Clinical depression is considered one of the most treatable mood disorders, but neither the condition nor the drugs used against it are fully understood. First-line SSRI treatments (selective serotonin reuptake inhibitors) likely free up more of the neurotransmitter serotonin to improve communication between neurons. But the question of how SSRIs enduringly change a person’s mood has never returned completely satisfying answers.

In fact, SSRIs often don’t work. Scientists estimate that over 30 percent of patients don’t benefit from this class of antidepressants. And even when they do, the mood effects of SSRIs take several weeks to kick in, although chemically, they achieve their goal within a day or two. (SSRIs raise the levels of serotonin in the brain by blocking a “transporter” protein that decreases serotonin levels.) “It’s really been a puzzle to many people: Why this long time?” says Gitte Knudsen, a neurobiologist and neurologist at the University of Copenhagen, Denmark. “You take an antibiotic and it starts working immediately. That’s not been the case with the SSRIs.”

Experts have proposed theories about what causes the delay, but to Knudsen, the most compelling involve our brains’ ability to physically readjust over time: a characteristic called neuroplasticity. In adulthood, brains rarely create new neurons, but they do sprout new interconnections between existing ones, called synapses. Essentially, they adapt by rewiring. “That’s exactly what happens when we exercise and learn something,” Knudsen says. This transformation improves cognitive function and emotional processing. Knudsen thinks rewiring could also break someone free from cycles of negative rumination—a hallmark of depressive episodes.

Knudsen believes that SSRIs owe their efficacy at least in part to boosting neuroplasticity. Writing in Molecular Psychiatry earlier this month, her team showed how they had tested this theory on people, thanks to a special kind of PET scan developed in the past few years. They recruited 32 people to take the SSRI escitalopram (also known by the brand name Lexapro) or a placebo for one month. Then they asked the people to take a PET scan at the end of the trial, and used radioactive tracers to track where in the brain new synapses were forming.

The more time someone spent on the antidepressant before their brain scan, the more synaptic signals the team detected—a proxy for increased connections. “This is one of the first pieces of evidence that these drugs do take time to work, and they do work through increasing the number of synaptic contacts between nerve cells,” Knudsen says.

The finding suggests that SSRIs improve neuroplasticity during the first weeks or months of treatments, and that neuroplasticity contributes to the drugs’ benefit—and to the delay before users feel better. “It has been a paradox,” says Jonathan Roiser, a cognitive neuroscientist at University College London who was not involved in the work. Because the drugs’ chemical effects happen on a scale of days, he says, “there needed to be this extra bit of explanation about why the mood change does not happen immediately.”

“This is really important not just for general scientific understanding, but for actually improving our ability to treat patients,” says Camilla Nord, a cognitive neuroscientist at the University of Cambridge in the UK, who was not part of Knudsen’s team. “This could help us target the treatment at particular subgroups of patients—or maybe help us understand why it doesn’t work in some people.”

Since SSRIs were invented about 40 years ago, neuroscientists and psychologists have wanted to know exactly how they work. Studies clarified serotonin’s role about 20 years ago by proving that when serotonin levels rise, the brain veers away from negative biases in processing emotions. But these momentary changes in perception aren’t enough to relieve symptoms. “You need a cumulative exposure to a more positive input over time to break out of the depressive state,” says Roiser. “Previously, this had been the end of the explanation.”

One theory for why there’s a lag between the start of SSRI treatment and mood change is that the brain takes weeks to recalibrate serotonin levels. Think of it as a feedback system: Initially, after an SSRI spikes a person’s serotonin levels, their brain responds by pumping the brakes on producing the neurotransmitter. Instead of maintaining a boost, their serotonin levels fall again. “It’s like a thermostat,” Knudsen says. It takes a while before the brain adjusts.

“It is a fairly simplistic explanation that has helped doctors try to explain to patients why it takes time and what these drugs do,” Knudsen says. But, as a neurologist hoping to improve treatment, Knudsen wasn’t satisfied with this answer, partly because studies in rats suggested that a more complicated story was unfolding. These studies showed that in female rats given daily doses of SSRIs, new synapses formed in their visual cortex and hippocampus, brain areas linked to learning and memory. This indicated that SSRIs induce neuroplasticity.

But until about seven years ago, scientists couldn’t replicate these studies in humans, since there was no way to measure synaptic density without cutting out brain tissue. Then in 2016, researchers developed a way to detect synaptic activity in live human brains during PET scans. These scans detect light emitted by radioactive “labels” designed to stick to specific proteins. The patient receives an injection of these radioactive markers, which diffuse to the target proteins in the brain. The scan reveals a map of where exactly those proteins are.

