Figure 4. Cerebello-hippocampal interactions during goal-directed behavior.

(A) Mice learned to traverse a 1 m linear track to receive a medial forebrain bundle stimulation (lightening symbols) upon reaching invisible goal zones (n = 8 mice). Representative trajectories from early (session 1) and late (session 20) training show the transition from exploratory to goal-directed behavior. (B) Mice improved their performance in the task across days as shown by increases in the mean number of rewards obtained (average of the three sessions per day, repeated measures Friedman test with FDR correction, Friedman statistic = 37.91, p < 0.0001; solid line, day 1 vs days 4–7, p < 0.01) and mean speed (average of the three sessions per day, repeated measures Friedman test with FDR correction, Friedman statistic = 36.32, p < 0.0001; solid line, day 1 vs days 4–7, p < 0.05). (C) Overall cerebello-hippocampal theta coherence per day (average of the three sessions per day) during learning of the linear track task (when both hippocampal recording electrodes were on target we averaged the coherence obtained with left and right hemispheres; Crus I, n = 4; lobule II/III, n = 6; lobule VI, n = 6). Hippocampus-Crus I coherence increased significantly compared with first day (day of training x cerebello-hippocampal combination two-way ANOVA with FDR correction, day effect F_6,84 = 3.873, p = 0.0018; solid line, day 1 vs days 6–7, p < 0.01). (D) i, Top: Mean speed aligned by distance from the reward location (position 0) averaged across runs during session 1. Bottom: Mean power spectrogram aligned by distance from reward and averaged across runs during session one for hippocampus LFP (n = 8, mean between left and right hemisphere LFPs when both hippocampal recordings were on target) and cerebellar cortical regions (Crus I, n = 4; lobule II/III, n = 6; lobule VI, n = 6). ii, Mean coherogram aligned by distance from reward location (position 0) averaged across runs during session one for each hippocampal-cerebellar combination (when both hippocampal recording electrodes were on target we averaged the coherence obtained with left and right; Crus I, n = 4; lobule II/III, n = 6; lobule VI, n = 6). iii, Mean theta coherence aligned by distance from reward and averaged across runs during session one for each hippocampal-cerebellar combination (mean between coherence with left and right hemisphere LFPs when both recordings were on target; Crus I, n = 4; lobule II/III, n = 6; lobule VI, n = 6). No significant differences between hippocampal-cerebellar combination or distances from reward were observed, but a significant interaction effect was obtained (distance x combination two-ways repeated measures ANOVA with FDR correction; combination effect, F_2,13 = 2.33, p = 0.1365; distance effect, F_84,1092 = 1.043, p = 0.3772; interaction effect, F_168,1092 = 1.332, p = 0.0053). (E) Same as D for session 20. Significant differences were observed between hippocampal-cerebellar combinations (distance x combination two-ways repeated measures ANOVA with FDR correction; combination effect, F_2,13 = 6.145, p = 0.0132; distance effect, F_84,1092 = 1.682, p = 0.0002; interaction effect, F_168,1092 = 1.271, p = 0.0163). Post-hoc analysis revealed sustained (at least for five consecutive cm) differences between Crus I-HPC and lobule II/III-HPC coherence at distances from −60 to −20 cm from reward (solid green/black line), between lobule VI-HPC and lobule II/III-HPC coherence between −44 and −31 cm from reward (solid red/black line) and also between Crus I-HPC and lobule VI-HPC coherence between −59 and −36 cm from reward (solid green/red line). (F) Top: Averaged LFP power between −60 and −20 cm from reward (session 1). Bottom: Averaged coherence between −60 and −20 cm from reward (session 1). The spurious peak in the 49–51 Hz band generated for the electrical noise has been removed. (G) Same as J for session 20. (H) Averaged theta coherence between −60 and −20 cm from reward between cerebellar recordings and hippocampus during session 1 (left) and session 20 (right, coherence averaged between left and right hemisphere LFPs when both hippocampal recording electrodes were on target; Crus I, n = 4; lobule II/III, n = 6; lobule VI, n = 6). Crus I-HPC coherence was significantly higher than that observed with lobule II/III in the session 20 (*, Kruskal-Wallis with FDR correction, Kruskal-Wallis statistic = 7.989, p = 0.0103; Crus I-HPC vs lobule II/III-HPC, p = 0.0110).

Figure 4—figure supplement 1. Cerebello-hippocampal coherence patterns are conserved across hemispheres during goal-directed behavior.

(A) Averaged z-score LFP power from HPC left (light blue, n = 6) and HPC right (dark blue, n = 7) from Figure 4J (session 1). Inset: No significant difference in the mean θ power was observed between both hemispheres (Mann-Whitney test, U = 18, p = 0.7308). (B) Same as A but for session 20 (related to Figure 4K). Inset: No significant difference in mean LFP theta power was observed between hemispheres (Mann-Whitney test, U = 20, p = 0.9452). (C) i: Averaged coherence between HPC left and cerebellar regions (color coded) from Figure 4J (Crus I, n = 3; lobule II/III = 4; lobule VI, n = 6). ii: Same for coherence with HPC right (Crus I, n = 3; lobule II/III = 5; lobule VI, n = 5). (D) Same as C for session 20 (related to Figure 4K). (E) Comparisons between cerebello-HPC left (light blue) and cerebello-HPC right (dark blue) theta coherence from panel C (Session 1). No significant differences were observed for any cerebello-hippocampal combination (Crus I, Mann-Whitney test, U = 4, p > 0.99; lobule II/III, Mann-Whitney test, U = 10, p = 0.7302; lobule VI, Mann-Whitney test, U = 10, p = 0.4286). (F) Same as E for session 20 (panel D). No significant differences were observed for any cerebello-hippocampal combination (Crus I, Mann-Whitney test, U = 3, p = 0.7; lobule II/III, Mann-Whitney test, U = 4, p > 0.99; lobule VI, Mann-Whitney test, U = 12, p = 0.6623).

