Causal Evidence for a Role of Cerebellar Lobulus Simplex in Prefrontal-Hippocampal Interaction in Spatial Working Memory Decision-Making

Abstract

Spatial working memory (SWM) is a cerebrocerebellar cognitive skill supporting survival-relevant behaviors, such as optimizing foraging behavior by remembering recent routes and visited sites. It is known that SWM decision-making in rodents requires the medial prefrontal cortex (mPFC) and dorsal hippocampus. The decision process in SWM tasks carries a specific electrophysiological signature of a brief, decision-related increase in neuronal communication in the form of an increase in the coherence of neuronal theta oscillations (4-12 Hz) between the mPFC and dorsal hippocampus, a finding we replicated here during spontaneous exploration of a plus maze in freely moving mice. We further evaluated SWM decision-related coherence changes within frequency bands above theta. Decision-related coherence increases occurred in seven frequency bands between 4 and 200 Hz and decision-outcome-related differences in coherence modulation occurred within the beta and gamma frequency bands and in higher frequency oscillations up to 130 Hz. With recent evidence that Purkinje cells in the cerebellar lobulus simplex (LS) represent information about the phase and phase differences of gamma oscillations in the mPFC and dorsal hippocampus, we hypothesized that LS might be involved in the modulation of mPFC-hippocampal gamma coherence. We show that optical stimulation of LS significantly impairs SWM performance and decision-related mPFC-dCA1 coherence modulation, providing causal evidence for an involvement of cerebellar LS in SWM decision-making at the behavioral and neuronal level. Our findings suggest that the cerebellum might contribute to SWM decision-making by optimizing the decision-related modulation of mPFC-dCA1 coherence.

Keywords: Cerebellum; Cognition; Coherence; Decision-making; Electrophysiology; Hippocampus; Medial prefrontal cortex; Neuronal communication; Optogenetics; Prefrontal cortex; Purkinje cell; Spatial working memory.

Fig. 1

Evaluation of plus-maze exploration behavior, spatial working memory, and classification of individual decisions. a Mice implanted with recording electrodes in the mPFC and dCA1 and an optical fiber above cerebellar LS were placed in a plus maze and allowed to freely explore for 12 min. Movement trajectories were video-captured and digitized for offline analysis. b The sequence of arm entries was analyzed to detect any sequence of four entries without a repeat (i.e., spontaneous alternations). In the example shown, arm entries that are part of a spontaneous alternation are printed in green font. A total of five spontaneous alternations are marked with grey brackets labeled SA1–5. Note that spontaneous alternations can be overlapping (SA2–5). The first occurrence of a spontaneous alternation in the example is a single sequence of four entries (CDAB, SA1). The second occurrence (CBADCBA) contains four overlapping spontaneous alternations (CBAD, BADC, ADCB, DCBA, SA2–5). c The same sequence as in (b) but with individual arm entry decisions classified as correct (green arrows) or incorrect (red arrows) based on the position of the decision in a sequence of four entries. For the analysis of decision-outcome–related neuronal activity only decisions that were preceded by three choices without repetition were classified as correct. As a consequence, the first two or three decisions of a spontaneous alternation (CBAD) were classified as incorrect depending on the choice immediately preceding the spontaneous alternation

Fig. 2

Illustration of recording locations, example data, and lesion sites. a Schematic drawing of the top view of a mouse brain with recording locations in the left mPFC and dCA1 marked with blue and red circles, respectively. The illustration above the cerebellum depicts an optical fiber coupled to an LED (465 nm) that was mounted over LS. Dashed lines indicate the approximate locations of coronal sections shown in panel (c). b Top traces are LFP signals recorded in the mPFC (blue) and dCA1 (red). The left bottom panel shows an enlarged sections of the dCA1 LFP that includes sharp-wave ripple activity indicated by black arrows. The bottom right panel shows a band pass filtered (130–200 Hz) version of the LFP on the left, highlighting the two sharp-wave ripple events. c Examples of electrolytic lesions (arrows) at recording sites in Nissl-stained sections of the mPFC and dCA1 region and a coronal section of cerebellar LS showing YFP expression in PCs of the L7-ChR2-GFP mice

Fig. 3

Optogenetic activation of PCs in the LS impairs SWM performance and decision-outcome–related modulation of LFP activity. a During control trials mice freely explored the maze for 12 min without optical stimulation (left panel). During trials with LS optical stimulation (right panel), LS PCs were photoactivated for 1 s at the time the mouse reached the center of the maze. Photoactivation was started manually when the mouse’s head reached the area marked in blue. b Compared to control trials, the percentage of spontaneous alternations was significantly reduced in LS stimulation trials (**p = 0.0012; two-way ANOVA). c Average time courses of decision-related LFP potentials in the dCA1 and mPFC for correct (green traces) and incorrect (red traces) decisions. Time zero in each plot corresponds to the time the mouse left the maze center with all four paws. Average peri-decision LFP activities in both dCA1 and mPFC reached values at around − 0.4 s that were higher during incorrect than correct decisions. Red triangles mark peak LFP values for incorrect decisions that exceeded baseline values calculated by averaging pre-decision (− 1.5 to − 1.25 s) values (p < 0.05; Wilcoxon signed-rank test). Green and red circles mark LFP minima that fall below pre-decision average for correct and incorrect decisions, respectively (p < 0.05; Wilcoxon signed-rank test). Gray-shaded rectangles mark time periods with significant differences between the traces (*p < 0.05; Wilcoxon signed-rank test). d As in (c) but for LS stimulation trials. LS stimulation eliminated the decision-outcome–related differences in dCA1 and mPFC LFP activity that preceded decision completion (~ − 0.4 s) during control trials

Fig. 4

Optical stimulation of LS Purkinje cells alters mPFC-dCA baseline coherence across different frequency bands. The baseline coherence was calculated for each frequency band by averaging time-resolved mPFC-dCA1 coherence during the initial 0.5 s after mice entered the center area with all four paws. LS Purkinje cell optical stimulation reduced baseline coherence for LFP oscillations in the theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), low-gamma (30–60 Hz), and mid-gamma (60–80 Hz) bands and increased baseline coherence for oscillations in the high-frequency (100–130 Hz), and very high-frequency (130–200 Hz) bands. LS Purkinje cell stimulation had no effect on baseline coherence in the high-gamma (80–100 Hz)-frequency band. Two-sample t tests were used to compare the differences between control and LS stimulation trials for each frequency band. Data were expressed as mean ± standard error. **p < 0.01; ***p < 0.001

Fig. 5

Optical stimulation of LS Purkinje cells disrupts decision-related mPFC-dCA coherence modulation. a Time-resolved peri-decision coherence averages for correct (green traces) and incorrect (red traces) decisions for control trials. Peri-decision coherence was analyzed for eight separate frequency bands between 4 and 200 Hz with time 0 making the completion of the decision process defined by the mouse entering the next maze arm with all four paws. Baseline coherence values for each mouse were subtracted from peri-decision coherence functions before averaging. Green and red triangles mark coherence peaks for correct and incorrect decisions, respectively, that exceed baseline coherence (Wilcoxon signed-rank test p < 0.05). Red circles mark coherence minima falling below baseline coherence values for incorrect decisions (Wilcoxon signed-rank test; p < 0.05). No significant coherence minima were found for correct decisions. Decision outcome was reflected in differences in time course of coherence functions. Peri-decision times where coherence functions for correct and incorrect decisions differed are marked by gray-shaded rectangles (two-sample t test, *p < 0.05). b As in (a) but for LS stimulation trials. LS stimulation eliminated all differences between peri-decision coherence functions for correct and incorrect decisions