The hippocampus is important for memory of place and events. It is also one of the few areas in the adult mammalian brain that exhibits neurogenesis, the continuous generation of new neurons. Much evidence indicates that adult neurogenesis contributes to hippocampal-dependent cognition, but the nature of this contribution remains elusive. Moreover it has remained unclear to what extent hippocampal neural code that underlie spatial memory are stable over long periods of time and to what extent ongoing neurogenesis affect the stability of such codes. The study of such coding dynamics requires longitudinal recordings of neuronal ensembles over periods of weeks, since this is the timescale on which new hippocampal neurons mature. In this project we used and further developed optical recording of Ca2+ dynamics from hundreds the same genetically defined neurons in the hippocampus of freely behaving mice over periods of multiple weeks. We established a probabilistic method (CellReg) for the automated identification of the same neurons (cell registration) across multiple imaging sessions (Scheituch et al, Cell Reports 2017). We have found that hippocampal ensemble dynamics time-stamp experienced events via unique patterns of neuronal activity for specific days, suggesting a role for representational drift in forming a mental timeline in which experiences could be mnemonically associated or dissociated based on their temporal proximity (Rubin et al, eLife 2015). By tracking the activity of hippocampal ensembles across a time span of minutes to weeks, we discovered that multiple distinct representations of the same spatial context can stably coexist in the mouse hippocampus, suggesting that a memory of a given context could be associated with multiple representations, rather than just one (Sheintuch et al, Current Biology 2020). We have further revealed that different aspects of neural code stability are differentially governed by time and experience, and that these effects differ across brain areas that support spatial memory. With respect to the effects of ongoing neurogenesis in the dentate gyrus on downstream CA1 codes, we have found, unexpectedly, that promoting neurogenesis in the dentate gyrus (i.e. increasing circuit rewiring) does not increase representational drift in CA1, but rather contributes to a more informative and stable place code. By conducting large-scale recordings of neuronal activity, we developed a novel approach to studying and interpreting neuronal activity data (Rubin et al, Nature Communications 2019). Using this approach, we showed that the internal structure of neuronal activity itself can be used to reconstruct neuronal representations and discover new encoded variables hidden within a neural code. This approach offers a new means to quantifying the quality and stability of the neural code, with minimal a-priori assumptions, and irrespective of the identities of the neurons that participate in the population activity or of the variables they encode. Overall, our work promotes our understanding of how the hippocampus supports long-term memory, by revealing basic mechanisms and dynamics of it neural code.
