A key task for animals navigating the world is to distinguish between similar yet different objects, places, or contexts. The neurocomputational function thought to underlie this task is referred to as “pattern separation.” Theoretical work suggests that sparse firing and sparse coding by granule cells in the dentate gyrus (DG) could be the basis for pattern separation. Clinical imaging and lesion experiments also support the idea that the DG is important for pattern separation. However, physiological support for such function has been hampered by an inability to reliably distinguish the two major types of excitatory cells in the DG: granule cells, which send strong outputs to CA3 and local neurons; and mossy cells, which are involved in the recurrent excitation of DG neurons. 

In our recent work (Senzai and Buzsáki, 2017), we identified electrophysiological criteria that can unequivocally distinguish granule cells from mossy cells and validated the classification by optogenetic tagging of mossy cells. This classification scheme allowed us to subsequently characterize the network properties and behavioral correlates of these two key neural cell types.

Specifically, excitatory neurons (i.e., granule cells and mossy cells) can first be distinguished from dentate gyrus interneurons based on their wide action potential waveform and increased tendency to fire in bursts. Three main characteristics then separated the excitatory neurons into granule cells and mossy cells: one, cell body location relative to a characteristic local field potential pattern called “dentate spike type 2”; two, relative firing rates in non-rapid eye movement sleep and waking states; and three, waveform shape. We validated these physiological characteristics with optogenetic tagging of mossy cells, establishing firm criteria for granule cell and mossy cell identification in vivo.

Our classification method allowed us to investigate how these cell types interact and encode spatial information. Granule cells typically had low firing rates and had a single place field, whereas mossy cells had higher firing rates and had two or more place fields. These results demonstrate that granule cells have properties theoretically suggested to be the basis for pattern separation: namely, sparse spatial coding. In order to assess each cell type’s involvement in pattern separation, we quantified the remapping of place fields across different testing apparatus in the same room. To our surprise, granule cells showed weaker remapping compared to their downstream targets mossy cells and CA3 pyramidal cells.

This raised another question: how is remapped spatial information computed by mossy cells and CA3 pyramidal cells, which receive super-strong inputs from less-remapped granule cells? To answer the question, we detected granule cell–mossy cell putative synapses in vivo. Despite the strongly facilitating nature of granule cell–mossy cell synapses, place fields of most mossy cells could not be explained by the “inheritance” of place fields from single granule cells.

Our results suggest that granule cells are not the only cell type responsible for pattern separation. Instead, the joint action of granule cells, mossy cells, and CA3 pyramidal cells may be critical for pattern separation.

—Yuta Senzai and György Buzsáki, MD, PhD

Read the paper “Physiological properties and behavioral correlates of hippocampal granule cells and mossy cells” in Neuron, published February 8, 2017.