Grades assigned against primary sources verified 2026-09-27: Grumelot and Mo full-text XML staged and hashed; Akinc, Niemietz, Dong 1996, Dong 1998 abstracts verified character-level by efetch; the 2017 and 2018 supporting papers staged full-text but quotes not yet pulled. Absence claims carry pinned queries with dates.
| Claim | Tier | Study types | Confidence |
|---|---|---|---|
| The LNP protein corona is apolipoprotein-dominant and forms by lipoprotein fusion into the membrane, not shell adsorptionPMID 42677364 (cryo-TEM, no peripheral shell, ∼25–55% of LFQ intensity as apolipoproteins); PMID 41711348 (composition-dependence of ApoE-LNP structure). In vitro, one formulation set. | [MECHANISTIC] | PP | MODERATE |
| ApoE is an endogenous targeting ligand for ionizable (not cationic) LNPs, demonstrated in vitro and in apoE-/- and LDLR-/- micePMID 20461061, abstract verified character-level. Scope: DLin-MC3-DMA-generation siRNA LNPs. | [MECHANISTIC] | ANPP | HIGH |
| ApoE receptor binding is isoform-sensitive in principle (Arg158Cys mechanism shifts the receptor-binding region)PMID 8756331 mechanism; PMID 9643348 quantifies apoE-Leiden variants only (25% and 51.4% figures). Intact-apoE2 binding magnitude is NOT pinned from any held source and is not quoted. | [MECHANISTIC] | PP | MODERATE |
| Human LNP-drug efficacy stratified by APOE genotype has been tested once and was nullPMID 31648569, patisiran, E3/3 E3/4 E3/2 strata, no significant TTR differences; per-genotype N absent from abstract, E3/2 rare. Null-until-N-read. | [MECHANISTIC] | PP | LOW-MODERATE |
| Corona composition by ApoE genotype on any LNP in humans has never been measuredAbsence pinned 2026-09-27: PubMed exact-form hits 1/0/0 for the three pinned queries, the single hit being PMID 31648569; OpenAlex counts (143/126/319) are full-text reconstruction neighbourhoods, not exact-phrase evidence. | [HYPOTHESIS] | PP | MODERATE |
ApoE4 carriers write in with one question in various forms: do these lipid nanoparticles treat me differently because of my genotype? It is a fair question asked at the exact seam between two literatures that rarely cite each other. The ApoE isoform literature, built by the Gladstone group from the mid-1990s onward on the held citations, established that the three common ApoE proteins differ in receptor binding and lipid handling. The LNP delivery literature, built largely by the siRNA industry, established that ApoE is the ride these particles use to enter hepatocytes. The 2026 fusion papers connected the two more tightly than anyone had before. And the vaccine safety conversation keeps assuming the connection means something specific for E4 carriers without anyone stating which links are measured and which are not.
This article is that statement. Five links, each at its verified grade, one weak weld named, one human datapoint read honestly, one absence pinned with dates. Nothing more is claimed.
The isoform biochemistry is the Gladstone group's: Weisgraber, Innerarity and colleagues determined the receptor-binding structures and the E2 binding defect's mechanism (Dong et al. 1996, PMID 8756331; Dong et al. 1998, PMID 9643348). The demonstration that ApoE is an endogenous targeting ligand for ionizable LNPs is Alnylam's (Akinc et al. 2010, PMID 20461061). The fusion finding and the ApoE-dominant corona are Grumelot, Mohammed and colleagues, with Whitehead and Mahmoudi as corresponding authors (PMID 42677364), and the composition-dependence of ApoE-LNP structure is Mo and colleagues (PMID 41711348). The only human genotype-stratified LNP study is Niemietz and colleagues' (PMID 31648569). This article assembles a chain other people built and states where its welds are thin.
The 2026 cryo-TEM finding changed the picture of what the corona even is. LNPs "displayed no peripheral protein shell"; instead the membranes undergo "localized thickening and electron-dense remodeling consistent with lipoprotein integration rather than surface adsorption" (Grumelot et al., PMID 42677364). The proteomics in the same paper: "Apolipoproteins accounted for a mean of ∼25–55% of total LFQ intensity across the LNP fractions (varying by LNP formulation), compared with approximately 18% in neat plasma processed in parallel", which is selective association, not passive contamination, and the share varies by formulation. Mo and colleagues showed the sharp edge of that variance: ApoE-LNP behaviour "is highly dependent on the LNP composition", with a 10 mol% DSPC replacement reshaping morphology, uptake, endosomal escape and viability endpoints, in porphyrin-lipid LNPs, not the vaccine lipids (PMID 41711348). GRADE: [MECHANISTIC], MODERATE, in vitro, one formulation set.
