Skip to content

Genetics

TL;DR — Alzheimer's disease is highly heritable — 58% in the full model and 79% in the best-fitting model in 11,884 Swedish twin pairs — but the architecture is layered rather than Mendelian for almost everyone (Gatz 2006, PMID 16461860). APOE remains the dominant common variant: in the founding meta-analysis of 5,930 cases and 8,607 controls, ε4/ε4 carried OR 14.9 (95% CI 10.8–20.6) and ε3/ε4 OR 3.2 (2.8–3.8) versus ε3/ε3 in white participants, with the effect stronger in Japanese and weaker in African American and Hispanic participants (Farrer 1997, PMID 9343467), a pattern re-quantified across 68,756 individuals three decades later (Belloy 2023, PMID 37930705). APOE4 homozygosity now has a case for being reclassified as a genetic form of AD rather than a risk factor: by age 65 nearly all APOE4 homozygotes had abnormal CSF amyloid and 75% a positive amyloid scan, with symptom onset at 65.1 years and a prediction interval as narrow as autosomal dominant disease (Fortea 2024, PMID 38710950). Common-variant GWAS has grown from a handful of loci to 75 (Bellenguez 2022, PMID 35379992) and then 91 (EADB 2026, PMID 42237039), converging on amyloid processing, tau, microglia/immunity and lipid handling. Autosomal dominant APP/PSEN1/PSEN2 disease accounts for a small fraction of cases but supplies the field's timeline: CSF Aβ42 falls about 25 years, amyloid PET and CSF tau change about 15 years, and hypometabolism and episodic-memory impairment about 10 years before expected symptom onset (Bateman 2012, PMID 22784036).

Architecture at a glance

Tier Examples Effect size Population frequency Clinical use today
Autosomal dominant APP, PSEN1, PSEN2 Near-complete penetrance; age at onset largely variant-determined 279 pathogenic/likely-pathogenic variants catalogued globally (PMID 39903689) Diagnostic and predictive testing with counselling; trial eligibility
Chromosomal Trisomy 21 (extra APP copy) ~95% lifetime risk of AD; ~70% diagnosed with dementia by ~54 y Down syndrome population Screening and dedicated trial programmes (PMID 40818475)
Strong common variant APOE ε4 (dose-dependent), ε2 (protective) ε4/ε4 OR 14.9 (10.8–20.6) in white clinic/autopsy samples ε4 allele common worldwide Genotype required before anti-amyloid therapy for ARIA risk stratification
Rare coding variants TREM2 R47H, SORL1, ABCA7, ATP8B4, ABCA1 OR ~2–3 for TREM2 R47H R47H 0.46% in Icelandic controls ≥85 y Research; occasional diagnostic yield in early-onset cases
Common polygenic background 91 ADRD loci beyond APOE Individually small; decile-level risk ratios ~1.6–2.0 Ubiquitous Research risk scores only

Heritability

The largest twin study screened all Swedish Twin Registry members aged 65+ (11,884 pairs; 392 pairs with at least one AD case) and estimated heritability at 58% in the full model and 79% in the best-fitting model, with the remaining variance attributed to non-shared environment and no significant sex difference after age adjustment. Within concordant pairs, the intrapair difference in age at onset was larger in dizygotic than monozygotic pairs, indicating genetic influence on timing as well as occurrence (Gatz 2006, PMID 16461860). High heritability and a substantial non-shared environmental term coexist — which is why the genetics on this page and the modifiable-risk agenda on risk reduction and prevention are not competing claims.

APOE

Dose, direction and the reference numbers

Genotype OR vs ε3/ε3 (95% CI), white clinic/autopsy samples Source
ε2/ε2 0.6 (0.2–2.0) Farrer 1997, PMID 9343467
ε2/ε3 0.6 (0.5–0.8) PMID 9343467
ε2/ε4 2.6 (1.6–4.0) PMID 9343467
ε3/ε4 3.2 (2.8–3.8) PMID 9343467
ε4/ε4 14.9 (10.8–20.6) PMID 9343467
ε3/ε4, Japanese 5.6 (3.9–8.0) PMID 9343467
ε4/ε4, Japanese 33.1 (13.6–80.5) PMID 9343467

The ε4 effect was present at all ages from 40 to 90 but diminished after 70, and varied by sex (PMID 9343467).

