Spatial modulation of hippocampal activity in freely moving macaques

Abstract

The hippocampal formation is linked to spatial navigation, but there is little corroboration from freely moving primates with concurrent monitoring of head and gaze stances. We recorded neural activity across hippocampal regions in rhesus macaques during free foraging in an open environment while tracking their head and eye. Theta activity was intermittently present at movement onset and modulated by saccades. Many neurons were phase-locked to theta, with few showing phase precession. Most neurons encoded a mixture of spatial variables beyond place and grid tuning. Spatial representations were dominated by facing location and allocentric direction, mostly in head, rather than gaze, coordinates. Importantly, eye movements strongly modulated neural activity in all regions. These findings reveal that the macaque hippocampal formation represents three-dimensional (3D) space using a multiplexed code, with head orientation and eye movement properties being dominant during free exploration.

Keywords: motion tracking; non-human primate.

Fig. 1.. Freely-moving monkey setup and electrode localization.

(A) The cylindrical arena. (B) Monkey K’s drive and skull models. (C) Segmented models of the brain, hippocampus (blue) and entorhinal cortex (green). (D) Segmented models of the hippocampus, entorhinal cortex, and blood vessels. Electrodes are visible as white tracks in the co-registered CT images. (E) CT, MRI, and co-registered images for sagittal (left) and axial (right) views. White tracks and dots are electrodes. (F) Models of the hippocampus and entorhinal cortex color-coded by dorsal-ventral position. Four axial planes are highlighted. (G) CT and MRI images corresponding to the 4 axial plans shown in (F). Hippocampus is circled in blue; entorhinal cortex is circled in green. Electrodes are shown as white dots. Deeper electrodes are also visible in more dorsal planes. (H) Four markers were placed on the head cap. Monkey K was trained to wear a wireless eye tracker. (I) Neural logger recordings and wireless transmission of eye tracking images. (J) A short segment of the markers’ trajectories in 3D.

Fig. 2.. Speed modulation of hippocampal local field potential.

(A) Left, absolute power spectrum; middle, aperiodic fit to the left; right, power spectrum of the oscillatory component (absolute - aperiodic). Shaded areas: SEM over sessions. (B) Scalogram of a short segment of hippocampal LFP. Speed trace is shown in white. (C) Mean partial correlation between speed and LFP power in different bands. Error bars: SEM over sessions. (D) Examples of theta and beta activity. (E) Colormaps of peak theta (top) and beta (bottom) power-triggered, normalized speed, shown for each session separately and as mean±SEM across sessions. (F) Left, colormap of LFP power plotted in frequency vs. speed. Middle, relative power (normalized such that the power ranges from 0 to 1 in each speed bin) in the theta and beta bands. White dots indicate the peak frequency in each speed bin. (G) Theta and beta frequencies as a function of acceleration shown as mean±SEM across sessions.

Fig. 3.. Diverse spatial tuning across hippocampal regions.

(A) Cumulative distribution of the firing rates for neurons recorded in each HF region. (B) Example raw tuning curves for head height, translational speed, azimuth head direction, and LFP phase (1-10 Hz). (C) Example raw data (top) and raw tuning curves (bottom) for horizontal position (place fields), facing location (where the head points on the arena surface), egocentric boundary, head tilt, and angular velocity. Red dots: spikes; gray: behavioral variables. Colormaps (bottom) show the raw tuning curves (bottom). Peak firing rate (yellow) is indicated. Tuning curves are all from different neurons. (D) Fraction of neurons modulated by position, grid, azimuth head direction, and speed for each region, computed using traditional analyses. Chance level is at 0.01. Bottom, Venn diagram showing the number of neurons encoding each variable and combination of variables.

Fig. 4.. Mixed selectivity of diverse spatial variables.

