Figures and data
Distinct PFC and CA1 beta frequency activity after the reward location entry during spatial navigation tasks.
a. Spatial learning tasks on Y and F shaped mazes with spatial alternation rules. Animals were trained on the mazes across interleaved sessions. Reward locations (1, 2 and 3) are indicated. The order of visits required to ensure reward on every trial is shown above the maze schematics. The daily training schedule shows the duration, order and separation between sessions. b. Example PFC (orange) and CA1 (blue) LFP traces aligned to goal location entry (black vertical line). The beta frequency-filtered LFP is shown below the raw LFP. The corresponding continuous wavelet transform spectrogram is included. Beta bursts are indicated by horizontal lines. Dotted vertical lines mark 1 s intervals. c. Power spectral density for time intervals when the animal is at the reward locations. Left panel shows the PSD, with the aperiodic component shown in gray. Right panel show PSD with aperiodic component subtracted. The peak marked by * in the CA1 PSD at ∼15Hz corresponds to the first harmonic of theta (Scheffer-Teixeira & Tort, 2016). d. Frequency profile for beta bursts grouped by peak power frequency for PFC (orange) and CA1 (blue). Top: the vertical histogram of the mean frequency profiles for bursts at each peak power frequency. Bottom: histograms with the count distribution for bursts within 15-30Hz, grouped by peak power frequency. Shading corresponds to the probability value at each frequency. Bursts occurring within 5s after reward location entry are included. e. Cumulative distribution of peak burst frequencies for PFC (orange) and CA1 (blue). Kolmogorov-Smirnov test, p=2.9×10−103. f. Distribution of burst times relative to goal location entry grouped by peak frequency for PFC (orange) and CA1 (blue). g. Cumulative distribution of burst occurrence relative to goal location entry for PFC (orange) and CA1 (blue). Kolmogorov-Smirnov test, p=1.1×10−4.
Location of recording tetrodes.
a. Representative histology showing location of tetrodes in PFC. b. Representative histology showing location of tetrodes in CA1.
Burst detection.
a. Two example bursts with corresponding LFP and continuous wavelet transform from PFC (1.5 s each, top row). The burst is marked by the black horizontal line. The corresponding continuous wavelet transform is shown in the middle row. The white cross marks the frequency and time corresponding to the maximum power in the burst interval. The mean power in the beta band (15-30Hz) from the continuous wavelet transform is shown by the histogram in the bottom row. The solid black line corresponds to 1 standard deviation (z) above the mean (1z), and the dotted black line corresponds to 0.5 standard deviation above the mean (0.5z). To detect bursts, we identify intervals during which the beta-band power exceeds 0.5 standard deviation above the mean. Intervals are included if the maximum power in the interval exceeds 1 standard deviation above the mean, and the duration of the interval is greater than 100ms. Intervals separated by less than 100ms were merged. The peak frequency for each burst is the frequency corresponding to the maximum power within the burst interval.
Power and duration for beta bursts by frequency.
a. Beta burst power by peak burst frequency. b. Beta bursts duration by peak burst frequency.
Beta bursts occur independently in both regions.
a. Beta frequency bursts (shaded) detected across multiple electrodes in PFC (orange) and CA1 (blue) relative to reward location entry (black vertical line). Three example trials are shown. b. Coherence (mean±SEM) within PFC (orange), between PFC and CA1 (green), or within CA1 (blue) for periods at the goal location. c. Mean coherence (15-30Hz) across the groups. Pairwise rank sum test with Benjamini-Hochberg false discovery rate correction: p<0.001 for all pairs. d. Cross-correlation (mean±SD) of burst peak times for bursts detected on tetrodes with PFC (orange), across PFC-CA1 (green) and within CA1 (blue). e. Proportion of the cross-correlation density within ±50ms lag for PFC (orange), across PFC-CA1 (green) and within CA1 (blue). Pairwise rank sum test with Benjamini-Hochberg false discovery rate correction: PFC versus PFC-CA1 (p=0.003**), PFC versus CA1 (p=0.24n.s.) and PFC-CA1 versus CA1 (p=0.003**). f. Frequency distribution of bursts by peak frequency for independent and coincident bursts in each region. g. Burst peak frequency distribution for coincident and independent bursts. Cumulative distribution burst occurrence relative to goal location entry. Kolmogorov-Smirnov test: PFC (p=0.96n.s.) and CA1 (p=0.002**).
