From adult dentate gyrus neurogenesis to pattern separation

For a few decades, adult dentate gyrus neurogenesis has been widely recognized by the neuroscience community as an intriguing phenomenon. Two observations are particularly puzzling. At the cellular level, the switch from excitation to inhibition of the GABAergic input onto newborn cells has been shown to be crucial for their proper integration into the existing network of dentate gyrus cells. At the behavioral level, adult-born dentate granule cells have been shown to promote pattern separation of similar stimuli in various tasks, while not playing a role in discrimination of distinct stimuli. It is still unclear, however, how these functionalities arise in the network of dentate gyrus cells. Several models of adult dentate gyrus neurogenesis have been designed with various levels of abstraction, and have suggested different roles of newborn cells. Yet, none of these models could explain how newborn cells promote pattern separation of similar stimuli, and not distinct stimuli. Moreover, none of the previous studies modeled the actual integration of adult-born dentate granule cells in the preexisting circuit, but rather initialized their inward connections to random, but fully grown, weights.

In my thesis work, I bridge the gap between biological and theoretical knowledge on adult dentate gyrus neurogenesis. I address the puzzling experimental observations and explain for the first time with a model: (i) how newborn cells integrate into the preexisting dentate gyrus network, and (ii) how they promote pattern separation of similar stimuli.

More specifically, I propose that the early phase of maturation of newborn cells, when GABAergic input has an excitatory effect, drives the synaptic weights towards the subspace of configurations of familiar stimuli through a cooperative effect. In the late phase of maturation, when GABAergic input switches to inhibitory, the synaptic weights move towards novel features of the presented stimuli through a competitive effect. This theory of newborn cells integration also explains why adult-born dentate granule cells promote better pattern separation of similar stimuli, but not distinct stimuli. Indeed, in the late phase of maturation, newborn cells can only learn novel features that are similar enough to familiar features, because the configuration of their synaptic weights makes them sensitive to familiar features at the end of the early phase of maturation.