SUMMARY

At present, long-term potentiation (LTP) of synaptic transmission is the leading neurophysiologic model for learning and memory processes, despite controversial results regarding its behavioral correlates. The evidence we present in this chapter demonstrates lasting plasticity at the level of local neuronal assemblies in both hippocampus and amygdala. Local circuit plasticity (LCP) is induced by tetanic stimulation of afferent fibers and is mediated, in the hippocampus, by a reduction in GAB A release. Different interneuronal populations are suggested to be involved in the LCP and LTP and at least one type of LCP correlates with age-related spatial memory abilities while the levels of LTP that can be induced initially were found unchanged in this respect. The results suggest that GABAergic interneurons play a major role in LCP and that the involved molecular/cellular modifications do not necessarily occur at the synaptic level. Overall, these data support the conception of LCP as a candidate mnemonic device that may be involved in more than one type of memory.

Introduction

A general principle of biology is that any given behavior of an organism depends on a hierarchy of levels of organization. As applied to the brain, it means that one needs to identify the main levels of organization in order to provide a framework for understanding the principles underlying its construction and function. The study of brain and mind has led to the recognition of several important levels of analysis from large information processing blocks down to the finest details of molecular structure and subcellular biophysics.

SUMMARY

Long-term potentiation (LTP) remains the most attractive model for learning-related plasticity. Of course, we must always test alternative hypotheses and be prepared to abandon those that no longer fit the existing data or provide predictive validity. Nonetheless, the LTP hypothesis for learning and memory has so much more data behind it than any alternative, I believe we are justified in continuing to test it rigorously using all available methods.

Among the most useful methods are those that involve direct manipulation of genes to disrupt LTP and learning. The use of genetic mutations to study learning and memory has a long history, but within the past decade its use has increased dramatically, driven by the development of techniques for targeted gene manipulation in mammals. Although these techniques are still essentially lesions, they have nonetheless increased the range of experiments that can be used to test the hypothesis that some forms of LTP underlie some forms of learning.

Introduction: A Brief History of Gene Manipulation in Long-Term Potentiation, Learning, and Memory

The intentional manipulation of genes has been going on for as long as humans have bred domesticated animals to select desirable physical or behavioral characteristics (Darwin, 1859). In the laboratory, this process dates back at least to Tryon (1934), who bred successive generations of rats on the basis of their performance in a complex maze. After several generations, there was nearly complete separation of performance in the maze, with “maze-dul “ rats making significantly fewer errors than “maze-bright” ones.

SUMMARY

In several forms in the mammalian brain, long-term potentiation (LTP) displays certain neurophysiologic characteristics that are commonly regarded as evidence that this mechanism for neuronal plasticity subserves the induction and/or storage of associative memories. In particular, the associative nature of LTP (induction by conjoint pre- and postsynaptic activity) and its dependence on relatively short interstimulus intervals between pre- and postsynaptic activity suggest certain qualitative similarities between LTP induction and the processes that define associative learning. Despite the popular acceptance of this cursory evidence, the processes that govern the formation and expression of associative memories are far more malleable and heterogeneous than those suggested by such mechanistic analogies. Here we describe in some detail the fundamental characteristics of the associative learning process, with an emphasis on those characteristics that are relevant to the assertion that LTP is an appropriate device for the storage of associative memories. With this as our point of reference, we are left to conclude that the mechanistic properties of LTP are in some instances irrelevant, and in some cases contraindicate, a role for LTP in the induction or storage of associative memories.

Introduction

Associative learning, an example of which is classical conditioning, is often described as a “simple” form of learning, limited in application to the modification of reflexive behaviors. This colloquial description is an unfortunate characterization of the phenomenon that misrepresents the fundamental nature and range of influence of associative learning. The insights that could be derived from a complete appreciation of the associative learning process were obvious to Ivan Pavlov, the Nobel Prize winner in 1904 for his studies on the physiology of digestion.

Overview of the Different Chapters

The opinions expressed in the different chapters show a great variety and span the entire range from researchers who believe that long-term potentiation (LTP) is not a model of learning at all and could be an artifact (chapters by McEachern and Shaw, McNaughton, and Matzel and Shors) to those who do not see any problems with the concept of LTP as a model for learning mechanisms (Abraham, Cho and Eichenbaum, and Rogan et al.). Most authors, however, voice a more diversified opinion and suggest a “revised and improved” model of LTP and memory formation that tries to integrate our increased knowledge of how neurons communicate in living brains. Such a model could account for discrepancies observed between LTP inducibility and learning abilities that have been published so far.

