 Good morning. I hope everyone had a nice weekend. Do we have any announcements from the TAs? Do we have any TAs? The TAs had a nice weekend. This is the second of our two lectures on memory and learning. Ralph gave the first one. I'm giving this one, and Ralph will give the third one. So this one is on synapses and synaptic plasticity. A more reductionist viewpoint. We start with this interesting vignette that occurred at Harvard in the late 40s and the early 50s in the 20th century. This was a discussion between B.F. Skinner and D.O. Hebb. You've all heard of Skinner boxes, presumably. Skinner was the psychologist who said, We can't know what goes on in the brain. So we are simply going to look at behavior. We're going to put animals in boxes that reward them for some behavior or critique them for other behavior or punish them for other behavior. He actually did this with his children as well. And so his book was entitled, The Organization of Behavior. His colleague at Harvard, Donald O. Hebb said, No, we actually do need to understand the physiology of behavior. And so his book was entitled, The Organization of Behavior Rather than the behavior of organisms. Hebb said, What is learning? Skinner said, We can't ever tell what learning is. We have to put people in boxes and reward them and punish them. Hebb had an idea. The synapse was just being discovered, was just being studied. It was about the same time that Roger Sperry was working on his experiments to cut axons to the optic tectum, et cetera. So Hebb had an idea that perhaps synapses were involved in memory and learning. And so his formalism, now called the Hebbian synapse, basically we knew about axons then we knew about synapses. And so Hebb basically said, Well, if axon A fires is near enough to excite cell B, and this takes place a whole lot of time, then there is a growth process or a metabolic change that occurs so that A's efficiency is increased in order to fire B. So this was pretty much the same time period that Sperry was thinking about this vague term, chemo-athenity. And so Hebb had this vague term joint firing of synapses. And this of course said, Hebb could be the basis for learning. And so now we would call a Hebbian synapse a coincidence detector. Two events occur at the same time, the presynaptic axon and the post-synaptic cell A and B. And we know that in many cases, not all cases, the coincidence detectors in MDA receptors, backpropagating action potentials into dendrites, and summation of excitatory post-synaptic potentials, all of which were just being discovered, well, we knew about synapses back in the late 40s, early 50s, the rest were discovered later on. And so these are the components that confer Hebbian behavior on the synapse. Now, there are lots of molecules involved, lots of ips, ands, or butts, lots of places in the brain, and we're going to talk about them today. We usually think of synaptic plasticity. Plastic, of course, means changeable, changing form. We usually think of synaptic plasticity as occurring in the selected regions of the brain. In the hippocampus, in the cerebral cortex, and the striatum are the best places we know that plasticity occurs. This cerebellum, which has lots, which probably has a plurality of synapses in the brain, is not strongly plastic. So I've been telling you at the beginning of the course about the synapse being a biophysical machine, specialized function on a time scale of milliseconds, distance of microns, all that is true. However, the synapse also needs to adapt to the changing needs and requirements of the organism and its changing activity levels, and therefore synapses need to be regulated in almost every brain region in order to maintain homeostatic balance. Now, homeostatic, and of course, at the same time, this regulation of the synapses allows the synapses to change so that they're storing information in neural circuits. I don't know if I've given you my rant about homeostasis yet, but I will do so now. There are terms that we throw around in neuroscience, homeostasis, adaptation, plasticity, compensation. These are very general concepts, but they really are summaries for a large number of processes. They are not mechanisms in themselves, and so I like to use those words more as adjectives. There is a homeostatic process, there is a compensatory process or mechanism, and so I have had very distinguished neuroscientists say to me in the last year, well, the brain changes because of homeostasis or because of plasticity. Well, we could say the brain changes because of homeostatic processes, and during this century, I do believe, and other neuroscientists believe, that we're going to continue to discover the molecular and mechanistic instantiation of homeostatic processes. Okay, rant has ended, you can take notes again. The key organ that is very often involved in memory and in learning is the hippocampus, which Ralph introduced to you in the previous lecture, and I have mentioned to you several times as well. You remember HM's brain and HM's hippocampus, which proved really elegantly that the hippocampus is required for memory. Now I'm going to give you a little more detail on the hippocampus. Remember that hippocampus itself means seahorse, and so this wound around region looks like a seahorse. The region of the hippocampus that looks like a tooth here, called the dentate region, is called a tooth-like region. Dentate