APOE

In one line: APOE is a humble cholesterol courier that comes in three versions — and which two you inherit is the single biggest common genetic dial on Alzheimer's risk at one end of life, and on the odds of reaching 100 at the other.

---

The gene

APOE (apolipoprotein E) sits on the long arm of chromosome 19, at 19q13.32. Its everyday job is lipid transport: it wraps cholesterol and other fats into lipoprotein particles and shuttles them around the body, acting as a ligand for the LDL-receptor family that pulls those particles back out of circulation. Inside the brain it is the dominant lipid carrier — produced mostly by support cells called astrocytes — ferrying cholesterol to neurons. Nothing about that delivery-truck role hints at dementia, which is exactly what made the discovery so surprising.

The twist that opens the episode: the three common versions — ε2, ε3, ε4 — are not equal, and ε4 is the ancestral allele, the one our lineage carried first. ε3 (now the most common, in more than half of people) and ε2 (the rarest) arose later in human evolution. "Oldest" turned out to mean "riskiest."

The hunt

By the early 1990s Alzheimer's research was fixated on the amyloid plaques seen in patients' brains. The APOE link came from a different direction — the families. A chromosome-19 linkage signal (Pericak-Vance, 1991) pointed the way, and in 1993 Allen Roses' group at Duke University landed it: ApoE bound amyloid-β, and the ε4 version was strikingly over-represented in people with late-onset Alzheimer's (Strittmatter et al.; Saunders et al.). The capstone was Corder et al. (1993), which showed the effect was a gene dose: across 42 late-onset families, risk climbed from ~20% to ~90% and the mean age of onset fell from ~84 to ~68 years as ε4 copies went from zero to two. A cholesterol gene was the largest common genetic risk factor ever found for the disease.

The mechanism

The three versions differ by almost nothing — two amino-acid swaps (cysteine↔arginine) at positions 112 and 158 of the mature protein. (Variant databases sometimes number these from the precursor, as 130/176; same residues, plus the 18-residue signal peptide.) ε3 = Cys112/Arg158; ε2 = Cys112/Cys158; ε4 = Arg112/Arg158. Those tiny changes reshape the protein and what it does in the brain.

As the brain works, neurons continually shed amyloid-β, a sticky peptide that must be cleared before it accumulates. APOE helps with that cleanup, but isoform-dependently: amyloid-β accumulation and plaque deposition follow the order ε4 > ε3 > ε2. ε4's shape makes it a poorer partner for clearing and a better one for aggregation, so the amyloid lingers and hardens into plaques. ε4 doesn't stop there — it also worsens tau pathology (partly independent of amyloid), drives neuroinflammation, and impairs cerebrovascular function. The courier meant to keep the space clear becomes part of the reason it clogs.

The other end of life

Turn the gene around and the same dial points the opposite way. ε2 is over-represented in centenarians (Schächter et al., 1994), and APOE consistently surfaces as a top locus in human-longevity GWAS. Meanwhile ε4 thins out of the oldest age groups — not because the gene changes, but because the people who carry it tend to die younger (a survivorship effect). One small gene tilts both ends of a human life.

The frontier

You can't trade in the alleles you were born with, but APOE has become one of the most pursued targets in Alzheimer's research, and the leads are tantalizing:

• The woman who resisted. A Colombian carrier of an autosomal-dominant early-onset Alzheimer's mutation (PSEN1 E280A) — who should have declined in her 40s — stayed cognitively intact into her 70s. She was homozygous for a rare APOE variant, APOE3-Christchurch (R136S), which appears to blunt tau spread despite a very high amyloid burden (Arboleda-Velasquez et al., 2019).
• Deliver the good version. Gene therapies are attempting to introduce the protective APOE2 into the brains of high-risk patients (e.g. an AAV-APOE2 program, early-phase/biomarker-stage).
• Two ε4 copies as its own disease. A 2024 study argued that ε4 homozygotes — roughly 2% of people — show such near-universal Alzheimer's pathology that the genotype behaves less like a risk factor and more like a distinct genetic form of the disease (Fortea et al.).

A practical note from the clinic: the new anti-amyloid antibodies (lecanemab, donanemab) carry a higher risk of brain-swelling side effects (ARIA) in ε4 carriers — highest in homozygotes — which is why APOE genotyping is now done before treatment.

The important caveat the episode keeps front and center: for carriers in general, ε4 is a risk factor, not destiny — many ε4 carriers never develop Alzheimer's, and many patients carry no ε4. That is categorically different from the rare autosomal-dominant early-onset genes (APP, PSEN1, PSEN2), where the mutation is essentially causative. (The ε4/ε4 homozygote case above is the emerging exception that proves the rule.)

Sources

Full claim-by-claim evidence is in references.md. Primary anchors:

• Strittmatter et al. 1993, PNAS 90:1977 — ApoE binds amyloid-β; ε4 enriched in late-onset familial AD.
• Corder et al. 1993, Science 261:921 — ε4 gene dose raises risk and lowers age of onset.
• Saunders et al. 1993, Neurology 43:1467 — ε4 association with late-onset AD.
• Farrer et al. 1997, JAMA 278:1349 — meta-analysis of ε4 odds ratios (and their ancestry/sex dependence).
• Liu, Kanekiyo, Xu & Bu 2013, Nat Rev Neurol 9:106 — APOE in AD: risk, mechanisms, therapy (canonical review).
• Liao, Yoon & Kim 2017, Curr Opin Lipidol 28:60 — APOE metabolism/function in the brain; isoform effects.
• Fullerton et al. 2000, Am J Hum Genet 67:881 — ε4 is the ancestral allele; ε3/ε2 derived.
• Schächter et al. 1994, Nat Genet 6:29 — ε2 enriched, ε4 depleted in centenarians.
• Arboleda-Velasquez et al. 2019, Nat Med 25:1680 — APOE3-Christchurch resistance case.
• Fortea et al. 2024, Nat Med 30:1284 — ε4 homozygosity as a distinct genetic form of AD.

Accuracy notes (the traps the fact-gate caught): (1) The ">tenfold" figure is for ε4 homozygotes (~12–15× vs ε3/ε3); one copy is ~2–3×, and all of these are ancestry-dependent — so the episode keeps numbers qualitative and ties "more than tenfold" to two copies. (2) ε4 is ancestral but not most common (ε3 is). (3) Amino-acid positions are given in mature-protein numbering (112/158); databases may use precursor numbering (130/176). (4) The mechanism is framed as impaired clearance / promoted aggregation, not amyloid over-production. (5) ε4's decline with age is survivorship, not the gene "changing." (6) The Christchurch protective variant is on an ε3 background (APOE3ch), not ε2. (7) Gene therapy is early/biomarker-stage, not a proven treatment.