Neurogenesis and pattern separation: time for a divorce

Abstract

The generation of new neurons in the adult mammalian brain has led to numerous theories as to their functional significance. One of the most widely held views is that adult neurogenesis promotes pattern separation, a process by which overlapping patterns of neural activation are mapped to less overlapping representations. While a large body of evidence supports a role for neurogenesis in high interference memory tasks, it does not support the proposed function of neurogenesis in mediating pattern separation. Instead, the adult-generated neurons seem to generate highly overlapping and yet distinct distributed representations for similar events. One way in which these immature, highly plastic, hyperactive neurons may contribute to novel memory formation while avoiding interference is by virtue of their extremely sparse connectivity with incoming perforant path fibers. Another intriguing proposal, awaiting empirical confirmation, is that the young neurons’ recruitment into memory formation is gated by a novelty/mismatch mechanism mediated by CA3 or hilar back-projections. Ongoing research into the intriguing link between neurogenesis, stress-related mood disorders, and age-related neurodegeneration may lead to promising neurogenesis-based treatments for this wide range of clinical disorders. WIREs Cogn Sci 2017, 8:e1427. doi: 10.1002/wcs.1427

This article is categorized under:

• Psychology > Memory
• Neuroscience > Cognition
• Neuroscience > Computation

Graphical Abstract

The generation of new neurons in the adult mammalian brain has led to numerous theories as to their functional significance. One of the most widely held views is that adult neurogenesis promotes pattern separation, a process by which overlapping patterns of neural activation are mapped to less overlapping, sparse representations. The schematic model of the hippocampus shown in (a) uses sparse coding to distinguish between similar patterns, whereas the model shown in (b) and (c) uses neurogenesis and sparse connectivity to overcome interference. The model input is a distributed pattern of activation across the entorhinal cortex pyramidal cells (PCs; activated cells shown in pink), generating a sparse pattern of activation across the mature dentate granule cells (GCs; active cells shown in blue). In (b) and (c), the model with neurogenesis also includes immature GCs (active cells shown in green), and this model is presented with two different input patterns and, in response, generates identical sparse activation patterns in mature GCs and overlapping but distinct distributed activation in immature GCs. While the two input patterns overlap by 40%, the two patterns of activation in the immature GC overlap by 50%, hence a decrease in pattern separation. In spite of this, the model with neurogenesis is able to generate distinct neural codes for the two similar inputs, in spite of a high plasticity in the immature cell population, due to the sparse connectivity of the immature GCs.