Scientists have long known that Alzheimer’s disease is driven by the buildup of toxic protein fragments in the brain. But until now, understanding exactly when and where these fragments appear — and how to stop them — has remained elusive.
It’s about 10 years of labor, driven by a very simple question: what goes wrong at the very beginning of Alzheimer’s disease?
—Jeffrey Savas, Northwestern University
Now, a new study from Northwestern University has identified a decades-old, FDA-approved anti-seizure drug that can prevent neurons from producing one of the most toxic protein fragments, amyloid-beta 42 (Aβ42), before it starts accumulating. The work, which examined engineered mouse models, human neurons, and brains from patients with Down syndrome, suggests a possible strategy for preventing Alzheimer’s in people at high genetic risk.
“This project really started in 2015,” Jeffrey Savas, senior author of the study and an associate professor of behavioral neurology, told DDN. “It’s about 10 years of labor, driven by a very simple question: what goes wrong at the very beginning of Alzheimer’s disease?”
Finding the earliest breakdown
Rather than starting with plaques or dementia, Savas’s team focused on protein turnover — the cell’s ability to make, use, and degrade proteins. Impaired protein turnover has long been associated with Alzheimer’s hallmarks like plaques and tau tangles, but it wasn’t clear which proteins go awry first. If they could identify proteins with impaired turnover during the very initial stages of pathology, it could highlight new mechanisms of causation for the disease.
Using gene-edited mouse models of Alzheimer’s, the researchers tracked how quickly proteins were made and broken down in the brain. By feeding the animals a harmless, non-radioactive isotope, they were able to track protein turnover across the brain, and pinpoint which pathways began to break down at the earliest stages of disease.
“Basically, what we found was that very few proteins have impaired turnover, but all of the proteins that had impaired turnover were associated with presynaptic terminals and synaptic vesicles,” Savas said.
Presynaptic terminals are the ‘sending’ side of neurons, where chemical signals are packaged into tiny sacs called synaptic vesicles and released to communicate with other cells. The findings pointed to a problem not with dying neurons, but with how neurons communicate, long before classic Alzheimer’s pathology appears.
Amyloid starts at the synapse
At the heart of the discovery is amyloid precursor protein (APP), a protein essential for normal brain development and communication between neurons. This protein can be processed in two ways: a harmless, non-amyloidogenic pathway, or an amyloid-producing pathway that generates amyloid-beta peptides.
Using the mouse models, the researchers found that the most toxic form, Aβ42, accumulates inside synaptic vesicles alongside other presynaptic proteins that fail to degrade properly. This buildup disrupts synaptic function early in the disease — before plaques form and before neurons die.
“We saw changes happening at the presynaptic terminal very early on — even before amyloid-beta began to build up. Our results pointed to this processing step as a key driver of the disease,” said Savas.
While the team focused on Aβ42, Savas noted that the mechanism likely affects all forms of amyloid-beta. “APP can be processed in different ways,” he explained. “When it goes down the harmless, non-amyloidogenic pathway, it’s cleared. When it’s processed by beta and gamma secretases, amyloid-beta is produced. Gamma secretase is a bit sloppy, sometimes generating different amyloid-beta lengths, like 37, 38, 40, or 42, in a largely random way. So, our findings could be relevant across these forms, not just Aβ42 — though the exact effects may vary.”
An unexpected intervention
The key turning point came from revisiting an old drug: levetiracetam, a generic anti-epileptic medication best known by the brand name Keppra.
This drug was first explored based on work by Michela Gallagher at Johns Hopkins University in 2012. She found that people with mild cognitive impairment who took the drug showed reduced abnormal brain hyperactivity, restoring activity to levels comparable to healthy controls and improving memory performance. Around the same time, Leonard Mucke at the University of California, San Francisco, showed that Alzheimer’s mouse models often develop seizure-like brain activity.
