A study of chicks' eating tastes has identified a molecule that affects their memory, which could provide a breakthrough in the quest to treat Alzheimer's disease. We are our memories. In old age we retain memories of our childhood even though every molecule in our bodies has been recycled thousands, perhaps millions of times. This is why diseases of memory, such as Alzheimer's, are so devastating.

Animals also make memories and can change their behaviour as a result of experience. Animals use their wits to capture their prey, and to avoid being eaten at least long enough to be able to reproduce. This demands the development of effective sense organs and the capacity to compare the data provided by those sense organs with past experience. This is what memory is about. We can learn a great deal about the brain mechanisms of memory, and about strategies for repairing damaged memory, by studying animals.

Until recently accidents provided virtually the only way to study the processes of human memory. Loss of recent memory but persistence of well-established memories following accidents like concussion, pointed to a distinction between short and long-term memory. But it is only within the last decade that powerful imaging techniques to explore the human brain have developed, such as PET (positron emission tomography) and MEG (magnetoencephalography). Neither is perfect; one, because it involves radioactivity, cannot be used repeatedly on the same subject; the other, in its infancy as yet, gives superb time resolution but poor spatial location.

What both show is the dynamics of learning and memory. For instance, when a person is asked to learn and recall a word list, the initial learning processes "light up" regions of the left frontal cortex - these regions show enhanced blood flow and glucose and oxygen utilisation, which are taken as surrogate measures for neural activity. By contrast, when the subject is asked a brief time later to recall the same list, regions of the right hemisphere become engaged. The problem with all these measures is that, dramatic and beautiful as the pictures produced are, we do not yet know what they actually mean in terms of the brain structures involved.

É«ºÐÖ±²¥

To go beyond this means to enter the brain, to explore its nerve cells and the connections between them, the synapses. Those working in this terrain owe their theoretical perspective to the Canadian psychologist Donald Hebb. Impressed by the durability of memory, he argued that it must result from permanent changes in the circuitry of the brain, on its synapses. Memory depends on the physical growth or realignment of synapses, so creating novel pathways. Granted there are perhaps 100 billion nerve cells, neurons, in the human brain, and something like 1014 synapses between them, that is an awful lot of possible circuits.

To turn Hebbian theory into experimental evidence requires a reliable animal model, to which modern techniques of molecular biology, neurophysiology and image analysis can be applied. I have been working with one for many years - young chicks. Chicks explore their environment energetically. They have a lot to learn, and need to do so fast. If offered a small bright object such as a bead, chicks will rapidly peck vigorously at it. If the bead is made to taste bad by dipping it in bitter liquid, the chicks will peck once, show their disgust, and avoid beads of a similar size and colour for several days. By contrast chicks which have pecked a dry or water coated bead will continue to peck at it.

É«ºÐÖ±²¥

What happens in the brain of the chick that has experienced the bad tasting bead? Two regions of the chick brain in particular "light up" - use more energy - when the bird pecks the bitter bead. They have the usual cumbersome dog-Latin names; I refer to them by their initials - IMHV and LPO.

Do synapses change in this region? The short answer is yes: 24 hours after training chicks, there is a reorganisation of synaptic connectivity in both IMHV and LPO, detectable by microscopy. If the new synapses are physiologically "meaningful" we might also expect that their electrical properties would change as a consequence of training.

This proves to be the case. If chicks are trained and later anaesthetised and a recording electrode placed into the IMHV or LPO, we pick up characteristic patterns of novel neural electrical activity - bursts of high frequency firing by the neurons. These burst patterns shift in space and time, beginning in the left IMHV and shifting to the right IMHV and LPO. Importantly, there is a time interval, about six to eight hours after training, when the bursts are particularly strong and during which their focus shifts between brain regions.

What are the molecular mechanisms involved in these physiological and neuroanatomical events? Following training, a veritable cascade of biochemical reactions occurs, involving a complex flow of chemical messages between pre- and post-synaptic cells. At both sides of the synapse, the signal cascade culminates in the activation of genes and the synthesis of new proteins. It is these proteins which are responsible for maintaining and changing the configuration of the synapses. Pre- and post-synaptic sides of the junction between neurons are held together by a class of proteins known as cell adhesion molecules. These molecules span the synaptic membrane, with a tail in the interior of the cell and an external head projecting out of the cell into the extracellular space.

This head end consists, like all proteins, of a long sequence of amino acids, but attached to the amino acids are also chains of sugar molecules. The sugars make the molecule sticky. This means that the extracellular portion of one of the cell adhesion molecules sticking out from the presynaptic side of the junction can glue on to the matching sugar sequence of another cell adhesion molecule sticking from the post-synaptic side, thus holding the two sides of the synapse in place.

