1. Neuroscience
Download icon

Anatomical and physiological foundations of cerebello-hippocampal interaction

  1. Thomas Charles Watson
  2. Pauline Obiang
  3. Arturo Torres-Herraez
  4. Aurélie Watilliaux
  5. Patrice Coulon
  6. Christelle Rochefort
  7. Laure Rondi-Reig  Is a corresponding author
  1. CNRS UMR 8246, INSERM, UMR-S 1130, Sorbonne Universités, University Pierre and Marie Curie, France
  2. CNRS and Aix Marseille Université, France
Research Article
  • Cited 3
  • Views 1,115
  • Annotations
Cite this article as: eLife 2019;8:e41896 doi: 10.7554/eLife.41896

Abstract

Multiple lines of evidence suggest that functionally intact cerebello-hippocampal interactions are required for appropriate spatial processing. However, how the cerebellum anatomically and physiologically engages with the hippocampus to sustain such communication remains unknown. Using rabies virus as a retrograde transneuronal tracer in mice, we reveal that the dorsal hippocampus receives input from topographically restricted and disparate regions of the cerebellum. By simultaneously recording local field potential from both the dorsal hippocampus and anatomically connected cerebellar regions, we additionally suggest that the two structures interact, in a behaviorally dynamic manner, through subregion-specific synchronization of neuronal oscillations in the 6–12 Hz frequency range. Together, these results reveal a novel neural network macro-architecture through which we can understand how a brain region classically associated with motor control, the cerebellum, may influence hippocampal neuronal activity and related functions, such as spatial navigation.

https://doi.org/10.7554/eLife.41896.001

Introduction

The cerebellum is classically associated with motor control. However, accumulating evidence suggests its functions may extend to cognitive processes including navigation (Petrosini et al., 1998; Burguière et al., 2005; Rondi-Reig and Burguière, 2005; Buckner, 2013; Koziol et al., 2014; Stoodley et al., 2017). Indeed, anatomical and functional connectivity has been described between cerebellum and cortical areas that are engaged in cognitive tasks (Kim et al., 1994; Kelly and Strick, 2003; Ramnani, 2006; Watson et al., 2009; Watson et al., 2014; Stoodley and Schmahmann, 2010). Furthermore, the cerebellum has recently been found to form functional networks with subcortical structures associated with higher order functions, such as the basal ganglia (Kelly and Strick, 2004; Hoshi et al., 2005; Bostan et al., 2010; Chen et al., 2014; see Bostan and Strick, 2018 for review), ventral tegmental area (Rogers et al., 2011; Carta et al., 2019) and hippocampus (Rochefort et al., 2013; Iglói et al., 2015; Yu and Krook-Magnuson, 2015; Babayan et al., 2017).

In the hippocampus, spontaneous local field potential (LFP) activity (Iwata and Snider, 1959; Babb et al., 1974; Snider and Maiti, 1975; Krook-Magnuson et al., 2014) and place cell properties (Rochefort et al., 2011; Lefort et al., 2019) are profoundly modulated following cerebellar manipulation (Rondi-Reig et al., 2014 for review). A recent study has also described, at the single cell and blood-oxygen-level-dependent signal level, sustained activation in the dorsal hippocampus during optogenetic enhancement of cerebellar nuclei output in head-fixed mice (Choe et al., 2018). These data point toward the existence of an anatomical projection from the cerebellum to the hippocampus. However, the direct or indirect nature of the connection between the two structures remains unclear. The suggestion of a direct connection between these two structures has been claimed by a recent tractography study in humans (Arrigo et al., 2014) and the presence of short-latency evoked field potentials (2–4 ms) in cat and rat hippocampi after electrical stimulation of the cerebellar vermal and paravermal regions (Whiteside and Snider, 1953; Harper and Heath, 1973; Snider and Maiti, 1976; Heath et al., 1978; Newman and Reza, 1979). However, secondary hippocampal field responses have also been described, at a latency of 12–15 ms following cerebellar stimulation, suggesting the existence of an indirect pathway (Whiteside and Snider, 1953).

Taken together, even if these studies provide compelling physiological evidence of cerebellar influences on the hippocampus, they do not provide direct evidence of neuroanatomical connectivity between the two regions. Given the known complex, modular functional and anatomical organization of the cerebellum (Apps and Hawkes, 2009) this represents a major gap in our understanding of the network architecture linking the two structures. In addition, these studies provide no measure of dynamic physiological communication or associated synchronization between the two structures, which is thought to be essential for maintaining distributed network functions (e.g. Fries, 2005).

Therefore, this study addresses two fundamental, unanswered questions: which regions of the cerebellum are anatomically connected to the hippocampus and what are the spatio-temporal dynamics of synchronized cerebello-hippocampal activity during behavior? To address these unresolved questions, we used rabies virus as a retrograde transneuronal tracer to determine the extent and topographic organization of cerebellar input to the hippocampus. Based upon the anatomical tracing results, we then studied the levels of synchronization between LFPs recorded simultaneously from the cerebellum and dorsal hippocampus in freely moving mice as a proxy for potential cross-structure interactions. We reveal that specific cerebellar modules are anatomically connected to the hippocampus and that these inter-connected regions dynamically synchronized their activity during behavior.

Results

To study the topographical organization of ascending, cerebello-hippocampal projections, we unilaterally injected rabies virus (RABV), together with fluorescent cholera toxin β-subunit (CTb), into the left hippocampus. The extent of the injection site, identified by the presence of fluorescent CTb, included both CA1 (stratum lacunosum-moleculare) and DG (molecular layer, granule cell layer and hilus, Figure 1—figure supplement 1).

Cerebellar projections to the hippocampus are precisely topographically organized

We characterized the presence of retrograde transneuronally RABV-infected neurons after survival times of 30, 48, 58 and 66 hr (Suzuki et al., 2012; Aoki et al., 2017Coulon et al., 2011). At 30 hr post-infection, staining was found predominantly in the medial septum diagonal band of Broca (MSDB), the lateral and medial entorhinal cortices, the perirhinal cortex and the supramammillary nucleus of the hypothalamus (SUM). Sparse staining was also found in a subset of mice, in the raphe nucleus and in other hypothalamic and thalamic regions (Figure 1—figure supplement 2). All these structures correspond to first-order relays as CTb staining was also systematically associated with each of the sites containing rabies infected neurons (Figure 1—figure supplement 3). Importantly, no RABV or CTb labeling was found in the cerebellum at this time-course ruling out the existence of a direct cerebello-hippocampal DG pathway in mice. Interestingly, among these putative first relays, the septum, the hypothalamus and the raphe nucleus receive direct projections from the deep cerebellar nuclei in cat (Paul et al., 1973), rat (Teune et al., 2000) and Macaca (Haines et al., 1990). In accordance with this, some RABV-infected neurons were observed in the deep cerebellar and vestibular nuclei 48 hr post-hippocampal infection and some labeled cells were found in the cerebellar cortex at 58 hr p.i. (Figure 1A, inset, Figure 1—figure supplement 4) suggesting the existence of single-relay pathways (i.e. di-synaptic connections) between the cerebellum and hippocampus.

Figure 1 with 5 supplements see all
Topographically restricted regions of cerebellar cortex are connected to the hippocampus.

(A) Left, mean number of labeled cells in the deep cerebellar nuclei (DCN), vestibular nuclei (VN) and cerebellar cortex at different survival times following rabies injection in left hippocampal dentate gyrus. Box shows a magnification of the labeling at 48 and 58 hr (n = 5 mice for 58 hr and 66 hr, 30 hr, n = 4 mice and 48 hr, n = 3 mice). Middle, schematic representation of rabies injection site and survival times required to reach the cerebellar and vestibular nuclei (58 hr, dashed blue line), and cerebellar cortex (66 hr, orange line). Upper right, schematic view of the posterior cerebellar cortex indicating regions of highest labeling following rabies virus injection (red, vermis lobule VI; green, Crus I; gray, paraflocculus). (B-E) Representative photomicrographs showing labeling in the cerebellar and vestibular nuclei 58 hr post-infection. Left panels show low-magnification view, right panels show magnified view of area indicated by dashed box. Solid arrow heads indicate the presence of the very few labeled cells in the IntP. (F) Pooled, normalized counts of rabies labeled cells in the ipsi- and contralateral cerebellar and vestibular nuclei 58 hr post-infection (n = 5 mice). No significant differences were found between ipsi- and contralateral nuclei (nuclei x hemisphere two-way ANOVA, hemisphere effect F (1, 4)=1.31×10−5, p=0.99, interaction effect F (3, 12)=2.79, p=0.09, nuclei effect F (3, 12)=9.38, p=0.002). (G) Normalized cell counts in the fastigial nucleus (left) and dentate nucleus (right), according to their rostro-caudal position relative to bregma. Open circles, contralateral count; filled circles, ipsilateral count (n = 5 mice). (H-K) Representative photomicrographs of the resultant labeling in lobule VI, Crus I, paraflocculus and lobule II at 66 hr post-infection. (L) Normalized count of rabies labeled cells in anterior (black bar; lobule II to lobule IV/V); central (dark gray bar; lobule VI to Crus II); posterior (clear bar; lobule VIII and lobule IX) and flocculonodular (Floc. Nod., light gray bar; lobule X, flocculus and paraflocculus) cerebellum 66 hr post-infection (n = 5 mice; one-way ANOVA with FDR correction, F (3, 16)=19.11, p<0.0001). (M) Normalized cell count of rabies labeled cells in all assessed lobules 66 hr post-infection. Color coding of bars indicate assignment of lobules to either anterior, central, posterior or vestibular cerebellum as indicated in L. Abbreviations: Dent., Dentate nucleus; Fast., fastigial nucleus; Fast. DL, dorsolateral fastigial nucleus; Floc. Nod., flocculonodular lobe; Interp., nucleus interpositus; IntA, nucleus interpositus anterior; IntDL, dorsolateral nucleus interpositus; IntP, nucleus interpositus posterior; i-Sim, ipsilateral simplex lobule; c-Sim, contralateral simplex lobule; i-Crus I, ipsilateral Crus I; c-Crus I, contralateral Crus I; i-Crus II, ipsilateral Crus II; c-Crus II, contralateral Crus II; i-Par, ipsilateral paramedian lobule; c-Par, contralateral paramedian lobule; i-CP, ipsilateral copula; c-CP, contralateral copula; i-Floc, ipsilateral flocculus; c-Floc, contralateral flocculus; i-PF, ipsilateral paraflocculus; c-PF, contralateral paraflocculus; Vestib., vestibular nuclei. ** q < 0.01, *** q < 0.001.

https://doi.org/10.7554/eLife.41896.002

At 48 hr post-infection, RABV labeling was also found in mammillary bodies, amygdala and several midbrain and pontine regions such as the periaqueductal gray (PAG), the nucleus incertus and the laterodorsal tegmental nucleus (LtDG). These regions, which are known to receive direct projections from the DCN (Teune et al., 2000) were neither stained with CTb nor RABV-positive at 30 hr post-infection and therefore represent a putative additional pathway involving two relays between the cerebellum and the hippocampus (Figure 1—figure supplement 4).

58 hr post-infection, abundant and strong staining was found in the fastigial and dentate nuclei, with almost no staining in the interpositus (Figure 1B–F). Within the fastigial and dentate, RABV-labeled cells were found to be topographically restricted to caudal and central regions, respectively (Figure 1G).

Following 66 hr of incubation, the number of strongly labeled cells increased in the DCN and vestibular nuclei (Figure 1A); however, the topographical distribution remained unchanged (Figure 1—figure supplement 5). At the level of the cerebellar cortex, longitudinal clusters of RABV+ Purkinje cells (PCs) were found bilaterally across highly restricted central and flocculo-nodular regions (Figure 1L). In the central cerebellum (i.e. lobules VI, VII and associated hemispheres), clusters were particularly concentrated in lobule VI and Crus I (Figure 1H,I,M). In the flocculo-nodular cerebellum, RABV-labeled cells were found in the dorsal and ventral paraflocculus (Figure 1J and M and Figure 2B). Within the vermis we identified a single cluster of RABV+ Purkinje cells that extended across both lobule VIa and lobule VIb-c (Figure 2). In contrast, within Crus I, RABV-labeled Purkinje cells were arranged in two spatially isolated clusters, one located rostro-laterally and the other caudo-medially (Figure 2).

Different cerebellar modules project to the hippocampus.

(A) 3-D reconstruction showing the location of RABV+ Purkinje cells in the most labeled cerebellar lobules at 66 hr post-infection. Red, blue and green dots represent RABV+ Purkinje cells in lobule VI, Crus I and paraflocculus, respectively. (B) Photomicrographs from case S18 showing double staining against zebrin II (green, left column), RABV (red, central column) and merge (right column) in lobule VI (i), Crus I (ii and iii) and paraflocculus (iv). RABV+ Purkinje cells were also zebrin positive and were organized in clusters of strongly labeled RABV+ cells (filled arrow-heads) surrounded by weakly labeled RABV+ Purkinje cells (unfilled arrow-heads). (C) Assignment of the RABV+ clusters to specific cerebellar modules for case S18 in the anterior (AZ; left), central (CZ; central column) and posterior (PZ; right column) zones. First row shows stacked sections with zebrin-positive Purkinje cells (white dots) and RABV+ Purkinje cells, which were also zebrin positive (purple dots, strong and weakly labeled cells included); central row shows reconstructed principal zebrin bands (delineated by yellow dashed lines and named from 1+ to 7+; nomenclature from Sugihara and Quy, 2007) and cerebellar modules (capital letters; defined as in Sugihara and Quy, 2007); and bottom row shows the location of the RABV+/zebrin Purkinje cells (purple dots) in relation to reconstructed zebrin bands and modules. Abbreviations, I, lobule I; III, lobule III; IV/V, lobule IV/V; VIa and VI b-c, lobule VIa and VI b-c, respectively; IX, lobule IX; X, lobule X; Sim, lobule simplex; Cr I, Crus I; Cr II, Crus II; Par, paramedian lobule; CP, copula, PFL, paraflocculus, FL, flocculus.; dPFC and vPFC, dorsal and ventral paraflocculus, respectively.

https://doi.org/10.7554/eLife.41896.009

The topographical arrangement of RABV-labeled PCs in longitudinal clusters is in keeping with the well-described modular organization of the cerebellum (e.g. Apps and Hawkes, 2009). Mapping of molecular marker expression patterns, such as zebrin II banding, provides a reliable basis from which modules can be defined and recognized in the cerebellar cortex of rodents. Thus, to further assign the observed PC clusters to previously described cerebellar zones, we used a double immunohistochemical approach to stain for both RABV and aldolase C (zebrin II) in one animal (case S18) after 66 hr of infection (Figure 2B) (Brochu et al., 1990; Sugihara and Shinoda, 2004). Lobule VI, Crus I and paraflocculus are mostly zebrin-positive regions (Sugihara and Quy, 2007), and we found that RABV-labeled Purkinje cells co-localized with zebrin II in all the observed clusters (Figure 2B). In the vermis, lobule VIa RABV-labeled PCs were mostly located in the a+ band. The few RABV-labeled cells found in lobule VII were confined to the 2+ band. Thus, together, these labeled cells belong to the a+//2+ pair that constitutes part of the cerebellar A module (Figure 2C) (Sugihara, 2011). In Crus I, the rostrolateral cluster of RABV-labeled PCs was aligned with the anterior 6+ zebrin band corresponding to module D2. The caudomedial cluster was in continuation with the posterior 5+ zebrin band suggesting that it is part of the paravermal module C2 (Figure 2C). In the paraflocculus, the assignment of the RABV-labeled cells to specific modules was not addressed given the complex morphology of this region. However, the presence of RABV-labeled cells both in the dorsal and ventral paraflocculus suggests the involvement of more than one module (Figure 2B–C) (Voogd and Barmack, 2006).

