In the middle of the twentieth century, scientists began to hypothesize about the connections between neurons. In 1949, the Canadian psychologist Donald O. Hebb speculated that a structural memory trace in the brain is the foundation for long-term memory formation. Hebb’s idea was that learning causes growth in brain structures, thereby establishing memory traces. His thinking was influenced by Spanish anatomist Santiago Ramón y Cajal who in 1894 wrote that “mental exercise” likely resulted in growth on axons and dendrites. Hebb adopted this idea and took it further. In considering what happens at a synapse when one neuron talks to its neighbors, Hebb proposed that when one cell excites another repeatedly, tiny structures on both sides of the synapse swell. (In current terminology, the structures on the axon are called axonal varicosities and those on the dendrites are called dendritic spines.) This growth makes it more likely that, in the future, the first cell will activate the second cell again. When animals and humans learn new information, several cells in close proximity are repeatedly excited at the same time, forming a closed circuit that strengthens gradually as learning progresses. This postulate, known as Hebb’s rule, pinpointed the synapse as a critical location to uncover the physiological basis of learning and memory. At that time, Hebb had no direct physiological proof that closed pathways, or loop circuits, participated in behavior or learning. As it turns out, however, his visionary hypothesis about the brain’s flexibility was correct—Hebbian plasticity does exist, and Hebb’s influence continues as neuroscientists investigate whether this kind of plasticity is, in fact, responsible for learning and memory.23
This line of inquiry advanced considerably in the late 1960s with the discovery of the phenomenon of long-term potentiation (LTP), which many neuroscientists now believe to be the physiological underpinning of learning and memory. In 1966, Terje Lømo, a PhD student at the University of Oslo, performed experiments on anaesthetized rabbits to explore the role of the hippocampus in short-term memory. When Lømo applied a series of fast pulses of electrical stimulation to the axons that carried information into a rabbit’s hippocampus, he found that after each successive zapping, neurons on the other side of the synapse in the hippocampus responded to the same input more quickly, more strongly, and in greater number than they had previously. The stimulation strengthened the transmission of information from one cell to another, much like turning up the volume on a radio. A critical characteristic of the enhancement was that it lasted for more than an hour. Lømo called this new finding frequency potentiation, and showed that it could be induced by repeatedly activating the axon of one cell, which generated a signal that crossed the synapse and caused a firing increase in the cell receiving this input. After further studies in rats, researchers in the early 1970s changed the name of this phenomenon to long-lasting potentiation, and later in the same decade to long-term potentiation.24
The discovery of LTP provided a versatile model for the study of memory formation in many species of animals. Thousands of researchers on several continents continue to explore the molecular and cellular mechanisms by which specific patterns of activation (that is, different experiences) alter the strength of connections between neurons. LTP provides striking evidence of neuroplasticity, the brain’s ability to change with experience. Neuroscientists cite two key concepts in studying the changing brain: structural plasticity and functional plasticity. Examining structural plasticity has shown us that the anatomy of the hippocampus is not rigidly fixed for life: dendrites and their synapses change continuously in response to experience. Functional plasticity illustrates the property of synapses in the hippocampus and other brain areas to increase or decrease in strength—in essence, the capacity for the activity of one neuron to excite other neurons. At the core of memory is the brain’s ability to change as the result of experience, and LTP is an excellent laboratory example of both structural and functional plasticity.25
A wealth of research conducted after the discovery of LTP fleshed out its three basic features. First, potentiation is lasting and may persist from a few hours to a few days, and even up to a year (persistence). Second, potentiation is restricted to neural pathways that are active when the delivery of specific patterns of stimulation begins the process of encoding new information (input specificity). Third, the neuron on the sending side of the synapse and that on the receiving side of the synapse must be active simultaneously (associativity).26
In the mid-1980s, a major question remained unanswered: is LTP responsible for learning, as observed in the performance of an animal engaged in a memory task? In other words, could learning and memory take place when researchers prevented LTP from happening? A 1986 study addressed this question and provided further evidence that deficient LTP is linked to spatial amnesia. Neuroscientists at the University of Edinburgh, in collaboration with colleagues at the University of California, Irvine, trained normal rats to swim to a platform hidden in a pool of opaque water (later christened the Morris water maze). After a few days of practice, one group of rats figured out where the platform was located in the pool and was able to use it to climb out. In another group of rats, the researchers delivered a drug that shut down LTP in their hippocampi as they tried to learn the task. These rats had trouble finding the platform. This result, a clear indication of a deficit in spatial memory related to the blockade of LTP, is similar to Henry’s trouble finding his way back to his new house after his family had moved.27
A major advance over the pharmacologic blockade method came in 1996 when several papers from the laboratories of two Nobel Laureates, Susumu Tonegawa and Eric Kandel, heralded a revolution in the quest to understand the kind of learning and memory that depends on the hippocampus. These distinguished researchers and their numerous collaborators used the powerful gene knockout technology to exclusively remove the NMDA (N-methyl-D-aspartate) gene from a specific kind of neuron—pyramidal cells—located in three separate parts of the mouse hippocampus. When LTP occurs, active cells on the sending side of synapses release a neurotransmitter, glutamate, which opens doors called NMDA receptors on the receiving side of the synapse—if these receiving cells are simultaneously active. This communication initiates processes that result in the protein synthesis and structural changes that help nail down the to-be-remembered event and make the synapse more effective (potentiated).28
By disabling the NMDA gene selectively in one area at a time, Tonegawa, Kandel, and their colleagues could describe what roles this gene played in memory formation, and they found that the CA1 and CA3 areas in the hippocampus had distinct specializations. When the NMDA receptor deletion targeted CA1, the mice were impaired on the Morris water maze, requiring more time than sibling mice with intact NMDA receptors to swim to and climb up on the hidden platform. By contrast, the researchers discovered that another part of the hippocampus, the CA3 region, plays a different role in memory. There, NMDA receptors were necessary for pattern completion, which occurs when animals have to retrieve an entire memory after they are given only a fragment of that memory as a cue. These experiments using the genetic blockade method signified a major advance because they targeted specific cells in the hippocampus and conclusively established the role of receptor-dependent synaptic plasticity in spatial memory.