means tooth. The region that looks like a ram's horn, so part of the hippocampus that just has one curve, is indeed named after Amon, the Egyptian ram-like god, and so it is called Amon's horn, and the later neuroanatomists gave it a Latin term, cornus amonus, which means simply Amon's horn. We abbreviate that term these days into the three subregions, CA1, CA2, which is rarely shown because it's rather small, and CA3. Now on the slide at the bottom, there's been a slate of hand, it's been flipped right to left. So here on top, the entorhinal cortex is at the left, but at the bottom, the entorhinal cortex is at the right. Another interesting aspect of the hippocampus is that it's so easy to picture because actually, it curves around so that you can take a coronal section or a sagittal section, and it still looks like a seahorse. So that's a little confusing because it looks slightly differently in each way. In any case, there is a famous circuit in the hippocampus called the trisynaptic circuit. The input from the entorhinal cortex, that part of the cortex that's a little lower than the rest, that's why it's called ento below, into the dentate region, also called the dentate gyrus, synapsing onto the granule cells. The granule cells send mossy fibers that not shown in this picture, but they are sort of wisp-like with lots of projections into the CA3 region, and the CA3 region send the axon collaterals to the CA1 region. Now for today's quiz, who discovered the Schaefer collaterals? So, there is another input to the hippocampus from the dentate gyrus. It is the direct input, and this input is also called the perforant pathway because it actually runs directly from the neocortical region, the entorhinal cortex, to the archicortical region, the hippocampus, which is only three-layered cortex. In order to do that, it needs to cut across a structure, it needs to perforate that structure. Alright, I've given you a little more anatomy than you thought you knew until now, and the reason that we are so interested in the hippocampus is that, of course, as your elf explained to you, it is very important for spatial memory. Now, what does it mean that an organ is important for spatial memory? Well, that's the 1914 Nobel Prize for Physiology or Medicine. And in the hippocampus, there are so-called place cells and grid cells. The place cells are in CA1, the largest of the regions. The grid cells are upstream in the entorhinal cortex. This is an experiment that was first done by John O'Keefe several decades ago in which he implanted electrodes into the CA1 region of a rat. At the time, it was easier to use rats than mice. TV had been invented, but I think he just looked at the rat at that time and led the single unit recording to an oscilloscope and noted that when the rat took a defined trajectory in the field, the particular cell that he was recording from fired. And so this would not be a Morris water maze because it's a little tricky to record from a rat or a mouse while it's swimming. You don't want to get the electrodes wet. So this was in a box that had landmarks around the edge of the box. But the rat knew where he was and lo and behold, these place cells fire when the rat is only in a specific position in the environment. Now later on, within the last 10 years, the Mosers in Norway, who shared the Nobel Prize with O'Keefe, found that in addition in the entorhinal cortex, there are cells that fire in a grid. Now what is this picture? This picture, the black lines show the trajectory of a rat walking through an environment. In a place cell in CA1, there's only one region in the environment where that cell fire. And the red is the particular cell that's being monitored and every time the cell fires a spike, there's a red dot. So here's a place cell in CA1. In the entorhinal cortex, the cells actually define a grid so that here is the rat's trajectory, the fine black lines. As the rat moves through the environment, this one cell, which is being monitored, fires here and then over here and then over here. And so it fires along a grid. Now the grid, the coordinates and granularity of the grid changes from dorsal to ventral, medial to lateral. In fact, this is usually called, you usually see this abbreviated MEC, medial entorhinal cortex. And the grid also expands when the rat or the mouse is in a different environment. So there's obviously a whole lot of interesting processing that relates grid to place and back and forth. And the processing there is about as interesting as the processing that leads from simple to complex cells and visual cortex. The key is that this gives us the physiological slash anatomical substrate for an animal to process place in the hippocampus. And these inputs are known to be plastic and depend on experience. And so the stories I'm going to tell you about synapses now have a lot to do with animals modifying their place fields. In fact, the stories I'm going to tell you about synapses revolve around a picture that you've already seen. Now we go into the hippocampus. We go into the CA1 region. We look at the dendrites of the CA1 pyramidal neurons. And they are very complex machines. So this is a single neuron. We're not even seeing the cell body. We're seeing only the dendrites. The dendrites expand out and they also have spines on them, little protrusion from each dendrite, which vastly increases the area of the dendrite and allows