Cerebellar modules are also defined by their outputs through the deep cerebellar and vestibular nuclei (Apps and Hawkes, 2009; Ruigrok, 2011). The presence of RABV-labeled cells in the fastigial nucleus is consistent with the involvement of module A. Similarly, the D2 module is routed through the dentate nuclei in which we found robust RABV labeling. We also found RABV+ cells in the nucleus interpositus posterior, which provides the output of module C2. Finally, RABV labeling was observed in the vestibular nuclei, which may represent the output of RABV+ Purkinje cell clusters observed in the ventral paraflocculus.

Together, our neuroanatomical tracing data indicate that cerebellar projections to the hippocampus emanate from three distinct cerebellar modules. It also suggests the existence of multiple, convergent pathways from the DCN to the hippocampus.

Cerebello-hippocampal physiological interactions in a familiar home-cage environment

In order to question the potential functional relevance of cerebello-hippocampal anatomical connectivity, we implanted mice (n = 21) with arrays of bipolar LFP recording electrodes in bilateral dorsal hippocampus (HPC) and unilaterally in two highly RABV-labeled regions of the central cerebellum, lobule VI (midline) and Crus I (left hemisphere). For comparison, we also simultaneously recorded LFP from cerebellar regions with minimal RABV labeling (lobule II or lobule III; Figure 1M; Figure 3A and B). Data were excluded from further analysis in cases where postmortem histological inspection revealed that electrode positions were off-target (Figure 3—figure supplement 1).

Figure 3 with 4 supplements see all
Assessment of cerebello-hippocampal interactions during active movement in the homecage.

(A) Schematic representation of recording and stimulation electrode implant positions. (B) Representative simultaneous raw LFP recordings (colored lines) and LFP filtered in the theta frequency range (θ, 6–12 Hz, overlaid gray lines) recorded during active movement in the homecage condition as defined by instantaneous speed. Solid magenta line indicates data epochs in which speed was above the required threshold for inclusion in further analysis (3 cm/s, dashed line (20 mice). In one mouse in which speed data was not available we used neck electromyograph (EMG) data to define periods of active movement). (C) Averaged z-score of the power spectra from hippocampal LFP (n = 21, mean between left and right hemisphere LFPs when both recording electrodes were on target) and cerebellar cortical regions Crus I (n = 12), lobule II/III (n = 13) and lobule VI (n = 16) during homecage exploration (speed above 3 cm/s). (D) Probability distribution of the instantaneous z-scored theta power for each of the recorded regions. Hippocampal theta power followed a negatively skewed distribution while cerebellar theta power followed a positively skewed distribution. (E) Correlation between the instantaneous z-scored theta power for each of the cerebellar recorded regions and instantaneous speed. Horizontal line indicates mean. Gray-shaded bars represent the confidence levels obtained from bootstrapped data with α = 0.05. Theta power was not significantly correlated with speed in any of the cerebellar recordings (Crus I, n = 11; lobule II/III, n = 12; lobule VI, n = 15). (F) Averaged coherence between cerebellar cortical recordings (color coded) and hippocampus (when both hippocampal recording electrodes were on target we averaged the coherence obtained with left and right hemispheres; Crus I, n = 12; lobule II/III, n = 13; lobule VI, n = 16) during homecage exploration (speed above 3 cm/s). (G) Probability distribution of the instantaneous hippocampal-cerebellar theta coherence (color coded). All the recording combinations followed normal distributions. (H) Correlation between the instantaneous hippocampal-cerebellar theta coherence (color coded) and the instantaneous speed. Horizontal line indicates mean. Gray-shaded bars represent the confidence levels obtained from bootstrapped data with α = 0.05. HPC-cerebellar theta coherence was not significantly correlated with speed in any of the recording combinations. (I) Theta coherence between cerebellar recordings and hippocampus (average between coherence with left and right hemisphere LFPs when both recordings were on target; Crus I - HPC, n = 12; lobule II/III - HPC, n = 13; lobule VI - HPC, n = 16). Lobule VI-HPC coherence was significantly higher than that observed with lobule II/III (horizontal line indicates median coherence for each combination; *, Kruskal-Wallis with FDR correction, Kruskal-Wallis statistic = 6.75, p = 0.0342; lobule VI-HPC vs lobule II/III-HPC, p = 0.0246). J, Mean lobule VI-HPC theta coherence plotted against the estimated medio-lateral position of the recording electrode in lobule VI (red dots; n = 16, when both hippocampal recording electrodes were on target we averaged the coherence obtained with left and right hemispheres; Spearman rho = 0.5718, p = 0.0225). In gray, number of RABV+ cells counted across lobule VI 66 hr after injection in the left HPC as a function of medio-lateral position (0.2 mm bins; n = 5 mice). Shading indicates S.E.M. Abbreviations, LFP, local field potential; HPC, dorsal hippocampus; lob II/III, lobule II/III, lob VI, lobule VI, MFB stim, medial forebrain bundle stimulation.

https://doi.org/10.7554/eLife.41896.010

The spectral profile of cerebellar and hippocampal LFP activity was first assessed during active movement in a familiar home-cage environment (epochs with speed >3 cm/s; see Materials and methods). Within the HPC, a dominant 6–12 Hz theta oscillation was similarly observed in both hemispheres (Figure 3—figure supplement 2B; left HPC: mean 6–12 Hz z-score power = 4.17 ± 0.20, n = 17; right HPC: mean 6–12 Hz z-score power = 4.52 ± 0.16, n = 19; unpaired t test, t34 = 0.007, p=0.1731), thus, when LFPs from both hippocampi where available we averaged the spectral power from left and right hemispheres (Figure 3C, mean 6–12 Hz z-score power = 4.35 ± 0.14). In the cerebellar recordings, low-frequency oscillations in the delta (2–4 Hz) and high-frequency activity in the mid-high gamma (50–140 Hz) range were prevalent. A clear peak in the 6–12 Hz band was not detected in the power spectrum of any of these recordings, which had similar levels of 6–12 Hz z-score power (Figure 3C; Crus I: 1.57 ± 0.10, n = 12 mice; lobule II/III: 1.57 ± 0.13, n = 13 mice; lobule VI: 1.47 ± 0.07, n = 16 mice; one-way ANOVA, F (2, 38)=0.34, p=0.7131).

In order to explore any potential state-dependent organization of theta oscillations in our recordings, we computed the distributions of instantaneous power in this frequency range (Figure 3D). If the power of theta oscillations was organized in differential states, we would expect to observe a multimodal distribution; however, we found unimodal power distributions in both the hippocampal and cerebellar recordings. The former was negatively skewed suggesting that the peak observed in the hippocampal power spectra is representative of sustained theta activity in the analyzed epochs. In contrast, the cerebellar theta distribution profile was positively skewed suggesting that activity in this frequency range, although of lower power than in the hippocampus, is also sustained within the cerebellar cortical regions we recorded from.

Previous studies have described correlations between locomotion speed and purkinje cell discharge in cerebellar vermal lobules V and VI (Sauerbrei et al., 2015; Muzzu et al., 2018). Furthemore, cerebellar nuclei - prefrontal cortex theta coherence has also been found to increase during active locomotion compared to rest (Watson et al., 2014). Therefore, we next asked if running speed could be modulating theta power in the cerebellar cortex. To do so, we computed the correlation between instantaneous speed and instantaneous theta power in each cerebellar region; however, none were significantly correlated (Figure 3E; Crus I, n = 11, mean Spearman rho = - 0.02 ± 0.04, bootstrap confidence level = [- 0.09 0.08]; lobule II/III, mean Spearman rho = - 0.07 ± 0.03, bootstrap confidence level = [- 0.08 0.08]; lobule VI, mean Spearman rho = - 0.04 ± 0.03, bootstrap confidence level = [- 0.08 0.08]).

As an indicator of cross-structure interaction (Fries, 2005), we next calculated coherence between LFP recorded from the different cerebellar subregions and left or right HPC. We found no statistically significant influence of hippocampal laterality on the measured cerebello-hippocampal coherence (Figure 3—figure supplement 2; Crus I-HPC left, n = 11, Crus I-HPC right, n = 11, Mann-Whitney test, U = 59, p=0.9487; lobule II/III-HPC left, n = 11, lobule II/III-HPC right, n = 12, Mann-Whitney test, U = 63, p=0.8801; lobule VI-HPC left, n = 15, lobule VI – HPC right, n = 14, Mann-Whitney test, U = 105, p>0.99). Therefore, for further analysis, when both hippocampal recording electrodes were on target we first calculated coherence with the cerebellar LFP for each hemisphere then averaged the two. Thus, for each mouse we obtained one coherence value per cerebello-hippocampal recording combination. In cases in which only one of the hippocampal recording electrodes was on target, we excluded the off target recording in our calculations of cerebello-hippocampal coherence.

A clear peak in coherence was observed for all cerebello-hippocampal combinations in the theta frequency range (6–12 Hz, Figure 3F; Crus I-HPC, n = 12; lobule II/III-HPC, n = 13, lobule VI-HPC, n = 16; frequencies between 48 and 52 Hz have been excluded due to notch filtering to remove electrical contamination of the LFP signals; see Materials and methods). We found that significant variations in coherence level were restricted to those within the theta frequency (frequency band (theta, 6–12 Hz; beta, 13–29 Hz; low gamma, 30–48 Hz) x cerebello-hippocampal combination two way repeated measures ANOVA, frequency band effect F2,76 = 22.42, p<0.0001; combination effect F2,38 = 2.843, p=0.0707; interaction effect, F4,76 = 3.825, p=0.0069; post-hoc multiple comparisons with FDR correction revealed significant differences between combinations only in the theta band). In addition, as theta coherence has already been reported as a potential mechanism for long-range network interactions between the hippocampus and the cerebellum (Hoffmann and Berry, 2009; Wikgren et al., 2010) and theta oscillations are known to play an important role in intra-hippocampal network organization, in particular for spatial navigation (see Buzsáki and Moser, 2013 for overview), we focused our analysis on activity within this frequency range. As for the instantaneous LFP power, we next asked if theta coherence between the different cerebello-hippocampal LFP combinations was organized in a state-dependent manner. To address this question, we computed the distribution of instantaneous coherence within this bandwidth. In line with the analysis of LFP power distributions, we did not observe any multi-modality and all combinations followed gaussian distributions (Figure 3G). Correlation analysis between instantaneous speed and instantaneous theta coherence failed to show significant relationships for any cerebello-hippocampal combination (Figure 3H; Crus I-HPC, mean Spearman rho = 0.06 ± 0.03, bootstrap confidence level = [−0.08 0.08]; lobule II/III-HPC, mean Spearman rho = - 0.02 ± 0.02, bootstrap confidence level = [- 0.08 0.08]; lobule VI-HPC, mean Spearman rho = 0.01 ± 0.03, bootstrap confidence level = [−0.09 0.08]). Significant differences across recording combinations were observed within the theta bandwidth (Figure 3I; Crus I-HPC, median [25–75 interquartile range (IQR)] coherence = 0.471 [0.461–0.480]; lobule II/III-HPC, median [IQR] coherence = 0.467 [0.462–0.471]; lobule VI-HPC, median [IQR] coherence = 0.481 [0.471–0.486]; Kruskal-Wallis with FDR correction, Kruskal-Wallis statistic = 6.75, p=0.0342) and post-hoc analysis revealed that LFP oscillations were significantly more synchronized between hippocampus and lobule VI than with lobule II/III (corrected p = 0.0246). Within lobule VI, theta coherence was significantly correlated to the mediolateral position of the recording electrode, which was consistent with the mediolateral location of greatest RABV-labeled PCs (Figure 3J; Spearman rho = 0.572, p=0.0225).

Next, to ascertain if local cerebellar spiking activity is coordinated or modulated by hippocampal theta oscillations we recorded single, photo-identified Purkinje cells from lobule VI of head-fixed L7-ChR2 mice (selectively expressing channelrhodopsin in Purkinje cells; n = 6) simultaneously with hippocampal LFP during periods of active movement. This allowed us to calculate the degree of phase-locking between the cerebellar spikes and hippocampal LFP, which circumvents volume conduction issues associated with LFP-LFP correlations (e.g. Vinck et al., 2011; Nolte et al., 2004; Sirota et al., 2008; Stam et al., 2007). We recorded a total of 22 units, of which 16 (from four mice) were classified as Purkinje cells based upon their responsivity to blue light illumination (see Figure 3—figure supplement 3). Of these 16 Purkinje cells, 31% (5 units) were significantly phase locked to the hippocampal 6–12 Hz oscillation (Figure 3—figure supplement 4) during periods of active movement (see Materials and methods). Of the six non-photo responsive units, 1 (16.7%) was significantly phase locked (Figure 3—figure supplement 4D,E). The mean vector angle of the significantly phase-locked units was 231 ± 18°.

Cerebello-hippocampal interactions during the learning of a goal-directed behavior

To further characterize the dynamics of cerebello-hippocampal interactions, we quantified cerebello-hippocampal theta coherence during a goal-directed task. A subset of mice (n = 8) were trained to traverse a linear track to get a reward (medial forebrain bundle stimulation, see Materials and methods) at a fixed position (Figure 4A).

