While you sleep, specialized neurons in your brain help you forget

Forgetting has long been considered a passive process in the brain, but new research puts that idea to bed

Kamila Kourbanova

Neuroscience and Molecular Biology

Johns Hopkins University

While you sleep, the brain forgets. But, until recently it was not clear how the brain decides to forget.

Scientists analyze sleep by measuring the electrical activity of neurons near the outer layer of the brain. By quantifying brainwave changes, scientists have already determined sleep is not just one process. There are two basic types of sleep: REM (rapid eye movement) and non-REM, or NREM. In NREM, your heartbeat slows, your muscles relax, and your brainwaves fall into a constant rhythm producing slow sleep waves. In REM, voluntary body motion is paralyzed and the brain's activity suddenly jumps

We don't remember every detail of our lives: Our brains decide which events are important for long-term storage and which can be purged. So, how does the brain divide memories between long-term safekeeping and the garbage bin?

A collaboration between Japanese and US researchers has revealed an unheard-of method of the brain actively "forgetting" under the lens of REM sleep. The researchers, authors of a recent study led by Shuntaro Izawa and published in the journal Science, have spent years studying sleep and wakefulness. They examined a group of neurons that produce melanin concentrating hormone (MCH) near a pea-sized area in the brain called the hypothalamus, which produces various types of hormones, including those needed for sleep. MCH neurons are known to control sleep and appetite, but more recently they have been studied for their involvement in memory. 

Izawa and the other researchers first confirmed that MCH-producing neurons sent blocking signals into the hippocampus, a brain structure important for memory. Once they visualized the interaction between MCH neurons and the hippocampus, they had to test the memory of the mice. They did this with a test called Novel Object Recognition. In this test, a researcher introduces two identical objects (say, film canisters) to the mouse and allows it to become familiar with the objects. The next day, the mice are exposed to the same two items but this time, after ten minutes, one of the items is replaced with a new item (a Lego block, for example). If the mouse remembers the film canister, it will spend less time exploring it, instead exploring the new item. 

 Neuroscientists commonly use the Novel Object Recognition test to assess the memory of mice

CC0 on PeakPx 

Izawa and the other researchers tested the memories of mice with and without active MCH neurons. Surprisingly, the mice lacking MCH neurons had a considerable improvement in memory. These mice remembered objects more quickly and for longer periods of time. In contrast, when testing mice with MCH neurons turned on, their memories plummeted. These mice spent equal time sniffing and licking the old objects (which they had already been introduced to) as with the new objects. In other words, it seemed like the mice did not remember old objects at all. 

In the world of sleep science, MCH neurons have been quite popular due to their predominant activity in REM. Since REM is known to be important in memory consolidation, the researchers tried to see if memory retention (the period after learning something new but before long-term storage) was affected by manipulating MCH neurons during sleep. The researchers could precisely control when the MCH neurons were active in a sleeping mouse, effectively creating a time-specific "light switch," using lasers. 

The results were phenomenal: by temporarily “turning off” MCH neurons during the REM phase, mice showed significantly increased memory during the memory tests, while “turning off” the neurons while the mice were awake or in NREM had no effect on their memory. 

It's easy to think of forgetting as a passive process, where things slip through the cracks. The results of this study suggest that MCH neurons can substantially impair memory and prompt forgetting during REM. Understanding the mechanisms of forgetting during sleep may help answer some questions about memory degeneration in disorders like Alzheimer’s Disease. For example, how is it an Alzheimer’s patient is able to process and respond to new experiences but have trouble retaining them? Is this due to a neuronal “off” switch or does this entail multiple processes researchers are not fully aware of yet?

Your most vivid dreams probably happen during REM sleep

Sleep, particularly REM sleep, has many functions. For many of us, the dreams we would describe as deeply emotional and perceptually vivid happen during REM sleep. Scientists speculate dreaming may help us regulate our emotions better, process fears and trauma, and assist us in consolidating our memories and forgetting negative events. Studying memory and emotional modulation of dreaming can shed some light on disorders such as Post Traumatic Stress Syndrome as well as the neuronal plasticity involved.

Sleep is a regular biological phenomenon that we all experience. Research investigating the relationship between memory, dreaming, and emotions can help us understand the significance of sleep. The importance of these results points out the fact that little is known about the process of forgetting, despite the mass of knowledge we have about memory. This new pathway can help scientists clarify how and why we forget, and how sleep may help us in trashing the "extra" memories while keeping the useful ones.

