the concept of a memory seems pretty intuitive to us it is the ability to store information about particular experiences and reconstruct them at later times through memory retrieval but where and how is this information stored inside the brain on a physical level in this video we are going to explore the idea of an engram a fundamental unit of physical memory substrate as well as biological and computational principles that govern memory formation and linkage if you're interested stay tuned everyone the study of memory can be traced back to early 20th century when a German biologist Richard
Simmons coined the term engram referring to last in physical changes in the brain that occur after learning or experience he envisioned that once formed the Ingram whatever it is becomes dormant and can be awakened by similar experiences or parts of the original event this reactivation is what we experience as memory recall but notice that the original phrasing in Salmon's definition is very vague it doesn't really tell us anything about the biological nature of these lasting changes as decades passed it became known that the brain consists of nerve cells that generate pulses of electricity encoding incoming
stimuli so memories must be stored as changes in patterns of how neurons talk to each other but before we dive deeper into what these changes are let's discuss how we can study memory formation in the first place to address what happens in the brain during memory encoding we need two crucial pieces a proper behavioral task and the means of experimentally monitoring the changes that occur in the brain it is only when you are working with human subjects you can explicitly ask them if they remember something or not but looking at the level of single cells
is only possible with modal organisms such as rodents and they for sure can't tell you the contents of what they remember to overcome this problem we can use a clever experimental design one of the most popular setups is called a fear conditioning paradigm there are many different variations but the main idea is that memory is defined as forming an association between the two stimuli and neutral stimulus also referred to as a conditional one for example a sound tone or a spatial context and an aversive stimulus or unconditional for example a mild food shock during training
the sound key and the shock are presented either simultaneously or with a short delay and the animal learns to associate previously neutral stimulus with pain the effects of such conditioning are usually tested on the next day when the mouse is presented with the same conditional stimulus if the mouse freezes in response to it and freezing is the type of a defense mechanism it means that the associative memory was successfully formed since the animal could retrieve that this sound leads to pain here I've used the sound queue as an example of conditional stimulus but it is
also possible to pair the foot shock with almost any type of sensation such as the particular spatial context or a smell great now we just need to find a way to experimentally see the memory formation now because this is a biologically complex process there are a lot of changes that occur in nerve cells from DNA modifications to the synthesis of new receptors just like any cells neurons contain DNA a genetic set of instructions of what proteins to make and how to make them it turns out that there is a very special class of genes called
immediate early genes that rapidly and selectively get activated in neurons that undergo plastic changes during learning these genes the two most common ones being Foss and Arc do a whole lot of different things on a molecular level regulating the amount of neurotransmitter receptors and inducing synaptic plasticity the details of which is not fully understood but for our purposes you can think of these genes as markers for memory encoding triggering the downstream Cascade of learning related changes but gene expression on itself is not something that we can easily observe and measure the idea is to functionally
link the activation of these immediate early genes to some kind of reporter process that could be addressed with our experimental toolbox for example it is possible to inject cells with a harmless virus that could genetically modify the neurons by spreading pieces of DNA containing a gene that encodes a fluorescent protein these DNA fragments also bear a special region called a promoter that basically controls whether the gene for the fluorescent protein is turned on or off importantly this promoter region is identical to the promoter of the phos gene controlling its expression in the intact cells as
a result the production of this protein and the expression of the false Gene become closely coupled so whenever the first Gene turns on to carry out its learning related changes the gene for the fluorescent protein also activates and we can visually see the neurons responsible for memory encoding because they are glowing under the microscope pretty cool right but there is a very big problem with this simple configuration I've just described namely it is impossible to carry out the fear conditioning right after the surgery since the animals have to recover and the virus needs to spread
around which can take up to several weeks but because it is impossible to tell when exactly the modification kicks in by the time you put an animal into the experimental apparatus it may have already formed a multitude of other memories for example while sitting in his home cage so under the microscope you'd see many glowing neurons marked for immediate early genes and it would be impossible to tell which of them correspond to the associative memory we are interested in so ideally we need to be able to control the timing of this tagging process and transiently
