A neuron that has stopped forming memories might not be dead. It might just be hungry. That possibility, once a fringe idea in Alzheimer’s research, is gaining serious traction after three independent mouse studies published in 2024 and 2025 each found a way to restart the brain’s failing energy supply and, in doing so, reversed cognitive deficits that had already taken hold.
The findings arrive against a backdrop of growing unease with the field’s long-dominant theory: that amyloid plaques are the primary cause of Alzheimer’s symptoms. Anti-amyloid antibodies like lecanemab and donanemab have reached the clinic, but their cognitive benefits have been modest, slowing decline by roughly 25 to 35 percent in trials without restoring lost function. Meanwhile, brain imaging studies have been quietly building a case that something else goes wrong even earlier than plaque accumulation, and that something is energy.
The decade-long blackout before symptoms appear
The human evidence anchoring this story comes from FDG-PET imaging, a scanning technique that tracks how much glucose the brain consumes region by region. In families carrying rare inherited Alzheimer’s mutations, researchers in the Dominantly Inherited Alzheimer Network (DIAN) have shown that glucose metabolism begins dropping roughly a decade before the person notices any memory trouble. That is not a subtle dip. By the time symptoms emerge, entire cortical networks are running well below normal metabolic output.
A separate analysis of postmortem Alzheimer’s brain tissue found that levels of thiamine diphosphate, a cofactor that cells need to extract energy from glucose, tracked closely with the degree of metabolic impairment. Critically, amyloid plaque burden in those same brains did not correlate with the metabolic decline. Plaques were present, but they were not the variable that predicted how starved the tissue looked. The implication is uncomfortable for a field that has spent billions targeting amyloid: the thing that tracks best with neuronal dysfunction is not the protein deposits but the energy deficit.
None of this proves amyloid is irrelevant. Plaques may act as an upstream trigger, setting off inflammation or vascular changes that eventually choke the fuel supply. But the timing data suggest that by the time a patient walks into a memory clinic, the metabolic crisis has been underway for years, and that crisis, not the plaques themselves, may be what makes neurons stop firing properly.
Three ways to refuel a starving brain, all tested in mice
If energy failure is the proximate cause of cognitive symptoms, then restoring energy should restore function, at least partially. Three research teams tested that logic using very different tools.
Turning up mitochondrial voltage. A team led by researchers at the National Institutes of Health, publishing in Nature Neuroscience, engineered a synthetic receptor called mitoDREADD-Gs that sits on the surface of mitochondria, the organelles that generate most of a cell’s energy. When activated by a designer drug, the receptor increases mitochondrial membrane potential and oxygen consumption, essentially forcing the cell’s power plants to work harder. In mice whose cognition had been pharmacologically impaired, switching on the receptor rescued performance on memory tasks.
Bypassing glucose with lactate. A separate group reported in Scientific Reports that supplementing lactate improved brain metabolism in 5XFAD transgenic mice, a widely used Alzheimer’s model, as well as in healthy controls. Lactate is an alternative fuel that neurons can burn when glucose processing falters. The treated animals showed improved synaptic markers and better performance on learning tasks, suggesting their neurons were not just surviving but functioning more normally once given a fuel they could actually use.
Unblocking the astrocyte supply chain. A third study, published in Nature, found that inhibiting an enzyme called IDO1 restored hippocampal memory in Alzheimer’s model mice by rescuing astrocyte metabolism. Astrocytes are support cells that normally convert glucose into lactate and shuttle it to neurons. In Alzheimer’s brains, a metabolic pathway that diverts the amino acid tryptophan into potentially toxic byproducts appears to jam that shuttle system. Blocking IDO1 freed astrocytes to resume feeding neurons, and cognition improved.
Each study attacked the energy problem from a different angle: one boosted the engine inside neurons, one supplied a different grade of fuel, and one cleared a blockage in the delivery pipeline. All three produced measurable cognitive recovery. That convergence matters because it points to a shared bottleneck, cellular energy failure, rather than a quirk of any single experimental setup.