a single neuron then to synthesize inputs for many different regions and for many different inputs. And as we'll see the size, shape of the dendrites, as well as the molecules in the dendrites change during synaptic plasticity. Any questions so far? All right. This is a blow-up, again that I've shown you in previous lectures, of a dendrite. Here's a dendrite. Here is presumably a spine. Two spines on the dendrite. Here's one of them. Here's another one. They have these dark regions here. An electron micrograph. These dark regions are the so-called post-synaptic densities. These post-synaptic densities, you may remember, are chock-full of interesting proteins. That's why they stand dark. In the pre-synaptic terminal, we have the synaptic vesicles, which you know well. There are mitochondria that make sure that we have enough energy. There are probably glia around. Microtubules in the dendrite. Now, in the pre-synaptic, the post-synaptic density are the molecules thought to be responsible for most of synaptic plasticity. And of course, the hero of this story is the NMDA receptor. And the hero of the NMDA receptor story is really calcium. We've come back to calcium again and again in this talk. During the very first lecture, we reminded you that organisms, once they decided that they were going to be using high-energy phosphate as energy, had no choice but to keep intracellular calcium on average very low. This then allowed organisms, cells, to signal by transient local increases in calcium with an enormous number of signaling molecules that take this calcium. So that naturally there is a fairly well-understood and well-proven calcium hypothesis that controls synaptic plasticity. And I've shown you examples using either small molecules or proteins of measuring calcium influx into cells, mostly using fluorescence. And we're going to talk a little bit about the control of post-synaptic calcium using spike timing later on. I'm simply going to give you a quick review of the NMDA receptor as a coincidence detector. Remind you that it is permeable to sodium and potassium like most excitatory ligand transmitter-gated channels but also to calcium, that its kinetics are slower than those of AMPA receptors, that it has a special pharmacology. Indeed, NMDA is not a natural transmitter. It is N-methode aspartate. And the natural transmitter is glutamate. A subset of glutamate receptors are activated by NMDA. And as we'll see, the NMDA receptor has this enormous intracellular region that seems to act as a scaffold for assembling other proteins, some of which are involved in sensing the calcium that flows through the NMDA receptor. We're going to introduce the term long-term potentiation and long-term depression as very much the physiologist's instantiation of the Hebb's rules. And we're going to show that the calcium comes in through the mouth of the NMDA receptor. And then we're going to talk a little about the biochemical pathways that do mediate synaptic strength. So just a reminder that the NMDA receptor is a coincidence detector, that the two coincidences are number one, that glutamate binds. And number two, that there is a strong post-synaptic membrane depolarization. That strong post-synaptic membrane depolarization, this part gets positive, kicks magnesium out of the channel, where it's there as a blocker. And you will remember that magnesium gets stuck in the channel for this weird, fundamental biophysical reason, that it can't exchange its waters of hydration. So it cannot flow through the channel. It simply gets stuck in the channel. So naturally, magnesium is this channel blocker that gets stuck in the channel, and that is very useful for the body and has the selective advantage that it's part of the coincidence detection. And the result is that NMDA receptors flux calcium into the cell when these two events occur simultaneously. Now, the post-synaptic density, here is the NMDA receptor. It's one of the glutamate receptors. There are also amper receptors, which are glutamate receptors as well. The amper receptors lack this intracellular region that helps the NMDA receptor serve as a scaffold for assembling other proteins. And then there are also metabotropic glutamate receptors, GPCRs, that have their own signaling apparatus. But it's really the NMDA receptor that we want to follow today. And in particular, the two molecules that seem to sense the calcium that comes in through the NMDA receptor. One of those molecules is CAM kinase 2. I'll give you the definition for CAM kinase 2. The other is called calcium neuron. CAM, of course, is calmodulin. So CAM kinase is a calmodulin-activated kinase. Calcium neuron is another calcium-activated kinase. Now, the CAM kinase is a rather interesting molecule, as discovered by Mary Kennedy, because it gets activated by binding calmodulin. So here is a calmodulin molecule. You may remember that calmodulin is the universal, very well-conserved calcium binding protein in all of life. It has four calcium binding sites. I know that Albert's made calmodulin look a little bit like two schmoos joined together. But in fact, these are the four calcium binding sites. When calcium binds, when calmodulin binds calcium, it undergoes a conformational change, which allows it to wrap around many molecules, including CAM kinase. This allows CAM kinase to become phosphorylated and activated at its active site. It's as though the