You MUST REMEMBER THIS

Finding the master switch for long-term memory

BY JOHN B. CONNOLLY AND TIM TULLY

Alfredo Castaneda, Retablo of the Forgotten, 1994

N BOOK NINE OF HOMER'S EPIC THE ODYSSEY with a contrived, complex mathematical formula. After a hurricane carries the hero, Odysseus, and his several minutes' study Shereshevsky reproduced the for­ fleet of ships far off course, to the land of the mula with complete fidelity. Astoundingly, fifteen years lat­ I lotus-eaters. When the storm finally subsides, er, when Shereshevsky was asked to generate the formula, Odysseus sends two ofhis best men and a runner ashore to he did so without error. Such a "gift"-commonly called reconnoiter. The men fail to return, and so Odysseus sets photographic memory-was a double-edged sword for the offfrom his ships to rescue them. But the rescue party is ill Russian mnemonist. He had difficulty combining memo­ prepared for what it finds on shore. The missing men are ries ofthe same individual and thus struggled with person­ neither dead nor hostages; instead they survive in a dream­ al interaction. Indeed, Shereshevsky's memory so inter­ like state, devoid ofall recollections, feasting with the na­ fered with his ability to work that he ended his days as a tives on the fruits and blossoms of the lotus flower. They "memory man" in a music hall. have lost all memory ofwho and what they are: they have Most people, fortunately, inhabit the more hospitable lost their psyches. middle ground between lotus-eater and mnemonist. That Homer clearly understood that memory is an integral felicitous state turns out to be-like many other dynamic part ofwho a person is. Previous experiences inextricably biological processes-the net result of countervailing ac­ link a person to the perception ofselfand others, and they tivities that either activate or repress. In our work at the serve to color almost all behavior. Fortunately, the plight Cold Spring Harbor Laboratory in Cold Spring Harbor, of Odysseus's men is temporary; dragged back to their New York, we have devised ways of studying memory in ships at last, they regain their memories, their senses and Drosophila melanogaster, the common fruit fly. In principle, their identities. our experiments combine the classic experimental design For people suffering from genuine memory disorders, of the turn-of-the-century Russian physiologist Ivan however, the fates are not so kind. In 1968 the Russian Petrovich Pavlov with late-twentieth-century genetic en­ neuropsychologist Alexander R. Luria published The Mind gineering. First we try to create associations between pre­ ofa Mnemonist, a book devoted to the remarkable case of viously unrelated stimuli in the insects, measure the one Shereshevsky. Apparently Shereshevsky could remem­ strength of the associations and determine how long the ber everything he had ever encountered in his life. Luria memory ofthe associations persists. Then we seek the ge­ described one occasion when he presented Shereshevsky netic underpinnings of the associative process.

MaY/]lIl1e /996· THE SCIENCES 37 Rene Magritte, Souvenir de Voyage III, 1951

E HAVE CONFIRMED IN FLIES WHAT But our findings may go far beyond the humble fruit fly. psychologists and neurobiologists have All animals, it seems, from invertebrates to vertebrates, W long suspected in larger animals: memo- form long-lasting memories in basically the same way. Fur­ ries are formed in distinct phases, each new phase over­ thermore, the genes identified in fruit flies also occur in the lapping the preceding one. Short-term memory (STM) other staples of the biological laboratory-in mollusks, gives rise to middle-term memory (MTM), which under chicks, mice and rabbits-and in humans. To the extent certain conditions becomes consolidated into a long­ that their functions have not changed over the long evolu­ lasting memory. But more, by analyzing flies that carry tionary span since insects and mammals diverged, those single-gene mutations, we have discovered that each genes may hold the key to the understanding ofmemory memory phase is closely associated with the function of in humans. Such a prospect offers both promise and threat. certain discrete sets of genes. In particular, we have been There is a hope that people with disorders of memory­ able to identify a gene that can enhance and suppress long­ Alzheimer's disease, for instance-could one day be helped lasting-memory formation in flies: We can make a fly into with drugs developed through our genetic insights. But a mnemonist or a lotus-eater. Our work is the first exam­ there is concern as well-concern that, as with all drugs ple of a genetic manipulation that enhances long-lasting ushered in from the new era of medical genetics, the use memory in any organism. Thus it begins to show what or abuse ofa pill ofmemory or a shot offorgetfulness will really makes memory tick. be solely at the discretion ofthe drug administrator.