(A) A model-based statistical framework was used to quantify the spatial coding. (B) An example cell showing tuning to facing location and position. (C) Log-likelihood (LLH; goodness of fit) increase as a function of model variables for the example cell shown in (B). (D) Pie chart shows the name and the fraction of select variables that were best encoded in each region. ‘None’ cluster corresponds to neurons not encoding any variables. (E) Breakdown of the fraction of encoding neurons for each variable in each region. A single neuron can encode none, one, or more than one variable. (F) Fraction of neurons tuned to single variable (single) or to more than one variable (mixed) in each region. (G) Circular graph representation of the degree of conjunction between variables. Line thickness and color correspond to how often two variables are co-coded in a single neuron. (H) Fraction of putative interneurons and principal cells encoding each variable. (I) Left, fraction of encoding neurons along the hippocampal long axis for variables FL, EB, and HT. Right, fraction of neurons showing single and mixed encoding along the long axis.

Fig. 5.. Spatial representations in the hippocampal formation are heterogeneous.

(A) Example model-based tuning curves for 10 neurons coding facing location (FL). Peak firing rates are indicated. (B) Example model-based tuning curves for 9 neurons coding head tilt (HT). (C) Preferred firing locations (red dots) for all neurons encoding FL superimposed on the average occupancy colormap across all monkeys. Dot size corresponds to neuron count. (D) Preferred firing fields for all neurons encoding HT superimposed on the average occupancy colormap across all monkeys. Dot size corresponds to neuron count. (E) and (F) The same as (C) and (D) for position and EB encoding neurons, respectively. (G) Left, speed occupancy across monkeys. Shaded area: 1x SD across sessions. Middle, distribution of preferred speed. Right, clustering analysis for the speed tuning curves using PCA. (H) The same as (G) for azimuth head direction. Gray bar indicates the direction of the exit/entrance door.

Fig. 6.. Spike-LFP phase coding.

(A) Spike-LFP phase distribution for 6 frequency bands for 3 example neurons. (B) Distribution of preferred LFP phase for all significantly phase-locked neurons. (C) Bar plot of the fraction of phase-locked neurons in each frequency band for spatially-tuned and other neurons. Out of all position tuned neurons, 54%, 21%, 28%, 44%, 61%, and 82% were phase tuned to each band, respectively. (D) Example filtered LFP trace and spike raster. (E) Spike phase autocorrelation for the example neuron shown in (D). Dashed lines indicate theta cycle. Red arrows indicate peak autocorrelation in consecutive theta cycles. (F) Power spectrum of the autocorrelation shown in (E). Gray dashed line shows the 99% percentile of the shuffled distribution. Spike-LFP relative frequency larger than 1 indicates spikes oscillate at a higher frequency than theta. (G) and (H) The same as (E) and (F) for an example phase-locking neuron without phase precession. (I) Fraction of neurons showing significant phase precession for low theta and theta.

Fig. 7.. Allocentric facing location and head direction tuning predominantly reflects head but not gaze properties.

(A) Example raw traces showing eye-in-head position (EP), facing location (FL), and spatial view (SV). (B) Fraction of neurons in each region encoding FL, SV, azimuth head direction (HD) and gaze direction (GD) when fitting all head- and gaze-related variables simultaneously. (C) Colormap representation of the firing as a function EP for an example neuron. (D) Colormap representation of the firing as a function of eye-in-head velocity (EV) for an example neuron. (E) Fraction of neurons in each region (color-coded bars, same as (B)) encoding EP and EV. (F) Mean firing rate from saccade onset for an example neuron. Error bars, SEM over saccade events. (G) Colormap of the normalized firing rate for all saccade-tuned neurons (sorted). (H) Fraction of neurons in each region encoding saccade event. (I) Confusion matrix showing the number of neurons encoding any combinations of FL and EP. (J) Average saccade onset-triggered LFP for an example session. (K) Average saccade onset-triggered LFP for all regions. Shaded area, SEM over sessions. (L) Theta band power vs. saccade magnitude for an example session. Error bars, SEM over saccade events. (M) Average, normalized theta band power vs. saccade magnitude for all sessions. Error bars, SEM over sessions. Data from monkey K.