The ancestry gradient, re-measured

Belloy's analysis of 68,756 individuals (21,852 East Asian, 5,738 Hispanic, 7,145 non-Hispanic Black, 34,021 non-Hispanic White) found a stepwise attenuation of ε3/ε4 risk: East Asian OR 4.54 (95% CI 3.99–5.17), White 3.46 (3.27–3.65), Black 2.18 (1.90–2.49), Hispanic 1.90 (1.65–2.18). The protective ε2 effect showed a parallel gradient — White OR 0.53 (0.48–0.58), Black 0.69 (0.57–0.84), Hispanic 0.89 (0.72–1.10, non-significant) — and, deviating from the pattern, ε2 was not protective in East Asian individuals (OR 0.97, 0.77–1.23). Within Black participants, lower global African (or higher European) ancestry tracked with stronger ε4 dose effects, but global ancestry did not explain the attenuated Hispanic associations. The sex-by-age interaction (higher ε3/ε4 risk in women) replicated but shifted to ages 60–70 (OR 1.48, 1.10–2.01) and was reproduced in a Black/Hispanic meta-analysis (OR 1.72, 1.01–2.94) (Belloy 2023, PMID 37930705). The practical consequence: an APOE-based risk estimate generated from European-ancestry data is not transportable, and neither is an ε2-based reassurance in East Asian populations.

APOE4 homozygosity as a genetic form of the disease

Analysing 3,297 individuals pathologically and 10,039 clinically, Fortea found that almost all APOE4 homozygotes showed AD pathology; biomarker levels diverged from APOE3 homozygotes from age 55; by 65 nearly all had abnormal CSF amyloid and 75% a positive amyloid scan; symptom onset was earlier (65.1 years) with a narrower 95% prediction interval than in APOE3 homozygotes; and the sequence of biomarker change mirrored autosomal dominant AD and Down syndrome. At the dementia stage, amyloid and tau PET did not differ across haplotypes despite the earlier trajectory (Fortea 2024, PMID 38710950). Combined with the amyloid-onset ages from the Jansen meta-analysis — 15% amyloid positivity reached at ~40 years in ε4ε4, ~55 in ε3ε4, ~65 in ε3ε3 and ~95 in ε2ε3 (Jansen 2015, PMID 25988462) — the case for treating ε4/ε4 as a distinct entity for prevention trials and counselling is quantitative rather than rhetorical.

APOE genotype is also the principal modifier of ARIA risk, which is why genotyping moved from research tool to pre-treatment requirement.

Autosomal dominant AD

A global catalogue assembled from DIAN, published literature and public databases identified 550 variants in APP, PSEN1 and PSEN2, of which 279 were classified pathogenic or likely pathogenic by ACMG/AMP criteria; symptomatic age-at-onset estimates were derived for 227, and 226 met eligibility criteria for disease-modifying trials. Mean variant age at onset and parental age at onset both predicted symptomatic onset, validated against DIAN Observational Study converters (Liu 2025, PMID 39903689).

Autosomal dominant disease is not exclusively familial. In a French screening programme of 129 sporadic early-onset cases (mean onset 45 ± 2 years), 17 carried a PSEN1 mutation and one an APP duplication (13% yield); where parental DNA was available in 10 such cases, the mutation was de novo in every one. Across the whole series, 90 distinct mutations were found and definite pathogenicity was established for only 77% of them, and among 53 mutation carriers with CSF, 46 (87%) had all three biomarkers abnormal and none had all three normal (Lanoiselée 2017, PMID 28350801). Two operational conclusions: restricting PSEN1 screening to familial cases misses de-novo carriers, and variant-classification uncertainty is a live problem in counselling.

The DIAN timeline

DIAN's cross-sectional analysis of 128 participants indexed to estimated years from expected symptom onset produced the field's canonical sequence (Bateman 2012, PMID 22784036):

Change Years before expected symptom onset
CSF Aβ42 concentration declines ~25
Amyloid deposition detectable (PiB PET) ~15
CSF tau rises; brain atrophy begins ~15
Cerebral hypometabolism; impaired episodic memory ~10
Global cognitive impairment (MMSE, CDR) ~5
Meets dementia criteria ~3 years after onset estimate

The authors' own caveat is important and often dropped: these are cross-sectional data requiring longitudinal confirmation and "may not apply to patients with sporadic Alzheimer's disease" (PMID 22784036). The sequence nevertheless underpins how fluid biomarkers and imaging are ordered in staging schemes.

A natural experiment in resistance

A PSEN1 E280A carrier from the large Colombian kindred did not develop MCI until her seventies — roughly three decades after the expected clinical onset for that variant. She was homozygous for the APOE3 Christchurch (R136S) variant and had unusually high brain amyloid with limited tau and neurodegeneration (Arboleda-Velasquez 2019, PMID 31686034). It is a single case, but it dissociates amyloid burden from tau and clinical expression in a genetically deterministic setting, and it made APOE-directed therapeutics a mainstream target.

