Investigations of the cellular bases of memory.

Part 2:

The behavioural assessment of interference with memory formation.

Transfer experiments such as those inspired by McConnell's work with flatworms are fraught with difficulty. Not only is there the problem of establishing exactly what has been transferred functionally (memory or something else) but there is also the problem of isolating the active ingredient of the material being transferred. The extraction and identification of these ingredients is the realm of biochemists rather than psychologists and the dominance of the field by biochemists in the late 1960's may have had something to do with the problems of interpreting and replicating the behavioural tests necessary to demonstrate memory transfer. As these problems became more and more apparent a different experimental approach began to dominate the search for the cellular basis of memory. Instead of trying to transfer memories researchers investigated the effects of various manipulations interspersed between learning and recall to see which interfered with recall and when.

The most common paradigm in these interference studies was one-trial passive avoidance in the chick. Day-old chicks are presented with a visually distinctive (e.g. red-coloured) bead at which they will instinctively peck. The bead is coated with an aversive chemical (e.g. quinine). In the absence of any interference chick will learn to associate the visually distinctive properties of the bead with an unpleasant taste in a single trial. If they are subsequently presented with a bead with the same visual properties previously paired with an aversive taste they will refuse to peck even though the bead is not coated with the aversive chemical in this recall test. Control chicks who are initially exposed to a bead not coated in an aversive chemical will still peck at a similar bead on a recall trial. Different mechanisms for encoding memory can be investigated by interpolating manipulations such as drug administration or electric shock between the training and recall phases.

With this paradigm it is not possible to differentiate between manipulations which specifically effect memory and those which have a generalised effect on pecking behaviour. For example, a drug which has no effect on memory but which stimulates all behaviour will increase the likelihood of pecking even after aversion. The effect of a stimulant and of a drug which disrupts memory will therefore be the same - a higher rate of pecking after aversion than is found in control birds. Similarly, sedative drugs and drugs which enhance memory will also both produce similar decreases in post-aversion pecking. Variants of the passive avoidance task help remedy these problems.

In a simple discrimination paradigm chicks are initially given a pre-training trial in which they are presented with red and blue beads, neither of which is aversive. Any manipulation falls between this pre-training trial and a learning trial where the red bead (alone) is presented paired with the aversive chemical. On the subsequent recall trial neither bead is aversive. If an interposed manipulation has no effect on (or enhances) memory then the chicks should peck the blue bead but not the red bead at recall. If the manipulation inhibits pecking then the chicks should peck neither colour of bead. Their memory that the red bead is aversive has not been disrupted - they learned this after the manipulation. This task identifies manipulations which generally inhibit pecking, but it does not identify manipulations which disrupt memory.

Pre-train  Manipulation	Train			Test

Red Blue		Red+Aversion		Red	Blue

						avoid	peck 	= intact memory
						avoid	avoid 	= peck inhibition
						peck	peck   	= memory disruption
									OR
								   peck enhancement
In the simple discrimination task we have just discussed an enhancement of the peck response would still produce similar results to a memory disruption. These possibilities can be distinguished using a three-choice discrimination design. Here the chick is presented with three coloured beads, none aversive, in pre-training. Training takes place in two phases with a manipulation between them. In the first phase one colour of bead (e.g. red) is paired with the aversive chemical. This is followed by a memory manipulating drug. Following this a second colour of bead (e.g. chrome) is paired with the aversive chemical. The third colour (e.g. blue) is not paired with the aversive chemical at any stage. In recall, if pecking is enhanced then the chick will peck at all three beads, on the other hand, if memory is disrupted then the chick should peck at the red bead but not the blue one.
Pre-train	  Train (1)	Manipulation	Train (2)	Test	

Red Blue Chrome	  Red+Aversion			Blue+Aversion	Chrome	Red  	Blue		

								peck	avoid	avoid 	= intact memory
								peck	peck	peck 	= peck enhancement
								avoid	avoid	avoid   = peck inhibition
								peck	peck	avoid   = memory disruption
These procedures can be used to provide evidence for the involvement of a number of different processes in memory formation. It is possible to distinguish both the number of temporally distinct phases of memory and whether these phases operate in series or in parallel.

