Alzheimer's Genes Aren't the Same for Everyone: Study Finds 133 New Risk Variants

When it comes to dementia, genetics has long looked almost exclusively at Europe. A new paper in Nature Communications breaks this optics: scientists have compiled the largest multinational catalog of variants associated with Alzheimer's disease and related dementias (AD/ADRD) to date, and have shown that the effects of key genes are highly dependent on origin. Most importantly, how does the famous APOE ε4 "behave" and what other variants can enhance or, conversely, mitigate its impact.

Background of the study

Alzheimer's disease and related dementias (AD/ADRD) have a strong genetic component: from rare highly penetrant variants in the APP, PSEN1/2 genes (familial forms) to dozens of "common" alleles of low and medium strength, forming a polygenic risk. Against this background, one gene - APOE - remains a "heavy anchor": the ε4 allele significantly increases the probability of the disease and shifts the age of onset, while ε2 more often protects. But the magnitude of the effect is not universal: it depends on the genetic origin, the linkage structure around APOE and neighboring modifiers.

Historically, the vast majority of genetic studies of dementia have been conducted in samples of European descent. This “Eurocentrism” reduces the transferability of results: markers, panels, and polygenic indices work worse in people of African, Latin American, South Asian, and other ancestry; rare variants simply do not come to the attention of the eye because they are few or absent in Europeans. As a result, clinicians receive “biased” lists of risk alleles, and patients receive less accurate estimates of individual risk and weaker prerequisites for targeted prevention.

Multinational, biobank-scale projects are designed to fill this gap. They allow simultaneously: (1) to refine the map of rare and splicing variants in the "core" AD/ADRD genes ( TREM2, MAPT, GRN, GBA1, SNCA, TBK1, TARDBP, etc.); (2) to search for risk modifiers in APOE ε4 carriers (alleli in TOMM40 and adjacent regions, as well as loci outside chromosome 19); (3) to re-evaluate the "pathogenicity" of variants taking into account frequencies and effects in different populations. This provides more honest genetic panels, improves the portability of polygenic scoring, and opens a window to the search for "resistant" alleles - those that mitigate the vulnerability of ε4.

The clinical context is clear: the more accurately we understand the population-specific architecture of risk, the better we can design screening, stratify patients for testing, and target preventive interventions. For science, this is a step away from “average European genetics” to an individualized ancestral risk picture, where the same phenotype is made up of different genetic combinations – and therefore requires different diagnostic and therapeutic solutions.

What did they do?

• We combined 5 biobanks (All of Us, ADSP, UK Biobank, 100K Genomes, AMP PD).
• We analyzed 25,001 cases of dementia and 93,542 controls from 11 genetic ancestors (European, African, Latin American admixtures, Ashkenazi, etc.).
• We scanned 11 “core” AD/ADRD genes: APP, PSEN1, PSEN2, TREM2, MAPT, GRN, GBA1, SNCA, TBK1, TARDBP, APOE.

The study did more than just “put together” the databases. The team specifically looked for rare and splicing variants, checked their pathogenicity using ClinVar/ACMG/CADD, calculated polygenic risk in the best-powered sample (ADSP), and — most importantly — looked at risk modifiers in APOE ε4 carriers in different populations. The result is a working map for future targeted therapies and fair, inclusive clinical trials.

Main findings

• 156 variants were identified, 133 of which were new. This is the largest "replenishment" of the AD/ADRD panel at one time.
• 26 potentially causal variants were found in non-European groups, with 18 completely absent from Europeans - another argument why we cannot limit ourselves to one population.
• APOE does indeed "play differently": for example, rs449647-T increased risk in ε4 carriers of African descent but decreased it in Europeans; TOMM40:rs11556505-T was associated with greater risk in ε4 carriers, especially in Europe.
• Potential risk mitigators have been identified in ε4 carriers: NOCT:rs13116075-G, CASS4:rs6024870-A, LRRC37A:rs2732703-G - candidates for ancestry-dependent protective or modifying effects.
• The controls contained 23 variants that were previously considered “pathogenic” - a reminder that annotations need to be rethought taking into account the origin and large databases.

To understand the scale and “texture” of the findings, the authors provide examples of “migrants” between diagnoses: PSEN1 p.R269H was found not only in early Alzheimer’s, but also in late-onset Alzheimer’s, and TARDBP p.G287S, known from ALS, was first seen in early dementia – such crossroads help explain mixed phenotypes in patients.

Why this is important now

• More precisely, targets: different ancestors - different risk combinations. Therapies and preventive panels should take this into account.
• Fair trials: To ensure drugs work “for everyone,” RCTs need multiethnic cohorts and stratification by APOE modifiers.
• Correct genetic counseling: the “pathogenic in some, neutral in others” option ceases to be a paradox and becomes the norm of clinical genomics.

How exactly was it researched?

• WGS with short reads (NovaSeq; aligned to GRCh38), screening for missense/frameshift/stop variants and splicing, followed by filtering for CADD>20 and “cases only” frequency.
• In the UK Biobank, 815 variants of target genes were found at the discovery stage; verification was carried out in ADSP and 100KGP.
• The overlap of phenotypes (AD, DLB, FTD, etc.) was assessed - hence the stories about GRN, MAPT, TBK1, GBA1.

What does this change for practice and science?

• Diagnostic panels must “live” and be localized: the same “family” of genes, but different priorities by ancestors.
• Biobanks ≠ "data dump": the authors opened an online browser (MAMBARD) with ancestry frequencies/associations - a tool for rapid verification of rare findings by clinicians and researchers.
• New prevention hypotheses: the search for “resistant” variants (which delay the onset of the disease in ε4 carriers) is the path to genetically motivated intervention strategies.

Nuances and limitations

• This is a genetic association map, not functional biology: experiments are needed on "newbies".
• Not all markers (for example, C9ORF72 expansions) are caught by WGS short reads - some of the "complex" variants remain "behind the scenes".
• Standardization of phenotypes across biobanks and quality of annotations are a perennial challenge, but scale and replication across multiple databases make inferences more robust.

Summary

The work does not simply expand the list of genetic "suspects" in dementia - it teaches us to read genetics in the context of ancestors. For the clinic, this means more accurate selection of tests and targets, for science - to build inclusive RCTs and look for risk modifiers that can "cover up" the vulnerability of APOE ε4.

Source: Khani M., Akçimen F., Grant SM, et al. Biobank-scale genetic characterization of Alzheimer's disease and related dementias across diverse ancestries. Nature Communications (2025) 16:7554. DOI: 10.1038/s41467-025-62108-y