Eric Kandel, Nobel Prize in Physiology or Medicine 2000: Nobel Prize lecture

thank you sten for those marvelous comments i also want to thank the nobel assembly for honoring me in this marvelous way

and for giving me the opportunity to share this honor with paul and with abit whose work i enormously admire i've been told that the opportunity of giving a nobel lecture

is one of the high points of the visit to stockholm as part of the nobel prize but i must say seeing people hang out over the balconies is something that even i was

not prepared for this is really just a marvelous moment for me may i have the first slide please as the title of my talk which then

introduced you to indicates the aim of my research over the years has been to develop a reductionist approach to learning and memory that would allow me to explore the underlying mechanisms in

molecular terms learning as you well know is the process by which we acquire new knowledge about the world and memories the process by which we retain that knowledge over time

for me learning and memory have proven to be endlessly fascinating mental processes because they address one of the most remarkable aspects of human behavior our ability to acquire new ideas from

experience most of the ideas we have about the world and our civilizations we have learned so that in good measure we are who we are because of what we've learned

and what we remember conversely many psychological and emotional problems are thought to result at least in part from experience and specific disorders of learning and

distort disturbances of memory haunt the developing infant as well as the mature adult down syndrome the normal weakening of memory with age and the devastation of

alzheimer's disease are only the more familiar examples of a large number of diseases that affect memory for a biologist interested in the mind

the study of learning has the further appeal that unlike thought language and consciousness learning is the mental process that is most accessible to molecular analysis

elementary forms of learning and memory have been well characterized by classical psychology since the first half of the 20th century and they represent

the most clearly delineated and for the experiment the most easily controlled of any mental process i initially became interested in the study of memory while an undergraduate

at harvard college in 1950 motivated by a long-standing interest in psychoanalysis but as i became immersed in biology doing medical training i began to find

the psychoanalytic approach limiting because it tended to treat the brain the organ that generates behavior as a black box in the mid 1950s while in medical school

i began to appreciate that during my generation the black box of the brain would in fact be opened and gradually demystified i realized that the problems of memory

storage once the exclusive concern of psychologists and psychoanalysts could become approachable with the methods of modern biology as a result my interest in memory

shifted from a psychoanalytic to a biological approach i spent three years 1957 to 1960 as a postdoctoral fellow at the national institutes of health in bethesda

learning more about the biology of the brain had the vague hope that with time i might contribute to translating some of the central unresolved questions in the psychology of learning and memory into

the empirical language of biology i was interested in knowing what sort of changes does learning produce in the neural networks of the brain

how is memory initially stored and once stored how is memory maintained what are the molecular steps whereby a transient short-term memory is converted

into an enduring self-maintained long-term memory my purpose in attempting this translation was not to replace psychological or psychoanalytic thinking

with the logic of molecular biology but to contribute to a new synthesis that would do justice to the interplay between the mentalistic psychology of memory storage and the molecular biology

of signaling with time i developed the further hope that the biological analysis of memory mechanisms might in themselves reveal new aspects of neuronal signaling

indeed this is proven to be true time and again the study of memory has exposed us to important new aspects of biological phenomena at first thought somebody interested in

learning in memory might be tempted to tackle the problem in its most complex and interesting form this was the orientation that my colleague alden spencer and i originally

had in 1958 one of the start of our scientific careers we joined forces at the nih to study the cellular properties of the hippocampus that part of the mammalian

brain thought to be most directly involved in complex aspects of memory despite the good start to be able to make in the cellular analysis of hippocampal neurons it seemed to it soon

became clear to us that to understand how the cells of the hippocampus participate in organismic behavior was a formidable challenge because the mammalian brain has a large

number of neurons and immense variety of interconnections it seemed unlikely that one could work out how sensory information about learning reached the hippocampus and how learned information

processed by the hippocampus might influence behavioral motor output i therefore became convinced that if one were to bring the power of modern biology to bear on the study of learning

it would be necessary to take a very different approach a radically reductionist approach were needed to study not the most complex case but the simplest case of

memory storage and to study them in the simplest and technically most tractable systems available were needed to develop experimental systems in which a simple behavioral act

that was modifiable by learning was controlled by a small number of large and accessible nerve cells so that one could relate the animal's overt behavioral

to molecular events that occurred in nerve cells that control the behavior such a reductionist approach based upon the selection of technically congenial

systems is traditional within biology but when it came to mental processes such as learning and memory many investigators were reluctant to consider reductionist approach

however from the outset i believe that a mechanism of memory storage will likely to be evolutionary conserved and that in molecular analysis of learning no matter how simple the animal of the task was

likely to reveal mechanisms that would be of general relevance and so after an extensive search i focused in on the blizzard californica

which you can tell it at moments glance is not only a very beautiful animal but exactly the sort of animal that you would select for the study of learning and memory

for the structural biologists i should simply point out that this is the head of the animal and this is the tail of the animal this animal is not only very beautiful

