neurodegeneration in a mouse model of tauopathy. Nature 2017; 549: 523–527.

20.

Shi

Manis

Long

, et al. Microglia drive APOE-dependent neurodegeneration in a tauopathy mouse model. J Exp Med 2019; 216: 2546–2561.

21.

Mishra

Blazey

Holtzman

, et al. Longitudinal brain imaging in preclinical Alzheimer disease: impact of APOE ε4 genotype. Brain 2018; 141: 1828–1839.

22.

Shi

Gutman

, et al. Influence of APOE genotype on hippocampal atrophy over time - an N=1925 surface-based ADNI study. PLoS One 2016; 11: e0152901.

23.

Smith

Andersen

Kryscio

, et al. Altered brain activation in cognitively intact individuals at high risk for Alzheimer’s disease. Neurology 1999; 53: 1391–1396.

24.

Filippini

MacIntosh

Hough

, et al. Distinct patterns of brain activity in young carriers of the APOE-epsilon4 allele. Proc Natl Acad Sci U S A 2009; 106: 7209–7214.

25.

Jones

Knopman

Gunter

, et al. Cascading network failure across the Alzheimer’s disease spectrum. Brain 2016; 139: 547–562.

26.

Damoiseaux

Prater

Miller

, et al. Functional connectivity tracks clinical deterioration in Alzheimer’s disease. Neurobiol Aging 2012; 33: 828.e19–30.

27.

Sheline

Morris

Snyder

, et al. APOE4 Allele disrupts resting state fMRI connectivity in the absence of amyloid plaques or decreased CSF Aβ42. J Neurosci 2010; 30: 17035–17040.

28.

Huntenburg

Bazin

P-L

Margulies

. Large-scale gradients in human cortical organization. Trends Cogn Sci 2018; 22: 21–31.

29.

Liu

Chen

Cheng

, et al. Functional connectivity gradients and neurotransmitter maps among patients with mild cognitive impairment and depression symptoms. J Psychiatry Neurosci 2025; 50: E11–E20.

30.

Hong

S-J

Nikolaidis

, et al. Toward a connectivity gradient-based framework for reproducible biomarker discovery. Neuroimage 2020; 223: 117322.

31.

Yang

, et al. Subsystem mechanisms of default mode network underlying white matter hyperintensity-related cognitive impairment. Hum Brain Mapp 2023; 44: 2365–2379.

32.

Vos de Wael

Benkarim

Paquola

, et al. Brainspace: a toolbox for the analysis of macroscale gradients in neuroimaging and connectomics datasets. Commun Biol 2020; 3: 103.

33.

Schaefer

Kong

Gordon

, et al. Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cereb Cortex 2018; 28: 3095–3114.

34.

Coifman

Lafon

. Diffusion maps. Appl Comput Harmon Anal 2006; 21: 5–30.

35.

Yan

C-G

Wang

X-D

Zuo

X-N

, et al. DPABI: data processing & analysis for (resting-state) brain imaging. Neuroinformatics 2016; 14: 339–351.

36.

Tzourio-Mazoyer

Landeau

Papathanassiou

, et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 2002; 15: 273–289.

37.

Jones

Machulda

Vemuri

, et al. Age-related changes in the default mode network are more advanced in Alzheimer disease. Neurology 2011; 77: 1524–1531.

38.

Lin

Y-H

Young

Conner

, et al. Anatomy and white matter connections of the inferior temporal gyrus. World Neurosurg 2020; 143: e656–e666.

39.

Y-Q

Wang

Y-N

Zhang

L-J

, et al. Regional homogeneity in patients with mild cognitive impairment: a resting-state functional magnetic resonance imaging study. Front Aging Neurosci 2022; 14: 877281.

40.

Convit

de Asis

de Leon

, et al. Atrophy of the medial occipitotemporal, inferior, and middle temporal gyri in non-demented elderly predict decline to Alzheimer’s disease. Neurobiol Aging 2000; 21: 19–26.

41.

Franzmeier

Neitzel

Rubinski

, et al. Functional brain architecture is associated with the rate of tau accumulation in Alzheimer’s disease. Nat Commun 2020; 11: 347.

42.

Wang

Yao

, et al. Shared and differing functional connectivity abnormalities of the default mode network in mild cognitive impairment and Alzheimer’s disease. Cereb Cortex 2024; 34: bhae094.

43.

Raichle

MacLeod

Snyder

, et al. A default mode of brain function. Proc Natl Acad Sci U S A 2001; 98: 676–682.

44.

Janelidze

Teunissen

Zetterberg

, et al. Head-to-head comparison of 8 plasma amyloid-β 42/40 assays in Alzheimer disease. JAMA Neurol 2021; 78: 1375–1382.

45.

Dark

Shafer

Cordon

, et al. Association of plasma biomarkers of Alzheimer disease and neurodegeneration with longitudinal intra-network functional brain connectivity. Neurology 2025; 104: e210271.

46.

Koutsodendris

Nelson

Rao

, et al. Apolipoprotein E and Alzheimer’s disease: findings, hypotheses, and potential mechanisms. Annu Rev Pathol 2022; 17: 73–99.