Learning and Memory

Humans adapt to new environments, and they do so by changing their behavior. How does this behavior change happen? Through learning, which, by changing the neural mechanisms underlying behavior, changes behavior. By definition, learning is the storage of information in the brain as a function of experience, resulting in a relatively permanent change in behavior. The storage of information referred to above is memory, which implies that learned information is encoded in the neural mechanisms of the brain in some way and can be retrieved when access of the learned material is necessary.

Types of memory: Long term memory vs short term memory:

1. Short-term (working) memory:

STM stores a very limited amount of memory for a very short period of time, usually no more than a few seconds. STM is used in comprehending, reasoning and problem solving.

2. Long-term memory: memory of the past. There appears to be no limit to its capacity in that everything that is and could be remembered is stored in LTM. LTM is undoubtedly produced by structural changes within the brain.

Declarative vs. nondeclarative memory:

1. declarative: things that one can bring to mind and declare or describe in words, such as the memory of general facts, concepts and knowledge and also the personal memory for events in one's past. All refer to 'what'.

2. nondeclarative: things than cannot be explained in any straightforward way, such as how to ride a bicycle (although they may be motoric skills, cognitive skills or perceptual skills). All these memories refer to 'how'.

Types of Learning:

1. non-associative: only a single stimulus or environmental event is involved. Examples of non-associative learning are habituation (decrease in behavioral response with repeated presentation of a stimulus) and sensitization (increase in behavioral response with a repeated stimulus).

2. associative: learning the relationship between two stimuli or between a stimulus and behavior so that animals can learn causal relationships (predictive as in classical conditioning or consequences as in operant conditioning). Our understanding of the causal relationship between events are constantly adjusted and refined. Thus associative learning provides the basis for behavioral adaptation by providing increasingly accurate internal representations of causal relationships in the external environment.

One way to approach the study of the neural basis of learning and memory is to look at individuals in whom only one kind of memory has been lost - declarative vs. nondeclarative, for example. If only one kind of memory is destroyed when brain tissue is destroyed, that suggests that different kinds of memory are stored in different places in the brain. Two famous cases of the loss of declarative memory (HM and NA), but not nondeclarative, suggest that these two processes are stored in different neural systems (or it could simply be that nondeclarative memory is simpler and more robust than declarative and thus survives better following brain damage).

H.M.: Amnesia is the loss of long-term declarative memory. H.M. was an assembly line worker suffering from intractable epilepsy. In an attempt to cure his epilepsy, a neurosugeon, William Scoville removed the medial portion of his left and right temporal lobes. This surgery thus removed two thirds of the hippocampus and all of the amygdala, entorhinal, perirhinal and parahippocampal cortex. While his seizures were remarkably and immediately improved (lessened), he developed a mild retrograde memory loss (loss of memory for events occurring before the surgery) and a profound and continuous anterograde amnesia (inability to form memories of events occurring after the surgery). Otherwise, H.M. shows no cognitive impairment, has retained an above-normal intelligences and adequate working memory. He has no impairment of non-declarative memory and can learn new motor skills.

We now know that these structures, the hippocampus and associated structures above serve as a temporary repository for newly learned information that will later be stored in other cortical regions. Thus, they hold information while memory consolidation is occurring.

Diencephalic brain lesions can also destroy memory. One patient, N.A. was a young 22 year old when in a mock duel a miniature fencing foil entered his right nostril and punctured the base of his brain, damaging the left dorsal medial thalamus. N.A. has severe anterograde amnesia, especially for verbally presented material, although he also shows superior intelligence. In contrast to H.M., N.A. can learn new declarative information, but only with great difficulty. N.A. can also learn non-declarative tasks with great ease, like H.M.

In Korsakoff's syndrome, caused by nutritional deficits due to acute alcoholism (but also head trauma in some cases) lesions are found in the walls of the 3rd and 4th ventricles, regions of the cerebellum, there is cortical atrophy and shrinkage, and also the dorsal medial thalamus and /or mammillary bodies are damaged. People with Korsakoff's syndrome also show profound loss of declarative memory.

Some experimental studies have shown that declarative and non-declarative memory can be dissociated, that is, you can show there is a different neural basis for each kind of learning.

Here is the study (Knowlton, Mangels and Squire, 1996). Subjects were required to learn associations between a set of four visual cues and two possible outcomes. This was presented as a game in which the subject learns to predict the weather so that the two outcomes were either sunshine or rain and the four cards were either more or less predictive for either outcome. Thus, each subject was shown 1, 2 or 3 of the cards on each trial and asked to make a prediction of sunshine or rain based on probabilities. Over time, people become more accurate at this task, while remaining unaware of what information they are actually using to make the prediction, thus this is considered to be non-declarative learning because it is not factual learning.

Three groups of subjects were tested on their ability to learn to assess these probabilities and make accurate predictions. One group had hippocampal, temporal or diencephalic lesions (amnesiacs), one group had Parkinson's disease (neural degeneration of substantia nigra which disrupts a major input to caudate and putamen) and the third group was matched, normal controls. The reason these two groups were chosen is that animal studies have shown that the hippocampus and related structures are necessary for declarative learning (spatial learning in animals), while nondeclarative learning (habit learning) is affected by lesions of the caudate and putamen.

Both the amnesiacs and normal controls learned the task equally well, so that at the end of training they were about 70% accurate (chance performance would have been 50%). The Parkinson's patients learned almost nothing, with about 55% accuracy, with a subgroup of the severest patients showing chance levels of accuracy.

Knowlton suggested that these findings show dissociation between declarative and nondeclarative memory in humans.

Subjects were also asked factual questions about the weather game after training was completed (declarative memory). The patients with Parkinson's disease remembered just as much about the procedure as did the controls, despite the fact that the controls had learned how to predict weather, while the Parkinson's patients had not. In contrast, the amnesiac group, could recall almost nothing about the factual procedures of the group, despite the fact they were as good at predicting as the controls.

(Learning task = non-declarative: Amnesiacs (hippocampal lesions) and normal controls equally as good. Factual questions about task = declarative: Parkinson (caudate-putamen) and control equally as good.)

This study is strong evidence of the separation of declarative and at least one type of non-declarative memory in the human CNS, and indicates that the limbic-diencephalic regions (declarative) and the caudate-putamen (non-declarative) support separate and parallel learning systems that are used by human beings to acquire different kind of knowledge about the world.

All the evidence we have suggests that physical changes in neurons maintain memory and that these physical changes alter the synapse. For STM, the synaptic alterations are brief, but for LTM, they must be either permanent or extraordinarily robust and stable. LTM thus must arise from protein synthesis at synapses. One way to study memory at the level of the cell is to look at very simple nervous systems in which individual neurons or synapses can be identified and labeled and examined.

Cellular Basis of Simple Learning:

Many simple invertebrates have a CNS that has only 10,000 - 100,000 cells in them. In addition, these cells and axons are often very large, so that it is possible to identify corresponding neurons from animal to animal, trace neuronal circuits and develop an understanding of how specific neuronal events control behavior. Many of these invertebrates are also capable of several forms of learning so that the cellular basis of learning can be investigated.

Eric Kandel has investigated two forms of learning, habituation and sensitization in the sea slug, Aplysia. Aplysia has a nervous system with about 20,000 neurons. It also has a large exposed gill and siphon, which are reflexively controlled. The gill is used for obtaining oxygen from the water and the siphon is a small spout above the gill that is used to eject seawater and waste. If the siphon or the mantle shelf that covers it is lightly touched, the animal defensively retracts both the gill and the siphon, protecting those organs from harm.

The circuitry mediating the defensive gill response is now almost completely understood, with the circuit consisting of 13 central motor cells and 30 peripheral motor cells. The peripheral motor neurons project directly to the muscles that produce the reflex movement. The central motor neurons receive input from about 48 sensory neurons located in the gill and siphon. In addition to the sensory and motor cells, there are several interneurons, which modulate the reflex. Excitatory input from the sensory neurons to both the motor neurons and the interneurons initiate the reflex. The strength of the response can be modified by three types of non-associative learning; habituation, sensitization and classical conditioning.


Habituation (a decrease in a response to a repeated stimulus) is seen when the siphon is repeatedly touched. The habituation is the result of a decrease in the amount of NT released at the synapses between the siphon sensory neurons and the motor neurons of the gill and siphon. This depression of synaptic output from the sensory neuron appears to be mediated by a decrease in calcium influx at the axon terminals of the sensory neuron.

