Picower Institute for Learning and Memory Inaugural Symposium: “The Future of the Brain," pt. 3

MODERATOR: Hello, there. This is your friendly voice at the front. Please have a seat. I hope your filet mignon was as good as mine. Want to welcome you all to the first afternoon session we're going to have. It's called Change your Mind, Memory and Disease, and it will focus on one theme and that is, unraveling the mechanisms that drive our capacity to remember and to learn, as well as related functions like perception, attention, and that bugaboo word that we've all avoided today-- consciousness.

Memory is the thread, really, that ties our next speakers together. And we're going to start talking about memory disorders as a health problem, in general. And then we're going to focus like a laser beam on two specific disorders-- Alzheimer's and post-traumatic stress disorder. So these will be great opportunities for your questions and answers because-- so after each of our speakers are finished here, we'll be moseying on over to those chairs over there and do our best to imitate Mr. Lipton as he sits down and talks to celebrities on his show. And we'll take questions from the audience as we do that.

Our first speaker is Dr. Thomas Insel. He is director of the National Institute of Mental Health at the NIH. He has a long history of public service, joining NIMH in 1979, then after 15 years, moving on to Emory University, where he worked on obsessive compulsive disorders. Now back at NIMH, he controls a $1.3 billion budget. And he will share with us, I'm hoping, his overview of the state of memory diseases and research, possible research to combat them. Please welcome Dr. Thomas Insel.

INSEL: Well, thank you, Ira, for the introduction, and thank you, Susumu, for the invitation. It's a real honor and pleasure to be here. Congratulations on the opening of this Institute. And for all of those who work here, I think it's an exciting beginning and a chance to do some of what we heard a little bit about this morning in terms of moving into the future. I guess I am the first of what we'll call the post-Nobel speakers today. I've been informed from my colleagues that they want this to be called the pre-Nobel part of the session.

And my comments are actually going to be quite focused on something that we call social cognition or social memory. There's so much we could talk about. But I thought it would be best, given that the morning was, shall we say, free ranging, that we'll try it in the afternoon to dig into a few stories where we get a little more of the detail, have a chance to chew a bit more on the intricacies of some very finite questions to get a sense of what the options are and what the tools are currently and where we're going in some of these areas that have to do with memory and may have implications for disease.

My brief is to talk about social memory, and we're going to put this in the framework of what we call social neuroscience. It's an exciting new area of neuroscience. There are really only three points that I want to make, that I want you to take from the next 20 minutes.

The first is that what we call social memory is different than all other forms of memory. It uses different pathways in the brain and probably responds to different kinds of signals. The second is that, though I'm mostly going to talk about mice, it turns out that humans also have very unique circuits, specialized circuits for social cognition. And finally, the growing evidence that the disease that we call globally autism-- which is probably many different disorders under one name-- all share a deficit in social cognition. It's really fundamental to this disorder.

Now, as you had a sense from this morning, a lot of what happens in science-- and it's no different in neuroscience-- is that we follow where the leads are, where the traction is. And in this field, we've been really fortunate in that there is a small group of neuropeptides, that were mentioned this morning, that nature seems to have selected through a long period of evolution to mediate many aspects of social behavior or social cognition. Who would have thought that that would have been one of the ways in which neuromodulators or neurotransmitters would have evolved?

But it does seem to be the case that, whether you're looking at worms or snails, reptiles, birds, fish, or even in mammals, this small group of neuropeptides has an important role either in social behavior generally or in sexual behavior, in particular. And in the case of mammals, there are really only two members of this neuropeptide family. One is called oxytocin, and the other is vasopressin.

If you've given birth in the United States, you've probably received oxytocin in the form of its synthetic version, which is called Pitocin, that's given usually during labor and delivery for most women who deliver in hospital in the US. We'll focus on this one oxytocin, and it's probably best known as a kind of maternal peptide because it's known in peripheral organ systems to be very important for labor and delivery. It causes the uterus to contract. It's also important for lactation. It causes mammary myoepithelium to contract to push milk out through the lactation ducts. And it's also important for the sexual response, and that's true in both males and females.

Now the key for this story today is that this is also found in the brain. And what does it do in the brain? It's not entirely clear what the peptide does in the brain, but the initial evidence is that what it does in the brain, in some ways, is very congruent with what it's doing in the periphery. And the evidence for that comes from studies, mostly in sheep and in rats, that suggests that in animals that otherwise would never become maternal, the injection of this peptide into select brain areas essentially turns on the whole maternal pathway.

