Long-term synaptic potentiation (LTP) is a leading candidate for a synaptic mechanism of rapid learning in mammals. LTP is a persistent increase in synaptic efficacy that can be quickly induced. The biophysical process that controls one type of LTP is formally similar to a synaptic memory mechanism postulated decades ago by the psychologist Donald Hebb. A key aspect of the modification process involves the N-methyl-D-aspartate (NMDA) receptor-ionophore complex. This ionophore allows calcium influx only if the endogenous ligand glutamate binds to the NMDA receptor and if the voltage across the associated channel is also sufficiently depolarized to relieve a magnesium block. According to one popular hypothesis, the resulting increase in the intracellular calcium concentration activates protein kinases that enhance the postsynaptic conductance. Further biophysical and molecular understanding of the modification process should facilitate detailed explorations of the mnemonic functions of LTP.

We have examined the evolution of the concept of a Hebbian synaptic modification and have suggested a contemporary definition. The biophysical mechanism demonstrated in vitro to control the induction of one type of hippocampal LTP has been shown to satisfy our definition of a Hebbian synaptic modification. Whether this biophysical mechanism is involved in the organization of behavior in the manner that Hebb originally envisioned remains to be seen. We have also summarized several modification algorithms that have been explored in theoretical studies of learning in adaptive networks. These algorithms also satisfied our definition of a Hebbian modification, but their relationships to known neurobiology require further exploration. By reviewing the biophysical mechanisms and formal algorithms together, we have exposed obvious similarities and differences. Such comparisons may help bridge the gap between computational theory and knowledge of the neurobiology of use-dependent synaptic change. Current models of LTP reveal that the activity-modification relationships are extremely sensitive to the biophysical/molecular details. The activity-modification relationships obviously can have a major influence on adaptive neurodynamics at the network level. As more accurate representations of the biological complexity and diversity are introduced into adaptive network simulations, we expect to gain new insights into the classes of computation that particular networks are capable of performing.

The manner in which the circuitry of the amygdala computes its suspected mnemonic functions has been a mystery, partly because the cytoarchitectual complexity of this nuclear group has impeded the necessary cellular analysis. Here we report in vitro methods and results that may help elucidate cellular learning mechanisms in amygdala neurons. The amygdala brain slice preparation was combined with the single-electrode clamp (SEC) technique for intracellular analysis of membrane properties and synaptic responses. With respect to the active and passive membrane properties, we found considerable diversity among the population of cells that were sampled in the lateral and basolateral nuclei (n = 85). Synaptic inputs to these neurons were studied by stimulating the external capsule (EC), which was shown to produce a complex response that typically consisted of an excitatory followed by an inhibitory component. Based on several criteria, the excitatory component appeared to reflect a monosynaptic connection from the EC. One immediate goal was to discover whether the excitatory component displays the phenomenon of long-term potentiation (LTP)--a persistent increase in synaptic strength that can be induced by brief periods of the appropriate synaptic stimulation. Indeed, we found that high-frequency (100 Hz) stimulation of the EC induced LTP in 80% of the cells from which suitable recordings were obtained (n = 20). This finding of LTP in the amygdala is significant in regard to current efforts to explore linkages between this use-dependent form of synaptic plasticity and rapid kinds of associative learning.

1. The quantal mechanism underlying the expression of long-term potentiation (LTP) was studied in the mossy-fiber (mf) synapses of the rat hippocampus. Whole-cell recordings were used to measure the excitatory postsynaptic currents (EPSCs) before and after LTP induction in brain slices maintained at 31 +/- 1 degrees C. 2. Evoked EPSCs were recorded from 473 CA3 pyramidal neurons. The mf synapses were stimulated using paired pulses (40-ms interpulse interval) repeated every 2-10 s. At least 400 pairs of mf responses were obtained before and during the expression of LTP, which was produced by high-frequency (100 Hz) mf stimulation. Sufficiently stationary data were obtained from five neurons that exhibited LTP and that also satisfied strict criteria and procedures that are necessary for eliciting and identifying unitary mf responses. 3. Three independent lines of evidence implicated a presynaptic component to the mechanism underlying mf LTP. The first was based on a graphical version of the classical method of variance. The graphical variance (GV) method was evaluated by clamping the cell at two different holding potentials during paired-pulse facilitation (PPF). The results indicated that the GV method can distinguish changes in mean quantal content m and mean quantal size q in rat mf synapses. The same analysis, when applied to PPF before and after LTP induction, indicated that both result from an increase in m. 4. The second line of evidence was based on the classical method of failures. Consistent with the inference that mf LTP is due to an increase in m, there was a statistically significant reduction in the number of quantal release failures.(ABSTRACT TRUNCATED AT 250 WORDS)

No abstract