Hypothesis

Aβ accumulation alters neuronal membrane lipid order, which shifts the equilibrium of tau liquid‑liquid phase separation (LLPS) toward solid amyloid aggregates preferentially in high‑degree hubs of the structural connectome.

Mechanistic rationale

1. Aβ oligomers insert into lipid rafts, increasing local cholesterol and saturated phospholipid content, which raises membrane order (as shown for prion protein anchoring suppressing aggregation) [3].
2. Elevated membrane order promotes LLPS of tau by reducing the effective concentration of water and enhancing hydrophobic interactions within nascent condensates [2].
3. However, when membrane order exceeds a threshold, the same environment destabilizes the liquid state, driving LLPS‑to‑aggregate transition via secondary nucleation on fibrillar seeds, a process accelerated by Aβ‑induced CDK5/GSK3β activation [1].
4. Network neuroscience shows that hub regions exhibit higher baseline metabolic stress and express greater levels of membrane‑anchoring proteins, making them susceptible to this lipid‑order‑driven switch [4,5].
5. Neural manifold learning can capture the resulting spatial pattern of tau aggregation as a low‑dimensional deformation of the healthy connectome manifold, providing a geometry‑based readout of pathology spread [6].

Testable predictions

• In vitro, exposing synthetic lipid vesicles with increasing cholesterol/tau ratios will show a biphasic curve: LLPS droplet formation rises then declines as aggregates appear, detectable by turbidity and ThT fluorescence.
• In APP/PS1 mice, pharmacological depletion of membrane cholesterol (using methyl‑β‑cyclodextrin) will reduce tau aggregate burden in hub regions (e.g., hippocampus) without affecting overall tau phosphorylation levels.
• Mapping individual patient diffusion MRI-derived connectomes onto a neural manifold will reveal that nodes with highest predicted lipid‑order stress (estimated from PET‑measured Aβ and CSF cholesterol) exhibit the greatest tau‑PET signal, a relationship absent in cognitively normal controls.
• Genetic knockdown of membrane‑anchoring proteins (e.g., flotillin‑1) in human iPSC‑derived neurons will shift tau LLPS toward soluble condensates and decrease seeding competence, even in the presence of Aβ oligomers.

Potential falsification

If cholesterol manipulation fails to alter the LLPS‑to‑aggregate transition of tau in hub‑enriched cultures, or if manifold‑based predictions of tau spread do not correlate with lipid‑order biomarkers, the hypothesis would be refuted.