[Applause] but it is highly intelligent and exceedingly accomplished now the remarkable thing about its

accomplishments is it has done this with a rather simple nervous system your brain consists of 10 to the 12 neurons a million million cells by

contrast the plane of the brain of a plasmid contains about 20 000 nerve cells these are clustered in in groups called ganglia each containing about 2

000 nerve cells a single ganglion like the abdominal ganglion i'm going to tell you about controls not one behavior but a set of behaviors so the number of cells committed to simple behavioral

acts can be quite small 100 even less i will also point out to you later that not only are there few cells but they're really gigantic so given the fact that one has a

tractable nervous system how does one go about than thinking about developing a cell a molecular biological approach that was clear from the beginning and these are discussions i had with alden spencer and the irving cooperman that

were needed to meet a number of criteria in order to develop an intellectually satisfying approach to learning and memory and four criteria particularly important

one is one needed to delineate a behavior in this simple organism that could be modified by learning we needed to do extensive behavioral analysis two of them were needed to define in

cellular detail the neural circuit that mediated the behavior that was being modified by learning third we needed to locate within that neural circuit the critical sides that were modified by learning and having

then located it one could bring the tools of modern cell and molecular biology to bear upon the analysis of the mechanisms that mediated the learning and that stored the memory so i'm going to follow this outline and

let me begin with the delineation of the behavior so this is a plesia we used this simple animal and continued with the reductionist approach selecting the simplest behavior that the simple

animal could generate so the animal has a respiratory organ called the gill which is normally covered by a sheet of skin called the mantle shelf which ends in a fleshy

spout called the siphon if you apply a weak tactile stimulus to the siphon you get a brisk withdrawal of the gill this simple defensive reflex is like the withdrawal of a hand from a hot object

and the surprising thing we found from the very beginning is this elementary behavior in this elementary organism can be modified by a number of different learning processes it can be modified by

habituation by sensitization and by classical conditioning and what was even more interesting was the fact that not only could it be modified but the modifications resemble

more complex forms of learning in more complicated organisms which i will tell you about later on for example with each of these forms of learning there was a short-term memory that

lasted minutes and a long-term memory that lasted days weeks or even longer number one number two in each of these cases practice made perfect you

transferred from short-term to long-term information by repetition and three in each of these cases long-term memory differed fundamentally from short-term memory and requiring the

synthesis of new proteins something which had been characterized for complex forms of learning in mammals this made one thing at the very beginning that if one could define some of the proteins that are critically

involved in the switch from short term to long term for any form of learning what might be defining proteins that have general importance so with this idea in mind we focused on one particular form of learning called

sensitization which i will now describe to you sensitization is a form of learned fear in which the animal learns about the properties of an aversive stimulus so if you give a noxious stimulus

to the animal's tail it recognizes the stimulus as being unpleasant as being aversive and it learns to enhance its reflex responses so if you previously gave a weak tactile stimulus that

produced a modest withdrawal now give a noxious stimulus to the tail the same weak tactile stimulus will produce a more powerful withdrawal and the animal will remember this

aversive event as a function of number of repetitions if you give it a single training trial it will remember it for minutes and this does not require new protein synthesis if you give five training trials it'll remember it for

days and this requires new protein synthesis if you give further training it'll remember it for weeks and this of course also requires protein synthesis i want to focus on the difference

between one training trial a short-term memory and five training trials minimum long-term memory that requires new protein synthesis to look at the relationship between them so the next thing we want to do is sort

of delineate the neural circuit for this behavior and our initial step was to show that the abdominal ganglion was the critical mediator of the behavior

now the abdominal ganglion which i show illustrate here shares features with other uh ganglionic plesia and the most remarkable feature is that the neurons are not simply large they're gigantic

a cell like r2 is a millimeter in diameter prior to my presbyopia days i could recognize it with my naked eye some of you might still be able to do it

these cells are of course large enough and jimmy schwartz was the first one to show this that they can be dissected out by hand you could develop biochemical approaches to transmit a biochemistry in single cells you could generate cdna

libraries from single cells as rich and axis to do later on you could do all kinds of molecular manipulations that are difficult to do in other animals we soon recognize that not only were the

cells gigantic the largest cells in the animal kingdom nerve cells in the animal kingdom but they're also characteristic in their pigmentation and in their location so we could

recognize them as unique individuals and we could give them names like paul and minus but i was not created enough to use those names we used prosaic names like r2 and l2

but one could identify many cells in the ganglion in fact in the whole nervous system as being absolutely unique moreover after while we realized that we could not only

realized that these cells were unique and returned to the same cell in every animal of the species both naive and trained but we also realized that we could map connections between cells and

