Protein Synthesis in Long-Term Memory in Vertebrates
PROTEIN SYNTHESIS IN LONG-TERM MEMORY IN VERTEBRATES
Most modern theories assume that memories are stored in the brain in the form of changed patterns of synaptic connections within ensembles of neurons, although it remains debated whether such patterns are stable once formed or subject to dynamic change. Any such growth or reorganization of synapses requires the synthesis of the molecules comprising them, especially the proteins and lipids of the synaptic and dendritic membranes. The idea that memory formation involves protein synthesis has been around for a long time—certainly since the days of Santiago Ramón y Cajal (Spanish histologist, 1852-1934) at the beginning of the twentieth century—but serious experimental tests of the idea became possible only with techniques available beginning in the 1960s. The adult brain has one of the highest rates of protein synthesis of any body organ, and also shows the greatest diversity of protein molecules. Some 30,000 different proteins are synthesized in the primate brain, only a small fraction of which have been characterized or ascribed functional roles. Many protein species are likely to be involved in memory consolidation, and it is unlikely that any of them are uniquely required; what provides the specificity for the memory trace is not the particular protein but the new pattern of synapses it helps to construct.
Two general approaches, sometimes described as interventive and correlative, have been used to identify these proteins. In the first, the behavioral consequences of either blocking protein function or preventing protein synthesis from occurring during or after training an animal on a particular task have been studied. If protein synthesis is necessary for memory formation, animals in which the synthesis has been blocked at the time of training should be amnesic for the task when subsequently tested. The second approach attempts to measure an increase in the synthesis or turnover of particular proteins as a result of the training experience. Both approaches have methodological pitfalls. Preventing protein expression or blocking function may have general effects on behavior, sensory or motor performance, or arousal, and therefore exert an effect on memory nonspecifically. Similarly, increases in protein synthesis following training could be the consequence of nonspecific behaviors such as the motor activity involved in carrying out the training task. Ruling out such alternative explanations requires devising careful control experiments.
Proteins are synthesized in a series of steps beginning with the copying (transcribing) of strands of DNA into messenger RNA (mRNA), after which the mRNA is translated into the sequence of amino acids that constitutes the protein. Many proteins, especially those of the cell and synaptic membranes, are further modified posttranslationally. For instance, to synthesize glycoproteins, which are key constituents of the cell membrane, it is necessary to add sugar molecules to the protein chains before transporting them to the cellular sites at which they will function. Inhibiting any one of the steps in this synthetic sequence from the DNA to the finished molecule may block the synthesis of a protein or render it nonfunctional, and hence interfere with memory formation. The first experiment of this sort was made in the mid-1960s, using an inhibitor of RNA synthesis, 8-azaguanine. Rats were trained to swim a water maze and then injected with the inhibitor. It had no effect on the performance of animals that had already learned the maze, or on their ability to swim in general. However, if the inhibitor was injected during the training trials, and the rats were tested the following day, they showed impaired memory for the maze.
In the decades that followed, experimenters employed various antibiotics that interfere with particular steps in protein synthesis: puromycin, cycloheximide, acetoxycycloheximide, and anisomycin. The consensus observation, in a variety of appetitive and aversive paradigms and with several species including rodents, birds, fish, molluscs and insects, is that concentrations of antibiotic sufficient to inhibit more than eighty percent (e.g., 80%) of all protein synthesis in the brain for several hours are, perhaps surprisingly, without effect on performance of already-learned tasks or other aspects of behavior. However, if the inhibitors are injected within specific time windows relative to the time of training (or testing), they will produce amnesia in animals tested twenty-four hours or more later. Behavioral controls rule out the possibility that these amnestic effects are due to some form of state dependency, and although the inhibitors have a variety of less specific biochemical effects (notably increasing the concentrations of intracellular amino acids, including several that are neurotransmitters and can be neurotoxic), it is generally agreed that they do indeed exert their amnestic effect by preventing the synthesis of proteins necessary for memory formation. More recently attention has turned to the blockade of specific proteins or protein classes. Glycoprotein synthesis can be blocked with the metabolic analogue 2-deoxygalactose. The expression of specific proteins can be prevented in mice or fruit flies by transgenic techniques such as the use of inducible knockouts, and in many species by antisense RNAs targeted against the specific transcribed RNA sequences coding for a particular protein. Appropriate antibodies can also functionally block proteins, especially membrane proteins, whose structure makes them accessible in vivo. Thus a monoclonal antibody raised against a key synaptic membrane constituent, the neural cell adhesion molecule N-CAM, a central player in the processes of cell-cell recognition, is amnestic if injected into rats trained in a passive avoidance task some six to eight hours after training. Antisense molecules that prevent the synthesis of the amyloid precursor protein, APP, or N-CAM, are also amnestic. Mouse knockouts lacking the cyclic AMP response element binding protein (CREB) also show characteristic learning deficits.
