Matt Ridley

In fruit flies the cyclic A M P system seems to be especially active in brain regions called mushroom bodies, toadstool-shaped extrusions of neurons in the fruit fly brain. If a fly has no mushroom bodies in its brain, then it is generally incapable of learning the association between a smell and an electric shock. C R E B and cyclic A M P seem to do their work in these mushroom bodies. Exactly how is only now becoming clear. By systematically searching for other mutant flies incapable of learning or memory, Ronald Davis, Michael Grotewiel and their colleagues in Houston came up with a different kind of mutant fly, which they called volado. ('Volado', they helpfully explain, is a Chilean colloquialism meaning something akin to 'absent-minded' or 'forgetful', and generally applied to professors.) Like dunce, cabbage and rutabaga, volado flies have a hard time learning.

But unlike those genes, volado seems to have nothing to do with C R E B or cyclic A M P . It is the recipe for a subunit of a protein called an alpha-integrin, which is expressed in mushroom bodies, and which seems to play a role in binding cells together.

To check that this was not a 'chopstick' gene (see the chapter on chromosome 11) that had lots of effects beside altering memory, the Houston scientists did something rather clever. They took some flies in which the volado gene had been knocked out, and inserted a fresh copy linked with a 'heat-shock' gene - a gene that becomes switched on when suddenly heated up. They had carefully arranged the two so that the volado gene only worked when the heat-shock gene was on. At cool temperatures, the flies could not learn. Three hours after a heat shock, however, they suddenly became good M E M O R Y 2 2 7

learners. A few hours after that, as the heat shock faded into the past, they again lost the ability to learn. This means that volado is needed at the exact moment of learning; it is not just a gene required to build the structures that do the learning.6

The fact that the volado gene's job is to make a protein that binds cells together raises the intriguing hint that memory may consist, quite literally, of the tightening of the connections between neurons.

When you learn something, you alter the physical network of your brain so as to create new, tight connections where there were none or weaker ones before. I can just about accept that this is what learning and memory consist of, but I have a hard time imagining how my memory of the meaning of the word 'volado' consists of some strengthened synaptic connections between a few neurons. It is distinctly mind-boggling. Yet far from having removed the mystery from the problem by reducing it to the molecular level, I feel that scientists have opened before me a new and intriguing mystery, the mystery of trying to imagine how connections between nerve cells not only provide the mechanism of memory but are memory. It is every bit as thrilling a mystery as quantum physics, and a great deal more thrilling than Ouija boards and flying saucers.

Let us delve a little deeper into the mystery. The discovery of volado hints at the hypothesis that integrins are central to learning and memory, but there were already hints of this kind. By 1990 it was already known that a drug that inhibited integrins could affect memory. Specifically, such a drug interfered with a process called long-term potentiation ( L T P ) , which seems to be a key event in the creation of a memory. Deep in the base of the brain lies a structure called the hippocampus (Greek for sea-horse) and a part of the hippocampus is called the Ammon's horn (after the Egyptian god associated with the ram and later adopted as his 'father' by Alexander the Great after his mysterious visit to the Siwah oasis in Libya). In the Ammon's horn, in particular, there are a large number of 'pyramidal' neurons (note the continuing Egyptian theme) which gather together the inputs of other, sensory neurons. A pyramidal neuron is difficult to 'fire', but if two separate inputs arrive at once, 2 2 8 G E N O M E

their combined effect will fire it. Once fired, it is much easier to fire but only by one of the two inputs that originally fired it, and not by another input. Thus, the sight of a pyramid and the sound of the word 'Egypt' can combine to fire a pyramidal cell, creating an associative memory between the two, but the thought of a sea-horse, although perhaps connected to the same pyramidal cell, is not

'potentiated' in the same way because it did not coincide in time.

That is an example of long-term potentiation. If you think, too simplistically, of the pyramidal cell as the memory of Egypt, then it can now be fired by the word or the picture, but not by a sea-horse.

Long-term potentiation, like sea-slug learning, absolutely depends on a change in the properties of synapses, in this case the synapses between the inputting cells and the pyramidal cells. That change almost certainly involves integrins. Oddly, the inhibition of integrins does not interfere with the formation of long-term potentiation, but it does interfere with its maintenance. Integrins are probably needed for literally holding the synapse closely together.

I glibly implied a few moments ago that the pyramidal cell might actually be a memory. This is nonsense. The memories of your childhood do not even reside in the hippocampus, but in the neo-cortex. What resides in and near the hippocampus is the mechanism for creating a new long-term memory. Presumably, the pyramidal cells in some manner transmit that newly formed memory to where it will reside. We know this because of two remarkable and unfortunate young men, who suffered bizarre accidents in the 1950s. The first, known in the scientific literature by the initials H.M., had a chunk of his brain removed to prevent the epileptic seizures caused by a bicycle accident. The second, known as N.A., was a radar technician in the air force, who one day was sitting building a model when he happened to turn round. A colleague, who was playing with a miniature fencing foil, chose that moment to stab forward and the foil passed through N.A.'s nostril and into his brain.

Both men suffer to this day from terrible amnesia. They can remember events from their childhood quite clearly and from right up to a few years before their accidents. They can memorise current M E M O R Y 2 2 9

events briefly if not interrupted before being asked to recall them.

But they cannot form new long-term memories. They cannot recognise the face of somebody they see every day or learn their way home. In N.A.'s (milder) case, he cannot enjoy television because commercials cause him to forget what went before them.

H.M. can learn a new task quite well and retains the skill, but cannot recall that he has learnt it - which implies that procedural memories are formed somewhere different from 'declarative' memories for facts or events. This distinction is confirmed by a study of three young people with severe amnesia for facts and events, who were found to have gone through school, acquiring reading, writing and other skills with comparatively little difficulty. All three, on being scanned, proved to have unusually small hippocampuses.7

But we can get a little more specific than just saying that memories are made in hippocampuses. The damage that both H.M. and N.A.

suffered implies a connection between two other parts of the brain and memory formation: the medial temporal lobe, which H.M. lacks, and the diencephalon, which N.A. partly lacks. Prompted by this, neuroscientists have gradually narrowed down the search for the most vital of all memory organs to one principal structure, the perirhinal cortex. It is here that sensory information, sent from the visual, auditory, olfactory or other areas, is processed and made into memories, perhaps with the help of C R E B . The information is then passed to the hippocampus and