Neuronal circuit of habituation:

Siphon >---------sensory neuron -----------< * interneuron (habituation I occurs at this synapse, with a I decrease in the amount of NT I released from sensory neuron) I I I I I I I----------------------------< motor neuron I I I gill

* = site of habituation


In Aplysia, sensitization (an increase in responding with a repeated stimulus) is demonstrated by presenting a noxious stimulus to the tail of the animal; after such a stimulus, defensive gill reflexes are enhanced - or occur again after habituation has occurred to a touch of the siphon.

The neuronal mechanism mediating sensitization of the gill reflex is at the synapses between the sensory neurons and their target cells. Sensitization results from an increased release of neurotransmitter at these synapses. Noxious stimulation of the tail activates a group of facilitator neurons that synapse on the axon terminals of the sensory neurons and act to increase the amount of NT through a process of presynaptic facilitation. One of the NTs released by the facilitator neurons and taken up by the sensory neuron is serotonin. The uptake of serotonin initiates a series of molecular events that results in increased release of NT in the circuit mediating the gill reflex. It appears that what happens is that some potassium channels are closed, so that fewer channels may be opened in the recovery phase of an action potential and thereby elongating subsequent action potentials arriving at that axon terminal. The elongated action potentials permit increased amounts of calcium to enter the neuron so that release of NT is enhanced.

Tail I I -------------- sensory neuron I facilitating interneuron I I I Siphon >---------sensory neuron -----------< * interneuron (habituation I I occurs at this synapse, with a I I decrease in the amount of NT I released from sensory neuron) I I I I I-------------- I I---------------------------**< motor neuron I I I gill

** sensitization occurs when, after habituation has occurred to the siphon touch (so that gill is no longer withdrawn), the tail is touched (or shocked). Now, when you touch the siphon again, it is withdrawn (sensitization), but not because NT is released in normal amounts from the sensory neuron connected to the siphon (that is still depressed from the habituation; *), but because of the excitatory NT that is released from the facilitory interneuron onto the axon terminal of the siphon. Thus, the motor neuron, in turn, is excited and the gill is withdrawn.

Cellular Basis of Complex Learning:

Most everyone agrees that information is acquired, stored and retrieved by the brain. Memory is a thing that happens in a place in the brain. However, the complexity of the brain precludes our understanding of how a memory trace, what is known as an 'engram', is formed. All brains consist of individual cellular elements. Most neurons have the same parts, and the majority of neurons communicate with each other across a synaptic space via neurotransmitters and neuromodulators. In human brains, billions of neurons interconnect in vast networks via even more billions of synapses. The brain accomplishes all of its remarkable activity though a network of neurons, thus, a single neuron is unlikely to encode a specific memory. Hebb (1949) increased our understanding of how networks of neurons might store information with the provocative theory that memories are represented by 'reverberating assemblies of neurons'. Hebb recognized that a memory so represented cannot reverberate forever and that some alteration in the network must occur both to make the assembly a permanent trace and to make it more likely that the trace could be reconstructed as a remembrance. Because neurons communicate with each other only at synapses, the activity of the assembly is most easily and perhaps only, altered by changes in synaptic function. Hebb formalized this idea in what is known as Hebb's Postulate "when an axon of cell A is near enough to excite cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A's efficiency as one of the cells firing B is increased". Hebb's Postulate is very close to a modern-day definition of what we call long-term potentiation.

The search for the cellular basis of declarative memory (memory of general facts, concepts and knowledge and also the personal memory for events in one's past) has focused on the hippocampus and related structures. Long-term-potentiation and long-term depression are terms for two distinct kinds of changes in neuronal synaptic efficiency that are assumed to underlie some types of human memory.

LTP is the strengthening of the effect of a presynaptic input on a postsynaptic cell as a consequence of using that synapse in a particular way. Book definition (page 582) is "A stable and enduring increase in the magnitude of the response of neurons after afferent cells to the region have been stimulated with bursts of electrical stimuli of moderately high frequency. Recent studies suggest that LTP is responsible for some important aspects of declarative learning.

The hippocampus is composed of two enfolded tissues, the hippocampus proper and the dentate gyrus. The hippocampus proper is also called Ammon's horn (or cornu Ammonis) (Ammon was an ancient Egyptian deity represented in Greek mythology by the horns of a ram, which curl in a manner similar to the structure of the hippocampus). It is divided into four separate regions, designated CA1, CA2, CA3, CA4, where CA signifies cornu Ammonis. The pyramidal cells of CA1 are particularly important in the study of LTP and learning in the hippocampus. The hippocampus proper is continuous with both the subicular complex and the entorhinal cortex, which in turn is adjacent to the neocortex of the cerebral hemispheres.

Input to the hippocampus originates in the adjacent entorhinal cortex, which receives information from many cortical association areas, thus providing the hippocampal formation with a rich source of data to construct memory from. The entorhinal cortex relays information to the granule cells of the dentate gyrus through projections called the perforant pathway.

In turn, the granule cells of the dentate gyrus project to the CA3 field of the hippocampus proper. Axons of the granule cells are the mossy fibers, and together, they form the mossy fiber pathway.

The pyramidal cells of area CA3 have axons that bifurcate or branch in two directions. One branch stays within the hippocampus and synapses on the pyramidal cells in area CA1; these axons form the Schaffer collateral pathway. The remaining branch of the CA3 axons leaves the hippocampus and terminates elsewhere.

Finally, the axons of the CA1 pyramidal cells project back to the entorhinal cortex by way of a neighboring structure, the subiculum. This completes the circuit of information flow within the hippocampal formation. All neurons in the hippocampal formation except those of the dentate gyrus, send axons to other regions of the cerebral cortex. Thus, the hippocampus is positioned to contribute to memory function in widespread regions of the brain.

LTP in CA1 Pyramidal Cells:

LTP (Bliss and Lomo) can be produced by applying a brief train of high-frequency electrical stimulation to any of the three hippocampal pathways: the perforant pathway, the mossy fiber pathway or the Schaffer collateral pathway. In each case, the excitatory synaptic response of the postsynaptic hippocampal neurons is markedly increased. Because this effect persists for considerable periods of time - often weeks or months - the name long term potentiation was picked.

At present, LTP at CA1 pyramidal cells seems the most interesting, because of the real possibility that it might be the mechanism by which at least some types of declarative memory are formed.

In the CA1 region, LTP can be produced at synapses between Schaffer collaterals and CA1 pyramidal cells, either in hippocampal brain slices or in intact animals. A bundle of Schaffer collaterals that provides input to the cell are used for stimulation. The post-synaptic cell is tested periodically with brief shocks to the Schaffer collaterals (fiber input into it) and the size of the excitatory postsynaptic potential is measured. Without specific stimulation to produce LTP, the response of the CA1 pyramidal cell does not change.

LTP can be induced by stimulating the Schaffer collateral fibers at a high frequency, such as 100 stimuli in a 1 second period (tetanic activation; representative of some naturally occurring phenomena in the brain).

The effect of this high frequency stimulation is to produce a profound depolarization of the CA1 pyramidal cell that - in turn- induces a change the cell's response to future input from the stimulated fibers. In the future, nontetanic activation of these same Schaffer collaterals will produce a larger excitatory postsynaptic potential, making a response by the CA1 cell more likely. Furthermore, the size of the CA1 cell's response to other inputs not active during the tetanus is unaffected. For this reason, LTP in CA1 hippocampal cells is said to be selective.

LTP in the CA1 pyramidal cells is also associative, meaning that it can establish a relationship between two of its inputs. Specifically, other inputs to the CA1 pyramidal cells that were activated during tetanus are also potentiated. LTP affects both the tetanic input and any simultaneously active nontetanic inputs. Thus, LTP provides a cellular mechanism in which two simultaneously activated inputs to the cell are associated, which is essential for associative learning.

Associative LTP in CA1 Pyramidal Cells:

Most excitatory synapses within the hippocampal formation - as in the brain more generally - use the amino acid glutamate as a NT. This single NT can have markedly different effects at different synapses, however, depending on the properties of the postsynaptic receptor system with which it binds. These postsynaptic receptors are categorized by the agonists with which they bind. The principal distinction for glutamate receptors is whether or not they bind with NMDA (N-methyl-D-aspartate).

