APOE-Stratified GWAS Uncovers New Alzheimer's Risk Loci and Supports A Priori APOE Stratification in AD Trials
核心洞察
A large APOE-stratified genome-wide association meta-analysis identified 25 loci reaching genome-wide significance, including nine new genomic signals for Alzheimer's disease (搜索) risk that appear only within specific APOE strata.
Nine loci were exclusive to APOE ε4 carriers, including new signals HP1BP3 (搜索), PTPRC (搜索) and DDHD1 (搜索), while seven loci emerged only in the ε33 stratum, including new signals SLC50A1 (搜索), NPAS3 (搜索) and CHST9 (搜索).
Pathway analysis showed complement and immune pathways over-represented in ε4 carriers, whereas amyloid and neurofibrillary tangle biology was highlighted in both the ε33 and ε44+ε43 strata.
A large genome-wide association study (GWAS) meta-analysis stratified by apolipoprotein E (APOE) genotype has identified new genomic signals for Alzheimer's disease (搜索) (AD) risk that manifest only in specific APOE contexts, findings the authors say may inform how genetics is used in designing randomized clinical trials of future AD medicines.
The study, published in Nature Genetics, brought together European, Asian, Asian American (AAC), African American (AFR) and admixed American (AMR) ancestry cohorts with clinically diagnosed AD. Individuals were grouped into ε22 + ε32, ε33 and ε44 + ε43 strata to maximize statistical power, and individuals with the ε42 genotype were excluded.
Scale of the Stratified Analysis
The ε22 + ε32 stratum meta-analysis was based on 2,606 AD cases, 70,388 controls and 10,336,311 variants. The ε33 stratum included 24,033 AD cases, 363,161 controls and 17,127,662 variants, while the ε44 + ε43 stratum comprised 29,122 AD cases, 164,206 controls and 14,672,059 variants. A separate meta-analysis restricted to ε44 carriers included 5,814 AD cases, 14,415 controls and 9,723,486 variants.
No signal reached the genome-wide significance threshold of 2.5 × 10⁻⁸ in the ε22 + ε32 stratum. In total, 25 loci reached genome-wide significance in the ε33 or ε44 + ε43 strata, or in both.
Nine loci were found in both strata and are known AD risk loci: CR1, BIN1 (搜索), HLA, TREM2 (搜索), CLU (搜索), MS4A64, PICALM, APH1B and ABCA7. Nine loci were observed exclusively in the ε44 + ε43 stratum, including six known AD loci (PILRA, SORL1, ADAM10, ACE, LILRA5, CASS4) and three new loci (HP1BP3 (搜索), PTPRC (搜索), DDHD1 (搜索)). Among seven loci reaching genome-wide significance only in the ε33 stratum, four are known AD risk loci (TMEM106B (搜索), SHARPIN (搜索), GRN (搜索), MAPT (搜索)) and three are new (SLC50A1 (搜索), NPAS3 (搜索), CHST9 (搜索)).
In the ε44-restricted meta-analysis, only one well-established locus, BIN1 (搜索), was observed.
Interaction Testing and Effect Modification
To formally test for single nucleotide polymorphism (SNP) × APOE interactions, the researchers used two complementary approaches: stratified effect comparisons using additive and dominant models, and a genome-wide SNP × APOE ε4 carrier interaction GWAS with a formal interaction term.
In the dominant stratified effect comparison, seven significant differences emerged. Three signals showed attenuated effect sizes in the presence of an ε4 allele (SLC50A1 (搜索), TMEM106B (搜索), NPAS3 (搜索)), while four showed augmented effects (BIN1 (搜索), HLA-DRA-1, CLU (搜索), DDHD1 (搜索)). In the additive mixed-effect model, SHARPIN (搜索) was additionally identified as interacting with APOE ε4, with effect size attenuated as the number of ε4 alleles increased.
In the genome-wide interaction GWAS, two new loci, SMYD2 (搜索) and ANK3 (搜索), reached genome-wide significance in the SNP term. The APOE ε4 carrier term was highly significant (P < 1 × 10⁻³⁰⁰) and SNP independent. In the interaction term, only the APOE locus was significant, although the DDHD1 (搜索) locus reached an interaction P value of 1.62 × 10⁻⁶.
Biological Pathways Differ by APOE Stratum
Pathway enrichment analysis on the APOE-stratified GWAS results from eight European ancestry studies found 12 and 33 pathways reaching statistical significance (q < 0.05) for the APOE ε33 and ε44 + ε43 strata, respectively. Complement and immune system pathways were over-represented in the ε4 carrier stratum compared with ε33, whereas amyloid and neurofibrillary tangle biology was highlighted in both strata. No pathway analysis reached significance for the ε22 + ε32 or ε43 strata.
Summary-data-based Mendelian randomization identified four genes passing the significance threshold and HEIDI filtering: STAG3 (PILRA locus) in the ε44 + ε43 stratum in cortex, and LRRC37A, ARL17B and LRRC37A2 (MAPT (搜索) locus) in the ε33 stratum across multiple brain regions.
DDHD1 Emerges as a Drug Discovery Focus
The DDHD1 (搜索) signal drew particular attention. The lead variant rs10131116 was associated with decreased risk of AD specifically in APOE ε4 carriers, and the variant has been identified as an expression quantitative trait locus (eQTL) associated with decreased DDHD1 expression in the ROSMAP dorsolateral prefrontal cortex (β = −0.098; eQTL P = 1.67 × 10⁻⁵, n = 560) and in the GTEx v10 brain putamen (β = −0.23; eQTL P = 3.6 × 10⁻⁵, n = 253).
DDHD1 (搜索) encodes a member of the phospholipase A1 family important in lipid and phospholipid metabolism and is associated with the rare neuronal disease spastic paraplegia 28 (搜索). The gene lies in the same biological pathway as the genome-wide significant known signal PLCG2 (搜索). The authors note that the combination of decreased AD risk in ε4 carriers and the eQTL association with decreased DDHD1 expression "highlights DDHD1 as an interesting focus for drug discovery."
The DDHD1 (搜索) signal is close to the FERMT2 (搜索) locus, a recognized AD genetic risk factor, but conditional testing showed the DDHD1 signal is independent of FERMT2.
Cross-Ancestry Consistency
The researchers evaluated 29 regions of interest in additional cohorts of East Asian ancestry, representing Japanese (JADNI, CL, NP), Chinese (HKS) and Korean (GARD) populations, as well as AAC, AFR and AMR multi-ancestry populations. Despite limitations including differences in linkage structure across ancestries, limited statistical power and lead variants differing from causal variants, similar signals were observed for several variants. Meta-analyzed results for HLA-DRA-1 and DDHD1 (搜索) were similar in East Asian and European ancestry cohorts.
Implications for Trial Design
The authors state that their findings suggest differential biological pathways dependent on APOE strata and "may inform how we use genetics in designing randomized clinical trials of future AD medicines." They point to the observed lower efficacy and higher risks of severe side effects in ε4 carriers in recent trials of anti-amyloid therapies, concluding that "it may now be timely to consider including APOE stratification a priori in future randomized clinical trials of emerging AD medicines."
The study also acknowledges that the meta-analyses do not allow full determination of whether there is a significant difference between signals observed in the ε33 and ε44 + ε43 strata due to sample and statistical power variations. The authors note that the effective number of independent tests was estimated at 1.3–1.4 for |z| > 5, consistent with one shared genetic dimension and a partial second dimension corresponding to the ε4-interacting minority of loci; a conservative threshold of Meff = 2 was applied.