Hippocampal firing fields anchored to a moving object predict homing direction during path-integration-based behavior

Homing based on path integration (H-PI) is a form of navigation in which an animal uses self-motion cues to keep track of its position and return to a starting point. Despite evidence for a role of the hippocampus in homing behavior, the hippocampal spatial representations associated with H-PI are largely unknown. Here we developed a homing task (AutoPI task) that required a mouse to find a randomly placed lever on an arena before returning to its home base. Recordings from the CA1 area in male mice showed that hippocampal neurons remap between random foraging and AutoPI task, between trials in light and dark conditions, and between search and homing behavior. During the AutoPI task, approximately 25% of the firing fields were anchored to the lever position. The activity of 24% of the cells with a lever-anchored field predicted the homing direction of the animal on each trial. Our results demonstrate that the activity of hippocampal neurons with object-anchored firing fields can predict homing behavior.

Methods
Mice were connected to the data acquisition system (RHD2000-Series Amplifier Evaluation System, Intan Technologies, analog bandwidth 0.09–7603.77 Hz, sampling rate 20 kHz) via a lightweight cable. The recording was controlled using ktan software (https://github.com/kevin-allen/ktan). Kilosort2 (https://github.com/jamesjun/Kilosort2) was used for spike extraction and clustering. Automatically generated clusters were visually inspected and manually refined with Phy (https://github.com/cortex-lab/phy).

The quality of spike clusters was estimated from the spike-time autocorrelation. A refractory period ratio was calculated from the spike-time autocorrelation from 0 to 25 ms (bin size: 0.5). The mean number of spikes from 0 to 1.5 ms was divided by the maximum number of spikes in any bin between 5 and 25 ms. Any cluster with a refractory period ratio larger than 0.25 was discarded.

In experiments with electrophysiological recordings, the position of the mouse was estimated from the position of infrared-LEDs (wavelength 940 nm) located on both sides of the head of the mouse. Three LEDs were used on one side and one on the other. The distance between the two LED groups was 3 cm. Two video cameras fitted with long-pass filters (Cut-On wavelength: 800 nm) and located directly above the recording environment monitored the position of the LEDs at 30 or 50 Hz. The location of the mouse was extracted online from the position of the LEDs (https://github.com/kevin-allen/positrack or https://github.com/kevin-allen/positrack2).