medium containing fucose. A bacterium grown under these conditions will respond to gradients of galactose, whereas a bacterium not subjected to this growth condition will be incapable of responding to galactose. Thus, exposure to fucose during youth has permanently modified the cell and would be a device for "learning" if appropriately generated in a more complex organism (Fahnestock and Koshland 1979). Overproduction of the receptor is achieved by the use of appropriate plasmids or chromosome manipulation. A cell with amplified receptor responds for much longer intervals to the same stimulus as a wild type cell. Hence, induction of new receptors could produce memory (Russo and Koshland 1983). Truncating the receptor to remove a few amino acids from the C-terminal end eliminates the capacity of the receptor to be methylated and hence to adapt. Such cells were found to be turned on by small amounts of signal that would be rapidly desensitized in a wild type cell. They could respond reversibly to the addition or removal of a stimulus, but could not adapt to an incessant stimulus. A proteolytic enzyme induced to remove a terminal peptide could thus record a learning experience (Russo and Koshland 1983). ● Finally, removal of the enzyme causing demethylation results in a cell that could potentiate a response (Rubik and Koshland 1978). Each of these modifications, if generated in a neuron in a complex organism, could result in permanent change in the firing pattern of that neuron. Hence, they are the ingredients from which a long-term memory system could be developed. Implications for the Future The understanding of memory is one of the most fascinating and important contemporary biological problems. Its loss through Alzheimer's disease, strokes, injuries, and other defects in elderly people is one of the great difficulties facing an aging population in the United States. The understanding of memory is the first step in devising methods to enhance it, to restore it in the case of loss, or to prevent its loss in the first place. Although memory seems perhaps one of the most complex biological processes, direct studies on neural cells by long-term potentiation (Teyler and DiScenna 1987) and studies of plasticity in neural systems (Levy and Desmond 1985) show many of the same biochemical responses as bacterial chemotaxis. That finding becomes less and less surprising as time goes on. The signaling systems of many types of cells are highly similar (Stryer 1988). Early hypotheses that there would be one type of signaling system for neural cells, another for kidney cells, a third for liver cells, and so on, seems clearly in error. The second-messenger systems of different types of cells are similar. The permutations by which the modules are put together make the essential differences between neural cells, hormonal cells, liver cells, etc. Cyclic AMP, phosphoinositide, and calcium second-messenger systems are present in all types of cells. Therefore, the understanding of a simple system in which memory can be manipulated and its correlation with long-term potentiation in mammalian cells can be established offers a major avenue by which the individual chips that serve to encode memory can be deciphered. In fact, work in our laboratory and elsewhere is already using tissue culture and cell lines to develop systems for studying long-term as well as short-term memory. The wiring diagram by which these neurons are held together requires different types of investigation, but it is impossible to establish that the cells are modified for long-term memory if one does not know what biochemical changes are needed to achieve such a property. The evidence that simple theories and manipulative experiments can start to throw light on this important problem indicates that progress is likely to be rapid and efficient in the decade ahead. Because memory is a sophisticated combination of relatively modest changes, it may be one of the first major sensory processes to be thoroughly understood. Its importance as a basis of mental function and its relevance to an aging population cannot be overemphasized. REFERENCES Bollag, G.E., and Koshland, D.E., Jr. A new model for signaling in bacterial chemotaxis. Abstracts of the Annual Meeting of the American Society for Microbiology, 1988. p. 192. Fahnestock, M., and Koshland, D.E., Jr. Control of the receptor for galactose taxis in Salmonella typhimurium. Journal of Bacteriology 137:758-763, 1979. Hess, J.F.; Oosawa, K.; Matsumura, P.; and Simon, M.I. Protein phosphorylation is involved in bacterial chemotaxis. Proceedings of the National Academy of Sciences of the United States of America 84(21):7609-7613, 1987. Koshland, D.E., Jr. Biochemistry of sensing and adaptation in a simple bacterial system. Annual Review of Biochemistry 50:765-782, 1981. Koshland, D.E., Jr. The bacterium as a model neuron. Trends in Neurosciences 6:133-137, Apr. 1983. Koshland, D.E., Jr. Chemotaxis as a model second-messenger system. Biochemistry 27(16):5829-5834, 1988. Koshland, D.E., Jr.; Goldbeter, A.; and Stock, J.B. Amplification and adaptation in regulatory and sensory systems. Science 217(4556):220-225, 1982. Levy, W.B., and Desmond, N.L. Synaptic Modification, Neuron Selectivity and Nervous System Organization. Hillsdale, NJ: Erlbaum, 1985. Macnab, R.M. Motility and chemotaxis. In: Ingraham, J.; Low, K.B.; Magasanik, B.; Schaechter, M.; Umbarger, H.E.; and Neidhardt, F.C., eds. Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology. Washington, DC: ASM Publications, 1985. Macnab, R.M., and Koshland, D.E., Jr. The gradient-sensing mechanism in bacterial chemotaxis. Proceedings of the National Academy of Sciences of the United States of America 69:2509-2512, 1972. Rubik, B.A., and Koshland, D.E., Jr. Potentiation, desensitization, and inversion of response in bacterial sensing of chemical stimuli. Proceedings of the National Academy of Sciences of the United States of America 75:2820-2824, 1978. Russo, A.F., and Koshland, D.E., Jr. Separation of signal transduction and adaptation functions of the aspartate receptor in bacterial sensing. Science 220(4601):10161020, 1983. Springer, M.S.; Goy, M.F.; and Adler, J. Sensory transduction in Escherichia coli: Two complementary pathways of information processing that involve methylated proteins. Proceedings of the National Academy of Sciences of the United States of America 74(8):3312-3316, 1977. Stryer, L. Biochemistry. New York: W.H. Freeman, 1988. Teyler, T.J., and DiScenna, P. Long-term potentiation. Annual Review of Neuroscience 10:131-161, 1987. Wylie, D.; Stock, A.; Wong, C.-Y.; and Stock, J. Sensory transduction in bacterial chemotaxis involves phosphotransfer between Che proteins. Biochemical and Biophysical Research Communications 151(2):891-896, 1988. Mechanisms Underlying the Initiation of Long-Term Potentiation of Synaptic Transmission in the Hippocampus R.C. Malenka, J.A. Kauer, and R.A. Nicoll Departments of Pharmacology and Physiology Learning and memory play a fundamental role in the development of behavior and the interactions of an organism with its environment. However, very little is known about the cellular mechanisms that underlie this form of plasticity. It is generally agreed that the change that occurs during the formation of a memory likely occurs at synapses. This proposal received strong support when Bliss and Lomo found that a substantial increase in synaptic strength that can last for many hours and, in intact animals, even for weeks, occurs following brief repetitive activation of excitatory synapses in the hippocampus (Bliss and Lomo 1973; Bliss and Gardner-Medwin 1973). This synaptic enhancement, called long-term potentiation, or LTP, is one of the most promising candidates for a cellular mechanism related to learning and memory in the vertebrate brain. A number of recent discoveries have enabled us to understand in mechanistic detail the steps involved in the initiation of LTP. This paper is divided into two parts. First, we review recent experiments on LTP, which have led to a simple and widely accepted model to explain the induction of LTP. Second, we discuss experiments in our laboratory that indicate that this model is not complete and that additional components are necessary to produce LTP. Two conditions must be satisfied to elicit LTP. First, high-frequency stimulation of the synapses is required. Second, the stimulus strength must be above a certain threshold intensity. Weak stimuli, even when delivered at high frequencies, fail to elicit LTP. Thus, a critical number of presynaptic fibers must be activated together to induce LTP, a property referred to as cooperativity (McNaughton et al. 1978). Cooperativity can also occur between two separate inputs, where it is called associativity (Levy and Steward 1979; Barrionuevo and Brown 1983). While tetanic stimulation of a weak input fails to elicit LTP, when this stimulation is paired with tetanic stimulation of an independent strong input capable of inducing LTP, the weak input will also show LTP. A number of studies have now demonstrated that depolarization of the membrane is responsible for the associative property of LTP. Thus, preventing the depolarization during a strong tetanus either by voltage-clamping the cell (Kelso et al. 1986) or hyperpolarizing the membrane (Malinow and Miller 1986) can block the induction of LTP. Although direct injection of depolarizing current into the postsynaptic cell by itself does not produce LTP, when a weak input is paired with this depolarization, LTP is elicited (Kelso et al. 1986; Sastry et al. 1986; Gustafsson et al. 1987). Interestingly, the weak input need not be fired at high frequency to induce LTP during strong depolarization of the postsynaptic neuron; even a few low-frequency stimuli are sufficient (Gustafsson et al. 1987). Taken together, these data suggest that the initiation of LTP requires activity of the presynaptic fiber concurrent with depolarization of the postsynaptic cell. To understand how the postsynaptic membrane potential influences the initiation of LTP, one must first understand the pharmacology of the excitatory synapses that show LTP. It appears that glutamate, the excitatory transmitter released from these synapses, acts at two subtypes of glutamate receptors: N-methyl-D-aspartate (NMDA) receptors and quisqualate/kainate (Q/K) receptors. The competitive antagonist of NMDA receptors, D-2-amino-5phosphonovalerate (APV), has only a modest effect on the EPSP elicited by low-frequency presynaptic stimulation (Collingridge et al. 1983), indicating that the Q/K receptors predominate for normal synaptic transmission in the hippocampus. However, APV completely blocks LTP following tetanic stimulation, indicating that NMDA receptors can be activated by synaptically released glutamate, and that this activation is required for LTP (Collingridge et al. 1983). Physiological levels of extracellular Mg exert a voltage-dependent block of the NMDA channel (Nowak et al. 1984; Mayer et al. 1984). At the normal resting potential, Mg blocks the channel, but when the membrane is depolarized, the Mg is expelled from the channel allowing current to flow. Unlike the Q/K channels, the NMDA channel has a high permeability to Ca++ as well as to Na+ and K+ (Jahr and Stevens 1987; Mayer et al. 1987; Mayer and Westbrook 1988). Thus, when the cell is depolarized by a strong tetanus, the Mg block of the NMDA channel is relieved, allowing Ca to enter the dendritic spine. The rise in Ca concentration within the dendritic spine may serve as a trigger for LTP. In support of this hypothesis, the induction of LTP can be prevented by loading the postsynaptic cell with the calcium chelator EGTA (Lynch et al. 1983). F+ ++ ++ ++ ++ The model for the induction of LTP that we have just reviewed predicts that only two components are required to induce LTP: activation of NMDA receptors and depolarization of the postsynaptic membrane. Thus, it should |