An agent which disrupts a specific stage of memory will be ineffective after that stage has been completed. If it is administered before the phase begins (and still active) then it will disrupt the phase. If it is administered and active part way through a stage it will have a partial effect. Thus, administration of manipulations at different times and recall tests at different times can uncover the time course of encoding and the time course of possible retrieval for a memory process. For example, intracranial injection of potassium or lithium salts into the brain effect recall between 10 minutes and 3 hours after training if they are given within 5 minutes of training. They do not effect recall if they are given later than this. Even if they are given in this critical period within 5 minutes of training they do not have a significant effect on recall earlier than 10 minutes after training. This implies that they disrupt a process in which encoding takes place in the period between training and five minutes post-training and that this process is responsible for a short to medium term memory stage starting a 10 minutes and lasting up to 3 hours after training. It is highly likely that the process being disrupted is the operation of the membrane sodium pump.

In the next section we will see the range of mechanisms which were implicated in memory encoding in research carried out in the 1970's and 80's.

Substances and mechanisms implicated in memory formation.

By using the methods we have just described and related designs it was possible to assess the ability of a wide range of substances to disrupt memory and to see both when these substances acted and when the disruption they caused occurred. The work which took this approach up to the mid 1970s was reviewed by Gibbs and Ng (1977). They divide the encoding of memory into four stages: Evidence for the first stage cannot be obtained using passive avoidance, but other studies implicate its dependence on ongoing electrical activity (action-potentials) by showing disruption by electro-convulsive shock administered in the period immediately after training.

Disruption of the signals being encoded themselves will obviously interfere with memory, but beyond that what kind of processes might be involved? The assumption behind this work is that memory involves changes in the synaptic connections between neurons, so the factors involved in neurotransmission are likely candidates. These include both the effects of neurotransmission, (changes in the membrane potential of the postsynaptic cell, changes in membrane permeability and so on) and the mechanisms of neurotransmission (the density and number of receptors, the metabolism and synthesis of transmitters and so on).

The dependence of memory on membrane depolarisation has been widely studied. One simple way of effecting this is to change the relative concentrations of ions inside and outside the cell, for example, by injecting solutions of LiCl or KCl into the ventricles. The presence of these salts in the extracellular space will alter the rate at which potassium ions (K+) diffuse through channels in the membrane, and hence alter the membrane potential. Intra-ventricular injection of these salts inhibits recall 5 minutes after learning provided the salt is administered 5 minutes before learning.

There are more complex ways of effecting the membrane potential. Chemicals such as oubain and ethacrynic acid disrupt metabolic processes within the cell, including the action of the sodium pump. This disruption leads to longer lasting changes in the membrane potential. Ouabain inhibits recall 15 minutes after learning, leaves recall 5 minutes after learning intact (again, if administered 5 minutes pre-learning). This suggests that there is a distinction between the short term membrane potential mechanism disruptable by salts and the slightly longer term mechanism disruptable by ouabain. This distinction is bolstered by the observation that ouabain induced amnesia, but not LiCl or KCl induced amnesia can be counteracted by diphenyl hydantoin (DPH) which stimulates metabolic (Na+/K+ ATPase) activity.

In the longer term protein synthesis inhibitors such as cyclohexamide and anisomycin produces amnesia between 60 minutes and 3 days after learning (probably longer). There is no evidence of disruption 30 minutes after learning. Administration can be anywhere between 30 minutes before and 30 minutes after learning.

In summary, the implication is that the mechanisms underlying the four stages are:

Since the 1970s other approaches to understanding the cellular bases of memory (which we will cover later in the course) have greatly influenced studies into passive avoidance in the chick. In particular, the importance of the role of the excitatory neurotransmitter glutamate and its receptors has been recognised. This has lead to some changes and some additions to the model outlined above.