That is, in a rat that may otherwise avoid newborn pups if she hasn't just gone through parturition, hasn't just delivered, if she's given oxytocin at the appropriate time, she will scurry around and build a nest. And if pups are anywhere nearby, she'll go over, retrieve them, and put them in the nest. And she'll crouch over them just as if her brain thinks that she's just delivered a litter. So it has an extraordinarily powerful effect within the brain to turn on maternal behavior, and we would suspect also maternal cognition in some way.

The last point about this is that this peptide seems to be also have different effects in different species, and that seems to depend on where the receptors for the peptide reside because they're somewhat different in the brains of mice, rats, voles, monkeys, and humans. What we do know is that in those species that pair bond, like the voles that Eric mentioned earlier, oxytocin seems to have a powerful effect on facilitating the pair bond, particularly in females.

Now we did what a lot of people do when they want to study a peptide. We had a knockout made. That is, we had a mouse that was genetically engineered to lack this peptide altogether. In the old days, meaning 10 years ago, we used to call that a highly selective genetic intervention. We now think of it as a sort of thermonuclear intervention because what you're doing is taking out this peptide or this gene all the way through development. So if you see a difference in behavior, it's hard to know what it means.

But here, we've got a very interesting observation because the mice that were made in this way, lacking oxytocin, actually look pretty normal. In fact, they don't lactate, but if they're given oxytocin, they can lactate. They don't show much of a deficit in maternal behavior or other aspects of maternal or social behavior initially. And you can imagine we were a little disappointed to find that. We later discovered why that was.

But they do have a specific deficit, which is around the social cognition. How do you study social cognition in a mouse? It's actually quite simple. This is what's called the social recognition test. In this case, we're testing this black mouse, who's in his own home cage. We put a white mouse in with them for five minutes, then take the white mouse out for 30 minutes. And we come back 30 minutes later and expose either to the white mouse the familiar animal or an unfamiliar animal.

And what you see here is that mice generally spend about, oh, 2 and 1/2 minutes in a kind of meet-and-greet ritual, which is a lot of anogenital and [INAUDIBLE] sniffing. They actually sniff wherever the scent glands are. So this is not something we'd recommend that any of you start doing in the near future, but this is one of the ways that, as Richard talked about this morning, that mice get to know each other very quickly.

And they develop this familiarity, and the way you can tell they're familiar-- interestingly, we're losing one of the bars here. Not sure why that is. Oh, there it is-- is that when you test them 30 minutes later, they spend a lot less time on investigation. And it's this difference between the 2 and 1/2 minutes here and the one minute here, about a 50% decrease in investigation time, that we believe reflects the fact that this has become a familiar mouse. Because if you put in a new mouse they've never seen before, they're back up to about two, 2 and 1/2 minutes.

So when you do this kind of a study, you can get some idea of whether mice are able to remember who they've been with. And in this case, we did this with the oxytocin knockout mice, as well as with controlled mice. They're called wild type mice. And we saw really no difference in this initial test. They spent about the same amount of time investigating initially. But when you look at them 30 minutes later, there is this actually quite interesting difference between the two groups.

The wild types, the normal mice, here they do a quick sniff say, oh, yes. You're probably the familiar mouse that I met 30 minutes ago, and walk away. But the knockout mice, while they first sniff in just the same way, they don't seem to be able to remember that this is a mouse that they've met before. In fact, they spend about the same amount of time investigating on the second trial as they did the first. And if they had a third and a fourth trial, you'd see much the same kind of behavior.

So these are what the data look like when you quantify this. The orange bars are what you see in normal wild type mice. And as I mentioned before, there's about a 50% decrease in investigation time between the first trial and the subsequent trial 30 minutes later. And note that in the knockout mice, there's simply no difference at all. They spend as much time the second interval as they did the first. It's just like they've never met before.

Obviously, there are a lot of reasons why that could be. Maybe they can't smell, and maybe they can't learn. It's not that they spend much more time just investigating, because as you can see, they're actually the same at baseline. But there's certainly lots of other possibilities.

Jennifer Ferguson, who did this research as a graduate student, did a very simple follow-up study where she, instead of giving them a mouse to remember, she gave them just a cotton ball that had been soaked in lemon. And what she found was they'd spend about two minutes investigating the cotton ball, and then 30 minutes later, both the knockouts and the wild type were perfectly able to realize that this was familiar. I.e., they are able to smell. And in fact, they're able to learn. But there seems to be a deficit in learning something that is a social as opposed to a non-social cue.

