Transethnic analysis identifies SORL1 variants and haplotypes protective against Alzheimer's disease

Abstract

Introduction: The SORL1 locus exhibits protective effects against Alzheimer's disease (AD) across ancestries, yet systematic studies in diverse populations are sparse.

Methods: Logistic regression identified AD-associated SORL1 haplotypes in East Asian (N = 5249) and European (N = 8588) populations. Association analysis between SORL1 haplotypes and AD-associated traits or plasma biomarkers was conducted. The effects of non-synonymous mutations were assessed in cell-based systems.

Results: Protective SORL1 variants/haplotypes were identified in the East Asian and European populations. Haplotype Hap_A showed a strong protective effect against AD in East Asians, linked to less severe AD phenotypes, higher SORL1 transcript levels, and plasma proteomic changes. A missense variant within Hap_A, rs2282647-C allele, was linked to a lower risk of AD and decreased expression of a truncated SORL1 protein isoform.

Discussion: Our transethnic analysis revealed key SORL1 haplotypes that exert protective effects against AD, suggesting mechanisms of the protective role of SORL1 in AD.

Highlights: We examined the AD-protective mechanisms of SORL1 in the general population across diverse ancestral backgrounds by jointly analyzing data from three East Asian cohorts (ie, mainland China, Hong Kong, and Japan) and a European cohort. Comparative analysis unveiled key ethnic-specific SORL1 genetic variants and haplotypes. Among these, the SORL1 minor haplotype, Hap_A, emerged as the primary AD-protective factor in East Asians. Hap_A exerts significant AD-protective effects in both APOE ε4 carriers and non-carriers. SORL1 haplotype Hap_A is associated with cognitive function, brain volume, and the activity of specific neuronal and immune-related pathways closely connected to AD risk. Protective variants within Hap_A are linked to increased SORL1 expression in human tissues. We identified an isoform-specific missense variant in Hap_A that modifies the function and levels of a truncated SORL1 protein isoform that is poorly investigated.

Keywords: APOE; East Asian; European; PET; Pittsburgh compound B; amyloid load; association; plasma biomarker; protective.

FIGURE 1

Comparative analysis of SORL1 genetic variants associated with Alzheimer's disease (AD) in the East Asian and European populations. (A) Regional association plot of meta‐analysis results for variants in SORL1 locus in East Asian population (N = 5249). The sentinel variant, rs11604897, is highlighted. (B) Comparison of effects of variants on modifying AD risks between East Asian (EAS) and European (EUR) populations. The p values for AD associations in European population are from Genome‐Wide Association Study summary statistics from Kunkle et al. Blue and red dashed lines represent Bonferroni‐corrected p value cutoffs of .05 in EUR population and meta‐p value cutoff of 1×10⁻⁴, respectively.

FIGURE 2

Identification of distinct variant clusters in SORL1 locus associated with AD in East Asian and European populations. (A) Linkage disequilibrium plots of candidate variants in East Asian population (upper panel) and European population (lower panel). Data are from high‐coverage, whole‐genome sequencing data from the 1000 Genomes Project Phase 3. Colors denote the pairwise R ² values among the selected variants. Middle panel: Attributes of variants from clusters EA_1 to EA_4 in East Asian population and clusters EU_1 to EU_6 in European population, specifically their associations with AD. Cluster EA_4 is subdivided into EA_4a and EA_4b owing to linkage disequilibrium patterns in corresponding variants. (B) Meta‐analysis results of effects of minor haplotypes with the lowest p values from each variant cluster on AD (N = 5249). Meta‐p values after Bonferroni correction calculated from Han and Eskin's random effects model are shown. (C) Meta‐analysis results of effects of minor haplotypes, with lowest p values from each variant cluster and five variants on AD association (*p < .05, Han and Eskin's random effects model). (B, C) Data are mean ± 95% confidence intervals. Adj, adjusted; Both, variants significantly associated with Alzheimer's disease in both East Asian and European populations; Chr, chromosome; EAS, East Asian population; EUR, European population; kb, kilobase.

FIGURE 3

Identification of AD‐protective SORL1 haplotype Hap_A. (A) Linkage disequilibrium plots in East Asian population (upper panel) and European population (lower panel) for 31 variants from variant clusters EA_3 and EU_5, four variants from EU_1 (including rs11218343), and an additional three variants (ie, rs9665907, rs3781832, and rs1784920) in linkage disequilibrium with rs11218343. Data are high‐coverage, whole‐genome sequencing data from the 1000 Genomes Project. Color denotes pairwise R ² values among the 31 selected variants. (B) Common haplotypes (ie, frequency > 1%) defined by the selected 31 variants identified in East Asian and European populations (ie, “Major” and Hap_A to Hap_G); the “Major” haplotype denotes the most common haplotype present in the general population. Lowercase letters denote the AD‐protective alleles of corresponding variants. Purple letters denote variants associated with AD in both the East Asian and European populations. Boldface letters denote the rs2282647 variant, which is present in both variant clusters EA_3 and EU_5. (C) Heatmap of association results of identified haplotypes in individual cohorts as well as meta‐analysis results in different ethnic groups and across all cohorts. Colors indicate association Z‐scores calculated as effect size divided by standard error. For individual cohorts, nominal p values obtained from logistic regression are displayed; for meta‐analysis results, p values from Han and Eskin's random effects model are displayed (***p < .001, **p < .01, *p < .05; p values from .05 to .1 are displayed as digits). AD, Alzheimer's disease; Both, variants exhibiting significant associations with AD in both East Asian and European populations; EAS, East Asian population; EUR, European population; kb, kilobase; LD, linkage disequilibrium.

