For thirty years the story of tau has been a story about shape. Tau lets go of the microtubule scaffolding it helps to steady, it collects chemical tags, it clumps, and eventually it knots into the neurofibrillary tangles that pathologists have counted in autopsied brains since the beginning. Tangles became the plot. A paper published online in Neuron on 6 August 2026 argues that some of the damage happens somewhere else entirely, in the dark interior of the mitochondrion, and that it happens before any knot forms. Wen Li and Suman Rimal share lead authorship, Bingwei Lu of Stanford's pathology department is the senior author, and colleagues at UCSF's Memory and Aging Center, including Lea Grinberg, Salvatore Spina and William Seeley, contributed the human tissue. The journal received the manuscript on 19 March 2026 and accepted it on 8 July; a preprint went up on bioRxiv on 7 April.

What the team reports is a piece of ordinary cell machinery running the wrong way. Electrons normally travel forward through the respiratory chain, and complex I is the first station on that line. Under certain conditions electrons can be pushed backwards through complex I, a process called reverse electron transport, which throws off reactive oxygen species and pulls down the ratio of NAD+ to NADH, two of the cell's basic accounting entries for energy and repair. The authors report that tau regulates this reversal, that ageing and stress switch it on, and that phosphorylated tau is the form that gets inside. They found p-tau in mitochondria purified from rTg4510 tauopathy mouse brain, still there after protease treatment that would have destroyed anything left on the outside, distributed where the inner membrane sits. Inside, it latches onto NDUFS3, a complex I subunit, and the normal contact between NDUFS3 and its neighbour NDUFV1 loosens. Reverse flow rises. Forward flow, notably, does not change.

The loop is the part that will interest clinicians. More reverse electron transport means more oxidative stress, and more oxidative stress means more tau phosphorylation, which means more tau able to enter mitochondria. The authors describe it as self-perpetuating, and they say the same signature appears in human brain: in post-mortem tissue from people who had Alzheimer's disease and from people who had progressive supranuclear palsy, mitochondria showed higher reactive oxygen species, a lower NAD+/NADH ratio, elevated reverse electron transport, and more p-tau bound to NDUFS3 than in normal brain. In flies carrying the FTD-linked tau mutations P301L and R406W, reverse flow ran higher than with ordinary human tau. And when tau was removed altogether, in flies, in mouse cells and in human iPSC-derived neurons, stress no longer triggered the reversal, and the cells and animals tolerated stress better. That inversion matters: the paper frames reverse electron transport as a job tau normally does, which goes wrong with age and disease, rather than as pure poison.

There is a compound. Called CPT in the paper, and CP-235 by Cerepeut, Inc., the company developing it, it blocks p-tau from docking on NDUFS3 and shuts down the reversal without disturbing normal forward electron flow. In tauopathy mice with marked cognitive deficits, an extended course of CPT suppressed reverse electron transport in brain mitochondria, and the treated animals did better across behavioural testing, with less inflammation and fewer signs of neurodegeneration. This is where a reader should slow down rather than speed up. Bingwei Lu is a co-founder and advisory board member of the company developing CPT, which he has disclosed, and he has been plain about the distance still to travel: "Much more work remains to be done before it can enter clinical trials." No person has received this compound. There is no trial, no dose, no safety data in humans.

So what does a family living with a tau diagnosis actually take from this? Not a treatment, and not a prognosis. What it offers is a different explanation for something people describe long before scans show much: the exhaustion, the way a bad night or an infection or a hot week can knock cognition sideways and then, sometimes, knock it back. If stress and ageing flip a switch inside mitochondria, and if that switch feeds the very tau chemistry that flipped it, then the fluctuation is not imagination and not weakness. It is also worth saying what the paper does not claim. It does not show that tangles are irrelevant, and it does not show that a person's decline is caused by this loop. Mitochondrial dysfunction has been documented in these diseases for decades without yielding a drug that works in people; a named binding partner and a compound that hits it are progress in the laboratory, not in the clinic.

The honest reading is that the field has been handed a new place to look, with a molecular address, NDUFS3, and a testable prediction: block the reversal, spare the neuron. Predictions of that kind have failed before, in mice that recovered and patients who did not. This one is at least specific enough to fail cleanly, which is more than most mechanisms offer.