At non-NMDA receptors (AMPA receptors), glutamate functions as an ordinary excitatory NT, controlling a simple ligand-gated ion channel. At these synapses, glutamate produces an excitatory postsynaptic potential that can lead to firing of the postsynaptic cell. Such AMPA receptors are responsible for all excitatory activity within the CA1 region; when they are blocked, all excitatory postsynaptic potentials disappear. In contrast, selective blocking of the NMDA receptors has minimal effect on postsynaptic excitation.

However, the binding of glutamate with NMDA receptors on the CA1 pyramidal cells is responsible for associative long-term potentiation. When CA1 NMDA receptors are blocked, all traces of associative LTP in CA1 disappear.

The NMDA glutamate receptor, which controls the entry of CA+ to the cell, is well suited for producing associative learning because it is doubly gated. Its activations requires that two criteria be met:

1. The membrane potential criterion is that the postsynaptic cell is substantially depolarized; this occurs when activation of the AMPA receptors or other excitatory receptors on the same neuron partially depolarize the membrane to less than -35 mV (EPSPs= -35 mV).

2. This partial depolarization removes the blockage of the calcium channels by magnesium, so that the NMDA receptors now respond actively to glutamate and admit large amounts of calcium though the channels. Thus the NMDA receptors are fully active only when they are gated by a combination of voltage and ligand. Both criteria must be fulfilled at the same time for the NMDA channel to open. For this reason, the NMDA receptor responds only to contiguously associated events.

LTP, then, is triggered by the opening of NMDA-calcium channels and the consequent increase in intracellular calcium levels. Increasing intracellular CA activates two related protein kinases (an enzyme that phosphoralates a specific amino acid of its target protein, thereby altering the function of the protein). These two activated enzymes are protein kinase C and calcium-calmodulin-dependent protein kinase II (CaMKII). Both kinases are essential for LTP, if either is blocked, LTP does not occur.

The ways in which these kinases act is now beginning to be understood. One major mechanism involves increasing the effectiveness of the postsynaptic AMPA receptors that are activated simultaneously with the NMDA receptors. When glutamate binds to the NMDA receptor, calcium enters the postsynaptic cell. Calcium ions switch CAMKII to its active form, which unleashes a biochemical cascade that permanently increases the responsiveness of AMPA glutamate receptors to the NT, by:

a) Increasing the conductance of NA+ and K+ at the receptor site

b) Moving AMPA receptors from the interior of the dendritic spine into the membrane, making more receptors available. By structurally altering the AMPA glutamate receptors, future inputs at these associated AMPA synapses will be enhanced.

c) There is also evidence of a presynaptic contribution to LTP. Protein kinases activated in the postsynaptic cell appear to release a retrograde messenger back into the synaptic cleft, which travels back to the presynaptic neuron. There, the retrograde messenger induces the presynaptic neuron to release more NT when that cell is activated in the future. One substance that may serve as the retrograde messenger is the gas, nitric oxide, which has neuroactive effects and can pass freely through membranes of both the pre- and postsynaptic neurons.

Thus, LTP in the CA1 pyramidal cells of the mammalian hippocampus is selective and associative. It is regulated by doubly gated NMDA receptors and initiated by synaptically produced increases of intracellular calcium in the postsynaptic cell. LTP requires both the protein kinase C and CaMKII be activated. Finally, LTP may result in both the postsynaptic modification of non-NMDA glutamate receptors and presynaptic changes in factors governing glutamate release at the modified synapses.

Insights from Transgenic animals:

A transgenic animal is one that has incorporated one or more genes from another cell or organism and can pass the altered gene on to its offspring. Thus, for example, transgenic animals can be used to study the effects that a particular protein plays in any biological process.

Transgenic animals are made in the following way:

1. first, the protein of interest is selected: for learning, CaMKII is a reasonable protein to select.

2. Next, the genes coding for the protein must be identified

3. the DNA sequence coding for that protein is altered, so that the mutant gene will create an altered protein when the gene is later expressed

4. Finally, the mutant gene is transferred to a fertilized egg.

Now, the new organism will manufacture the genetically altered protein whenever that gene is expressed and the altered gene has become a permanent part of the organism's genome. Animals such as these are often called 'knock-out animals' because one specific gene has been removed from the genome and its associated protein eliminated from the organism. Such transgenic animals have been used for a number of years to study the neurobiology of learning and memory - but there is one major drawback, and that is that gene expression is altered throughout the entire body, so that mutant organisms often show severe developmental defects and early death, and in surviving organisms, it is difficult to ascribe the effects of the gene alteration to any specific cell or tissue.

However, recently such problems have been solved by the development of technology that allows the researcher to delete a specific gene only in a specific cell type in a specific tissue. The initial application of this new technology was to examine spatial learning and the molecular biology of LTP in the CA1 pyramidal cells of the mouse hippocampus.

Disruption of spatial memory in transgenic mice:

Spatial learning is an important part of the behavior of rodents, since they need to move around their world to forage, find shelter and find mates; spatial information is stored by a form of explicit, fact-based learning that is similar to declarative learning in humans. Rats, for example, excel at spatial learning and quickly develop a cognitive map of their surroundings.

The hippocampus is known to be important for spatial learning in rodents. When it is bilaterally removed, rats and mice can no longer learn tasks that depend on spatial maps, such as the Morris water maze. The Morris water maze is a commonly used method to assess spatial learning in rodents. It consists of a small tub filled with mikl-colored water, which prevents the animal from seeing below the surface. Distant visual cues are placed around the pool to allow the animal to orient itself in space. Rodents are trained to find a hidden platform beneath the surface of the milky water, using the spatial information provided by the distant cues. Normal rodents learn such a maze quickly.

Non-spatial learning can also be tested by the same method, by what is called the 'landmark task', in which the platform is moved on every block of trials, but the platform's position is indicated by a clearly visible marker placed on the pool wall near the platform.

Several years ago, reports indicated that both spatial learning and CA1 LTP in mice were dependent on CaMKII. The gene for CaMKII was genetically altered, and although apparently normal in most respects, these transgenic mice showed no evidence of LTP in CA1 pyramidal cells. They also showed impaired spatial learning, as measured by the Morris Water Maze. Such evidence is consistent with the argument that hippocampal LTP is necessary for spatial memory in mice. But, because CAMKII was missing in every cell of the organism, other explanations could also be offered for the observed learning deficit. Thus, a strain of mice was created in which the NMDA receptor 1 gene was eliminated only in the pyramidal cells of the CA1 region hippocampus. As a result, the CA1 pyramidal cells in this strain of adult mice lack functioning NMDA receptors. They are called NMDAR1 CA1-KO or simply CA1-KO mice.

The CA1-KO mice differed both behaviorally and electrophysiologically. First and foremost, the CA1-KO mice were seemingly unable to learn the spatial water maze task. Both CA1-KO mice and a number of different groups of control mice were given 12 block of training in the Morris Water maze. At the end of the training, all animals were given a transfer test, in which the platform was removed from the pool. If the animals had formed a spatial map of the test apparatus, they should spend most of their time swimming in the portion of the pool where the platform was formerly located. This is precisely what the control mice did. In contrast, the CA1-KO mice swam around the entire pool, indicating they had not formed a spatial map of the platform's location.

Second, the CA1-KO mice showed no evidence of LTP in the CA1 region of the hippocampus when tested electrophysiologically. IN response to intense stimulation (100 stimuli in 1 second) of Schaffer collateral fibers, the CA1 pyramidal cells of the CA1-KO mice showed no evidence of LTP, whereas the response of the CA1 pyramidal cells in all control animals was potentiated as expected.

The CA1-KO mice were not impaired in other, non-spatial forms of learning. Although they learned somewhat more slowly than control mice, they were able to master the non-spatial, landmark version of the Morris Water Maze and equaled the performance of the control mice by the end of the experiment.

These findings give strong support to the theory that long-term hippocampal memory in rodents is produced - at least in part - by associative learning in the pyramidal cells of the CA1 region of the hippocampus. Spatial learning requires functioning NMDA receptors in those cells and is likely to be mediated by the same cellular processes that give rise to CA1 LTP.

Cortical plasticity:

One approach to learning and memory relies on the tendency of the cerebral cortex to form topographic maps of the information that it processes. A topographic map is a representation that preserves the surface structure and relations between its elements, at least to some extent. Thus, a visual cortical area contains a retinotopic map that preserves the spatial relations among retinal photoreceptors at the level of the cortex. Similarly, the primary and secondary somatosensory areas of the cortex are marked by somatotopic maps that preserve the local spatial relations among somatosensory receptors in their cortical representations.