She followed this up with what I think is a really interesting experiment where she said, rather than putting the lemon scent on a cotton ball, let's put it on a mouse and let's see how they perform in that case. And sure enough, if she, in this case, perfumes the mice with the lemon, they are able to recognize, that both the wild types and the knockouts seem to know 30 minutes later that this is indeed a familiar mouse.

But then, if you do the same experiment where you then 30 minutes later put the lemon-- you take a second mouse that they've never seen before, also scented with lemon, what you can see is the wild type mice very quickly say, this is a female I've never seen. She may be wearing the same perfume, but this is a novel female, whereas the knockouts can only follow the perfume. They can only follow the lemon, and they don't seem to be able to make this percept of this being a familiar individual, a familiar, conspecific female that I've known before.

Now, if we put this into human terms, we'd be talking about something that's called prosopagnosia, the inability to make social memories that are, in humans, visual instead of being olfactory. What would this mean? Let's use the example of the people that you've met this morning.

So this would be like having just met Eric Kandel this morning and then 30 minutes later, seeing him again. You'd probably have less reason to have the same conversation with him so you spend less time in discussion with him. But if he then did something quite odd and took his name tag off-- because you weren't able to remember his face but you can only follow his name tag-- and he put his name tag on Richard Axel, you would then greet Richard Axel very much as if he was Eric Kandel.

You wouldn't know. You probably should have been able to know this because only one would wear a bow tie, but in this case, that would be insufficient even for the mice. And you would be following now the semantic information and not the social information. So this is the kind of deficit that we're seeing here.

Now, there are two things about this that were intriguing to us. The first was that it's highly specific. So when you look at a whole range of other kinds of learning tasks, these mice look absolutely uninteresting. They're able to do spatial learning, habituation, other kinds of tasks without any deficits whatsoever. This seems to be unique to the social domain.

The second was that, for us, we really weren't interested in the genes. We were interested in the biology. And so for us, this just became a tool that we could use to now begin to interrogate, how does the brain normally make a social memory? How does the wild type mouse do it? And so we reasoned that we could use these knockout mice to help inform that because we could ask, what's not happening in the knockout mice that is happening in the wild type when they meet a new mouse, when they become familiar with a new mouse?

Now we know that the olfactory system for the mouse has been well-worked out by people like Richard and others. And there are essentially two parallel systems, one that's the main olfactory system that covers most of the olfactory world as we would know it, and then there's another one for non-volatile kinds of substances called the accessory olfactory system that picks up pheromones and a number of other compounds. And our assumption from the beginning was that they probably have a deficit in the way that they're processing information out in the olfactory bulb before it comes into the forebrain.

This is actually very analogous to the story that Richard Axel told this morning about Drosophila. So you can begin to either measure information processing at the nose, at the first synapse in the olfactory bulb, or here further up into the brain at the places where, whether there's a ghost in the machine or whether there's just some higher-order processing, that percept is then becoming something that has meaning.

And so we reasoned that somewhere along the lines the lights were not going on in the knockout mice when they were going on in the wild type. And so we tried to figure out how that was. We used a very simple method, which is to use Fos. It's an oncogene that turns on in cells when they're activated to do what a poor man's PET scan, I guess. And we found that indeed, all through this area, in both the olfactory bulb and in the amygdala, there was lots of activation when there was social interaction with a new mouse.

But in the knockout mice, though there was plenty of activation in the olfactory bulb-- both the accessory olfactory bulb and the main olfactory bulb-- when we got back to the medial amygdala and its projection sites, all the lights were off. So that told us that indeed it seemed to be that there was an area here in the medial amygdala that needed to have oxytocin to be able to make a social memory. And why that was intriguing to us is that we already knew from other studies we had done that an area right about here has the densest innervation or the densest expression of oxytocin receptors in the mouse forebrain, but there are only a few hundred cells that seem to be really critical at that point.

So the last study that Jennifer did as part of her thesis then was to say, if that's really true, we should be able to rescue social memory in the mice the way we rescued lactation by replacing oxytocin where it's missing. So she did one further experiment, and this is an example where it's a little difficult to follow perhaps. But let me just take you through it very quickly.

