Significance of glomerular compartmentalization for olfactory coding.

This paper deals with the dendritic signal processing by mitral cells in the olfactory bulb and its meaning for olfactory coding. The output signals of olfactory receptor neurones are sent to the olfactory bulb where they converge onto the secondary neurones, the mitral cells. On a short time scale, the connectivity between receptor and mitral cells can be assumed to be constant, whereas on a longer time scale, when considering the ongoing de- and regeneration, it is necessary to model the synaptical weights between receptor and mitral cells as variables. In a first approach we used Hebb's rule to this end and presumed that a mitral cell can be represented by one compartment only. In this case, and with a sequence of realistically modeled receptor activity signals, the synaptical weights of all mitral cells converged to the same point though every mitral cell had initial weights different from those of any other mitral cell. This means that a mitral cell, when modeled as one compartment, does not become sensitive to any particular odor quality. A similar lack of quality tuning turned out to occur when one-compartment mitral cells were connected among each other by laterally inhibiting interneurones. We therefore took into account the glomerular fine structure of mitral cell dendrites, assuming electrotonically decoupled dendritic subbranches. This feature together with local inhibitory circuitry at the subbranches led to a fundamentally different type of synaptical convergence pattern. In this case, mitral cells developed differential sensitivities for different odors. Mitral cells have thus to be regarded as multicompartment cells, and local, non-Hebbian learning rules for their afferent synapses are necessary to achieve a reasonable map of odors upon mitral cell activities.