Figure 4 with 4 supplements see all
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 F6,84 = 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, F2,13 = 2.33, p = 0.1365; distance effect, F84,1092 = 1.043, p = 0.3772; interaction effect, F168,1092 = 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, F2,13 = 6.145, p = 0.0132; distance effect, F84,1092 = 1.682, p = 0.0002; interaction effect, F168,1092 = 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).

https://doi.org/10.7554/eLife.41896.016

Across training, mice improved their performance as shown by the optimization of their path (Figure 4A), significant increase in the mean number of rewards obtained per day of training (Figure 4B; mean number of rewards obtained on 1 st day = 17 ± 4, mean number of rewards obtained on 7th day = 68 ± 11; repeated measures Friedman test, Friedman statistic = 37.91, p < 0.0001) and the significant increase in their mean speed (Figure 4B; mean speed on 1st day = 5.74 ± 0.54 cm/s, mean speed on 7th day = 13.94 ± 1.63 cm/s; repeated measures Friedman test, Friedman statistic = 36.32, p < 0.0001). Thus, we next explored the dynamics of cerebello-hippocampal theta coherence across this learning period. We confirmed the absence of a laterality effect on the power spectra calculated at the beginning (Figure 4—figure supplement 1A; Session 1; HPC left, n = 6, median [IQR] 6–12 Hz z-score power = 5.842 [5.683 6.384]; HPC right, n = 7, median [IQR] 6–12 Hz z-score power = 6.261 [4.919 6.587]; Mann-Whitney test, U = 18, p = 0.7308) or end (Figure 4—figure supplement 1B; Session 20; HPC left, median [IQR] 6–12 Hz z-score power = 5.850 [5.583 6.002]; HPC right: median [IQR] 6–12 Hz z-score power = 5.898 [5.511 6.110]; Mann-Whitney test, U = 20, p = 0.9452) of training. Consequently, we averaged the spectral power from left and right hemispheres when both were available. Similarly, no differences on coherence between left and right hippocampi and the different cerebellar recordings were observed at the beginning (Figure 4—figure supplement 1C,E; Session 1; Crus I-HPC left, n = 3, median [IQR] 6–12 Hz coherence = 0.4789 [0.4557 0.5326]; Crus I-HPC right, n = 3, median [IQR] 6–12 Hz coherence = 0.4802 [0.4687 0.5238]; Mann-Whitney test, U = 4, p > 0.99; lobule II/III-HPC left, n = 4, median [IQR] 6–12 Hz coherence = 0.4574 [0.4513 0.4709]; lobule II/III-HPC right, n = 5, median [IQR] 6–12 Hz coherence = 0.4633 [0.4549 0.4663]; Mann-Whitney test, U = 8, p = 0.7302; lobule VI-HPC left, n = 6, median [IQR] 6–12 Hz coherence = 0.4764 [0.4653 0.4904]; lobule VI-HPC right, n = 5, median [IQR] 6–12 Hz coherence = 0.4684 [0.4558 0.4857]; Mann-Whitney test, U = 10, p = 0.4286) or end (Figure 4—figure supplement 1D,F; Session 20; Crus I-HPC left, median [IQR] 6–12 Hz coherence = 0.5357 [0.4747 0.5574]; Crus I-HPC right, median [IQR] 6–12 Hz coherence = 0.5137 [0.4616 0.5500]; Mann-Whitney test, U = 3, p = 0.7; lobule II/III-HPC left, median [IQR] 6–12 Hz coherence = 0.4596 [0.4466 0.4769]; lobule II/III-HPC right, median [IQR] 6–12 Hz coherence = 0.4595 [0.4422 0.4736]; Mann-Whitney test, U = 10, p = 0.7302; lobule VI-HPC left, median [IQR] 6–12 Hz coherence = 0.4702 [0.4662 0.5124]; lobule VI-HPC right, median [IQR] 6–12 Hz coherence = 0.4665 [0.4610 0.5066]; Mann-Whitney test, U = 12, p = 0.6623) of training so the averaged coherence between cerebellum and both hippocampal hemispheres was computed when possible.

We first examined overall, mean cerebello-hippocampal theta coherence as learning progressed. We observed significant changes over training (Figure 4C; Crus I-HPC, n = 4; lobule II/III-HPC, n = 6; lobule VI-HPC, n = 6; day of training x cerebello-hippocampal combination two-way repeated measures ANOVA with FDR correction, day effect F6,84 = 3.873, p = 0.0018) and post-hoc analysis revealed that only Crus I-HPC coherence significantly increased when comparing with values observed on the first day of training (p < 0.01 for days 6 and 7). We next examined detailed power spectra and coherence dynamics at the level of individual sessions from first day of training (session 1), when animals exhibited an exploratory behavioral profile, and the last day of training (session 20), when animals performed efficient goal-directed behavior (Figure 4). This allowed us to investigate the spatial dynamics of both the cerebellar and hippocampal LFP profiles alongside coherence as mice traversed the linear track to reach the reward.

During session 1, mice approached the reward point with a sustained and low speed (Figure 4Di, top), which is in agreement with the exploratory behavioral profile illustrated by the distributed occupancy of their trajectories on the track (Figure 4A bottom). The hippocampal spectrogram was dominated by sustained activity in the theta band across the whole track. In contrast, clear activity in this frequency band was not apparent in the cerebellar recordings (Figure 4Di) and the LFP power profile was maintained at low levels across the track. This homogeneous pattern was consistent with the unimodal distributions of instantaneous hippocampal and cerebellar theta power (Figure 4—figure supplement 2A). Similarly, the coherograms did not reveal clear coherence in any of the cerebello-hippocampal combinations (Figure 4Dii). However, we found a significant interaction between the distance from reward and the theta coherence of different cerebello-hippocampal combinations (Figure 4Diii; distance from reward point x combination two-way repeated measures ANOVA; combination effect, F2,13 = 2.33, p = 0.1365; distance effect, F84,1092 = 1.043, p = 0.3772; interaction effect, F168,1092 = 1.332, p = 0.0053). Post-hoc, FDR corrected, multiple comparisons revealed that significantly higher theta coherence was present between hippocampus and Crus I compared to lobule II/III at certain positions on the track prior to the reward point location (Crus I-HPC vs lobule II/III-HPC, p < 0.05 from −29 to −21 cm from reward; lobule VI-HPC vs lobule II/III-HPC, p < 0.05 from −25 to −22 cm from reward). As for the homecage recordings, instantaneous theta coherence for all cerebello-hippocampal combinations also followed gaussian, unimodal distributions during session one in the linear track (Figure 4—figure supplement 2E).

On the last day of training, during session 20, mice displayed goal-directed behavior on the track. This goal-directed profile was illustrated in the efficient running trajectories (Figure 4A bottom) and by the acceleration-plateau-deceleration speed profile observed along the track (Figure 4Ei top). As in session 1, the hippocampal spectrogram was dominated by sustained theta activity. On the other hand, cerebellar LFP power profiles were notably different from session one with the appearance of sustained activity in the theta and delta (2–4 Hz) frequency bands, particularly in lobule VI and lobule II/III (Figure 4Ei bottom). These differences were not related to transient bouts of theta activity as the instantaneous theta power probability distributions continued showing unimodal profiles (Figure 4—figure supplement 2C). Coherogram analysis also revealed changes in session 20, which were mainly reflected in sustained Crus I-HPC theta coherence spanning multiple positions on the track as animals actively approached the reward (Figure 4Eii). This pattern was not as apparent in the other cerebello-hippocampal combinations (Figure 4Eiii; distance from reward x cerebello-hippocampal combination two-way repeated measures ANOVA, distance from reward effect F84,1092 = 1.682, p = 0.0002; combinations effect F2,13 = 6.145, p = 0.0132; interaction effect F168,1092 = 1.271, p = 0.0163) and post-hoc FDR corrected multiple comparisons revealed significantly higher sustained theta coherence between hippocampus and Crus I than with lobule II/III from −60 to −20 cm or with lobule VI from - 59 to −36 cm prior to the reward point. Lobule VI-HPC coherence was also higher than with lobule II/III at some positions on the track (−44 to −31 cm from reward) although it was not as sustained and prominent as Crus I-HPC coherence. Importantly, we also reproduced these findings when the imaginary part of coherency was computed, which is robust against contamination by volume conduction (Figure 4—figure supplement 3).

In order to better explore the differences between sessions 1 and 20 we then pooled power spectra (Figure 4F and G top panels) and coherence (Figure 4F and G bottom panels) between −60 and −20 cm from reward point (i.e. the range of distances where differences across recording combinations were most apparent in the coherograms). The appearance of a theta peak in the power spectra from all cerebellar recordings in session 20 compared with session one can be clearly seen (Figure 4G) as well as the increase in coherence limited to this frequency band (Session 20: frequency band (theta, beta, low gamma) x cerebello-hippocampal combination two way repeated measures ANOVA, frequency band effect F2,26 = 13.42, p < 0.0001; combination effect F2,13 = 6.545, p = 0.0108; interaction effect, F4,26 = 5.242, p = 0.0031; post-hoc multiple comparisons with FDR correction revealed significant differences between combination only in the theta band). We found that in session 20 theta coherence between hippocampus and Crus I was significantly higher than that obtained with lobule II/III (Figure 4H right; Crus I-HPC, n = 4, median [IQR] coherence = 0.525 [0.480–0.549]; lobule II/III-HPC, n = 6, median [IQR] coherence = 0.461 [0.452–0.468]; lobule VI-HPC, n = 6, median [IQR] coherence = 0.470 [0.465–0.503]; Kruskal-Wallis with FDR correction, Kruskal-Wallis statistic = 7.989, p=0.0103; Crus I-HPC vs lobule VI-HPC, corrected p = 0.0110) while no significant difference was found in session 1 (Figure 4H left; Crus I-HPC, n = 4, median [IQR] coherence = 0.480 [0.466–0.516]; lobule II/III-HPC, n = 6, median [IQR] coherence = 0.462 [0.457–0.466]; lobule VI-HPC, n = 6, median [IQR] coherence = 0.476 [0.463–0.494]; Kruskal-Wallis with FDR correction, Kruskal-Wallis statistic = 4.779, p = 0.0887).

Given that the speed profiles of mice changed significantly between sessions 1 and 20 on the linear track (Figure 4B,D,G) and seemed to mirror the observed modulation in theta power and coherence, we next correlated instantaneous cerebellar theta power (Figure 4—figure supplement 2B,D) and instantaneous theta coherence (Figure 4—figure supplement 2F,H) with instantaneous speed. In contrast to home-cage recordings, for all recorded cerebellar LFPs, theta power was significantly positively correlated with speed during both session 1 (Figure 4—figure supplement 2B; Crus I, median [IQR] Spearman rho = 0.213 [0.209 0.258], bootstrap confidence level = 0.081; lobule II/III, median [IQR] Spearman rho = 0.239 [0.065 0.343], bootstrap confidence level = 0.082; lobule VI, median [IQR] Spearman rho = 0.235 [0.090 0.284], bootstrap confidence level = 0.083) and session 20 (Figure 4—figure supplement 2D, Crus I, median [IQR] Spearman rho = 0.235 [0.056 0.465], bootstrap confidence level = 0.081; lobule II/III, median [IQR] Spearman rho = 0.310 [-0.010 0.490], bootstrap confidence level = 0.081; lobule VI, median [IQR] Spearman rho = 0.194 [0.105 0.550], bootstrap confidence level = 0.085). Cerebello-hippocampal theta coherence was not correlated with instantaneous speed during session 1 (Figure 4—figure supplement 2F; Crus I-HPC, median [IQR] Spearman rho = 0.024 [-0.033 0.237], bootstrap confidence level = 0.082; lobule II/III-hippocampus, median [IQR] Spearman rho = 0.0278 [-0.105 0.050], bootstrap confidence level = 0.075; lobule VI-HPC, median [IQR] Spearman rho = −0.001 [-0.052 0.045], bootstrap confidence level = −0.090). However, during session 20, lobule II/III-HPC theta coherence was anticorrelated with speed (Figure 4—figure supplement 2H; lobule II/III-HPC, median [IQR] Spearman rho = −0.086 [-0.182 0.012], bootstrap confidence level = −0.080) while lobule VI-HPC was weakly but significantly correlated with it (lobule VI-HPC, median [IQR] Spearman rho = 0.091 [-0.140 0.173], bootstrap confidence level = 0.090). Interestingly, the Crus I-HPC recording combination, which presented higher and sustained levels of theta coherence during this task, was not significantly correlated with instantaneous speed (Crus I-HPC, median [IQR] Spearman rho = 0.020 [-0.051 0.094], bootstrap confidence level = 0.083). Together, these findings suggest that the observed Crus I - HPC theta coherence dynamics during goal-directed behavior in the linear track cannot be explained by changes in running speed.

We also observed similar cerebello-hippocampal coherence dynamics in mice navigating for rewards in a virtual reality based linear track (Figure 4—figure supplement 4A–C, n = 6). A marked spatial re-organization of cerebello-hippocampal theta coherence was apparent in those mice that showed behavioral modulation across training (as evidenced by increases in the number of rewards across day of training and speed modulation on a run by run basis during approach to the reward location; Figure 4—figure supplement 4G–I). In contrast, mice that failed to show behavioral modulation across training, and displayed rather homogenous speed profiles, did not present such coherence dynamics (Figure 4—figure supplement 4D–F).

To examine whether the observed changes in coherence across learning of the linear track were specifically related to performance of the goal-directed task itself, we next conducted pairwise analysis of cerebello-hippocampal theta coherence levels across the following conditions: home-cage prior to any linear track training (HC pre, considered as baseline), the 1st and 20th linear track trials, and home-cage following the end of training in the linear track task (HC post) (see Figure 5).

Hippocampal-Crus I theta coherence is dynamic.

Comparisons of hippocampal-cerebellar theta coherence before and after acquisition of a goal-directed behavior in the linear track task. Levels of coherence during homecage active exploration before first session in the linear track were taken as a baseline. HPC-Crus I theta coherence became significantly different from baseline by session 20 in the linear track task and returned to baseline levels during a second homecage recording immediately following linear track session 20 (n = 4, repeated measures Friedman test with FDR correction, Friedman statistic = 11.1, p = 0.0009; LT session 20 vs baseline, p = 0.0021). No significant differences across conditions were observed for the other hippocampal-cerebellar combinations.

https://doi.org/10.7554/eLife.41896.022

From the three cerebello-hippocampal recording configurations, only Crus I-HPC theta coherence varied significantly across task conditions (Figure 5, n = 4, repeated measures Friedman test with FDR correction, Friedman statistic = 11.1, p = 0.0009). At the outset of linear track learning (session 1), HPC-Crus I coherence values did not significantly differ from home-cage (HC pre, median [IQR] = 0.471 [0.461 0.495]; session 1, median [IQR] = 0.480 [0.466 0.516], corrected p = 0.1196). However, during late stage linear track learning, the level of coherence was significantly higher than in home-cage recordings (HC pre, median [IQR] = 0.471 [0.461 0.495]; session 20, median [IQR] = 0.525 [0.480 0.550], corrected p = 0.0021) and when mice were returned to the home-cage environment following completion of linear track training (HC post) the level of HPC-Crus I coherence dropped back to pre-training levels (HC pre, median [IQR] = 0.471 [0.461 0.495]; HC post, median [IQR] = 0.493 [0.471 0.505], corrected p = 0.0580).

Discussion

Taken together, our findings reveal previously undescribed cerebellar inputs to the hippocampus and offer novel physiological insights into a long-range neural network linking disparate brain regions initially assumed to support divergent behavioral functions, namely spatial navigation (hippocampus) and motor control (cerebellum). Projections from topographically restricted regions of cerebellar cortex discretely route through restricted parts of their associated nuclei en-route to the hippocampus through multiple, convergent pathways, involving one or two relays, from the DCN. Interestingly, the possible single-relay pathways we described points toward involvement of the medial septum and the supramammillary nucleus, two structures crucial for theta generation. Congruently, our physiological data suggest that these connected cerebellar regions may dynamically interact with the hippocampus during behavior, via theta (6–12 Hz) LFP coherence. Our findings thus offer an anatomical and physiological framework for cerebello-hippocampal interactions that could support cerebellar contributions to hippocampal processes (Burguière et al., 2005), including spatial map maintenance (Rochefort et al., 2011; Rondi-Reig et al., 2014; Lefort et al., 2019).