inhibitory domain of CAM kinase gets pushed aside by calmodulin. CAM kinase becomes activated either partially or fully. It allows CAM kinase. Remember, a kinase is a molecule that phosphorylates another molecule. In this case, CAM kinase phosphorylates itself. That is, it autophosphorylates. And this stabilizes CAM kinase in the active form so that CAM kinase can now go and phosphorylate other molecules. So this calcium influx hits CAM kinase at the mouth of the NMDA receptor. The CAM kinase actually hits the calmodulin. The calmodulin wraps itself around the CAM kinase. This activates the CAM kinase, which has now been switched on because it autophosphorylates itself. And so we have a calcium-independent activity, calcium-independent kinase activity. And as you've learned, kinases are very important switches inside neurons. Ultimately, the activity does die away, hours or so, depending on how strongly it was activated. Any questions? Okay, so we've said all of that. So now we talk about the fact that synapses can actually be modulated either presynaptically or post-synaptically. We can modulate the number of vesicles released. Remember, they're called quanta. We discussed this when we discussed the nerve muscle synapse. Or we can also regulate the quantal size, the effect, the post-synaptic effect of every quanta. Now, at the nerve muscle synapse, the only way to regulate the best way to regulate the amount of calcium released, sorry, the amount of acetylcholine released, the number of vesicles released, is by changing the calcium in the external solution, which changes the amount of calcium that comes in presynaptically. At the neuromuscular junction, the only way to modulate the sensitivity to an individual quantum that is post-synaptically is by putting a blocker, a partial blocker, such as curary on the synapse. But the nerve muscle synapse really exists to be an all-or-none synapse. It doesn't have to grade itself, whereas synapses in the brain do need to grade, modulate, et cetera. So synapses in the brain control either the number of vesicles released or by regulating the quantum size. There are several components of this modulation. They are usually divided into short-term and long-term. Now, obviously, short-term depends on what sort of scientist you are. It could range from picoseconds, if you're that kind of biophysicist, to millennia, if you're a geologist. But we're talking about short-term from milliseconds to minutes. Usually, these short-term plasticities are presynaptic. In turn, the short-term modulation is divided into numerous subtopics that the physiologists have discovered. There is so-called paired pulse facilitation, which I will show you. That's the briefest. There is synaptic depression, which takes a little longer. And there's potentiation. And in almost every case, the synapse can be either depressed or potentiated. And this varies with the synapse and varies with the stimulation. And then we talk about long-term plasticity. And typically, to a neuroscientist, long-term plasticity is something longer than 30 minutes. There is long-term potentiation. Here, the P means potentiation. And long-term depression. And as Ralph told you, LTP, when it lasts for more than a couple of hours, is usually associated with protein synthesis. We'll get to that as well. So the briefest part then of synaptic plasticity is so-called paired pulse facilitation. Probably the best example of paired pulse facilitation is if we look at the trace at 25 milliseconds. We've given two pre-synaptic stimuli in the cortex. This is a layer 2-3 cortical neuron. It's one of the densest layers in the cortex. We've given two stimuli. Now, you'll remember, when we discussed integration at synapses, we discussed purely electrical stimulation that two post-synaptic potentials that occurred near each other could integrate simply because of the currents. Or if they were far from each other, they were at other ends of the cable. Here we're talking about more subtle processes in which the numbers of pre-synaptic quanta are released are different. So you can see pretty easily that post-synaptic potential number 2 is larger than post-synaptic potential number 1. And this continues to be true for about 100 milliseconds. And then, after about 100 milliseconds, there is no residual effect of stimulus 1 on stimulus 2. And it's been shown pretty much consistently that this residual effect, so-called paired pulse facilitation, has to do with the fact that that pre-synaptic terminal has calcium left over inside it from that first pulse. So the pre-synaptic terminal has more calcium. There's a very nonlinear relationship between the amount of calcium and the amount of transmitter release. As I think I mentioned earlier, it's probably something like the fourth power. And so a little bit of calcium left over produces a big change in transmitter release. Any questions? Of course, if one goes too heavily, the other aspects of transmitter release, the releasable pool of vesicles, for instance, begin to be fatigued during mini-stimuli, and so if one stimulates quite rapidly, one actually sees the opposite, short-term synaptic depression. This varies among synapses, whether one is going to see facilitation, potentiation, or depression. And here is a graph of the EPSP amplitude as a function of the rate of stimulation at a low rate of stimulation when gets a normalized amplitude, whatever