In this case, it was giving oxytocin directly into the brain, because you can't just inject it. It doesn't cross the blood-brain barrier. And giving it into the ventricle of the brain, she was able to do a dose response curve to find a dose that worked. Now the way these data are arrayed, it's each experiment is just shown as one bar. So this is the ratio of investigation on the second test versus the first test. So if you have social amnesia and you can't learn, you're spending as much time on the second test as you are on the first. And that's what you see in these bars. But if you are able to learn, like the wild type mice, you're showing about a 50% decrease.

So she took a dose that doesn't seem to work very well when given centrally, and she injected that directly into the amygdala in the knockout mice. And what you see is that when she injected specifically into this tiny area where the lights didn't seem to be going on with social interaction, the knockout mice were, in fact, able to learn. So very quickly, the black shows the control injection in knockouts.

So the control injection here is just CSF, cerebral spinal fluid. And that's failure to learn because they're spending as much time on the second test as the first test. This is when given into the olfactory bulb. Giving oxytocin into the bulb doesn't seem to help but giving it into this one area that's missing is really sufficient to rescue them and to now give them olfactory memory back.

Now the final experiment here is just the converse of this, to take the wild type animals and to say, if this is really the critical area, then if we just remove the peptide from that region, that should be really enough to give us the full knockout phenotype. And that's what you see. So again, if it's under 0.4 here, it means that they're perfectly able to learn. That's normal behavior. If you knock out oxytocin using an antagonist, in this case-- so doing it pharmacologically-- in the olfactory bulb, it has no effect whatsoever. But given into the medial amygdala-- again, this one spot where you've got just a few hundred, maybe a few thousand cells with the receptors-- seems to be really critical.

So to sum up, what I've told you is that at a behavioral level, these mice appear to have a selective deficit in social recognition. Other aspects of learning and memory seem to be retained. The cellular basis of this seems to be due to those receptor-containing cells in a small part of the media amygdala because you can rescue the deficit entirely by injecting into that area. And finally, at the molecular level, it appears that this gene is really critical for processing social information. But the take-home message here is that there's something really quite unique about social memory and social recognition that doesn't follow the rules, or at least doesn't follow the same circuitry and probably has different sets of mediators and different sets of underlying molecular substrates, than other forms of learning and memory. Now is this relevant to humans?

Well, as it turns out, in the human brain, as I mentioned at the outset, there are also very specific pathways for social information processing. But we are not olfactory mammals. We are really-- excuse me-- visual mammals. And so the pathways that we're dealing with here are in the visual system. And actually, Nancy Kanwisher, who works in this building, has probably done more than almost anyone to help to delineate these areas.

There are areas that are higher-order processing areas for visual information-- one that's called the fusiform gyrus being the main one. And the point of telling you this story is that in the human brain, information is parsed in a way that might be different than what we would imagine. And I think really part of the power of modern neuroscience is not just that we now know that mental life is neuro life, that you can find the evidence of mental activity in neuro activity, but you can begin to interrogate the brain to find out, how does the brain make sense of the world? How does it categorize the world? How is information actually processed?

And this is an interesting example of that because here, what you find is that almost all visual scenes are going to be processed through this fusiform area. But the brain separates out fairly early on animate from inanimate objects. And it doesn't matter actually whether you show a picture, whether you show the object itself, or even whether you just give this information semantically giving a card with the name hammer, saw, shovel on it. You're going to activate mostly medial regions of the fusiform. If you show another card with camel, dog, face, animate objects, you're going to be more lateral. So this is a so-called face area that's in the brain, and it's been now interrogated in the disease of autism to ask whether there's something abnormal in the way that autistic people or people with Asperger's syndrome-- part of the autistic spectrum-- whether they're processing facial information differently.

This is work from the Yale group led by Bob Schultz. It's published about three or four years ago. It's now been replicated a total of nine times using a fMRI study. And what was done here is people were shown faces, objects, or just matched visual activity. And what you see when you do that is that, indeed, this area of the brain, the fusiform gyrus on the bottom of the brain-- very important active for encoding face information-- is activated in healthy individuals but not in people who have Asperger's or who have autism. So in some way, the information simply isn't being processed where it should be processed in the brain of someone with autism.

Now the initial interpretation of that information was that there could be a lesion in this area, as if these people had had failure of a region to develop in the brain and therefore, weren't able to process the faces. It's actually much more interesting than that. It turns out that if you really look carefully at what's going on when people are in the scanner and shown pictures of faces, the problem is that they're not looking at the face. They're never encoding it because they're looking someplace else. And there was no control in any of these studies to make sure that people really were looking directly at the part of the face that they needed to to activate the fusiform gyrus.

