Putting a bright idea to the test

The first big surprise came sometime around 2014, Li-Huei Tsai says. A doctoral student in her lab at the Massachusetts Institute of Technology had come to her with an idea for an experiment: Hunter Iaccarino wanted to induce gamma rhythms in a mouse model of Alzheimer’s disease and measure the effects on amyloid beta levels in the brain.

Tsai’s lab had previously discovered how to generate gamma oscillations in the brain by way of optogenetic stimulation. But before Iaccarino proposed it, no one had attempted to associate those rhythms with any molecular, cellular or biochemical changes, says Tsai, director of MIT’s Picower Institute for Learning and Memory.

The experiment felt novel, Tsai recalls. And it took Iaccarino just one assay to show that inducing gamma oscillations optogenetically in the hippocampus at 40 hertz lowered amyloid beta levels in the mice—a finding so exhilarating that Iaccarino “ran down the hallway, ran into my office and showed me the plate readout,” Tsai says.

The next eye-opener came when Tsai’s group tested a new, noninvasive way to induce gamma waves—a light flickering at 40 hertz—in 5XFAD mice, a common model of Alzheimer’s disease. The light also lowered amyloid beta levels in the animals’ visual cortex before they developed any of the condition’s hallmark plaques. And even in mice with plaque deposits, the amyloid beta load was reduced.

When they published the results in Nature in 2016, they almost immediately garnered attention in the popular press and ignited discussions on Twitter (now X). “Holy s**t: evidence that 40 Hz visual flicker can treat Alzheimer’s,” wrote Mark Humphries, a systems neuroscientist at the University of Nottingham. The surprises kept coming: Tsai’s paper has since been cited more than 1,100 times, and at least two dozen entities have filed or been granted patents focused on treating the brain with some sort of stimulation around 40 hertz.

The frontrunner in that group is Cognito Therapeutics, co-founded by Tsai and MIT neurotechnology professor Ed Boyden, who co-led the 2016 publication with Tsai. The company has raised $128 million in venture capital funding, been awarded 11 U.S. patents protecting its technology and has advanced a therapeutic device—called Spectris—into a pivotal trial, with plans to report results to the U.S. Food and Drug Administration by the end of next year.

Nearly eight years after the initial finding, however, it still isn’t clear—to the MIT researchers or anyone else—just how 40 hertz stimulation might help reduce amyloid in the brain, and failed attempts to replicate the initial experiment have observers wondering if the idea of noninvasive stimulation for Alzheimer’s disease is just more false hope in a field long plagued by it.

There have been “numerous attempts to affect this biology in a lot of different ways,” says Bryce Mander, associate professor of psychiatry and human behavior at University of California, Irvine, and many of them looked good in mice but sputtered in humans. “We’ve cured mouse-heimer’s thousands of times,” he says. “We fail to cure Alzheimer’s disease.”

T

he work in Tsai’s lab was based on something researchers had realized years prior: People with Alzheimer’s have disturbances in gamma brain waves. Although these waves fall in the 25 to 100 hertz range, Tsai and her group found that stimulating the brain in mice at 40 hertz proved to be a kind of sweet spot: Optogenetic stimulation not only lowered levels of amyloid beta in the hippocampus, but also altered the shape of microglia in that region.

The effect on microglia in particular seemed intriguing. Over recent decades, increasing evidence has suggested that microglia play a role in neurodegeneration, and it was shocking that noninvasive 40 hertz light might somehow thwart that process, Mander says. “That, to me, was amazing.”

Microglia also seemed to form clusters in the hippocampus, auditory cortex and prefrontal cortex when Tsai and her colleagues later induced gamma waves there in 5XFAD mice, using 40 hertz auditory tones in combination with visual stimulation. And the cells showed reduced inflammation after 40 hertz visual stimulation in two additional mouse models of neurodegeneration.

Despite these tantalizing clues, the link unifying gamma entrainment in the brain and a microglia response—never mind its apparently positive effects on plaques, memory, spatial learning and synapses—remained elusive. In the beginning, Tsai says, she and her team were convinced microglia were involved in clearing amyloid from the brain. But because a mouse model with depleted microglia still showed amyloid reduction after the 40 hertz exposure, now she thinks microglia “probably are not very important.”

Meanwhile, Cognito Therapeutics began developing a therapeutic device based on Tsai’s research in 2016. The Spectris headset combines glasses and over-ear headphones to produce flashes of light and pulses of sound, both at a frequency of 40 hertz. In 2018, the company launched a feasibility trial at Emory University and a phase I/II multicenter clinical trial dubbed OVERTURE. The company’s first patents came through in 2019.

When the OVERTURE study ended in 2020, 53 of the 76 participants enrolled had completed the trial, and it did not meet its primary endpoint: There was no statistical difference between the treatment arm and sham on the MADCOMS scale, a modified version of a clinical survey that assesses cognition and function in people with Alzheimer’s disease. Also, there was no change in amyloid in the brain, based on three PET scans over six months.

