Integrating rare pathogenic variant prioritization with gene-based association analysis to identify novel genes and relevant multimodal traits for Alzheimer's disease

Abstract

Introduction: Increasing evidence has highlighted rare variants in Alzheimer's disease (AD). However, insufficient sample sizes, especially in underrepresented ethnic groups, hinder their investigation. Additionally, their impact on endophenotypes remains largely unexplored.

Methods: We prioritized rare likely-deleterious variants based on whole-genome sequencing data from a Chinese AD cohort (n = 988). Gene-based optimal sequence kernel association tests were conducted between AD cases and normal controls to identify AD-related genes. Network clustering, endophenotype association, and cellular experiments were conducted to evaluate their functional consequences.

Results: We identified 11 novel AD candidate genes, which captured AD-related pathways and enhanced AD risk prediction performance. Key genes (RABEP1, VIPR1, RPL3L, and CABIN1) were linked to cognitive decline and brain atrophy. Experiments showed RABEP1 p.R845W inducing endocytosis dysregulation and exacerbating toxic amyloid β accumulation, underscoring its therapeutic potential.

Discussion: Our findings highlighted the contributions of rare variants to AD and provided novel insights into AD therapeutics.

Highlights: Identified 11 novel AD candidate genes in a Chinese AD cohort. Correlated candidate genes with AD-related cognitive and brain imaging traits. Indicated RABEP1 p.R845W as a critical AD contributor in the endocytic pathway.

Keywords: Alzheimer's disease; biomarkers; rare variant; the endocytic pathway; whole‐genome sequencing.

FIGURE 1

The overall study design. WGS analysis was conducted on a Chinese cohort of 988 individuals. European participants from the ADNI and the ADSP cohorts were included as replication samples. Comprehensive variant prioritization strategies were employed to identify rare likely‐deleterious protein‐coding variants. First, as a proof of concept, an initial screening evaluated the rare LoF variants in 25 known AD core genes. Next, cross‐cohort SKAT‐O between the AD and NC groups identified 11 candidate genes. The associations with AD were further supported by AD classification models, biological network analyses, and endophenotype analyses. Finally, cellular experiments validated the functional impact of a top locus in AD. AD, Alzheimer's disease; ADNI, Alzheimer's Disease Neuroimaging Initiative; ADSP, Alzheimer's Disease Sequencing Project; CI, cognitive impairment; DMis, predicted deleterious missense; LoF, predicted loss‐of‐function; MAF, minor allele frequency; NC, normal control; SKAT‐O, optimal sequence kernel association test; WGS, whole‐genome sequencing.

FIGURE 2

Rare likely‐deleterious variants in the AD core and novel genes. (A) The burden of rare LoF variants in the AD core genes in case groups compared to NC. Midpoints and error bars represent the OR and their 95% confidence intervals. The dashed vertical red line indicates an OR of 1. (B) The distribution of different types of LoF variants exclusive to AD cases across AD core genes. Each bubble's position reflects the presence of a specific mutation type in a gene. The size of each bubble corresponds to the number of mutations observed, with the numerical values inside providing the exact count of mutations for each gene‐mutation type combination. (C,D) Example of the PSEN1 gene splice‐site LoF variant c.869‐2A > G. (C) This variant alters the basic region at the 3′ boundary of intron 8, resulting in the skipping of exon 9 and introducing an aberrant exon 8–10 junction. (D) The carrier of this variant was only 50 years old but suffered from unusually severe hippocampal atrophy. The box spans the first to third quartiles; the whiskers extend 1.5 times the interquartile range; and the middle line represents the median. P values between the AD and NC groups were calculated using Mann‐Whitney U tests. The red dots represent this specific carrier. (E) The MetaSKAT‐O results for the 11 novel candidate genes. Each bar represents the negative logarithm MetaSKAT‐O FDR for each gene, with the MACs annotated on top. The color of the bar indicates the direction of each gene's effect. The eight genes that remained significant in the three‐cohort meta‐analysis are highlighted by gray diagonal stripes. The red dashed horizontal line indicates an FDR of 0.05. (F) Performance of random forest models used to distinguish AD cases from NC. Models with different types of variant sets as features are represented by differently colored receiver operating characteristic curves. ***p < 0.001. AD, Alzheimer's disease; AUC, area under the receiver operating characteristic curve; FDR, false discovery rate; LoF, loss‐of‐function; MACs, minor allele counts; Mut, mutation carrier; NC, normal controls; OR, odds ratio; SKAT‐O, sequence kernel association tests.

FIGURE 3

Validation of the AD candidate genes via network and endophenotype analysis. (A) The density plot shows the average distances from AD core genes for all genes, tested across 10,000 permutations; the vertical dashed line denotes the average distances between the 11 AD candidate genes and the AD core genes. (B) Hierarchical clustering (left) of AD candidate genes (red) and core genes (blue) according to HGC. Different background colors represent gene modules divided based on hierarchical clustering. Heatmaps illustrate the negative logarithm of p values for the gene‐based association tests of MMSE scores, subcortical volumes, and differential gene expression across multiple brain regions. Only tests with unadjusted p values less than 0.05 are colored and indicated by varying degrees of transparency. Asterisks indicate the significance levels of associations after controlling for FDR in candidate genes. Borderless blocks signify either missing data or tests that do not meet the SKAT‐O test criteria. (C) The top significantly enriched biological process terms for each gene module, displaying up to the top 10. ***FDR < 0.001; ***FDR < 0.01; *, FDR < 0.05. AD, Alzheimer's disease; CBE, cerebellum; DLPFC, dorsolateral prefrontal cortex; FDR, false discovery rate; FP, frontal pole; HGC, human gene connectome; IFG, inferior frontal gyrus; MMSE, Mini‐Mental State Examination; PHG, parahippocampal gyrus; STG, superior temporal gyrus; SKAT‐O, sequence kernel association tests; TCX, temporal cortex.

FIGURE 4

Functional validation of the RABEP1 p.R845W variant. (A) The proliferative capacity of cells measured by the CCK8 assay. (B,C) Images (B) and quantification (C) of the apoptosis assay by flow cytometry. (D,E) Western blot analysis (D) and quantification (E) in RABEP1 ^R845W/+ cells and wild‐type HEK293T cells. (F,G) Immunofluorescence (F) and quantification (G) of EEA1‐positive early endosomes. Representative immunofluorescence images were stained with EEA1 (red) and nuclei (blue). The diameters of EEA1‐positive endosomes were quantified by fluorescence intensity. Scale bar, 50 µm. (H,I) Concentration of Aβ42 and Aβ40 (H), and the ratio of Aβ42 to Aβ40 (I) detected by ELISA. P values were calculated using unpaired two‐sample t‐tests if the data conformed to a normal distribution; otherwise, the non‐parametric Mann–Whitney U test was used. *p < 0.05; **p < 0.01; ns, not significant; CCK8, Cell Counting Kit‐8; WT, wild‐type HEK293T cells; R845W, HEK293T cells carrying RABEP1 ^R845W/+.