And if you do that experiment and you ensure that people with autism are in fact encoding that part of the face with the eyes, that seems to be essential for activating this area, you can indeed activate the fusiform gyrus. But what you do at the same time is you increase the activity in the amygdala at a very high rate. So there is something about gaze avoidance and face avoidance that's really intrinsic to this disorder, something that we don't really understand very well. But it does tell us that there's something about when people with this disease do begin to look at the face and encode that information, they're able to do it. But they find it in some way aversive, or at least it's one of the things that seems to really activate the amygdala.

Now, a final story about this-- which isn't yet published but will be out very soon-- is that when you have people actually take oxytocin, to get back to the oxytocin story-- and this is now not with autism but with just healthy volunteers-- and you show them faces but now not just any face but threatening faces, you can greatly activate-- well, if you do this in any one of us, you'll activate the amygdala if the face is threatening in some way. It was any face might be that way for someone with autism. But if you take actual threatening faces and you have a person stare at those faces while they're in the fMRI scanner, you will turn on the amygdala big time. And that's what you see here, and you can see that again even with threatening scenes or looking at really catastrophic scenes.

What's interesting is that if you have healthy volunteers take oxytocin-- and it has to be done intranasally-- you very selectively turn off this activation in that particular part of the brain. So you see it both with faces and scenes, but it actually seems to be more true in this case-- this is the subtraction image-- more true with faces than with scenes, suggesting that indeed there is something quite specific, again in humans, about how oxytocin alters social information processing.

So to sum up, I think what we have here is a real opportunity to understand-- both through these genetic techniques and through neuroimaging techniques-- a lot about how social information and social cognition gets processed. We're beginning to identify the pathways in the brain that really count. And I think the comments that were made this morning were right on, that the real heart of the matter here is not going to be just the genes. It's going to be understanding the circuitry, but the genes help us to get to the circuitry. And that's what we found here.

So the information comes in to primary sensory pathways-- maybe olfactory, maybe visual maybe auditory, depending on the species-- and the question is answered. Is this social? At that point, subsequent pathways within the brain or subsequent weigh stations, like the amygdala, temporal cortex, prefrontal cortex, began to instantiate this then with value, to ask the question, who is this? Familiar? Novel? Is this somebody important to me, not important? Friend? Enemy? All of those kinds of issues that really require a lot more integration of information based on past experience and based on being able to decode what's coming in from a primary sensory pathway.

You go from there to a much more interesting area that Eric began to allude to this morning, to parts of the brain that have to do with reward, when this now becomes not only instantiated with meaning but with also hedonic value. Is this someone who is very important to me who I would die for, somebody who I'm pair bonded to, somebody who will take care of me? All of those issues, which happen, actually, in a different part of the brain altogether where we know if you have oxytocin and vasopressin receptors, you're more likely to show very high affiliative behavior and to again, infer that social information is extremely important.

And finally, there's a response axis that we haven't talked about. I think someone very affectionately called it the id this morning. But it's that piece that comes out of the hypothalamus that provides both the endocrine and the motor responses that are really important for the our interaction with the social world. So that is my social memory story. I thank you very much for your attention, and I look forward to a bit of time for questions. Forgot my water.

MODERATOR: Okay. We'll just spend some time over here answering questions if you have in the audience. I'll get the ball rolling while people make their way to the mic. One question I had was, are there other hormones that lend themselves to the same kind of investigation?

INSEL: Well, so there's a lot we don't know. And I think one of the things to remember is that we're at such an early phase in neuroscience in general. We know that there are 20,000 genes in the human genome, and we know that there are probably 500,000 proteins that are encoded by this genome. And almost everything that we've written about and talked about up until the last two or three years is based on probably less than 1% of what's there.

So we're really very much in a discovery era. I would be very sad if, in 2053, all we had to talk about were the very things that we had discovered in this first 1%. So I would assume that there is a lot more out there to work with. And I have to say I'm very envious of all the people up there who are going to have a lot longer to do that than most of us down here.

MODERATOR: You have a question up there? Yes?

AUDIENCE: [INAUDIBLE]

MODERATOR: Yes.

INSEL: Yes.

AUDIENCE: A quick question for [INAUDIBLE]. Has oxytocin been tested in autism?