Whilst previous studies provide compelling physiological evidence of cerebellar influences on the hippocampus (Cooke and Snider, 1955; Iwata and Snider, 1959; Babb et al., 1974; Snider and Maiti, 1975; Krook-Magnuson et al., 2014; Choe et al., 2018), they do not provide the spatial resolution afforded by neuroanatomical tracing. Indeed, to the best of our knowledge, anatomical tracing studies have failed to report a mono-synaptic ascending cerebello-hippocampal projection. This is consistent with our rabies virus tracing study, in which incubation periods of 48–58 hr were required before cell labeling was seen in the cerebellar nuclei. Such a timescale is indicative of a multi-synaptic pathway (Kelly and Strick, 2000; Ugolini, 2010; Suzuki et al., 2012; Jwair et al., 2017). Single relay pathways can be envisioned through the septum, the hypothalamus (potentially including the supramammillary nucleus (SUM)) and the raphe nucleus. Other pathways including two relays through either the lateral and medial entorhinal cortex and/or the perirhinal cortex are also possible. Interestingly, among the different regions labeled at 48 hr post-infection, several midbrain and pontine regions such as the PAG, the nucleus incertus and the LtDG are known to receive direct projections from the DCN and could therefore represent putative first-order relays between cerebellum and hippocampus.

Our anatomical results highlight three main inputs to the hippocampus emanating from the cerebellum. The first input we reveal originates from the vestibulo-cerebellum, specifically from the dorsal and ventral paraflocculus, which is likely routed via the vestibular and dentate nuclei (Voogd and Barmack, 2006). This anatomical connection between the vestibulo-cerebellum and the hippocampus reinforces the already well-described influence of the vestibular system on hippocampal-dependent functions (Stackman et al., 2002; Goddard et al., 2008; Zheng et al., 2009).

In addition to the classically described vestibular pathway, our data reveal that the central cerebellum also provides inputs to the hippocampus from vermal lobule VI, routed through caudal fastigial nucleus, and from Crus I, routed through the dentate. Using a combination of RABV expression and zebrin II staining, we identified three specific cerebellar modules involved in these inputs: (1) the A module in lobule VI, (2) the hemispheric Crus I D2 module and (3) the Crus I paravermal C2 module. Of the latter two modules, C2 is likely less prominently anatomically connected with the hippocampus since the number of RABV+ cells in the nucleus interpositus posterior, its output nucleus (Apps and Hawkes, 2009), was minor compared with the other cerebellar nuclei. The convergence of inputs from disparate cerebellar zones (flocculo-nodular and central zones) and modules from vermal (A), paravermal (C2) and hemispheric (D2) regions in to the hippocampus suggest that its optimal function requires the integration of multiple aspects of sensory-motor processing carried out at these distinct cerebellar locations.

According to Voogd and Barmack (2006), based upon evidence from a wide range of species, the oculomotor cerebellum can be most broadly described to include lobule V, VI and VII. Given the highly conserved structure-function relationships of the vermal cerebellum (Sillitoe et al., 2005), it seem likely that the mouse A module could also be considered part of the oculomotor cerebellum. In rat, it receives climbing fibers from the caudal medial accessory olive, and sends mainly ascending projections through the caudal portion of fastigial nucleus (Apps, 1990; Apps and Hawkes, 2009). The oculomotor vermis receives multiple sensory inputs which include visual, proprioceptive, vibrissae, vestibular and auditory inputs conveyed by both climbing and mossy fibers (Voogd and Barmack, 2006). The D2 module receives its climbing fiber input from the dorsal cap of the principal olive and projects out of the cerebellum through the rostromedial dentate nucleus (Herrero et al., 2006). It receives mossy fiber inputs carrying somatosensory, motor (Mihailoff et al., 1981), and visual (Edge et al., 2003) information; along with inputs from the prefrontal cortex (Kelly and Strick, 2003). Climbing fiber inputs to this module relay information from the parvocellular red nucleus, which receives projections from premotor, motor, supplementary motor and posterior parietal areas. The majority of these cortical areas also receive projections from the D2 module after a thalamic relay in the ventro-lateral nucleus (Kelly and Strick, 2003; Glickstein et al., 2011).

Complementary to these anatomical results, our electrophysiological findings reveal coherent activity between the hippocampus and those cerebellar lobules that are anatomically connected with it (lobule VI and Crus I). This synchronization was restricted to the 6–12 Hz frequency range in the awake, behaving animal and showed dynamic profiles that were lobule dependent. Oscillations can align neuronal activity within and across brain regions, suggesting a facilitation of cross-structure interactions (e.g. Singer, 1999; Fries, 2005). Cerebellar circuits support oscillations across a range of frequencies (for review see De Zeeuw et al., 2008; Cheron et al., 2016). Of particular relevance to the current study are reports of oscillations within the theta frequency (~4–12 Hz), which have been described in the cerebellar input layers at the Golgi (Dugué et al., 2009) and granule cell (Hartmann and Bower, 1998; D'Angelo et al., 2001) level, and also in the cerebellar output nuclei (Wang et al., 2014; Watson et al., 2014).

Neuronal coherence has been described across the cerebro-cerebellar system at a variety of low frequencies (O'Connor et al., 2002; Courtemanche and Lamarre, 2005; Soteropoulos and Baker, 2006; Rowland et al., 2010; Frederick et al., 2014; Watson et al., 2014; Chen et al., 2016) and oscillations within the theta range are thought to support inter-region communication across a wide variety of brain regions (Colgin, 2013). Our finding that cerebello-hippocampal coherence is limited to the 6–12 Hz bandwidth is in keeping with previous studies on cerebro-cerebellar communication in which neuronal synchronization has been observed between the cerebellum and prefrontal cortex (Watson et al., 2014; Chen et al., 2016), primary motor cortex (Soteropoulos and Baker, 2006; Rowland et al., 2010), supplementary motor area (Rowland et al., 2010) and sensory cortex (Rowland et al., 2010). Furthermore, LFPs recorded in the hippocampus and cerebellar cortex are synchronized within the theta bandwidth during trace eye-blink conditioning in rabbits (Hoffmann and Berry, 2009; Wikgren et al., 2010). Human brain imaging studies have also described co-activation of blood oxygen level dependent signals in both cerebellar and hippocampal regions during navigation (Iglói et al., 2015) and spatio-temporal prediction tasks (Onuki et al., 2015), thus highlighting putative neuronal interactions between the two structures. Regarding studies in mice, a recent study has demonstrated the existence of statistically significant co-activation of the dorsal hippocampus and cerebellar lobules IV-V, lobule VI and Crus I after the acquisition of a sequence-based navigation task (Babayan et al., 2017).

Multiple lines of evidence suggest that the coherence described here is unlikely to have resulted from volume conduction: 1) Rather than using a common reference electrode, our recordings were bipolar, with each recording electrode being locally and independently referenced (Kajikawa and Schroeder, 2011). 2) If volume conduction of theta oscillations was emanating from a hippocampal source then it could be assumed that cerebellar regions in closer proximity to the hippocampus would show higher levels of coherence (Figure 3—figure supplement 2). However, we found that coherence values were not related to the relative distance between the hippocampus and cerebellar recording site. 3) By recording simultaneously from hippocampus and multiple cerebellar regions, we have been able to demonstrate that the observed coherence is non-homogenous among the different cerebellar lobules in contrast to what one would expect if theta was volume conducted from a common location. 4) We calculated the imaginary part of coherency, which is not influenced by volume conduction (Nolte et al., 2004), in the linear track paradigm and found results that were remarkably similar to those obtained using standard coherence analysis. 5) We showed that Purkinje cell spikes in lobule VI of the cerebellum can phase lock to hippocampal theta oscillations.

Importantly, we have shown for the first time that theta rhythms in the hippocampus preferentially synchronize with those in discrete regions of the cerebellum and that the degree of this coupling changes depending upon the behavioral context. Lobule VI-HPC coherence was dominant during active movement in the home-cage and remained stable during learning of the real world linear track task. On the other hand, Crus I - HPC coherence was highly dynamic, showing a significant increase over learning of the real world linear track task and becoming dominant after the acquisition of a goal-directed behavior. Furthermore, we reveal that high Crus I-HPC coherence was sustained across the linear track when the animal performed goal-directed behavior but not during the early training, exploratoration phase. Such sustained coherence may be related to the ability of the cerebellum, and particular Crus I, to link internal and external sensory context with specific action toward the goal. In line with this hypothesis, a recent study has suggested that Purkinje cells in the Crus I region and neurons in its output, the dentate nucleus, exhibit firing rate modulation in anticipation of expected reward in a VR task (Chabrol et al., 2018). Interestingly, they show that the activity of Purkinje cells in lateral Crus I was modulated by running and/or visual flow speed. This is consistent with our examples showing that Crus I-HPC coherence remained present when animals performed goal-modulated behavior in VR, although multiple streams of sensory input, such as vestibular, whisker and olfactory, become irrelevant and even confounding in the head-restricted virtual environment task.

We next consider our results within the well characterized, modular understanding of cerebellar function. Within lobule VI, the A module receives multi-modal sensory information, mainly arising from collicular and vestibular centers (Voogd and Barmack, 2006). The superior colliculus plays a role in visual processing and generation of orienting behaviors (Basso and May, 2017), which might be relevant for the establishment and maintenance of the hippocampal spatial map, and thus may be required constantly during active movement, independent of the specific behavioral task. The persistent and similar levels of lobule VI-HPC coherence during active movement in the homecage and linear track task, in both real world and virtual reality environment tasks is in agreement with such a hypothesis.

In monkeys and humans, Crus I is anatomically and functionally associated with prefrontal cortex (Kelly and Strick, 2003; Iglói et al., 2015). In mouse Crus I, the D2 module receives convergent sensory and motor information (Proville et al., 2014). Furthermore, this module has been found to contain internal models, a neural representation of one’s body and the external world based on memory of previous experiences, that are used for visuo-motor coordination (Cerminara et al., 2009). Similarly, the C2 module has been found to also participate in visuo-motor processing related to limb coordination during goal-directed reaching (Cerminara and Apps, 2011). Both modules might be particularly important during the acquisition of a goal-directed behavior such as our real-world linear track task in which animals needed to reach non-cued reward zones. Our finding that Crus I-HPC coherence increases during task learning fits with this hypothesis.

In summary, our results suggest the existence of anatomically discrete hippocampal-cerebellar network interactions with a prominent involvement of Crus I during goal-directed behavior. Both anatomical and electrophysiological data point toward involvement of the theta generating pathway in cerebellum-hippocampus interactions. The topographical dynamic weighting of these interactions may be tailored to the prevailing sensory context and behavioral demands.

Materials and methods

Anatomical tracing studies were performed under protocol N°00895.01, in agreement with the Ministère de l’Enseignement Supérieur et de la Recherche. RABV injections were performed by vaccinated personnel in a biosafety containment level two laboratory.

All behavioral experiments were performed in accordance with the official European guidelines for the care and use of laboratory animals (86/609/EEC) and in accordance with the Policies of the French Committee of Ethics (Decrees n° 87–848 and n° 2001–464). The animal housing facility of the laboratory where experiments were made is fully accredited by the French Direction of Veterinary Services (B-75-05-24, 18 May 2010). Surgeries and experiments were authorized by the French Direction of Veterinary Services (authorization number: 75–752).

A total of 44 adult, male mice were used for this study. Seventeen adult male C57BL6-J mice were used for the anatomical tracing study, (Charles River, France) and 21 for the electrophysiology study (Janvier, France). Six adult male CD-L7ChR2 mice were used for the dual hippocampal LFP and cerebellar unit-recording study (in-house colony derived from Jackson labs stock, USA).

Mice received food and water ad libitum, were housed individually (08: 00–20: 00 light cycle) following surgery and given a minimum of 5 days post-surgery recovery before experiments commenced.

Anatomy

Rabies virus injections

Request a detailed protocol

All the RABV (the French subtype of Challenge Virus Standard; CVS-N2C) inoculations were performed in the Plasticity and Physio-Pathology of Rhythmic Motor Networks (P3M) laboratory, Timone Neuroscience Institute, Marseille, France. Mice (n = 17) were injected intraperitoneally with an anesthetic mixture of ketamine (65 mg/kg; Imalgene, France) and xylazine (12 mg/kg; Rompun, Bayer) to achieve surgical levels of anesthesia, as evidenced by the absence of limb withdrawal and corneal reflexes and lack of whisking and were then placed in a stereotaxic frame (David Kopf Instruments, USA). The scalp was then incised, the skull exposed and a craniotomy drilled above the hippocampus.

Mice were injected with 200 nL of a mixture of one part 1% CTb Alexa Fluor 488 Conjugate (Invitrogen, distributed by Life Technologies, Saint Aubain, France) and four parts RABV in the left hippocampus (AP −2.0, ML +2.0, DV 1.97; Figure 1—figure supplement 1 and ) using a Hamilton needle with internal diameter of ~200 µm (Hamilton, USA). Injections (200 nL/min) were performed using a pipette connected to a 10 μL Hamilton syringe mounted on a microdrive pump. Following infusion, the pipette was left in place for 5 min. The incision was then sutured and the animals allowed to recover in their individual home cage for either 30 hr (n = 4); 48 hr (n = 3), 58 hr (n = 5) or 66 hr (n = 5). All animals were carefully monitored during the survival period and, in line with previous studies using these survival times, were found to be asymptomatic (Ugolini, 2010).

Tissue preparation

Request a detailed protocol

At the end of the survival time, mice were deeply anesthetized with sodium pentobarbitone (100 mg/kg, intraperitoneal) then transcardially perfused with 0.9% saline solution (15 mL/min) followed by 75 mL of 0.1M phosphate buffer (PB) containing 4% paraformaldehyde (PFA; pH = 7.4). The brain was then removed, post fixed for 2–3 days in 4% PFA and then stored at 4°C in 0.1 M PB with 0.02% sodium azide. Extracted brains were then embedded in 3% agarose before being coronally sectioned (40 μm) on a vibratome. Serial sections were collected and divided in 4 vials containing 0.1 M PB so consecutive slices in each vial were spaced by 160 μm.

Injection site visualization

Request a detailed protocol

Sections from vial one were used to visualize the injection site by the presence of CTb. In most of the cases, the injected CTb was fluorescent and sections were directly mounted with Dapi Fluoromount G (SouthernBiotech, Alabama, USA). In the other cases (S4-5, S11-13 and S17-18), the sections were first rinsed with PB 0.1 M and then permeated with PB 0.1 M and 0.3% Triton X-100. They were then incubated overnight in a choleragenoid antibody raised in goat (goat anti-CTb, lot no. 703, List Biological Laboratories, USA) diluted 1: 2000 in a blocking solution (PB 0.1 M, 5% BSA). Subsequently, the sections were rinsed in PB 0.1 M and incubated for 4 hr at room temperature with donkey anti-goat secondary antibody (1: 1000 in the blocking solution; Alexa Fluor 555, Invitrogen, distributed by ThermoFisher Scientific, Massachusetts, USA). Finally, the sections were mounted with Dapi Fluoromount G.

The injection site was then visualized using a fluorescence microscope equipped with a fluorescein isothiocyanate filter (Axio Zoom V16, Carl Zeiss, France).

