PET Scans Reveal How Ketamine Reshapes the Depressed Brain

For about 30 percent of people with depression, antidepressants simply don’t work. They try one drug, then another, then perhaps a third, and the illness persists, grinding through years and sometimes decades. Ketamine, the anaesthetic repurposed as a rapid-acting antidepressant over the past two decades, has offered many of them their first relief. Nobody, until now, could quite say why.

The puzzle wasn’t that ketamine works. It manifestly does, often within hours. The puzzle was the mechanism: animal studies had pointed at a particular class of receptor in the brain, but there was no way to confirm those findings in living humans. “Although ketamine has shown rapid antidepressant effects in patients with treatment-resistant depression,” says Takuya Takahashi at Yokohama City University Graduate School of Medicine, “its molecular mechanism in the human brain has remained unclear.”

What Takahashi’s team had, which almost nobody else in the world has, is a PET tracer they built themselves. Called [¹¹C]K-2, it’s a radioactive molecule that binds selectively to AMPA receptors, a class of glutamate receptor that sits on the surface of neurons and plays a central role in how synapses strengthen or weaken. When you inject [¹¹C]K-2 into a person and slide them into a scanner, you get a picture, in reasonably fine detail, of how densely those receptors are distributed across the living brain. No comparable tool existed before they developed it.

The receptors matter because of what animal experiments had suggested. Ketamine is primarily known as an NMDA receptor blocker, but that’s not the whole story. Downstream of that initial block, AMPA receptors appear to be reshuffled, their density changing in different regions, in a way that seems to be where the antidepressant effect actually lives. The question was whether any of that happened in people.

It does. The team scanned 34 patients with treatment-resistant depression before and after a two-week course of ketamine infusions (four sessions, 0.5 milligrams per kilogram each time), comparing them against 49 healthy participants. What they found was a brain that had been considerably rearranged, region by region, in ways that tracked closely with how much each patient improved.

The pattern wasn’t uniform. Ketamine increased AMPA receptor density in cortical regions, including the precuneus and the parietal and occipital lobes; those increases predicted symptom improvement. In the habenula, a small subcortical structure deep in the brain, the effect ran the other way: receptor density fell, and the more it fell, the better patients tended to do. That detail matters because the lateral habenula has an established role in encoding something like negative expectation, the signal that fires when an anticipated reward fails to arrive; in animal depression models it tends to run hyperactive, and ketamine quietens it. Now there’s evidence the same quietening happens in people. “Ketamine’s antidepressant effect in patients with TRD is mediated by dynamic changes in AMPAR in the living human brain,” Takahashi says. “Using a novel PET tracer, [¹¹C]K-2, we were able to visualize how ketamine alters AMPAR distribution across specific brain regions and how these changes correlate with improvements in depressive symptoms.”

There’s a separate, perhaps stranger finding tucked into the occipital data. Patients with depression often describe their visual world as somehow dimmed, colourless, a bit washed out, and neuroimaging studies have found disrupted processing in the visual cortex. The new data show that ketamine’s antidepressant effect correlates with AMPA receptor increases in that region too, possibly pointing toward a neurobiological account of why the world starts looking like itself again after treatment. It’s tentative, but it’s the kind of convergence that makes neuroscientists sit up.

The clinical hook is this: you can, it turns out, predict response. Patients whose baseline AMPA receptor profiles (before a single dose of ketamine) showed certain patterns in the frontal, temporal, parietal and insular cortex went on to improve more. The receptors aren’t just markers of what ketamine does; they may indicate, in advance, who it will help. That’s the sort of biomarker the field has been hunting for rather a long time.

Limitations apply, as they always do with imaging work this ambitious. The sample was 34 patients, all Japanese, all from a single clinical trial, and the imaging analysis wasn’t pre-registered with statistical power to match. The team also couldn’t use an active placebo (something that mimics ketamine’s dissociative side effects), which means some patients may have guessed which arm they were in. Takahashi himself holds the patent on [¹¹C]K-2 and is a founder of the company that licences it, a conflict of interest the paper declares plainly.

None of that changes the basic finding, which is that the AMPA receptor story, long established in rodents, now has its first direct human support, visualised region by region in the living brain.

What comes next is harder to predict. Ketamine works, but its effects last only a few weeks, and long-term safety remains genuinely uncertain. If the mechanism is better understood, cleaner drugs that target AMPA receptor dynamics more selectively, without the dissociative effects, become at least conceivable. Whether [¹¹C]K-2 or something like it ever becomes a clinical prediction tool is another question entirely. For now, it’s enough that the black box has opened a little, and what’s inside looks more like what the animals were always suggesting.

Study link: https://www.nature.com/articles/s41380-026-03510-w