“What we uncovered is how the drug works. Levetiracetam binds to SV2A and, in both mouse models and human neurons, blocks the production of amyloid-beta. It does this by keeping APP on the cell surface, where it is more likely to be processed in a harmless, non-amyloid-forming way. That’s the short of it,” said Savas.
Following this, the team examined postmortem brain tissue from young adults with Down syndrome, who are prone to developing an aggressive, early-onset form of Alzheimer’s due to carrying an extra copy of the APP gene. They found the same early buildup of presynaptic proteins seen in the mouse models, showing that the pathway targeted by levetiracetam is also active in humans.
Levetiracetam is not a perfect drug, Savas noted, but it is widely used and well understood. It is one of the most commonly prescribed neurological drugs worldwide. Originally designed to stop seizures, the drug works by reducing neuronal hyperexcitability rather than broadly dampening brain activity. “In normal people, it doesn’t do much beyond some mild mood effects,” Savas explained. “What it really does is slow down the synaptic vesicle cycle — the process by which neurons release and recycle neurotransmitters — and reduces the size of overactive vesicle pools.”
The key idea behind the paper is that levetiracetam or levetiracetam analogues serve as a powerful tool to understand how Alzheimer’s begins — and point to a new way of intervening very early, before irreversible damage occurs.
—Jeffrey Savas, Northwestern University
“The key idea behind the paper is that levetiracetam or levetiracetam analogues serve as a powerful tool to understand how Alzheimer’s begins — and point to a new way of intervening very early, before irreversible damage occurs,” said Savas. The drug does have limitations, including a short half-life that requires daily dosing, but the researchers are exploring related compounds that could improve its stability and efficacy.
To develop better drugs, his team now has a simple readout: they can test which molecules keep APP on the cell surface in cultured mouse and human neurons. This allows them to screen other compounds, including endocytosis inhibitors or molecules similar to levetiracetam, as potential new therapies.
Timing is everything
Despite the promise, Savas is clear that levetiracetam is not a cure — and would not work if given after dementia has set in. “Alzheimer’s unfolds over decades,” he said. “Amyloid rises first, then tau, and dementia comes later. If you give this drug once neurons and synapses are already lost, it’s too late.”
However, when the researchers analyzed existing clinical data from the National Alzheimer’s Coordinating Center, they found that Alzheimer’s patients who happened to be taking levetiracetam lived longer after diagnosis than those taking other anti-epileptic drugs or none at all.
“It’s a modest effect — a few years,” Savas said. “But it supports the idea that blocking amyloid production can slow disease progression.” He suggested the drug could one day be used alongside plaque-clearing therapies, such as lecanemab, to prevent new amyloid from forming after existing plaques are removed.
Savas explained that the team is currently focusing on genetic forms of Alzheimer’s, such as in Down syndrome patients, people with APP or PSEN1 mutations, and possibly APOE4 carriers — groups who are highly likely to develop the disease. For them, starting a low-cost drug like levetiracetam in their 20s or 30s could make sense as a preventative approach.
For the majority of people who develop sporadic Alzheimer’s, the situation is more complicated. “We don’t currently have a way to predict who will benefit,” Savas said. Not everyone with mild cognitive impairment progresses to full Alzheimer’s, and existing FDA-approved diagnostic tests only work once dementia has already set in — by which point it’s too late for preventative treatment.
A new frontier in Alzheimer’s treatment
Savas explained that in healthy young brains, APP is processed safely and degraded. But as we age or cells are stressed, the process shifts, producing more amyloid-beta peptides like Aβ42. This change is likely a normal part of aging. Researchers are exploring whether presynaptic proteins and phosphoinositide lipids could serve as early biomarkers for Alzheimer’s. While promising, it’s still very early research.
The findings suggest a shift in how scientists think about Alzheimer’s: not focusing on clearing plaques after the fact but instead preventing their formation at the synaptic level. While still early, the research raises the possibility that combining early preventive therapy with synapse-supporting strategies could one day help at-risk individuals, particularly those with genetic predispositions, stave off the disease decades before symptoms appear.