Rabies virus labeled cell quantification

Request a detailed protocol

Sections from vial two were used for quantification and 3 D reconstruction of the RABV labeled cells. Sections mounted on gelatin-coated SuperFrost Plus slides (Menzel-Glaser, Braunschweig, Germany) were first rinsed with PB 0.1 M and pre-treated with 3% H2O2for 30 min in the blocking reaction against endogenous peroxidase. Following pretreatment, the sections were incubated overnight at room temperature with an anti-rabies phosphoprotein mouse monoclonal antibody (Raux et al., 1997) diluted at 1: 10000 in a blocking solution (PB 0.1 M, 0.1% BSA, goat serum 2% and 0.2% Triton X-100). The next sections were rinsed in PB 0.1 M and incubated 2 hr with a biotinylated affinity-purified goat anti-mouse IgG (1: 2000 in blocking solution; Santa-Cruz, Heidelberg, Germany). Then, they were also incubated using an avidin-biotin complex method (Vectastain Elite ABC-Peroxidase kit R.T.U. Universal, Vector Laboratories, Burlingame, CA, USA) to enhance sensitivity. For visualization, the sections were incubated in a 3,3’-diaminobenzidine-tetrahydrochloride (DAB) solution (0.05% DAB and 0.015% H2O2 in PB 0.1 M). Finally, they were counterstained with cresyl and cover-slipped.

Quantitative analyses of rabies-positive cells were performed using a computerized image processing system (Mercator, Exploranova, France) coupled to an optical microscope. The quantification of rabies-positive cells was carried out at 10x magnification. Structures were defined according to a standard atlas (Franklin and Paxinos, 2007). Immunoreactive neurons were counted bilaterally. Representative images were obtained using an Axio Zoom V16 microscope (Carl Zeiss, France).

3-D reconstruction

Request a detailed protocol

A Nikon Eclipse E800 microscope equipped with a digital color camera (Optronics, USA) was used to visualize mounted cerebellar sections under brightfield illumination. The contour of every fourth section was then manually drawn using Microfire software (Neurolucida, MBF Bioscience, USA) and cell counts were performed. The sections were then aligned and stacked (160 µm spacing).

Rabies virus-zebrin II double immuno-staining

Request a detailed protocol

For case S18, sections from vial three were mounted on gelatin-coated SuperFrost Plus slides (Menzel-Glaser, Braunschweig, Germany), rinsed with PB 0.1 M and then permeated and blocked in a solution of PB 0.1 M, 0.2% Triton X-100 and bovine serum 2.5% for 30 min. Then they were incubated for 48 hr at 4°C in a mix of rabbit polyclonal anti-Aldolase C primary antibody (a kind gift from Izumi Sugihara (Sugihara and Shinoda, 2004); No. 69075; 1:500000) and the mouse anti-rabies antibody used for the single RABV staining (1:5000) in a blocking solution (PB 0.1 M, 0.1% Triton X-100 and bovine serum 1%). Subsequently, the sections were first rinsed with PB 0.1% and then incubated in a mix of Rhodamine Red-XGoat anti-rabbit IgG (1: 5000; ref 111-295-144, Jackson Immuno Research) and donkey anti-mouse secondary antibody (1: 5000; Alexa Fluor 647, Invitrogen distributed by ThermoFisher Scientific, Massachusetts, USA) in blocking solution. Finally, the sections were mounted with Dapi Fluoromount G.

Images were obtained using an Axiozoom v16 microscope (Carl Zeiss, France) then cerebellar contours and labeled neurons were manually drawn for reconstruction of zebrin bands,cerebellar modules and location of the RABV+ cells.

Electrophysiology procedures

Preparation 1: Dual hippocampal – cerebellar LFP recording

Request a detailed protocol
Subjects and surgical protocols

Bipolar LFP recording electrodes (interpolar distance of ~0.5 mm; 140 µm diameter Teflon coated stainless-steel, A-M system, USA) were stereotaxically targeted to hippocampus (AP −2.2, ML +2.0, DV 1.0), lobule VI (AP −6.72, ML 0.0, DV 0.1), lobule II/III (AP −5.52, ML 0.0, DV 1.8) and Crus I (AP −6.24, ML 2.5, DV 0.1) of 21 C57BL6-J mice. Pairs of flexible stainless-steel wires were used to also record neck EMG (Cooner wire, USA).

In 15 C57BL6-J mice, bipolar stimulation electrodes (140-μm-diameter stainless steel; A-M system, USA) were also implanted at the left medial forebrain bundle [MFB; to serve as a reward signal; AP −1.4, ML +1.2, DV +4.8 (Carlezon and Chartoff, 2007; de Lavilléon et al., 2015). All electrode assemblies were fixed to the skull using a combination of UV activated cement (SpeedCem, Henry Shein, UK), SuperBond (SunMedical, Japan) and dental cement (Simplex Rapid, Kemdent, UK). Four miniature screws (Antrin, USA) were also attached to the skull for additional support and to serve as recording ground.

In six mice, a lightweight metal head fixation device (0.1 g) was also affixed to the implant. The total implant weight did not exceed 2.5 g (including head fixation post and cement).

Recording

Electrodes were attached to an electronic interface board (EIB 18, Neuralynx, USA) during surgery. Differential recordings were made via a unity-gain headstage preamplifier (HS-18; Neuralynx, USA) and Digital Lynx SX acquisition system (Neuralynx, USA). LFP and EMG Signals were bandpass-filtered between 0.1 and 600 Hz and sampled at 1 kHz. Mouse position was tracked at 30 Hz using video tracker software and infra-red LEDs attached to the headstage (Neuralynx, USA).

Medial forebrain bundle (MFB) stimulation

Intracranial rewarding stimulation consisted of a 140 Hz stimulation train lasting 100 ms delivered through the headstage to the implanted electrodes (SD9k, Grass Technologies, USA). The optimal voltage for intracranial MFB was determined for each mouse with a nose-poke task prior to training (range, 1–6 V [de Lavilléon et al., 2015]).

Preparation 2: Simultaneous cerebellar single unit and hippocampal LFP recording

Subjects and surgical protocols

Request a detailed protocol

General surgery procedures were similar to those used for preparation 1 (dual hippocampal – cerebellar LFP recording experiments). CD-L7ChR2 mice (selectively expressing ChR2 in Purkinje cells; n = 6 mice) were implanted with bipolar LFP recording electrodes in the left hippocampus (AP −2.2, ML +2.0, DV 1) and a lightweight (<0.1 g) recording chamber was constructed over the cerebellum. A silicon elastomer (QuickSil, World Precision Instruments, USA) was used to seal the chamber following surgery and between recording sessions. A stainless steel post (0.1 g) was also affixed to the skull and used for head fixation. To measure movements, flexible stainless steel EMG electrodes were implanted in the front left forelimb.

Recording

Request a detailed protocol

Animals were positioned in a custom-built head fixation device and placed on a floating Styrofoam ball (see virtual reality Materials and methods for further details). Hippocampal and EMG electrodes were connected to an electronic interface board (EIB 18, Neuralynx, USA). Filter and recording settings were the same as in preparation 1. For cerebellar single unit recordings, the silicon elastomer was removed from the recording chamber allowing the exposed cerebellar cortex to be viewed under a microscope (Leica, USA). Quartz based electrodes or tetrodes (impedance 1 mOhm at 1 Khz; Thomas Recording, Germany) were inserted in to the cerebellar cortex using a custom manipulator mounted on a stereotaxic frame (Kopf, USA). An optical fiber (600 µm diameter, Prizmatix, Israel) was mounted on a separate stereotaxic manipulator and positioned inside the chamber just above the cerebellar cortical surface within close proximity to the recording electrode. A low impedance silver ball reference electrode was also positioned within the chamber and a skull screw served as ground. A hydraulic micromanipulator was used to lower the optetrodes through the cerebellar cortical layers (Narishige, Japan).

Once units were identified, blue light pulses were used to photo-identify putative Purkinje cells (50 mW/mm2, ~460 nm, 100 ms delivered using a commercial LED driver, Prizmatix, Israel. Chaumont et al., 2013). Cerebellar signals were recorded using an EIB 18 and headstage (HS-18, Neuralynx, USA) connected to the electrodes via a custom made adapter. Unit signals were band-pass filtered between 0.3 and 9 kHz while sampling was set at 32 kHz.

Histology

Request a detailed protocol

After completion of all the experiments, mice were deeply anesthetized with ketamine/xylazine solution (150 mg/kg) and electrolytic lesions created by passing a positive current through the electrodes (30µA, 10 s). With the electrodes left in situ, the animals were perfused transcardially with saline followed by paraformaldehyde (4%).

Brains were extracted and post-fixed in paraformaldehyde (4%; 24 hr) then embedded in agarose (24 hr). A freezing vibratome was used to cut 50 μm thick sagittal cerebellar and coronal hippocampal sections. The sections were mounted on gelatinized slides and stained with cresyl violet. Recording locations were identified by localized lesions in the cerebellum and hippocampus and plotted on standard maps with reference to a stereotaxic atlas (Franklin and Paxinos, 2007).

Behavioral procedures

Familiar environment

Request a detailed protocol

All recordings were made in the animal’s home-cage (30 cm x 10 cm x 10 cm plastic box), with the lid removed and lasted a maximum of 4 hr. Recordings were made during the day between the hours of 10 am and 6 pm.

Linear track – real world

Request a detailed protocol

The linear track was made in-house from 100 cm x 4 cm x 0.5 cm of black plastic positioned 20 cm above the surface of the experimental table. The behavioral assembly was located in a separate room from the experimenter and was surrounded on four sides by black curtains. Three salient visual cues were placed at fixed locations along the edge of the track (10 cm from the edge). Mice were trained to run in a sequential manner from one end of the track to the other in order to receive a reward, which consisted of an electrical stimulation of the MFB. The reward stimulation was delivered automatically when the mice reached a 5 cm wide goal-zone, which was located 10 cm from the end of the track. Timing of the reward signal was logged on the electrophysiological recordings via TTL signals. Sessions lasted 12 min and were repeated three times per day with an inter-session time of 5 min over 7 days. Between sessions, the track was cleaned with 20% ethanol.

Linear track – virtual reality environment

Request a detailed protocol

A commercially available virtual-reality environment was used (Jet Ball, Phenosys, Germany), utilizing an air cushioned Styrofoam ball (200 mm), which served as a spherical treadmill for head restrained mice (Lasztóczi and Klausberger, 2016) (Figure 4—figure supplement 2). The floating ball assembly was positioned 20 cm from a series of six octagonally arranged TFT surround monitors (19 inch) such that the head restrained mice had an unobstructed view of the visual scene. Rotation of the Styrofoam ball was detected by an optical sensor (sampling frequency 5700 dots per inch at 1 kHz). The vertical axis signals were interpreted by the VR software as the forward and backward movement of the virtual position of the animal. Position within the VR was then translated to a voltage signal (zero to five volts, with five volts indicating the end of the track), and sent to the Digital Lynx SX (Neuralynx, USA) electrophysiology system via a DACQ interface (DACQ 6501, National Instruments, USA). The start of the VR display was logged on the electrophysiology recordings via a TTL signal. To provide a reward signal, when the mice reached a given location within the VR (10 cm from the end of the track) a TTL marker was sent to both the electrophysiological recording system (to provide a timestamp-marker of the event) and an electrical stimulus generator linked to the HS-18 headstage (in the same manner as for real world linear track experiments).

The virtual scene consisted of a 1 m long track with gray walls and included three salient visual cues. After 3 × 12 mins sessions of habituation to the head fixation on the floating-ball assembly, mice were trained to run on the linear track in 12 min sessions, three times per day with an inter-session interval of approximately 5 mins during 7 days. The number of rewards received by the animal was logged in the electrophysiology software (Cheetah 5.6.3, Neuralynx, USA).

Behavioral and electrophysiological analysis

All data were processed in Matlab (Mathsworks, USA), Spike 2 (Cambridge Electronic Design, UK) and Prism (Graphpad, USA).

Behavior

Request a detailed protocol

In all conditions, behavioral data were analyzed using custom-made Matlab scripts. Instantaneous speed (or virtual speed) was derived from video tracking data (or virtual X and Y coordinates recorded as voltage signals in Neuralynx Cheetah software) and downsampled to 10 Hz for consistency with spectrogram and coherogram data (see below). Only epochs of active movement, defined by a sustained speed above 3 cm/s for a minimum of 4 s, were selected for further analysis.

In the home-cage environment, there was one case in which video tracking was not available and the epochs of active movement were thus defined by an amplitude threshold in EMG signal and the presence of a high theta/delta ratio in the hippocampal recordings.

For linear track and virtual reality-based experiments, active movement periods (instantaneous speed >3 cm/s) were analyzed from each 12 min session (three sessions per day) for overall calculations. Each session was then subdivided in to trials (runs), that is, each time the animal traversed the track from one reward site to the other, and measurements of speed, power and coherence were averaged by distance from reward location in 1 cm bins normalized by the occupancy (time spent in each bin).

Electrophysiology

Request a detailed protocol

LFP data was z-scored, notch filtered (filter centered to 50 Hz to remove electrical line noise) and detrended (using local-linear regression with a moving window of 1 s in 0.5 s steps to remove the DC component) prior to subsequent analysis. Multi-taper Fourier analyses (Chronux toolbox [Bokil et al., 2010]) were used to calculate power and coherence of the LFP data. We used a 1 s sliding window in 0.1 s step and four tapers for all analysis. Time points with large-amplitude, low frequency artifacts, identified by threshold crossings of the mean z-score power in 0.5–5 Hz band, were removed from further analysis. Similarly, peri-MFB electrical stimulation times (±0.5 s) in linear track and virtual environment conditions were also excluded from analysis.

For overall power spectra and coherence calculations, the means of the spectrograms and coherograms were respectively computed. Spectral power between 0.1 and 500 Hz was z-scored to homogenize values and reduce the impact of inter-animal and inter-region global variations. Frequency axes in both spectral power and coherence plots were presented in logarithmic scale to facilitate visualization across a large frequency band (1–300 Hz). Frequencies between 48–52 Hz were removed from coherence plots due to the presence of a spurious peak related to the notch filtering. Data duration of recordings made in the home-cage environment, varied across mice (range, 12 to 132 min). Therefore, to reduce the impact of data length on subsequent analyses and also to match with subsequent linear track experiments (duration of 12 min), for each mouse we concatenated the LFP in to 12 min blocks. When multiple 12 min blocks were available (number of data blocks ranged from 1 to 11), we calculated the average coherence across all blocks.

Combined recordings of cerebellar units and hippocampal LFP

Request a detailed protocol

Cerebellar cell recordings were sorted using Spike2 software (CED, UK) where single units were verified with principal component analysis. We did not separate simple and complex spikes when calculating firing rate. For each cell, the firing rate was normalized against a 1 s pre photo-illumination period. Firing rate was then computed in 10 ms bins. A change of 1.96z was used to classify a significant change in firing rate during photo-illumination (Chaumont et al., 2013). For phase locking analysis, hippocampal LFP was bandpass filtered from 6 to 12 Hz. Circular statistics were used to quantify phase distribution (e.g. Jones and Wilson, 2005) of each cell and to determine significant phase-locking at p < 0.05.

Statistical analysis

Request a detailed protocol

Statistical analyses were conducted using Matlab Statistical Toolbox and Prism (Graphpad, USA). Normality was assessed using a Shapiro-Wilk test. Parametric and non-parametric tests were then used accordingly.