Like the interventive approach, correlative studies began in the 1960s. Because it is unlikely that there will be measurable increases in the total amount of proteins in general or of specific proteins in particular during memory formation, the initial approach was to measure the rate of protein synthesis during training by the use of a radioactive precursor. A radioactively labeled amino acid, injected into the bloodstream, is taken into the brain and there becomes incorporated into protein. The amount of radioactivity found in the protein after a fixed time interval then depends, among other factors (which need to be controlled for), on the rate of synthesis of the protein. If more radioactivity is found in brains of trained than of control animals, this is assumed to indicate that the training procedure has resulted in enhanced synthesis (or turnover), and the behavioral question then becomes that of ensuring that it is memory formation rather than some other aspect of the task that has increased the synthesis.
Some of the clearest evidence for enhanced protein synthesis using this approach came from studies of early learning in the chick, especially imprinting and one-trial passive avoidance training. These tasks involve strong and biologically programmed learning in an otherwise naive animal, and therefore maximize the chance of finding changes. For example, training on the passive avoidance task results in a long lasting increase in incorporation of radioactive amino acids into the proteins of specific brain regions.
To study such increases in more detail, it is necessary to know which brain regions might be involved. Here, an autoradiographic mapping technique can be employed in which, after incorporation of the radioactivity, the brains are sectioned and apposed to X-ray film, and the specific regions showing training- related increases identified by image analysis. Such techniques have been used to show enhanced protein synthesis in the rat hippocampus during a variety of learning paradigms.
The next task becomes that of identifying which of the many proteins are involved. Various approaches are possible. Subcellular fractionation can show in which cellular compartments the new proteins are most concentrated. In both rat and chick there are increases both in soluble proteins such as tubulin and in synaptic membrane constituents such as N-CAM. In an unusual training task in goldfish—in which the animal has a float attached to its belly that inverts it, so that it has to learn to swim correctly once more—a different class of low-molecular weight soluble glyco-proteins, named ependymins, has been identified by similar techniques. This class of molecules has also been shown to be relevant in mammalian learning. Other imaging techniques, such as in situ hybridization or immunocytochemistry, can define specific cellular regions, such as dendritic spines, as showing increased levels of particular messenger RNAs or proteins.
Memory Consolidation and Memory Phases
The demonstration of the time windows during which protein synthesis inhibitors are amnestic has been an important piece of evidence in developing stage theories of time-dependent processes in memory formation. Thus it was early demonstrated that if protein synthesis inhibitors were administered around the time of training in, for instance rodents or chicks, then animals could learn and show memory retention for periods of an hour or so, after which a progressive amnesia set in. This was taken to indicate a distinction between short-term memory, not requiring new protein synthesis, and long-term memory, for which such synthesis was necessary. Giving the inhibitor one to two hour posttraining no longer results in amnesia and hence memory was said to be now stable and protein synthesis independent. However, this proved to be simplistic; a second time window four to six hours downstream of the training experience was found in a number of paradigms during which memory consolidation once more becomes sensitive to protein synthesis inhibitors.
