Temporal Features of Spike Trains in the Moth Antennal Lobe Revealed by a Comparative Time-Frequency Analysis

The discrimination of complex sensory stimuli in a noisy environment is an immense computational task. Sensory systems often encode stimulus features in a spatiotemporal fashion through the complex firing patterns of individual neurons. To identify these temporal features, we have developed an analysis that allows the comparison of statistically significant features of spike trains localized over multiple scales of time-frequency resolution. Our approach provides an original way to utilize the discrete wavelet transform to process instantaneous rate functions derived from spike trains, and select relevant wavelet coefficients through statistical analysis. Our method uncovered localized features within olfactory projection neuron (PN) responses in the moth antennal lobe coding for the presence of an odor mixture and the concentration of single component odorants, but not for compound identities. We found that odor mixtures evoked earlier responses in biphasic response type PNs compared to single components, which led to differences in the instantaneous firing rate functions with their signal power spread across multiple frequency bands (ranging from 0 to 45.71 Hz) during a time window immediately preceding behavioral response latencies observed in insects. Odor concentrations were coded in excited response type PNs both in low frequency band differences (2.86 to 5.71 Hz) during the stimulus and in the odor trace after stimulus offset in low (0 to 2.86 Hz) and high (22.86 to 45.71 Hz) frequency bands. These high frequency differences in both types of PNs could have particular relevance for recruiting cellular activity in higher brain centers such as mushroom body Kenyon cells. In contrast, neurons in the specialized pheromone-responsive area of the moth antennal lobe exhibited few stimulus-dependent differences in temporal response features. These results provide interesting insights on early insect olfactory processing and introduce a novel comparative approach for spike train analysis applicable to a variety of neuronal data sets