Down syndrome

Trisomy 21 gives a third APP copy and, with it, a genetically determined AD. Nearly all individuals with Down syndrome have AD pathology by age 40; about 70% are diagnosed with dementia by around age 54; lifetime risk is approximately 95%; and AD is the leading cause of death in adults with Down syndrome older than 35 (Rafii 2025, PMID 40818475).

Meta-analysis puts age at onset at 53.8 years (95% CI 53.1–54.5, n=2,695), age at death at 58.4 years (57.2–59.7, n=324) and disease duration at 4.6 years (3.7–5.5, n=226), with coefficients of variation comparable to autosomal dominant AD. US mortality data show median age at death rising from 1 year (IQR 0.3–16) in 1968 to 57 years (49–61) in 2019, but with a ceiling in the highest percentiles (90th percentile: 63 years in 1990, 65 years in 2019). Modelled fully-penetrant AD matched up to 80% of deaths, against dementia being mentioned on only 30% of death certificates — and 78.9% in the DABNI cohort. Racial disparities persisted, most sharply at the lower percentiles (10th percentile age at death: 1 year for Black individuals versus 30 years for White individuals in 2019) (Iulita 2022, PMID 35604690). The authors' conclusion is stark: life expectancy in Down syndrome will not rise further until AD is treatable.

Common and rare variants beyond APOE

Study Design Result
Kunkle 2019 (PMID 30820047) GWAS meta-analysis, 94,437 individuals with clinically diagnosed LOAD Confirmed 20 loci, identified 5 new (IQCK, ACE, ADAM10, ADAMTS1, WWOX); HLA-DR15 haplotype confirmed as risk; pathways implicate immunity, lipid metabolism, tau-binding proteins and APP metabolism; rare-variant enrichment (P=1.32×10⁻⁷)
Bellenguez 2022 (PMID 35379992) Two-stage GWAS, 111,326 diagnosed/proxy cases, 677,663 controls 75 risk loci, 42 new; microglial implication; a new genetic risk score yielding a 1.6–1.9-fold risk increase from lowest to highest decile beyond age and APOE ε4
EADB 2026 (PMID 42237039) European-ancestry consensus meta-analysis, 128,681 (proxy) cases, 849,833 controls 91 loci (16 new; 56 detected specifically in clinically diagnosed AD), plus 18 loci flagged for external validation; non-APOE polygenic score associated primarily with AD rather than non-AD pathology — top decile had a twofold higher risk of Braak stage >IV with moderate-to-severe neuritic plaques at death versus median
Holstege 2022 (PMID 36411364) Exome sequencing, 16,036 cases vs 16,522 controls Rare damaging variants in ATP8B4 and ABCA1 associated with AD, alongside TREM2, SORL1, ABCA7; suggestive for ADAM10; largest-effect variants (especially loss-of-function) enriched in early-onset cases
Jonsson 2013 (PMID 23150908) Whole-genome sequencing of 2,261 Icelanders with imputation and replication TREM2 R47H (rs75932628-T) OR 2.92 (95% CI 2.09–4.09) in Iceland; 2.90 (2.16–3.91) combined; control frequency 0.46% at age ≥85; non-demented carriers aged 80–100 had poorer cognition (P=0.003)
Guerreiro 2013 (PMID 23150934) Sequencing in 1,092 cases vs 1,107 controls plus meta-analysis and replication Excess exon-2 TREM2 variants in cases (22 vs 5 variant alleles, P<0.001); R47H association confirmed in imputation meta-analysis and in 1,887 additional cases vs 4,061 controls

The convergence matters more than any single locus: three independent analytic strategies (common-variant GWAS, exome burden testing, rare-variant sequencing) all point to amyloid processing, lipid handling and microglial function — the substrate of neuroinflammation and glia.

Ancestry coverage is a structural weakness

African American individuals have roughly twice the AD risk of non-Hispanic White individuals from the same community, yet the largest GWAS have been conducted in European-ancestry samples. In 5,896 African American participants, ABCA7 rs115550680 reached genome-wide significance with OR 1.79 (95% CI 1.47–2.12, P=2.2×10⁻⁹), an effect size comparable to the APOE ε4-determining SNP in the same sample (OR 2.31, 2.19–2.42) (Reitz 2013, PMID 23571587). A larger meta-analysis (2,784 cases, 5,222 controls) using the African Genome Resources panel found four novel common loci (EDEM1, ALCAM, GPC6, VRK3) and rare-variant signals including an intergenic locus near IGF1R; of 25 loci known in non-Hispanic White individuals, only APOE, ABCA7, TREM2, BIN1, CD2AP, FERMT2 and WWOX replicated at nominal significance. Pathways overlapped (immunity, lipid processing, intracellular trafficking) even though the specific loci differed, and a kidney-system pathway emerged as novel (Kunkle 2021, PMID 33074286). A polygenic score derived in Europeans is therefore expected to underperform in exactly the populations with the highest burden — see epidemiology and burden.