INSEL: Oxytocin been tested in autism. It has been tested giving it intravenously, which is underwhelming. I'm not sure how much of that would even get into the brain. It's not been given yet intranasally, which we know does get into the brain but not at a very high efficiency. But that does work. It hasn't been done.

AUDIENCE: A quick question is [INAUDIBLE] social recognition by overtraining the animals or [INAUDIBLE] between the training and the testing?

INSEL: Right. So can we overtrain by changing the interval. It's really remarkable. In the knockout mice, this is an extremely dense social amnesia. It's so dense that even if you have given them only a few seconds of time apart, they still show no evidence of recognition.

MODERATOR: Okay. We'll go to this side of the room. Well, while you're getting over there, let me ask you this question. Why do you think so little research has been going on in this field?

INSEL: Oh, why has so little--

MODERATOR: Is it an NIMH priority to--

INSEL: Well, we're interested in social neuroscience, generally, as is the drug abuse agency and others. I think that probably the-- as with any other area of science-- you go where the action is, where you have good candidates, where you have good models, and where you have good tools. And frankly, we have not had a lot of that for something as complex as social behavior. From my perspective, I think Cori Bargmann really got us started with the C. elegans work and her gene-- which, you'll appreciate this. The gene's called NPR, but not related to the radio.

MODERATOR: So he's looking for money.

INSEL: But that really was the first evidence I think that any of us had that you could attack what seemed to be complicated behavior with a rather simple genetic approach, that even a single gene could have fairly profound effects on a complicated behavior. And I think-- just to add one thing to that-- I think it helped us to realize that we're not even smart enough to know what is a complicated behavior, that we shouldn't be thinking complicated, simple we should possibly be thinking conserved, derived. If it's a conserved behavior, it may be more likely to have a relatively simple molecular basis. And I think her work got many of us going, thinking much more clearly about that.

MODERATOR: Last question here.

AUDIENCE: Back in the dark ages of psychoanalysis, some of us maybe remember, there was, of course, the concept of cathexis. And you would cathect somebody if you developed strong emotions directly with that person. And one of the goals of analysis was to decathect. And of course, one wonders now whether the decathexis process might have had some reality to it and might have been related to changes in the numbers of oxytocin receptors. And in that vein, I wonder whether there is the possibility that oxytocin antagonists might be useful in treating panic disorders that are occasioned by individuals that the person sees.

INSEL: Well, I think before we jump into thinking about how to use this in people, we need to know a lot more about what it does. But one of the things that's so important about this system is the species' specificity of the peptides' effect. So what you see in mice is not quite what you see in rats, and what you see in rats is not what you see in voles. So I'm a little loathe to jump immediately from the mouse to the human. We've got the first human study, which isn't even yet published, on what the peptide seems to do in the brain. That will be out very soon.

But I have to say, that is promising. I think it does tell us that there are effects. I'm a little bit almost speechless about this. I've heard more about psychoanalysis here at MIT this morning than I have in three years at NIH. So my last thought is to refer your question to Dr. Kandel.

MODERATOR: Nothing personal. Thank you very much for taking the time to talk with us.

INSEL: Thank you.

MODERATOR: All right. Moving along. Our next speaker is looking at a different aspect of neurology. Let's talk about Alzheimer's disease for this period. It's one that is surely affecting more and more of us as the population ages, yours truly included. Our next speaker, Dr. Li-Huei Tsai of Harvard, has a laboratory that is trying to understand the underlying mechanisms of Alzheimer's. And if I read her research-- well, she'll explain it to you herself. Please welcome Dr. Li-Huei Tsai.

TSAI: So I feel extremely honored and humbled to be on this podium today to speak with you. So I just go right into my talk. Alzheimer's disease is a irreversible neurological disorder that progressively attenuates the cognitive abilities of those afflicted. I think many of us in this room have seen the effect of this disease in a elderly family member or a friend.

It was first documented almost exactly one century ago by this man, a German physician, Dr. Alois Alzheimer, describing a female patient, Auguste Deter, who was not able to care for herself, disoriented, with impaired memory. This disease really represent a great human tragedy, which can be appreciated by the following quote. "In Alzheimer's disease, the mind dies first-- names, dates, places-- the interior scrapbook of an entire life-- fade into mists of nonrecognition."

The disease is the most common among all human neurodegenerative disorders, with about 20 million people afflicted worldwide. And age is the single most significant risk factor for Alzheimer's. Currently, about 25% of those over 75 years old have the disease. So with life expectancies continue to rise, the victim of this disease is expected to triple by the year of 2050.

