In Search of Memory: The Emergence of a New Science of Mind (35 page)

T
he new discoveries in the hippocampus—place cells, the NMDA receptor, and long-term potentiation—raised exciting prospects for neuroscience. But it was not at all clear how the spatial map and long-term potentiation were related to each other or to explicit memory storage. To begin with, although long-term potentiation in the hippocampus was a fascinating and widespread phenomenon, it was a highly artificial way of producing changes in synaptic strength. This artificiality caused even Lømo and Bliss to wonder “whether or not the intact animal makes use in real life of a property which has been revealed by synchronous, repetitive volleys….” Indeed, it seemed unlikely that the same pattern of firing occurs in the course of learning. Many scientists questioned whether the changes in synaptic strength produced by long-term potentiation play any role at all in spatial memory or the formation and maintenance of the spatial map.

I began to realize that the ideal way to explore these relationships would be through genetics, much as Seymour Benzer had used genetics to study learning in
Drosophila
. In the 1980s biologists began to combine selective breeding and the tools of recombinant DNA to produce genetically modified mice. These techniques made it possible to manipulate the genes underlying long-term potentiation and thus to answer some pressing questions that interested me. Does long-term potentiation, like long-term facilitation in
Aplysia
, have different phases? Do those phases correspond to short- and long-term storage of spatial memory? If they do correspond, we could interfere with one or the other phase of long-term potentiation and thereby determine what actually happens to the spatial map in the hippocampus when an animal learns and remembers a new environment.

It was exhilarating for me to return to the hippocampus, an old love found again. I had kept up with the advances in research, so it did not seem as though thirty years had passed. Per Andersen was a good friend, as was Roger Nicoll. But most of all I was motivated by memories of my experiments with Alden Spencer when we were both at NIH. I was feeling once again the excitement of being on the edge of something new—but this time armed with molecular genetic techniques whose power and specificity Alden and I could not have imagined in our wildest dreams.

THESE ADVANCES IN MOLECULAR GENETICS HAD THEIR
intellectual roots in the selective breeding of mice. Experiments at the turn of the twentieth century showed that various lines of mice differ not only in their genetic makeup but also in their behavior. Some lines proved extremely gifted at learning a variety of tasks, while others were exceptionally dull at those tasks. Such observations showed that genes contribute to learning. Animals differ similarly in their degree of fearfulness, sociability, and parenting ability. By inbreeding and creating some lines that are abnormally fearful and others that are not, behavioral geneticists overcame the randomness of natural selection. Selective breeding was thus the first step in isolating the genes responsible for particular behaviors. Recombinant DNA now made it possible both to try to identify the specific genes needed and to examine the role of those genes in the alteration of the synapses that underlie each behavior, emotional state, or learning capability.

Until 1980 molecular genetics in the mouse relied on a classical analysis known as forward genetics, which is the technique Benzer used in
Drosophila
. It begins by exposing mice to a chemical that usually damages only one of the 15,000 genes in the animal’s genome. The damage occurs at random, however, so which gene is affected is anyone’s guess. The animals are given a variety of tasks to see which, if any, are affected by the randomly altered gene. Because mice must be bred for several generations, forward genetics is very demanding and time-consuming, but it has the great advantage of being unbiased. There are no hypotheses and therefore no biases involved in screening for genes in this manner.

The recombinant DNA revolution enabled molecular biologists to develop a less demanding, less time-consuming strategy, reverse genetics. In reverse genetics, a specific gene is either removed from or introduced into the mouse’s genome and the effects on synaptic change and learning examined. Reverse genetics is biased—it is designed to test a specific hypothesis, such as whether a particular gene and the protein it encodes are involved in a particular behavior.

