Many inhibitory interneurons display tonic activation over the entire duration of odor stimulation in contrast to others that AZD5363 remain hyperpolarized (Figure 7A, bottom). Such patterns of activity are well described by our model (Figure 7A, top). In this example, for the network with nonunique coloring, one group of LN neurons remained active

during the entire duration of stimulation while the two other groups of neurons switched between active and silent states. Why do LNs in the AL exhibit only a subset of the broad repertoire of patterns that the networks simulated here are capable of generating? The formalism we developed in our manuscript points us toward several possibilities. These dynamical patterns are likely to result from an intrinsic asymmetry within the AL subnetwork that gets activated in response to a specific odor. If only a subset of neurons receives strong activation during a particular odor stimulation, these neurons will dominate the response. Asymmetries in coupling strength can also result in the predominance of one group that would prevent switching between groups to occur. In addition, if the number of colors is large, a trajectory may never recur during odor stimulation. Hence the same

LN may not generate multiple bursts of spikes. We have shown that fast GABAergic inhibition mediated by GABAA receptors transiently synchronizes PN activity over a few cycles of the ensemble oscillatory response. A second important PCI-32765 price form of inhibition found in the AL mediated by slow GABAB receptors acts over a timescale in the range of hundreds of milliseconds (Wilson and Laurent, 2005). Experiments (MacLeod and Laurent, 1996) and models (Bazhenov et al., 2001a) have demonstrated that this type of interaction leads to lengthy epochs of time wherein individual

PNs are hyperpolarized and do not spike at all. Picrotoxin applied to the AL spares patterning caused by slow inhibition while abolishing oscillatory synchronization on a fast timescale. The timescales Sodium butyrate separating the two forms of inhibition differ by approximately an order of magnitude. To explore how network structure leads to transient synchrony, a key dynamical variable involved in fine discrimination in the olfactory system (Stopfer et al., 1997), we focus here on fast inhibition while minimizing the effects of slow inhibition in the model. The repertoire of patterns generated by the inhibitory subnetwork in the locust AL forms a subset of the full range of patterns that can be generated by the networks simulated here. Feedforward architecture and coincidence detection mechanisms like those illustrated here are not unique to the insect olfactory system.