APOE genotype moderates the relationship between LRP1 polymorphism and cognition across the Alzheimer's disease spectrum via disturbing default mode network

Abstract Aims This study aims to investigate the mechanisms by which apolipoprotein E (APOE) genotype modulates the relationship between low‐density lipoprotein receptor‐related protein 1 (LRP1) rs1799986 variant on the default mode network (DMN) and cognition in Alzheimer's disease (AD) spectrum populations. Methods Cross‐sectional 168 subjects of AD spectrum were obtained from Alzheimer's Disease Neuroimaging Initiative database with resting‐state fMRI scans and neuropsychological scores data. Multivariable linear regression analysis was adopted to investigate the main effects and interaction of LRP1 and disease on the DMN. Moderation and interactive analyses were performed to assess the relationships among APOE, LRP1, and cognition. A support vector machine model was used to classify AD spectrum with altered connectivity as an objective diagnostic biomarker. Results The main effects and interaction of LRP1 and disease were mainly focused on the core hubs of frontal‐parietal network. Several brain regions with altered connectivity were correlated with cognitive scores in LRP1‐T carriers, but not in non‐carriers. APOE regulated the effect of LRP1 on cognitive performance. The functional connectivity of numerous brain regions within LRP1‐T carriers yielded strong power for classifying AD spectrum. Conclusion These findings suggested LRP1 could affect DMN and provided a stage‐dependent neuroimaging biomarker for classifying AD spectrum populations.


| INTRODUC TI ON
Low-density lipoprotein receptor-related protein 1 (LRP1) is a large cell surface transmembrane receptor, highly expressed in the neurons, astrocytes, and vasculatures of the brain. It regulates the pathogenesis of Alzheimer's disease (AD). 1 Reportedly, up to 50 structurally diverse proteins including β-amyloid (Aβ) and apolipoprotein E (APOE) are ligands of LRP1. 2 Notably, LRP1 promotes Aβ clearance, maintains synaptic integrity, and regulates lipid metabolism in the brain. 3,4 LRP1 gene rs1799986 polymorphism in exon 3 has a silent mutation of C allele to T allele and generates three isoform genotypes including CC, TC, and TT. The T allele potentially confers a risk factor for developing sporadic AD. Nonetheless, the relationship of this variant with AD is elusive. Several previous studies reported that LRP1 rs1799986 polymorphism was connected with late-onset AD. [5][6][7][8] However, recent three genome-wide association studies failed to discover a significant impact of this variant on AD risk. [9][10][11] Despite the conflicting findings in polymorphism, one human postmortem study with brain tissues revealed that LRP1 levels from the middle frontal cortex were significantly reduced in AD patients compared with healthy controls. Also, LRP1 levels progressively decreased with the increasing age in controls, whereas a higher level of LRP1 correlated with later age of AD onset. 12 Elsewhere, another postmortem study found significantly decreased LRP1 levels in the hippocampus of mild cognitive impairment (MCI), an early stage of AD, compared with age-matched controls. 13 These observations preliminarily suggest that the disrupted LRP1 levels might partly reflect brain function.
Accumulating evidence from animal studies indicates that LRP1 has been implicated in the process of Aβ and Tau pathology and related to cognitive function. Besides, LRP1 potentially acts predominantly over Aβ clearance in a mouse model of AD. 14 In cerebral blood vessels, LRP1 importantly mediates the rapid removal of Aβ from the brain to transport across the blood-brain barrier; also, endothelial LRP1 may be treated as a potential target for the treatment of AD. 15 Additionally, recent research identified that knockdown LRP1 significantly reduced tau uptake in neurons and tau propagation between neurons. This implies a master regulatory role of LRP1 in tau pathology, thereby providing a novel therapeutic target for tau-related neurodegenerative diseases. 16 Drug trials also have revealed that low-dose pioglitazone ameliorates learning and memory impairment by upregulating LRP1 expression in the hippocampus. 17 Moreover, APOE-ε4 mediates Aβ pathology based on its neuronal receptor LRP1 18 and LRP1 knockout prevents the increase of Aβ pathology caused by APOE-ε4 expression. 19 Therefore, LRP1 is a common factor modulating Aβ and tau metabolism to maintain brain homeostasis.
Default mode network (DMN) is an intrinsic brain networks that most consistently exists in healthy and disease populations; its deterioration acts as a tracking tool to monitor AD progression. 20,21 Aβ accumulation preferentially commences in several core regions of DMN, including the precuneus, medial orbitofrontal, and posterior cingulate cortex, and further affected brain connectivity within DMN. 22 Nevertheless, information on how the LRP1 gene polymorphism affects DMN in the AD spectrum is scant, despite being the regulatory effect of LRP1 on Aβ. Additionally, considering that APOE exacerbates Aβ pathology in an LRP1-dependent manner, 18 investigating the role of APOE on the relationship between LRP1 and cognition in AD spectrum is of importance.
Herein, we first assessed whether the LRP1 genotype disturbed functional connectivity (FC) within DMN and affected cognitive performance across all subjects. Secondly, this work explored the relationships among APOE, LRP1, and cognition in the AD spectrum with moderation analysis. Thirdly, a support vector machine (SVM) model was employed to classify AD spectrum with the altered connectivity within DMN as an objective diagnostic biomarker, based on LRP1 genotypes.

