WP#1: Past results suggested that the frontal association cortex (FrA) of rodents could participate in the computation of action-selection when animal face a decision task. To test this hypothesis, a communal effort in the Gambino lab (Univ. Bordeaux, CNRS) has been to characterize the activity of FrA neurons in mice during decision-making. We have implemented in vivo microscopy of Ca2+ dynamics to simultaneously monitor the activity of dozens of neurons in FrA, while imaged animals took decisions, over weeks of behavior. We have also designed new computational models of cognition for the decision-making task we used and exploited them to correlate changes in behavior with that of neural activity. As a result, our analysis of the activity of hundreds of neurons in FrA revealed that subsets of cells code for dierent stages of the behavior and that these codes can change with learning the values of different choices. Results from us, and others, therefore point to some decision-making computation occurring in FrA. We have then focused our attention on the basolateral amygdala (BLA), a region widely implicated in learning associations (eg between sensory stimuli and fearful/rewarding events), which we hypothesized to transmit rich sensory information and task parameters to the FrA. We recorded the activity of axons coming from BLA neurons at their output sites in FrA by using in vivo microscopy. Our experiments revealed that BLA activity is modulated by most events relevant to decision-making. Interestingly, the amplitude of the BLA activity changed with the expertise of the animal and its knowledge of the value of the different possible decisions. Overall, it confirmed that BLA encodes varied information about decision-making tasks and that it relays it to the FC, where optimal decision making could take place.
WP#2: Two-photon laser scanning microscopy (2PLSM) in vivo is emerging as a prime method for the detailed investigation of cognitive processes since it can record the activity of micrometer-scale objects, genetically target specific cell groups, and monitor the same features over extended periods of time. However, the few methods which could automatically analyze 2PLSM Ca2+ images when the project started were not fit to the low level of light and the complex shapes that are typical of studies of neural sub-compartments (such as axons in WP#1). To solve this problem, we focused our efforts in on two main aspects: 1) extracting faint signals from noisy 2PLSM images, 2) avoiding morphological constraints in order to handle complex shapes, in particular those of axonal and dendritic trees. Thanks to the design of new mathematical methods for those two steps, we were able to build a combined, fully-automated, analysis pipeline which proved accurate and versatile. It allowed to process varied images, from oval-shaped neuronal cell bodies to complex, disjointed axonal and dendritic trees.
For dissemination purpose, preliminary results were communicated to the Neuroscience community at the NeuroFrance conference in 2019. A poster presentation of latest results was scheduled at the Federation of European Neuroscience Societies (FENS) meeting 2020, but was cancelled due to the COVID-19 outbreak. Two manuscripts summarizing our results for WP#1 and WP#2, respectively, are being prepared for publication in peer-reviewed journals. The software programs resulting from WP#2 will be openly accessible to all via a web platform (eg github.com) at the time of publication.
