Can a Peptide Break the Alzheimer's Tau Loop?

Tau is the protein that forms the tangles in Alzheimer's disease. It turns out that tangling is not all it does.

When tau picks up phosphate groups, some of it gets inside the mitochondrion, the compartment that makes your cells' energy. There it finds one component of the energy machinery, a subunit called NDUFS3, and grabs hold of it.

And the power plant starts running backwards.

Tau gets inside and grabs one part. Drawn from Li et al., Neuron 2026.

Your mitochondria normally pass electrons forward along a chain of protein complexes, using the energy released to build a gradient that makes ATP. Forward flow, energy out.

When the electron carrier pool gets too loaded, or the membrane charge runs too high, that flow can reverse. Electrons run backwards into Complex I instead of forward out of it. It has a name: reverse electron transport.

Two things come out of the machine when it runs that way.

Reactive oxygen species, in bulk. The damaging molecules that oxidative stress is made of.

And NAD+ burned in reverse, converted back to NADH, dropping the NAD+/NADH ratio your cells run their repair and signaling on.

In plain English: the same machine, pointed the other way, produces damage instead of energy while draining the currency you need to clean it up.

Same machine, pointed the wrong way.

Here is the part that turns a mechanism into a trap.

Those phosphate groups are what let tau do this in the first place. A tau mutant that cannot be phosphorylated at the key spot did not trigger reverse transport at all. Two enzymes attach those groups to tau: GSK-3beta and MARK. Treat ordinary tau with GSK-3beta, and it drives the backwards flow hard.

Now look at what the backwards flow does to those two enzymes. In the brains of mice with tau disease, both sat switched on: GSK-3beta's off-switch disabled, MARK activated.

More phosphorylated tau, more reverse electron transport, more ROS and less NAD+, more active tau kinases, more phosphorylated tau.

It only turns one way. Call it a ratchet.

A pawl on a tooth. Phosphorylated tau, reverse flow, damage up and NAD+ down, kinases on, round again.

And they found it in a lot of places: human stem-cell-derived neurons carrying a tau mutation, flies, two different tauopathy mouse lines, and, the part that matters most, mitochondria purified from the brains of people who died with Alzheimer's disease and progressive supranuclear palsy. Elevated ROS, lower NAD+/NADH, elevated reverse transport, with forward transport relatively untouched. In the Alzheimer's tissue they could see the phosphorylated tau stuck to NDUFS3.

The researchers broke the ratchet three ways in animals: a small molecule called CPT, the old lab poison rotenone, and knocking down NDUFS3 itself. Mice on oral CPT held onto memory and, on MRI, had more cortex, bigger hippocampi and smaller ventricles than untreated animals carrying the same mutation.

All three are preclinical. None is available to anyone.

I have a specific reason for reading that last sentence twice.

My whole genome came back in April. One result outranked everything else on it: rs429358 TC, rs7412 CC. APOE e3/e4.

One copy of e4, the allele most strongly tied to late-onset Alzheimer's disease. The meta-analysis in my notes puts the odds around 3.2 times those of the common e3/e3 genotype. It is the finding that reorganized my priorities, and it is why I do not read a tau paper as neutral science.

So when a paper tells me the disease runs on a loop inside mitochondria, my first question is not when CPT is coming. It is this: which peptides already aim at that machinery?

There is a whole shelf of them, and I take one already.

Three candidates, and they are not equal.

SS-31, now approved as elamipretide. A four-amino-acid peptide that concentrates in the inner mitochondrial membrane and binds cardiolipin, the lipid the electron transport machinery is built on. It stops cardiolipin peroxidation and electron leak, and in isolated mitochondria it increased respiration and the efficiency of ATP synthesis. In September 2025 the FDA gave it accelerated approval as Forzinity for Barth syndrome. It is a real drug you can point at.

Humanin. This one is almost too on the nose. It is a 24-amino-acid peptide encoded inside your mitochondrial DNA, discovered in 2001 by screening for whatever was keeping neurons alive in an Alzheimer's brain that should have lost them. Its analog S14G-HN prevented memory impairment in amyloid-injected mice, and in 9-month-old Alzheimer's-model mice with plaques already built, three months of it improved learning and reduced plaque burden. A mitochondrial peptide found in surviving neurons, in a paper about mitochondria driving neuron death.

MOTS-c. Also mitochondrially encoded. Enhances memory in mice, works through AMPK, and comes with a hard limit the others do not: the one paper that measured it found peripheral MOTS-c does not cross the blood-brain barrier. They had to bolt a cell-penetrating tag onto it to get brain effects. Anyone selling MOTS-c for your brain is contradicting the only experiment that checked.

So: a cardiolipin-stabilizing approved drug, a peptide your own mitochondria write that was found in the neurons that refused to die, and one that cannot get into the brain.