The current working hypothesis is that the initial stages of memory formation involve transient synaptic membrane events. These events include the phosphorylation of membrane proteins (including the presynaptic protein known as B50 or GAP43) by a membrane-bound enzyme, protein kinase C. This phosphorylation step activates a cycle of intracellular second messengers, including calcium ions, and transcription factors, such as CREB, which in turn trigger a genomic response in the neuronal cell nucleus. The initial genomic response is to switch on a family of specific immediate early genes whose protein products include in particular c-fos, c-jun, and zif. These proteins are expressed only during the early phases of neuronal plasticity, when neurons are growing or actively differentiating, and they have been shown to increase dramatically in concentration in a number of learning tasks, including brightness discrimination in the rat and passive avoidance in the chick, as well as hippocampal long-term potentiation. But c-fos and cjun, although they are excellent markers to show where in the brain neural plasticity is occurring, are themselves only intermediates; their production acts as a trigger for the activation of further genes (late genes) whose products include the proteins and glycoproteins already mentioned. It is the enhanced synthesis of these proteins, in the so-called second synthetic wave, four to six hours downstream of training, which makes consolidation once more sensitive to the general synthesis inhibitors as well as to specific antisense or antibodies against, for instance, the families of cell adhesion molecules such as N-CAM. These adhesion molecules have sticky glycosylated extracellular domains. Projecting both from pre- and postsynaptic sides, they can bind together rather like a molecular form of Velcro, holding the synaptic junction in a specific configuration. It is therefore assumed that these are the ultimate effector proteins required for lasting remodelling of synapses.
Consolidation and Reconsolidation
Recently even this conclusion concerning stable memory has been called into question. If an animal is trained on a task, and some time—hours to days later—given a reminder by being exposed once more to the original training situation, then the previously stable memory again becomes labile and sensitive to protein synthesis inhibitors. For some researchers, reminder plus inhibitor results in lasting amnesia, implying that in some way reactivating a memory requires retraversing the same biochemical pathway that was initially engaged. Others find that the effect of the inhibitors is transient, suggesting a temporary blockade of access to an otherwise unimpaired memory. The phenomenon is intriguing but its interpretation remains an arena of active debate.
Protein synthesis is universally necessary for the formation of long-term memory, although some forms of newly acquired memory may persist for long periods in the absence of such synthesis. The proteins involved are synthesized in increased amounts in specific cells in the hours following a training experience and during the period of memory consolidation. Different proteins are synthesized at different times after training in a time-and space-dependent sequence; in the first phase they include members of the immediate-early family, such as c-fos and c-jun. Later stages involve the synthesis of the microtubular protein tubulin and glycoprotein components of synaptic and dendritic membranes, including members of the cell adhesion family. Others certainly remain to be identified. If their expression or function is prevented, amnesia results. Some may also be required for the expression of old but newly activated memories. These proteins are not, however, in themselves specific to the particular memory. What conveys specificity is the pattern of neurons whose connections are modified by the learning experience; the proteins are part of the housekeeping processes involved in modifying those connections. Identifying more precisely the proteins involved, their locations, and their functions will help substantially in the development of theories of memory formation, in addition to being of major intrinsic neurobiological interest, and of potential relevance to the treatment of such conditions as Alzheimer's disease.
Davis, H. P., and Squire, L. R. (1984). Protein synthesis and memory: A review. Psychological Bulletin 96, 518-559.
DeZazzo, J., and Tully, T. (1995). Dissection of memory formation: From behavioral pharmacology to molecular genetics. Trends in Neuroscience 18, 212-217.
Dudai, Y. (1989). Neurobiology of memory. Oxford: Oxford University Press.
Goelet, P., Castelluci, S., Schachner, S., and Kandel, E. R. (1986). The long and the short of long-term memory—a molecular framework. Nature 322, 19-423.
Izquierdo, I., and Medina, J. H. (1997). Memory formation: The sequence of biochemical events in the hippocampus and its connection to activity in other brain structures. Neurobiology of Learning and Memory 68, 285-316.
Mileusnic, R., Lancashire, C., and Rose, S. P. R. (1999). Sequence specific impairment of memory formation by NCAM antisense oligonucleotides. Learning and Memory 6, 120-127.
Nader, K., Schafe, G. E., and LeDoux, J. E. (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature 406, 722-726.
Radyushkin, K. A., and Anokhin, K. V. (1999). Recovery of memory in chicks after disruption during learning: The reversibility of amnesia induced by protein synthesis inhibitors. Neuroscience and Behavioral Physiology 29 (1), 31-36.
Rose, S. P. R. (1992). The making of memory. London: Bantam.
—— (2000). God's organism? The chick as a model system for memory studies. Learning and Memory 7, 1-17.
Steven P. R.Rose