In primates, both types of maps show large-scale distortions in their topographic mappings. In primate vision, the extreme variations in photoreceptor density between fovea and periphery necessitate a near-logarithmic global transformation between the retinal and cortical maps. In somatosensation, global discontinuities between the skin surface and the cortex necessarily result from the fact that the skin enclosing the body is a three-dimensional structure, whereas the cortical area containing the somatosensory map is a two-dimensional surface. But, in both cases, local spatial relations are maintained in the cortex, despite the presence of these global transformations.

Cortical maps provide an excellent means for studying learning because these maps are plastic, that is, they change as a result of experience.

Deafferentiation Studies:

Monkeys, like humans, have an orderly somatotopic map in the primary somatosensory cortex that devotes a considerable cortical space to the sensitive digits of the hand. This mapping can be demonstrated by recording from a large number of cortical cells using many closely spaced microelectrode penetrations and determining the receptive field of each cell.

These somatotopic maps are so consistent from animal to animal that it tempting to think of them as genetically determined and unlikely to changed by experience, but instead, the map appears to be dynamically maintained and subject to substantial reorganization. Here is one experiment: Deafferentiation (elimination of sensory input) was performed on the third (middle) finger. The cortical representation of the finger was mapped before deafferentiaition and after. Following surgery, the moneky's cortex had reorganized itself so that cortical neurons in the region that once corresponded to the missing finger now responded to input from the adjacent unamputated digits. The new map was orderly, but four-fingered.

The receptive fields of neurons in the finger areas of the primate somatotopic map are characteristically small, so that the information processed by such cortical neurons arises within a discrete, well-defined continuous region of the skin. The monkey hand, like the human, is dexterous and flexible with the separate fingers normally moving independently of each other. Thus, the sensory input to the individual fingers in usually uncorrelated. However, if two fingers are temporarily sutured together, the hand area of the cortex rearranges itself so that the receptive fields of individual cells grow larger and extend from one finger to the other. There is no explanation for this finding other than that sensory receptors in both fingers experience highly correlated patterns of temporal input after being joined. From the perspective of the monkey's cortex, there is no difference in the message being sent from touch receptors in the two fingers joined together.

Amputation and surgical suturing of fingers together are obviously not normal events in the life of a monkey. Thus, we must ask, during the normal course of events, does the cerebral cortex rearrange itself? In one experiment, monkeys were allowed to learn fine motor coordination. They could enjoy banana pellet hors d'oeuvres each day before mealtime, but these snacks were served in five 'cups' of varying size, the smallest of which required that the monkey make careful finger movements to extract the pellets from the container. Each of the 9 monkeys was given the opportunity to snack on 100 pellets distributed among the various cups on 3 days a week preceding normal feeding time, for a total of between 24 - 42 practice sessions. Training ended for each monkey when it could retrieve a food pellet from any of the cups on the first try.

Following training, the area 3b hand-maps for each monkey were measured. Cortical magnifications of the fingertips used by each individual monkey to remove the pellets were more than twice as large as those on untrained fingers. Cortical magnification is the ratio of the cortical area to receptor surface area; here, it was the ratio of the size of the cortical region devoted to each fingertip to the skin surface area of that fingertip. Similarly, the receptive fields for those digits shrunk, so that cortex was reorganized in such a way that provided them with a larger number of cortical cells each with smaller receptive fields, thereby increasing the precision of the somatosensory information available to the animal. Many other studies have corroborated this finding, including in humans.

Such results indicate that cortical plasticity is a normal property of everyday life.

Learning: New Synapses and New Cells:

How can such rapid and extensive reorganization of the cerebral cortex take place? There exist 3 possibilities:

1. the effectiveness of existing synapses between cortical cells could change with learning

2. new synapses are formed among existing neurons

3. new neurons might be created

Strengthening Existing Synapses:

Such processes are well documented, as in LTP. Thus, such changes might consist of an alteration in the efficacy of NT release; the speed of reuptake; the mechanisms that alter binding of the NT.

Creating new Connections:

Less is known about the formation of new synapses as a result of learning, but recent work suggests that learning may produce new synaptic growth that strengthens connections between neurons in the hippocampal LTP paradigm.

Over the years, many investigators have suggested that learning may proceed by both the remodeling of existing synapses and the formation of new synapses between neurons showing LTP. Recently, electron microscopy was used to examine the structure of synapses between neurons before and after the induction of LTP in a hippocampal slice preparation; and to identify the specific neurons actually involved in LTP within the slice by staining the cells for accumulated calcium (a sign of LTP) within the dendritic spines of synapses in the slice.

It was found that within 1 hour following the induction of LTP, there was a marked increase in the number of postsynaptic neurons within the hippocampal slice showing double-spine synapses (overhead), that is, synapses in which two, rather than one, dendritic spine extends from the postsynaptic cell to make contact with the axon terminal of the presynaptic neuron. This arrangement only occurred at synapses involved in LTP, as marked by heightened calcium levels within the postsynaptic dendritic spines. Thus, new synaptic growth was demonstrated between neurons activated by a physiological process, LTP, that is believed to be responsible for at least some forms of learning within the primate brain. Thus, the long-lasting changes in synaptic efficiency produced by LTP are likely to be the result of new synaptic spines increasing the connectivity between an activated axon terminal and its target cell.


Of the broad spectrum of diseases that can produce dementia, Alzheimer's disease is the most common. It is a devastating disorder characterized by progressive loss of memory and intellectual abilities, affecting more than 40% of people older than 85. It has been recognized as a clinical entity for nearly 100 years , but therapies were unavailable until about 20 years ago.


The primary pathological hallmarks of AD, described by Alois Alzheimer in 1906 consists of neuritic plaques and neurofibrillary tangles. Neuritic plaques are complex structures that consist of extracellular aggregates of the amyloid-beta peptide, surrounded by swollen dystrophic neuritis (abnormal growth of hairlike projections from neurons) and infiltrated by microglia and activated astrocytes (the resident immune systems of the CNS; these plaques are an inflammatory response). Neurofibrillary tangles are intracellular accumulations of hyperphosphorylated tau, a cytoskeleton-associated protein organized as paired filaments. Grossly, there is often atrophy of the frontal, parietal and temporal lobes. The medial temporal lobe structures, such as the amygdala, hippocampus and entorhinal cortex are usually markedly shrunken.

In addition to plaques and tangles, there are other changes in the brain. There are amyloid deposits in leptomeningeal blood vessel walls (leptomenengeal - pia and arachnoid), knowns as amyloid angiopathy, which results in an increased frequency of hemorrhage. Neuropil threads are short neuronal processes that are found mostly in cortical regions associated with tangle pathology. Hirano bodies are rod-like filaments composed of actin and other microfilaments, mostly occurring in the pyramidal neurons in the hippocampus.

Granulovaculoar degeneration is the appearance of large cytoplasmic membrane-delimited vacuoles found in the same regions. The vacuoles also contain cytosketon-associated proteins. All of these occur in normal aging, to a certain extent, but they accumulate to a much greater degree in the hippocampus, neocortex and other areas in AD.

There are also cellular and synaptic alterations. There is a loss of large neurons, particularly in layer II of the entorhinal cortex, pyramidal neurons in layers III and V of neocortex and cholinergic neurons in the basal forebrain. Loss of synapses is an important change because this pathology presumably results in the disconnection of cortical association areas, and synapse loss best correlates with cognitive deficits. Finally, there are changes within the cell machinery itself, with Gogi apparatus shrunken and an expansion of endosomal compartments. These changes occur very early in the disease, even at a preclinical stage.

The wide variety of pathological changes raises many questions about their sequence of occurrence. That is, in order to understand what causes the disease, we need to understand the primary changes versus secondary changes that occur as a result of the disease. This is currently unresolved.

Neurochemistry and Neuropharmacology:

The most consistent neurochemical change that occurs involves the loss of acetylcholine, serotonin and glutamate. Loss of serotonin occurs due to loss of serotonin neurons in the brainstem; the loss of acetylcholine occurs because of loss of neurons in the basal forebrain and hippocampus and the loss of glutamate seen is the direct result of the degeneration of cortical pyramidal cells.