Now the part I did not expect when I started looking.

I went looking for any study testing any peptide against reverse electron transport, against Complex I, against NDUFS3, or in a tauopathy animal.

There are none.

Not SS-31. Not humanin. Not MOTS-c. Every SS-31 brain study is an amyloid model, or normal aging, or sepsis, or anesthesia. Humanin's only tau data is two cell-culture papers where it blocked tau hyperphosphorylation, and even those work through a phosphatase called PP2A rather than through the kinases or the mitochondrial loop in this paper. There is no registered clinical trial of any mitochondrial peptide in Alzheimer's, MCI or any tauopathy. Zero.

The mechanism lines up, the molecules exist, one of them is FDA approved, and the experiment has simply never been run.

Nobody has tested one against this loop.

That is not a reason to lose interest. It is the most interesting thing on the page. A mechanism paper this specific is exactly how you decide which experiment to run next, and the tools are sitting right there.

Here is the tension I cannot resolve, and I would rather hand it to you than hide it.

The researchers broke this loop with inhibitors. CPT blocks reverse transport at Complex I. Rotenone blocks Complex I outright. Knocking down NDUFS3 does the same thing by removing the part.

SS-31 does the opposite. Its whole pitch is making electron transport more efficient. Reverse transport is driven by an over-loaded carrier pool and a high membrane charge, which is not obviously a state you want to push a struggling mitochondrion toward.

Nobody has measured which way that lands. It could go either way.

The same question hangs over something I have taken every morning for years. CoQ10 is sold as a mitochondrial supplement, and the form I take, ubiquinol, is the loaded version of that same electron carrier. An honest reading of this paper says I do not know whether that helps or hurts here, and neither does anyone else. That unknown is the thing most likely to make "peptides fix this loop" wrong.

Skip the peptides for a second, because the other arm of the loop, the NAD+ side, does have human data.

In a crossover trial of 37 people with subjective cognitive decline or mild cognitive impairment, eight weeks of nicotinamide riboside, a NAD+ precursor, lowered pTau217 by 7% while the placebo group rose 18% (p = 0.02). pTau217 is the blood marker that currently tracks Alzheimer's pathology best.

That is a NAD+ molecule moving a tau number in living humans, which is the exact relationship this paper predicts.

Now the part the supplement ads leave out: cognition did not change. Not in that trial, and not in a second one where 20 people with MCI took a gram a day for ten weeks, tripled their blood NAD+, and scored the same. A third study found brain NAD+ takes about four weeks of daily dosing to move at all, and moves by wildly different amounts in different people.

A biomarker moved. Nobody has shown a person thinking better. Both of those are true and you need both.

Given all of that, here is my honest inventory.

Urolithin A, 500 mg every morning. Mitophagy: the process that identifies broken mitochondria and recycles them. This paper is about mitochondria turned into damage factories. Clearing the worst ones is the most direct thing I own.

Creatine, 10 g every morning. A phosphate buffer for ATP, which is the currency this loop drains. It does nothing to tau. It gives the system slack.

Sleep, defended like an asset. Magnesium and taurine at 9 pm and a hard line on the hours. My own note when the genome came back said sleep moves to the top, and I have not changed my mind.

Running, about 25 miles a week. Which is also doing the other half of my APOE job: ApoB from 102 to 82, best HDL in a year.

Nicotinamide riboside, 1 g every morning. Tru Niagen. This is the arm with the only human tau-biomarker signal, the same molecule and close to the same dose as the trial above, and I was already on it before I read this paper. I also take niacinamide, a second NAD+ precursor, and apigenin at night, which blocks CD38, one of the enzymes that burns NAD+ as you age.

And the honest gap. I take CoQ10, which is a reduced electron carrier, and after this paper I am genuinely less sure about it than I was a week ago. I also take none of the three peptides this post is about, because nobody has tested them against tau in a person.

For thirty years, tangles and broken mitochondria sat on the same list of Alzheimer's hallmarks with an unexamined "and" between them. This paper replaces that "and" with a grip on a named protein, and shows the ratchet turning in flies, in mice, in human neurons and in human brain tissue.

That is a target. Targets are how drugs happen.

The peptides that live in that exact machinery have never been tested against it, and that gap is not a reason to shrug. It is a to-do list, and it is the kind of thing I would like to be early to rather than read about in five years.

I carry one copy of e4. I am watching this one closely.

Source: Li W, Rimal S, Bhurtel S, et al. "Tau-induced mitochondrial reverse electron transport drives neurodegeneration." Neuron, online August 6, 2026. doi:10.1016/j.neuron.2026.07.012.

Rebody runs peptides the way I wanted them run for myself: pharmacy sourced, batch tested, with a clinician on the dose.

For entertainment and education only. Not medical advice. Talk to your doctor before starting anything.