Multisensory control of hippocampal spatiotemporal selectivity

Abstract

The hippocampal cognitive map is thought to be driven by distal visual cues and self-motion cues. However, other sensory cues also influence place cells. Hence, we measured rat hippocampal activity in virtual reality (VR), where only distal visual and nonvestibular self-motion cues provided spatial information, and in the real world (RW). In VR, place cells showed robust spatial selectivity; however, only 20% were track active, compared with 45% in the RW. This indicates that distal visual and nonvestibular self-motion cues are sufficient to provide selectivity, but vestibular and other sensory cues present in RW are necessary to fully activate the place-cell population. In addition, bidirectional cells preferentially encoded distance along the track in VR, while encoding absolute position in RW. Taken together, these results suggest the differential contributions of these sensory cues in shaping the hippocampal population code. Theta frequency was reduced, and its speed dependence was abolished in VR, but phase precession was unaffected, constraining mechanisms governing both hippocampal theta oscillations and temporal coding. These results reveal cooperative and competitive interactions between sensory cues for control over hippocampal spatiotemporal selectivity and theta rhythm.

Fig 1

Large reduction of track active cells in VR without comparable reduction in firing rates. A) Schematic of the task environment and distal visual cues in VR and RW. Rats turned themselves around in RW, while the scene was passively reversed in VR. B) Running speed (mean± STD) of rats as a function of position on a 2.2m long linear track for RW (blue) and VR (red). Similar color scheme is used throughout. While the rats were faster in RW, their behavior was similar, reliably reducing speed prior to reaching the end of the track (n=49 sessions in RW, n=128 sessions in VR). C) Example of a directional, stable place cell recorded in RW with firing rate (top panel), and raster plot (bottom panel). Arrows indicate running direction. D) A similar place cell recorded in VR. E) Comparison of activation ratio and firing rates of active cells on track (RW: 45.5%, 3.06 ± 0.12 Hz, VR: 20.4%, 2.71 ± 0.08 Hz) and at goal (RW: 20.1%, 2.68 ± 0.20 Hz, VR: 9.5%, 3.16 ± 0.18 Hz). F) Spatial information content across 432 track active cells in VR (1.23 ± 0.03 bits, n=432) was significantly lower (22%, p<10^-7) than in 240 RW cells (1.58 ± 0.05 bits, n=240).

Fig. 2

Bidirectional place cells exhibit position code in RW but disto-code in VR. A) Firing rate maps along both running directions for bidirectional cells in RW (top) and VR (bottom). Top panel depicts a position coding cell firing at the same position in both running directions. Bottom panel depicts a disto-coding cell firing at the same distance in both directions. B) Position code index is significantly positive in RW (0.27±0.05, p<10^-6, n=91) but significantly negative in VR (-0.11±0.04, p<0.05, n=127). C) The disto-code index is significantly positive in VR (0.14±0.04, p<0.001, n=127) but significantly negative in RW (-0.25±0.06, p<10^-4, n=91). The position code index is significantly greater in RW than VR (p<10^-7) while disto-code index is significantly greater in VR (p<10^-6). D) Similarity of the population of 91 bidirectional cells in RW between two movement directions was computed using the population vector overlap (see methods). Each colored pixel shows the vector overlap between two positions in opposite running directions. Note the clear increase in overlap along the -45° diagonal indicating spiking at the same position. E) As in D, for the population of 127 bidirectional cells in VR. Note the clear increase in overlap along the +45° diagonal, indicating spiking at the same distance in both running directions.

Fig. 3

Reduced theta frequency in VR without significant change in phase precession. A) Autocorrelation function of sample hippocampal LFPs in RW and VR showing significant increase in theta period in VR. Both LFPs are recorded from same electrode on the same day. B) Representative place field in RW showing clear phase precession. C) Representative place field in VR showing clear phase precession. D) Infield theta frequency of place fields in VR (7.53 ± 0.02 Hz, n=251) was significantly less (8.7%, p<10^-10, n=204) than that of place fields in RW (8.25 ± 0.03 Hz, n=204). E) Quality of phase precession, measured by position-phase linear-circular correlation, in RW (0.33 ± 0.01) was not significantly different (p=0.8) from precession in VR (0.33 ± 0.01).

Fig 4

Absence of theta frequency speed dependence in VR. A) Sample theta cycles during high (>50cm/s) and low (<10cm/s) running speed in RW and VR from the same electrode as Fig 3A. Bold lines are filtered between 4 and 12 Hz, narrow lines are unfiltered traces (see Fig S15). B) Population average (mean ± s.e.m) speed dependence of theta frequency from 287 LFPs in RW and 681 LFPs in VR. C) Density map of individual theta cycle frequencies and corresponding speeds from a single LFP in RW (correlation=0.13, p<10^-10). D) As in C, for the same electrode in VR on the same day (correlation=0.01, p=0.48). E) The population of LFPs in RW shows significant correlation between theta frequency and speed (0.21 ± 0.01, p < 10^-10), while the population of LFPs in VR shows no significant correlation (-0.01 ± 0.01, p<0.01). F) Theta cycle amplitude is similarly (p=0.8) correlated with speed in both RW (0.16 ± 0.01) and VR (0.16 ± 0.01).