they're measuring here, 0.8, but with higher and higher frequencies, the amplitude of an individual post-synaptic potential decreases remarkably. And so the pre-synaptic terminal gets tired, people have studied whether it gets tired because the vesicles can't be filled, that usually doesn't happen because there are a few vesicles around to be released, that does happen. And you may remember that when we studied the visual system, we had a structure in the photoreceptor called the ribbon. The function of the ribbon is to line up those vesicles near the synapse so that lots and lots of them are available, and that's appropriate because in the retina, the photoreceptors are constantly releasing transmitter and what the rodopsin does eventually is to stop releasing transmitter by hyperpolarizing the membrane because the cyclic GMP-activated channels are no longer activated. So varying with the synapse, we have various kinds of potentiation and depression. And probably the most famous example of potentiation is so-called post-titanic potentiation. Remember, a tetanus is a train of stimuli at something like 50 hertz or 100 hertz or a couple of 100 hertz. Their post-titanic potentiation is the earliest manifestation of giving a very strong stimulus. The EPSPs are highly potentiated for the first few minutes after the tetanus. This dies away, but one is left with a post-titanic potential that's greater than the control period and that's called long-term potentiation. So here we are back in the hippocampus. We are at the CA-3, CA-1 synapse. Here's the CA-3. Here are the Schaefer collaterals being stimulated by a presynaptic wire. And we are recording actually extracellularly the synaptic currents in CA-1. This is the classical hippocampal preparation in which we stimulate the Schaefer collaterals and record in CA-1. There are these two phases after a tetanus. Even though the tetanus lasts for just a couple of seconds, we then test the post-synaptic currents every two or three minutes so that the tests themselves do not evoke any changes in plasticity. It's only the tetanus at the very beginning, the tetanic pulse at the very beginning that provokes a change. Gradually then the potentiation dies down and one is left with long-term potentiation. So the post-tetanic potentiation right here again is presumably an accumulation of calcium but the LTP long-term potentiation which is very interesting is a model for memory and learning. It's different processes. So again, recording long-term potentiation in a hippocampal slice, so one takes a slice from an animal, puts it in a special bath, stimulates the CA-3, the Schaefer collaterals, records the post-synaptic currents in CA-1. Now the way this is typically done is not intracellularly but extracellularly actually recording the currents and there are technical reasons why this is easier. So from about 50 to 200 hertz, one gets this brief rise but then the long-term rather stable increase in post-synaptic potentials and after about two hours results say that this probably depends on new protein synthesis. So what's going on here? It's frequency dependent. There are probably many mechanisms and lots of neuroscientists spend their careers very happily tracking down mechanisms of LTP after titanic stimulation. Again, LTP lasting 30 minutes to a few hours don't require a new protein synthesis after a few hours new proteins do need to be synthesized. In addition to the LTP, which is produced by a high-frequency training, 50 to 200 hertz, there is another phenomenon called long-term depression. Again, many mechanisms, again it modifies circuits to store information and we're going to see that long-term depression actually typically occurs for stimulus frequencies that are lower than LTP. So we usually, neuroscientists usually talk about LTP as titanic stimulation, 50 to 200 hertz, LTD as 1 to 10 hertz. You can't keep on increasing your synaptic transmission all the time. You can't keep on learning. You have to forget some things too in order to sharpen up the learning. Now we're going to talk about actually the mechanism that Heb really postulated, which is coincidence, not frequency. Coincidence, we call these days, we measure it as spike-timing dependent plasticity, but it presumably arises from the same set of mechanisms as LTP and LTD. In every case, it is thought oversimplified that post-synaptic calcium levels are controlling LTP and LTD. So here is rest. When calcium levels rise very high post-synaptically because of tetanide, titanic stimulation, we get LTP long-term potentiation. When they rise less strongly, we get LTD, which is long-term depression. That's an oversimplified view, but it seems to hold more or less, especially for experiments on hippocampal slices. So, is that all of that? Okay. So when we record LTD in the hippocampus, we again use this extracellular field recording. We give stimuli at moderate frequencies 1 to 10 hertz, and after a moderate frequency stimulation, for just a couple of seconds, we then record at much lower rates so we don't perturb the system. We get a decrease, and then this decrease reaches a steady state that seems to represent depression. So what we're actually doing here is to say that the strength of synaptic transmission is what underlies memory and learning. There are very good experiments with mouse genetics and pharmacology suggesting that this is true, and so we are