But do humans exhibit LTP? Since the late 1990s, labs in Germany, Austria, Canada, Australia, and England have been inducing LTP in the human hippocampus, motor cortex, and spinal cord. In some people, LTP may be maladaptive, either when it is enhanced or decreased, and it is possible that some neurological and psychiatric disorders result from too much or too little LTP. This possibility opens up numerous treatment options for the millions of people made miserable by these conditions. An outstanding characteristic of the human brain is its plasticity—its capacity to change with experience. By capitalizing on this potential, it should be possible to correct dysfunctional LTP. An exciting prospect is that memory loss, epilepsy, chronic pain, anxiety, addiction, and other conditions that have been linked to dysfunctional LTP can be reduced by de
ploying, at strategic points in the nervous system, one of the many chemicals that control LTP.29
It appears more and more plausible that LTP is necessary for learning, but there is much we do not understand. So far, scientists have been unable to prove that LTP is as enduring as our memories; LTP lasts for weeks at most, whereas long-term memories may last for decades. Neuroscientists are also working to understand how the cellular-molecular mechanisms that they have observed directly in the laboratory relate to the encoding, storage, and retrieval of specific memories in everyday life. We have a lot of ground to cover before we can bridge the gap between processes in our hippocampal cells and how well we do on a written test for a driver’s license.
Some people deny they dream, and others say they always forget their dreams. This is because fully recollecting our dreams requires some work. To document our dream content, we have to have a pad and pen next to our bed so we can record each dream immediately upon awakening, before it slips away. Dreams typically incorporate experiences from our past and may also play a role in memory consolidation. Still, we do not yet have direct evidence that dreams are required for memory consolidation, so we must interpret experiments relating dreams and memory cautiously.
To better understand how memories are consolidated—anchored—researchers in the mid 1990s embarked on sleep studies in rats. These experiments documented mental content during sleep by means of electrodes placed in the rats’ hippocampus. The specific patterns of neural activity recorded during sleep were compared with recordings made when the same animals were awake. This comparison often showed a clear correspondence between the two sets of recordings, giving some insight into the role of sleep in remembering awake experiences.
This research evolved from the major discovery of place cells in the hippocampus. In 1971, neuroscientists at University College London identified specialized neurons in the rat’s hippocampus that signal the animal’s current location in space. Each place cell corresponds to a particular zone in the rat’s space—the cell’s place field. When the cell fires, it signals to the rat where it is and the direction in which it is facing. Place cells are activated, for instance, when a rat is placed in a maze and has to find the way to get a treat. Together, these cells map the rat’s environment. Place fields provide the best example we have of how the rat’s world is represented inside its hippocampus.30
Since this discovery, place cells in rats and mice have attracted enormous experimental interest. When these animals perform in maze experiments, the cells fire in patterns and sequences that correspond to different locations in the maze, pinpointing where the animal is running or stopping. Even more intriguing is that these place cells are reactivated in the same order after the rat is removed from the maze. That is, when the rats are quiet—sleeping or pausing—their place cells replay the pattern of neural activity that occurred during a previous trip.
How does the activity of place cells affect the formation of long-term memories? In 1997, neuroscientists at the University of Arizona proposed that the hippocampus facilitates the reactivation of cortical activity patterns during offline periods, such as sleep or quiet wakefulness, when the cortex is less engaged in processing incoming information. To explore this possibility, the researchers inserted electrodes next to place cells in the rats’ hippocampus to record their activity. Each recording session had three phases—sleep, maze running, and sleep. The researchers predicted that during the second sleep period, neuronal firing in the hippocampus would resemble that in the cortex, and, in addition, the pattern of activity would correspond to that during maze running. The results verified their predictions. During sleep, the patterns of neuronal firing when the rat was in the maze were re-expressed in the hippocampus and cortex, and the mental representations, the paths, in the two areas resembled the maze-running representation that preceded the second sleep period. This correspondence suggests that circuits in the hippocampus and cortex were interacting during sleep, but the question lingered: Did this transient activity play a role in long-term memory?31
With the Arizona study as a springboard, MIT neuroscientist Matthew Wilson and his colleagues conducted experiments in freely moving rats and mice, from which they simultaneously recorded the firing of about one hundred cells. The main question driving this research was, how do large populations of cells in the hippocampus form and retain memories? The answer came from recordings of the activity of place cells in the animal’s hippocampus. These animals wore little hats that held a number of miniature recording electrodes—tetrodes—that allowed the scientists to eavesdrop on many neurons simultaneously. Together, these data gave researchers a realistic picture of the electrical activity of numerous place cells at any given moment.
