Recurrent synaptic inhibition in an oscillatory neuronal network

The PD neurons of the lobster stomatogastric ganglion are bursters which receive synaptic input from only the inhibitory LP neuron. This inhibition is recurrent, in that it is triggered by, and follows at a fixed latency after, every PD burst. The timing, duration, and strength of this input was controlled by manipulating LP; the PDs were also perturbed by injecting pulses and pairs of pulses of current intracellularly after removing input from LP. The results were summarized in a simple, linear model of the information processing properties of the PDs: the PDs compute a weighted average from endogenously and synaptically-supplied temporal information. This computation determines the timing of bursting discharge, but does not affect the duration or intensity of discharge if the input occurs in the interburst interval.; The recurrent input from LP normally makes the burst rhythm produced by the PDs more robust, in that the rhythm is less sensitive to perturbing input and undergoes less spontaneous variation in burst period. This effect of synaptic inhibition is a consequence of hyperpolarization, in that one hyperpolarizing input makes the PDs less sensitive to another perturbing input. Normal synaptic input does not affect the period of the rhythm because of its timing.; A new technique is described which makes it possible to functionally potentiate LP's synapse onto the PDs. In this technique, action potentials produced by LP trigger pulses of hyperpolarizing current injected intracellularly into the PDs. Potentiating LP's recurrent synapse onto the PDs in this way induces a qualitative change in the burst pattern produced by the PDs and LP: instead of producing regular, alternating bursting discharge, they produce patterns which are either irregular or which show complex periodicities. This occurs because strong hyperpolarization of the PDs in any one cycle increases the PDs' release of inhibitory transmitter onto LP in the following cycle.; These results show that a rhythm can be made more robust by incorporating recurrent inhibition, but that that input must not be so strong that it alters the dynamics of the network.