Building Blocks of Memory in the Brain

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]