There were positive signs, however, including a 69 percent reduction in whole brain volume loss in the treatment arm versus the sham arm, and Cognito progressed to a pivotal trial, called HOPE, with two new primary endpoints; it began enrollment in early 2023. The company has “not used MADCOMS going forward,” says Brent Vaughan, Cognito’s CEO, and the new primary endpoints—gauging a participant’s basic function and cognition—reflect guidance from the FDA, he says.

In January 2024, Cognito announced it had closed a $35 million Series B funding round, bringing its total raised to $128 million. And in April, the company presented additional results from the open-label extension of the OVERTURE trial at a conference, showing that the statistically significant difference on the brain volume measure seen at the end of the 6-month trial persisted in the 12-month extension.

The HOPE trial is currently enrolling. Earlier this year, Cognito also launched a substudy of HOPE to measure unspecified biomarkers related to Alzheimer’s disease and neurodegeneration. The company plans to submit all relevant data to the FDA’s Center for Devices and Radiologic Health by the end of 2025, Vaughan says.

Yet the company also went through a round of layoffs in April. It has “decided to focus as much of our efforts as possible around execution of the Alzheimer’s study,” Vaughan says.

I

f there is uncertainty on the industry side, the scientific community is perhaps even less sure about the future of 40 hertz effects, with some researchers questioning if Tsai’s original study does what it says it does. For Tsai’s part, she points to other researchers who deployed the 40 hertz flicker and found positive effects on amyloid beta clearance, and outside of Alzheimer’s research, other groups have also seemingly found success with 40 hertz to treat insomnia, stroke and traumatic brain injury in mice.

Audio and visual flicker also reduced interictal epileptiform discharges, a biomarker of epilepsy, in trial participants in a study published in May of this year. The study was led by Annabelle Singer, associate professor of biomedical engineering at the Georgia Institute of Technology and Emory University, and co-first author of the 2016 Nature paper. There is promise in the technology, she says, but more data and trials in humans are needed. (Singer is the sister of The Transmitter’s opinion and community editor, Emily Singer, who was not involved in editing or reporting this story.)

But a skeptical contingent points to work done by György Buzsáki, professor of neuroscience at New York University. Last year, he published a study showing that his team could not entrain 40 hertz gamma oscillations in the three regions of the brain in both the 5XFAD and APP/PS1 mouse models of Alzheimer’s disease using flashing light. This, he and his team wrote, indicated “that 40-Hz flickering light does not reliably reduce Aβ load in the neocortex or hippocampus, potentially due to the large variability of AD pathology.” The authors concluded “the hypothesized flicker light entrainment of native gamma oscillations is not a strong candidate mechanism for reliably affecting AD pathology.” (Buzsáki declined to be interviewed for this article.)

Tsai says she had gotten used to skepticism about the published work—people sometimes tell her directly that the idea seems too good to be true—but after the paper from Buzsáki, a “heavyweight” in the field of brain rhythms, “we always have to defend, so to speak.”

Part of her defense is a preprint that points out issues she and her colleagues have with Buzsáki’s replication attempt. One problem, they write, is that Buzsáki pooled results from mice of different ages, sexes and even strains, which could have altered the results.

Buzsáki’s replication attempt might have worked if the group had used a bigger sample size, according to a preprint from scientists at OptoCeutics, a Berkeley, California-based company with a vested interest in 40 hertz as a treatment: It is recruiting for a clinical trial of its own device, a table-top light box, and in the meantime selling it for about $2,000 as a product for “general wellness.”

Whether the effect of 40 hertz in the lab is replicable is one thing. How it might work is another. Once Tsai lost enthusiasm for her ideas about microglia, she developed a new theory—one that implicates another kind of glial cell and the glymphatic system. In this new hypothesis, which she and her colleagues described in Nature in February, the 40 hertz stimulation promoted the influx of cerebrospinal fluid that can then clear out amyloid in 5XFAD mice, they reported. That finding provides “direct evidence that neuronal activation can accelerate glymphatic waste clearance from the brain,” Lauren Hablitz and Maiken Nedergaard of the University of Rochester wrote in an accompanying News and Views article. Work in this area is still ongoing.

Tsai hoped this new paper would open people’s minds about 40 hertz stimulation, she says. But doubt still exists. A PubPeer thread questioned the statistics and power behind the study. One commenter wondered if the work involved pseudoreplication, or correlated samples, because it should not be possible to get the p-value listed from an experiment in which n equals 5. Stanley Lazic, who writes about pseudoreplication in neuroscience, told The Transmitter via email he agrees with this commenter after looking at the data and says there is “likely” pseudoreplication.

In response, Tsai points out that both the statistical analysis and the methodology were peer reviewed, and her group did everything “by the book.” But trying to define the mechanism behind a new theory can be like putting “a puzzle together,” she says, and her group is still trying to produce the pieces.

When Mander allows himself to “walk in the realm of possibilities,” he finds himself hopeful about 40 hertz stimulation. If it worked, it would be “pretty amazing” he says. And the kind of excitement Iaccarino felt upon seeing the initial plate readout is not lost on Mander. For him, “it’s always exciting to see basic science work and see how people find different ways to manipulate these mechanisms.”

But that only goes so far, Mander says. “At the end of the day, phase 3 clinical trials are going to be needed to show this is effective,” he says. “And until we get there, you know, it’s not going to be different from a lot of other failed trials.”