Rare coding variation makes the immune pathway causal, not merely reactive

A three-stage association study of 85,133 people identified protein-altering variants in three microglia-expressed genes: PLCG2 p.Pro522Arg was protective (OR 0.68, P=5.38×10⁻¹⁰), ABI3 p.Ser209Phe increased risk (OR 1.43, P=4.56×10⁻¹⁰), and TREM2 p.Arg62His increased risk (OR 1.67, P=1.55×10⁻¹⁴) (Sims 2017, PMID 28714976). The bidirectionality matters: microglial signalling can either reduce or increase risk depending on the perturbation, so “activate microglia” and “suppress inflammation” are both genetically under-specified therapeutic strategies.

This result also connects common-risk architecture to mechanism more cleanly than a locus-count does. PLCG2, ABI3 and TREM2 form an immune-related interaction network enriched for previously known AD genes; the finding arose from rare coding variants with interpretable protein changes, not solely from non-coding proximity. It strengthens the case developed in neuroinflammation and glia while leaving the direction and disease stage for intervention unresolved.

Testing, disclosure and behaviour

The REVEAL programme is the main randomised evidence base for what happens when APOE risk is disclosed to asymptomatic adult children of people with AD (Roberts 2005, PMID 16306249).

  • Perceived risk depends on the framing, not only the number. Among two groups of women given identical 29% lifetime risk estimates, 73% of those whose estimate incorporated an ε4-negative genotype judged their risk lower, versus 25% of those given a family-history-based estimate (P<0.0001); 63% versus 9% said the assessment removed uncertainty (LaRusse 2005, PMID 15654228).
  • Disclosure changes behaviour even without proven prevention. One year after disclosure — with participants explicitly told no proven preventive measures existed — ε4-positive participants were more likely to report AD-specific health-behaviour change (adjusted OR 2.73, 95% CI 1.14–6.54, P=0.02) (Chao 2008, PMID 18317253).
  • Adding actionable pleiotropic information is safe and may help. In a randomised equivalence trial of 257 asymptomatic adults, disclosing APOE-related coronary-artery-disease risk alongside AD risk produced equivalent 12-month anxiety (mean BAI 3.5 in both arms) and depression (CES-D 6.4 vs 7.1), lower test-related distress among ε4 carriers (difference −4.8, 95% CI −8.6 to −1.0), and more health-behaviour change regardless of genotype (Christensen 2016, PMID 26810768).

These trials studied volunteers without severe anxiety or depression who sought risk information, so they bound the harm in a self-selected group rather than in the general population. The parallel evidence on biomarker disclosure is on diagnostic criteria and the biological definition.

Open questions

  • Should APOE4 homozygosity be operationalised as a distinct diagnostic entity with its own prevention trials and counselling pathway, as Fortea's data imply (PMID 38710950)?
  • Why does the ε2 protective effect vanish in East Asian populations and the ε4 effect attenuate in Hispanic populations without an ancestry explanation (Belloy 2023, PMID 37930705)?
  • What fraction of "sporadic" early-onset AD is de-novo autosomal dominant disease, and should screening criteria change accordingly (Lanoiselée 2017, PMID 28350801)?
  • Can polygenic scores derived in European-ancestry samples be recalibrated for African-ancestry populations, given that only 7 of 25 European loci replicated nominally (Kunkle 2021, PMID 33074286)?
  • Does the APOE3 Christchurch mechanism generalise beyond one case, and can it be pharmacologically mimicked (Arboleda-Velasquez 2019, PMID 31686034)?
  • Do the ~23% of APP/PSEN variants without established pathogenicity represent benign variation, reduced penetrance, or classification failure (Lanoiselée 2017, PMID 28350801; Liu 2025, PMID 39903689)?
  • Would the psychological safety demonstrated in REVEAL volunteers hold if APOE genotyping became a routine prerequisite for treatment eligibility rather than an elective disclosure (Roberts 2005, PMID 16306249; Christensen 2016, PMID 26810768)?

