“I alone have more memories than all mankind has probably had since the world has been the world,” said the title character in “Funes the Memorious,” a short story by Jorge Luis Borges (1). In the story, Funes falls from a horse and loses the ability to forget, although whether this is a blessing or a curse is never entirely clear.

“The brain has evolved to store information,” said Ronald Davis, a neuroscientist at the University of Florida Scripps Biomedical Research Institute. “But just as learning and memory are important, forgetting is also important. Too much storage can actually be a bad thing.”

Organisms do not benefit from storing information that is irrelevant or outdated. In fact, Davis and other researchers in the field hypothesize that forgetting extraneous details is actually crucial for the brain to function efficiently (2,3).

Although written decades before most research on the neurobiology of forgetting, the story of Funes illustrates this concept quite well.  The story’s narrator observes that, “with no effort, he had learned English, French, Portuguese, and Latin. I suspect, however, that he was not very capable of thought. To think is to forget differences, generalize, make abstractions. In the teeming world of Funes, there were only details, almost immediate in their presence.”

While scientists once thought that memories simply faded over time, the real story is more complex: the brain may have several mechanisms for actively pruning away memories. Although there have been decades of research on the mechanisms organisms use to acquire, stabilize, and access memories, the scientific understanding of the mechanisms that govern forgetting is still in its infancy.

Now, a handful of dedicated scientists are investigating the processes by which the brain forgets and even exploring whether remembering and forgetting can be manipulated to treat disorders that involve pathological memories, such as post-traumatic stress disorder (PTSD).

How flies forget

Like many scientists in the field, Davis spent years studying learning and memory. Then in the course of his research on scent memories in fruit flies, he and his team discovered something remarkable.

The researchers were exploring the role of dopamine-producing neurons that send signals to the mushroom body, a brain region that is crucial for olfactory learning and memory in the fly. In these experiments, they taught flies to associate a specific odor with a mild electric shock. Thereafter, flies that remembered this lesson avoided this odor.

“When we blocked dopamine transmission, the flies still forgot but at a very, very slow rate compared to normal flies,” said Davis. “Then we did the converse experiment, where we trained animals and then we activated those neurons in the living fly. And those animals forgot very, very quickly.” This held true even when the neurons were activated after the window of time that flies require to consolidate, or stabilize, a memory, supporting the idea that these neurons promote forgetting and actively removing information, rather than simply interfering with consolidation.

Just as the brain has multiple ways of learning information — multiple signaling pathways and so forth — there's going to be multiple ways to forget.
- Ronald Davis, University of Florida Scripps Biomedical Research Institute

“That told us this innervation of the mushroom bodies was doing something very profound with respect to flies’ behavior,” said Davis. “That was the beginning of my lab’s foray into the mechanisms underlying forgetting.”

A fruit fly has two mushroom bodies, one on each side of the brain, and each is innervated by twelve dopaminergic neurons. The researchers determined that just two of these neurons seemed to govern forgetting. Fittingly, Davis dubbed these “forgetting neurons,” although he noted that they may serve other functions as well.

By removing a tiny part of a fly’s exoskeleton on top of its head to expose the brain, the researchers were able to observe the activity of these forgetting neurons in real time. This window into the fly brain revealed that the forgetting neurons were chronically active, said Davis. Perhaps, he hypothesized, forgetting might be the default fate for information, and things are only remembered if another part of the brain judges them to be worthy of keeping and overrides the forgetting mechanism. He acknowledged that this hypothesis may be a bit extreme, and it’s unclear what this judging part of the brain would be, but he hopes to explore this concept in future experiments.

Davis has also worked on identifying some of the downstream processes triggered by this dopaminergic forgetting signal. Once dopamine is released from the forgetting cells, it binds to specialized receptors on mushroom body neurons. Davis discovered that a scaffolding protein called Scribble was crucial for passing along this forgetting signal inside the mushroom body neuron (4).

This work dovetailed nicely with discoveries made by Yi Zhong, a neuroscientist at Tsinghua University. In 2010, Zhong’s group discovered that a protein called Rac1 mediated forgetting in flies; more Rac1 activity meant quicker forgetting, while inhibiting Rac1 activity preserved memories for longer (5). Unbeknownst to them, Davis and Zhong had been working on the same pathway. Scribble was a link in the network that connects Davis’s forgetting cells and Zhong’s Rac1 (4).