Persisting with LTP As a Model for Learning and Memory Formation

While not discussing the lack of correlations between LTP inducibility and learning abilities in numerous publications, Abraham suggests in Chapter 1 that LTP [and long-term depression (LTD)] is still a useful model and the dominating theory for mechanisms of memory formation. He notes that LTP is not homogeneously expressed in areas of the brain. LTP in the dentate gyrus is usually detrimental, whereas CA1 LTP tends to be robust and nondetrimental. LTP in the cortex is difficult to induce but appears to be very stable once it has been induced. To him, these data indicate that genuine differences exist that might reflect different roles in different areas for LTP as a process of memory formation.

SUMMARY

As described in the introduction of this book, the induction of long-term potentiation (LTP) of synaptic transmission in the hippocampus using high frequency stimulation (HFS) cannot be seen as a physiologic method that models synaptic plasticity that might occur during learning. Several other stimulation protocols have been developed that emulate natural neuronal firing patterns more closely, and LTP obtained with such a technique may therefore provide a better model of learning-related plasticity. The novel stimulation protocol presented here makes use of the reduction of local inhibition during theta rhythm. This stimulation technique (five pulses at 200 Hz phase-locked with peaks of theta oscillations in vivo) is able to induce stable LTP with as few as fifteen stimuli. Such bursts of stimuli are comparable to natural activity that can be recorded in the living brain. This type of stimulation protocol appears to induce LTP in a different way than HFS does. The metabotropic glutamate receptor agonist 1S,3S-ACPD blocks HFS-induced LTP, but does not affect the ability of rats to learn spatial tasks. However, LTP induced by a stimulation protocol of short bursts that are given phase-locked with theta rhythm was not blocked by 1S,3SACPD. Hence, the novel stimulation protocol appears to induce LTP in a more physiologic way that correlates with learning abilities of rats. It is suggested that this protocol, which not only resembles natural firing patterns found in the hippocampus in a living brain, but also models changes in synaptic transmission that are linked to memory formation more closely than HFS-induced LTP, has the potential to be a better model for putative synaptic changes that occur during learning.

SUMMARY

In order to draw a closer connection between spatial memory and hippocampal long-term potentiation (LTP), the activity of a place cell that fires selectively when an animal is located in a certain area of a maze was examined in freely behaving knock-out mice that exhibit spatial learning deficit and altered hippocampal plasticity. This approach seeks to examine the effects of altered synaptic plasticity on the information coding mechanism by characterizing firing properties of certain hippocampal neurons directly underlying perception and memory. Available recording data indicated that synaptic plasticity in the hippocampus is not an essential mechanism for the formation of place selective firing of hippocampal pyramidal neurons, but is important for fine tuning and stabilizing its neural activity across time.

Together with genetic, pharmacologic, and in vitro electrophysiologic studies, this type of approach should allow us to examine whether experimental manipulations that block LTP change the capacity of neurons to process critical sensory stimuli and to code relevant cues into memory, and thus should provide a unique and promising avenue in bringing further insight into the cellular and physiologic mechanisms underlying learning and memory.

Introduction

A major focus of neuroscience research, and the central topic of this volume, is the identification of cellular and molecular plasticity mechanisms that mediate memory. Most studies aimed at this goal involve attempts to relate physiologic or molecular indices of neural plasticity to behavioral performance in learning and memory tests.

What Is LTP?

At present, long-term potentiation (LTP) of synaptic transmission is the most widely studied model for neuronal change that occurs during learning and that stores information in the brain. When Bliss and Lømo discovered the phenomenon of LTP 30 years ago, it must have been a very exciting moment (see Bliss and Lømo, 1996, for a detailed account). The finding was an important step in the long ongoing search for the neuronal basis of memory formation in the nervous system. Even though anatomists such as Exner showed changes in neuronal connections that were induced by exposure to rich environments as early as 1884 (Exner, 1884), it was by no means clear how exactly neurons change in order to store information. Exner and others (Hebb, 1949; Hinton, 1992; Jodar and Kaneto, 1995) proposed activity-dependent, selective plastic changes in the connections between neurons that could “freeze” the pattern of activity of the neurons and therefore preserve information. The theoretical basis for LTP existed for a long time, and when LTP was described by Bliss and Lømo in a full paper (Bliss and Lømo, 1973), researchers in the field were more than ready for the actual empirical evidence that neuronal network processes might be used in the brain.

