How vision and movement combine in the hippocampal place code

Abstract

How do external environmental and internal movement-related information combine to tell us where we are? We examined the neural representation of environmental location provided by hippocampal place cells while mice navigated a virtual reality environment in which both types of information could be manipulated. Extracellular recordings were made from region CA1 of head-fixed mice navigating a virtual linear track and running in a similar real environment. Despite the absence of vestibular motion signals, normal place cell firing and theta rhythmicity were found. Visual information alone was sufficient for localized firing in 25% of place cells and to maintain a local field potential theta rhythm (but with significantly reduced power). Additional movement-related information was required for normally localized firing by the remaining 75% of place cells. Trials in which movement and visual information were put into conflict showed that they combined nonlinearly to control firing location, and that the relative influence of movement versus visual information varied widely across place cells. However, within this heterogeneity, the behavior of fully half of the place cells conformed to a model of path integration in which the presence of visual cues at the start of each run together with subsequent movement-related updating of position was sufficient to maintain normal fields.

Fig. 1.

Place cells firing on a virtual linear track. (A) Six representative examples of place cells firing on a virtual linear track. Color maps of spike rates are created from multiple runs in the eastward direction (Right) and westward direction (Left). Peak firing rates (Hz) are indicated above each plot. Theta cycle phases of spikes from the place field of cell 1 on the westward runs (B) and on the eastward runs (C) are plotted against position (the underlined segments in A). (D) Average spatial information from complex spike cells (n = 56) on the first three training days (**increase from day 1 to day 2, P < 0.01). (E) Spike phase precesses within each firing field, being highest in the early third, lower in the middle third, and lowest in the late third (n = 57). Vertical bars represent ± SEM.

Fig. 2.

Visual control of place fields in the virtual environment. (A) Five representative cells showing different types of responses to visual cue manipulations (left to right: spatial firing that requires only side cues, only end cues, either side or end cues, both side and end cues, or some other stimuli). (B) Percentage of response types corresponding to A (see color key above each type, and SI Materials and Methods for details). Peak firing rate (C), spatial information (D), and spatial correlation (E) decreased as more visual cues were removed. n = 73. *P < 0.05; **P < 0.01; ***P < 0.001 in the comparison with baseline trials.

Fig. 3.

Sensory and motor control of place fields in the virtual environment. (A) Two representative cells showing responses to the manipulation of passive movement: the firing field on the left was maintained, whereas that on the right scattered. (B) Sixteen of 63 cells (25%) show similar firing in the baseline and passive movement trials (spatial correlation within 2 SDs of the mean baseline–baseline correlation, i.e., R > 0.51, red line). (C) Four representative cells showing responses to the manipulation of half speed: the firing fields of cells 1 and 2 moved backward against the direction of movement, whereas that of cell 3 did not move and that of cell 4 scattered. (D) Thirty-two of 73 cells show significant centroid backward shift (shifting by more than 2 SDs from the mean baseline–baseline shift, 2.93 cm). (E) The cells from the backward-spatial-shift group (n = 32) have significantly bigger centroid shift in the half speed trial than between baseline trials. (F) Classification of place cell responses to the half-speed manipulation.

Fig. 4.

Comparison of theta power and mean firing rates between half speed, passive move trials, and baseline trials. (A and B) Two examples of the raw LFP traces in the baseline condition (in blue). Red traces are the filtered theta (4–12 Hz). (C and D) Two examples of LFP power spectra (A and C are from the same mouse; B and D are from the same mouse). (E) Comparison of theta index (the ratio of mean power within 1 Hz of the theta peak and mean power over 2–40 Hz; n = 5) across conditions. (F) Comparison of mean firing rates (n = 63, from the five animals). (G) Comparison of spatial information (n = 63, from the five animals). **P < 0.01, ***P < 0.001.

Fig. 5.

Place cells are driven by heterogeneous combinations of visual and movement-related inputs. (A) A total of 25% of place cells can be driven solely by visual inputs. Cell 1 is an example showing that a place field remained in position in the passive move and half-speed trials. The majority (75%) of place cells also required movement-related inputs. Two representative cells showing dependence on movement-related inputs; their fields disappeared in the passive movement trial, but either moved against the direction of movement in the half-speed trial (cell 2) or maintained their position (cell 3). (B) Relationship of spatial correlation between the passive-movement and baseline conditions to the amount the field shifts in the half-speed condition (r = −0.26, P = 0.04).

Fig. 6.

Movement-related control of place fields in the virtual environment. (A) Two representative cells showing responses to the manipulation of light-off after run start: the firing field on the left was maintained, whereas that on the right scattered. (B) Thirty-six of 73 cells (49%) show similar firing patterns in baseline and light-off trials (spatial correlations within 2 SDs of the mean baseline–baseline correlation, r > 0.51, red line).