É«ºÐÖ±²¥

A few hours downstream of the bead-peck there is a surge in the sythesis of cell adhesion molecules which are transported to the "activated" synapses and inserted into them, thus, presumably, stabilising them in their new configuration. Blocking the synthesis of these molecules, or preventing them sticking to one another by introducing into the brain specific antibodies which will interrupt the binding process, results in amnesia. This amnesia occurs about five to eight hours after training, coinciding almost exactly with the increased high-frequency electrical bursting activity and with the apparent shift in the "memory sites" between IMHV and LPO.

There is reason to believe that while the initiating events that trigger the biochemical cascade in the chick will differ from those in you or I when we learn a number sequence or are introduced to a new colleague, those that occur downstream - the molecular biological processes and involvement of cell adhesion molecules - are likely to be common across species.

Although the processes I have described give the appearance of supporting the idea that the memory circuits are in some sense hard-wired once the new patterns have been established, other experiments from our laboratory make it clear that the memory "store" is not fixed but dynamic. Furthermore, it does not reside in a discrete brain location. Indeed, "it" is not a single entity at all. Although when I train the chick on a bead, I present it to the bird as a single unified object, it is clear that it is not represented this way in either my brain or the chick's. Brains classify data presented to them. In the case of a bitter bead classification may be by colour, shape or size, locale and context of presentation, and of course taste. When the chick first tastes the bitter bead it does not know which of these features are important, so it will tend to classify all of them. But the sites of classification within the brain are not coterminous. We believe, for instance, that colour representation may reside in the IMHV, while shape and size may be represented in the LPO.

É«ºÐÖ±²¥

Whereas it was once believed that the brain was organised hierarchically, with "lower" regions each reporting up to some homunculus in the brain, this is not how brains work. There are no "grandmother neurons" which integrate inputs from many regions so as to decide this is or is not your grandmother. Instead, grandmother, and bitter bead, representations are distributed. The brain is not a Stalinist command economy, but a well-integrated anarchistic commune. A major problem for neuroscience is to understand how these distributed properties can be organised, so that we, and chicks too presumably, perceive the world in a unified manner. The partial representations of fragmentary characteristics of the bead or the grandmother must come together. This has become known as the "binding problem". The high frequency neuronal bursting I have talked about may have something to do with it.

So far, I have described the effect of training the chick by offering it a bitter bead to taste. This memory persists for at least several days. However, if we train the chick on a less bitter taste, such as that of quinine, the chick will remember and avoid the bead for only a relatively short time, and then apparently forget, so that it pecks rather than avoids a dry bead when tested. The cut-off point for this short lasting memory appears to be about eight hours - the time when, with strong learning, the cell adhesion molecules are made and inserted into the synaptic membranes. We have evidence that, in weak learning, the "second wave" synthesis of the cell adhesion molecules does not occur. However, it is possible to stimulate the synthesis of these molecules. A number of drug treatments will do so, and so will steroid hormones - and will also enhance the retention of the memory.

Why does this matter? I have referred to the tragic consequences of Alzheimer's disease, characterised as it is in its early stages by a progressive loss of memory. Much effort has been devoted to developing drugs which might alleviate this memory loss, but to date most of the attention has been devoted to agents which affect neurotransmitters. I believe this may be the wrong place to look. Recent molecular biological studies have demonstrated that one of the key molecules whose metabolism is disrupted in Alzheimer's is a substance known as the amyloid precursor protein, APP. In the disease, parts of this molecule get chipped off and accumulate in the spaces between synapses, resulting in the characteristic plaques which can be seen in the brains of Alzheimer's sufferers. It turns out that APP is a cell adhesion molecule, and its extracellular head with attached sugars presumably plays a part in holding synapses together. I do not yet know how directly it is involved in these processes, but if it acts like those adhesion molecules we have already studied, it becomes obvious why one of the early deficits in Alzheimer's is loss of long-term memory. And we may be able to follow up our observations of the effects of hormones on both the synthesis of the molecules and memory retention by developing a class of pharmaceuticals which, acting via APP, may be beneficial in the treatment, though not alas the cure, of Alzheimer's disease.

The article is a shortened version of the lecture given in the Darwin College Series on "Memory". Steven Rose is professor of biology, director of the Brain and Behaviour Research Group at the Open University, and author of The Making of Memory.

É«ºÐÖ±²¥