| Participants
Cross-sectional data were downloaded from the Alzheimer's disease Neuroimaging Initiative (ADNI) database (http://adni.loni.usc.edu) before March 21, 2021. Moreover, resting-state functional magnetic resonance imaging (rs-fMRI) images were downloaded from ADNI-1, ADNI-GO, ADNI-2, and ADNI-3 projects. If two or more rs-fMRI scans were performed at baseline, the first available scan was included for analysis (n = 184). Corrupted image sequences (n = 6), image quality control failure (n = 4), and excessive head motion (n = 6) were excluded. In total, 168 participants including 55 CN, 45 subjective cognitive decline (SCD), 42 MCI, and 26 mild AD participants were enrolled for the final analysis. Mini-Mental State Examination (MMSE) was adopted as a measure of general cognition since it was available across all participants. 23 In addition, demographic and genetic information were obtained from the ADNI database. Genetic genotyping for APOE and LRP1 was performed as previously described. 24 Participants with at least one ε4 allele were categorized into APOE-ε4 carriers (APOE ε4 + ), while others without ε4 alleles were categorized into APOE-ε4 non-carriers (APOE ε4 − ). Similarly, those with at least one T allele were categorized into LRP1-T carriers (LRP1 T + ), while others without T allele were categorized into LRP1-T non-carriers (LRP1 T − ). Hardy-Weinberg equilibrium (HWE) test for each gene was calculated. Flow chart was shown in Figure S1.

| Demographic and neuropsychological data analysis
First, the Shapiro-Wilk test was adopted to assess the data normality of continuous variables. A non-parametric test was analyzed when data distributions were not normal. Levene's test was examined to assess the homogeneity of variance. One-way analysis of variance (ANOVA) was separately used to compare the group differences of age, years of education, and MMSE scores. Nonparametric Kruskal-Wallis test was used if the Shapiro-Wilk test or Levene's test p < 0.05. Chi-square tests were applied to compare the group differences of gender, APOE-ε4 status and LRP1-T status. The significant level was set at p < 0.05. Post hoc analyses with Bonferroni correction (p < 0.05/6 = 0.0084) were essential in establishing the significance between any two groups. All statistical analyses were performed using SPSS 22.0 software (SPSS, Inc., Chicago, IL, USA).

| DMN functional connectivity analysis
A voxel-wise one-sample t-test was performed on the subjectspecific maps to achieve a t-map (DMN pattern, FDR corrected, p < 0.001), which was shown in Figure S2. This pattern was converted to a binary map. Of note, voxels outside of gray matter would be excluded. Subsequently, an overlap mask was generated by combining the above binary map and gray matter mask to prevent any spurious effects from white matter and ventricles. After controlling nuisance variables of age, gender, education, and APOE genotype, multivariable linear regression analysis was employed to investigate the effects of LRP1 genotype, disease status, and LRP1 × disease interaction on the DMN within the above overlap mask (3dRegAna, AFNI).
The cluster-level threshold corrected for multiple comparisons was derived using Monte Carlo simulation of the random noise distribution in the data using the latest 3dClustSim program with the -acf function in AFNI [overlap DMN mask correction (39,

| Neuroimaging biomarker of FC for classifying AD spectrum disease
The extracted mean FC from each region of interest was considered a predictive variable to classify the AD spectrum among all subjects.
Further, this regional FC was adopted to classify the AD spectrum in separate LRP1-T carriers and non-carriers. The receiver operating characteristic (ROC) curve was used to calculate the area under the curve (AUC) with the better predictive effect as AUC close to 1.
To verify the accuracy of the classification, a leave-one-out crossvalidation method was used in the linear support vector machine (SVM) model implemented under MATLAB LIBSVM library 26 and repeated 10,000 times permutation tests for the limited sample size.
Unless specifically emphasized, the statistical significance was set at p < 0.05.