38 THE SCIENCES· May/Julle 1996 O ST PEOPLE ARE AWARE OF THEIR ABILITY The second feature oflong-lasting memory, derived from to remember both recent and long-past experiments on animals, is that its formation depends on M events-the weather yesterday, as well as protein synthesis. The emerging view among neurobiolo­ the toys and stuffed animals they had as children. But how, gists is that memory is ultimately stored as a permanent exactly, do they remember such events? Do memories ex­ change in the way certain neurons communicate with each ist as isolated facts or discrete processes, quanta of infor­ other in the brain. Neurons connect to each other by way mation stored in some well-demarcated region of the of synapses, which they have in abundance. When long­ brain? Or are memories in some way holographic, diffused lasting memory appears, existing synaptic connections seem over the entire nervous system? to strengthen or grow. Proteins, synthesized within the neu­ In the 1950s, after years ofeffort to find the seat ofmem­ rons, are necessary raw materials for that process, in the same ory in the brain, the American psychologist Karl S. Lash­ way bricks are necessary for extensions to a brick house. ley rejected the idea that memories are localized, and he The classic experiments demonstrating the need for conjectured that memories protein synthesis in long­ are "statistical features of THE GENE WE IDENTIFIED CAN ENHANCE lasting-memory formation temporal patterns." Since were done in 1963. Wesley then, however, numerous or suppress long-term memory. C. Dingman, a psychiatrist studies of people whose We can make a fruit fly at Chestnut Hill Hospital in brains are partly incapaci­ Rockville, Maryland, and tated have led to what is into a mnemonist or a lotus-eater. Michael B. Sporn, a phar- now the prevailing view macologist at the Dart­ among neurobiologists: that mouth School of Medicine specific memories are indeed stored in specific sites. In in Hanover, New Hampshire, injected rats with a drug that 1904 the German biologist Richard Semon coined the inhibits protein synthesis. When the rats were injected just term engram for the site ofmemory storage-where mem­ before being trained to negotiate a water maze, they quick­ ory would physically manifest itself as a "change of the ly forgot what they had learned about the maze. But when organic substance." the rats were injected only after the training period, they re­ The classiccase study ofan epileptic man known as H.M. membered more about the maze; and the later the injec­ demonstrated one ofthe most striking features ofmemory tions, the more they remembered. After a certain interval, storage: whatever underlies long-lasting memory appears to what they learned about the maze was "committed to be physically distinct from learning and from short-term memory," and the subsequent injection of the inhibitor memory. In 1953, to quell his severe bouts ofepilepsy, sur­ drug had no effect. Thus a long-term memory (LTM) de­ geons removed significant portions of the hippocampus, pendent on protein synthesis became progressively resistant amygdala and temporal lobe ofH.M.'s brain. Although the to inhibitors after training. effects of the epilepsy were attenuated, the surgery left H.M. unable to transfer new information into permanent OTH FEATURES OF LONG-LASTING-MEMORY memory, though he could remember new information for formation-its resistance to disruption by a short time. As a result, H.M. lives in a perpetual present B anesthesia and its dependence on protein syn- where "every day is alone by itself, whatever enjoyment I've thesis-exist in the subject ofour investigations, the fruit fly. had, and whatever sorrow I've had... ." His memories from We first train our flies, as Pavlov trained his dogs, to associ­ before the surgery, however, remain intact. ate a neutral stimulus with a stimulus that usually elicits a Long-lasting memory forms in many animal species, strong behavioral response. To do so, we trap about a hun­ and psychologists and neurobiologists have often pointed dred flies at a time in a cylindrical chamber much like a test to two general features ofthe process. One feature is retro­ tube, whose inner surface is covered with an electrifiable grade amnesia: newly acquired information can be lost if grid. At one end ofthe chamber we attach an "odor cup" one is subjected to head trauma, shock treatment, hypo­ that gives off one of two odors: octanol, which smells like thermia, anesthetics or insults that lead to unconscious­ licorice, or methylcyclohexanol, which smells a lot like ten­ ness. Typically, the amnesia reaches backward in time from nis shoes inJuly. By passing air through the chamber, we ex­ the moment ofthe unconsciousness to an earlier moment pose the flies to one of the odors, and we simultaneously before which memory is unaffected by the trauma. What electrify the grid on which they rest. We expose the flies to appears to take place is that as time passes and new infor­ the second odor in the absence ofelectroshock, as a control. mation gets "committed to memory," the memory be­ Once the flies are trained to associate one of the odors comes progressively resistant to disruption. For example, with electroshock, we test them at various times afterward after being knocked unconscious in a sledding accident on in a T maze. We place the flies at the junction ofthe T, be­ Christmas Day 1968, one of us (Tully) could remember tween converging air currents that carry one odor or the past events until as recently as the preceding December 12. other. Untrained flies show no preference for either odor; The ensuing two weeks, however, including the gift ex­ they distribute themselves in a fifty-fifty ratio in the two change earlier on Christmas Day, were permanently lost. arms of the T maze. But the trained flies are far from in­ The appearance of that so-called anesthesia-resistant different: 90 percent ofthem avoid the shock-paired odor memory has generally been interpreted as the earliest by running into the opposite arm of the maze. As time manifestation ofa stable long-lasting memory. passes, however, the flies' memories slowly fade, and after