after a while we can map connections between cells and the sensory and motor periphery so for example if you stimulated one of the six cells that we identified to be motor neurons to the

gill just stimulating a single cell by itself produced a detectable movement of the gill we can then identify sensory neurons

that innervated the siphon skin and when we stimulated them with a single action potential they produced a nice fast juicy synaptic potential glutamate mediated of the kind that paul

spoke about in the motor neurons so if you now stimulated a sensory neuron repetitively it's occurred in the behavior it would fire the motor neurons and it would cause a guilt contraction so not only

could we work out a neural circuit in terms of specific identified cells but these identified cells had significant controls over behavior and that's because there was such a limited number

of cells involved and in this way we're able to work out the neural circuit of the behavior which i indicate here in very simplified terms so if you stimulate the siphon

skin you activate 24 sensory neurons that make direct connections to motor neurons and the motor neurons the six motor neurons make direct connections to the gill the sensory neurons also excite

inhibitory and excitatory interneurons that modulate the firing of the motor neurons so the first thing that struck vincent castellucci and irv cookford and myself as we looked at this neural circuit was

its invariance not only were the cells invariant but the connections were invariant certain cells only connected to some cells and not the others and this of course posed the first

interesting question in the study of learning in memory how do you reconcile the invariance of the connections in fact the high specificity one sees in

the brain in general with the modifiability of behavior and karl ashley and many people are worried about this in fact the challenge first posed the problem at the beginning of the century

and by the time we came along there were really a lot of confused thinking going on about the mechanisms of memory storage for example carl ashley had suggested along the lines that paul greengard had earlier developed that

there were electrical fields that were generated by learning that somehow modified behavior so we thought we'd just explore it empirically we would produce learning

and see what happens and so we looked at various forms of learning examining the neural circuit of the reflex we found that in every case what learning did was alter the strength of pre-existing

synaptic connections in the brain in some forms of learning this was enhanced in other forms of learning this was the synaptic connections were decreased

in strength moreover we found that the the persistence of the change of the synaptic change was the mechanism whereby memory was stored and i

show this as an example using sensitization but as a general principle this has held up almost every form of learning that has been explored in any detail so genetic and developmental processes give you the precision of

interconnections between nerve cells what they do not specify what they do not give you is the exact strength of connections what learning does is to alter the strength of these connections

let me illustrate what happens with the case of sensitization if you stimulate the tail you activate three different modulatory systems in the brain very much like the modulatory systems in the

base of your brain that you heard about from avid carlson the serotonergic pathway is particularly important this pathway ends in the sensory neurons including on the presynaptic terminals

and acts there to strengthen the synaptic connections by enhancing transmitter release from the presynaptic terminal if you give one tail shock you get a transient enhancement of transmitter

release the last four minutes this does not require new protein synthesis but if you give five training trials you get a persistent enhancement of synaptic transmission that lasts for days and

this does require new protein synthesis now the modulatory pathways produce changes at other points within the neural circuit but i want to focus on this one for several reasons one is that this is an important locus which

mediates a significant part of the memory storage number one number two it has a representation of both the short-term and long-term memory in fact one of the interesting things

that has emerged from this study that again is proven to be quite general is that the same synapse can mediate both short and long-term memory and finally we were able to find with sam schacker

that you can reconstitute this component of the neural circuit in dissociated cell culture you can take a single sensory neuron a single motor neuron and a single serotonergic cell put it in culture they form perfectly good

connections with one another you don't even need the serotonergic cell you know it releases serotonin you can just puff on serotonin so if you puff on serotonin once you get a transient enhancement

that lasts for minutes this does not require new protein synthesis if you puff on serotonin disconnection five times you get a more persistent enhancement that lasts for several days

this does require new protein synthesis so i want to look with you at the conversion of short-term to long-term because this illustrates in paradigmatic form what we want to understand how do you set up the

short-term process and how do you convert it to the long-term process i want to simply to emphasize to you in order to connect this with what avid and paul talked about we're looking here at

rapid synaptic transmission glutamate mediated between the sensory neurons and the motor neurons and the motor neurons in the gill on the other hand the serotonergic pathways i will show you in a moment

produce a slow synaptic action which modulates this rapid synaptic action so we're going to look at a blow-up of the connections between the sensory neuron and the motor neuron in order to

consider how the short-term process is set up and how it's converted to the long-term process this is the sensory neuron this is the motor neuron this is the tail and i'm only showing you the case of the

serotonergic facilitators but the others work the same way the search energy facilitators engage as seven transmembrane spanning receptor which on activation activates an

adenocyclase which increases the level of cyclic amp in the sensory neurons and as paul pointed out to you the cyclic gamep level acts the increasing cyclic gameplay

cyclic amp dependent protein kinase and this is a very interesting enzyme that has two regulatory subunits these spindle-shaped structures and two catalytic subs