To compare brain activity during sleep and wakefulness, researchers monitor EEG recordings and divide sleep into separate stages characterized by different electrical activity. In humans, slow-wave sleep, a deep sleep, is most common during the first half of the night, and REM (rapid eye movement) sleep, a lighter sleep, occurs mainly during the second half. Animals also experience slow-wave and REM sleep.32
In a 2001 experiment, the Wilson Lab recorded the activity of place cells in the rats’ hippocampus, first for ten to fifteen minutes as the animals ran a maze to get a treat, and later from the same cells for one to two hours as the rats were sleeping. When the researchers compared the behavior of cells in the hippocampus during maze running and during subsequent REM sleep, they found a remarkable correspondence—the two sets of data were strikingly similar, suggesting that the sleeping rats were replaying the behavior learned previously in the maze. Hippocampal neurons fired in the same order during REM sleep as they had during learning.33
These patterns of neural activity, representing a behavioral sequence, lasted as long as the real experience did. Repetition of the firing patterns of a collection of individual cells is known as memory replay. The awake rats encoded the sequences in a specific part of their hippocampus (area CA1), and this electrical activity was still detectable twenty-four hours later during REM sleep. This recapitulation of neural activity linked to a previous behavioral experience is strong evidence of a persistent memory. CA1 place fields encoded the location of the animal in space, and from this information, its brain assembled sequences of locations. Cross talk between the hippocampus and brain areas in the cortex that are specialized for spatial abilities likely contributed to this achievement. The replay of awake activity during REM sleep may enhance memory consolidation in the cortex as a result of interactions between cortical and hippocampal circuits. This and many other hypotheses still need to be explored before we can definitively nail down the role of memory replay in learning and consolidation, but a large body of evidence supports the view that hippocampal place cells remember.34
Further evidence that memory replay is an important component of learning and long-term memory came from Wilson and colleagues’ examination of brain activity during slow-wave sleep. They found that the effect on memory was different from REM sleep. They trained rats to run to and fro on a track, rewarding them with a piece of chocolate every time they reached one end or the other. As the rats indulged their sweet tooth, the researchers recorded the activity of many cells in the hippocampus simultaneously. They then monitored the same cells while the animals slept. The rats’ awake experience was replayed in the hippocampus during slow-wave sleep, only now at high speed—a four-second lap on the track was replayed fifteen to twenty times faster in the brain during slow-wave sleep. The memories that were reactivated reflected the order of the events, not their actual duration. These ordered events could then be played back over a shorter time than the animal needed to execute them, making them appear compressed.35
If we create a picture in our mind’s eye of the route we take from our home to the supermarket, and then imagine ourselves traversing this route, our mental trip will take much less time than it actually takes to go by car. We no doubt compress our dream content in the same way. In Wilson’
s experiment, this kind of memory replay was most likely to occur in the first few hours of slow-wave sleep following the experience, so its role may be in the early processing of information, prior to the establishment of long-term memory.36
Because memories are stored throughout the cortex, cortical activity must be part of the process of memory consolidation that occurs during sleep. We know that the hippocampus and cortex must work in partnership to receive, organize, and retrieve knowledge, but how do remote brain areas collaborate? Wilson’s team contributed another important insight in 2007 when they observed a close association, during memory replay, between cellular activity in the hippocampus and in the cortex. The researchers trained normal rats to run in a figure-eight-shaped maze. The rats began each experimental session in the middle of the eight and, to receive a food reward, had to run alternately to the left and to the right. After three weeks of training, the researchers implanted small electrodes in the rats’ hippocampus and visual cortex—a sensory processing area in the back of the brain that receives information from the eyes—and recorded patterns of neuronal firing from the two areas. The scientists found evidence of memory replay during slow-wave sleep in both, suggesting that rats, like people, have visual dreams. The activity patterns in the hippocampus resembled those in the visual cortex.37
It would be wonderful if all we needed to have better memory was a good night’s sleep packed with memory replay. Although that possibility is remote, growing evidence shows that memory consolidation and synaptic plasticity benefit from sleep. Experiments in humans that examine the effects of different sleep stages on consolidation now focus on the links between particular kinds of memory performance—declarative and nondeclarative—and the type and duration of sleep. This knowledge enhances our understanding of memory, allowing us to go beyond behavioral experiments by looking inside the brain for the neural processes that support consolidation. By examining the many physiological changes that accompany sleep and sleep disorders, researchers may be able to formulate novel remedies for insomnia and memory impairment.
Permanent Present Tense Page 16