May/June 1996' THE SCIENCES 39 about a day their preference for the odors in the T maze re­ the protein-synthesis inhibitor had no effect on the mem­ verts to indifference. ories that developed. The drug seemed to work only when Although a day is a long time in the life ofa fruit fly, it the training sessions were spaced out. Crucially, flies sub­ is not forever. We were able to induce long-term, perma­ jected to massed training seemed to acquire anesthesia­ nent memory in our flies by drawing upon work done resistant memory, which also seemed unaffected by the more than a century ago by the German psychologist Her­ protein-synthesis inhibitor. Those results were curious, be­ mann Ebbinghaus. In his 1885 book, Ober das Gedachtnis cause for thirty years neurobiologists and psychologists had (On memory), Ebbinghaus reported his discovery that a list assumed that the appearance ofARM and the requirement ofnonsense syllables can be memorized more accurately if for protein synthesis were two aspects ofthe same process: several training sessions are spaced out over time than ifthe the consolidation oflong-term memory. training is crammed into a single long session. School­ But how could we prove the obvious hypothesis: that teachers, of course, have been aware of the phenomenon contrary to the Widespread belief, anesthesia-resistant for years. Cramming before a test helps students only in the memory and long-term memory are physically distinct? short term. They retain more ifthey parcel out their study The key was the vast and detailed knowledge geneticists time over several intervals. have gained in the past century about the genetics of the We applied the same principle to our flies. We trained fruit fly. One strain ofthe fly, in particular, possesses a de­ them in ten sessions, with a fifteen-minute rest interval be­ fective, mutant copy ofa single gene known as radish. Flies tween each session. The memories we were able to create with an intact radish gene developed anesthesia-resistant in the flies then persisted indefinitely. memory, but flies with the defective radish gene did not. In a way, such a result was typical. Over time we and our UR PAVLOVIAN TRAINING SHOWED THAT colleagues at Caltech and the Massachusetts Institute of fruit flies exhibit many of the kinds of Technology have identified other strains offly with muta­ O memory seen in other animals. Most no- tions in single genes that are unable to form short-term or tably, memory forms in increasingly stable phases. Imme­ middle-term memories. But we also discovered that muta­ diately after training, flies have a burst ofshort-term mem­ tions in genes involved with STM or MTM disrupt all ory, which lasts for several minutes. STM is followed by downstream memory phases. That observation suggested middle-term memory, which lasts several hours and is fol­ there is sequential processing, at the genetic level, of the lowed by anesthesia-resistant memory (ARM). various memory phases: STM induces MTM, which in­ We have a simple trick for showing that flies develop duces ARM, which leads to LTM. ARM: we "cold-shock" them at progressively longer in­ But the radish mutant proved an important exception to tervals after one training session. We induce cold shock by that simple picture. Although flies with the mutant radish placing the flies in a test tube and submerging the test tube gene did not develop ARM, after spaced training they still in ice water until the flies become unconscious. After two developed memory that, apparently, was permanent. minutes we warm them up again (they regain conscious- Could such permanent memory reflect the appearance of LTM, for which proteins must be synthesized? The answer IN FRUIT FLIES, AS IN PEOPLE, is yes. When mutant radish flies were fed a protein-synthesis inhibitor, spaced training failed to give rise to long-term cramming helps only in the short term. memory. The observation that the mutant radish gene Both species retain more if they rest disrupts ARM but not LTM demonstrates a clear genetic dissection of the two properties of consolidated memory: between training sessions. anesthesia resistance and the dependence on protein syn­ thesis. Because the two kinds ofmemory can be indepen­ dently disrupted, the ARM and LTM phases of memory ness quickly, with no side effects), and we test their mem­ appear to be parallel processes rather than sequential ones. ories in the T maze three hours after the training. We find that ifthe flies are cold-shocked immediately after training, F ARM AND LTM ARE SO DISTINCT, WHAT ARE their three-hour memory is severely disrupted. As the in­ the genes specific to the formation oflong-terrn terval between training and cold shock becomes progres­ I memory? The responses ofnormal (that is, non- sively longer, however, the flies' memories three hours af­ mutant) flies to massed-training versus spaced-training reg­ ter training become more resistant to disruption. Those imens supplied the clue. We discovered that whereas features indicate the appearance ofARM. spaced training of normal flies gives rise to both anesthe­ Finally, we were able to generate long-term, nondecay­ sia-resistant memory and long-term memory, massed ing memory in our flies by repeating the training sessions training leads only to ARM. Even after forty-eight massed­ and spacing them out. More, we have shown that for that training cycles, done without intervals ofrest, the flies still final memory phase to form, proteins must be synthesized: formed no long-term memories. when we fed our flies a protein-synthesis-inhibiting drug, What is so important about the rest? Whatever proteins we found, just as Dingman and Sporn had with their rats, turn out to be associated with LTM, their concentrations that long-term memory failed to appear. somehow increase during the rest period. The concentra­ One finding that aroused our interest was that when we tion of a protein in a cell is controlled by the degree to trained our flies in one massed training session without rest, i which the gene coding for that protein is expressed. Typi-