subunits the oval-shaped structures the regulatory subunits normally inhibit the catalytic subunit a point i want to come back to in a moment when the level of cyclic amp rises the

regulatory subunit binds the cyclic mp undergoes a conformational change frees the catalytic subunit and the catalytic subunit phosphorylates substrates in the cytoplasm ion channels

and machinery for exocytic release of transmitter and leads to a trans enhancement of transmitter release glutamate police from the pre-synaptic terminal together with vincent castellucci we

were able to show that if you just inject cyclic amp you could simulate this in a very nice series of experiments with paul and angus nerd we showed that cyclic mp produces all of

its action through the catalytic subunit if you inject the catalytic subunit by itself you simulate the short-term process with a transient increase in cyclic amp

as you see with one tail shock or one pulse of serotonin the level of cyclic amp rises only transiently and only a small amount of the catalytic subunit

translocates from the cytoplasm to the nucleus but with rajatian we were able to show that if you give repeated training five pulses of serotonin the regulatory subunit stays off the

catalytic subunit for a longer period of time because as jimmy schwartz showed the cyclic gamep level rises for a longer period of time and that allows the catalytic subunit to translocate to

the nucleus and in so doing it also recruits the map kinase and this is the first time we got any insight into why repeated training is necessary for long-term memory one of

the reasons is necessary it allows the translocation to the nucleus of the appropriate kinases necessary to set up the long-term process the setting up of the long-term process

involves three steps an initiation step a consolidation step and a stabilization step i want to say a word about each of them the initiation step is an extremely

interesting and nuanced step because it turns out that what you need to do when you translocate to the nucleus is not only activate an activator of transcription you have to remove a

repressor in order to trigger the long-term process you have to activate a transcription factor called kreb 1 cyclic amp response element binding

protein because it binds to sequence in dna called the cre the cyclic amp response element but this but the cyclic amp response element binding protein

kreb is normally inhibited by a repressor kreb 2.

and in order to activate the activator you have to get rid of the repressor and this is the first time we realized that there are in fact inhibitory constraints on long-term memory

when you think of how difficult it is to put information in long-term memory you realize the actual elements in opposition to putting information into long-term memory you

the ease with which you put things into long term not only varies from period to period but really varies from time to time during the course of the day and we think that one

of the reasons this is so is there are a number of inhibitory constraints in which i'm just showing you the first if you remove this inhibitory constraint you have both in a plasia and in flies

where this has been looked at flash bulb memories where a single training trial will immediately give you a long-term memory this really has a lot of significance because it explains to you

why you can forget things so easily why you don't put it in the long term and it also tells you you know why some people remember things so well i have a friend

steve segelbaum siegel bomb remembers everything i used to think he was just a bright kid at columbia and now i'm beginning to worry that he might be a mutant he might

he might have a specific defect in kreb 2 that allows him to put these things into long-term memory so readily once you free kreb 1 of its inhibitory

constraints you activate a number of immediate response genes of which i'm going to only going to focus on two ubiquitous hydrolase and cbp

first ubiquitous hydrolase jimmy schwartz first showed that ubiquinone hydrolase binds to the ubiquitin proteasome and leads to the cleavage of the regulatory subunit which he shows

establishes persistent kinase activity so the ubiquitin proteolysis cleaves the regulatory subunit and removes a second inhibitory constraint it now feeds the

catalytic subunit it allows it to phosphorylate substrates in the cytoplasm but now without requiring any signaling it does no longer requires either serotonin or cyclic amp

so this is if you will the simplest case of long-term memory excuse me you activate in a transcriptionally dependent fashion

a second messenger kinase is activated by the short term process keep it going now for long term process without requiring any further signaling we have found this carries the memory

for the first 10 to 12 hours what gives the memory its persistence is the fact that cbp acting by itself and together with another transcription factor i'm not going to describe for you gives rise

to the growth of new synaptic connections a typical sensory neuron has in the basal state about 1200 synaptic connections after long-term sensitization training craig bailey

showed he goes to about 26 to 2800 synaptic connections so if you leave this beautiful amphitheater remembering anything from what paul avitz told you it is because you will leave here with a

somewhat different brain that you walked into this room with and that's because anatomical changes have occurred in your brain this general sequence this core signaling sequence

has turned out to have a certain generality almost identical sequence of steps have been delineated independently in drosophila using a completely different approach based on genes and behavioral

analysis and initial analysis from the honeybee suggested a similar kind of process is present there but if you think about it the finding that a transcriptional mechanism

is involved in long term solves one problem for you but it poses another it solves the problem because it gives you the initial insight as to why generally new protein synthesis is necessary for

long-term but not short-term memory new protein synthesis interfering with new protein synthesis blocks the expression of these immediate response genes