Figure 4—figure supplement 2. Distributions and correlations during goal-directed behavior.

(A) Probability distribution of the instantaneous LFP theta power for each of the recorded regions during session one in the linear track. Hippocampal theta power followed a negatively skewed distribution while all cerebellar LFP theta power followed a positively skewed distribution. (B) Correlation between the instantaneous LFP theta power for each of the cerebellar regions and instantaneous speed during session one in the linear track. Gray-shaded bars represent the confidence levels obtained from bootstrapped data with α = 0.05. Cerebellar theta power was significantly positively correlated with speed across all recorded regions (Crus I, n = 4, median rho Spearman = 0.2125; lobule II/III, n = 6, median rho Spearman = 0.2391; lobule VI, n = 6, median rho Spearman = 0.2353). (C) Same as A for session 20. (D) Same as B for session 20. Cerebellar theta power was significantly positively correlated with speed across all recorded regions (Crus I, n = 4, median rho Spearman = 0.2349; lobule II/III, n = 6, median rho Spearman = 0.3097; lobule VI, n = 6, median rho Spearman = 0.1941). (E) Probability distribution of instantaneous hippocampal-cerebellar theta coherence (color coded) and instantaneous speed during session one in the linear track. All combinations followed gaussian distributions. (F) Correlation between the instantaneous hippocampal-cerebellar theta coherence (color coded) and instantaneous speed during session one in the linear track. Gray-shaded bars represent the confidence levels obtained from bootstrapped data with α = 0.05. Hippocampal-cerebellar theta coherence was not significantly correlated with speed across any of the recorded combinations. (G) Same as E for session 20. (H) Same as F for session 20. Lobule II/III-HPC theta coherence was significantly negatively correlated with speed (n = 6, median rho Spearman = −0.0859) and lobule VI-HPC theta coherence was significantly positively correlated with speed (n = 6, median rho Spearman = 0.0911). Crus I-HPC theta coherence was not correlated with speed (Spearman rho = 0.024).

Figure 4—figure supplement 3. Calculation of the imaginary part of coherence during goal-directed behavior.

(A) Mean coherogram plotted using the imaginary part of coherence, aligned by distance from reward and averaged across runs during session one for each hippocampal-cerebellar combination (mean coherence between left and right hemisphere LFPs when both hippocampal recording electrodes were on target; Crus I, n = 4; lobule II/III, n = 6; lobule VI, n = 6; related to Figure 4E). (B) Same as A but for session 20 (related to Figure 4H). (C) Mean imaginary theta coherence aligned by distance from reward averaged across runs during session one for each hippocampal-cerebellar combination (mean coherence between left and right hemisphere LFPs when both recording electrodes were on target; Crus I-HPC, n = 4; lobule II/III-HPC, n = 6; lobule VI-HPC, n = 6). No significant differences between hippocampal-cerebellar combinations was observed (distance x combination two-ways repeated measures ANOVA with FDR correction; combination effect, F_2,13 = 1.065, p = 0.3730; interaction effect, F_168,1092 = 0.6713, p = 0.9993). (D) Same as C for session 20. A significant interaction between hippocampal-cerebellar combination and distance from reward was observed (distance x combination two-ways repeated measures ANOVA with FDR correction; interaction effect, F_168,1092 = 1.588, p < 0.0001). Post-hoc analysis revealed sustained differences between Crus I-HPC and lobule II/III-HPC coherence at distances from −42 to −40 and from −34 to −23 cm from reward (solid green/black line) and between Crus I-HPC and lobule VI-HPC coherence from −33 to −23 cm (solid green/red line). (E) Imaginary coherence from C (session 1) averaged between −42 and −23 cm from reward. (F) Imaginary coherence from D (session 20) averaged between −42 and −23 cm from reward.

Figure 4—figure supplement 4. Cerebello-hippocampal coherence patterns during running or goal-directed movement in a virtual environment.

(A) Schematic of the virtual reality system and recording setup. Head-fixed mice were trained to move an air-cushioned Styrofoam ball in order to navigate through a virtual environment displayed on six TFT monitors surrounding the animal. (B) Example recording of the virtual position as a mouse traversed a virtual linear track to receive rewards (MFB stimulation indicated by a lightning symbol, n = 6). (C) Behavioral performance across training as illustrated by the mean number of rewards. Each mouse represented by a different line color. Two mice, ME21 and ME11, showed an increase in the number of rewards obtained over training days. (D) i, speed (top) and spectrograms (bottom; HPC, n = 6; Crus I, n = 2; lobule II/III, n = 5; lobule VI, n = 3) averaged by distance to reward from a session in early training day 2 (session 5); ii, coherograms for non learning mice averaged by distance to reward during early training day 2 (Crus I-HPC, n = 2; lobule II/III-HPC, n = 5; lobule VI-HPC, n = 3). (E) same as D for late training day 5 (session 14). (F) i, power spectra (upper) and coherence (lower) for non-learner mice during early training day 2 pooled from −60 to −20 cm from reward. ii, power spectra (upper) and coherence (lower) for non-learner mice during late training day 5 pooled from −60 to −20 cm from reward. (G-I). Same as E-F but for the two mice that showed an increase in the number of rewards obtained over training (ME21 and ME11, Crus I, n = 1; lobule II/III, n = 1; lobule VI, n = 1).