Short-term stage. It has been shown that the excitatory neurotransmitter glutamate is released rapidly following passive avoidance learning in the chick. The postsynaptic effect of glutamate is to hyperpolarise the cell by opening receptors through which K+ ions can escape. Drugs which deplete presynaptic glutamate (L-aspartic acid beta-hydroxamate LAA§H) blocks STM (as does monosodium glutamate). Both are only effective if given within a few minutes of learning. The calcium channel through which K+ ions escape can be blocked with lanthanum chloride which also prevents STM formation.

Intermediate labile stage. Results which were earlier taken as evidence for a specific role of the sodium pump involved drugs which disrupted metabolism within the cell. These results may implicate metabolic processes not necessarily associated with the sodium pump. In particular, it has been shown that the conversion of glutamine - the breakdown product of glutamate - back into glutamate is necessary for recall between 20 and 50 minutes after learning.

Long-term stage. There are a number of different types of glutamate receptors. The drug N-methyl-D-aspartate (NMDA) selectively binds to one type which is therefore referred to as the NMDA receptor (even though it is really a glutamate receptor). The NMDA receptor is a voltage dependent ionotrophic receptor - that is, when glutamate binds to it it allows the passage of calcium and potassium ions in and out of the cell respectively, provided the cell is already depolarised (i.e. the action of glutamate depends on the membrane potential). Blockade of this type of glutamate receptor with the drug D-2-amino-5-phosphonovalerate (APV, sometimes referred to as AP5) produces recall deficits 180 mins after training if administered within 10 seconds of training (but not later)!

A second ionotrophic glutamate receptor, this one not voltage-sensitive, is characterised by binding the chemical AMPA (alpha-amino-3-hydroxy-5-methyl-4-isoxalone proprionate !). The AMPA receptor antagonist DNQX also produces recall deficit 180 mins after training, but only if administered between 10 and 25 minutes after training. So, although both NMDA and AMPA glutamate receptor binding appears necessary for long-term memory formation, the action of the NMDA receptors starts much earlier than that of the AMPA receptor. The AMPA dependent process is occurring at the same time as the intermediate/labile phase has reached the stage where it is available for recall.

Finally, there is also evidence that activity of a quite different kind of glutamate receptor is crucial for long-term memory. The metabotropic glutamate receptors (mGluR) do not open ion channels (in contrast to both the NMDA and AMPA receptors), but rather give rise to much slower changes within the cell. Protein kinase C (PKC) activation and the release of stored calcium are among the consequences of mGluR. Protein kinase C itself is involved in the protein synthesis implicated by the effects of protein synthesis inhibitors in earlier experiments.

A revised, 1990s version of our summary is therefore:

These studies have given us a good idea of the various processes which contribute to memory formation. As they involve behavioural tests of the animals ability to recall we can, with appropriate testing procedures, have confidence that the manipulations discussed above are really involved in memory formation. The disadvantage of this approach, however, is that we cannot really tell whereabouts, both with and between neurons, all these processes are taking place. The next two approaches we will consider to studying the cellular basis of memory allow many more of these details to be uncovered.

Notes

This lecture covers material from two review papers:

Gibbs, M.E. and Ng, K.T. (1977) Psychobiology of memory: Towards a model of memory formation. Biobehvioral Reviews 1 113-116.

Ng, K.T.; O'Dowd, B.S.; Rickard, N.S.; Robinson, S.R.; Gibbs, M.E.; Rainey, C.; Zhao, W.-Q.; Sedman, G.L. and Hertz, L. (1997) Complex roles of glutamate in the Gibbs-Ng model of one-trial aversive learning in the new-born chick. Neuroscience and Biobehavioral Reviews 21 45-54.

together with some additional background from:

Hammond, C. (1996) Cellular and molecular neurobiology. San Diego CA: Academic Press.