Influence of local spontaneous activity and hippocampal volume on the relationship between APOE genotype, ATN biomarkers, and cognitive function across the Alzheimer’s disease spectrum

To the Editor: Alzheimer’s disease (AD) is a progressive neurodegenerative disorder involving complex interactions among genetics and structural and functional changes in the brain.^[1] While the APOE ε4 allele is the strongest genetic risk factor for sporadic AD, its influence on cognition varies across the AD spectrum.^[2–4] In this study, we investigated whether hippocampal volume and spontaneous local brain activity, measured by four-dimensional consistency of local neural activities (FOCA), moderate or mediate the relationship between the APOE genotype and cognitive performance.

We analyzed multimodal neuroimaging and cerebrospinal fluid (CSF) data from 154 individuals across the AD spectrum (46 cognitively normal [CN]), 35 with subjective cognitive decline [SCD], 41 with mild cognitive impairment [MCI], and 32 with AD from the Alzheimer’s Disease Neuroimaging Initiative [ADNI]). Carriers of APOE ε4 (n = 75) were compared with non-carriers (n = 79). This study used de-identified public data from the ADNI database, with the core ethical approval granted by the Institutional Review Board (IRB) of the University of California, San Francisco (UCSF) (Approval No. 10-001117); the complete list of IRBs that approved the ADNI study for all participating institutions has been provided in the Supplementary Material, http://links.lww.com/CM9/C884. Cognitive function was assessed using a battery of neuropsychological tests. Mediation and moderation analyses were conducted using PROCESS macros with bootstrap confidence intervals (CIs). Additionally, we used a support vector machine (SVM) model incorporating APOE genotype and magnetic resonance imaging (MRI) biomarkers to distinguish individuals across the AD spectrum. Data on demographics, amyloid/tau/neurodegeneration (ATN) biomarkers, and neuropsychological parameters are listed in Table 1. The predominant effect of the APOE genotype on FOCA was observed in the frontal-temporal-occipital system, as illustrated in Supplementary Figure 1A–B, http://links.lww.com/CM9/C884. Compared with non-carriers of APOE ε4, APOE ε4 carriers exhibited significantly higher intrinsic FOCA in the left middle frontal gyrus (MFG.L), but lower FOCA in the right middle occipital gyrus (MOG.R) and right superior/middle temporal gyrus (STG.R/MTG.R). The major effect of the disease was seen in the frontal-temporal system, as shown in Supplementary Figure 1C–D, http://links.lww.com/CM9/C884. Notably, FOCA distribution indicated a disrupted pattern with dynamic changes (U-shape and inverted U-shape) throughout AD progression. The effects of the interaction between the APOE genotype and disease on FOCA were observed in the left frontal gyrus (FG.L), left inferior parietal lobe (IPL.L), right inferior frontal gyrus (IFG.R), and bilateral precuneus (PCUN.B), as illustrated in Supplementary Figure 2A, http://links.lww.com/CM9/C884. Compared with non-carriers of APOE ε4, APOE ε4 carriers exhibited opposite changes in FOCA trajectories during the progression of AD, particularly in the transitions from the CN to SCD stages and from the MCI to AD stages, as illustrated in Supplementary Figure 2B, http://links.lww.com/CM9/C884. Further details on these brain regions are provided in Supplementary Table 1, http://links.lww.com/CM9/C884. As shown in Supplementary Figure 3A–E, http://links.lww.com/CM9/C884, linear regression analyses revealed that the STG.R FOCA, hippocampal volume, and 18F-fluorodeoxyglucose (FDG, Siemens Healthineers, Erlangen, Germany) standardized uptake value ratio (SUVR) were inversely correlated with global cognitive scores (ADAS-Cog), whereas 18F-Florbetapir (AV45, Avid Radiopharmaceuticals, Philadelphia, PA, USA) SUVR and CSF Tau showed positive correlations with ADAS-Cog scores. Additionally, significant correlations were observed between FOCA, hippocampal volume, FDG SUVR, CSF Tau, and scores in other cognitive domains [Supplementary Figures 4–6, http://links.lww.com/CM9/C884]. Classification of the AD spectrum population using an SVM-trained model that integrated APOE genotype with MRI biomarkers, specifically, gray matter volume (GMV), hippocampal volume, and modified FOCA, showed that the model was highly effective in classifying the different stages of the disease. Supplementary Figure 7A, http://links.lww.com/CM9/C884 shows that all AUC values exceeded 0.70, indicating the robust predictive capacity of these combined variables in the differentiation of disease stages. The AUC values from the classification model, cross-validated through permutation testing, were statistically significant, as shown in Supplementary Figures 7B–G, http://links.lww.com/CM9/C884. Both FOCA and the hippocampal volume were found to moderate the relationship between the APOE genotype, ATN biomarkers, and cognition. The moderation analysis revealed that hippocampal volume significantly moderated the effects of APOE genotype and FDG SUVR on the ADAS-Cog scores across all groups [Supplementary Figure 8A–F, http://links.lww.com/CM9/C884]. The relationship between APOE genotype/FDG SUVR and ADAS-Cog scores was contingent upon the hippocampal volume. Initially, the APOE genotype exerted a significant positive predictive effect on cognitive performance (β = 33.52, 95% CI: 11.82–55.21, t = 3.05, P = 0.003, Supplementary Figure 8B, http://links.lww.com/CM9/C884). Notably, as indicated in Supplementary Figure 8C, http://links.lww.com/CM9/C884, the hippocampal volume showed a significant negative moderating effect on the impact of the APOE genotype on cognitive performance, suggesting that larger hippocampal volumes reduce the effect of the APOE genotype on cognitive function. This significant moderating effect was observed only when the hippocampal volume fell between 3.6 and 7.0. Supplementary Figure 8E, http://links.lww.com/CM9/C884 indicates that FDG SUVR significantly affected the ADAS-Cog scores (β = −19.43, 95% CI: −25.7 to −13.1, t = −6.27, P <0.001). To further explore the moderating effect of hippocampal volume, a simple slope analysis was conducted. The hippocampal volume was categorized into high and low levels based on the mean ± 1 standard deviation [Supplementary Figure 8F, http://links.lww.com/CM9/C884]. Using ADAS-Cog scores as the dependent variable and FDG SUVR as the independent variable, the analysis showed that FDG SUVR negatively predicted ADAS-Cog scores in both groups: the low hippocampal-volume group (simple slope = −10.95, 95% CI: −14.25 to −7.71, t = −8.42, P <0.001) and the high hippocampal-volume group (simple slope = −7.57, 95% CI: −10.11 to −5.02, t = −6.11, P <0.001). This implies that, for the same FDG SUVR, a larger hippocampal volume is associated with better cognitive function. The effect of FDG SUVR on cognitive function appears to be less pronounced among individuals with larger hippocampal volumes, as indicated by the increasing divergence between the two lines as the FDG SUVR values increased. Significantly, our findings indicated that neither hippocampal volume nor STG FOCA played a significant mediating role in the relationship between APOE genotype/FDG SUVR and cognitive performance, as illustrated in Supplementary Figure 9, http://links.lww.com/CM9/C884. Moreover, no moderation effect of STG FOCA on the relationship between APOE genotype/FDG SUVR and cognitive performance was observed [Supplementary Figure 9, http://links.lww.com/CM9/C884]. Finally, as shown in Supplementary Figure 8G–H, http://links.lww.com/CM9/C884, both hippocampal volume (indirect effect = 9.20, 95% CI: 4.40–14.49) and STG FOCA (indirect effect = 2.95, 95% CI: 0.84–5.74) significantly mediated the effect of AV45 SUVR on cognitive performance across all groups. The AV45 SUVR was found to have a positive influence on ADAS-Cog scores. However, we discovered that neither the hippocampal volume nor STG FOCA significantly moderated the relationship between AV45 SUVR and cognitive performance [Supplementary Figure 9, http://links.lww.com/CM9/C884].