but it poses a very deep and fascinating problem in the cell biology of memory that i want you to think about for a moment having a transcriptional mechanism for long term means that now

there's ready communication between the nucleus and the synapse what does that mean does that mean the unit for long-term information storage is in

fact the whole neuron and not the single cell this is not an academic question because simply to remind you as paul pointed out a single cell in the central nervous

system of vertebrates makes not one but a thousand connections and a lot of those connections can be on different target cells so the question comes up having a transcriptional mechanism

can one reconcile that with synapse restriction as one has in the short term where individual synapses can be regulated or does the long term process commit each neuron

to responding as a whole so every single synaptic connection is modulated in the same way to address that question kelsey martin in the lab worked at a new culture system

rather than culturing a single sensory neuron with a single motor neuron she took bifurcated sensory neurons and cultured them with two symmetrical motor

neurons now you could puff on serotonin on one set of uh of terminals without in any way affecting the cell body or the other set of terminals

and we found that if you applied a single pulse of serotonin you produced a transient facilitation at this set of terminals synapse restricted that lasted for minutes and

did not require new protein synthesis there was no facilitation here if you now applied five pulses of serotonin you produced a persistent facilitation that lasted for days but

that again was synapse restricted there was no facilitation here this facilitation was associated with a increase in number of synaptic varicosities with growth of new synaptic

connections there was a doubling in the number of synaptic connections and it required transcription it involves creme-mediated transcription you could block it by injecting anti-crab

antibody into the sensory neuron cell body so this raised the question how does this come about clearly the five pulses are somehow sending a signal back to the

nucleus to activate kreb all the proteins then selectively targeted to only this set of terminals to give you synapse restriction or all

proteins sent to both sets of terminals but only those terminals that have been marked in some way can utilize those proteins productively to give rise to the growth of new synaptic connections

so we tested the second hypothesis whether a marking signal is necessary in order to capture the proteins and we thought that perhaps one or two pulses

of serotonin might be adequate to do this so we test it again in a minimalist way applying just a single pulse of serotonin in the experiments that's illustrated on the next slide

so if you apply five pulses of serotonin to one set of terminals you produce a synapse specific facilitation that lasts for days this is 72 hours associated

with a doubling of synaptic connections and crab mediated if you apply a single pulse of serotonin to the other branch just before or just

after there's a restricted time window you before after you give the five pulses you can capture the long-term process for the other set of synaptic connections this is about half the

amplitude and instead of having a doubling you have a 30 to 40 increase in number of synapses but it has exactly the same time force

so this was really a fascinating result because it indicated that the short term process has really two very different functions

acting by itself and acting in conjunction with the long-term process acting by itself it produces a short-term synaptic facilitation that

contributes to short-term memory storage but acting in conjunction with a long-term process in any part of the synaptic tree it marks that synapse so it can utilize proteins coming down from

all terminals in a way so it could give rise to a new synaptic terminal which other parts of the neuron which do not receive this mocking signal cannot do they

cannot use those proteins in such a productive way so this obviously has raised the question what is the nature of the marking signal and we have spent really a lot of time analyzing that recently

and i will simply outline to you our results we have found there are two components to the marking signal one is covalent modification serotonin

activating cyclic game p and pka mediating that covalent mark that captures the signal and allows you to set up the long-term facilitation but interestingly we found that to

stabilize that mark requires local protein synthesis and i want to remind you and one of the reasons i drew this dendritic spine here

is to remind you that input into neurons as paul earlier indicated comes on to dendritic spines and we have known for years that denveritic spines have in them

the machinery for local protein synthesis they have ribosomes they have messenger rnas and they synthesize proteins but there's been no idea whatsoever of what the function of that

local protein synthesis is we've now provided direct evidence that one of the functions of local protein synthesis is to stabilize the mark so that the growth

of synaptic connections persists if you inhibit protein synthesis locally the processes grow out but they retract after about 24 to 36 hours

we've now dissected out isolated process of the sensory neurons generated cdna libraries from just that process so for the first

time we have all the messages in that compartment we've sequenced those sequences and we found a very interesting feature that i just want to emphasize for one second we

found that a number of those messages have a special mechanism for translation which is sensitive to the inhibitor rapamycin these are messages that have been shown in other contexts to be recruited for growth

and if you use rapamycin you in fact block this protein synthesis dependent component of stabilization so not only do we have some insight into the importance of local

protein synthesis in the stabilization of the mark but we've shown that there's a specific sub-category of translational mechanism that is particularly important

now the ability to mark synapses on the one hand and to activate transcription on the other really sets up a new mechanism of signaling within neurons long-range

signaling which has significant implications for the integrative action of the neuron but i've considered with you only the simplest case i've shown you that facilitatory input can set up

specific facilitation and that can be captured but clearly you want to make sure that you don't simply grow additional connections you want to be able to prune them in some ways so you want to know