40 THE SCIENCES· M"y/J""e 1996 cally, the expression ofsuch a gene is controlled in turn by training sessions to acquire long-term memory, flies that activator and repressor proteins that enable or prevent the synthesized higher than normal levels ofCREB activator de­ information carried by the gene from being transcribed. A veloped the usual amount oflong-term memory after only good guess about the control mechanism might be that the one training session. functional levels ofboth activator and repressor proteins in­ Such discoveries should hold in a general way for crease during the training. The functional level of repres­ people. As the wags sor protein should then fall more rapidly than that ofacti­ have it at Cold vator protein during the rest. The differential buildup of Spring Har­ activator protein would then account for the subsequent bor: "Flies buildup ofthe proteins directly associated with LTM. are flies and Workers in our laboratory have recently found precisely rmce are such a mechanism in the fruit fly. It has been known for people." some time that some of the earliest biochemical events in­ The ge­ volved in learning are mediated by the cyclic AMP, or nome of a cAMP, signal-transduction pathway within the cell. (The mouse IS cAMP pathway is a well-studied messenger system.) The virtually neurobiologist Eric R. Kandel and his colleagues at the identical to Columbia University College ofPhysicians and Surgeons in the New York have shown that learning in the mol­ lusk Aplysia is disrupted by perturbing cAMP signaling. The signals mediated by cAMP closely resemble the observed phases in the formation of memory: both are sequential, and both lead to increasingly stable molecular changes within neurons. It turns out that what is lacking in the ge­ netically mutant flies that form no short-term memory are enzymes that either generate or break down cAMP. Thus the biochemistry of learning appears to be virtually the same in mollusks as it is in flies. Farther down the cAMP­ signaling pathway is a molecule known as cAMP­ responsive element-binding protein, or CREB. The molecule dictates whether a cell will make new proteins in response to cAMP signaling. And therein lay the clue for LTM formation in flies. If the early events in cAMP signaling were respon­ sible for the early events of memory formation, perhaps molecules acting later in the same path­ way mediated the protein synthesis on which long-term memory depends. Ifso, CREB was an obvious candidate.

FTER YEARS OF PAINSTAKING molecular research, Jerry C. P. A Yin, a colleague at Cold Spring Harbor Laboratory, identified a gene encod­ ing CREB in flies, known as dCREB2. The dCREB2 gene is complex and encodes many forms ofCREB protein, including a repres­ sor form to turn off CREB activity and an activator form to switch it on. Yin showed that the activities of each form are highly specific. Flies in which the repressor form ofthe protein was produced at artificially high levels showed no sign oflong-term memory, though short-term, middle­ term and anesthesia-resistant memory all formed normally. Even more dramatic, whereas normal flies required ten spaced- Memory board (lukasa), Luba, Zaire, early-twentieth century