Table 1.

Demographic information and neuropsychological performance of all participants.

Variables	CN (n = 46)	SCD (n = 35)	MCI (n = 41)	AD (n = 32)	Statistics	P value
Age (years)	74.18 ± 5.88	73.22 ± 5.96	71.49 ± 8.14	73.48 ± 7.11	1.42*	0.16
Education (years)	16.13 ± 2.56	16.71 ± 2.71	16.70 ± 2.70	15.75 ± 2.79	1.16*	0.32
Gender (Female/Male)	21/25	16/19	24/17	17/15	1.92†	0.59
APOE ε4 (+/-)	14/32	22/13	17/24	22/10	12.05†	0.002§
GMV (cm3)	589.05 ± 47.45	573.21 ± 47.57	562.71 ± 42.77	532.61 ± 67.29	8.76*	<0.001‡ §|| ¶ **
Hippocampus volume (cm3)	7.62 ± 0.81	7.61 ± 0.92	7.03 ± 1.09	6.07 ± 1.19	9.14*	<0.001‡ §|| ¶ **
PET
AV45 SUVR	1.11 ± 0.17	1.14 ± 0.19	1.26 ± 0.24	1.44 ± 0.20	10.21*	<0.001‡ §|| ¶ **
FDG SUVR	6.68 ± 0.46	6.58 ± 0.49	6.19 ± 0.61	5.24 ± 0.70	9.88*	<0.001‡ §|| ¶ **
General mental status
MMSE	28.80 ± 1.53	28.52 ± 1.96	27.66 ± 1.64	22.53 ± 2.42	11.34*	<0.001‡ §|| ¶ **
ADAS-cog	9.28 ± 3.83	10.92 ± 5.56	16.93 ± 6.89	34.05 ± 9.14	10.05*	<0.001‡ §|| ¶ **
Memory
RAVLT-I	43.95 ± 9.99	43.57 ± 10.51	33.14 ± 7.49	21.93 ± 7.03	9.63*	<0.001‡ §|| ¶ **
RAVLT-L	6 (4–8)	6 (4–7)	3 (2–5)	2 (0–3)	26.53††	<0.001‡ §|| ¶ **
RAVLT-sum	20 (10–30)	39 (20–58)	65 (45–85)	93 (85–102)	31.47††	<0.001‡ §|| ¶ **
Visuospatial function
CDT	4.77 ± 0.55	4.67 ± 0.70	4.27 ± 1.03	3.19 ± 1.27	8.55*	<0.001‡§||¶**
Information processing speed
TMT-A	33 (25–42)	36 (28–48)	48 (36–68)	80 (65–110)	25.63††	<0.001‡§||¶**
Executive function
TMT-B	71 (55–88)	85 (65–120)	135 (100–180)	230 (200–280)	29.85††	<0.001‡§||¶**
Language
BNT	27.89 ± 1.58	27.65 ± 1.74	23.17 ± 5.99	20.20 ± 5.63	8.92*	<0.001‡§||¶**
CSF biomarkers (pg/mL)
Aβ1-42	191.41 ± 48.74	189.36 ± 53.26	169.44 ± 51.38	139.95 ± 38.51	10.45*	<0.001§¶**
t-tau	58 (38–90)	72 (48–98)	88 (52–130)	122 (78–185)	26.79††	<0.001‡§¶**
p-tau181	32 (22–46)	39 (24–56)	43 (28–62)	54 (36–74)	15.32††	0.003‡§¶