what happens to inhibition can inhibition also be synapse specific can it be captured can it prune connections that have been enhanced by facilitatory input

well a number of years ago jimmy schwartz steve siegelbaum and i found that the peptide for maphamid related to encephalin working through arachidonic acid can produce the mirror image

process to serotonin it can produce presynaptic inhibition decreasing transmitter release and we've shown that with repeated pulses from aphromite can produce long-term presynaptic inhibition

that this is associated with a retraction of synaptic connections more recently we've shown that this can be synapse specific you can get it at one synapse and not the other if you give five pulses of

serotonin but you can capture it if you give a single pulsar from afromide to the other branch this of course raised the question how is this mediated transcriptionally is

this also mediated by kreb 1 or does this involve other uh transcriptional mechanism and so we explored this in the following way we first applied in one uh set of

cultures five pulses of serotonin to produce synapse-specific facilitation and we showed we could block that with kreb 1.

we then gave five pulses from affirmative we produced synapse specific inhibition this was not blocked by kreb 1 but this was blocked by specific antibody to kreb

2.

so this was a fascinating result because i told you earlier that we had demonstrated that krep2 is a repressor of probe one and with these experiments others that we've carried out we've

shown that crep 2 has a second function it can also mediate transcriptional activation and specifically can be transcriptional

activation that participates in long-term presynaptic facilitation so here we had a situation in which kreb 2 mediates two functions it acts as a repressor of prep one

and it acts as a transcription activator of its own important for presynaptic inhibition and this raises the fascinating question what happens if you interact

facilitation and inhibition in the same neuron what happens if you give five pulses of serotonin here and five pulses from afromite here will they balance each

other out will crab one will win out or will crept two went out so we carried out the experiments in the following way we first showed that if you apply five pulses of formalfamily to one set of

synapses you get synapse specific inhibition nothing at the other branch if in another set of cultures you give five pulses of serotonin to one branch you

get synapse specific facilitation at that branch and not at the other but if you now do the competition experiment you give five pulses of serotonin to one branch and five pulse

of formaphmite to the other branch you find that the inhibition is completely expressed normally expressed but that the facilitation is completely obliterated

this is a completely new kind of logic for synaptic integration simply to remind you short-term synaptic integration the way it's classically described in the textbooks like those

written by schwartz and jessel describe synaptic inhibition and synaptic excitation as equals competing with one another and in fact in the short term

process if you give single pulses of formative serotonin a single pulse of serotonin gives you perfectly good expression of the facilitation and for mathematically the other branch gives you a perfectly good expression of

inhibition but in the long term the crep mediated repression dominant decrep to mediated repression dominates and completely overrules the synapse

specificity so it acts as if you even hadn't applied any serotonin at all so in this particular case the logic of the neural network is not determined by the history of usage of the synapse it is

determined by the logic of the transcriptional factors and i think simply summarize that for you here that when one deals with nuclear-mediated integration

one is dependent not only on the history of the synapse but also on the logic of the transcriptional circuitry that you've established so when you interact serotonin and phermaphamide you really

have the logic of kreb 2 determining the system and i think this is really quite likely that we have great practical importance because it can

really explain why the consolidation of long-term memories can so easily be disrupted by stimuli that interrupt your thinking or disturb you in some way that can

activate inhibitory pathways and shut off the ability of the excitatory synapse to really store a still long-term memory so so far i've indicated to you that

using a reductionist approach in a very simple animal one has been able to delineate a behavior that can be modified by learning characterize the neural circuitry pinpoint an important site

that is modified by learning show how that site mediates both short and long-term memory and define a signaling sequence that contributes importantly both to the short-term and the long-term

process but it raises the final point that i want to consider with you and that is how general is this process i've so far indicated to you the simplest case we could consider of

learning in memory this is a form that is called implicit or procedural learning and memory and it refers to essentially the modification

of reflex behavior of improvements in perceptual and motor skills this you find in invertebrates as well as vertebrates it does not involve consciousness for recall it is really an

enhancement of reflex behavior and it as i've shown you and has been shown now in a number of cases this vol involves alterations of synoptic strength within the reflex circuit itself that mediates

the behavior there is no learning system superimposed in it the learning occurs within the system itself by modulatory input but the memories we hold most near and dear

are called explicit or declarative they involve conscious recall conscious recall of facts of events information about people places and objects and this

requires a special structure in the brain called the medial temporal lobe and a region deep to it called the hippocampus and about

