Most of what families are told about brain fuel fits on a fridge magnet: the brain runs on glucose. It is true, and it is incomplete. Alongside sugar, a small amount of ketone body circulates in the blood at all times, even in people eating ordinary meals, and during fasting or a low-carbohydrate ketogenic diet those levels climb ten to a hundred times higher. The question that has sat unanswered is who actually burns them inside the brain, and whether that burning is a luxury for lean times or a daily necessity. A preprint posted to bioRxiv on March 27, 2026, by Joyce Yang, Mitsunori Nomura, Ken Nakamura, John C Newman and colleagues at the Gladstone Institute of Neurological Diseases in San Francisco and collaborating labs, makes the case that it is a necessity, and that the neuron has to do the work itself.
The enzyme at the centre of the story is BDH1, beta-hydroxybutyrate dehydrogenase 1. It runs in both directions: it performs the last step when the liver makes the ketone body beta-hydroxybutyrate, and the first step when a cell wants to burn it, converting it back to acetoacetate before that fuel can enter the machinery that makes ATP. The team worked first in human neurons grown from induced pluripotent stem cells, silencing BDH1 with CRISPR interference and comparing those cells with non-targeting controls. Using carbon-13 labelling, they traced fuel from beta-hydroxybutyrate and acetoacetate into citrate, a marker that the fuel had entered the cell's central energy cycle, and they measured oxygen consumption directly. When the neurons were handed ketones as their fuel for three days and BDH1 was missing, fewer of them survived. Crucially, supplying acetoacetate, the metabolite just downstream of the blocked step, softened the damage. That is the shape of a real bottleneck rather than a coincidence: cut the pipe, the cell struggles; deliver fuel past the cut, and it partly recovers.
Then they took the question into living animals, and this is where it stops being a tidy biochemistry lesson. The researchers used mice in which Bdh1 could be deleted only in neurons and only in adulthood, giving tamoxifen at four months of age so that development was left untouched. These were mice on normal food, not fasting, not on a ketogenic diet. According to the preprint, losing the ability of neurons to oxidise ketones was enough to worsen mortality and to impair performance on a novel object recognition test at 24 months of age, with survival differences assessed by the standard log-rank and Gehan-Breslow-Wilcoxon methods. The authors' framing is blunt: ketone oxidation for energy is, in their words, "an essential mechanism for normal function in neurons and mice".
The last experiment is the one dementia readers will look at hardest. The team crossed the neuron-specific knockout into hAPPJ20 mice, a long-used model that overproduces human amyloid precursor protein. Removing the neurons' ability to burn ketones made things worse in that model too, increasing mortality and memory impairment. The interpretation the authors offer is that ketone oxidation is not merely a background housekeeping habit but a reserve that neurons lean on harder when they are already in trouble. Take the reserve away from a brain under amyloid stress, and the stress shows.
What this changes, conceptually, is the direction of the arrow. Much of the recent excitement around ketones has been about signalling: beta-hydroxybutyrate modifying proteins, altering gene expression, acting like a hormone. This preprint argues for the older and less glamorous story, that neurons need the calories, and that they have to run the conversion inside their own mitochondria rather than receiving pre-processed fuel from neighbouring astrocytes. For a family managing a diagnosis, the practical distance is still considerable. Nothing here tests a diet, a supplement, a ketone ester drink or a drug in a single human being. It identifies a step in the machinery and shows what happens when that step is deliberately broken in cells and in mice.
It is also worth being precise about what "aging breaks that rule" can and cannot mean today. This study removed the enzyme by design and watched the consequences unfold across an animal's lifespan. It is a demonstration that the capacity matters with age, not a measurement showing that human aging strips it away. Those are different claims, and the second one still needs its own evidence.