Alzheimer's patients who inherit the disease from family members are known as familial Alzheimer cases, and this is extremely rare, only representing less than 5% of all patients. So most of the patients are sporadic. Therefore, that means the disease is very complex. And to date, there is essentially no therapeutic strategy against this disease.

Brain is known to be severely reduced in size, with about 30% to 40% reduction of total brain volume in severe cases. But if you look more closely, there are two signature pathological features-- amyloid plaques and neurofibrillary tangles. The plaques are protein deposition in the brain composed of a-beta peptides with 39 to 42 amino acid in length. The tangles are intracellular aggregates of straight and paired helical filaments composed of hyperphosphorylated tau protein.

But one very important feature of Alzheimer's disease that sets it apart from most of the other neurodegenerative disorders is the specific brain areas that are affected by Alzheimer's pathology. Neurofibrillary tangle, as well as brain atrophy, always initiate in the hippocampal formation, part of the limbic system essential for learning and memory-- that you heard a lot about this morning. The pathology then spread into the adjacent cortical area, including the temporal lobe and subsequently to the prefrontal lobe-- important for other higher cognitive function. And eventually, in the terminal stage of the disease, the entire cortical area is affected.

And our initial understanding of the disease actually comes from the small percentage of the familial cases. The important role of the a-beta peptides, especially a-beta 42, in the early stage of the disease is supported by the familial Alzheimer disease mutations that preferentially elevated the production of a-beta 42. Therefore, the amyloid hypothesis predicts that accumulation of a-beta in brain, plaque formation, can trigger cascades of events eventually leading to dementia.

Recently, it is proposed that it is the oligomeric soluble form of a-beta that's toxic. In any case, elevated a-beta can change the cellular properties of neurons, induce neurofibrillary tangle formation, caused neuronal death, culminating in dementia. So the amyloid hypothesis nicely explain the familial cases of Alzheimer's disease.

However, the cause for the sporadic Alzheimer disease cases, which account for the greater than 95% of all patients, remain to be elucidated. And also, it is really not clear why only certain areas of the brain are selectively vulnerable to Alzheimer's and also, why age is the single most important factor. So I think it is reasonable to speculate that there must be other important factors contributing to the development of Alzheimer's disease.

So based on the pathogenic features of Alzheimer's, several years ago we predicted that the other factors contributing to Alzheimer's must fulfill the following criteria. We speculate that the factor must be abundantly present in the vulnerable area of the brain, and it should be able to cause neuronal death and brain atrophy, as well as amyloid and neurofibrillary tangle pathology, at least in an animal model. And finally, since the most vulnerable areas of the brain for Alzheimer's are exactly the same area essential for learning and memory, one can speculate that the factors might be normally involved in learning and memory but somehow can lead to altered learning and memory in the disease.

In the past several years, our research, and later on joined by other laboratories, provide compelling evidence that a small protein kinase, Cdk5, may fit the bill. Recently, polymorphisms and a splice variant of Cdk5 have been linked to Alzheimer's disease. As a protein kinase, Cdk5 is highly unusual in that it has to associate with a regulatory activator to become [INAUDIBLE] active.

And there are two such activators present for Cdk5, and we all have these two proteins-- p35 and p39. Cdk5, p35, and p39 are all specifically expressed in neurons. In fact, the highest levels of p35 and cdk5 expression are in the hippocampal formation and certain cortical areas. This is a in situ hybridization using p35 probe of a sagittal section of the mouse brain.

And similarly, the coronal section showed the same thing, that several areas in the hippocampal formation express high level of p35 as well as the piriform cortex, which is equivalent to the temporal lobe in human brain. Now it is important to know that p35 comes in two flavors. A protease known as calpain can chop of the end terminal one third of p35 and produce a smaller species, known as p25. P25 is more potent in activating cdk5, and it can cause hyperactivation of cdk5.

Interestingly enough, we found that p25 actually accumulates during aging. Here, I'm showing you p25 and p35 protein levels in a mouse brain in different ages. And you can see that while the p35 protein level remain more or less constant, p25 is much higher in a 28 months old mouse brain-- which is very, very old for a mouse-- than a seven-month-old mouse brain.

