Representation of Abstract Event Units in the Hippocampus

Post by Cody Walters 

What’s the science?

There is a growing body of literature on how the brain represents information about continuous variables such as space and time. However, it remains comparatively unclear how experiences (which are continuous) become integrated and represented as unitary events (which are discrete). This week in Nature Neuroscience, Sun et al. provide evidence suggesting that individual neurons in the mouse hippocampus encode event information in a generalizable and abstract format. 

How did they do it?

The authors trained mice to run laps in a square maze. Before running the first lap, mice received a reward (a sugar pellet) in a reward box attached to the maze. After running four laps, the mice once again received a reward. Therefore, four laps were treated as one trial, with the reward marking the transition between trials. 

The authors injected an adeno-associated viral vector encoding the calcium indicator GCaMP6F into layer CA1 of the dorsal hippocampus (dCA1) of Wfs1-cre transgenic mice. Throughout the study, they used a microendoscope positioned over dCA1 to measure calcium activity in freely behaving mice. Because dCA1 neurons are known to respond to a variety of situations and behaviours, the authors fit a linear regression model (with regressors for 1) spatial location, 2) head direction and 3) running speed) to the activity of each neuron. They then subtracted the model-explained calcium activity from raw calcium activity. The result of this subtraction (the ‘model-corrected calcium activity’) effectively reflected the activity of dCA1 neurons while minimizing the effect of the three regressors. 

What did they find?

The authors found a subset of dCA1 neurons (~30%) that exhibited an increase in calcium activity during one of the four identical laps. This phenomenon was called event-specific rate remapping (ESR). If a reward was delivered at the outset of every lap instead of every fourth lap, the number of these lap-modulated ESR cells decreased significantly. ESR activity within individual neurons was preserved across days - the correlation between the ESR activity profiles (which were model-corrected) for day 1 and day 2 was significantly higher than chance (i.e., compared to shuffled data). To investigate whether cells were simply tracking time elapsed or distance traveled, they engineered a variation of the experiment that involved two versions of the maze: the original shorter square track and an elongated rectangular track. Under various combinations of track length per trial (e.g., short, short, long, long), they still observed ESR correlation values (across different track length trial types) that were higher than chance. This result provides support in favor of the view that these lap-modulated neurons are responding to laps as fundamental events regardless of their length or duration. 

To further test this hypothesis, the authors conducted an experiment in which they used the square maze on day 1 and a circular maze on day 2. Once again, the correlated ESR activity across days remained significantly above chance levels despite place fields (the region in the environment that a spatially selective neuron responds to) remapping to new preferred positions in the circular maze. In another experiment, the authors extended each trial to include 5 laps instead of 4 and found that neurons typically active on the third lap or fourth lap respectively were now active on the fourth or fifth lap respectively.

Due to prior evidence suggesting that the medial entorhinal cortex contributes to the structuring of experience in the hippocampus, the authors decided to optogenetically inhibit medial entorhinal cortex neurons that terminated in dCA1. They found that the ESR activity correlation between light-on (inhibited) and light off trials was reduced to chance levels once these neurons were inhibited, while place field positions remained unaffected. In the final experiment, mice ran on a treadmill in the first arm of the square maze and the authors measured ESR activity during a 12 second period of time (representing a continuous but essentially non-spatial experience). They discovered that there was a lap-modulation effect wherein dCA1 neurons (~20%) exhibited increased activity when mice were walking on the treadmill only on specific laps. 

What’s the impact?

These data provide evidence that spatial representations and event representations are jointly expressed in individual hippocampal cells, yet these representations are dissociable from one another: one can have persistent event representations despite place field remapping (e.g., the circular track experiment) as well as stable spatial representations despite disruption in event-specific rate remapping (e.g., the fifth lap experiment and the entorhinal inhibition experiment). This study marks a step forward in our understanding of how the brain processes continuous experience and segments it into generalized event units.  

Hippocampal neurons represent events as transferable units of experience. Nature Neuroscience, (2020). Access the original scientific publication here.