The basic principle of every neuronal net is that connections between neurons can change in a defined, use-dependent way thereby storing information. Such a system has certain attributes such as pattern completion, parallel processing, and self-instructing learning properties, all of which are characteristics of activity in working brains (Hinton, 1992; Jodar and Kaneto, 1995).

SUMMARY

Amnesics and hippocampally lesioned animals fail to suppress incorrect information rather than failing to store correct information. There is also strong evidence that anxiolytic drugs act on a behavioral inhibition system involving the hippocampus. This has led to a theory of the neuropsychology of anxiety (Gray and McNaughton, 2000) that has as a corollary, the view that all memories are stored outside the hippocampus. It follows that long-term potentiation (LTP) in the hippocampus cannot be the basis of memory. There is good evidence, however, that LTP is involved in hippocampal function in some way. Our theory would be consistent with hippocampal LTP storing hippocampal-specific data or reprogramming the hippocampal computer. In both cases LTP would be a means of altering the input–output relationships of the hippocampus (and hence indirectly performance on hippocampal-sensitive tasks). If this is true, the hippocampus may often function without the requirement for LTP.

Introduction

In the Introduction to this book, Hölscher notes that the bulk of studies, which see long-term potentiation (LTP) as the substrate for learning and memory, focus on the hippocampus. They manipulate LTP in the hippocampal formation and assess storage of spatial memories. In this chapter I argue that the hippocampus (unlike the cortex or amygdala) does not use LTP to store any type of memory and is not especially concerned with space. Further, I suggest that the hippocampus prevents rather than promotes the formation of connections within the brain. This view derives from a recent update of Gray's (1982) theory of the neuropsychology of anxiety (Gray and McNaughton, 2000).

SUMMARY

The neuronal code that is used by the brain to represent complex information is not yet identified. Two complementary representational strategies, smart neurons and neuronal assemblies, have been proposed as solutions to the coding problem. Neurons that respond selectively to particular feature combinations are called “smart.” Neuronal assemblies are large populations of neurons that represent information in distributed form by their correlated discharge. These different strategies seem to be optimized for different functional aspects of the neuronal processes underlying perception and sensorimotor integration. Smart neurons could serve the fast processing of familiar information by operating as coincidence detectors for converging inputs arriving on feed-forward connections. In contrast, neuronal assemblies appear to provide a flexible coding mechanism that can cope with the combinatorial complexity of new material during creation and reorganization of representations. Synchronized neuronal responses have been proposed as constituting functional assemblies that could code for the relations among individual features of an input pattern in the case of perceptual processes, or sensory information and motor activity in the case of sensorimotor integration. Neuronal oscillations in the gamma-frequency range are often observed to accompany precisely synchronized responses. Oscillatory modulation of neuronal discharge may therefore serve as a mechanism to synchronize responses even over large distances. Oscillatory discharge is induced by activation of the same modulatory systems like the reticular formation that control vigilance and improve sensorimotor performance of the awake performing brain. As the two described representational strategies function in rather different ways, the mechanisms by which they can be modified during learning and memory formation may also differ considerably.

SUMMARY

If the hippocampus is a site for spatial learning, then it should be possible to see changes occurring in its representation of space following learning. How would we recognize such changes? It is argued in this chapter that if the synaptic plasticity hypothesis of learning is true, then to attribute changes in neuronal activity to memory formation, we need (1) to know what the neurons' inputs were before and after learning, and (2) to show that these changed in a meaningful way. By “meaningful” is meant that they altered the cognitive representation in a manner congruent with the actual experience of the animal. Although it is not yet feasible to record single inputs onto hippocampal cells in awake, behaving animals, it is possible to infer the strengths of these inputs by recording the responses of the neurons to environmental stimuli. By showing that the inputs change in an appropriate way following experience, it is possible to derive a simplified model of memory formation that looks at the cognitive representation directly, independent of the animal's behavior. This approach may circumvent some of the difficulties involved in trying to relate very low-level processes, such as synaptic plasticity, with very high-level processes, such as behavior.

Introduction

The hypothesis that a long-term potentiation (LTP)-like process underlies memory formation has the drawback that it is difficult to think of an experimental result that could definitively refute it. For every piece of evidence that appears to falsify the hypothesis, there is either another piece of evidence or a hand-waving argument that explains it away.