Most of the evidence that we have collected indicates that the clinical symptoms we see are due to loss of transmission at the cholinergic synapses in the neocortex and the hippocampus. First, we know that presynaptic cholinergic markers are decreased in neocortex and hippocampus in AD, and this correlates with dementia severity. Second, basal forebrain neurons, which provide the majority of cholinergic innervation of neocortex and hippocampus, degenerate in AD. Third, basal forebrain lesions impair learning, memory and attention. Fourth, clinical trials of acetylcholinesterase (AChE; this is an enzyme that degrades acetylcholine) inhibitors show improved cognition and quality of life, reduced behavioral problems and delayed institutionalization.

AChE inhibitors are currently the primary symptomatic treatment of AD. It has been shown that complete disruption of the gene for this enzyme in mice results in overactivity of cholinergic systems in both CNS and PNS. Thus, the goal of AChE inhibitors in humans is to prevent the degradation of ACh once it is released from the surviving cholinergic neurons. However, the typical degree of AChE inhibition with available drugs is only about 30%; and while this limits their efficacy, at the same time, if any more inhibition was achieved, this would have a serious impact on cholinergic function in the gut, the heart and muscle function.

Since AChE inhibitors are limited by the above considerations, researchers have been working on finding drugs that will bind directly to the cholinergic receptors and thus mimic the action of Ach (this is an agonist). We know that there are ACh receptor subtypes (nicotinic and muscarinic), and the hope is that one particular subtype may prove to be the receptor that should be targeted with an agonist. Most studies have focused on a subtype family called the mAChR, because we know this receptor family has a role in learning and memory. Within this receptor family there are 5 distinct subtypes, M1 - M5, all encoded by different genes and all expressed in the brain. Several lines of evidence suggest that M1 may be the target subtype. It is the most abundant receptor in the cortex and hippocampus; it is post-synaptic on pyramidal neurons and has important signaling roles in enhancing the effect of glutamate at NMDA receptors and it activates protein kinase pathways involved in LTP. One of the first selective M1 agonists, xanomeline, showed significant clinical efficacy in a double-blind, multicenter study. Some of the most remarkable benefits were in the reduction of the agitation, delusions and hallucinations, providing evidence that the cholinergic system plays an important role in behavior as well as cognition (Current medications used are Tacrine (Cognex); Donepezil (Aricept); Rivastigmine (Execelon); Galantamine (Reminyl; side effects include nausea, diarrhea, abdominal cramping, anorexia).

Alterations in serotonin and glutamate also play a role in the symptoms of AD. Serotonin is implicated in mood and anxiety disorders, which commonly coexist with dementia in AD. Thus, selective serotonin reuptake inhibitors (SSRIs; inhibits the reuptake of serotonin into the presynaptic cell so it is available for use longer) are frequently used and have a demonstrated efficacy. Drugs that modulate glutamate seem to also have an effect at lessening symptoms of AD. For example, ampakines, a class of drugs that modulate the AMPA receptor are under development. In addition, memantine, a noncompetitive NMDA antagonist, is commonly used in Europe to treat patients with dementia and was approved for use as of 2004 in the U.S. Although other NMDA antagonists have considerable side effects, memantine seems to be well tolerated. Memantine seems to be most efficacious with those with a vascular dementia, and since cerebrovascular disease and AD often coexist, this seems to be a reasonable drug to use with AD patients. Memantine is used only for moderate to severe cases of AD, it seems to have little or no effect on early cases, In addition, it doesn't alter the course of the disease, it merely slows the course down. One effect of AD appears to be a kind of hyperexcitability of neurons via release of glutamate, this hyperexcitability is so great that it can lead to the death of the neuron. Memantine prevents this by binding to the NMDA glutamate receptor and reducing neuronal excitation. Since it is a weak agonist, it reduces excitability while still allowing synaptic responsiveness so that learning, memory and cognition can occur.

Risk Factors for AD:

Aging is the biggest risk factor for AD. The prevalence of AD doubles about every 5 years after age 65 and increases to about 40% by age 85. This has led some researchers to suggest that AD is simply a consequence of aging and is actually accelerated aging of the brain and that everyone would develop it if they lived long enough. However, the risk actually declines after 90, indicating that AD is not an inevitable consequence of aging. Thus, other risk factors must interact with age to produce AD. One risk factor that has been identified is that people who have sustained severe traumatic brain injury are at a higher risk for AD than expected. Head injury seems to interact with genetic predisposition to decrease the age of onset of AD. Down syndrome is another well established risk factor and virtually all Down syndrome patients who are older than 40 exhibit the neuropathology of AD, although clinically evident AD is less common; interestingly, the mothers of Down syndrome children are also at a higher risk of AD.

Epidemiological studies suggest that AD is more common in women and that postmenopausal estrogen replacement is protective for AD; there is experimental evidence that estrogen modulates the production of the amyloid-beta peptide. Currently, the role of gender in AD pathology remains to be clarified, since it is not clear why women may be at higher risk.

It appears that lifestyle factors, such as dietary habits, physical activity and leisure activities also play a role in AD. Subclinical deficiencies of vitamin B6, B12 and folate are associated with poorer cognitive function in the elderly and AD patients tend to have even lower levels of these vitamins. These vitamins are essential components in the generation of neurotransmitters and myelin. Alternatively, or additively, deficiencies in these vitamins may be related to the increases in homocysteine, a metabolite that causes vascular and ischemic damage in the brain; beginning several years before diagnosis, patients who develop AD have elecated homocysteine levels. AD is also linked to high fat consumption, namely total and saturated fat and lower levels of fish oil (which is polyunsaturated fat).

Decreased leisure activity is also a risk factor for AD, although it is possible that this decrease is actually an early sign of subclinical cognitive impairment.

Uncertain Risk Factors:

The effects of smoking and alcohol consumption are unknown at this point. Heavy alcohol users develop a syndrome that is similar to dementia, but moderate consumption of red wine is associated with lower risk of AD. Moderate alcohol use is linked to improved cardiovascular health, which may account for this effect. The picture about smoking is less clear, with some studies showing an increased rate of AD with smoking, and other a decreased rate (perhaps due to the stimulatory effects of nicotine at the nicotinic Ach receptors).

Preventative Factors:

Vitamins C and E are antioxidants and are associated with lower levels of AD, and one clinical study found that vitamin E delays disease progression.

Several studies have suggested that education and early cognitive and linguistic abilities are associated with lower risk of developing AD. In addition, participation in activities that require cognitive ability over the lifespan are also associated with a lower risk. One explanation may be that people who are more intelligent and who use their cognitive abilities, have a kind of "cerebral reserve" and that they don't show symptoms until there is more damage than would be true for those who don't use their cognitive abilities. In one study of AD in nuns, linguistic ability early in life strongly predicted cognitive ability in old age. This suggests the possibility that developmental mechanisms or early education may play a protective role for AD in later life. A study in Scotland showed that low performance on childhood school tests was associated with increased later dementia.

Epidemiological studies have also linked lower rates of AD with patterns of drug usage, particularly exposure to estrogen, nonsteroidal anti-inflammatory drugs (NSAIDs such as asprin, ibuprofen (Advil & Motrin), Tylenol, Celebrex, Vioxx, Bextra), and statins (3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors; used in cardiac patients to lower lipids). Depending on the study, the reduced risk for all of these drugs is somewhere in between 30 - 80%. Double-blind, placebo-controlled studies trying to show a causal effect of estrogen and NSAIDs on AD have thus far been negative, and because all of these drugs can have serious side effects (NSAIDs: gastrointestinal upset, toxic effects on liver; estrogen increases risk of breast cancer;), to date, none of them are recommended as either a prevention or treatment of AD. However, the associations seen implicate hormones, inflammation and cholesterol as playing some role in AD. Currently, studies are under way to determine how effect NSAIDs or COX-2 inhibitors (which do not have the same side effects as NSAIDs but are anti-inflammatory; they are asprin-like analgesics such as Celebrex, Vioxx and Bextra; they work by decreasing the production of prostaglandins, a hormone involved in producing inflammation) are underway (recently COX-2 inhibitors have been pulled from the market because of very serious side effects such as heart attack).