May/June 1996' THE SCIENCES 41 genome ofa person. Our laboratory colleague Alcino Sil­ factors involved in such aspects oflife as behavior, identi­ va studied memory formation in mice with a mutant ty, memory and thought becomes astronomical. It makes CREB gene and observed an outcome similar to the one little sense to characterize the "meaning" or "function" of we observed in flies. Short-term memory remained nor­ a gene, when in the vast majority of biological circum­ mal, but long-term memory did not form. CREB seems to stances its effect depends on its interactions in combination be the master memory switch in flies, mice and, by genet­ with hundreds ofother genetic and environmental factors. ic implication, people. In short, genes to not determine behavior. There is tantalizing evidence that CREB misregulation That principle is particularly true in the formation of may account for some human cognitive disorders. For memory. The discovery of a master switch for long-term CREB to activate the synthesis ofnew protein, it must team memory reveals nothing about which memories will be up with a molecule called CREB-binding protein (CBP). stored or what shape they will take in storage. Rarely are Recently, a link has been reported between disruption of events experienced in isolation; generally, a person brings a the CBP gene in people good part of the past to and Rubinstein-Taybi GENES DO NOT DETERMINE BEHAVIOR, bear on present circum­ syndrome. Among the stances. Memory making clinical features of the and, perforce, they cannot determine memories. in the present is undoubt­ syndrome are mental re­ But who would not rejoice if edly affected by the mem­ tardation and physical ab­ ories already established. normalities ofthe thumbs a memory pill became available? Science may never be able and toes. Analyses have to fathom the complexity shown that Rubinstein- ofpast, present and future Taybi patients carry mutant forms ofthe CBP gene or have events and interactions as they take place in the mind. The microdeletions in the region of the chromosome that most immediate danger ofmemory pills, if any become avail­ includes that gene. able, is that most oftheir effects will be unpredictable. We have only begun to consider how CREB functions The most serious worry about such a technology, in humans. We speculate that CREB may act as an infor­ though, may be what is entirely predictable. The burden of mation filter for most tasks, ensuring that only recurrent an overpowering memory, as the case of Shereshevsky events become committed to long-term memory. If so, shows all too clearly, may be unsupportable. Indeed, much many new questions arise: Exactly how does the differen­ is known about eidetic, or photographic, memory from tial activity between CREB repressors and activators unfold studies of its presence in elementary-school children. As during the rest intervals between training sessions? What many as half such children possess it up to puberty, after genes does CREB target? Where are the long-term mem­ which it disappears in all but a few. Eidetic memory prob­ ory cells, the elusive engrams? How does CREB activity in ably helps a developing mind assimilate new facts in early the nucleus of a neuron target only a small group of spe­ life. But as children reach adulthood, eidetic memory gives cific synapses while leaving unmodified thousands ofoth­ way to the unconscious process offiltering, sorting, evalu­ er synapses in the same neuron. ating and overlooking that is necessary for living in a world Answers to such questions could lead to enormous hu­ in constant flux. Chronically circumventing that process man benefits. Understanding the workings ofproteins en­ through pharmaceutical memory enhancement could coded by memory-related genes may enable pharmacolo­ bring about inconceivable difficulties. gists to develop drug treatments for patients suffering from memory lapses, including the terrible losses caused by dis­ T THE SAME TIME, FEW THINGS IN DAILY LIFE eases such as Alzheimer's. Because memory is such an in­ are more frustrating than the experience of tegral part of human identity, the hope exists that the far A forgetting. Who would not rejoice to be larger group of people suffering from such mental disor­ spared the embarrassment of blocking a person's name, of ders as anxiety and schizophrenia could also be helped. blanking on a speech, of stumbling over the performance ofa song? What but cause for celebration would it be ifag­ ET THAT HOPE SHOULD BE TEMPERED WITH ing did not bring the humiliation and loss ofselfthat comes caution. Basic genetic research has taught bi­ with deep forgetting? Ifwe honor Homer for his humane Y ologists that the activity of a gene is rarely understanding of memory, we also marvel at the virtuosi­ isolated from the environment or even from the rest ofthe ty ofhis memory in performance. Who today could com­ genome. For example, as long ago as 1907 experiments mit to memory 15,000 lines ofverse, as Homer did? In an with magnesium concentrations in water showed that the age ofscripts and teleprompters, how wonderful it would environment can dramatically alter gene expression. Living be to travel light, to kick away the mental scaffolds and in water with high concentrations ofmagnesium, fish de­ speak from memory, confident that our faculties will not velop not two eyes, but one. The "nature versus nurture" fail. That, in part, is the promise and vision ofour work. • debate in biology must be dismissed as oversimplified and JOHN B. CONNOLLY is a visiting scientist at the Cold Spring anachronistic by contemporary genetic research. Harbor Laboratory, on Long Island, New York; he is affiliated Interaction among many genes, for instance, is the rule with the Department oj Genetics at the University oj Cambridge rather than the exception. Ifone further considers the role in England. TIM TULLY is a senior staff scientist at the Cold the environment plays in gene expression, the number of Spring Harbor Laboratory.

42 THE SCIENCES· MayIJ""e 1996