PROJECT SUMMARY A critical feature of memory is mnemonic discrimination, the ability to distinguish between similar events in our past and prevent confusion between similar situations. The dentate gyrus (DG) of the hippocampus is a brain region known to play a central part in this cognitive function. For thirty years, the DG neuronal network has been hypothesized to support mnemonic discrimination by performing a computation called "pattern separation" during memory encoding. In this view, the role of the DG is to disambiguate upstream cortical representations of similar events, transforming similar patterns of incoming activity into dissimilar output patterns, before they reach downstream hippocampal areas to be stored as memories. Computational modelling suggests that the anatomy and physiology of the DG network could allow DG to perform pattern separation, and recent experimental studies demonstrated that the isolated DG is able to perform some forms of pattern separation. However, a direct demonstration that DG performs such computation in vivo remains elusive. Consequently, the central tenet that pattern separation in DG during memory encoding is the mechanism underlying mnemonic discrimination has never been directly tested. Moreover, the role of DG during memory retrieval or consolidation is also unclear. The general goals of this research project are thus to 1) test in vivo whether DG performs pattern separation or other computations and 2) determine how DG computations during encoding and retrieval support mnemonic discrimination. Measuring the computations of a network like DG requires knowledge about its simultaneous inputs and outputs, which has never been achieved in vivo at a single-cell resolution. To resolve this difficulty, I will use multiplane two-photon calcium imaging in behaving mice in order to simultaneously record activity dynamics of a large population of DG output neurons and the activity of the cortical axons that target their dendrites. To investigate the basis of mnemonic discrimination, one needs to compare neuronal representations of different but similar experiences. To this end, mice trained to navigate in a virtual environment will be placed in several novel environments of parametrically varied similarity. Experiment-1 will consist of recording DG inputs and outputs while mice explore a sequence of environments, repeated over several days to determine how computations evolve as familiarity increases. In experiment-2, an explicit mnemonic component will be added: a new environment will be associated to a fearful stimulus and the ability of individual mice to discriminate this context from another similar but neutral one will be measured. This will allow to determine what computations are performed by DG during encoding and recall and how they relate to context discrimination: if the theory is correct, the degree of pattern separation during encoding should correlate to the amount of discrimination. Finally, the ...