10 years ago when i reached my 60th birthday i screwed up my courage and together with a number of colleagues particularly mark mayford and seth grant returned to

the hippocampus one of the stimulus one of the stimuli for turning to the hippocampus was the fact that peki and smithey's had developed techniques for knocking out individual

genes in mice and it became clear that mice are terrific genetic systems for looking at the roles of how individual genes affect synaptic transmission in the one hand and behavioral performance

the other so we began to try to use both the pharmacological and the genetic approach to mice focusing specifically on spatial

memory mediated importantly by the hippocampus now we already knew at that time that as a result of the work of per anderson and many of his students one

had delineated within the hippocampus uh a mechanism of enhancing synaptic strength called long-term potentiation which for some similarity to the

facilitation we had encountered in the plesia there are three major synaptic pathways in the hippocampus i will focus only on one called the schaefer collateral pathway because it's been

well studied and it's very important clinically one has shown that lesions of this pathway give rise to important memory disturbances in humans now earlier work had shown that if you

give a train of stimuli like that to one of these pathways schaefer collateral pathways you produce an enhancement of synaptic strength in slices that last for a couple of hours

an analysis of that showed that this was a covalent modification mediated by kinase called cam kinase different one encountered in in the plesia but the logic was the same a covalent

modification of pre-existing proteins but we had found in the blizzard that there was a big difference between one training trial and repeated training trials there was a representation on the

cellular level of long-term memory so we wanted to see what happens when used repeated trains and we found that when you used four repeated trains you produced a new phase a late phase of ltp

that had very different properties it required cyclic gamep it required the cyclic game p-dependent protein kinase map kinase and it required transcription

and translation so the next law just gives you a an outline of current thinking of how ltp is uh is set up shape for collateral

pathways if you activate them they release glutamate like in the plasia activate a receptor called the nmda receptor that activates that allows

calcium to come in activates an enzyme calcium calmodulin which activates calcium carbonate independent kinase which phosphorylate substrates in the cytoplasm and gives rise to

short-term synaptic enhancement but with repeated trains the amount of calcium coming in also activates an adenylyl cyclase which activates the cyclic gain p-dependent

protein kinase map kinase leading to crep phosphorylation and the activation of downstream genes that are thought and now the evidence is increasingly good give rise to the growth of new synaptic

connections now initially other people in our cells worked on this primarily pharmacologically but we of course were interested in seeing whether we could bring a genetic approach to bear upon us

and i simply want to show you one set of genetic experiments the attempt to separate the late phase from the early phase using a genetically modified mice and particularly we wanted to see what

happens if you compromised the cyclic amp signaling pathway and we expressed in the mouse in a way that was restricted to the forebrain and to the

hippocampus and inhibitory constraint on the cyclic amp-dependent protein kinase a regulatory subunit that inhibited the catalytic subunit but did not recognize cyclic amp

this simply shows you the construct we used a cam kinase promoter to restrict the expression to the forebrain this compares two lines of mutant mice with one line of mild wild-type mice and you

see with one train of ltp ltp is perfectly normal there's no difference between mutant and white type but if you use four trains this is the mutant and this is the wild type you see

a selective defect in the late phase of ltp so here we had something which was really very nice we had an animal that had a perfectly normal early phase and a

selectively compromised late phase and we could ask do these animals learn well do they retain short-term memory and what about their long-term memory

we looked at this in a number of different tasks and we found selective defects in long-term memory and all the tasks we looked at but the nicest one that lies the cleanest temporal

separation is a task called contextual conditioning a form of learned fear but much more complicated than sensitization if you put a mouse into a

closed container like a skinner box it walks around and becomes familiar with the space you then can sound a tone and electrify the grid and shock the animal and it now learns a number of things

most specifically learns that this space is bad news anytime you put it in this space it will anticipate the shock and it will freeze if you put it in another space it will

not freeze and then you can see how long does it remember to freeze one hour short term memory of 24 hours long-term memory and you can see that these animals this

is mutant and this is wild type learn perfectly well they have a perfectly good short-term memory consistent with the fact that the early phase of ltp is normal but the late phase of ltp

selectively compromised this is exactly what you see when you're in protein synthesis so protein synthesis inhibitors interfere with the same step that pka

leads to the turning on of transcription and this of course raises the question why does the animal not learn this task

what is wrong with its ability to recognize this space and it dawned on us that maybe what is happening is that this animal no longer has a memory for the cognitive map of space because

you've interfered with ltp and with spatial memory storage in the early 1970s o'keefe in london first pointed out that the

very cells that participate in ltp also generate a cognitive map of space they have an internal representation of the space they move around with just

like you have a map in your head so the mouse has a map in their head of their space individual cells will code for different positions in that space

so if an animal walks around and you record from cells in the hippocampus and take a picture of its movements you can now take a top-down view of that

and this is illustrated here in the pseudo-color images recording from three cells simultaneously the yellow pseudocolor indicates the animal is moving around but the cell is not firing

so this cell flies at six o'clock this is 12 o'clock and this cell fires here so in this way if you record from lots of cells you see that the animal has in fact formed a three dimensional map a

