Tracking receptive field dynamics in CA1 and the entorhinal cortex with an adaptive point process filtering algorithm

Loren M. Frank1, U.T. Eden1, R. Barbieri1, V. Solo2 Matt A. Wilson3 and Emery N. Brown1

2University of New South Wales, Sydney, Australia
3Riken - MIT Neuroscience Institute

Neural receptive fields are dynamic: the nature of a neuron's response to a stimulus can change over time as a result of experience. In the CA1 region of the hippocampus, place cells' spatial receptive fields increase in total firing rate, and move and become skewed in the direction opposite the animal's direction of motion as a result of repeated passes through the place field (Mehta et al. 1997, 2000). Here we extend those findings by developing an adaptive point process filtering algorithm that allows us to estimate receptive field shape on a millisecond time scale without binning over time or space (Brown et al. 2000). Our algorithm allowed us to estimate a firing rate function for each cell that depends on both the animal's position and the history of spiking. We applied this algorithm to simultaneously recorded neurons from the CA1 subregion of the hippocampus and from the superficial and deep layers of the entorhinal cortex (EC) of rats running along a U shaped track. The EC was chosen as a recording site because superficial EC cells provide the main source of neocortical input to the hippocampus and deep EC cells are the primary targets of neocortically bound hippocampal outputs. We found that including history dependence in our model improved the goodness-of-fit to the data, suggesting that examining firing rate as a function of behavioral correlates alone may not accurately describe neural activity patterns. Overall, our results for CA1 were consistent with those of Mehta et al. (1997, 2000). At the same time, the place fields of superficial and deep EC cells did not show a tendency to move backwards along the animal's direction of motion. The most striking finding was that, on average, the spatial firing rate functions of superficial and deep EC cells, like those of CA1 cells, increased in total area while the variance of these functions either decreased or remained constant. In addition, the increase in total firing was significantly faster in CA1 than in the deep EC. These results suggest that place specific activity in both CA1 and the EC sharpens over the course of the animal's experience and that receptive field plasticity may occur on a faster time scale in CA1 as compared to the EC. Thus, our findings support the hypothesis that the hippocampus is a site of rapid learning as compared to the EC. More generally, we believe that our algorithm may provide a new and powerful tool for analyzing the dynamics of receptive field plasticity.