Art - Faber (R)

Learning and memory are crucial for establishing and storing predictive relationships between stimuli, which can serve as the basis for adaptive behavior. As a prerequisite, learning should lead to structural and functional changes in the brain, which may occur at primary sensory centers as well as at central and efferent brain structures1–6. Recordings of single central neurons in vertebrates and invertebrates show that associative learning can change cellular response properties7–10. However, the consequences of learning for neural networks in vivo are less well understood. A possible approach to the study of learning-based modifications in neural networks involves imaging techniques1,3, which can record many neurons in parallel and may thus be used to investigate whole neural networks. Here we used optical imaging to distinguish neural representations for different stimuli during associative conditioning, to study whether associative learning induces persistent changes in sensory networks and, if so, how neural stimulus representations are altered in a sensory center. To investigate this question, we developed a preparation that allows us to monitor distributed brain activity while training the animal to distinguish odors. We used the calcium-sensitive dye Calcium Green-2 AM to record topographical activation patterns in the antennal lobe (AL) of honeybees11, which is the functional and structural analogue of the vertebrate olfactory bulb and has a similar glomerular organization12. Different odors are coded as different glomerular activity patterns in the AL, and one glomerulus can be involved in the representations of different odors11,13. We recorded glomerular activation patterns corresponding to three odors, before and after training the animals to such odors. In the differential conditioning procedure, which is well characterized14,15, the bees learned to respond by extending their proboscis to an odor paired with sucrose reward (rewarded odor), but not to a different, unpaired odor (unrewarded odor). A third odor (control odor), not presented during training, was tested before and after conditioning as a generalization control. Results Before training, each odor activated specific patterns of glomeruli (Fig. 1a). After differential conditioning, activity for the rewarded odor increased (Fig. 1b). Activity for the control odor was also enhanced, but to a lesser extent than for the rewarded odor (Fig. 1b). The activity for the unrewarded odor decreased in this case. Activated regions corresponded well with the underlying glomerular structure of the AL (Fig. 1c and d), determined by confocal reconstruction16. This method, however, does not allow identification of individual glomeruli. The learning-induced activity changes were not due to a simple offset added to baseline activity, possibly introduced by the sensitizing effect of the reward17. A comparison of the spatially averaged actitivity before and after training showed no significant change for the rewarded, unrewarded or control odor. However, evaluation of active glomeruli as functional units of the AL showed spatially heterogenous changes. To clearly extract active glomeruli, we defined a threshold at the top 25% of the signal range before training and considered only those signals lying above the threshold (Fig. 2a and b). Before training, the rewarded odor induced activity in two glomeruli (Fig. 2a). After training (Fig. 2b), activity of both glomeruli increased (maximum signal before training ∆F/Fmax = 1.54%, after training ∆F/Fmax = 2.23%). To quantify changes due to associative learning, we calculated the integral of the volume of signals above threshold (Fig. 2c). In this analysis, activity for the rewarded odor increased, whereas that for the unrewarded did not change. Activity for the control odor was also enhanced after training, but remained lower than that for the rewarded odor. Parallel behavioral experiments demonstrated that bees do learn to respond differentially to the rewarded and unrewarded odor under the conditions of our imaging experiments. Bees prepared and treated exactly as in the imaging experiments (handling, dissection, dye infiltration in chilled animals, wash, temporal sequence of conditioning and testing) were trained with Associative learning modifies neural representations of odors in the insect brain