For bootstrap calculations on the correlation of theta power and theta coherence with speed, we computed the Spearman's correlation between these variables and a randomly shuffled rearrangement of instantaneous speed values. We repeated this 1000 times in order to obtain the cumulative probability distribution of the random correlation values. The confident limits for α = 0.05 were obtained by finding the correlation values at probabilities of 0.025 and 0.975.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
    The Mouse Brain in Stereotaxic Coordinates (Third Edition)
    1. K Franklin
    2. G Paxinos
    (2007)
    Elsevier.
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
    Cerebellar stimulation: effects on septal region, Hippocampus, and amygdala of cats and rats
    1. RG Heath
    2. CW Dempesy
    3. CJ Fontana
    4. WA Myers
    (1978)
    Biological Psychiatry 13:501–529.
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
    Cerebellar directed optogenetic intervention inhibits spontaneous hippocampal seizures in a mouse model of temporal lobe epilepsy
    1. E Krook-Magnuson
    2. GG Szabo
    3. C Armstrong
    4. M Oijala
    5. I Soltesz
    (2014)
    eNeuro, 1, 10.1523/ENEURO.0005-14.2014, 25599088.
  57. 57
  58. 58
    Impaired cerebellar purkinje cell potentiation generates unstable spatial map orientation and inaccurate navigation
    1. JM Lefort
    2. J Vincent
    3. L Tallot
    4. F Jarlier
    5. CI De Zeeuw
    6. L Rondi-Reig
    7. C Rochefort
    (2019)
    Nature Communications, 10, 10.1038/s41467-019-09958-5, 31113954.
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
    Is the cerebellum ready for navigation?
    1. L Rondi-Reig
    2. E Burguière
    (2005)
    Progress in Brain Research, 148, 10.1016/S0079-6123(04)48017-0, 15661192.
  75. 75
  76. 76
  77. 77
  78. 78
    Conservation of the architecture of the anterior lobe vermis of the cerebellum across mammalian species
    1. RV Sillitoe
    2. H Marzban
    3. M Larouche
    4. S Zahedi
    5. J Affanni
    6. R Hawkes
    (2005)
    Progress in Brain Research, 148, 10.1016/S0079-6123(04)48022-4, 15661197.
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84
  85. 85
  86. 86
  87. 87
  88. 88
  89. 89
  90. 90
  91. 91
  92. 92
    Topography of cerebellar nuclear projections to the brain stem in the rat
    1. TM Teune
    2. J van der Burg
    3. J van der Moer
    4. J Voogd
    5. TJ Ruigrok
    (2000)
    Progress in Brain Research, 124, 10.1016/S0079-6123(00)24014-4, 10943123.
  93. 93
  94. 94
  95. 95
    Oculomotor cerebellum
    1. J Voogd
    2. NH Barmack
    (2006)
    Progress in Brain Research, 151, 10.1016/S0079-6123(05)51008-2, 16221591.
  96. 96
  97. 97
  98. 98
  99. 99
  100. 100
  101. 101
  102. 102

Decision letter

  1. Richard B Ivry
    Senior and Reviewing Editor; University of California, Berkeley, United States
  2. Peter Strick
    Reviewer; University of Pittsburgh

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Anatomical and physiological foundations of cerebello-hippocampal interactions" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by Indira Raman as the Reviewing Editor and Richard Ivry as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal his identity: Peter Strick (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This manuscript examines the communication between the cerebellum and the hippocampus by tracing anatomical pathways connecting the two regions and examining the coherence of theta oscillations between the (dorsal) hippocampus and the cerebellum during spatial navigation in mice.

Essential revisions:

The three reviewers all found aspects of the work to be of interest. They all, however, had fairly extensive critiques, including comments on the limitations of the correlative data and the ambiguity of the link between cerebellar and hippocampal oscillations; the incompletely explained methods and uncertainties about how some methods/illustrations relate to results with occasional apparent mismatches; and statistical questions, including the validity of interpretations given the relatively small numbers of mice. Since many of the comments stem from the difficulty the reviewers had in following what the experiments where and why certain approaches were taken as being valid, the reviewers agreed that there may be multiple appropriate ways in which you may choose to address the comments. We are therefore not stipulating precisely what the explicit experiments, analyses, and or revisions should be. Nevertheless, for the revision, please address the concerns by making the following essential revisions:

1) Acknowledging the limitations of the correlative data address/explain more clearly the link between cerebellar theta and hippocampal oscillations;

2) Providing a clearer explanation of methods and justification of their validity (especially regarding temporal components of manipulations) and ensure that the text and interpretations are consistent with the data (as illustrated and as summarized);

3) Addressing the statistical queries in the reviewers' comments, including addressing the small numbers of mice on what conclusions were based in some experiments.

The specific (concatenated) comments in the reviewers' words, with editorial notes about which essential revision(s) they primarily pertain to in brackets, are given below to help guide you in your revision. Some points span multiple categories of revisions, but the indicators are given to try to provide some structure and clarity. Also, the full comments are included for completeness, but because some items were noted by more than one reviewer, some redundancy is present. We realize that in some cases a single response may answer more than one specific comment, which is fine, as long as all points are addressed in the revision.

The general assessments are also provided.

General assessments:

Reviewer #1:

This is an interesting paper that uses (1) anatomy to demonstrate that topographically restricted regions of the cerebellar cortex are connected to the hippocampus and (2) that local field potential measured oscillatory activity in the theta band (6 – 12 Hz) is coordinated between sub-regions of the cerebellum and the hippocampus during navigation. I find the topic timely and the results of broad interest but do have some concerns regarding the robustness of the effects and the limited presence of oscillatory activity in the theta frequency range in the cerebellum. The report is also primarily correlative in nature – although there are interesting novel insights (the topographic nature of cerebellar-hippocampal connectivity and the increase in coherence between specific cerebellar regions and the hippocampus with navigation/experience).

Reviewer #2:

This is an interesting study describing the interactions between cerebellum and hippocampus during a goal-directed behavior. To this end the authors first studied the topography of the cerebellum regions projecting to the hippocampus using rabies via a polysynaptic (disynaptic?) pathway. Then, they demonstrate a spectral coherence of the theta oscillations (an important neural mechanism that supports neural communication) between these regions during active movement in the home cage and during learning of a goal-directed behavior task in a linear track and in virtual reality conditions in mice; difference of coherence between the regions seem to match the density of connections, and changes of coherence intensity takes place during learning in freely moving mice but not in virtual reality. The overall results seem appealing for a broad readership interested in functional relationship between different structures, here the cerebellum and the hippocampus during behavior, although the cause of temporal variations in coherence remains elusive (caused by upstream gating of theta-modulated inputs? changes in cerebellar oscillation amplification?) and is not discussed. Cellular recordings in the cerebellum showing phase-locking to hippocampal theta oscillations to ascertain that LFP is reflecting intra-cerebellar activity (and not simply external entrainment) would have been helpful but this would require more than can be completed in a reasonable revision time. Anyway, this work could be much improved by the authors with new analysis and clarification of interpretation of their data.

Reviewer #3:

This is a fascinating manuscript in which the authors provide anatomical and physiological evidence for cerebellar interactions with the hippocampus. Based on the data presented, it is clear that regions of the cerebellar vermis and hemisphere, as well as their associated deep nuclei are a source of multisynaptic drive to the hippocampus. The anatomical results provide the most compelling data so far on this issue. Overall, I find the results believable, but requiring additional documentation. For example, the injection sites in the hippocampus are not clearly illustrated. The neurons that mediate the transneuronal transport have not been identified. It is not clear whether the connection between the cerebellar nuclei and the hippocampus is mediated by one or two neurons. The neurons that mediate this connection do not appear to have been analysed. At least the authors have neither presented nor illustrated this critical information. Thus, we are left with clear evidence that multiple regions of the cerebellum influence the hippocampus, but the subcortical or cortical regions that mediate this influence are uncertain. I believe that the authors should analyse and illustrate this critical information prior to publication.

Major points [concatenated from all reviewers]:

1) [Related to Essential Point 1] The authors indicate that transient oscillatory activity was observed in the 6-12 Hz range in the cerebellum. To navigate, sensory-motor information must be integrated nearly continuously and how the cerebellum meaningful contributes to this calculation in the hippocampus is not clear to me given the transient nature of theta in the cerebellum.

2) [Related to Essential Point 1] In addition, more information should be presented regarding this feature [the transience] of cerebellar theta. For example, what is the frequency or duration of transient theta oscillations in the home cage versus various tasks? Are the transitions between theta frequency and other frequencies discrete (i.e. what are the dynamics of preceding and following activity in the theta frequency range?). It's also not always clear whether analyses are focusing on identified transient periods of theta or longer periods of theta. It's possible I missed this information but it should be clear in the main text.

3) [Related to Essential Point 1] The authors show only power spectra and coherence until 40 Hz and they focused on theta but there is clearly increased activity in beta frequencies which are not commented nor discussed (while task-related changes in beta frequencies have also been reported in the hippocampus). Recording were performed with sampling frequency of 1kHz (from Materials and methods) so the authors could also analyze higher frequencies up to 200 Hz. A large fraction of the gamma band is currently not really studied.

4) [Related to Essential Point 1] In Figure 4G-H, "observed changes restricted to the theta band": around 30Hz there is definitely a peak, both in trial 1 and in trial 20 (even larger/higher in trial 20). A shift in 30Hz peak can also be seen on the z-score power of the HPC (Figure 4E and F). Notably, this peak seems absent from the homecage recordings (Figure 3D). In virtual reality, this peak at 30 Hz seems absent, except when selecting specific epochs comparable to the linear track trials (Figure 7C). There might be something to investigate here for the authors, please check the absence of significance in this frequency band.

5) [Related to Essential Point 1] The coherence analysis is classically difficult to interpret because of volume conduction. Indeed, the authors did not find a clear peak of theta in power spectrum in the cerebellum but only transient oscillations in traces. The authors make an argument on distance, but this may be misleading if the dipolar nature of the source is not taken into account. Imaginary part of coherency is robust against volume conduction and should be used to replicate the findings (see Nolte et al., 2004).

6) [Related to Essential Point 1] The comparison of freely moving and virtual-reality is tricky. The learning curves look very different in Figure 4B and Figure 6C. Please provide a comparison between learning curves between linear track and virtual reality conditions. This impact on the choice of the epochs with similar behavior performances to compare in Figure 7. How similar is this behavior? Is it simply running speed? The VR animals seem overtrained while the freely moving seem to be learning during the session; this may result in differences of hippocampal engagement, beyond differences of sensory context.

7) [Related to Essential Point 2] It is a bit hard to make sense of the time-lapse of the cerebellum; from the Figure 1—figure supplement 2, it seems that the cerebellar nuclei is infected 18 hours after the primary infections observed in the structures which project to the cerebellum, suggesting that the cerebellum is one synapse upstream these hippocampus-projecting regions. Is any of the structures labelled at 30h known to receive inputs from the cerebellar nuclei? Alternatively, is it conceivable that some infection took place in the cerebral cortex overlying the hippocampus? The size of the pipette used for the injection is not provided. A picture of the cortex over the infection site would be useful for the reader to judge of the risk of virus leaks to the cortex.

8) [Related to Essential Point 2] Regarding the anatomical study in subsection “A precise topography of the cerebellum regions projecting to the hippocampus”, second paragraph, the authors stated that the presence of labeling in the contralateral hippocampus at the earliest survival time is at 30h vs. 48h for only contralateral markings in cerebellar output nuclei DCN: that's more than the 12h required for an infection cycle, so they should start seeing cells as well in the ipsilateral cerebellar nuclei, no? The time lapse of the retroinfection isn't very clear: both hemispheres of the cerebellar cortex are infected at the same moment, while the DCNs are infected at different times (cf. Supplementary Figure 2)? The symmetry in the pattern of infection in the cerebellar cortex suggests connections going to the same area, but the timelapse of the infection installs some doubts. Please clarify.

9) [Related to Essential Point 2] I do not see some of the findings mentioned in the following text illustrated in Figure 1. "Rather, RABV/CTb-labeled neurons were found in two well described subcortical pathways leading to the DG of the hippocampus (first cycle of infection). One labeled pathway included the diagonal band of Broca and the septum. The other labeled hippocampal input pathway included the lateral entorhinal and perirhinal cortices (Figure 1—figure supplement 3)(Mosko et al., 1973; Dolorfo and Amaral, 1998; Witter, 2007.… "

10) [Related to Essential Point 2] While the tracing was performed by injections in the dentate gyrus, recordings were performed in another region, the CA1 area. Please clarify the motivation of these choices and discuss how this may impact the result.

11) [Related to Essential Point 2] Subsection “A precise topography of the cerebellum regions projecting to the hippocampus”, last paragraph: Several pieces of anatomical information are missing from the analysis.

a) What is the location of first order neurons in subcortical structures that are labeled by retrograde transport after hippocampal injections of RV?

b) What is the time course of the appearance of this first order labeling?

c) What is the time course of the appearance of neurons in fastigial, interpositus and dentate? Do they all appear at the same time in individual animals? Is the timing of their appearance fully consistent with all the neurons in the deep cerebellar nuclei being second order neurons labeled by retrograde transneuronal transport? Or is it possible that some of the neurons are third order neurons? In other words, do neurons in the fastigial nucleus label 8 hours before neurons in the dentate? If so, then this would be consistent with fastigial neurons being second order and dentate neurons being third order.

12) [Related to Essential Point 2] Is the time course of the first labeled neurons in the deep nuclei fully consistent with their being second order neurons? Or is it possible that even the earliest labeled neurons are third order neurons that are labeled by transneuronal transport through two subcortical neurons?

13) [Related to Essential Point 2] What is the time course of labeled Purkinje cells in different regions of the cerebellar cortex? Do neurons in the vermis label ~8 hours prior to neurons in the hemisphere?

14) [Related to Essential Point 2] As the anatomical data currently stand, their analysis and/or presentation are incomplete. The following statement is neither clearly stated in the Results section nor clearly illustrated. "Notably, all of these regions contained RABV+ cells at 48h p.i., and thus they cannot be excluded as potential routes towards the hippocampus."

15) [Related to Essential Point 2 and 3] Some additional methods are needed. It was not clear to me how recordings were consistently selected for inclusion in various analysis. For example, the last paragraph of the subsection “Cerebello-hippocampal physiological interactions in a familiar home-cage environment”, lists 23 values from 13 mice. How were these sessions selected? Which mice did these sessions come from (up to two sessions per mouse?) and why would session numbers used in analyses differ between mice? This is just one example but the comment applies to statistics throughout the paper.

16) [Related to Essential Point 3] The authors provided a measure of physiological interaction – namely theta oscillations between cerebellum and hippocampus. First they assessed the cerebello-hippocampal interactions with coherence in the home cage and then during the learning of a goal directed behavior in a linear track and in virtual reality conditions. In the third paragraph of the subsection “Cerebello-hippocampal interactions during the learning of a goal-directed behavior” and Figure 4E-F: The shift in mean theta power and peak frequency is reported significant in the text but not clear on the figure when no visible difference in the theta peak between trial 1 and 20 was observed. Also when looking at the numbers in this z-score power (subsection “Cerebello-hippocampal interactions during the learning of a goal-directed behavior”, third paragraph), the difference is barely visible and the SEM give overlapping numbers (1st trial between 1.59 and 1.69, 20th trial between 1.68 and 1.76 for example): the statistical test seems to be conflicting with these results. Please clarify.

17) [Related to Essential Point 3] Some of the effects appear to rely heavily on smaller numbers of mice, although the authors do have quantitative explanations for the variability they observe (i.e. the recording distance from the midline in Figure 3F). Even so, Figure 3F seems to be primarily driven by two mice and in Figure 5, I was unsure of how the authors interpreted the high degree of variability in mean coherence in the Crus I data set.

