Sparse versus dense coding of very large environments in hippocampal subregions CA3 and CA1
Shir R. Maimon, Tamir Eliav, Johnatan Aljadeff, Liora Las, Nachum Ulanovsky
The hippocampus plays a critical role in memory formation and contains spatially selective neurons known as place-cells. It is comprised of distinct subregions, including areas CA1 and CA3, which differ markedly in their anatomical connectivity: CA1 is largely a feedforward network, whereas CA3 is a recurrent network of densely interconnected pyramidal neurons. Despite these structural differences, prior studies conducted in animals navigating in small arenas, have reported similar spatial coding properties in CA1 and CA3 neurons, characterized by single place fields of similar sizes. In our previous work we showed that CA1 place cells recorded in bats flying in a very long 200-meter tunnel, exhibit multiple place fields with different field-sizes for different fields – a multifield multiscale spatial code which is fundamentally different from the single fields observed in confined laboratory environments. Here we show a dramatic difference in how CA1 and CA3 place-cells encode large-scale environments: Unlike CA1 place-cells, CA3 place-cells mostly had only single place-fields in the long tunnel; however, the sizes of individual place-fields were similar between the two subregions. Using a neural network model, we propose that the sparse single-field code for space in CA3 may facilitate fast learning of new spatial maps. These results suggest a fundamental functional difference in neural-coding between these two anatomical subregions of the hippocampus: Sparse coding in CA3 versus dense coding in CA1.