To date, four genes - APP (amyloid precursor protein), PS1 (presenilin-1), PS2 (presenilin -2) and APOE have been consistently linked to AD in studies around the world. There are two major presentations of the disease; familial and sporadic. About 10% of AD cases are familial, mostly with early onset (before age 65). In these families there are mutations in three distinct genes - APP on chromosome 21; PS1 on chromosome 14 and PS2 on chromosome 1. The PS1 mutations account for most of the familial cases with early onset. The other 90% of AD cases are sporadic, in that they do not have a definite family clustering, but in these cases, careful family histories suggest that relatives have been affected. Only one gene, APOE is associated with risk in all populations. APOE exists as three allelic variants (2, 3 and 4) and it is the inheritance of the APOE4 allele that confers the greater risk of AD. Individuals who are homozygous for APOE4 have a substantially greater risk of developing AD than those with no APOE4 alleles, or those with only one (who have an intermediate risk between the other two).


The cause of most cases of AD remains unknown. Postmortem observations that include changes in neurotransmitter systems, accumulation of amyloid plaques, intraneuronal neurofibrillary tangles, inflammatory changes, increased oxidative stress, mitochondrial dysfunction have spawned numerous theories about etiology. It is likely that many of these observations reflect different aspects of disease initiation and progression, and one of the key goals in AD research is to develop a unifying theory that accurately identifies the triggering event or events. One we know this, then therapeutic strategies can be developed for all stages of the disease. The hypothesis that has gained the most support in recent years for explaining the etiology of AD is the amyloid cascade hypothesis, but it is not yet proven. However, it is used as a rationale for developing and testing therapies in clinical trials.

Since its identification as the primary component of senile plaques, the amyloid beta peptide and its precursor molecule, amyloid-beta precursor protein (APP) have been the targets of intense research interest. The amyloid hypothesis of AD proposes that amyloid production and accumulation plays the central role in initiating a cascade of events that leads to cellular dysfunction, neuron loss, and eventual clinical disease manifestation. Downstream effects might include cytoskeletal dysfunction, inflammation, oxidative injury, neuronal apoptosis (death) and other mechanisms that result in neurodegeneration, but the hypothesis proposes that the critical triggering event is the abnormal amyloid-beta metabolism. Thus, interventions that reduce the production and deposition of amyloid may provide an effective means of ameliorating symptoms, halting progression, and possibly preventing disease. The amyloid hypothesis is so well accepted that almost all current efforts to develop new therapies directly or indirectly target this mechanism.

Production of amyloid-beta from APP is regulated by three enzymes and predominantly, the peptide is 40 amino acids in length, but a slightly longer form that is 42 amino acids in length can also be produced, and it is this longer form of the peptide that seems to provide a platform for additional amyloid-beta deposition and plaque formation. Pathological examination of brains of those who died in early AD show depositions of amyloid-beta 42 in diffuse plaques, while those who died in the later stages showed depositions of amyloid-beta 40 in mature, dense plaques.

Of perhaps equal interest is the clearance of amyloid-beta peptide. While we don't fully understand these mechanisms, in transgenic mouse models of AD, the administration of thiorphan significantly reduced plaque in the brains of those animals (thiorphan would increase the amount of zinc mettalopeptidease in the brain, and we know that at least some of these degrade amyloid-beta peptide). There is also evidence that there are abnormalities in the lysosome function in neurons, and these changes are some of the earliest neuropathological changes seen in AD brains.

The appearance of dense deposits of amyloid in senile plaques suggests that they are static, but there is evidence from transgenic mouse models of AD that deposits of extracellular amyloid can change rapidly, that is, there can be rapid increases in size. Vaccination to induce production of anti-amylolid-beta antibodies has shown a dramatic ability to prevent or reverse the accumulations of amyloid plaques in animal models and this suggests that vaccination for immunity to amyloid-beta may be a viable therapy. However, a recent clinical trial employing this approach was halted because some patients developed encephalitis; but promoting clearance of amyloid-beta remains a potentially fruitful area of research. Additionally, it has been noted that as people age, the turnover of CSF slows, part of the job of CSF is to clear by-products of metabolism, such as amyloid-beta. One recent study showed that the turnover of CSF in AD patients was very much slowed, in comparison to age-matched controls, and this slowing of the turnover of the fluid bathing the brain may also facilitate deposition of amyloid-beta. Other recent studies indicate that the effects of estrogen, NSAIDs and statins may change amyloid-beta metabolism, so their protective effects may be due to their effects on this peptide.

Another possible etiology:

Oxidative injury: The brain consumes large amounts of oxygen and is vulnerable to oxidative damage done by free radicals, which are normal products of cellular metabolism (the major source of oxygen-containing free radicals is incomplete reduction of oxygen by the mitochondria). There are a number of molecules that can be formed (superoxide (02.-); hydroxyl (OH.); and other molecuesl such as hydrogen peroxide (H202) and peroxynitrite(OONO-) can lead to the production of free radicals). These molecules are called reactive oxygen species (ROS), and problems occur when the production of ROS exceeds the ability of cells to defend themselves against these substances. This is called oxidative stress, and can cause cell damage via damage to lipid membranes, proteins and DNA, and ultimately cause cell death.

Numerous studies have shown higher levels of oxidized lipids, proteins and DNA in brains of AD patients and the damage is highest in those areas most heavily affected by AD (neocortex and hippocampus) and lowest in those areas that are spared (cerebellum). Neurofibrillary tangles and amyloid plaques show signs of oxidative injury. Oxidative damage may play a central role in the formation of the tangles and plaques by promoting protein linking and aggregation. Since oxidative injury increases with age, this may explain the association of AD with age. Increasing evidence suggests that oxidative damage is an early event in AD. In AD brains, there are increased amounts of a marker for recent oxidative damage (8-OHG; 8-hydroxy-guanosine). In contrast, neurons with neurofibriallary tangles have low levels of 8-OHG, implying that oxidative injury precedes tangle formation. In transgenic mouse models of AD, mice expressing a mutant form of APP show increased lipid peroxidation months before amyloid deposits are detected. Potential sources of early oxidative stress include mitochondrial dysfunction and changes in cytoplasmic transition metals such as copper and zinc.

The role of oxidative stress in AD has implications for both disease diagnosis and treatment. Increased levels of oxidized lipid metabolites are detectable in CSF, blood serum and the urine of patients with AD. Levels of one metabolite, an isomer of prostaglandin F2 are elevated in the CSF of patients with mild cognitive impairment, a condition that progresses to AD in 50% of cases. Thus, detection of increased oxidative stress may be useful for the diagnosis of existing or future AD. Antioxidants also play a role in AD therapy, since vitamin E has been shown to delay disease progression by 6 - 12 months. Studies to determine if vitamin E can prevent the development of AD in patients with mild cognitive dysfunction are under way.

Food that are preventative do so because they are anti-oxidants:3 - 4 glasses of red wine a day will reduce risk by 80%, compared to those who drink less or do not drink at all; Vitamin E & C, green tea; Ginkgo biloba; ; blueberries, strawberries, spinach.

Cognitive Enhancers

5 drugs used in AZ: (4 are cholinesterase inhibitors: donepezil, rivastigmine, tacrine, and galantamine; prevent the breakdown of acetylcholine to its component parts (acetylCoA and choline). The fifth drug, memantine is an uncompetitive antagonist to the NMDA receptor that is thought o help patients with Alzheimers by limiting calcium entry into neurons through the NMDA receptor.

This paper (Disterhoft and Oh, 2006, J of Physiology-Paris, Vol 99 (2-3)) is concerned with studying compounds that are thought to aid in learning and memory and is using behavioral tasks as the dependent variable. In particular, they are using behavioral tasks that depend on the hippocampus for their acquisition. The hippocampus has been shown to be critical to the formation of declarative memory (facts, concepts, personal memory of the past). In experimental animal work, tasks that depend on spatial or temporal learning are though to be declarative. One task that can be used is the trace eyeblink conditioning task. The subject is presented with a neutral, conditioning stimulus (such as a tone), that does not elicit an eyeblink prior to conditioning. The stimulus is follwed by a temporal gap before an unconditioned stimulus (usually a puff of air to the cornea) is presented. This is a simple classical conditioning task, but is difficult for both aging humans and non-humans to learn (that the tone predicts the puff of air, so one should close one's eye to prevent the puff of air on the eye). It is difficult to learn because of the temporal gap of time between the tone and the puff of air. Over half of aging rabbits and rats used in this experiment cannot learn it; and those that did so, acquired it slower than younger animals. An intact hippocampus has been shown to be necessary in order to bridge the temporal gap between the tone and the puff of air.