turn it on for the duration of the experiment to isolate the engram for its signal memory there exists a handful of approaches which I'm not going to describe in detail but on a high level it is possible to engineer this genetic Machinery to lie dormant and activate only upon the presence or absence of a certain chemical compound so in overall simplified setup might look like this at first we have a transgenic Mouse with inactive memory system so none of its neurons are glowing to perform a fear conditioning task we take them out put it into
the training environment administer the drug to turn on the tagging system and pair the foot shock to a sound inside the brain when this particular memory is formed some of the neurons undergo modifications controlled by the force Gene to encode this experience causing a subset of neurons involved in the membrane coding to express fluorescent proteins the effect of the drug quickly wears off so although the mouse certainly forms new memories as it's chilling in the home cage afterwards no new neurons are being tagged with the fluorescent proteins so we can now look under the microscope
and see the newly formed engram [Music] depending on additional genetic modifications it is possible to interrogate these memory coding neurons to uncover the properties of the memory trace the so-called tag and manipulate approach as we'll see shortly one of the first things that is compelling to test is what happens to these engram neurons during memory recall remember that according to Salmon's visionary definition engram can be reactivated by stimuli similar to their original experience indeed during the recall of this memory on the test day the same pool of engram cells become activated but does the activation
of these neurons actually cause the memory to be relived or is there activation merely a byproduct of recall well it turns out that if you selectively suppress their activity during test trial or even kill those tagged neurons the mouse won't show any signs of freezing when presented with the conditional stimulus importantly this disruption affects only the trained fear condition in memory and mice don't show any deficit in the recall of other memories and suppressing this same number of random neurons that are not tagged as engram cells doesn't affect the recall of the trained memory Animals
still freeze in response to the sound as before all this indicates that the activation of these specific neurons is necessary for memory recall so they are indeed responsible for storing that particular piece of information and if you selectively activate them in the absence of the conditional stimulus the mouse will freeze even though it's not presented with anything associated with the shock so the activation of engram neurons is not only necessary it is also sufficient to induce a memory recall notably similar results were obtained in other memory paradigms when mice learned to associate conditional stimuli with
rewards rather than pain suggesting that these principles of engram reactivation are not unique to fearful memories let's take a closer look at how engrams are formed in the first place the experience can generally activate a large number of neurons and there may be some background internal patterns of activity on top of that yet only a small proportion of these active neurons end up chosen to become a part of the engram so what determines which cells get to incur the memory is there some sort of selection process if so can we experimentally intervene with it to
control which neurons become allocated to the engram short answer to both of these questions is yes but let's first make a few observations consider the amygdala an emotional center of the brain which has been shown to play a key role in such fear conditioning tasks although the majority of amygdala neurons receive necessary sensory input and respond to both the food shock and the tone only between 10 and 20 of them become allocated to a given engram and just in comparison in the dentate gyrus a region of the hippocampus this number is much lower between two
and six percent in other words engrams are spores and this sparsity differs across brain regions perhaps a more exciting and counter-intuitive observation is that within one brain area the sparsity is highly conserved across different memories for example changing the strength of the stimulus in the conditioning task and even changing the memory content for example from Fear to reward doesn't affect the Ingram size it's compelling to think that more Vivid memories after a stronger shock would have larger engrams with more neurons encoding for it but that's just not the case this suggests that there must be
some internal mechanisms that keep the engram sparsity constant controlling the proportion of neurons that become allocated to storing each memory but why is that well there is a great amount of evidence that the brain implements a sparse distributed system for information coding and computations these representations with non-overlapping codes seem to be optimal having higher storage capacity and robustness to noise so the sparsity of the engram is a defining characteristic that must be kept within reasonable bounds one particular mechanism that ensures the constant engram size is the competition through neuronal excitability let's unpack what this means
as you probably know neurons are electrically excitable cells to send bits of information they generate brief boxes of electricity called Action potentials or spikes spikes are born when the voltage on the neurons membrane crosses a certain threshold the intrinsic excitability of neurons refers to their inherent ability to change membrane voltage and generate Action potentials in response to various stimuli it is like the Readiness of a neuron to Fire and transmit information imagine you have different cars each represented a neuron the acceleration of a car represents the excitability of a neuron which determines how quickly it
