Capsid formation: ARC expression levels rise in dendrites about two hours after neurons are exposed to a chemical that artificially triggers long-term potentiation.

Neurons making memories shush their neighbors

When neurons strengthen their synapses, they “infect” surrounding cells with a virus-like protein to weaken those cells’ excitatory connections, according to a new preprint.

Our understanding of memory is often summed up by a well-worn mantra: Neurons that fire together wire together. Put another way, when two brain cells simultaneously send out an impulse, their synapses strengthen, whereas connections between less active neurons slowly diminish.

But there may be more to it, a new preprint suggests: To consolidate memories, synapses may also influence neighboring neurons by using a previously unknown means of communication.

When synapses strengthen, they release a virus-like particle that weakens the surrounding cells’ connections, the new work shows. This novel form of plasticity may aid memory by helping some synapses to shout above the background neuronal hubbub, the researchers say.

The mechanism involves the neuronal gene ARC, which is known to contribute to learning and memory and encodes a protein that assembles into virus-like capsids—protein shells that viruses use to package and spread their genetic material. ARC capsids enclose ARC messenger RNA and transfer it to nearby neurons, according to a 2018 study. This leads to an increase in ARC protein and, in turn, a decrease in the number of excitatory AMPA receptors at those cells’ synapses, the preprint shows.

“ARC has this crazy virus-like biology,” says Jason Shepherd, associate professor of neurobiology at the University of Utah, who led the 2018 study and the new work. But how ARC capsids form and eject from neurons was unclear, he says.

As it turns out, synaptic strengthening spurs ARC capsid release, according to the preprint. When neuronal connections strengthen, ARC capsids are packaged into vesicles, which then bubble out of neurons through their interactions with a protein called IRSp53. Surrounding cells absorb the vesicles containing ARC, which tamps down their synapses, the new work suggests.

“They have taken the story a long way,” says Clive Bramham, professor of biomedicine at the University of Bergen, who was not involved in the work. “It’s a lot of work, nicely put together.”

But the results—all from in vitro experiments—need to be replicated in animals before any firm conclusions can be drawn, Bramham adds. “This is in neuronal cultures, so we don’t really know to what extent this is going to apply in adult tissue. It may be that in cultured neurons, the release of vesicles is way more important.”


hepherd and his colleagues extracted brain tissue from young mice reared in a cage filled with toys, an environment known to boost plasticity and increase ARC expression. ARC proteins isolated from the tissue were directly bound to IRSp53, the team found.

In a petri dish, purified ARC proteins formed more capsids when IRSp53 was also present, suggesting that the interaction between the two proteins prompts capsid assembly. And IRSp53 helped to launch the capsids from active neurons: Silencing its expression in cultured mouse neurons led to fewer ARC-containing vesicles exiting the cells, the study found.

Further experiments revealed that long-term potentiation (LTP), the process by which synapses strengthen in response to neuronal activity, induces IRSp53 to assemble and release ARC capsids. Applying a chemical that artificially triggers LTP enhanced expression of IRSp53 and ARC in dendrites. And the dendrites formed fleeting dendritic branches containing both proteins, live-cell imaging showed. Soon after, ARC levels abruptly dropped, suggesting that the capsids had ejected into the extracellular fluid, Shepherd says.

Applying this fluid to cultured neurons decreased the number of AMPA receptors at synapses, the study further shows. AMPA receptors remained unchanged, though, when the team used extracellular fluid from neurons that lack the ARC gene.

The new work chips away at the question of what mechanisms underly plasticity, which unlike other biological pathways, remains unresolved, says Travis Thomson, assistant professor of neurobiology at UMass Chan Medical School, who was not involved in the study. “Finding a pathway and understanding how plasticity is regulated is the fundamental search that neurobiologists are trying to undertake.”


he findings suggest that plasticity occurs not only between interconnected neurons, but also among those in the same neighborhood. By spreading out to surrounding cells, ARC may help to cement memories by boosting the “signal-to-noise” of active neurons, Shepherd says. In essence, ARC is telling the surrounding cells to “shut up,” he adds, ensuring that only active synapses are incorporated into the memory circuit.

The results also point to the brain’s tendency to domesticate rogue genes for neuronal signaling. ARC originated about 400 million years ago from a retrotransposon, a type of freewheeling genetic element that pastes itself into genomes, according to Shephard’s 2018 study. And about 50 million years ago, the brains of flies co-opted a similar capsid-forming protein derived from a retrotransposon. “Because evolution has done this [type of repurposing] at least twice, it suggests there is something very useful about that biology for the nervous system,” Shepherd says.

In fact, the brain may express several other capsid-forming proteins that mediate neuronal communication, Shepherd says. A gene called PNMA2—another repurposed retrotransposon that is highly expressed in the brain—encodes a protein that also forms capsids, according to a study published in Cell in January by Shepherd’s group. Recycling “junk” DNA into useful neuronal genes might explain why we have so much of it, Thomson says.

Repeating the ARC plasticity experiments in vivo is the next challenge—one that involves a complicated set of labeling, imaging and electrophysiology experiments, Shepherd says. “Those are some tough experiments that will probably take five years or so.”