Comment Peer Commentary

We ask other scientists from our Consortium to respond to articles with commentary from their expert perspective.

Sahana Sitaraman

Neuroscience and Behavior

National Centre for Biological Sciences, India

Really interesting article. I was wondering if MCH neurons show selectivity in which memories they erase. How many times would the subject need to be exposed to the same  stimulus before it becomes consolidated as a long term memory? And would that be mediated via some sort of feedback inhibition onto the MCH neurons? 

Kamila Kourbanova responds:

Very good questions, MCH neurons have been recently in the big spotlight for memory, particularly for object-recognition. The Novel Object Recognition test has been well-used paradigm in research for testing memory. As far as selectivity, I think it would be worth exploring what other tests are available for memory testing in animals, and they would have to answer specific questions, ie. What part of memory are they testing (encoding, storage, retrieval, all three)? Can we be sure they are testing specific memories? Are they repeatable? And of course, how translatable are they for human studies?

In this study, NOR testing shows the mice really only need 5-10  minutes to become familiar with an object and be able to recognize the object later after just a single exposure. We use the same test in our lab with different objects and we’ve found the more intricate the object  is (ie. variation in patterns and textures), the more time is spent exploring the object in the first exposure. The mice seem to be more interested in say, a falcon tube with different colored stripes taped onto it than a simple black film canister. You can explore the variation of this test in this review, but overall the conclusion is: mice will explore the novel objects more than objects they are already familiar with. The single exposure is enough to consolidate the memory into long-term storage as long as testing is done within a few days.

Regarding the feedback inhibition theory, I think a lot more work will need to be done to identify if MCH neurons have a role in not just “forgetting” but also actively “labeling” memories for long-term storage, or if perhaps there is a different mechanism MCH neurons rely on for this. Specifically, there needs to be a focus on which subpopulation could be involved with this. This paper focuses on MCH neurons projecting to the lateral hypothalamus, and theyfound bursts of activity in mice exploring novel objects suggesting  these neurons might have a role in encoding new experiences. Because of the number of vast projections MCH neurons have throughout the brain, it could be difficult to determine if indeed a direct feedback exists.

Dori Grijseels


University of Sussex

I really enjoyed reading this, very interesting  write-up! As a scientist who studies the hippocampus, I can confirm that  we spend a lot of time thinking about how to make memories, and not a lot of time thinking about how and why things are forgotting. I believe memory reconsolidation relies (at least partially) on reactivation of neurons that were active during the encoding of a memory. Do you know if the researchers have looked at whether the MCH neurons specifically impair this replay?

I think the point on Alzheimer’s disease is very interesting too. I assumed that neuronal death and impaired hippocampal function would be the main reason the memory function is impaired. How do you reckon MCH neurons might play into this?

Kamila Kourbanova responds:

Based on the past research that has been done on MCH neurons, it appears they actually assist in the encoding and reactivation process. This recent paper showed some pretty specific results on the silencing of these neurons in object memorization. Their findings indicate MCH neurons have to be active upon first exposure to the object in order for the mice to recognize the object in a second encounter; therefore the lack of activation impairs remembering.

In the Alzheimer’s brain: we know know NMDAR activity is essential for learning but amyloid-beta plaques cause synaptic depletion of NMDAR-dependent axons due to excessive stimulation by the toxic environment. I believe studies pinpointing these molecular process could help shed some light on how this specific dysfunction occurs. Although I do not have any expertise in electrophysiology, it’s hard not to admire the the depths this paper went into describing effects of a global MCH-R1 deletion on the hippocampus. They found a large decrease in the synaptic NMDAR and AMPAR currents; the authors suggest MCHR-1 activation may be partially  responsible for NMDAR synthesis. Perhaps future studies need an artillery of both behavioral and electrophysiological approaches in order to understand what mechanisms are at play here and how they  influence memory specifically.

Anisha Kalidindi

Molecular Biology

Ohio State University

Very interesting and well-written article! I find the biological basis of sleep to be quite fascinating as it is an integral process yet there are still many unknown aspects to it. Also, sleep disruption is integral to many memory disorders, thus  studies that explore its basic function are essential. Research  indicates that some memory consolidation happening during REM sleep, in addition to this data supporting forgetting occurring during REM sleep What are your thoughts on how research in the future will tease these  two processes (almost opposing processes) apart when studying memory during REM sleep? 

Kamila Kourbanova responds:

Sleep is a very important process indeed! There are so many different aspects of sleep and everyone seems to experience it differently. 