using synaptic plasticity, either potentiation or depression, as a proxy for memory and learning. It's a simple proxy. It doesn't always work, but it is remarkable the extent to which it does work. So the two cellular processes that seem to underlie the major changes during LTP and LTD are, first of all, that the post-synaptic cell inserts new amper receptors into the membrane right at the synapses or withdraws them from the synaptic membrane in LTD. Remember, amper receptors are a type of glutamate receptor. And then in addition, the entire spine can grow or shrink changing the size of the post-synaptic potential. There have been some remarkable facts discovered about LTP. One of them is that it's probably input-specific as it needs to be for the Hebbian hypothesis. There are thousands of synapses on a given neuron, and we're talking about only one. And so if we have normal synaptic transmission with a weak active input, we do not get LTP. If we have some cooperativity between neighboring synapses that are all weak, we begin to get LTP, but they need to be near each other so that the molecules produced can indeed communicate with each other. There is perhaps some associativity in LTP in the sense that a strongly-potentiated synapse may also potentiate a weakly-activated synapse next door. But if two synapses are far apart, they do not potentiate each other. They have LTP separately. And so these are ways in which the Hebbian rules get down to the level of individual synapses. In terms of Hebbian rules, this very interesting neuroscientist, Mooming Poo, who now directs the Shanghai Institute of Neuroscience, found a phenomenon called spike-timing-dependent plasticity. Spike-timing-dependent plasticity is really a refinement of Hebb's rules. The natural synapse has the presynaptic cell firing before the post-synaptic cell. And so if you were Hebb, you would say, well, the natural course of events needs to take place pre-before-post. And under some circumstances, this so-called spike-timing-dependent plasticity holds. That is, if the Hebbian situation were pre-fires-before-post, but they both do fire, if that occurs, then the synapse gets potentiated. But in the anti-Hebbian situation, where the post actually fires before the pre, see here's the little post-synaptic potential, but the post-synaptic spike is coming first, this oftentimes, not all the time, but oftentimes produces LTD, long-term depression. And so if the experiment is done under certain circumstances, then one can embroider the Hebbian rules with additional timing rules that appear to make sense in terms of what the organism needs to know. So if the pre-fire is 5 to 30 milliseconds before the post, one gets LTP. If the pre-fire is 5 to 30 milliseconds after the post, one gets LTD. Another refinement on Hebb's rules or on the calcium hypothesis is supra-linear influx of calcium. So here is a case where the experimenters produced glutamate locally by uncaging photo-activated glutamate. They also produced glutamate locally with pre-synaptic stimulation. They asked about using calcium measurements, a calcium dye. They asked whether you could get a linear sum of the calcium influxes produced with those two simulations or a non-linear influx. And it turns out that you can get a greater than summed influx of the calcium. And so this is thought also to play a role in the important elements of synaptic plasticity. Again, by saying that calcium is the major intracellular messenger for synaptic plasticity. So I told you about insertion of new receptors. Now I've actually modified the picture that's in Candela a little bit here to make it clearer. Here again is the experimental setup where we have a hippocampus. We're stimulating the Schaefer collaterals in CA3, recording in CA1. And in early LTP, and what we're doing here is to record the quantal size, the amount of current produced by a single quantum, the miniature post-synaptic currents. In early LTP, which typically arises from more quantum being released, the quantal size does not change much. But in late LTP, we do have a vast increase in the quantal size, and that is thought to be a post-synaptic mechanism where the cell actually puts more receptors in post-synaptically. So as you can see, I've added a receptor here. In late LTP, you know why I'm going to skip this because I think it's wrong. We'll just skip it. So the overall view of LTP is fairly complex, although calcium is involved. This now is one of the figures in the book. It is practically the last figure in the main text of Candel. Synaptic plasticity is Candel's Nobel Prize. He likes to think about it, and he likes to go even beyond the experiments that gave him recognition. And so, although we all love Candel and think he's very important, sometimes when he's writing a book, he does go a bit beyond the experiments. So this is his view, Candel's view, of the sum of synaptic plasticity. There is involved in addition to the synapse here and the calcium influx, which activates Cal modulant. There is a retrograde signal that goes back to the presynaptic terminal. There needs to be a retrograde signal if synaptic plasticity is also going to increase the amount of transmitter release. This is typically thought to be nitric oxide, the gas, but in some cases, the opposite occurs with endocannabinoids. There is also dendritic protein synthesis, which we recalled is very important for the