Why mouse memory rescue does not yet mean human therapy
The distance between a mouse navigating a water maze and a person trying to remember a grandchild’s name is vast, and every section of that distance is littered with failed Alzheimer’s drugs that worked beautifully in rodents.
The mitoDREADD-Gs receptor does not exist in human brains. Translating that approach would require either gene therapy to install the receptor or a conventional drug that mimics its effect on mitochondria, neither of which has been tested in people. The original paper also does not publish detailed behavioral score tables or exact group sizes for the neurodegenerative arm of the experiment, making it difficult for outside scientists to independently gauge how large and reliable the effect was.
The lactate and IDO1 studies face a different limitation: neither reports head-to-head comparisons on the same behavioral tasks, so there is no way yet to rank which metabolic lever pulls hardest. More importantly, none of the three studies tracked whether cognitive rescue lasted beyond the acute testing window. A treatment that works for hours or days in a mouse cage is a long way from one that sustains human cognition across years of progressive disease, and repeated dosing could surface safety problems that short experiments never reveal.
The human correlation data linking thiamine diphosphate to hypometabolism, while compelling, also has limits. The published analysis synthesizes findings across multiple small studies rather than presenting a single large dataset with individual patient-level FDG-PET values matched to thiamine concentrations. The strength of the relationship at the single-patient level is therefore hard to pin down from the published record alone.
And there is a broader question of generalizability. The human imaging data showing early hypometabolism comes largely from families with rare inherited Alzheimer’s mutations. The far more common late-onset form of the disease involves a messier tangle of vascular damage, chronic inflammation, and age-related metabolic decline. A patient with small-vessel disease, for instance, might not be able to deliver extra fuel to vulnerable brain regions even if those neurons remain capable of burning it. Whether metabolic rescue strategies would work in that more complex biological environment is genuinely unknown.
Where metabolism fits in the bigger picture
The metabolic hypothesis does not erase the amyloid and tau hypotheses. It intersects with them. Plaques and tangles may still initiate or accelerate damage, but the point at which neurons lose their energy balance could be where symptoms become clinically visible and, potentially, where therapies have the best chance of making a difference. That framing helps explain why anti-amyloid drugs can clear plaques without dramatically improving cognition: if the downstream energy crisis is already entrenched, removing the original trigger may be too little, too late.
Timing is likely critical. The FDG-PET data suggest that hypometabolism precedes overt dementia by years, which implies that any metabolic therapy would need to start early, perhaps before symptoms, to deliver the largest benefit. That raises practical and ethical questions about screening, risk prediction, and treating people who feel healthy but carry biomarkers of future decline.
Readers familiar with popular claims about ketogenic diets or MCT oil supplements as “brain fuel” should note an important distinction. While those approaches are loosely inspired by the same biology, providing ketones as an alternative energy source, rigorous clinical trial evidence that they slow or reverse Alzheimer’s progression in humans remains thin. The mouse studies described here used precisely targeted interventions under controlled conditions, a very different proposition from over-the-counter supplements.
As of June 2026, no metabolism-targeting drug has demonstrated robust disease modification in a large, randomized human trial. Early-stage clinical work is underway, and recent reporting on experimental metabolic therapies reflects both the excitement among researchers and the caution from investigators who have watched promising mouse results collapse in human testing before.
What this changes about how we think about Alzheimer’s
The most grounded takeaway from this body of work is conceptual, not clinical. In Alzheimer’s disease, neurons that appear to have failed may be more salvageable than anyone assumed a decade ago. They may not be dead. They may be starving. And three independent research teams have now shown, in mice, that feeding them can bring function back.
That does not mean a cure is imminent, or that any supplement on a pharmacy shelf will replicate what engineered receptors and targeted enzyme inhibitors achieved in a lab. It means that energy, not just protein deposits, deserves a central place in the search for treatments. And it means the window for intervention might be wider than the field once believed, stretching back into that long, silent decade when the brain is losing power but the person has not yet lost a single memory.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