The hippocampus pyramidal neurons can be kept alive in vitro in the laboratory and what is called the post-burst afterhyperpolarization (AHP) is studied. AHP is a post-synaptic membrane property of a neuron that limits the firing of action potentials during a long period of membrane depolarization caused by high frequency sustained excitation of the neuron. This is mediated by the influx of calcium that activates an outward potassium current. AHP is greatly enhanced in the hippocampus of aging animals as compared to the young. Thus, enhanced AHP in aging subjects makes them less excitable and limit those neurons from firing at high, sustained rates.

The authors have examined the relationship between learned behaviors and AHP in a variety of contexts. They initially reported that AHP was reduced in CA1 hippocampal neurons from rabbits that learned the delay eyeblink conditioning task as compared to those from nave rabbits. They suggested that this reduction in the AHP would increase neuronal excitability and should contribute to the increased firing rate of single CA1 pyramidal neurons that had been observed in vivo during eyeblink conditioning. These findings have been extended to trace eyeblink conditioning in rabbits and rats. Reductions in AHP were observed in CA1 neurons from aging animals that were trained and learned the trace eyeblink conditioning task, such that the AHP and accommodation observed in these neurons were nearly identical to that observed in CA1 neurons from young animals that learned the task. However, aging animals that received more paired training trials but did not learn, showed no reduction in the AHP as compared to nave animals. These findings support the working hypotheses that the AHP of CA1 neurons from aging rabbits are potentially "plastic" and may be reduced; and that aging rabbits that are trained but fail to acquire the task may have neurons with too large an AHP to allow learning to occur and/or reduced capacity for reducing the AHP. This hypothesis is presumably relevant to explaining the mechanism for age-associated learning deficits in other mammals, including humans.

Nimodipine, a calcium channel antagonist

Nimodipine is an L-type calcium channel antagonist that readily crosses the blood brain barrier. It has been demonstrated in rabbits to enhance blood flow in the brain as a direct consequence of vasodilation. In double-blind clinical trials, nimodipine was found to be beneficial in improving the cognitive deficits observed in elderly patients with dementia. Therefore, we investigated the potential benefits of calcium channel blockade, with nimodipine, in reversing the age-related learning deficit on the trace eyeblink conditioning task.

Nimodipine reversed the age-related learning impairment of aged animals on the trace eyeblink conditioning task. Treatment with nimodipine allowed the aged rabbits to learn the task at a very similar rate as young rabbits. No significant impact was observed in young animals treated with nimodipine. The behavioral rescue in aged animals may in part be due to the enhanced neuronal activity of hippocampal pyramidal neurons in vivo, as administration of nimodipine greatly enhanced the basal firing rate of CA1 pyramidal neurons. Thus, we examined the effects of nimodipine on the biophysical properties of CA1 hippocampal pyramidal neurons in vitro.

Nimodipine significantly enhanced the neuronal excitability of CA1 neurons in vitro by reducing the post-burst AHP of these neurons. Therefore, the behavioral effects of nimodipine may be in part be due to the enhanced activity of hippocampal pyramidal neurons via reduced calcium entry through the L-type calcium channel, which led to the significant reduction of the normally enlarged post-burst AHP of CA1 neurons in aged animals.

Metrifonate, a cholinesterase inhibitor

Metrifonate is an organophosphate which produces the long lasting inhibition of both acetylcholinesterase (AChE) and butyrylcholinesterase. Treatments with metrifonate have reversed the behavioral deficits observed in acquiring passive and active avoidance, Morris water maze, and radial-arm maze tasks by normal aging, medial-septum lesioned, or scopolamine-treated subjects. More importantly, in double-blind clinical trials, the cognitive impairments observed in Alzheimer's disease patients were alleviated by metrifonate-treatment. Thus, we were interested in observing metrifonate's effects on acquisition of trace eyeblink conditioning in aging rabbits and on biophysical properties of CA1 hippocampal pyramidal neurons in vitro.

Chronic, oral treatment with metrifonate ameliorated the learning deficit observed in aging rabbits. This amelioration was dependent on AChE inhibition. However, the memory of the task was not dependent on the AChE inhibition, as the aging animals that learned the task still remembered the CS-US association even when the treatment with metrifonate was stopped and the AChE activity returned to basal levels This demonstrated that modulation of cholinergic transmission was essential for learning the task, but not for retrieval of the learned association. Interestingly, the level of steady-state AChE inhibition (40-60%) necessary for the amelioration of the age-related learning deficit was achieved after 3 weeks of metrifonate treatment. This gradual buildup to the final, target range of ChE inhibition was found to be beneficial and necessary for behavioral improvements in humans.

Bath application of metrifonate dose-dependently reduced the AHP of CA1 hippocampal pyramidal neurons from both young and aging animals. These reductions of the AHP and accommodation were effectively reversed by addition of a muscarinic receptor antagonist, atropine, to the perfusate; suggesting that the reductions of the AHP and accommodation involved modulation of muscarinic cholinergic transmission.

A key question that needed to be addressed was 'could the biophysical state of CA1 neurons be altered after 3 weeks of metrifonate treatment that may have led to the behavioral amelioration of the learning deficit observed in aging rabbits?' Thus, we treated aging rabbits with either metrifonate or saline for 3 weeks, after which the biophysical properties of CA1 neurons from these animals were compared. We found that CA1 hippocampal pyramidal neurons from aging rabbits chronically treated with metrifonate had significantly reduced spike-frequency accommodation as compared to that from vehicle-treated rabbits. Surprisingly, the accommodation of CA1 neurons from chronically metrifonate-treated aging rabbits was similar to that observed in neurons from nave, young rabbits. Thus, it appears that 3 weeks of metrifonate treatment produced a steady-state inhibition of cholinesterase activity which, among many things, altered the biophysical properties of CA1 pyramidal neuronal of the aging subjects to a 'young' like state. This change may have enabled these metrifonate treated aging animals to learn the trace eyeblink conditioning task like young animals.

CI-1017, a muscarinic receptor agonist

CI-1017 is a muscarinic agonist designed to selectively activate the M1 receptor. The direct stimulation of the muscarinic receptor may be beneficial, because it does not depend on the presence of endogenous acetylcholine in the brain for action like the cholinesterase inhibitors. In addition, post-mortem examination of brains from Alzheimer's disease patients revealed that muscarinic acetylcholine receptors appear to remain intact although they may not be all functional.

Animals given CI-1017 improved their performance on the Morris water maze task and on a continuous performance task. In contrast, animals given the M1 antagonist pirenzepine were impaired on inhibitory avoidance, water maze and working memory tasks, and representational memory. Furthermore, a muscarinic receptor antagonist, scopolamine, impaired acquisition of the trace eyeblink conditioning task. These data suggest that modulation of M1 receptors impacts learning and memory. Thus, we were interested in observing CI-1017's effects on acquisition of trace eyeblink conditioning in aging rabbits and on biophysical properties of CA1 hippocampal pyramidal neurons in vitro.

CI-1017 ameliorated the learning deficit observed in aged rabbits. It significantly increased the rate and amount of learning without any evidence of pseudoconditioning. Thus, our data suggest that CI-1017 acts on associative sites to increase the probability of evoking a CR, and not on unconditioned reflex sites.

CI-1017 also enhanced the excitability of CA1 hippocampal pyramidal neurons from young and aging naive rabbits; via reductions of the AHP. The AHP reductions were reversed with addition of either atropine, a muscarinic receptor antagonist, or pirenzepine, a selective M1 muscarinic receptor antagonist, to the perfusate. These results suggest that M1 agonists ameliorate age-related learning and memory impairments at least in part by reducing the AHP of hippocampal pyramidal neurons, and that M1 agonists may be an effective therapeutic compound for reducing the cognitive deficits that accompany normal aging and/or Alzheimer's disease.

Galantamine, a cholinesterase inhibitor and an allosteric potentiating ligand

Galantamine is a third generation cholinesterase inhibitor. Galantamine also potentiates the activity of nicotinic acetylcholine receptors (nAChR) and, thus, is called an allosteric potentiating ligand of nAChRs. Galantamine treatment reversed learning impairments observed after various insults to the brain: nucleus basalis magnocellularis lesion, ischaemia, and ACh deficit due to prolonged alcohol treatment. Galantamine treatment also facilitated the acquisition of the delay eyeblink conditioning task in aging rabbits. More importantly, Alzheimer's disease patients treated with galantamine had their disease progression temporarily reversed and slowed as compared to placebo treated patients.