“Just as the brain has multiple ways of learning information — multiple signaling pathways and so forth — there's going to be multiple ways to forget,” said Davis. “We're really just at the very, very beginning of getting insights into the mechanisms of forgetting.”

“Forgetting neurons” have not yet been identified in mice or humans, although this doesn’t necessarily mean that they don’t exist. The fly brain contains only about 100,000 neurons, while estimates put the mouse brain at around 70 million neurons (6,7). Larger and more complex brains may hold many more unknowns.

Furthermore, rodents and humans obviously do forget information. “There's every reason to believe that these mechanisms — or at least the general idea of these mechanisms — are conserved,” Davis said.

“In the human brain, it may not be dopamine that is the culprit like it is in flies. It may be another neuromodulator, like serotonin or norepinephrine. But the general principle of having this mechanism of chronic forgetting, some kind of signal onto cells that are encoding a memory, which we call engram cells, it hasn't been shown yet, but I think it will prove to be true,” said Davis.

Eternal sunshine of the rodent mind

Other researchers don’t only want to figure out how forgetting happens, they want to make it happen. 

Stephen Maren, a neuroscientist at Texas A&M University, studies the brain circuits underlying fearful memories. For the most part, the persistence of fear memories is beneficial; these memories help organisms survive in a dangerous world. However, Maren said, in disorders such as PTSD, “one could argue that the system has become dysfunctional and pathological.”

To suppress dysfunctional memories, “We have some behavioral strategies, such as extinction learning or other types of inhibitory learning, where we're trying to superimpose a safety memory on top of a fear memory. But the issue there is that the fear memory never goes away,” said Maren. These safety memories may fade with time, and they may not generalize to different contexts.

For example, if veterans with PTSD experience fear when they hear sounds similar to gunshots, therapists might expose them to this sound in a safe context, in the hopes that they will learn that this sound no longer associates with immediate danger, creating a safety memory. However, this safety memory may not persist in a different context, such as outside the therapist’s office, so if the person hears a similar sound somewhere else, perhaps fireworks at the park, the fear memory may dominate.

“The inhibitory process that therapy tries to establish to dampen the retrieval of fear memories tends not to work particularly well. It does provide relief for some individuals and can be lasting in some cases. But these traumatic memories are really stubborn,” said Maren.

“We hope to figure out if we can manipulate that memory, if we can delete it. Depending on your perspective, and depending on your ethics around manipulating minds, this is kind of the Holy Grail,” said Maren.

To delete the fear memory, researchers first needed to find where it was stored; they needed to find the engram, the physical memory trace within the brain. The term engram was first coined by Richard Semon, an evolutionary biologist and memory researcher in 1904 (8). However, the engram proved tricky to pin down until relatively recently, when researchers developed new techniques for labeling, activating, and silencing specific populations of neurons.

In 2007, neurobiologist Mark Mayford at the Scripps Research Institute (now at the University of California, San Diego) tagged neurons that were active during the creation of a fear memory in rodents and showed that some of these same neurons were active during memory retrieval, indicating that these brain cells may store this memory (9).

Two years later, neuroscientists Sheena Josselyn and Paul Frankland at the Hospital for Sick Children showed that they could selectively kill specific populations of neurons in the amygdala, a brain region important for emotion and memory, to erase a particular fear memory (10). Afterwards, the animals still created new fear memories, indicating that researchers hadn’t destroyed the capacity to create or store fear memories or respond appropriately in fearful situations; they had deleted the memory itself.

While this is a great proof of concept, said Maren, it’s probably not something that can be translated to humans. “I don't know that anyone would be on board with the idea of just killing cells in the brain,” he said.

We hope to figure out if we can manipulate that memory, if we can delete it. Depending on your perspective, and depending on your ethics around manipulating minds, this is kind of the Holy Grail. 
- Stephen Maren, Texas A&M University

Fortunately, killing cells is not the only way to disrupt memories. When an organism learns something, the information is consolidated for storage. When the organism actively accesses that memory, it becomes temporarily destabilized. At this stage, certain interventions such as protein synthesis inhibitors seem to prevent the memory from being restabilized and put back into storage, effectively deleting the memory in rodent studies (11).

Although he works in rodent models, Maren wants this work to eventually translate to human patients. One problem with rodent studies that attempt to abolish fear memories is that they largely use direct exposure by putting the animal back in the exact context in which the fear memory was acquired. When memories are reactivated in this way, scientists can interfere with reconsolidation to weaken or erase the memory.

