Cognitively healthy centenarians are genetically protected against Alzheimer's disease

Abstract

Background: Alzheimer's disease (AD) prevalence increases with age, yet a small fraction of the population reaches ages > 100 years without cognitive decline. We studied the genetic factors associated with such resilience against AD.

Methods: Genome-wide association studies identified 86 single nucleotide polymorphisms (SNPs) associated with AD risk. We estimated SNP frequency in 2281 AD cases, 3165 age-matched controls, and 346 cognitively healthy centenarians. We calculated a polygenic risk score (PRS) for each individual and investigated the functional properties of SNPs enriched/depleted in centenarians.

Results: Cognitively healthy centenarians were enriched with the protective alleles of the SNPs associated with AD risk. The protective effect concentrated on the alleles in/near ANKH, GRN, TMEM106B, SORT1, PLCG2, RIN3, and APOE genes. This translated to >5-fold lower PRS in centenarians compared to AD cases (P = 7.69 × 10^-71), and 2-fold lower compared to age-matched controls (P = 5.83 × 10^-17).

Discussion: Maintaining cognitive health until extreme ages requires complex genetic protection against AD, which concentrates on the genes associated with the endolysosomal and immune systems.

Highlights: Cognitively healthy cent enarians are enriched with the protective alleles of genetic variants associated with Alzheimer's disease (AD). The protective effect is concentrated on variants involved in the immune and endolysosomal systems. Combining variants into a polygenic risk score (PRS) translated to > 5-fold lower PRS in centenarians compared to AD cases, and ≈ 2-fold lower compared to middle-aged healthy controls.

Keywords: Aging; Alzheimer's disease; Cognitively healthy centenarians; Endolysosomal; Genes; Genome‐wide association studies; Heritability; Immunity; Protection; Resilience.

FIGURE 1

Single variant associations summary. A (top), Raw minor allele frequency in AD patients (red circles), age‐matched controls (blue squares), and cognitively healthy centenarians (green triangles). Black and white annotation squares refer to whether the plotted allele (the minor allele) was associated with an increased risk of AD (risk allele, white) or a decreased risk of AD (protective allele, black). B (bottom), Change in effect size comparing observed effect sizes (AD cases vs. cognitively healthy centenarians) to the reference effect sizes (Bellenguez et al. ¹⁰ ). Blue genes refer to novel SNP–AD associations discovered by Bellenguez et al. for the first time, while red genes were known before Bellenguez et al. The dashed red line at 1 indicates the published effect size from the literature. *: P value of association < 0.05; **: FDR‐corrected P value of association < 0.05; pink bars indicate SNPs for which observed effect size is significantly different from published effect size. AD, Alzheimer's disease; FDR, false discovery rate; SNP, single nucleotide polymorphism.

FIGURE 2

Summary of PRS analyses. A (top left), Distribution of the PRS including the two APOE SNPs (86 SNPs in total) in AD cases (red), age‐matched controls (blue), and cognitively healthy centenarians (green). B (top right), Distribution of the PRS excluding the two APOE SNPs (84 SNPs in total). C (bottom), Association statistics (OR, 95% CI, and corrected P value) and forest plot of the PRS including and excluding APOE SNPs. For the comparisons, we used logistic regression models in a pairwise manner (i.e., AD cases vs. cognitively healthy centenarians, AD cases vs. age‐matched controls, and age‐matched controls vs. cognitively healthy centenarians), controlling for population substructure. AD, Alzheimer's disease; APOE, apolipoprotein E; CI, confidence interval; OR, odds ratio; PRS, polygenic risk score; SNP, single nucleotide polymorphism.

FIGURE 3

Relationship between cognitively healthy centenarians and age‐matched controls. A (left), Number of individuals (age‐matched controls on the y axis and centenarians on x axis) necessary to achieve 80% power for a SNP association at P = 0.05, assuming 8000 AD cases. We restricted this analysis to common variants (MAF > 1%) with expected direction of effect in our comparisons (N = 67 SNPs, see Methods). Note that, for this reason, some variants enriched in cognitively healthy centenarians such as rs13237518 (TMEM106B) and rs13237518 (SORT1) could not be represented here. Each dot represents a SNP: dark green dots identify the eight SNPs that did not converge using both age‐matched controls and centenarians (i.e., the power did not reach 80%). Light green dots indicate the two APOE SNPs, for which N = 200 individuals (age‐matched controls and centenarians) were enough to guarantee 80% power. Light blue dots identify SNPs for which the number of cognitively healthy centenarians (to achieve 80% power) was lower than the number of age‐matched controls. Of these, N = 13 SNPs did not converge using age‐matched controls. Conversely, dark blue dots identify SNPs for which the number of age‐matched controls was lower than the number of cognitively healthy centenarians. Of these, N = 8 SNPs did not converge using cognitively healthy centenarians. B (right), Ratio between the number of age‐matched controls and the number of cognitively healthy centenarians, for each SNP. Color code is the same as (A). SNPs larger than the blue dotted line (N = 31, ratio > 2) were used for functional annotation and gene‐set enrichment analysis. AD, Alzheimer's disease; APOE, apolipoprotein E; GWAS, genome‐wide association study; MAF, minor allele frequency; SNP, single nucleotide polymorphism.

FIGURE 4

Functional annotation of SNPs with the largest effect in cognitively healthy centenarians. The figure shows the result of the functional annotation of 31 SNPs for which the number of cognitively healthy centenarians required to achieve 80% power was at least half of the number of age‐matched controls required to achieve the same power. Functional annotation analysis was performed using snpXplorer. A, Result of the gene‐set enrichment analysis followed by REVIGO analysis, which clusters enriched pathways based on a semantic similarity measure. B, Dendrogram of the main enriched pathways along with their cluster (branches color code for cluster assignment) and word clouds showing the main terms enriched in the underlying pathways. C, Mapping between significant pathways (x axis), AD‐associated SNPs (y axis, labeled with the name of the gene as provided by Bellenguez et al. ¹⁰ ), and the relative gene‐set enrichment cluster. AD, Alzheimer's disease; SNP, single nucleotide polymorphism.