responds to stimuli car a has a low acceleration meaning it takes a lot of effort on the gas pedal to make it move faster similarly a neuron with low excitability requires a strong and sustained input to generate an action potential car B has high acceleration it quickly responds to even a light touch on the gas pedal rapidly increasing its speed similarly a neuron with high intrinsic excitability can generate Action potentials readily even with the weak input it turns out that from this pool of all eligible neurons the ones with elevated excitability have a higher probability
of being recruited into a memory trace this presents opportunities to control the very process of memory allocation experimentally since it is relatively simple to alter the excitability of neurons for example it is possible to genetically modify the cells with special ion channels that would open when you shine colored light on them and let positive ions in bringing the voltage closer to the threshold and thus making the neuron more excitable indeed if in this configuration we turn on the light to excite a certain sparse population of neurons during training then later same population of neurons would
be activated during memory recall and just like before it is possible to selectively activate them in the absence of sensory context to trigger freezing or inhibit their activity blocking this particular memory in such allocate and manipulate approaches we can control where exactly memory gets stored but let's get back to the concept of intrinsic excitability for a second more excitable neurons seem to indeed get preferably recruited to the memory Trace but what are the mechanisms of such competition apparently this gating is carried out by local micro secretary with the help of inhibitory neurons consider this circuit
where the principal neurons the ones that compete for the memory are also connected to interneurons that locally inhibit other neighboring principal neurons the potential competitors this ensures that the small pool of most excitable cells essentially indirectly suppress their neighbors there is indeed some experimental data supporting this View for example it has been shown that blocking inhibitory interneurons results in increased engram size okay great so far we have only looked at the ingrams formed in one particular rigid of the brain the lateral amygdala similar studies were conducted with other brain regions observing separate engrams for fearful
memories in hippocampus and cortical areas but there remains very little understanding of how these memory traces throughout the brain actually interact with each other are they separate memories or just small components of a larger distributed engram the paper published last year in nature Communications sought to answer this question they employed a novel technique called tissue clearing essentially making the neural tissue transparent which allowed them to image neurons in the entire brain using similar approaches to tag neurons with active plasticity genes and later activate or inhibit them the authors discovered that a single fear memory elicited
an engram that was distributed across a wide range of brain regions from areas that were known to hold memories such as the hippocampus and amygdala to a few surprising regions including Thalamus hypothalamus and even the brain stem this supports the idea of a so-called engram complex that memories are not localized in one specific region instead they are distributed across sparse in samples of neurons scattered throughout the brain this has led to the idea that different brain regions are likely to encode specific aspects of the full memory for example part of the engram in amygdala holds
information about emotional valence hippocampus is responsible for this spatial context and the cortex might encode the particular sensory experience of the food shock up until this point we have only talked about a single memory in isolation which is only one piece of a puzzle because the brain has to utilize stored information memories need to somehow be connected to each other in order to be later unified into something like an abstract concept or a principle so the question is how can different memories be linked the important thing to realize is that the very existence of a
link between two memories is also a piece of new behaviorally relevant information kind of a memory of itself and because it needs to be stored somewhere this information about the connection between the two memories also should have some sort of physical manifestation inside the brain so how can the brain not only store the contents of individual memories but also links between them well a very elegant solution is to encode the connection between memories as the degree of overlap between the populations of engram neurons let's see what this means remember we talked about how neuronal excitability
controls the process of memory allocation so that more excitable cells are more likely to become recruited to an engram but the excitability of a given neuron is not set in stone instead it constantly waxes and wanes throughout the lifetime of a cell these time Windows of elevated excitability last several hours before the sales excitability goes down if you think about it it makes perfect sense otherwise if the excitability levels of all the neurons were fixed the same small pool of cells would win the memory competition over and over again mixing memories together and drastically decreasing
information storage capacity for instance consider the experiment with two different memories being associations between food shock and two different sound tones when the two corresponding fear conditioning sessions occur closely in time with less than 6 hours in between neurons that were most excitable during the allocation of the first memory several hours later still retain their elevated levels of excitability and a large portion of them can out-compete the rest of the population during the allocation of this second experience as a result n grams for the two events become overlapping and functionally linked this means that if