This direct exposure isn’t an option for many patients with PTSD; therapists can’t bring soldiers back to a war zone for each therapy session, for example. Instead, said Maren, therapists often ask patients to imagine the traumatic incident or use trauma-related cues as reminders as a form of indirect exposure. Maren and his team wanted to see if these indirectly retrieved memories could be disrupted in the same way as directly retrieved memories in animal models. In other words, could they prevent reconsolidation of these indirectly retrieved fear memories to erase the memory itself?

In order to do this, they paired a negative stimulus, a small electric shock, with a sound that occurred after the shock, a process known as backward conditioning. Later, when the played this sound, the animals froze, which is an expression of the fear memory. This showed that they could indirectly reactivate this fear memory by playing this sound without directly exposing the animal to the context in which it received the shock (12). Researchers then attempted to manipulate the memory.

They labeled the neurons that were active during the indirect memory retrieval and used a special viral cocktail to express designer receptors on these neurons. These designer receptors are engineered to only respond to one drug and not to any of the brain’s natural neurotransmitters. The drug does not activate any of the brain’s naturally occurring receptors, so administering the drug should only activate the neurons with the designer receptors.  

Later, when researchers administered this drug to activate the designer receptors, the animals froze, indicating that this memory could be reactivated by manipulating this subset of neurons.  

To find out if they could weaken a memory, they injected a protein synthesis inhibitor into the hippocampus, which interferes with memory reconsolidation of directly retrieved memories after indirect memory retrieval. When they put the animals back into the original context where they had received the shocks, the animals should have been afraid. However, animals that had received the injections froze much less, showing that the fear memory had been weakened.

Weakening fear memories in a model that is likely more translatable to humans is an important step, but Maren said that this is only the beginning. Administering these protein synthesis inhibitors to the whole brain, or even into a certain area of the brain may have unintended consequences, so Maren wants to develop a more specific approach.

“Really, the next technological innovation will be figuring out how we can directly target these inhibitory molecules to the neurons and the synapses that hold the information,” said Maren. Maren noted that others have been able to inhibit specific neuronal populations to suppress fear memories, but this was only temporary. “We think we can piggyback on those methods to actually erase or interfere with the memory on a long-term basis.”

Maren noted that there is an important caveat. “Memories are highly associative,” he said. “It will be a challenge to see if we can really target and delete, erase, or manipulate a specific piece of information that might be actually woven together with lots and lots of other pieces of information.”

References

1. Borges, J. L. in Ficciones 107–115 (Grove Press, 1962).
2. Nader, K., Hardt, O. & Lanius, R. Memory as a new therapeutic target. Dialogues in Clinical Neuroscience  15, 475–486 (2013).
3. Davis, R. L. & Zhong, Y. The Biology of Forgetting-A Perspective. Neuron  95, 490–503 (2017).
4. Cervantes-Sandoval, I., Chakraborty, M., MacMullen, C. & Davis, R. L. Scribble Scaffolds a Signalosome for Active Forgetting. Neuron  90, 1230–1242 (2016).
5. Shuai, Y. et al. Forgetting is regulated through Rac activity in Drosophila. Cell  140, 579–589 (2010).
6. Zheng, Z. et al. A Complete Electron Microscopy Volume of the Brain of Adult Drosophila melanogaster. Cell  174, 730-743.e22 (2018).
7. Erö, C., Gewaltig, M.-O., Keller, D. & Markram, H. A Cell Atlas for the Mouse Brain. Frontiers in Neuroinformatics  12, (2018).
8. Josselyn, S. A., Köhler, S. & Frankland, P. W. Heroes of the Engram. J Neurosci  37, 4647–4657 (2017).
9. Reijmers, L. G., Perkins, B. L., Matsuo, N. & Mayford, M. Localization of a Stable Neural Correlate of Associative Memory. Science  317, 1230–1233 (2007).
10. Han, J.-H. et al. Selective erasure of a fear memory. Science  323, 1492–1496 (2009).
11. Josselyn, S., Köhler, S. and Frankland, P. Finding the engram. Nat Rev Neurosci  16, 521–534 (2015). 
12. Ressler, R. L., Goode, T. D., Kim, S., Ramanathan, K. R. & Maren, S. Covert capture and attenuation of a hippocampus-dependent fear memory. Nat Neurosci  24, 677–684 (2021).