Two methods of modifying individual genes made reverse genetics in mice possible. The first, transgenesis, involves the introduction of a foreign gene, called a transgene, into the DNA of a mouse egg. Once the egg is fertilized, the transgene becomes part of the genome of the baby mouse. Adult transgenic mice are then bred to obtain a genetically pure strain of mice, all of which express the transgene. The second method of genetically modifying mice involves “knocking out” one gene in the mouse genome. This is done by inserting a segment of genetic material into the mouse’s DNA that renders the chosen gene dysfunctional and thus eliminates the protein encoded by that gene from the mouse’s body.

IT WAS BECOMING EVIDENT TO ME THAT WITH THESE ADVANCES
in genetic engineering, the mouse would be a superb experimental animal for identifying the genes and proteins responsible for the various forms of long-term potentiation. One could then relate those genes and proteins to the storage of spatial memory. Although mice are relatively simple mammals, they have a brain that is anatomically similar to that of humans and, as in humans, the hippocampus is involved in storing memories of places and objects. Moreover, mice breed much faster than larger mammals such as cats, dogs, monkeys, and people. As a result, large populations with identical genes, including identical transgenes or knockout genes, can be bred within months.

These revolutionary new experimental techniques also had major biomedical ramifications. Almost every gene in the human genome exists in several different versions, called alleles, which are present in different members of the human population. Genetic studies of human neurological and psychiatric disorders had made it possible to identify alleles that account for behavioral differences in normal people as well as alleles that underlie many neurological disorders, such as amyotrophic lateral sclerosis, early onset Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, and several forms of epilepsy. The ability to insert disease-causing alleles into the mouse genome and then study how they wreak havoc on the brain and on behavior revolutionized neurology.

The final stimulus for me to move into studying genetically engineered mice was the presence in our lab of several talented postdoctoral fellows, among them Seth Grant and Mark Mayford. Grant and Mayford knew mouse genetics much better than I, and they greatly influenced the direction of our research. Grant was the driving force behind my beginning to work on genetically modified mice, while Mayford’s critical thinking became important later, as we began to improve on the methodology we and others had used in our first generation of behavioral studies in mice.

Our original methods for producing transgenic mice affected every cell in the mouse’s body. We needed to find a way to restrict our genetic manipulations to the brain, specifically to the regions that form the neural circuits of explicit memory. Mayford developed ways of limiting the expression of newly implanted genes to certain regions of the brain. He also developed a method for controlling the timing of gene expression in the brain, thereby making it possible to turn the gene on and off. These two feats inaugurated a new stage in our studies and were widely adopted by other researchers; they remain cornerstones of the modern analysis of behavior in genetically modified mice.

THE FIRST ATTEMPT TO RELATE LONG-TERM POTENTIATION TO
spatial memory was made in the late 1980s. Richard Morris, a physiologist at the University of Edinburgh, had shown that by blocking the NMDA receptor pharmacologically, one could block long-term potentiation and thus interfere with spatial memory. In independent experiments, Grant and I at Columbia and Susumu Tonegawa and his postdoctoral fellow Alcino Silva at the Massachusetts Institute of Technology carried this analysis one important step further. We each created a different line of genetically modified mice that lacked a key protein thought to be involved in long-term potentiation. We then observed how learning and memory were affected in the genetically modified mice, compared with normal mice.

We tested the animals’ performance on several well-established spatial tasks. For example, we placed a mouse in the center of a large, white, well-illuminated circular platform surrounded by a rim of forty holes. Only one of the holes led to an escape chamber. The platform was in a small room, each of whose walls was decorated with a different, distinctive marking. Mice do not like being in an open space, especially a well-lit one. They feel defenseless and try to escape. The only way a mouse could escape from the platform was to find the hole that led to the escape chamber. Ultimately, the mouse found that hole by learning the spatial relationship between the hole and the markings on the wall.