Scientists quickly began using the PET method to study disorders like Alzheimer’s and schizophrenia, convincing Knudsen of its power for mental health studies. So her team organized a double-blind, randomized clinical trial in which healthy participants would receive a standard 20-milligram SSRI or a placebo daily. After three to five weeks, the team would collect PET scans of the synapses in each person’s neocortex and hippocampus. In this case, the labels were designed to stick to a protein at the connection between neurons. Tracing them would map out the brain’s synapses, allowing scientists to measure synaptic density.

Their hypothesis was simple: The participants who took the drug instead of the placebo would show more synaptic density. That hypothesis was wrong.

“At first glance, it seemed slightly disappointing,” Knudsen says. There was no significant difference between synapses in the drug and placebo cohorts. But an imperfection in the study became its life raft. For logistical reasons, each person’s PET scan varied from 24 to 35 days after their first drug dose. This introduced a new variable into the experiment—duration—and it let the researchers perform a new analysis.

“It was only when we started to look closer at the timing that we could see they had an increase,” Knudsen says. Participants who spent longer on the drug had more synapses than those who spent less time. And for those on the placebo, timing didn’t matter whatsoever. Knudsen thinks this means that these synaptic changes accumulate during the weeks it takes for SSRIs to ramp up.

Nord says that the Danish team’s biological explanation nicely complements the psychological theory that augmenting positive emotions has a cumulative effect on mood. “The two explanations are compatible,” says Nord, whose book The Balanced Brain: The Science of Mental Health was released in September. “They’re explaining it at different levels.”

“It’s a different perspective to what’s come before,” agrees Roiser. “It gives the additional weight to this idea that you need the cumulative changes over time in order to shift the environment to be more positive, which can then explain how people are then going to recover from depression.”

Neuroplasticity may be an antidote to the distressing recurring thoughts that are often present in depression. “It’s almost as if the brain is fixed in an unhealthy pattern that is reinforcing itself,” Knudsen says. If rumination reinforces negative thinking, then forming new connections offers a way out, she says, “like having a reset button that makes you think differently.”

But Mark Rasenick, a neuroscientist at the University of Illinois Chicago, hesitates to draw broad conclusions based on how the healthy individuals in Knudsen’s study responded to SSRIs. Antidepressants affect mood more in a depressed person, he says: “What do they do to healthy people? The answer is not much.”

Knudsen agrees that healthy participants may respond less to the neuroplastic effects than people diagnosed with clinical depression, and she says that the project’s next phase, including participants with depression, is ongoing.

Rasenick imagines a PET study of only depressed patients, all receiving the same SSRI for the first time. Some participants won’t benefit from the drug, so this setup could compare neuroplasticity in those who benefit versus those who don’t.

In 2016, Rasenick’s team proposed another biological explanation for why the effects of antidepressants lag, when they observed that SSRIs gradually accumulate in the membrane of certain brain cells in rats. They may not have an effect until they build up to a critical level. Based on a pilot study Rasenick published last year, this facet of SSRI action may one day allow doctors to use blood tests to quickly measure whether a patient is responding to the drugs. Still, Rasenick believes that neuroplasticity is an important factor too. “To have the evidence from living human brains is critical,” he says.

PET scans are becoming an unparalleled resource for measuring wiring in human brains. “It’s very rare to have the ability to do an experiment like this,” Nord says. “They’re giving us a quite unusual window into the processes that are happening in this treatment.” Knudsen’s team has also used them to investigate psilocybin’s effects, and another team has studied ketamine’s. “What this paper is really showing is that you can detect these new connections being formed,” Roiser adds.

The problem is, PET scans and radioactive labels cost researchers thousands of dollars per participant. (About $4,500 per scan in this study, according to Knudsen.) Yet the benefits might pay off if they improve treatment. Nearly one in five adults in the United States have been diagnosed with clinical depression, making it “a major contributor to mortality, morbidity, disability, and economic costs,” according to the CDC.

Rosier says this new study suggests it could be beneficial to speed up synapse formation, perhaps with an accelerant drug that could complement SSRIs. “One can imagine encouraging these neuroplastic changes during antidepressant treatment, maybe making them happen faster,” Roiser says. That could help the many people who spend months trying drugs to find the one that works. But there’s still much to uncover about why depression varies from person to person, and how to predict the best treatment. (A fast-acting antidepressant that acts on the neurotransmitter GABA instead of serotonin recently earned approval to treat postpartum depression, but not for general depression.)

Knudsen likens treating depression to treating a fever. Antibiotics can’t kill every kind of bacterial infection, and they do nothing if the fever is caused by a virus. So doctors have to know the exact cause of the fever if they want to be confident about the medication they give. Neuroscientists yearn for the same grasp of the biological causes of depression. “To expect that the same kind of medication would be helpful for all patients with depression is perhaps a bit naive,” Knudsen says. “It makes a whole lot of sense to rethink what depression really is and how it should be treated.”