18) [Related to Essential Point 3] The authors recorded the activity in both left and right hippocampus in all mice and they noted no difference (subsection “Cerebello-hippocampal physiological interactions in a familiar home-cage environment”, fourth paragraph) in the analysis of cerebello-hippocampal coherence. However, in Figure 4 to Figure 7 the data from different animals are indifferentially reported from data from the same animal. Indeed, the significant effects, for example for Crus I in Figure 5 and Figure 7D may be only due to a few values, possibly coming from only one or two animals; moreover it seems that some of the statistics are not performed using repeated-measure procedures. Please provide a clear view on left/right recording from each animal and clarify the statistics (e.g. averaging values from single animals). Overall, the number of mice seems a bit small for the task-related coherence analysis to evaluate reproducibility of the results.

https://doi.org/10.7554/eLife.41896.026

Author response

Essential revisions:

The three reviewers all found aspects of the work to be of interest. They all, however, had fairly extensive critiques, including comments on the limitations of the correlative data and the ambiguity of the link between cerebellar and hippocampal oscillations; the incompletely explained methods and uncertainties about how some methods/illustrations relate to results with occasional apparent mismatches; and statistical questions, including the validity of interpretations given the relatively small numbers of mice. Since many of the comments stem from the difficulty the reviewers had in following what the experiments where and why certain approaches were taken as being valid, the reviewers agreed that there may be multiple appropriate ways in which you may choose to address the comments. We are therefore not stipulating precisely what the explicit experiments, analyses, and or revisions should be. Nevertheless, for the revision, please address the concerns by making the following essential revisions:

1) Acknowledging the limitations of the correlative data address/explain more clearly the link between cerebellar theta and hippocampal oscillations;

2) Providing a clearer explanation of methods and justification of their validity (especially regarding temporal components of manipulations) and ensure that the text and interpretations are consistent with the data (as illustrated and as summarized);

This essential point 2 concerns the anatomical part of the manuscript. Following reviewer’s suggestions we have now illustrated the injection site in more detail and also performed a detailed analysis of the neurons that mediate the transneuronal transport. To provide a detailed documentation of these points, Figure 1 now has 6 associated supplementary figures:

- Figure 1—figure supplement 1 illustrates the injection site for all the mice. We also illustrated this injection site with CTB labeling pictures. Following this new analysis we have removed one 30h animal due to leaks into the cortex.

- Figure 1—figure supplement 2 corresponds to RABV labeling at 30h post injection.

- Figure 1—figure supplement 3 corresponds to CTB labeling. This figure confirms that the predominantly RABV labeled structures at 30h post injection are first order structures.

- Figure 1—figure supplement 4 corresponds to RABV labeling at 48h post injection. Structures labeled at 48h that were not labeled at 30h are potential second order structures (written in bold in the table)

- Figure 1—figure supplement 5 is a summary table of RABV labeling in cerebellum and vestibular nuclei at 58h post injection.

- Figure 1—figure supplement 6 illustrating the topographical distribution of DCN labeling at 66h.

3) Addressing the statistical queries in the reviewers' comments, including addressing the small numbers of mice on what conclusions were based in some experiments.

The specific (concatenated) comments in the reviewers' words, with editorial notes about which essential revision(s) they primarily pertain to in brackets, are given below to help guide you in your revision. Some points span multiple categories of revisions, but the indicators are given to try to provide some structure and clarity. Also, the full comments are included for completeness, but because some items were noted by more than one reviewer, some redundancy is present. We realize that in some cases a single response may answer more than one specific comment, which is fine, as long as all points are addressed in the revision.

[…]

1) [Related to Essential Point 1] The authors indicate that transient oscillatory activity was observed in the 6-12 Hz range in the cerebellum. To navigate, sensory-motor information must be integrated nearly continuously and how the cerebellum meaningful contributes to this calculation in the hippocampus is not clear to me given the transient nature of theta in the cerebellum.

We agree with the reviewer that understanding the temporal dynamics of cerebellum and hippocampus activity is a major question. Therefore, our spectrogram and coherogram analyses have been adapted to use parameters that allow for better temporal resolution (1s window in 0.1s steps compared with the previous 10s with 1s step). With these parameters we were able to explore temporal and spatial dynamics at the level of single trials.

In the linear track condition we have a constrained behavior and the task can be divided into individual trials (each run from one end of the corridor to the next goal zone) inside a session (each 12 minute block). We have now analysed the spectrogram averaged by the distance to the reward in late training and we observed rather continuous theta band activity during the whole goal-directed behavior (see new Figure 4Ei). We also computed coherograms averaged by the distance to the reward. Our new analysis reveals sustained coherence between hippocampus and Crus I, but not Lob II/III or Lob VI, in the theta band (see new Figure 4Eii, Eiii, 4H). Interestingly, activity and synchronization in this frequency band is also absent when analyzing trial 1 (new Figure 4D, F, H).

These results (subsection “Cerebello-hippocampal interactions during the learning of a goal-directed behavior”) are in line with the general findings described in the previous version of the manuscript, i.e. the increase of coherence between the Crus I region and the hippocampus when the animal performs a goal directed behavior (Figure 4C), and offer further insights on the dynamics of this interaction during ongoing behavior.

In the homecage, the animal performed a mixture of different behavioral states that are difficult to separate or divide on a “trial by trial” basis. To analyze the nature of the theta activity in the home cage, we looked at the distribution of instantaneous theta power. We found a skewed but unimodal distribution arguing against the existence of two differentiated states (presence versus absence of theta activity) (new Figure 3D) (see also new Results subsection “Cerebello-hippocampal physiological interactions in a familiar home-cage environment”).

Altogether, these new analyses sustain the idea that the cerebellum may provide the hippocampus with integrated sensory-motor information within the contextual framework of the goal directed behavior (see new Discussion section, eighth paragraph), which should occur in a nearly continuous manner (as pointed out by the reviewer).

2) [Related to Essential Point 1] In addition, more information should be presented regarding this feature [the transience] of cerebellar theta. For example, what is the frequency or duration of transient theta oscillations in the home cage versus various tasks? Are the transitions between theta frequency and other frequencies discrete (i.e. what are the dynamics of preceding and following activity in the theta frequency range?). It's also not always clear whether analyses are focusing on identified transient periods of theta or longer periods of theta. It's possible I missed this information but it should be clear in the main text.

As discussed in point 1, we now show evidence for a continuum in the cerebellar theta activity rather than transient activity. Briefly, across the all conditions, we analysed LFP power and coherence during epochs of active movement (speed above 3 cm/s). Furthermore, for the plots in Figure 4F-H we have averaged the spectra and coherence within the range of distances from the reward that showed significant differences across combinations in session 20 (-60 to -20 cm, see Figure 4Eiii).

3) [Related to Essential Point 1] The authors show only power spectra and coherence until 40 Hz and they focused on theta but there is clearly increased activity in beta frequencies which are not commented nor discussed (while task-related changes in beta frequencies have also been reported in the hippocampus). Recording were performed with sampling frequency of 1kHz (from Materials and methods) so the authors could also analyze higher frequencies up to 200 Hz. A large fraction of the gamma band is currently not really studied.

We have now analysed a much larger frequency profile (2-300 Hz) for both our LFP power and coherence analyses. We have analysed the differences between cerebello-hippocampal combinations in the frequency bands theta (6-12 Hz), beta (13-29 Hz) and low-gamma (30-48 Hz) and then used a two-way repeated measures ANOVA for frequency band and combination factors. In all conditions the only significant values or changes across combinations were found within the theta range (new Figures 3C, F, I; 4D-H) (see below for detailed statistical analysis and Results section).

Homecage:

Frequency band effect F2,76 = 22.42, p < 0.0001

Combination effect F2,38 = 2.843, p = 0.0707 Interaction effect, F4,76 = 3.825, p = 0.0069

FDR corrected multiple comparisons between combinations for each frequency band:

Theta:

Lob VI-HPC vs. Crus I-HPC, corrected p = 0.003

Lob VI-HPC vs. Lob II/III-HPC, corrected p < 0.0001

Crus I-HPC vs. Lob II/III-HPC, corrected p = 0.0964

Βeta:

Lob VI-HPC vs. Crus I-HPC, corrected p = 0.8314

Lob VI-HPC vs. Lob II/III-HPC, corrected p = 0.8314

Crus I-HPC vs. Lob II/III-HPC, corrected p = 0.8314

Gamma:

Lob VI-HPC vs. Crus I-HPC, corrected p = 0.7015

Lob VI-HPC vs. Lob II/III-HPC, corrected p = 0.5968

Crus I-HPC vs. Lob II/III-HPC, corrected p = 0.5968

Linear track:

Session 1:

Frequency band effect F2,26 = 9.283, p = 0.0009

Combination effect F2,13 = 2.193, p = 0.1512

Interaction effect, F4,26 = 2.044, p = 0.1176

Session 20:

Frequency band effect F2,26 = 13.42, p < 0.0001

Combination effect F2,13 = 6.545, p = 0.0108 Interaction effect, F4,26 = 5.242, p = 0.0031

FDR corrected multiple comparisons between combinations for each frequency band:

Theta:

Lob VI-HPC vs. Crus I-HPC, corrected p = 0.0016

Lob VI-HPC vs. Lob II/III-HPC, corrected p = 0.0178

Crus I-HPC vs. Lob II/III-HPC, corrected p < 0.0001

Beta:

Lob VI-HPC vs. Crus I-HPC, corrected p = 0.4051

Lob VI-HPC vs. Lob II/III-HPC, corrected p = 0.9882

Crus I-HPC vs. Lob II/III-HPC, corrected p = 0.3978

Gamma:

Lob VI-HPC vs. Crus I-HPC, corrected p = 0.5312

Lob VI-HPC vs. Lob II/III-HPC, corrected p = 08208

Crus I-HPC vs. Lob II/III-HPC, corrected p = 0.4084

We have now also modified our plots (new Figure 3C, F and Figure 4F and G) to a logarithmic scale which to display the full frequency range (2-300 Hz).

4) [Related to Essential Point 1] In Figure 4G-H, "observed changes restricted to the theta band": around 30Hz there is definitely a peak, both in trial 1 and in trial 20 (even larger/higher in trial 20). A shift in 30Hz peak can also be seen on the z-score power of the HPC (Figure 4E and F). Notably, this peak seems absent from the homecage recordings (Figure 3D). In virtual reality, this peak at 30 Hz seems absent, except when selecting specific epochs comparable to the linear track trials (Figure 7C). There might be something to investigate here for the authors, please check the absence of significance in this frequency band.

As mentioned in point 3 (above), we have analysed the whole frequency profile. The significant difference is restricted to the theta band (6-12 Hz, see statistics above).

5) [Related to Essential Point 1] The coherence analysis is classically difficult to interpret because of volume conduction. Indeed, the authors did not find a clear peak of theta in power spectrum in the cerebellum but only transient oscillations in traces. The authors make an argument on distance, but this may be misleading if the dipolar nature of the source is not taken into account. Imaginary part of coherency is robust against volume conduction and should be used to replicate the findings (see Nolte et al., 2004).

Whilst we accept that volume conduction is indeed an inherent caveat of coherence analysis, we have made many efforts in our approach to minimize its impact on our findings. Distance between recording sites is not our only control and we would like to list them below (see also Discussion section) as well as describing the two new analyses we have now performed.

1) Rather than using a common reference electrode, our recordings were bipolar, with each recording electrode being locally and independently referenced. This method has been described as being an effective method in reducing the general reference electrode issue (Kajikawa and Schroeder, 2011).

2) By recording simultaneously from hippocampus and multiple cerebellar regions we have been able to demonstrate that the observed coherence is non-homogenous among the different cerebellar lobules in contrast to what one would expect if theta was volume conducted from a common location.

In addition to these previous points and following reviewer’s comments and suggestions, we now have calculated the imaginary coherence and performed single cell recordings from the cerebellum.

3) The findings described in Figure 4 are replicated with high similarity when imaginary coherence is calculated (see new Figure 4—figure supplement 3), again suggesting that the described coherence is not generated by volume conduction.

4) Finally, we have conducted a new series of technically challenging experiments in which we have recorded single-unit activity in the cerebellar cortex alongside with LFP in the hippocampus (see new Figure 3—figure supplement 4) of head fixed L7-ChR2 mice. We took advantage of this mouse line to allow positive photo-identification of cerebellar Purkinje cells during recordings (determined by cell responses to blue light illumination; Cf Chaumont et al., 2013) (Figure 3—figure supplement 3). Interestingly, we found that 16 of 22 cells (6 mice) recorded in cerebellar lobule VI showed a robust response during optical illumination. Of these 16 cells, almost one-third (31%) were found to be significantly phase-locked to hippocampal theta oscillations. This finding also argues against a volume conduction origin of the coherence described in this manuscript (see new Results section).

6) [Related to Essential Point 1] The comparison of freely moving and virtual-reality is tricky. The learning curves look very different in Figure 4B and Figure 6C. Please provide a comparison between learning curves between linear track and virtual reality conditions. This impact on the choice of the epochs with similar behavior performances to compare in Figure 7. How similar is this behavior? Is it simply running speed? The VR animals seem overtrained while the freely moving seem to be learning during the session; this may result in differences of hippocampal engagement, beyond differences of sensory context.

Following reviewer’s suggestion we performed an individual analysis of the behaviour of each mouse during the virtual reality training. Among the 6 trained mice, two improved their performances and four remained with stable performances. Therefore, we now have analysed these mice separately (see Figure 4—figure supplement 4). The different examples show different patterns of activity and coherence. Interestingly, as it was the case during the linear track training, coherent activity between Crus I and hippocampus appeared when mice learned the task.

7) [Related to Essential Point 2] It is a bit hard to make sense of the time-lapse of the cerebellum; from the Figure 1—figure supplement 2, it seems that the cerebellar nuclei is infected 18 hours after the primary infections observed in the structures which project to the cerebellum, suggesting that the cerebellum is one synapse upstream these hippocampus-projecting regions. Is any of the structures labelled at 30h known to receive inputs from the cerebellar nuclei?

The fact that at 48h, labeled cells in the DCN are sparsely located and their processes weakly stained could suggest the beginning of a third order infection cycle and thus tends to suggest that the main infection of the DCN starts at 58 h post infection and involves two relays to reach the hippocampus (Figure 1—figure supplement 4).

However, at 30h post infection, staining is mainly found in the medial septum diagonal band of Broca (MSDB), the lateral and medial entorhinal cortices and the perirhinal cortex as well as the supramammillary nucleus (SUM) (Figure 1—figure supplement 2). We now confirm that these structures correspond to first order as CTB staining was also systematically found in the same structures (see Figure 1—figure supplement 3).

Among these structures, the septum has been described to receive direct projection from the fastigial nucleus in cats (Paul et al., 1973). In addition, we also found, in two mice out of four, RABV staining in the hypothalamus (including SUM in three animals) and raphe nucleus at 30h as well as CTB staining (see below). As these regions receive direct projections from the DCN (Teune et al., 2000), the few cells observed in the DCN 48h post infection could be related to the hypothalamic (potentially including the supramammillary nuclei (SUM)), and/or raphe nucleus staining observed at 30h.