In trying to escape, mice use three strategies in sequence: random, serial, and spatial. Each strategy allows the animals to find the escape hatch, but each varies greatly in efficiency. The mice first go to any hole at random and quickly learn that this strategy is not efficient. Next, they begin with one hole and then try each consecutive hole until they find the right one. This is a better strategy but still not optimal. Neither strategy is spatial—neither requires the mice to have an internal map of the spatial organization of the environment stored in their brain—and neither strategy requires the hippocampus. Finally, the mice use a spatial strategy that does require the hippocampus. They learn to look to see which marked wall is aligned with the target hole, and then they make a beeline for that hole, using the marking on the wall as a guide. Most mice go through the first two strategies quickly and soon learn to use the spatial strategy.

We then focused on long-term potentiation in an area of the hippocampus called the Schaffer collateral pathway. Larry Squire of the University of California, San Diego had shown that damage to this one pathway produces a memory deficit similar to that experienced by H.M., Brenda Milner’s patient. We found that by knocking out a particular gene encoding a protein that is important for long-term potentiation, we could compromise synaptic strengthening in the Schaffer collateral pathway. Moreover, the genetic defect was correlated with a defect in the mouse’s spatial memory.

Each year the Cold Spring Harbor Laboratory holds a meeting devoted solely to one major topic in biology. The topic for 1992 was “The Cell Surface,” but because Susumu Tonegawa’s and our work relating genes to memory in the mouse was deemed so interesting, a new slot, unrelated to the cell surface, was created for the two of us so we could give back-to-back talks. Tonegawa and I presented our separate experiments on how knocking out a single gene inhibits both long-term potentiation in a pathway of the hippocampus and spatial memory. At the time, it was the most direct correlation that had been made between long-term potentiation and spatial memory. Soon thereafter, both of us went one step further and examined how long-term potentiation relates to the spatial map of the external environment represented in the hippocampus.

By the time of that meeting, Tonegawa and I already knew each other a bit. In the 1970s he had determined the genetic basis of antibody diversity, an extraordinary contribution to immunology for which he received the Nobel Prize in Physiology or Medicine in 1987. With this accomplishment behind him, he wanted to turn to the brain for new scientific worlds to conquer. He was a good friend of Richard Axel, who suggested that he talk to me.

The problem that most interested Tonegawa when he came to see me in 1987 was consciousness. I tried to encourage his enthusiasm for brain research while dissuading him from taking on consciousness, which was too difficult and too poorly defined for a molecular approach at that time. Susumu had started to use genetically modified mice to study the immune system, so it was natural and much more realistic for him to turn to learning and memory, which he did when Silva joined his lab.

Since 1992 many other research groups have obtained results parallel to our own. Although there are occasional, important exceptions to the link between disrupted long-term potentiation and deficient spatial memory, it has nevertheless proven to be a good place to start examining the molecular mechanisms of long-term potentiation and the role of these molecules in memory storage.

I KNEW THAT SPATIAL MEMORY IN MICE, LIKE THE IMPLICIT
memory studied in
Aplysia
and
Drosophila
, has two components: a short-term memory that does not require protein synthesis and a long-term memory that does. Now I wanted to find out whether storage of explicit short- and long-term memory also has distinctive synaptic and molecular mechanisms. In
Aplysia
, short-term memory requires short-term synaptic changes that rely only on second-messenger signaling. Long-term memory requires more persistent synaptic changes that are based on alterations in gene expression as well.

My colleagues and I examined slices of the hippocampus taken from genetically modified mice and found that in each of the three major pathways of the hippocampus, long-term potentiation has two phases similar to those of long-term facilitation in
Aplysia
. A single train of electrical stimuli produces a transient, early phase of long-term potentiation that lasts only one to three hours and does not require the synthesis of new protein. The reaction of neurons to those stimuli was just as Roger Nicoll had described: NMDA receptors in the postsynaptic cell are activated, leading to the flow of calcium ions into the postsynaptic cell. Here calcium acts as a second messenger; it triggers long-term potentiation by enhancing the existing AMPA receptors’ response to glutamate and by stimulating the insertion of new AMPA receptors into the membrane of the postsynaptic cell. In response to certain patterns of stimulation, the postsynaptic cell also sends a signal back to the presynaptic cell, calling for more glutamate.