Beta burst aligned power and coherence
a. Multi-taper power spectrum within a 1 s window centered on the peaks of independent or coincident beta bursts (columns) in the PFC (top row) and CA1 (bottom row). The heatmap shows the mean across all bursts in each category. b. Power (18-25Hz) within ±100 ms of the peak. PFC (orange, top) and CA1 (blue, bottom). Pairwise rank-sum test with Benjamini-Hochberg false discovery rate correction: p<0.001*** and p<0.05* for all significant comparisons. c. Change in power within ±100 ms of the peak relative to baseline (125 ms from - 0.5 s). PFC (orange, top) and CA1 (blue, bottom). Wilcoxon signed-rank test with Benjamini-Hochberg false discovery rate correction: p<0.001***. d. Multi-taper coherence in PFC (upper row) and CA1 (lower row) within a 1s window centered on the peaks of independent or coincident beta bursts. The heatmap shows the mean across all bursts in each category. e. Coherence (18-25Hz) within ±100 ms of the peak in PFC (upper row) and CA1 (lower row). Pairwise rank sum test with Benjamini-Hochberg false discovery rate correction: p<0.001*** and p<0.05* for all significant comparisons. f. Change in coherence within ±100ms of the peak relative to baseline (125ms from -0.5s). PFC (orange, top) and CA1 (blue, bottom). Wilcoxon signed-rank test with Benjamini-Hochberg false discovery rate correction: p<0.001***. g. PFC-CA1 coherence within a 1 s window centered on the peaks of independent or coincident beta bursts. The heatmap shows the mean across all bursts in each category. h. PFC-CA1 coherence between (18-25Hz) within ±100 ms of the peak in PFC (top row) and CA1 (bottom row). Pairwise rank sum test with Benjamini-Hochberg false discovery rate correction: p<0.001*** and p<0.01** for all significant comparisons. i. Change in PFC-CA1 coherence within ±100 ms of the peak relative to baseline (125 ms from - 0.5 s). PFC (orange, top) and CA1 (blue, bottom). Wilcoxon signed-rank test with Benjamini-Hochberg false discovery rate correction: p<0.001*** and p<0.01**.
Distinct changes in PFC and CA1 burst properties across days of learning.
a. Performance in each maze task by day. Linear model with fixed (day) and random (animal) effects. Y maze (βday=0.073, SE=0.001, 95% CI=(0.071, 0.075), z=71.9 and p<0.001***) and F maze (βday=0.04, SE=0.001, 95% CI=(0.038, 0.042), z=42.4 and p<0.001***). b. Burst power in each maze task by day. Linear model with fixed (day) and random (animal) effects. Y maze PFC (βday=347.7, SE=38.2, 95% CI=(272.9, -422.5), z=9.1 and p<0.001***) and F maze PFC (βday=300.3, SE=49.4, 95% CI=(203.5, 397.1), z=6.1 and p<0.001***). Y maze CA1 (βday=-1133.2, SE=121.5, 95% CI=(-1371.3, -895.0), z=-9.3 and p<0.001***) and F maze CA1 (βday=-711.7, SE=158.6, 95% CI=(-1022.6, -400.9), z=4.5 and p<0.001***). c. Burst peak frequency variance in each maze task by day. Linear model with fixed (day) and random (animal) effects. Y maze PFC (βday=-1.1, SE=0.3, 95% CI=(-1.7, -0.48), z=-3.6 and p<0.001***), F maze PFC (βday=-0.37 SE=0.3, 95% CI=(-.94, .22), z=6.1 and p=0.22n.s.), Y maze CA1 (βday=-0.89, SE=0.30, 95% CI=(-1.5, -0.31), z=-3.0 and p=0.003**) and F maze CA1 (βday=-0.38, SE=0.32, 95% CI=(-1.0, 0.25), z=-1.2 and p=0.24n.s.) .