you try to extinguish one memory for example by presenting sound a without the shock so that the mouse gradually dissociates them and stops freezing in response to the tone a it will also affect the other memory and the mouse will show decreased levels of freezing in response to the tone b as well you can see how the paired memories got extinguished together in contrast if during training the fear conditioning sessions are separated by 24 hours instead of 6. populations recruited to the two engrams are mostly non-overlapping and it is possible to extinguish one memory without
affecting the other this demonstrates that the larger the overlap between the populations of Ingram neurons the stronger the link between memories so that thinking of one experience automatically makes you think of the other notice that in this example the link is established only during the initial memory allocation when the overlapping pool of neurons is recruited but this offers very limited capabilities because the two events May not seem related at the time of the memory formation and the link may become apparent only later once the dots are connected so to speak it turns out that memories
can also be linked by co-retrieval when the two engrams that are initially non-overlapping become repeatedly reactivated together for instance researchers trained mice on two different tasks taster version where the animals learned to associate saccharine with the feeling of sickness and fear conditioning Paradigm pairing sound to the food shock essentially the two memories saccharine equals bad and sound equals shock were formed independently with four days between them resulting in two non-overlapping engrams after this animals were repeatedly presented with the two conditional stimuli saccharine and sound simultaneously thus reactivating the two engrams together as a result mice
started to freeze when they tasted saccharine which normally shouldn't happen analysis of the neural populations revealed a greater overlap of engrams after such simultaneous retrievals this core retrieval reorganizes memory traces and generates an assemble of neurons that are shared by both engrams intuitively it means that if some external cue repeatedly reactivates two engrams simultaneously it is advantageous to physically couple those memory traces inside the brain so that the next time retrieval of only one of them would read to a recall of the other interestingly this shared pool of neurons that emerges as a result of
the co-retrieval is not essential for storing individual memories if you selectively silence these neurons fear conditioning and tasterversion memories in isolation are recalled properly however I after silencing suckering fails to induce freezing this means that the emerging overlapping memory Trace holds information about the link between the two memories rather than the content of them alright let's try to tie everything together in this video we have seen how information about experiences is encoded as synaptic changes in sparser rates of neurons known as memory traces or engrams this parsity of engrams is highly conserved and is kept
at optimal values through a kind of a competition where most excitable cells are preferably chosen to be recruited to the memory Trace activation of these chosen neurons is both necessary and sufficient for memory recall and by manipulating the activity of these cells we can evoke memories delete them and even create a new memory in the absence of experience engrams are believed to be scattered throughout the brain encoding different aspects of each experience like pieces of a puzzle and multiple engrams can further become linked by sharing some of their neurons this linkage applied to a multitude
of different individual memories is likely to underlie abstraction and general principles of learning understanding the intricate workings of engrams is crucial not only for unraveling the mysteries of memory but also for unlocking the potential of our own cognitive abilities this is where our today's sponsor relent.org comes into play brilliant is an Innovative educational platform that can help you advance in stem fields in a fun and engaging way with an emphasis on Active Learning their courses contain a wide range of problem-solving exercises and beautiful interactive visualizations so instead of Simply memorizing facts you get to develop
a deep understanding of the concepts and apply them in real world situations and the courses are fully self-paced so you can learn in bite-sized chunks just by dedicating 15 minutes a day brilliant offers over 80 courses covering a variety of topics in math physics computer science and more whether you're a student a professional or just someone who loves to learn they have something for everyone for example if you enjoyed this video you may be interested in brilliant's newly released course on algorithms and data structures which explores how memory is organized and allocated inside computers don't
miss out on this opportunity and take a learning to the next level go to brilliant.org artem carsonov to get a 30-day free trial of everything brilliant has to offer and the first 200 people to use this link will get 20 off the premium membership if you enjoyed this video press the like button share it with your friends and colleagues subscribe to the channel if you haven't already and consider supporting me on patreon to suggest video topics and enjoy the bonus content stay tuned for more interesting topics coming up goodbye and thank you for the interest
in the brain foreign [Music]