Akinc and colleagues, in the paper that explained why Onpattro works: "apolipoprotein E (apoE), which plays a major role in the clearance and hepatocellular uptake of physiological lipoproteins, also acts as an endogenous targeting ligand for ionizable LNPs (iLNPs), but not cationic LNPs (cLNPs)" (Mol Ther 2010, PMID 20461061). The demonstration ran both in vitro, with recombinant ApoE, and in vivo, "in wild-type and apoE(-/-) mice", with receptor dependence explored in "low-density lipoprotein receptor (LDLR(-/-))-deficient mice". That is the load-bearing sentence for everything downstream in this article, and its scope is explicit: ionizable lipids, siRNA generation, hepatocyte delivery. GRADE: [MECHANISTIC], HIGH, within that scope.
A staged 2017 study extends the point from presence to mechanism: an ApoE-enriched corona switches the uptake mechanism and trafficking route of the particle (doi 10.1039/c7nr06437c; full text staged, quotes to be pulled before publish). The corona does not just decorate the particle; it changes how the cell takes it in.
The Gladstone structural work settled that receptor binding is isoform-sensitive. For E2, the defect is mechanistic and precise: "replacement of Arg 158 with Cys, disrupting the naturally occurring salt bridge between Asp 154 and Arg 158. A new bond between Asp 154 and Arg 150 is formed, shifting Arg 150 out of the receptor binding region" (Dong et al. 1996, PMID 8756331). The 1998 follow-up quantified binding for the apoE-Leiden family of variants: intact apoE-Leiden and its truncations "displayed similar activities (approximately 25% of apoE3's)", with a solid-phase assay reading "51.4+/-9.4%" (PMID 9643348).
One caveat is enforced here and travels with this article everywhere: those percentages concern apoE-Leiden variants, not intact ApoE2, and the commonly quoted magnitude for intact E2 binding has not been pinned from any held primary source in this workstream. No binding percentage for any isoform is quoted beyond what the two held abstracts state. The E4-specific side of the structural question is covered by a staged 2018 computational structure of lipidated ApoE4 receptor binding (doi 10.1371/journal.pcbi.1006165; quotes to be pulled before publish). What the held evidence supports is the principle, that single-residue differences move the receptor-binding region and change affinity. It does not support any specific claim about E4's binding magnitude on an LNP surface. GRADE: [MECHANISTIC], MODERATE, structural principle only.
Niemietz and colleagues did the only stratification anyone has done: patisiran-treated ATTR amyloidosis patients, genotyped. The mechanism control came back positive: "Downregulation of APOE suppressed efficacy of Patisiran in HepG2 cells." The clinical readout: "TTR levels were found to be robustly reduced (84.7% +/- 1%) following Patisiran treatment in patients plasma", and across the three most frequent genotypes, E3/3, E3/4 and E3/2, "APOE stratification of patients did not show significant differences of TTR plasma concentrations following treatment". Their own conclusion: "Our results suggest that APOE is an important mediator of TTR silencing by Patisiran, however efficacy is independent of the APOE genotype" (Amyloid 2020, PMID 31648569).
Read that carefully in both directions. It is a genuine human null on genotype, and it is the single study standing between the mechanism chain and any clinical claim. But the abstract carries no per-genotype numbers, and E3/2 is the rare genotype in any cohort, so a small E3/2 stratum cannot be excluded from where it sits; and the study tested the MC3-lipid siRNA product, not an mRNA vaccine. The honest grade is [MECHANISTIC], LOW-MODERATE: null result, power unassessable until the per-stratum numbers are read from the full text.
The question an ApoE4 carrier actually asks, does my corona look different, has never been measured in humans on any LNP. This is not a rhetorical absence. It was pinned with exact-form PubMed queries on 2026-09-27: "APOE genotype" AND "lipid nanoparticle" returns one hit, the Niemietz null study itself; "ApoE isoform" AND corona AND uptake returns zero; "ApoE4" AND "lipid nanoparticle" AND uptake returns zero. Broad database counts that look like they contradict this (OpenAlex returns 143, 126 and 319 for the same strings) match full-text reconstructions, not exact phrases; the neighbourhood they describe is ApoE-functionalised targeting particles and computational ApoE4 structure, not isoform-stratified corona measurement. GRADE: [HYPOTHESIS], absence dated 2026-09-27, queries reproducible.
Links 2 and 4 were demonstrated on DLin-MC3-DMA siRNA particles, the patisiran lineage. The authorised mRNA products use ALC-0315 and SM-102. No held study tests ApoE-dependence directly on the vaccine lipids. The fusion papers anchor the target side: they show what the corona of mRNA-type LNPs is made of, and that ApoE dominates it. Composition is not uptake-dependence. The chain is therefore: demonstrated mechanism in the siRNA generation, demonstrated composition in the mRNA generation, and the transfer between them is an inference, declared as one. Anyone quoting this article as "ApoE4 changes vaccine behaviour" or "proven irrelevant to vaccines" has quoted a link that does not exist in either direction.
Not self-treatment advice, and this article in particular should not be pushed in either direction.
The biodistribution article established that the particle is not neutral packaging and that its corona, not its manufactured surface, defines its biological identity. This article takes the corona layer one step personalised: whether one common host genotype changes that identity. The verified answer today: the mechanism is plausible at every link, the human evidence is a single null on the wrong lipid at unassessable power, and the direct experiment has simply never been run. When someone runs it, this page gets a dated update.
]]>