Distinct changes in PFC and CA1 burst properties with performance.
a. Performance in each maze task by performance quintile. Linear model with fixed (performance quintile) and random (animal) effects. Y maze (βday=0.084, SE=0.001, 95% CI=(0.083, 0.086), z=103.3 and p<0.001***) and F maze (βday=0069, SE=0.001, 95% CI=(0.068, 0.070), z=136.9 and p<0.001***). b. Burst power in each maze task by performance quintile. Linear model with fixed (performance quintile) and random (animal) effects. Y maze PFC (βperf.=484.4, SE=40.5, 95% CI=(405.1, 563.8), z=12.0 and p<0.001***) and F maze PFC (βperf.=242.6, SE=60.9, 95% CI=(123.2, 362.0), z=4.0 and p<0.001***). Y maze CA1 (βperf.=-1199.6, SE=127.2, 95% CI=(-1448.9, -950.3), z=-9.4 and p<0.001***) and F maze CA1 (βperf.=60.6, SE=198.3, 95% CI=(-328.1, 449.3), z=0.3 and p=0.76n.s.). c. Burst peak frequency variance in each maze task by performance quintile. Linear model with fixed (performance quintile) and random (animal) effects. Y maze PFC (βperf.=-0.80, SE=0.30, 95% CI=(-1.4, -0.22), z=--2.7 and p<0.007**) and F maze PFC (βperf.=-0.68, SE=0.54, 95% CI=(-1.74, 0.39), z=-1.2 and p=0.21n.s.). Y maze CA1 (βperf.=-1.2, SE=0.35, 95% CI=(-1.87, -0.50), z=-3.40 and p=0.001**.) and F maze CA1 (βperf.=-0.27, SE=0.67, 95% CI=(-1.57, 1.03), z=-0.41 and p=0.68n.s.).
Bursts occur sooner after goal entry over days of training.
a. Y maze burst distribution after goal entry for rewarded trials. PFC (orange) and CA1 (blue). Gray line indicates the smoothed histogram. Black plus marks the peak of the smoothed histogram. Linear model with fixed (day) and random (animal) effects on burst timing: PFC (βday=-0.127 SE=0.023, 95% CI=(-0.17, -0.081), z=-5.45 and p<0.001) and CA1 (βday=-0.11 SE=0.030, 95% CI=(-0.17, -0.048), z=-3.54 and p<0.001). b. Y maze burst distribution after goal entry for unrewarded trials. PFC (orange) and CA1 (blue). Linear model with fixed (day) and random (animal) effects on burst timing: PFC (βday=-0.14 SE=0.057, 95% CI=(-0.26, -0.031), z=-2.51 and p=0.012) and CA1 (βday=-0.23 SE=0.065, 95% CI=(-0.34, -0.099), z=-3.48 and p<0.001). c. Boxplot comparison for rewarded and unrewarded trials. Wilcoxon rank sum test: p=2.20×10−4 (PFC) and p=1.15×10−2 (CA1). d. F maze burst distribution after goal entry for rewarded trials. PFC (left) and CA1 (right). Gray line indicates the smoothed histogram. Black plus marks the peak of the smoothed histogram. Linear model with fixed (day) and random (animal) effects on burst timing: PFC (βday=-0.042 SE=0.037, 95% CI=(-0.12, 0.030), z=-1.15 and p=0.25) and CA1 (βday=-0.05 SE=0.047, 95% CI=(-0.18, 0.008), z=-1.78 and p=0.073). e. F maze burst distribution after goal entry for unrewarded trials. PFC (left) and CA1 (right). Linear model with fixed (day) and random (animal) effects on burst timing: PFC (βday=-0.082 SE=0.043, 95% CI=(-0.17, 0.003), z=-1.9 and p=0.057) and CA1 (βday=-0.056 SE=0.058, 95% CI=(-0.17, -0.057), z=-0.97 and p=0.33). f. Boxplot comparison for rewarded and unrewarded trials. Wilcoxon rank sum test: p=2.89×10−4 (PFC) and p=8.01×10−4 (CA1).
PFC and CA1 cells show stronger spiking modulation to local beta.
a. Example PFC (orange) and CA1 (blue) spike phase locking histogram, sorted by increasing phase locking strength. The white line indicates the peak and trough of beta. The distributions were duplicated across two cycles of beta to improve visualization. The strength of the concentration parameter kappa (k) is indicated for each cell. b. Distribution of Rayleigh z for PFC (upper) and CA1 (lower). The proportion of cells showing significant phase locking are shown. c. Distribution of k from a von Mises distribution fit for significantly beta modulated cells (Rayleigh p <0.05 in panel b) in PFC (upper) and CA1 (lower). d. Distribution of phase preference for cells in PFC (upper) and CA1 (lower). Red dotted line shows the mean population phase preference. The gray line indicates the peak and trough of beta. Watson-Williams circular means test for PFC and CA1: p=1.01×10−10. e. Beta phase locking strength (Rayleigh z) to local or remote beta for cells in PFC (upper) and CA1 (lower). Wilcoxon signed-rank test: p=3.17×10−15 (PFC) and p=3.64×10−31 (CA1). f. Beta phase locking strength (kappa) to local or remote beta for cells in PFC (upper) and CA1 (lower). Wilcoxon signed-rank test: p=2.20×10−11 (PFC) and p=6.76×10−24 (CA1).