FIGURE 4

Effects of SORL1 AD‐protective haplotype on AD endophenotypes and biological processes. (A) Heatmap summarizing associations between AD‐protective haplotype (Hap_A) and AD‐associated endophenotypes, including cognitive function (MMSE: n = 1392 from mainland Chinese cohort, including 1082 individuals with AD and 310 NCs; MoCA: n = 1210 from Hong Kong population, including 429 individuals with AD and 781 NCs), volumes of specific brain regions (n = 216, including 110 individuals with AD and 106 NCs), and levels of plasma ATN biomarkers (n = 377, including 184 individuals with AD and 193 NCs). The color scale corresponds to t values obtained from the association analysis (robust regression, *p < .05; p values from .05 to .1 are displayed as digits). (B) Association between MoCA score and Aβ brain load detected by PET in APOE ε3 homozygous individuals harboring haplotype Hap_A (n = 32) or without haplotype Hap_A (n = 16) (robust regression, ***p < .001; Z‐score test, ^# p < .05). (C) Volcano plot showing association between plasma proteomes and allele dosage of haplotype Hap_A in NCs (n = 110). Proteins with a nominal p value less than .05 (robust regression) are marked in red (positive association) or blue (negative association). Key genes that exhibit strong associations or significant level changes are marked in the plot. (D) Bar charts displaying biological processes in plasma proteome that are associated with haplotype Hap_A. Red and blue denote biological processes that are positively or negatively associated with haplotype Hap_A, respectively. Aβ, beta amyloid; AD, Alzheimer's disease; ATN, amyloid, tau, and neurodegeneration; C‐PiB, carbon‐11 – labeled Pittsburgh compound B; MMSE, Mini‐Mental State Examination; MoCA, Montreal Cognitive Assessment; NC, normal control; NfL, neurofilament light chain; p‐tau181, phosphorylated tau protein at threonine 181; PET, positron emission tomography.

FIGURE 5

Effects of SORL1 AD‐protective and risk alleles on modulation of SORL1 transcript expression. (A) Visualization of SORL1 AD risk and AD‐protective variants that sit in candidate cis‐regulatory regions and brain cell active chromatin regions. The AD risk and protective variants are denoted in red and blue, respectively. The track visualizes the SORL1 isoforms, location of AD‐protective and risk variants, and signals of brain single‐cell ATAC‐seq data retrieved from Corces et al. (B) Association of AD‐protective variant, rs75279208, with elevated SORL1 transcript levels across different tissues in GTEx dataset. (C) SORL1 transcript levels in nerve tissue among rs75279208 carriers (n = 23) and non‐carriers (n = 509) (robust linear regression, ***p < .001). (D) Association of AD risk variant, rs1792125, with reduced SORL1 transcript levels across different tissues in GTEx dataset. (E) SORL1 transcript levels in esophageal tissue stratified by rs1792125 genotype (n = 105, 225, and 135 for homozygous non‐carriers, heterozygous carriers, and homozygous carriers, respectively; robust linear regression, **p < .01, ***p < .001). AD, Alzheimer's disease; ATAC‐seq, assay for transposase accessibility by sequencing; eQTL, expression quantitative trait loci; GTEx, Genotype‐Tissue Expression Project; NES, normalized effect size.

FIGURE 6

Impact of isoform‐specific Trp15Cys coding variant on modulation of truncated SORL1 protein levels. (A) Western blots depicting expression levels of WT and mutated (W15C) HA‐tagged truncated SORL1 proteins transfected into HEK293T cells with 1‐, 1.5‐, or 2‐µg plasmids. The truncated SORL1 protein was detected using HA antibody, with α‐tubulin concurrently blotted as loading control. (B) Dot plot illustrating relative expression levels between WT and W15C HA‐tagged truncated SORL1 proteins. The expression of mutated SORL1 protein is normalized to both α‐tubulin and WT SORL1 protein. A mixed‐effects model (ie, REML) was employed to examine the impacts of both batch effects and mutation on the modulation of the truncated SORL1 protein. **p < .01 indicates a significant effect of the Trp15Cys mutation on the modulation of protein level. SORL1‐HA, HA‐tagged truncated SORL1 protein; W15C, mutated truncated SORL1 protein with Trp15Cys mutation; WT, wild type.