Studies in recent years have indicated that REM sleep may be more important for non-declarative memories (ie. the how, muscle memory), whereas NREM sleep may be more important for declarative memory consolidation (ie.  the what). But recently this idea has been contested and that memory consolidation, both declarative and non-declarative, happens across sleep phases. Interestingly, some memories are only NREM-specific. For example, researchers showed that stimuli presented during NREM sleep promoted participant learning of object-location associations while a specific odor (say bubble-gum  candy) was present in the room. The participants could not find the  correct location if the smell was presented during REM sleep. This study had an interesting take on memory… it took advantage of the participant’s unconscious state and thus eliminated some potential bias.

When teasing apart memory consolidation vs. forgetting in REM sleep, I believe researchers will have to focus on these specific questions: Which type of memories are consolidated during REM sleep and to what extent? Is there a dual-consolidation (both NREM and REM-mediated) process involved as the current literature might suggest? Is there a  limit to what kind of memories are “deleted” in REM sleep? Are there other mechanisms we haven’t discovered yet which allow NREM to aid in  deletion of other types of memories? The hardest part will be creating  tests that can answer these specific questions with the limitation of  animal research and/or working to create tests in humans which limit  bias and false positives.

Thiago Arzua


Medical College of Wisconsin

Awesome article, and so many good commentaries too! I’m particularly curious and excited for more research on how  active the process of forgetting is.

Most of the studies we conduct on animals have been focused on different kinds of memory, but all with easily quantifiable endpoints -  remembering a maze, or interacting with an object. While we tend to treat forgetting as simply not remembering, there are some super cool studies  on “forgetting cells” that are physically changing neurons in order to  change a memory, or erase it. I’m curious about the communication between these dopamine/forgetting neurons and these new MCH neurons.

Furthermore, related to what you said, it’d be super interesting to go with different molecular tools to see how these MCH neurons are possibly distributed in the human brain. Looking for instance at healthy vs Alzheimer’s post-mortem tissue.

I’d also be curious to see if there are any extracellular signals  related to these cells that we could pick up on. We know, for example, that there are changes in the cerebrospinal fluid (CSF) that happen during REM sleep, even that sleep might play a protective role in AD. So it would be interesting to see if the MCH neurons are putting out specific biomarkers related to their newfound ability that we could pick  up with a spinal tap (I do not volunteer for that).

Kamila Kourbanova responds:

It seems like forgetting may be a chemical and physical process! The review you linked above is actually a super interesting read! Who can define exactly what ARE “molecular and cellular memory traces”? It almost dips its toes into the “meta” side of neuroscience. Furthermore,  if you are interested in CSF fluid clearance during sleep, Jeffrey Illif’s lab  at the University of Washington has done some pretty amazing studies in this field. They look at not only the CSF clearance but also the involvement of the vascular system to assist in this clearance of toxic  particles not just limited to AD brains but in addition to brains affected by PTSD. The PTSD studies are particularly interesting because they change the viewpoint of memories as a substance that is in and of  itself actually harmful. What an interesting perspective! 

Burcin Ikiz


Great article! It is true that when we think about memory, we most often focus on what we remember and eventually what we learn, but not much on what we forget. Forgetting is a crucial part of the learning process.

As a neuroscientist with research experience in neurodegenerative diseases, I was especially intrigued by the connection you have made between MCH neurons and Alzheimer’s disease. Sleep problems, such as sleep apnea and mixing day/night cycles, are frequently observed in  Alzheimer’s patients. My guess is that it is due to the selective loss of cholinergic neurons in the cortex. It is possible, however, that as the disease progresses and more neurons die, their interactions with other neuronal cell types are also altered. I am curious if you have any opinions on that.

Kamila Kourbanova responds:

This is the big “IT” factor in AD research! Once the machinery changes, all the cascades are substantially altered. We are no longer studying a brain under normal functioning, the entire system is altered. I think it is possible that among all the cell types that are altered in the AD brain, MCH neurons (and other neurons assisting in forgetting) are affected. Another big question is, how are MCH neurons affected in an extremely toxic environment? Are they more susceptible to “malfunction”  in these environments than their counterparts? Or perhaps the cellular cascades they rely on for information are changed and therefore they themselves are changed. The authors of this study are indeed interested in how these cells are changed in AD. I’m sure the field of memory study will have their hands full in addressing these questions in further studies! Are these cells affected? And if so, how do they respond to the change in neuronal environment?