References

  1. Gatz M, et al. Role of genes and environments for explaining Alzheimer disease. Arch Gen Psychiatry. 2006;63:168-74. PMID 16461860.
  2. Farrer LA, et al. Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease: a meta-analysis. JAMA. 1997;278:1349-56. PMID 9343467.
  3. Belloy ME, et al. APOE genotype and Alzheimer disease risk across age, sex, and population ancestry. JAMA Neurol. 2023;80:1284-1294. PMID 37930705.
  4. Fortea J, et al. APOE4 homozygozity represents a distinct genetic form of Alzheimer's disease. Nat Med. 2024;30:1284-1291. PMID 38710950.
  5. Jansen WJ, et al. Prevalence of cerebral amyloid pathology in persons without dementia: a meta-analysis. JAMA. 2015;313:1924-38. PMID 25988462.
  6. Bellenguez C, et al. New insights into the genetic etiology of Alzheimer's disease and related dementias. Nat Genet. 2022;54:412-436. PMID 35379992.
  7. Kunkle BW, et al. Genetic meta-analysis of diagnosed Alzheimer's disease identifies new risk loci and implicates Aβ, tau, immunity and lipid processing. Nat Genet. 2019;51:414-430. PMID 30820047.
  8. EADB, et al. Consensus meta-analysis of genome-wide association studies for Alzheimer's disease and related dementias. Nat Genet. 2026;58:1214-1225. PMID 42237039.
  9. Holstege H, et al. Exome sequencing identifies rare damaging variants in ATP8B4 and ABCA1 as risk factors for Alzheimer's disease. Nat Genet. 2022;54:1786-1794. PMID 36411364.
  10. Jonsson T, et al. Variant of TREM2 associated with the risk of Alzheimer's disease. N Engl J Med. 2013;368:107-16. PMID 23150908.
  11. Guerreiro R, et al. TREM2 variants in Alzheimer's disease. N Engl J Med. 2013;368:117-27. PMID 23150934.
  12. Reitz C, et al. Variants in the ATP-binding cassette transporter (ABCA7), apolipoprotein E ε4, and the risk of late-onset Alzheimer disease in African Americans. JAMA. 2013;309:1483-92. PMID 23571587.
  13. Kunkle BW, et al. Novel Alzheimer disease risk loci and pathways in African American individuals using the African Genome Resources panel: a meta-analysis. JAMA Neurol. 2021;78:102-113. PMID 33074286.
  14. Lanoiselée HM, et al. APP, PSEN1, and PSEN2 mutations in early-onset Alzheimer disease: a genetic screening study of familial and sporadic cases. PLoS Med. 2017;14:e1002270. PMID 28350801.
  15. Liu H, et al. The landscape of autosomal-dominant Alzheimer's disease: global distribution and age of onset. Brain. 2025;148:2429-2440. PMID 39903689.
  16. Bateman RJ, et al. Clinical and biomarker changes in dominantly inherited Alzheimer's disease. N Engl J Med. 2012;367:795-804. PMID 22784036.
  17. Arboleda-Velasquez JF, et al. Resistance to autosomal dominant Alzheimer's disease in an APOE3 Christchurch homozygote: a case report. Nat Med. 2019;25:1680-1683. PMID 31686034.
  18. Rafii MS, et al. Down syndrome and Alzheimer's disease: insights into biomarkers, clinical symptoms, and pathology. Lancet Neurol. 2025;24:753-762. PMID 40818475.
  19. Iulita MF, et al. Association of Alzheimer disease with life expectancy in people with Down syndrome. JAMA Netw Open. 2022;5:e2212910. PMID 35604690.
  20. Roberts JS, et al. Genetic risk assessment for adult children of people with Alzheimer's disease: the REVEAL study. J Geriatr Psychiatry Neurol. 2005;18:250-5. PMID 16306249.
  21. LaRusse S, et al. Genetic susceptibility testing versus family history-based risk assessment: impact on perceived risk of Alzheimer disease. Genet Med. 2005;7:48-53. PMID 15654228.
  22. Chao S, et al. Health behavior changes after genetic risk assessment for Alzheimer disease: the REVEAL study. Alzheimer Dis Assoc Disord. 2008;22:94-7. PMID 18317253.
  23. Christensen KD, et al. Disclosing pleiotropic effects during genetic risk assessment for Alzheimer disease: a randomized trial. Ann Intern Med. 2016;164:155-63. PMID 26810768.
  24. Sims R, et al. Rare coding variants in PLCG2, ABI3, and TREM2 implicate microglial-mediated innate immunity in Alzheimer's disease. Nat Genet. 2017;49:1373-1384. PMID 28714976.