Beta bursts occur earlier than hippocampal SWRs at goal locations and have inverse learning-related changes in amplitude compared with PFC beta bursts.
a. PFC (orange) and CA1 (blue) LFP relative to goal entry (black vertical line). Beta frequency filtered signal is shown below the LFP. Ripple frequency (150-250Hz) filtered CA1 signal is shown in black. b. Goal entry aligned histogram for SWRs. c. SWR size across days of training. Linear model with fixed (day) and random (animal) effects: βday=-0.15 SE=0.01, 95% CI=(-0.16, -0.13), z=-15.3 and p<0.001. d. Goal entry aligned histogram for PFC beta (orange) or CA1 beta bursts (blue). e. Cross-correlation between PFC (left) or CA1 (right) beta bursts and SWRs. Wilcoxon signed-rank test p=3.12×10−114 (PFC) and p=5.64×10−78 (CA1.)
SWR and beta modulation in PFC and CA1.
a. Four example PFC cells showing SWR aligned spiking raster and histogram, spike beta-phase locking histogram and goal entry aligned firing histogram. Top two cells show spiking excitation around SWRs, and bottom two cells show spiking inhibition around SWRs. In the spike beta-phase locking histogram, the distribution is duplicated to improve visualization. The peaks and troughs for the beta cycle is shown in white. For the goal entry aligned firing histogram, the median firing density is shown by the vertical black line. This corresponds to 50% of the firing density relative to goal entry. b. PFC beta modulation strength (kappa) versus SWR modulation index. The four combinations of SWR (S) and beta (β) modulation (+: modulated and -: not modulated) and their corresponding proportions are indicated. c. PFC SWR modulation index versus beta phase locking preference for beta and SWR modulated cells (β+S+) (left) and SWR modulated but not beta modulated cells (β+S-) (right). The regression line is show in cyan. The gray shading indicates the peak and trough of beta. Two cycles are repeated for clarity. d. PFC median firing density from goal entry versus beta phase locking preference for beta and SWR modulated cells (β+S+) (left) and SWR modulated but not beta modulated cells (β+S-) (right). Median firing density is the time from goal entry corresponding 50% of the firing density within 5 seconds. e. Four example CA1 cells showing SWR aligned spiking raster and histogram, spike beta-phase locking histogram and goal entry aligned firing histogram. f. CA1 beta modulation strength (kappa) versus SWR modulation index. The four combinations of SWR and beta (β) modulation (+: modulated and -: not modulated) and their corresponding proportions are indicated. g. CA1 SWR modulation index versus beta phase locking preference for beta and SWR modulated cells (β+S+) (left) and SWR modulated but not beta modulated cells (β+S-) (right). The regression line is show in red. The gray shading indicates the peak and trough of beta. Two cycles are repeated for clarity. h. CA1 median firing density from goal entry versus beta phase locking preference for beta and SWR modulated cells (β+S+) (left) and SWR modulated but not beta modulated cells (β+S-) (right). Median firing density is the time from goal entry corresponding 50% of the firing density within 5 seconds.
Permutation test for the significance of the correlation coefficient.
a. Distribution of the permuted (gray) and observed (vertical line) for PFC SWR modulation index versus beta phase locking preference in Fig. 8C. We performed this additional permutation test to verify the observed value of the correlation coefficient is different to the chance distribution. b. Distribution of the permuted (gray) and observed (vertical line) for PFC median firing density from goal entry versus beta phase locking preference in Fig. 8D. c. Distribution of the permuted (gray) and observed (vertical line) for CA1 SWR modulation index versus beta phase locking preference in Fig. 8G. d. Distribution of the permuted (gray) and observed (vertical line) for CA1 median firing density from goal entry versus beta phase locking preference in Fig. 8H.