late stages of LTP, and that is thought to be due to RNA in the dendrites engaging ribosomes in the dendrites making new proteins, which can be inserted into the synapse. There is thought to be cytoskeletal changes that expand the spine. And finally, the longest stages of synaptic plasticity are thought to involve new activation of genes. So the longest stages actually do get back to the nucleus, it is thought. To activate proteins like CREB, we discussed CREB when we discussed the G-protein pathway, calcium and cyclic AMP binding protein, which activate new genes, produce new gene products, and strengthen synapses. So all of these very definite, very widespread effects are thought to play a role in memory and learning and in synaptic plasticity. And just briefly, how these two types of calcium are sensed through the NMDA receptor. Remember, we have a weak calcium influx causing LTP, sorry, a weak calcium influx causing LTD and a stronger calcium influx causing LTP. It is thought that there are actually two types of calcium receptors that look at the calcium in the post-synaptic density and respond to those calcium. I circled them on the map of the post-synaptic density. Both of them depend on CalModulin. One of them is CampKinase2, the other is CalCneurin. So, CampKinase2, I showed you a map of CampKinase2 in which it autophosphorylates itself. That's CampKinase2. And it then can also phosphorylate the subunits of the amper receptor. Apparently, the phosphorylated amper receptor has a larger current, so that's a post-synaptic event. And this is likely short LTP. And then in addition, CampKinase2 apparently activates a process of inserting new amper receptors, also a post-synaptic event. And possible that this takes place in development as well. So, CalCneurin, which responds to smaller influx of calcium in LTD, again, calcium flows in. CalCneurin is a protein phosphatase, also called protein phosphatase 2B. If those of you who are at the symposium around a little more than a month ago remember that I gave the crystal structure, a glass cube with the crystal structure of PP2B to one of my former post-docs, because he studies that molecule intensively. It is a calcium-calmodulated, CalModule-independent protein phosphatase. That is, it takes phosphates off proteins. And it regulates an inhibitor called phosphatase 1. And the result is that amper receptors get removed from the post-synaptic membrane. And so it's a convenient hypothesis, and there's probably some truth in it, that the direction of long-term changes, LTP, LTD, depends on the relative levels of activating CampK2 and CalCneurin, in turn, because of the relative amounts of calcium that flow in. And so this slide really gives you a little bit more detail. Calcium flows in, activates CalCneurin, regulates an inhibitor. I've told you this, so I won't bother going through it again. Are we done? No, we are not done. And the reason we are not done with the effects on synaptic plasticity is that Candela and others have reminded us that there are other types of long-term potentiation. The CA3-CA1 pathway is the most famous, and I've given you that one. There are also involvement of calcium channels in the perforant path synapse. And there's also an involvement simply of amper receptors in the mossy fiber pathway. So, for instance, in the hippocampus, each of the three different types of synapses has a different molecular route of potentiation and depression. It typically involves calcium, typically flowing in through receptors, sometimes through calcium channels. But we're not even done yet, because there are epigenetic changes that also are involved in synaptic plasticity. So, this again is reason work from Candela, and it's not clear whether this will stand the test of time or not. Epigenetic typically means changes in the structure of DNA or of chromatin. And so, the most famous example, so here we have DNA wound around nucleosomes. And the basic idea is that when DNA is available, not wound around chromatin, it can be activated, and when it is covered by other proteins, it cannot be activated. So, we begin with phosphorylation of CREB. This, again, this recruits a protein called CREB binding protein. This acetylates lysine residues on histones. So, here we have the DNA wrapped around histones. We get a post-translational change, which essentially release the DNA. We get DNA released from the histone. This allows transcription during the later phases of LTP. And so, the way that CREB is said to work is by allowing histones to be deacetylated and allowing DNA to be revealed. In addition, LTP changes DNA methyl transferases. So, now we're talking about methylation of DNA. So, DNMT is a DNA methyl transferase. This recruits methyl DNA binding proteins to the DNA, which can now see the methyl where they couldn't before. This recruits histone deacetylases, which remove the acetyl groups, and this allows the binding of another CREB binding called CREB2, which can repress transcription. So, as a result of calcium influx, activating CREB, either CREB1 or CREB2, we can have further changes in gene activation. Well, all I can say is that this has been a whirlwind tour of a enterprise that occupies many neuroscientists trying to understand the mechanisms of changes in synaptic transmission. And we'll stop here and I'll be available for office hours today and Friday. Jared, any announcements? Okay, thanks.