Scalable Representation of Time in the Hippocampus
Post by Andrew Vo
What's the science?
Hippocampal place cells allow us to form map-like spatial representations of our environments, and these maps adaptively rescale themselves when our environments change. Whether hippocampal “time cells” can also form such scalable representations for information about time has yet to be systematically investigated. This week in Science Advances, Shimbo and colleagues examined if hippocampal CA1 activity of rats during encoding of time intervals scaled in response to expansions or contractions in elapsed time.
How did they do it?
The authors trained rats to perform a task during which they ran on a treadmill for either long (e.g. 10 s) or short (e.g. 5 s) time intervals before navigating a Y-maze (shaped like a Y). The left and right arms of the maze were associated with long and short treadmill time intervals, respectively, and so the rats had to discriminate between intervals to select the correct arm and receive a reward. The rats performed three blocks of trials, across which the sets of time intervals were scaled up or down. For example, rats would discriminate between 10 (long) and 5 s (short) intervals in block 1, which were scaled up to 20 (long) and 10 s (short) intervals in block 2, before returning to their original interval durations in block 3.
During the treadmill interval periods, the authors recorded activity from hippocampal CA1 to identify time cells (i.e., neurons whose activity represented information of elapsed time). The firing activity of these cells was compared across experimental blocks using peri-event time histograms (PETHs) that quantify the rise and fall of activity over time in relation to the event. Using this method, scaling factors in response to changes in time intervals could be quantified. To test whether time cell activity was specific to a time-based task, they trained a different set of rats on a light discrimination task that shared the same structure as their original task, except Y-maze performance was based on a light cue instead of interval times.
Next, the authors recorded theta sequences in the brain, which are patterns of neuron firing among cell assemblies that represent compressed time episodes. They also tested if these theta sequences scaled to changes in time intervals. Finally, they used a statistical method — Bayesian decoding — to decode these theta sequences and see if time cell activity predicted the rats’ Y-maze decisions.
What did they find?
The authors found that rat hippocampal CA1 activity during the treadmill interval period represented information on the elapsed time that scaled up or down depending on the expansion and contraction of the time intervals. This finding appeared to be related to task demands, as the number of time cells was significantly reduced when rats were not required to estimate time during the light discrimination task. This reduced number of time cells continued to display scalable representations of elapsed time, however. Examining the finer temporal structure of time cell ensembles, they noted the presence of theta sequences that were also scalable when time intervals were varied. The temporal information of these theta sequences could be decoded and reflected the rats’ decisions based on time estimation during test trials.
What's the impact?
In summary, this study demonstrated that time cells in the hippocampus can form scalable temporal representations of the environment, similar to how place cells code for spatial information. These findings suggest there is a common mechanism in the hippocampus underlying representations of temporal and spatial information by time and place cells, respectively. The ability to flexibly scale such representations might allow us to better navigate our complex and changing environments.
Shimbo et al. Scalable representation of time in the hippocampus. Science Advances (2021). Access the original scientific publication here.