| Demographic characteristics
No differences were observed in gender and education among all participants (all p values > 0.05). A significant decreasing trend was noted in age along the disease process, and the main difference existed between the MCI or AD groups and the CN group.
While acknowledging no significance in LRP1-T status, a slight marginal differential trend was noted (p = 0.052). Regarding APOE-ε4 status, a remarkable discrepancy was noted with a tendency that the proportion of ε4 allele increased as the disease progressed.
Furthermore, the MCI and AD groups showed a more significant decrease in MMSE scores; however, the SCD subjects had higher MMSE scores than the CN group. No genotype frequency deviated from HWE (LRP1, χ 2 = 2.455, p = 0.117; APOE, χ 2 = 1.580, p = 0.209). The characteristics of participants are summarized in Table 1.  Figure 1C). In contrast with LRP1-T non-carriers, LRP1-T carriers exhibited a stronger FC strength in these brain areas ( Figure 1D).

| Main and interactive effects of
The interactive effects of LRP1 genotype and disease status on the DMN were also discovered in the LMFG, LPCC, RPCC, LRSC, and right dorsolateral prefrontal cortex (RDLPFC) ( Figure 1E). More importantly, unlike LRP1-T non-carriers, the FC of the LRP1-T carriers displayed opposite trajectory changes in these brain regions across the entire disease process, specifically between the SCD and MCI stages. The LRP1-T carriers showed increased connectivity in the LMFG and decreased connectivity in the LPCC, LRSC, RDLPFC, and RPCC; nevertheless, the LRP1-T non-carriers demonstrated a relatively stable change with the progression of disease ( Figure 1F). The main effects and interaction of LRP1 genotype and disease status were majorly focused at the core hubs of the frontal-parietal network. Comprehensive descriptions of brain regions and their FC differences among different groups are illustrated in Tables S1 and S2.

| Relationship between altered FC and cognitive performance
As shown in Figure 2, the linear regression analyses indicated that disrupted FCs in brain regions of LMFG, RPCC, and LIPC were significantly correlated with the MMSE scores across all groups in the LRP1-T carriers but not in the non-carriers.

| Relationships among LRP1, APOE, and MMSE
The moderation effect analysis revealed that APOE genotype and LRP1 genotype regulated the effects of each other on cognitive performance across all subjects. Figure 3A shows the moderation model of APOE genotype affecting the effect of LRP1 genotype on MMSE scores. As presented in Figure 3B, a relationship between LRP1 genotype and MMSE scores was moderated by Furthermore, the interactive effect analysis disclosed that the effect of LRP1 genotype on MMSE scores was dependent on different levels of APOE genotype. Figure 3E presents the main effect of LRP1 on MMSE was insignificant (F = 0.510, p = 0.476), yet that of APOE was significant (F = 6.877, p = 0.010). As evident in Figure 3F,

| Stage-dependent neuroimaging biomarker of FC for classifying AD spectrum population
As displayed in Figure 4, ROC analysis indicated that FC of numerous brain regions produced a strong power for classifying different disease stages specifically in the LRP1-T carriers but not in non-carriers.
In the brain area of LMFG, FC was detected as the predictive vari-

| DISCUSS ION
This study demonstrates the interactive effect of with non-carriers. This dichotomous pattern suggests that these vulnerable brain regions across the AD spectrum undergo different temporal and spatial patterns of progression. More importantly, we found that FCs correlated with cognitive performance in the

