A Single Rapamycin Dose Normalized Autism-Like Behavior In Mice Within Two Hours. Daily Dosing Stopped Working

Stick a mouse in a plain gray box with a rough patch of floor and it will tell you something about its nervous system. Normal mice explore both halves, then settle in the dark one, because mice like the dark. The animals in Janel Le Belle‘s colony at UCLA did the opposite. Put 80-grit anti-slip strip under the dark chamber and they wouldn’t stay in it, preferring the glare of the lit side to the feel of grit under their paws.

Those mice had never been touched by a drug, a gene edit, or a surgeon. Their mothers had received one small injection of bacterial cell wall on the ninth day of pregnancy, at a dose so low the moms never got visibly sick and their litters came out full size.

What happened next, over 52 litters and several years of work published this week in Nature Communications, is a fairly complete rendering of autism-associated biology in a mouse: chronic inflammation across the body and brain, mild early brain overgrowth followed later by cortical thinning, an enlarged hippocampus and caudate, hair-trigger seizure susceptibility, repetitive grooming and circling, and the sensory aversion that showed up on the rough floor. Not one litter failed to produce the behavior. Which is either reassuring, if you build animal models for a living, or a unsettling if you think about what a single mild fever in the second week of a mouse pregnancy is doing here.

All of which set up the question that actually drives the paper. The physical brain of an adult mouse like this is different: thinner cortex, altered volumes, wiring laid down decades ago in mouse-years. Can you do anything about the behavior without rebuilding the brain?

Two Hours, Not Two Months

Rapamycin, an mTOR inhibitor best known as an immunosuppressant, has been thrown at autism mouse models before. Nearly always for weeks or months, on the reasonable theory that the drug works by preventing or slowly correcting structural damage. Le Belle and her colleagues gave adult mice a single dose (5 mg/kg, into the abdomen) and started testing two hours later. That timing is the whole point. Two hours is far too short for synapses to be physically rebuilt.

Nearly everything moved. Cortical pyramidal neurons in the somatosensory cortex, which in these mice fired in a pattern you don’t normally see in the upper layers, quieted down. Seizure threshold rose. Under a standard chemical challenge, all eight of the inflammation-exposed mice had seized where only two of eight controls did. Repetitive behaviors dropped. The rough-floor aversion eased, as did the abnormally low paw-withdrawal threshold, and startle habituation, which had been simply absent, came back.

The brain-wide picture shifted too. Resting-state fMRI showed these mice had chopped their functional networks into four modules where controls had three, with a modularity index of 0.693 against 0.511. After a single dose: three modules, index 0.552. Some researchers have argued that this kind of modular reorganization in autism is compensatory, a scaffold the brain builds to cope, and that flattening it would make things worse. Here, normalizing it tracked with the behavior improving.

“The level of functional normalization achieved over this short time suggests new mechanisms by which possible treatments may act,” says senior author Harley Kornblum, who directs UCLA’s Intellectual and Developmental Disabilities Research Center.

What The Drug Didn’t Fix

It fixed none of the anatomy. The cortex stayed thin, the hippocampus stayed large, the chronic cytokine load barely budged (the team checked, and acute rapamycin had a negligible effect on the circulating inflammatory markers, which rules out the simplest explanation). Connectivity didn’t fully normalize either: cortical hyperconnectivity came down toward control levels while subcortical connectivity went up, further from controls, and the mice got better anyway. Function and structure came apart.

“These results reframe how autism-associated symptoms might be treated,” says Le Belle, first author and an associate professor of neurosurgery at UCLA. “If the adult brain remains capable of functional normalization, then some features of autism may be successfully addressed without needing to correct underlying structural differences.”

Two control experiments narrow it further. RapaBlock, a compound that neutralizes rapamycin everywhere except the brain, didn’t abolish the rescue, so this is central, not peripheral. And wiping out microglia, the brain’s resident immune cells and an obvious suspect given all that inflammation, worked in young adults but not older ones. Rapamycin worked in both. Microglia help build the problem, then apparently hand it off.

The Catch Is A Real One

By 72 hours the repetitive behavior was back. Give the drug daily instead and the benefit shrinks a little every week until, by week five, it has largely gone. Tolerance, in other words, on top of a drug already known for toxicity in chronic use. Anyone reading this as a treatment story should stop there.

Co-senior author Neil Harris, also of UCLA neurosurgery, is clear. “This points toward new therapeutic targets like sensory circuit neuromodulation or balancing neuronal inhibition and excitation, rather than toward rapamycin itself as a treatment,” he says.

So the targets are the interesting part. The connectivity changes showed up in sensory circuits, not motor ones — even though the repetitive movement was the most obvious behavior in the cage. That fits a long-held suspicion: repetitive behavior may develop partly as a response to sensory overload, not just alongside it. Sensory over-responsivity is common across the autism spectrum and there’s very little to offer for it. If the relevant machinery is a thalamocortical loop that can be nudged by focused ultrasound or magnetic stimulation, that is a different sort of proposition than a lifelong immunosuppressant.

Mice are not people, and a rough floor in a plastic box is not a supermarket at four in the afternoon. But the paper does one useful thing: it separates the question of whether a brain was built differently from the question of whether it can currently be run differently.

Frequently Asked Questions