Synapse Specificity and Long-Term Information Storage

would enable them to coopt proteins destined for other The dendrites of a typical neuron in the central nervous late LTP synaptic sites. To test this idea, a “two-pathsystem bear at least 1000 postsynaptic specializations way” experimental design was employed in which two indicating that in principal the individual neuron can independent sets of presynaptic afferents that converge process and store many bits of information. Electroon a common set of postsynaptic neurons were alterphysiological experiments have also revealed that ananately stimulated. (This paradigm allows one to examine tomically isolated groups of synapses on the same cell two separate populations of synapseson the same postcan separately process plastic events. Finally, many synaptic neurons.) Late LTP was induced in one pathstudies of behavioral and synaptic plasticity have demway, and then a protein synthesis inhibitor was introonstrated that long-lasting changes in synaptic transduced into the bath, and the second pathway was mission and behavior require both gene transcription tetanized. Under normal circumstances, the second and mRNA translation. To understand long-term inforpathway, tetanized in the presence of the protein synmation storage, then, the problem is to determine how thesis inhibitor, would only exhibit early LTP. Frey and the potentiated (or depressed) synapses of a single neuMorris found, however, that prior induction of late LTP ron become selectively modified during the long-lasting in the first pathway converted the usually decremental phases of synaptic plasticity. That is, how do the prodLTP in the second pathway to late LTP. Similar results ucts of transcription and/or translation reach the modiwere obtained when the second pathway was stimufied synaptic siteswithout affecting the unmodified sites lated (in the absence of a protein synthesis inhibitor) within the same neuron? There are at least three ways with a very weak tetanus, which normally gives rise to this could occur, including the selective capture of shorter-lived potentiation. broadly distributed newly synthesized proteins by poOne interpretation of these findings is that late LTP tentiated sites, the selective transport of newly syntheinduced in the first pathway generates new proteins that sized proteins to potentiated sites, and the site-specific can then be used by the synapses stimulated in the local production of new proteins by dendritic protein second pathway to generate late LTP. Because the secsynthesis machinery (Figure 1; reviewed by Sossin, ond pathway was only transiently treated with a protein 1996; Goelet et al., 1986). synthesis inhibitor, it is also possible that new mRNAs, In hippocampal slices, long-term potentiation of synrather than proteins, are transported out to dendrites aptic transmission can last between 1 hr (early LTP) during late LTP (e.g., Mayford et al., 1996). Alternatively, and up to 12 hr (late LTP), depending on the induction it might be that late LTP induction somehow induces a stimulation protocol. Other than the duration of the syncell-wide priming of synapses, which facilitates or lowaptic enhancement, a distinguishing feature of late LTP ers the threshold for the subsequent induction of late is that it requires both gene transcription and mRNA LTP at other sites on the same cell. translation (e.g., Nguyen et al., 1994). Hippocampal A curiosity of this finding is that some synaptic memoslices tetanized in the presence of either a transcription ries, which under ordinary circumstances would decay or a translation inhibitor exhibit potentiation that decays over time, now become long-lasting. As the authors back to baseline within 2–3 hr. These observations sugpoint out, this is a potential mechanism for establishing gest one of two possibilities: either there is constitutive associations between a significant event (e.g., events production and transport of inherently unstable mRNAs encoded by L-LTP at the first pathway) and other stimuli that must be translated in a plasticity-dependent manassociated with it in time (e.g., events encoded by stimuner, or inductive events at the synapse must somehow lation of the second pathway). It could also lead, howresult in information flow to the soma (e.g., Bito et al., ever, to the storage of noise or unnecessary information 1996). In the latter case, the protein products resulting since subthreshold synaptic stimulation capable of from this communication must then be sent back to the generating a marker could enable synapses to exhibit appropriate synaptic sites so that long-term synaptic enduring LTP. As such, there may well be molecular enhancement is maintained. In the absence of a mechasafeguards that decrease the likelihood of erroneous nism for selectivity, the induction of late LTP at one set long-term information storage.