two dimensional map of that space you can ask now how stable is that map in the wild type animal it forms that map in about 10 to 15 minutes and the

map is stable you can take it in and out and the map will retain itself what happens with the mutant mouse so we recorded from these cells i want to show you two examples here in the wild

type the animal forms the fields perfectly well you take it out you put it back uh an hour later the map is perfectly well retained you take it out you put it back 24 hours the map is

retained if you take the mutant mice you see that they form a perfectly good map if you take them out and put them back an hour later short term the map is really quite good but if you take them

out and put them back 24 hours later you see that each of these fields has rotated you can quantitate this in a blind way and you can show that although in the

mutant animal the map forms it has a perfectly good short-term memory in space the cognitive map is interfered with the long term exactly as you see with inhibitors of protein synthesis

so although this is really the beginning we're now for the first time looking at a cognitive process the kind that in you and me would involve conscious participation and we're seeing how

synaptic plasticity interferes with memory storage because it interferes with the stable maintenance of this cognitive map in space so let me conclude by just making two

general points i have indicated that studies of implicit and explicit memory storage indicate that there is a

conserved logic to signaling involved in memory storage on both the cellular and molecular level on the cellular level

in both implicit and explicit memory storage and evolutionarily ranging from a pleasure to mice short-term memory is

is stored as transient changes in synaptic strength and long-term memory is stored by persistent changes in synaptic strength and the growth of new synaptic

connections on the molecular level short-term memory involves in every case that has been looked at covalent modifications of pre-existing proteins leading to the strengthening of

pre-existing synaptic connections and long-term memory involves recruitment of transcription and the growth of new synaptic connections in its molecular detail short-term

memory recruits a number of different signaling systems suggesting that different learning processes can recruit specific different second messenger cascades but surprisingly the transition

from short to long term and the initiation of the long-term process is amazingly conserved and involves now in a large number of memory processes pka map kinase

phosphorylation of kreb the induction of immediate response genes and the growth of new synaptic connections the fact that this switch is so conserved has some clinical

relevance to begin with a number of disorders of memory including age-related memory loss that is not alzheimer's disease but the

normal of beginning of age of memory that occurs with age commonly involves converting of new information from short-term to

and in mice we've been able to demonstrate that in fact this involves a selective defect in the pka system and its ability to initiate the

transcriptional cascade and since this is modulated by dopamine we've been able to show we can reverse the physiological deficit and reverse the behavioral

deficit in mice by using either d1 agonists or inhibiting the cyclic game phosphodiesterase using roller pram so if you're a mouse we can really help you with the age of

related memory loss for people we're not at all sure as yet but in a larger sense

not only have we been able to use molecular biology to learn something about this nature of memory storage but also as one had hoped at the very beginning

the study of learning and memory has given us i think new insights into biological aspects of signaling within the nervous system and i just want to

use two or three examples to conclude and that is for example we have learned that one of the functions of modulatory transmitters of the kind that paul and

other talked about are recruit is that they're recruited by learning processes as reinforcing stimuli and the reason they're used as reinforcing stimuli because they can initiate

signaling through second messenger cascades of the kind that paul has delineated they can translocate to the nucleus activate genes to give rise to uh to the growth of new synaptic

connections one and two they can mark synapses so that those proteins can be used productively but most interestingly i think

the work on learning has shown that there is ready and easy communication an intimate dialogue if you will between the synapse and the nucleus and the

nucleus and synapses as a result of which synaptic transmission needs to be importantly governed or is importantly governed by the logic of the transcriptional machinery itself and

i've shown you when facilitation and inhibition interact it is the logic of the transcription machinery that wins out in the case of uh facilitation so

when i began my research 40 years ago i had hoped that a reductionist approach

in the simple system would give us some initial insight into learning and memory storage and that was clearly a leap of faith but a leap for which i've been

rewarded beyond my fondest dreams but clearly explicit memory storage is an enormously deep problem and we are really just at the beginning we're beginning to

understand aspects of the storage mechanism but we know very little about the system's properties of memory how storage mechanism at one side of the hippocampus communicates with storage

mechanisms at other sites how the hippocampus communicates with the neocortex but not to worry i think that

there is every reason to believe that these questions will be avidly explored in the years ahead because the

problems of biology of the mind in fact neuroscience in general have captured the imagination of the scientific community of this century

this 21st century in the way the biology of the gene has captured the imagination of the biological community at the end of the last century young people increasingly moving into

neurobiology because they find problems with neurobiology of the mind so extraordinarily interesting in that spirit i really believe that the three of us whom you honor here today

represent if you will the early messengers on mind brain to stockholm as the biological study of mind comes to

assume its central role within medicine and within biology i fully expect that you will continue to call to stockholm and honor as

graciously as you've honored us a whole succession of brain scientists and recognize them for their own leaps of

faith thank you very much

[Applause]

you