We observed that galantamine ameliorated the age-related learning impairment on the trace eyeblink conditioning task in rabbits. Galantamine treated aged rabbits met the learning criteria of 8 CRs in a 10 trial block much quicker than the vehicle treated age-matched controls, requiring a similar number of trials as the young rabbits. Additionally, the properties and timing of the eyeblink response resembled those of young rabbits, which differed significantly as compared to age-matched controls. These data suggest that the learning deficits associated with decreased cholinergic transmission in the aging brain is offset by enhancing nicotinic and muscarinic transmission with galantamine treatment.

Preliminary findings from our laboratory demonstrate that the post-burst AHP of CA1 pyramidal neurons from young and aging rabbits are reduced with bath application of galantamine. Atropine, a muscarinic agonist, significantly reversed these reductions. Additionally, galantamine significantly enhanced the excitatory post-synaptic potentials (EPSP) measured by Schaffer collateral stimulation. The amelioration of learning deficit observed in aging subjects with galantamine treatment may in part be due to the enhanced post-synaptic neuronal excitability via muscarinic receptor activation, as well as, the enhanced synaptic transmission via nicotinic receptor activation.

Knocking out BACE1 improves learning in the Tg2576 Alzheimer's mouse model

The amyloid-beta hypothesis of Alzheimer's disease has recently driven the search for the cure of the disease. It has been suggested that soluble, rather than insoluble, amyloid-beta is the most important pathogenic factor in Alzheimer's disease, as behavioral deficits precede amyloid-beta plaque formation in mice genetically engineered to overexpress the human form of the disease. Within the past several years, the APP cleaving enzyme 1 (BACE1) has become the primary therapeutic target candidate. Thus, in collaboration with Robert Vassar and colleagues, we examined the potential rescue of behavioral deficits observed in Tg2576 animals by eliminating the function of BACE1 by knocking it out in these animals.

Deletion of BACE1 prevented the behavioral deficit observed in the Tg2576 animals. The bigenic animals with BACE1_/_ and Tg2576 (BACE Tg2576) performed indistinguishably from their wild-type littermates in the hippocampus-dependent social recognition and spontaneous alteration Y-maze tasks; whereas, the Tg2576 animals were severely impaired. The behavioral rescue was corroborated with ELISA assays that showed the levels of amyloid-beta were very similar between the wild-type littermates and the BACE Tg2576 animals. Thus, we demonstrated for the first time that lowering amyloid-beta levels by inhibiting BACE1 is beneficial for AD-associated memory impairments mediated through the hippocampus.

It has been suggested that amyloid-beta can inhibit cholinergic signal transduction independent of apparent neurotoxicity. Additionally, as reviewed above, the cholinergic agonist CI1017 has been demonstrated to facilitate learning and enhance neuronal excitability, via reduction of the AHP. Thus, we examined the biophysical properties of CA1 neurons from BACE Tg2576 animals to cholinergic stimulation by bath applying carbachol, a cholinergic agonist. The capacity for post-synaptic plasticity evidenced by reduction of the AHP in CA1 pyramidal neurons was rescued in the BACE Tg2576 animals. Carbachol significantly reduced the slow AHP in CA1 neurons from BACE Tg2576 and wild-type litter mate animals as compared to those from Tg2576 animals. Therefore, the behavioral rescue through BACE1 knockout in Tg2576 animals may be due in part to the restored capacity for post-synaptic neuronal excitability increases in hippocampal neurons via cholinergic modulation. We assume that such a process is an important component of the cellular changes that occur during learning in wild type animals.

Concluding statement

The results from our work led us to formulate this working hypothesis: the enhanced excitability of hippocampal pyramidal neurons, via reductions in the slow AHP and accommodation, are important cellular changes that underlie hippocampus-dependent spatial and temporal learning. In support of this hypothesis, we have demonstrated that a transient, but not permanent, reduction of the AHP and accommodation is observed in hippocampal pyramidal neurons of animals that learned a hippocampus-dependent task. The post-burst AHP in CA1 hippocampal pyramidal neurons from aged animals is significantly enlarged as compared to that in neurons from young animals. Compounds that enhance excitability of CA1 neurons (by reducing the post-burst AHP) ameliorate the learning impairment observed in normal aging animals. It is very interesting to note that chronic metrifonate treatment altered the biophysical properties of CA1 neurons from aging rabbits to that usually observed in neurons from young, untreated animals. It is likely that alterations that we have observed in CA1 hippocampal pyramidal neurons with the cholinergic compounds are also occurring in other pyramidal neurons throughout the neocortex. Furthermore, these compounds have been demonstrated to be beneficial in alleviating the cognitive deficits observed in patients with Alzheimer's disease.

The alteration of post-synaptic membrane properties (more specifically, the relationship between AHP and learning) is also being pursued in other laboratories. Reduction of the AHP in piriform cortical neurons has been demonstrated. Recently, it has been shown that there is a direct, inverse relationship in aged rats between the size of the post-burst AHP of CA1 neurons and performance on the water maze task. They found that aged animals with pyramidal neuron AHP amplitudes that are similar to that of neurons from young animals were able to learn the location of the hidden platform; whereas, animals with significantly large pyramidal neurons slow AHPs were unable to learn the task. Thus, the post-burst AHP in hippocampal pyramidal neurons may be a cellular property that controls the capacity to learn a hippocampus-dependent task.

In this abbreviated review, we discussed our experiences with pharmacological compounds and a genetic manipulation that ameliorate the learning deficit observed in normal aging and in a transgenic animal model of Alzheimer's disease. A common denominator between normal aging and the transgenic AD model is the altered biophysical properties of hippocampal pyramidal neurons of these animals as compared to their respective counterparts (young and wild-type littermates). It remains to be seen if the rescue of transgenic animals with BACE knockout prevents the age-related enlargement of the post-burst AHP. The post-burst AHP is an indicator of cellular excitability. There are various neurotransmitter/neuromodulators that have been shown to alter it; in addition to numerous biochemical pathways. Understanding how the AHP is modulated by these pathways will help us come closer to understanding the cellular events that take place during learning. This may, in turn, help engineer future therapeutics to ameliorate the various learning deficits associated with normal aging and dementia. _

Effect of FK960, a Putative Cognitive Enhancer, on Synaptic Transmission in CA1 Neurons of Rat Hippocampus JOSEPH P. HODGKISS and JOHN S. KELLY

During the neuronal attrition, which underlies Alzheimer's disease, there is a progressive loss of neuronal projections to the cortex and hippocampus. The loss of cholinergic projections is marked and has been the subject of many studies. At present, no cure is in prospect and palliative treatment is aimed at up-regulating the function of the reduced pool of available neurons.

In the brain, as in the periphery, it is possible to enhance synaptic transmission either presynaptically, by increasing the amount of transmitter released, or postsynaptically, by modulating the behavior of the postsynaptic receptors or reducing the breakdown of transmitter.

In Alzheimer's disease, there is a loss of choline acetyltransferase, the enzyme involved in acetylcholine synthesis; thus treatments have concentrated on up-regulating cholinergic transmission in the brain, particularly by inhibition of cholinesterase.

FK960 has been shown to reverse scopolamine-induced cognitive deficits in rats in vivo, to increase the magnitude of long-term potentiation in guinea pig hippocampus in vitro, and improve visual recognition memory in primates.

These studies did not indicate whether FK960 acts pre- or postsynaptically nor which neurotransmitter is involved.

A role for glutamatergic transmission in memory has been advanced by studies with AMPAkines, such as BDP-12 and related compounds that enhanced memory, increased the degree and duration of long-term potentiation, and the amplitude and duration of the field EPSP in the hippocampus by changing the characteristics of AMPA receptor desensitization and/or deactivation.

The present study shows FK960 to enhance transmitter release at AMPAergic synapses on CA1 neurons in the hippocampus and raises the possibility that FK960 acts on the nerve terminal to increase transmitter release.

Male rats were decapitated, the brain was removed, and CA1 isolated by trimming the brain. Intracellular recordings (EPSPs) were made from CA1 pyramidal neurons. EPSP amplitudes and slopes were measured before exposure to FK960 and after exposture to either FK960 in the pretreatment solution or the pretreatment solution to which no FK960 was added.

Discussion This study demonstrates that the increase in EPSP amplitude and slope seen in hippocampal CA1 neurons following exposure to FK960 can be accounted for by an increase in transmitter release.

The views and opinions expressed in this page are strictly those of the page author. The contents of this page have not been reviewed or approved by the University of Minnesota.