How long-term memories are made

How do we make and store memories? A team led by Nobel prize winner Susumu Tonegawa at the RIKEN-MIT Center for Neural Circuit Genetics in Cambridge, Massachusetts, has zeroed in on how memory circuits, or engrams, are uploaded for long-term storage in the outer layers of the brain.

Remarkably the details fit with the engram concept first espoused more than a century ago. But as the Science paper published today shows, there’s a twist. Rather than being uploaded over time, engrams appear to be uploaded straight away into the brain cortex. What takes time is the cementing of the memory.

Unpicking these details of memory storage opens the door to finding new ways to tweak memory either when it fails or becomes hyperactive. “In principle this study shows how we might treat these cells that become overactive in PTSD,” says Pankaj Sah, director of the Queensland Brain Institute, who was not involved with the study. “In some ways it’s a surprise to find these very complete memories can be so discrete.”

The first experimental evidence of how human memories are formed and stored goes back to just 1953. That’s when 27-year-old American Henry Molaison had his hippocampi removed to cure seizures. To the horror of his surgeons, the operation also destroyed his ability to make new memories. Yet his old memories were fine. The unintended experiment revealed the hippocampus is needed to weave new memories – particularly the context-rich “episodic” memories made every day, like what you saw when you walked your dog in the park this morning.

These detailed memories aren’t stored in the hippocampus, though. Over time they are transferred to the brain’s outer shell – the cortex. We know this from patients who, when these parts of their brain have been electrically stimulated, recall particular memories. The uploading of these memories generally involves compression of information, somewhat like the way we upload computer files for long-term storage. It was also believed to take place over several days.

This coarse-grained picture was how largely how things stood until five years ago. That’s when Tonegawa’s lab, a collaboration between Japan’s RIKEN Brain Science Institute and MIT, reduced a couple of near-mythical ideas to practice by using a state-of-the-art technology known as optogenetics.

One of the ideas was that of the engram, first conceptualised by German biologist Richard Semon in the early 1900s. A memory, he posited, would leave a physical trace in the brain; and the brain, when stimulated, would replay the memory.

Semon proposed this idea decades before researchers understood neurons sent signals via electrical impulses.

Researchers have since decoded much of the electrical signalling that passes between neurons; and shown how learning and memory correspond to the strengthening of connections, or synapses, between individual neurons.

Yet no-one had ever been able to  match a particular ensemble of neurons in the brain to a particular memory. In 1999 Francis Crick, another Nobel prize winner who turned his talents to unpacking the mysteries of the brain, mused that, to make progress, pulses of light might be employed to activate individual neurons in a living animal. “This seems rather far-fetched,” he wrote, “but it is conceivable that molecular biologists could engineer a particular cell type to be sensitive to light.”

Just six years later Stanford neuroscientists Edward  Boyden and Karl Deisseroth, much to their own surprise, made it a reality with their pioneering work in optogenetics. They co-opted a light switch used by green algae – the channelrhodopsin protein. When zapped by blue light, the protein opens a pore, allowing positively charged ions to flow across the cell membrane. This flow of current signals the flagella at the opposite end of the algal cell to beat, propelling it towards the light. 

Researchers found they could insert a single channelrhodopsin gene into individual neurons by using an infecting virus as the courier. They also ensured that only cells that had recently made a memory produced the light switch gene; memory-making cells produce a protein called c-fos, so the gene was engineered to only be made in cells producing c-fos.

In 2012, Tonegawa’s group used this optogenetic technique to demonstrate the existence of a fear engram. A mouse was placed in a box with distinctive wall patterns and floor textures. Whenever it was placed in that box, it received an electric shock. Subsequently just placing it in the shock box was enough to make it cringe.

The researchers also identified a group of cells in the hippocampus actively making the light switch, the smoking gun indicating those cells had been involved in making a memory. To prove that was the case, the scientists then threaded an optical fibre through the brain to the hippocampus to target these cells. When they zapped the hippocampus with rhythmic flashes of blue light, the mouse froze as if were reliving the memory of being placed in the shock box. It was the first evidence for an engram –  a collection of a few hundred cells that, when stimulated, replayed the memory.

In the current study, the researchers wanted to see what happened to the hippocampus engram in the mice over time. Other studies had suggested it was a particular small patch of the cortex – the prefrontal cortex – where fear memories appeared to be stored. So the researchers infected the cells of the prefrontal cortex with the virus containing the light switch. 

They found something curious. As before, once the mice learnt to fear the shock room, the memory could be replayed by directing flashes of light at the hippocampus. The surprise was the memory could also be provoked by flashing lights at the prefrontal cortex cells. So the engram, it seemed, was simultaneously uploaded  to the prefrontal cortex. “This was surprising,” notes Tonegawa, “because it indicated that the cortical memory was likely created on the very first day, and not gradually as has been assumed.”

However, when the mice were placed in the shock room, cringing at the memory, those same cells of the prefrontal cortex were silent (as evidenced by checking the chemical activity in isolated brain tissue). It was only a couple of weeks after the experience that the cells of the prefrontal cortex fired when the mouse was placed in the shock room. Conversely, the engram in the hippocampus began to fade.

So when it comes to long-term memory storage, first a silent copy is made in the prefrontal cortex; only gradually does it become cemented while the hippocampal engram is erased.  Just what that long-term cement is remains to be determined, says Takashi Kitamura, the first author of the paper.

Another key to cementing the memory was that the prefrontal cortex needed to get inputs from both the hippocampus and the amygdala, the emotional centre of the brain. When the researchers blocked neuronal inputs from either (again employing light switches), the cortex memory failed to cement.

How might this information help people? While we can’t implant light switches, it is nevertheless possible to switch particular regions of the brain on or off by implanting fine electrodes using a technique known as deep brain stimulation, already used to treat disorders like Parkinson’s disease. Kitamura imagines it will one day be possible to use a similar technique to manipulate the engrams in the brain. “But first we need to map them out in mice.” 

Given the breakneck speed at which this field is progressing, the era of manipulating our engrams might not be that far off.