Sensitivity vs. Gain control

MARACHLIAN EMILIANO; FERNANDO LOCATELLI

Santa Barbara

Workshop; Deconstructing the Sense of Smell, 2015; 2015

Kavli Institute for Theorethical Physics

Sensitivity vs. Gain control Marachlian1, Huerta2,Locatelli11IFIByNE-CONICET- Universidad de Buenos Aires,Argentina2Biocircuits Institute, UCSD, USA. Odors in nature are very complex stimuli. In order to survive in natural environments,animals must be able to identify specific components embedded in complexmixtures and recognize them across a big concentrations range. Both challengesrequire solutions that are apparently opposite. For example, higher sensitivityto a relevant component could be achieved by boosting the gain of the olfactorysystem for that specific odor while on the other hand the ability to recognizethat odor across concentrations might require a strict gain control thatstabilizes odor specific representation irrespective of its concentration. Inthe poster we present experimental and modeling results that support thepossibility that the solution for both problems might be subject of experiencedependent plasticity. First, we performed calcium imaging experiments in thehoney bee antennal lobe to study the representation of different odorconcentrations under normal conditions and under the effect of picrotoxin whichblocks GABA-A inhibition. We found that without inhibition the magnitude ofcalcium signals in the antenal lobe increases together with the concentrationof odor stimuli, thus evidencing the role of inhibition in gain controlmechanisms. Second, we performed calcium imaging experiments combined withassociative learning to measure the representation of mixtures that contains arewarded component. We found that associative learning with a specificcomponent increases the relative weight of that component in the representationof the mixture. In comparison to the representation of the mixtures inuntrained animals, the activity pattern elicited by the mixture in trainedanimals is more similar to the rewarded component and less similar to thenon-rewarded component. Interestingly, the changes induced by training involvea global reduction of activity in the antennal lobe, but specially in theelements that encode the non-rewarded odor. No increased activity was observedin elements that encode the rewarded odor. These results indicate that changesinduced by training are caused by the strengthening of the inhibitory interactionsfrom the learned component towards the non-rewarded component. All together, itsuggests that modulation of inhibition may contribute to increase the salienceof relevant components and stabilize its representation among a certain levelsof activity. In addition, we performed computational simulations of theantennal lobe network that indicate that both results can be reproduced bymodifying the strength of inhibitory synapses. The results suggest thatregulation of inhibitory synaptic strength based on experience dependentplasticity might tune sensitivity to relevant odors and generalization acrossconcentrations.