Credit: Magictorch / Alamy Stock Photo

Patterns of connectivity differ among hippocampal CA3 subregions, indicating that there may be corresponding differences in function. Two new studies demonstrate functional heterogeneity along the transverse axis of the CA3 and suggest that proximal CA3 neurons (those close to the dentate gyrus) may contribute to pattern separation, whereas distal CA3 neurons may have a role in pattern completion.

During memory encoding, neural networks store 'representations' — associations between particular patterns of inputs and specific experiences — in a process known as pattern separation. For example, populations of hippocampal place cells use environmental and self-motion cues to form representations of an animal's location in its environment. When presented with an (often incomplete or slightly altered) input pattern, memory retrieval requires the neural network to perform pattern completion to retrieve these stored representations. To examine the contribution of CA3 neurons in different subregions to pattern separation and pattern completion, the authors of both studies recorded from large numbers of place cells across the CA3 and the adjacent CA2 in rats.

The ability of a neural network to perform pattern separation or pattern completion may be reflected by the extent to which its firing patterns 'remap' when its inputs are altered. In the experiments reported by Lee et al., the rats ran around a circular track. They were exposed to local cues (track textures) and global cues (objects placed on the walls). The authors evoked place cell remapping by rotating the local and global cues in opposite directions between trials. Some cells exhibited remapping, firing in only the standard or rotated configuration, whereas others fired in both configurations (with their place fields mostly maintaining their location with respect to the local cues). The proportion of cells that exhibited remapping was higher in the proximal CA3 than in the distal CA3 or CA2. Furthermore, neuronal population vector correlation analysis showed that a coherent representation of the animals' position in the environment was maintained in the rotated configuration by cell populations in the distal CA3 and CA2, but was lost in the proximal CA3.

Both rate and global remapping were greater in the proximal CA3 than in the distal CA3

Lu et al. used two protocols to induce place cell remapping. First, they changed the colour of the walls of the rats' chamber between trials. This induced 'rate remapping', in which the firing rate of the cells, but not the location in which they fired, changed. To induce 'global remapping', in which both the firing rate and firing location of the cells are altered, they recorded from animals moving in an open arena that was placed within two different rooms. Both rate and global remapping were greater in the proximal CA3 than in the distal CA3 or CA2, a trend that was also reflected when correlations in population vectors were examined.

These findings suggest that place cells in the proximal CA3 are better able to represent changes in the environment (a requirement for pattern separation), whereas those in the distal CA3 (and CA2) tend to generalize across different environments, making them well suited to a role in pattern completion. These differences may relate to the anatomical features of each region, such as the high level of recurrent connectivity present in the distal CA3.