Amyloid and tau, the two proteins that three decades of Alzheimer’s research have cast as rival villains, may be fighting over the same parking spot on the same cellular structure. That is the proposal in a new paper from a four-person chemistry group at UC Riverside, and if it holds, the field has been litigating the wrong question.

The proposal, out in PNAS Nexus from Thomas Shoff, Maxence Derbez-Morin, Peishan Cai, and chemistry professor Ryan Julian, is almost embarrassingly simple. Amyloid beta and tau are not running on separate tracks. They are competing for the same site, and the surface they are fighting over is the microtubule.

Picture a neuron. Microtubules are the cell’s railroad: the long structural beams that hold the axon open, and the tracks along which the cell’s cargo (vesicles, mitochondria, the works) gets shuttled from the cell body out to the synapse. Tau’s day job, the one everyone learns in cell biology, is to bind those tracks and stabilize them. Tau is the maintenance crew. When tau falls off and starts clumping into tangles, the tracks fall apart, transport collapses, and the neuron dies. Textbook.

Here is the new bit. Julian’s group ran fluorescence polarization assays and showed that amyloid beta, the protein everybody else has been obsessing over outside the cell, also binds microtubules, and binds them tightly. Aβ1–42, the species most associated with disease, latches on with a dissociation constant of 2.72 µM (lower means stickier). Aβ1–40, the less pathogenic species, was looser at 5.33 µM. When the team added tau, it knocked Aβ partially back off the tubules. The two proteins were fighting for similar real estate, and a sequence-homology analysis suggested they look enough alike at the binding region that they would. You can read it as: Aβ is a tau impersonator, sticky enough to elbow the real maintenance crew aside.

BINDING AFFINITY TO MICROTUBULES (µM (Kd, lower = tighter))
Aβ1–422.72Aβ1–405.33
Aβ1–42, the species most linked to disease, binds microtubules about twice as tightly as Aβ1–40 in the UC Riverside assays. Source: Shoff et al., PNAS Nexus 2026

Reframe the disease through that single fact and the whole picture rotates. If Aβ is competing tau off microtubules inside the neuron, then plaques outside the neuron might not be the engine of damage so much as the exhaust pipe, which is the implication the authors actually lean into. The toxicity in their model lives on the tracks: either Aβ riding the tubules and doing tau’s job badly, or tau getting displaced and free-floating until it tangles. Both halves of the field’s signature debate become symptoms of the same underlying collision. That is why the authors call it a “unifying theory.” It is also why a chemistry lab is the one publishing it: the question turns out to be a binding-affinity problem dressed in three decades of neuroscience.

Now the caveat, and I want to phrase it the way the authors did, because I almost never see this in a paper that’s getting press. This is an in-vitro biophysics study. The microtubules in the tube were porcine. There are no mice, no neurons, no living brain tissue, no human data of any kind. Shoff and colleagues write, verbatim, that “the results and discussion presented herein do not in any way prove the microtubule nexus hypothesis is correct, but the idea certainly appears to be sufficiently compelling to merit further exploration.” That is what scientific honesty looks like in print, and it is rare enough now that it deserves to be quoted.

The reason the paper is news anyway is the institutional rubble it is landing on. The amyloid hypothesis did not just win three decades of Alzheimer’s research; it ate the field. Grant priorities, drug pipelines, journal real estate, peer-review weight, all tilted toward the assumption that clearing plaques would clear the disease. The most cited paper undergirding that bet, Sylvain Lesné’s 2006 Nature claim that a specific oligomer called Aβ*56 caused memory loss in rats, was retracted in 2024 with the notice citing “excessive manipulation, including splicing, duplication and the use of an eraser tool.” Vanderbilt neuroscientist Matthew Schrag, the whistleblower who flagged the doctored figures, called the data an “elaborate mirage” when Science first published the suspicions in 2022. Every author on the 2006 paper except Lesné himself signed the retraction. Per Retraction Watch’s count, it became the second most highly cited paper ever pulled.

And the drugs that the amyloid hypothesis finally produced have been a sobering payoff. Aduhelm was approved by the FDA in 2021 over its own advisory committee’s objections, then pulled from the market. Lecanemab (Leqembi), the survivor of that bet, did clear traditional approval on a real signal in CLARITY AD: a 27% slowing of cognitive and functional decline at 18 months, paired with brain-swelling and microhemorrhage risks (ARIA-E and ARIA-H) serious enough to require ongoing MRI monitoring. Eisai and Biogen priced it at $26,500 a year. Set “27 percent slower decline, $26,500” against thirty years of explanation that amyloid clearance would be the cure, and the gap between what the field promised and what it has shipped is hard to wave away.

So when a four-person chemistry lab on federal grant money (NIA R01AG066626, no industry funding, no competing interests declared) publishes a paper saying the framing has been slightly off-axis the whole time, the right response is not to crown them the way the field crowned Lesné. The right response is to take the mechanism seriously and demand the next experiment. Does Aβ actually displace tau from microtubules inside a living neuron? In a mouse brain? Does blocking that competition rescue cognition in an animal model? Those are the studies that have to land before this becomes a treatment hypothesis instead of a theory paper, and none of them are in this one.

What I keep coming back to is how clean the unification is. Two proteins, structurally similar enough to share a binding pocket, both implicated in a disease that wrecks the cellular structure that pocket sits on. The field spent thirty years arguing over which protein was the real villain instead of asking what they were both doing on the same surface, and that says less about the underlying biology than it does about how a dominant hypothesis can monopolize a field’s research questions until somebody outside the priesthood asks an obvious one. The fact that it took a chemistry group, rather than a neurology lab, is the tell.

I would not move a Leqembi prescription on the strength of a fluorescence assay. I will be watching, hard, for the cellular and animal work that has to come next. If it holds, the drug-discovery target is not amyloid clearance at all; it is the microtubule binding site, and the molecules worth screening for are the ones that protect tau’s seat at the table. That would be a very different program from the one this field has been running for a generation, and given how the last one has gone, “different” is starting to sound like an underrated virtue.

Sources

  1. PNAS Nexus – Shoff, Derbez-Morin, Cai & Julian, “The microtubule nexus linking amyloid beta and tau” (2026)
  2. PubMed – Retraction Note: A specific amyloid-β protein assembly in the brain impairs memory (Nature 440, 2006), retracted 2024
  3. Science – Researchers plan to retract landmark Alzheimer’s paper containing doctored images (Schrag investigation, 2022)
  4. NIA – Statement on report of lecanemab reducing cognitive decline (CLARITY AD, 27% slowing at 18 months)
  5. Alzheimer’s Association – Lecanemab (Leqembi) overview and $26,500 annual pricing