Research ArticleNEUROSCIENCE

Hippocampal theta phases organize the reactivation of large-scale electrophysiological representations during goal-directed navigation

See allHide authors and affiliations

Science Advances  03 Jul 2019:
Vol. 5, no. 7, eaav8192
DOI: 10.1126/sciadv.aav8192
  • Fig. 1 Virtual navigation task and behavioral data.

    (A) Associative object-location memory task during virtual spatial navigation. At the beginning of the experiment, patients collected eight different objects from eight different locations within the virtual environment. Afterward, patients completed variable numbers of retrieval trials, during which they were first presented with one of the eight objects serving as cue (“cue presentation”). Patients then navigated to the remembered location of that object (“retrieval”) and made a response. Following this response, patients received feedback via an emoticon (“feedback”) and had to collect the object from its correct location (“re-encoding”). (B) Overhead view of the virtual environment (diameter, 9500 vu). Goal-directed navigation occurred after cue presentation, when patients started (“S”) navigating to the assumed object location. Starting locations were identical with ending locations from preceding trials and thus varied from trial to trial. The trial-wise drop error was calculated as the Euclidean distance between the response location (“R”) and the correct location (“C”). (C) Histogram of drop errors across all trials and all patients. Red dashed line, overall chance performance. (D) Change in mean drop error across objects between the first and the last trial. Gray lines, patient-wise data; thick red line, average.

  • Fig. 2 Identification of large-scale electrophysiological cue representations using tr-sRSA.

    (A) Colored brain surfaces showing the number of channels for each Montreal Neurological Institute (MNI) coordinate. Black coloring, no coverage. (B) Analysis principle (illustration). In both data halves, we obtained one cue representation (across channels) per time point (t = x) during cue presentation (each small brain, one cue representation). We estimated neural similarity between each pair of cue representations, giving an 8 × 8 confusion matrix. On-diagonal (green squares) and an equal number of off-diagonal values (blue squares; randomly chosen) were extracted, resulting in a time point–specific measure of neural similarity between identical and different cues. (C) Higher similarity values for identical as compared to different cue representations between 256 and 530 ms after cue onset (red shaded area). Multivariate iEEG activity contains cue-specific information during this time window, constituting a tROI for subsequent analyses. (D) More distinct neural representations of cues with subjectively more similar goal locations (left bar plot), driven by the second data half (right bar plot). Error bars represent SEM. (E) Contribution of brain regions to the cue representations (“relative engagement”). Light blue bars, number of implanted channels. *Pcorr < 0.05 (Bonferroni-corrected for 29 regions).

  • Fig. 3 Hippocampal theta oscillations during goal-directed navigation.

    (A) Depiction of selected hippocampal electrode channels, which were located in the anterior hippocampus. Each white dot represents one channel from a separate patient (n = 16). (B) Hippocampal ERP during cue presentation. (C) Exemplary time periods with theta oscillations during goal-directed navigation from different patients. Black, raw signal; red, low-frequency component of the raw signal (passband, 1 to 10 Hz); green shading, time periods with theta oscillations as detected by MODAL (see Materials and Methods). (D) Summary of frequency bands detected by MODAL, across patients, showing that they preferentially occurred at a frequency of 3 to 4 Hz. Dots represent mean, and vertical lines represent SEM.

  • Fig. 4 Dynamic reactivation of cue representations at distinct hippocampal theta phases during goal-directed navigation.

    (A) Analysis procedure of representation-to-theta-phase-clustering. For each cue, we extracted one neural representation vector (NRVi) within the tROI during cue presentation. For each trial, we then calculated the dynamically changing similarity between NRVi (i.e., the NRV of cue i whose goal location had to be retrieved during this trial) and all NVs (NV1−n) during the retrieval period. (B) This resulted in a sliding RSA time course (top subplot). In addition, we obtained one hippocampal (HC) theta phase for each time point during retrieval where power exceeded the background 1/f spectrum (bottom subplot). For each trial, the preferred theta phase of the sliding RSA values was extracted via the nonparametric Moore-Rayleigh test. (C) We assessed the significance of representation-to-theta-phase-clustering by comparing against a surrogate distribution that was obtained by circularly shifting the sliding RSA values against the hippocampal theta phases. Red dot and red line, empirical mean. (D) Preferred theta phases of one patient. Bold lines, preferred theta phases; shaded areas, circular SEM. (E) Stability of preferred theta phases across trials (within-stability). (F) Distinctiveness of preferred theta phases (between-similarity). (G) Lower between-similarity values are associated with better spatial memory performance.

