1. We have shown previously, with experimental and computer models, how a '40 Hz' (gamma) oscillation can arise in networks of hippocampal interneurones, involving mutual GABAA-mediated synaptic inhibition and a source of tonic excitatory input. Here, we explore implications of this model for some hippocampal network phenomena in the rat in vitro and in vivo. 2. A model network was constructed of 1024 CA3 pyramidal cells and 256 interneurones.AMPA (a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid), NMDA (N-methyl-D-aspartate), GABAA and GABAB receptors were simulated on pyramidal cells and on interneurones. 3. In both model and experiment, the frequency of network oscillations, in the gamma range, depended upon three parameters: GABAA conductance and decay time constant in interneurone-*interneurone connections, and the driving current to the interneurones. 4. The model of gamma rhythm predicts an average zero phase lag between firing of pyramidal cells and interneurones, as observed in the rat hippocampus in vivo. The model also reproduces a gamma rhythm whose frequency changes with time, at theta frequency (about 5 Hz). This occurs when there is 5 Hz modulation of a tonic signal to chandelier and basket cells. 5. Synchronized bursts can be produced in the model by several means, including partial blockade of GABAA receptors or of AMPA receptors on interneurones, or by augmenting AMPA-mediated EPSCs. In all of these cases, the burst can be followed by a 'tail' of transiently occurring gamma waves, a phenomenon observed in the hippocampus in vivo following sharp waves. This tail occurs in the model because of delayed excitation of the interneurones by the synchronized burst. A tail of gamma activity was found after synchronized epileptiform bursts both in the hippocampal slice (CA3 region) and in vivo. 6. Our data suggest that gamma-frequency EEG activity arises in the hippocampus when pools of interneurones receive a tonic or slowly varying excitation. The frequency of the oscillation depends upon the strength of this excitation and on the parameters regulating the inhibitory coupling between the interneurones. The interneurone network output is then imposed upon pyramidal neurones in the form of rhythmic synchronized IPSPs.EEG rhythms at frequencies of 30-100 Hz are prominent possible significance for information processing (for review,

We used transverse and longitudinal rat hippocampal slices to study the synchronization of γ frequency (> 20 Hz) oscillations, across distances of up to 4.5 mm. γ oscillations were evoked in the CA1 region by tetanic stimulation at one or two sites simultaneously, and were associated with population spikes. Tetanic stimuli that were strong enough to induce oscillations were associated with depolarization of both pyramidal cells and interneurones, largely produced by activation of metabotropic glutamate receptors. Computer simulations of γ oscillations were also performed in a model with pyramidal cells and interneurones, arranged in a chain of five cell groups. This model had suggested previously that interneurone networks alone could generate synchronous γ oscillations locally, but that pyramidal cell firing, by inducing spike doublets in interneurones, was necessary for the occurrence of highly correlated oscillations with small phase lag (< 2.5 ms), in a distributed network possessing long axon conduction delays. In both experiment and model, pyramidal cell spikes occurred in phase with local population spikes, as did the first spike of the interneurone doublet. The conductance of the interneurone α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazole propionic acid (AMPA) receptor‐mediated conductance was manipulated in the model, while the relation between oscillations at opposite ends of the chain was examined. When the conductance was large enough for doublet firing to be synaptically induced in interneurones, oscillation phase lags were < 2.25 ms across the chain. As predicted, experimental blockade of AMPA receptors resulted in increased phase lags between two sites oscillating simultaneously, compared with control conditions. Both in model and in experiment, when stimuli to the two ends of the network were slightly different, cross‐network synchronization occurred with a shorter phase lag at high frequencies than at lower frequencies. These data suggest that, while interneurone networks alone can generate locally synchronized γ oscillations, firing of pyramidal cells, and the synaptically induced doublet firing in interneurones, contribute to the stability and tight synchrony of the oscillations in distributed networks.

No abstract