While our muscles pump iron, our cells pump out something else: molecules that help maintain a healthy brain.
But scientists have struggled to account for the well-known mental benefits of exercise, from counteracting depression and aging to fighting Alzheimer's and Parkinson's disease.
Now, a research team may have finally found a molecular link between a workout and a healthy brain.
Much exercise research focused on the parts of our body that did the heavy lifting.
Muscle cells ramped up production of a protein called FNDC5 during a workout. A fragment of this protein, known as irisin, got lopped off and released into the bloodstream, where it drove the formation of brown fat cells, thought to protect against diseases such as diabetes and obesity. (White fat cells were traditionally the villains.)
While studying the effects of FNDC5 in muscles, cellular biologist Bruce Spiegelman of Harvard Medical School in Boston happened upon some startling results: Mice that did not produce a so-called co-activator of FNDC5 production, known as PGC-1α, were hyperactive and had tiny holes in certain parts of their brains.
Other studies showed that FNDC5 and PGC-1α were present in the brain, not just the muscles, and that both might play a role in the development of neurons.
Spiegelman and his colleagues suspected that FNDC5 (and the irisin created from it) was responsible for exercise-induced benefits to the brain - in particular, increased levels of a crucial protein called brain-derived neurotrophic factor (BDNF), which was essential for maintaining healthy neurons and creating new ones.
These functions were crucial to staving off neurological diseases, including Alzheimer's and Parkinson's. And the link between exercise and BDNF was widely accepted.
"The phenomenon has been established over the course of, easily, the last decade," said neuroscientist Barbara Hempstead of Weill Cornell Medical College in New York City, who was not involved in the new work.
"It's just, we didn't understand the mechanism."
To sort out that mechanism, Spiegelman and his colleagues performed a series of experiments in living mice and cultured mouse brain cells.
First, they put mice on a 30-day endurance training regimen. They didn't have to coerce their subjects, because running is part of a mouse's natural foraging behaviour.
"It's harder to get them to lift weights," Spiegelman noted.
The mice with access to a running wheel ran the equivalent every night to a human running 5km.
Aside from physical differences between wheel-trained mice and sedentary ones - "they just look a little bit more like a couch potato," said co-author Christiane Wrann, also of Harvard Medical School, of the latter's plumper figures - the groups also showed neurological differences.
The runners had more FNDC5 in their hippocampus, an area of the brain responsible for learning and memory.
Using mouse brain cells developing in a dish, the group next showed that increasing the levels of the co-activator PGC-1α boosted FNDC5 production, which in turn drove BDNF genes to produce more of the vital neuron-forming BDNF protein.
They have reported these results online in Cell Metabolism.
Spiegelman said it was surprising to find that the molecular process in neurons mirrored what happened in muscles as we exercised.
"What was weird is the same pathway is induced in the brain," he said, "and as you know, with exercise, the brain does not move".
So how was the brain getting the signal to make BDNF? Some have theorised that neural activity during exercise (as we coordinate our body movements, for example) accounted for changes in the brain.
But it was also possible that factors outside the brain, like those proteins secreted from muscle cells, were the driving force.
To test whether irisin created elsewhere in the body could still drive BDNF production in the brain, the group injected a virus into the mouse's bloodstream that caused the liver to produce and secrete elevated levels of irisin.
They saw the same effect as in exercise: increased BDNF levels in the hippocampus.
This suggested that irisin could be capable of passing the blood-brain barrier, or that it regulated some other (unknown) molecule that crossed into the brain, Spiegelman said.
Hempstead called the findings "very exciting," and believed this research finally began to explain how exercise related to BDNF and other so-called neurotrophins that kept the brain healthy.
"I think it answers the question that most of us have posed in our own heads for many years."
The effect of liver-produced irisin on the brain was a "pretty cool and somewhat surprising finding," said Pontus Bostrom, a diabetes researcher at the Karolinska Institute in Sweden.
But Bostrom, who was among the first scientists to identify irisin in muscle tissue, said the work didn't answer a fundamental question: How much of exercise's BDNF-promoting effects came from irisin reaching the brain from muscle cells via the bloodstream, and how much was from irisin created in the brain?
Though the authors pointed out that other important regulator proteins likely played a role in driving BDNF and other brain-nourishing factors, they were focusing on the benefits of irisin and hoped to develop an injectable form of FNDC5 as a potential treatment for neurological diseases and to improve brain health with aging.
Do you believe Bigfoot exists?Related story: (See story)
The cost of losing nature