  • Fig. 5 Exemplary trials of one patient depicting representation-to-theta-phase-clustering.

    The circle on the top left depicts the preferred theta phases from eight different trials (one for each cue representation). This circle is unfolded on the top right. Color coding corresponds to Fig. 4D. The eight subplots on the left side show sliding RSA time courses (white) and concurrent hippocampal theta phases (colored) during a 1-s time interval of each of the eight trials (corresponding to three to four theta cycles). Shaded areas indicate the time periods during which the phase bin of the preferred theta phase was present. Phase jumps (indicated by small black triangles above each subplot) occur because only periods were included when theta oscillations above the background 1/f spectrum were identified by MODAL (these periods were concatenated). The eight right-hand subplots depict the averaged sliding RSA values (separately for each phase bin) from the entire trial. Colored bold lines depict preferred theta phases. r* values obtained from the Moore-Rayleigh test indicate the strength of representation-to-theta-phase-clustering for the given trial. The eight subplots on the right and left side were ordered according to the preferred theta phase of the dynamically reactivating cue representations. sRS, sliding representational similarity; RS, representational similarity.

  • Fig. 6 Dynamic reactivation of cue representations at distinct prefrontal theta phases during goal-directed navigation.

    (A) Prefrontal electrode channels. Each white dot, one channel from a separate patient (n = 13). Blue, lateral orbitofrontal cortex; red, rostral middle frontal gyrus. (B) Exemplary time periods with theta oscillations during goal-directed navigation from different patients. Black, raw signal; red, low-frequency component of the raw signal (passband, 1 to 10 Hz); green shading, time periods with theta oscillations as detected by MODAL. (C) Prefrontal (“PFC”) theta oscillations occurred in temporal proximity to hippocampal theta oscillations as shown by an increased percentage of time with prefrontal theta oscillations when hippocampal theta oscillations were present as compared to when they were not present. (D) Prefrontal theta oscillations preferentially occurred at a frequency of 5 to 6 Hz. Dots represent mean, and vertical lines represent SEM. (E) We assessed the significance of representation-to-theta-phase-clustering by comparing against a surrogate distribution that was obtained by circularly shifting the sliding RSA values against the concurrent prefrontal theta phases. Red dot and red line, empirical t statistic. (F) Stability of preferred theta phases across trials (within-stability). (G) Distinctiveness of preferred theta phases (between-similarity). (H) Lower between-similarity values are associated with better spatial memory performance.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/7/eaav8192/DC1

    Table S1. Patient information.

    Table S2. MNI coordinates of hippocampal channels.

    Fig. S1. Layout of the virtual environment and patient-wise goal locations.

    Fig. S2. Stimulus specificity of neural cue representations.

    Fig. S3. Identification of large-scale electrophysiological cue representations using tr-sRSA based on gamma power.

    Fig. S4. Derivation of higher-order similarity.

    Fig. S5. Neural cue representations rely on large-scale electrophysiological signals.

    Fig. S6. Contribution of brain regions to the similarity of identical and different large-scale electrophysiological cue representations.

    Fig. S7. Second-level statistics across patients depicting which brain regions simultaneously increased the neural similarity of identical cues and decreased the neural similarity of different cues.

    Fig. S8. Phase coupling (3.5 Hz) between the hippocampus and prefrontal cortex during goal-directed navigation.

  • Supplementary Materials

    This PDF file includes:

    • Table S1. Patient information.
    • Table S2. MNI coordinates of hippocampal channels.
    • Fig. S1. Layout of the virtual environment and patient-wise goal locations.
    • Fig. S2. Stimulus specificity of neural cue representations.
    • Fig. S3. Identification of large-scale electrophysiological cue representations using tr-sRSA based on gamma power.
    • Fig. S4. Derivation of higher-order similarity.
    • Fig. S5. Neural cue representations rely on large-scale electrophysiological signals.
    • Fig. S6. Contribution of brain regions to the similarity of identical and different large-scale electrophysiological cue representations.
    • Fig. S7. Second-level statistics across patients depicting which brain regions simultaneously increased the neural similarity of identical cues and decreased the neural similarity of different cues.
    • Fig. S8. Phase coupling (3.5 Hz) between the hippocampus and prefrontal cortex during goal-directed navigation.

    Download PDF

    Files in this Data Supplement:

Navigate This Article