We have also previously shown that the p25, p35 ratio is generally elevated in post-mortem brain samples of Alzheimer's disease. And we recently obtained further evidence that this ratio is already elevated in the preclinical stage of Alzheimer's disease, known as mild cognitive impairment, and also in early stage of Alzheimer's disease. And the elevation is observed in the hippocampal formation and in the prefrontal cortex, areas susceptible to Alzheimer's disease. Therefore, p25 accumulation, leading to hyperactivation of cdk5, occurs during aging and in the development of Alzheimer's disease.

So if p25 contributes to the development of Alzheimer's disease, one will speculate that it should be able to induce Alzheimer disease pathology in an animal model. To this end, we created a p25 mouse model that specifically produces p25 in the vulnerable area of the brain. A more important feature of the mouse is the fact that we can actually turn on and off p25 expression in the brain just by feeding the mice with different kinds of diet, and this really gave us a lot of flexibility to test the potential role of p25 in neurodegeneration.

As it turned out, production of p25 in mouse brain has a very severe deleterious effect. First of all, brain is severely reduced in size. I'm showing you a small animal MRI scan of a control mouse brain and a p25-producing mouse brain. This is 10 weeks after p25 production. And you can see the volume of the brain is massively reduced.

In general, we see a 30% to 35% reduction of the brain volume after production of p25. And this is clearly due to a very massive death of neurons, first in the hippocampal formation and later on spreading to the cortex. The upper panel here representing control normal mouse brain, and the bottom panel here are the brains producing p25. In addition, there is also massive infiltration of inflammatory cells in the afflicted area, another indication for neurodegeneration.

After prolonged p25 production, we see evidence for neurofibrillary tangle formation. Here, I'm showing you hyperphosphorylation of the tau protein using antibody staining and also filamentous tau protein formation using electron microscopy. And after about 30 weeks of p25 production, we can see evidence of neurofibrillary tangles with silver staining or with Thioflavin S staining.

Finally, another very revealing piece of information about p25 and Alzheimer's is the fact that p25 actually has a profound effect on the production of a-beta peptides. Here, I'm showing you some biochemical evidence for the level of a-beta peptides in a mouse brain with p25 production in green and without p25 production in red. So here, if you look at a-beta 42, the species of a-beta peptides particularly relevant to the etiology of familial Alzheimer disease is elevated by several fold.

A-beta 40 is also somewhat elevated now to a similar extent of a-beta 42. This elevated a-beta peptides accumulate inside neurons and readily form the aggregates looks filamentous in appearance. And this is accompanied by the appearance of condensed or pyknotic nuclei, indicative of sick or dying neurons. So, so far, I show you that cdk5 and its regulatory activator are abundantly present in the vulnerable areas of the brain for AD. And in an animal model, it can cause neuronal death and brain atrophy. It can also cause amyloid and neurofibrillary tangle pathology. And I haven't addressed the question, whether cdk5 is normally involved in learning and memory and whether hyperactivation of cdk5 can lead to altered learning and memory in Alzheimer's disease.

In fact, there is already a large body of literature addressing the role of cdk5 in regulating learning and memory. Especially in several animal models, it has been shown that reduction of cdk5 in mouse with pharmacological inhibitors can impair learning ability. Also, pathological function of p35 in mice through genetic manipulation can also result in impaired fear conditioning learning.

Recently, our lab produce and publish result indicating that inducible mice with cdk5 loss of function in the hippocampus show learning impairment in the Morris water maze test. So these observations suggest that cdk5 is normally involved in regulating learning and memory. So the Morris water maze test is a very well-established paradigm testing the spatial learning in mice. And actually, Dr. Richard Morris is in the audience today.

And briefly, mice are placed in a swimming pool filled with murky water and a platform is inserted in a swimming pool submerged from the water so the mouse cannot see it. There are also various visual cues placed in the room to facilitate orientation. So the mouse can only escape from swimming if they can manage to locate the hidden platform by learning where the platform is in relation to those cues. So it usually takes very intensive training for the mouse to learn, takes on average 10 to 14 days. And this is a typical learning curve for a normal mouse. With each day of training, it takes less time for the mouse to find the platform.

I also prepare a couple of movie just to give you some idea what it's like. This is a mouse first day in training. And this is the same mouse 10 days after training. And you can see that it behave very differently since know exactly where to go and quickly escaped. And one can also perform a so-called probe trial, which involves removal of the hidden platform from its usual place and then place the mouse back to the swimming pool and just follow a similar trajectory. If a mouse learned where the platform is, then it tends to spend a lot more time in the area where the platform used to be placed.