SUMMARY

For almost a century we have been attempting to test the notion that information is stored in the brain as changes in the efficacy of synaptic connections on those neurons that are activated during learning. Since the discovery of long-term potentiation (LTP) in 1973, we have learned a great deal about the mechanisms underlying activity-dependent synaptic modification. To date, LTP is currently accepted as the most viable model of the type of modifications that would occur to process information and store the memory trace. Although over the past 25 years or so there has been a great deal of empirical data that allude to the possibility that this form of synaptic plasticity may well be one of the crucial mechanisms underlying learning and memory, we are still lacking definitive answers. We now know that LTP is the output of a series of biochemical and molecular events that in all likelihood results in a form of modification of neural networks. In this review we describe how the advances in molecular biology give us the tools to both investigate the mechanisms of synaptic plasticity and to apply these to investigations of the underlying mechanisms in learning and the formation of memories that have until now eluded us.

Introduction

One of the major concepts in our understanding of how memories are laid down in the brain lies in the notion that they (the memories) are encoded as spatiotemporal patterns of activity in cell networks, and not at the single cell level. An underlying principle is that the encoding and the storage of memory would therefore require some form of dynamic modification that is driven by the interaction between cells within these networks.

SUMMARY

Studies of long-term potentiation (LTP) induction and persistence in awake animals have revealed intriguing regional differences. LTP and heterosynaptic LTD in the dentate gyrus both inevitably appear to be decremental in nature, with the time constant of decay related to the strength of the induction stimulus. On the other hand, nondecremental LTP can be induced in area CA1 with a very mild stimulus that, when given in the dentate gyrus, evokes little or no LTP at all. CA1 synapses can also exhibit decremental LTP. LTP in neocortex has so far been observed only after multiple, spaced tetanization episodes, but this LTP can then be quite long-lasting. These data indicate intriguing differences in the pattern of LTP persistence across brain regions that may reflect different contributions to the memory or information storage process in behaving animals.

The dentate gyrus and CA1 appear to differ in a number of ways at molecular, cellular, and network levels that could account for the apparent differences in LTP persistence. These include the ability to produce voltage-dependent calcium channel (VDCC)-dependent LTP, the contribution by catecholamines to the persistence of LTP, the pattern of constitutive gene expression, and the pattern of tetanic stimulation commonly used to study LTP in these two regions. Few of these features appear to strongly differentiate the two hippocampal regions, however, making it conceivable that the dentate gyrus has the inherent capacity for nondecremental LTP, if only the right induction conditions could be met. Indeed, apparently nondecremental LTP has been reported for perforant path synapses in studies that have employed electroconvulsive shocks.

SUMMARY

Psychological studies of list learning provide a quantitative behavioral description of human episodic memory. Our goal here is to describe this literature and to attempt, insofar as possible, to relate these finding to underlying physiologic processes. One prominent hypothesis to emerge from psychological studies is that of a short-term memory (STM) buffer (e.g., Atkinson and Shiffrin, 1968). It is thought that this buffer stores a small number of items (e.g., 7 ± 2 digits) using maintained neural activity. The repetitive firing produced by the STM buffer is important for the transfer to long-term memory (LTM). The rapid formation of LTM is revealed by the pre-recency part of the serial position curve in free-recall experiments. The information stored in LTM include intraitem associations, asymmetric interitem heteroassociations, and associations of items to context. Despite the success of buffer models, some observations, particularly long-term recency, argue against two-store models and alternative models have been developed. Additional information relevant to this controversy comes from neuropsychological, pharmacologic, physiologic, and computational studies. In free-recall studies, hippocampal lesions selectively reduce the recall of early list items consistent with a selective effect on LTM. Furthermore, the rapid formation of LTM (within seconds) and the selective inhibition of this process by cholinergic antagonists is consistent with what is known about the induction of long-term potentiation (LTP) and further supports the distinction between LTM and STM.

A second hypothesis to emerge from behavioral studies (the Sternberg task) is the idea of rapid serial search of the STM buffer. A model has been developed that relates these findings to brain oscillations.

SUMMARY

Long-term potentiation (LTP) is a form of synaptic plasticity that has been proposed to mediate certain forms of learning and memory. In this chapter, it is argued that LTP in the hippocampus and amygdala plays a crucial role in the acquisition of a simple form of emotional learning and memory: Pavlovian fear conditioning. The distinct roles for hippocampal and amygdaloid LTP and the roles for short-term synaptic plasticity mechanisms in the acquisition of learned fear responses are discussed.

Introduction

Throughout history, humans have been fascinated by the nature of memory, the permanent storehouse of the mind's experience. The foregoing passage excerpted from Plato's Theaetetus (ca. 400 B.C.) represents an early attempt to describe the process of memory formation in the human brain. In this passage, Plato envisions that memories are established in the brain when perceptions or thoughts render lasting impressions in the mnemonic wax of the mind.