Data are shown as mean ± standard deviation (SD), n/N, or median (Q1–Q3). ^*F values; ^†χ² values; ^††H values. No significant differences were found in age, gender and education data. The source of significant differences is indicated by the following symbols: ^‡CN vs. MCI; ^§CN vs. AD; ^||SCD vs. MCI; ^¶SCD vs. AD; ^**MCI vs. AD. Aβ: Amyloid-β; AD: Alzheimer’s disease; ADAS-cog: 13-item Alzheimer’s Disease Assessment Scale-cognitive subscale; BNT: Boston Naming Test; CDT: Clock Drawing Test; CN: Cognitive normal; CSF: Cerebrospinal fluid; FDG: Fludeoxyglucose; GMV: Gray matter volume; MCI: Mild cognitive impairment; MMSE: Mini-mental state examination; p-tau: Phosphorylated tau; PET: Positron emission tomography; RAVLT-I: Rey auditory verbal learning test-immediate; RAVLT-L: Rey auditory verbal learning test-learning; SCD: Subjective cognitive decline; SUVR: Standardized uptake values ratio; TMT-A/B: Trail Making Test A part/ B part; t-tau: Total tau.

These findings enhance our understanding of the links between the APOE genotype, ATN biomarkers, brain structure, function, and cognition. The results also provide new insights into the identification of diagnostic biomarkers for categorizing the AD spectrum. However, the cross-sectional design precludes causal inference, and verification in larger, population-based, and ethnically diverse cohorts is warranted.

This study had several limitations. First, with single-time-point data, it is not possible to rule out reverse causation or reciprocal effects. Longitudinal panel designs are required before translation of these pathways into clinical action. Nevertheless, examining these pathways with cross-sectional data could lead to skewed results. Future research would benefit from assessing mediation and moderation using longitudinal data, preferably employing panel models that enable the analysis of different causal pathways. Second, as limited tau positron emission tomography (PET) imaging was only performed in subjects in the ADNI3 stage, these results were not included in the analysis, and tau levels in the CSF were used instead. Third, self-reported race/ethnicity data from the ADNI database showed that our cohort was predominantly composed of White individuals (85.7%, 132/154), with non-White subgroups (Black/African American, Asian, Latino/Hispanic, Other/Not specified) accounting for only 14.3% of the total sample and all having small cell sizes in each diagnostic group (CN, SCD, MCI, AD; all n ≤2) [Supplementary Table 2, http://links.lww.com/CM9/C884]. Thus, interaction analyses stratified by race/ethnicity were not performed due to insufficient sample size of each subgroup, which limited the generalizability of our findings, and future validation in non-ADNI cohorts with diverse ethnic/racial compositions is warranted. Fourth, the study was limited by a restricted participant pool, comprising individuals with different stages of AD, APOE genotypes, and PET scans. A larger participant pool is needed to confirm the findings. Lastly, the SVM model incorporating the APOE genotype and MRI biomarkers demonstrated strong classification power for the AD spectrum. Despite cross-validation analysis, future research should be conducted on a fresh dataset to confirm the current findings.

Funding

This study was supported by grants from China Postdoctoral Science Foundation (No. 2024M753141), Research Funds of Centre for Leading Medicine and Advanced Technologies of IHM (No. 2023IHM02010), Natural Science Foundation of Anhui Province (Nos. 2208085QH254 and 2208085MH227), Key Research Projects of Natural Science in Colleges and Universities of Anhui Province (No. 2022AH051252), Opening Foundation of State Key Laboratory of Cognitive Intelligence, iFLYTEK (No. COGOS-2023001), National Key Research and Development Program of China (No. STI2030-Major Projects 2021ZD0203100), and the ADNI (National Institutes of Health GrantU01 AG024904).

Conflicts of interest

None.

Supplementary Material

cm9-139-2839-s001.pdf ^{(1.6MB, pdf)}