Population Coding and Reconstruction of Complex Stimuli in the Locust Olfactory System

Citation

Broome, Bede Michael (2006) Population Coding and Reconstruction of Complex Stimuli in the Locust Olfactory System. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/MR29-EA69. https://resolver.caltech.edu/CaltechETD:etd-03142006-171138

Abstract

Odors evoke complex sequences of activity in antennal lobe projection neurons (PNs), the insect analogs of mitral/tufted cells. These PN activity patterns evolve over hundreds of milliseconds, are consistent across trials, and contain information about odor identity and concentration. However, in natural settings animals must often identify odor blends or multiple odorants that occur in short temporal succession. We explored the effects of such stimulus history on single cell PN activity, PN ensemble responses, and downstream Kenyon cell (KC) responses by performing two experiments with novel stimulus paradigms. In the first experiment two different odors with variable intervening delays were presented while recording from PN and KC ensembles in the locust. We found that PN ensemble representations rapidly tracked odor stimulus changes and, in conditions of temporary odor overlap, often corresponded to the representation of neither odor alone nor their binary mixture. KC responses tracked the specificity of the instantaneous state of the PN representation: for example, for stimulus conditions that lead to unique and unpredictable PN population activity, we could find KCs with correspondingly unique spiking. These results support the hypotheses that PN population dynamics are history dependent and that PN output is read broadly in space (over the PN population) and piecewise in time by its target population.

In the second experiment two different odors were presented either as single pulses or complex M-sequence odor stimuli while we recorded PN activity and the time-varying stimulus using a specially adapted electronic sensor. Our results describe the numerous similarities between the processing of stimuli with diverse frequencies and highlight several important differences. We found that PN response types are varied, low dimensional, and that individual PNs do not fall into stable functional classes. Also, PN ensembles occupy different regions of high dimensional coding space and follow trajectories that depend on stimulus history. Finally, an observer’s ability to reconstruct either the concentration or identity of an odor improves with PN number. These results add support to a growing body of evidence that Kenyon cells, the targets of PNs in the mushroom bodies, carry out a pattern classification on PN activity vectors sampled over large sets of PNs.

Thesis Files