LRP1-T carriers but not in the LRP1-T non-carriers. APOE and LRP1
could regulate the effect of each other on cognitive performance.
Furthermore, we confirmed that the disrupted FCs potentially classify the AD spectrum population and act as a potential neuroimaging biomarker, specifically in the LRP1-T carriers. These findings imply that the LRP1 gene rs1799986 variant polymorphism consistently affects the DMN FC changes across the AD spectrum population and provides a stage-dependent neuroimaging biomarker for early identification of the AD spectrum.
We detected the neural correlates of the LRP1 genotype and disease status on the DMN along the AD spectrum. These brain regions belong to the frontal-parietal network, which governs the cascade of attentional processes underlying the complex cognitive functions. [27][28][29] Besides, they are vulnerable areas of AD progression, as reported formerly. [30][31][32][33] Interestingly, the distribution pattern of FC in the LTPJ, LIPC, LPCC, and LRSC exhibited an inverted-U shape, suggesting that the altered FC strengths of these regions may compensate for cognitive decline in the MCI stage but decompensation occurred in the AD stage. 34 The FC changes in the LPCUN/LCUN, RPCC, and RPCUN/RCUN manifested a U-shape pattern, indicating a disruption in these regions-related networks at the early stage of AD; however, compensation occurred until the AD stage. 35,36 Then, the FC strength of regions affected by the LRP1 genotype disease. This LRP1 genotype-related distribution suggested that the increased FC in LMFG likely occurs in LRP1-T carriers at a higher risk of AD, which is considered compensatory reallocation of cognitive resources since it is associated with better cognitive performance. 42 However, the decreased FCs in LPCC, LRSC, RDLPFC, and RPCC seemingly occur in LRP1-T carriers at a higher risk of AD. This declining trajectory is broadly consistent with disease progression. On the other hand, the FC of LRP1-T non-carriers has a relatively stable change across four groups, indicating that the LRP1-C allele might be a protective factor that prevents functional deterioration.
Moreover, the FCs of LMFG, RPCC, and LIPC correlated with MMSE scores in the LRP1-T carriers but not in the non-carriers. This indicates that the presence of risk T allele might partly influence the connection of the brain network with behavior performance.
The FC of LMFG negatively correlated with MMSE, which might be explained as FC compensation for cognitive decline as above. 42 Besides, we detected positive correlations of the RPCC and LIPC FCs with MMSE. This increased FC is broadly consistent with better cognitive performance at earlier disease stages. These findings highlight the role of the LRP1 genotype in the intrinsic brain function of the AD spectrum population.
As mentioned above, LRP1 is a key receptor of APOE, and APOE mediates Aβ pathology depending on LRP1. 18 This means that their corresponding genes may be inextricably linked to each other.
Previous research also verified that APOE-ε4 is linked to cognitive phenotypes across the AD spectrum. 43 In the present study, two simple moderation analyses identified that the LRP1 genotype and Besides, a strong APOE effect may bridge the connection between the LRP1 genotype and the complex brain cognition as well as provide a novel outlook on the complex mechanism underlying genebehavior interaction.
Several studies have confirmed that brain network variables could be used as predictors in the classification of diseases. [44][45][46] The FC values have been applied in distinguishing AD, MCI from CN subjects. 47 Our group discovered that FCs of numerous brain regions including LMFG, LIPC, and RPCC could classify different disease stages of AD spectrum among all LRP1-T carriers but not in non-carriers. carriers, which might be beneficial for individuals at high risk of AD. 53 Therefore, future work will focus on cerebral vascular issues and design effective intervention strategies to delay the disease progression in preclinical stage of AD. Fourthly, hypotheses such as autophagy 54 and gut dysbiosis 55 represented distinct signaling pathways in AD, and more genes were implicated in Aβ clearance, including APOE, 56 clusterin, 57 α2-Macroglobulin, 58 and triggering receptor expressed on myeloid cells 2. 59 Thus, future research will focus on pathway-based polygenic effects on brain networks that can better translate the underlying mechanism of AD.

| CON CLUS ION
In conclusion, this paper first found the LRP1 gene rs1799986 variant polymorphism could consistently affect DMN patterns across the AD spectrum population. APOE regulated the effect of LRP1 on cognitive performance. The disrupted FCs might be used as a stage-dependent neuroimaging biomarker in classifying the AD spectrum. These findings provide novel insights into the potential mechanism underlying cognitive impairment in AD spectrum progression, and the identification of intervention targets based on genetic risk variants may offer a sneak peek at future research direction.

CO N FLI C T O F I NTE R E S T
No authors have any possible conflict of interest.