Altogether, these results suggest multiple convergent pathways from the DCN to the hippocampus. Single relay pathways could be envisioned through the septum, the hypothalamus (potentially including the supramammillary bodies (SUM)) and the raphe nucleus (see new Results and Discussion sections as well as new supplementary figures associated with Figure 1). Other pathways including 2 relays through either the lateral and medial entorhinal cortex and/or the perirhinal cortex are also possible.

Alternatively, is it conceivable that some infection took place in the cerebral cortex overlying the hippocampus? The size of the pipette used for the injection is not provided. A picture of the cortex over the infection site would be useful for the reader to judge of the risk of virus leaks to the cortex.

The size of the pipette diameter is around 200 µm. We have now added this information in the Materials and methods section. Please note that in the majority of the cases, the cortex overlying the hippocampus has been severely damaged during the injection, which renders a possible leak very unlikely. Nevertheless, in the other cases, no virus leak was detected in the cortex overlying the hippocampus. We have now added a supplementary figure to illustrate the cortex over the infection site (see Figure 1—figure supplement 1E).

8) [Related to Essential Point 2] Regarding the anatomical study in subsection “A precise topography of the cerebellum regions projecting to the hippocampus”, second paragraph, the authors stated that the presence of labeling in the contralateral hippocampus at the earliest survival time is at 30h vs. 48h for only contralateral markings in cerebellar output nuclei DCN: that's more than the 12h required for an infection cycle, so they should start seeing cells as well in the ipsilateral cerebellar nuclei, no? The time lapse of the retroinfection isn't very clear: both hemispheres of the cerebellar cortex are infected at the same moment, while the DCNs are infected at different times (cf. Supplementary Figure 2)? The symmetry in the pattern of infection in the cerebellar cortex suggests connections going to the same area, but the timelapse of the infection installs some doubts. Please clarify.

The labeling at 48h in the DCN is too sparse to allow any quantitative comparison between ipsi and contra DCN (see Figure 1—figure supplement 3). At 58h, the labeling in the DCN is more abundant and stronger. At this time course, we did not find any difference between ipsi and contralateral staining in fastigial and dentate nuclei. The labeling in the interpositus remained weak in both sides (Figure 1—figure supplement 6).

9) [Related to Essential Point 2] I do not see some of the findings mentioned in the following text illustrated in Figure 1. "Rather, RABV/CTb-labeled neurons were found in two well described subcortical pathways leading to the DG of the hippocampus (first cycle of infection). One labeled pathway included the diagonal band of Broca and the septum. The other labeled hippocampal input pathway included the lateral entorhinal and perirhinal cortices (Figure 1—figure supplement 3)(Mosko et al., 1973; Dolorfo and Amaral, 1998; Witter, 2007.… "

Indeed as mentioned by the reviewer, the CTB labelling was not shown in the previous version of the manuscript. We have now added CTB labeling corresponding to the first order labelled neurons (new Figure 1—figure supplement 3). We also replaced the table (previous Figure 1—figure supplement 2) by the different supplementary figures mentioned above. We have therefore replaced the sentence by a better and more detailed description of the structures labelled at the different survival time (see subsection “Cerebellar projections to the hippocampus are precisely topographically organized”).

10) [Related to Essential Point 2] While the tracing was performed by injections in the dentate gyrus, recordings were performed in another region, the CA1 area. Please clarify the motivation of these choices and discuss how this may impact the result.

We chose to inject into the dentate gyrus for two reasons: 1) to target the maximal number of potential afferent inputs of the hippocampus and 2) we wanted to minimize the risk of a leak in the regions overlying the hippocampus. It is important to note that the site of viral injection involves both the stratum lacunosum-moleculare and the molecular layer of dentate gyrus which correspond to the main axonal entrance to CA1 and DG. The LFP recording was centered on CA1 since we previously described the influence of cerebellar plasticity on CA1 neuronal activity (Rochefort et al., 2011). We now have added a paragraph in the Results section of the manuscript to add these precisions.

11) [Related to Essential Point 2] Subsection “A precise topography of the cerebellum regions projecting to the hippocampus”, last paragraph: Several pieces of anatomical information are missing from the analysis.

a) What is the location of first order neurons in subcortical structures that are labeled by retrograde transport after hippocampal injections of RV?

We now present two supplementary figures describing all the regions containing first order neurons, 30h after hippocampal infection by rabies virus (new Figure 1—figure supplement 2) and confirmed by CTB (see new Figure 1—figure supplement 3). The main subcortical regions infected at 30h and thus projecting directly to the hippocampus are the Medial septum and diagonal band of Broca (MSDB), the lateral and medial entorhinal as well as the perirhinal cortices and the supramammillary bodies (see also answer to point 7 and 9).

b) What is the time course of the appearance of this first order labeling?

In our manuscript, the time course used to label first order neurons is 30h. We did not test a shorter survival time since this time is classically used in rodents tracing studies (e.g. Coulon et al., 2011) and allows visualization of a maximal number of directly projecting structures towards the injection site with a good intensity of labeling as confirmed by CTB labeling.

c) What is the time course of the appearance of neurons in fastigial, interpositus and dentate? Do they all appear at the same time in individual animals? Is the timing of their appearance fully consistent with all the neurons in the deep cerebellar nuclei being second order neurons labeled by retrograde transneuronal transport? Or is it possible that some of the neurons are third order neurons? In other words, do neurons in the fastigial nucleus label 8 hours before neurons in the dentate? If so, then this would be consistent with fastigial neurons being second order and dentate neurons being third order.

The labeling at 48h in the DCN is too sparse to allow any quantitative comparison between ipsilateral and contralateral DCN (see Figure 1—figure supplement 3). At 58h, the labeling in the DCN is more abundant and stronger. At this time course, we did not find any difference between ipsi and contralateral staining in fastigial and dentate nuclei. The labeling in the interpositus remained weak in both sides (see Figure 1—source data 1).

12) [Related to Essential Point 2] Is the time course of the first labeled neurons in the deep nuclei fully consistent with their being second order neurons? Or is it possible that even the earliest labeled neurons are third order neurons that are labeled by transneuronal transport through two subcortical neurons?

According to CTB staining, we consider that 30h post-infection stained cells represent first order neurons and that 48h post-infection stained cells form second order neurons. It is thus possible that the few labeled neurons observed in the DCN at 48h post hippocampal infection represent second order neurons (see a detailed answer to this point in point 7 and in the Results section of the manuscript).

13) [Related to Essential Point 2] What is the time course of labeled purkinje cells in different regions of the cerebellar cortex? Do neurons in the vermis label ~8 hours prior to neurons in the hemisphere?

Labeled neurons were found in both the hemisphere and the vermis at 58h post infection (see new Figure 1—figure supplement 5).

14) [Related to Essential Point 2] As the anatomical data currently stand, their analysis and/or presentation are incomplete. The following statement is neither clearly stated in the Results section nor clearly illustrated. "Notably, all of these regions contained RABV+ cells at 48h p.i., and thus they cannot be excluded as potential routes towards the hippocampus."

The revised manuscript now contains a paragraph in the Results section presenting the different structures stained at 48h. The table (Figure 1—figure supplement 4) describing these regions is also updated. Interestingly, among the different regions labeled at 48h post infection, several midbrain and pontine regions such as the PAG, the Nucleus Incertus and the LtDG are known to receive direct projections from the DCN and could therefore represent putative relays between DCN and hippocampus.

15) [Related to Essential Point 2 and 3] Some additional methods are needed. It was not clear to me how recordings were consistently selected for inclusion in various analysis. For example, the last paragraph of the subsection “Cerebello-hippocampal physiological interactions in a familiar homecage environment”, lists 23 values from 13 mice. How were these sessions selected? Which mice did these sessions come from (up to two sessions per mouse?) and why would session numbers used in analyses differ between mice? This is just one example but the comment applies to statistics throughout the paper.

We apologized if the data used for the different analysis were not clear enough in the previous version of the manuscript. The difference between values and animals referred here it was due to the use of HPCright – cerebellar coherence and HPCleft – cerebellar coherence values from the same animal when both hippocampal electrodes were on target (after histological verification). In order to clarify this and reduce the potential impact of this pooling on the statistics we have now calculated and used the mean between these two values so each animal is represented only once. A detailed explanation has been added in the Results section of the new version of the manuscript.

16) [Related to Essential Point 3] The authors provided a measure of physiological interaction – namely theta oscillations between cerebellum and hippocampus. First they assessed the cerebello-hippocampal interactions with coherence in the home cage and then during the learning of a goal directed behavior in a linear track and in virtual reality conditions. In the third paragraph of the subsection “Cerebello-hippocampal interactions during the learning of a goal-directed behavior” and Figure 4E-F: The shift in mean theta power and peak frequency is reported significant in the text but not clear on the figure when no visible difference in the theta peak between trial 1 and 20 was observed. Also when looking at the numbers in this z-score power (subsection “Cerebello-hippocampal interactions during the learning of a goal-directed behavior”, third paragraph), the difference is barely visible and the SEM give overlapping numbers (1st trial between 1.59 and 1.69, 20th trial between 1.68 and 1.76 for example): the statistical test seems to be conflicting with these results. Please clarify.

Following our extensive re-analysis, the figures referred to by the reviewer are no longer present in the manuscript. We have endeavoured to ensure that all new analysis descriptions, statistics and results are clear to the reader. Indeed, changes in power are now better visualised after the trial by trial analysis averaged by distance from reward (Figure 4D-G). We have added information relating to the new findings in the Results section.

17) [Related to Essential Point 3] Some of the effects appear to rely heavily on smaller numbers of mice, although the authors do have quantitative explanations for the variability they observe (i.e. the recording distance from the midline in Figure 3F). Even so, Figure 3F seems to be primarily driven by two mice and in Figure 5, I was unsure of how the authors interpreted the high degree of variability in mean coherence in the Crus I data set.

Following re-analysis, we have now averaged coherence values obtained between the cerebellum and left/right hippocampus. Figure 3F (now Figure 3J) shows a strong correlation between theta coherence and electrode positioning within lobule VI. This correlation remains robust after our updated analysis in which we averaged down our coherence values so each data point corresponds to one animal. In terms of Crus I coherence variations, we have also conducted extensive reanalysis of this data set and we find that 3/4 animals show similar patterns of Crus I coherence change in the linear track task (Figures 4H and 5). Variations in recording electrode position within Crus I may have also contributed to variations in coherence levels as for lobule VI, but we were unable to accurately reconstruct the electrode positions with high enough resolution to systematically check this as we did for lobule VI.

18) [Related to Essential Point 3] The authors recorded the activity in both left and right hippocampus in all mice and they noted no difference (subsection “Cerebello-hippocampal physiological interactions in a familiar home-cage environment”, fourth paragraph) in the analysis of cerebello-hippocampal coherence. However, in Figure 4 to Figure 7 the data from different animals are indifferentially reported from data from the same animal. Indeed, the significant effects, for example for Crus I in Figure 5 and Figure 7D may be only due to a few values, possibly coming from only one or two animals; moreover it seems that some of the statistics are not performed using repeated-measure procedures. Please provide a clear view on left/right recording from each animal and clarify the statistics (e.g. averaging values from single animals). Overall, the number of mice seems a bit small for the task-related coherence analysis to evaluate reproducibility of the results.

As the reviewer points out, we observed no significant difference in values of coherence obtained between cerebellum and left or right hippocampus. The detailed comparison of HPC left vs. HPC right for each condition can be found in Figure 3—figure supplement 2 and Figure 4—figure supplement 1. Therefore, we have now changed the main figures and the statistical analysis presented in the manuscript to a single value of coherence between hippocampus and cerebellar regions per animal reflecting the average between left and right measurements when both are available.

Regarding the repeated-measure procedures they have been used when the data allowed for it; however, if this comment refers to its employment when comparing differences across combinations in a single condition (Figure 3 and Figure 4), each recording has been considered independent as they were locally referenced.

https://doi.org/10.7554/eLife.41896.027

Article and author information

Author details

  1. Thomas Charles Watson

    Neuroscience Paris Seine, Cerebellum, Navigation and Memory Team, CNRS UMR 8246, INSERM, UMR-S 1130, Sorbonne Universités, University Pierre and Marie Curie, Paris, France
    Present address
    1. Centre for Discovery Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom
    2. SimonsInitiative for the Developing Brain, University of Edinburgh, Edinburgh, United Kingdom
    3. Patrick Wild Centre for Autism Research, University of Edinburgh, Edinburgh, United Kingdom
    Contribution
    Formal analysis, Supervision, Validation, Investigation, Methodology, Writing—original draft
    Contributed equally with
    Pauline Obiang and Arturo Torres-Herraez
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5146-6981
  2. Pauline Obiang

    Neuroscience Paris Seine, Cerebellum, Navigation and Memory Team, CNRS UMR 8246, INSERM, UMR-S 1130, Sorbonne Universités, University Pierre and Marie Curie, Paris, France
    Contribution
    Validation, Investigation, Methodology, Writing—review and editing
    Contributed equally with
    Thomas Charles Watson and Arturo Torres-Herraez
    Competing interests
    No competing interests declared
  3. Arturo Torres-Herraez

    Neuroscience Paris Seine, Cerebellum, Navigation and Memory Team, CNRS UMR 8246, INSERM, UMR-S 1130, Sorbonne Universités, University Pierre and Marie Curie, Paris, France
    Contribution
    Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft
    Contributed equally with
    Thomas Charles Watson and Pauline Obiang
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6333-5623
  4. Aurélie Watilliaux

    Neuroscience Paris Seine, Cerebellum, Navigation and Memory Team, CNRS UMR 8246, INSERM, UMR-S 1130, Sorbonne Universités, University Pierre and Marie Curie, Paris, France
    Contribution
    Formal analysis, Validation, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Patrice Coulon

    Institut de Neurosciences de la Timone, CNRS and Aix Marseille Université, Marseille, France
    Contribution
    Resources, Supervision, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2996-405X
  6. Christelle Rochefort

    Neuroscience Paris Seine, Cerebellum, Navigation and Memory Team, CNRS UMR 8246, INSERM, UMR-S 1130, Sorbonne Universités, University Pierre and Marie Curie, Paris, France
    Contribution
    Validation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  7. Laure Rondi-Reig

    Neuroscience Paris Seine, Cerebellum, Navigation and Memory Team, CNRS UMR 8246, INSERM, UMR-S 1130, Sorbonne Universités, University Pierre and Marie Curie, Paris, France
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Writing—original draft, Project administration
    For correspondence
    laure.rondi-reig@upmc.fr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1006-0501

Funding

Fondation pour la Recherche Médicale (DEQ20160334907)

  • Laure Rondi-Reig

Agence Nationale de la Recherche (ANR-17-CE16-0019-03)

  • Laure Rondi-Reig

Centre National de la Recherche Scientifique

  • Patrice Coulon

Aix-Marseille Université

  • Patrice Coulon

Université Pierre et Marie Curie (ANR-10-LABX-BioPsy)

  • Laure Rondi-Reig

Université Pierre et Marie Curie (ANR-11-IDEX-0004-02)

  • Laure Rondi-Reig

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by the Fondation pour la Recherche Médicale DEQ20160334907-France, by the National Agency for Research ANR-17-CE16-0019-03 (LRR), by the CNRS and Aix-Marseille Université through UMR 7289 (PC). This work also received support under the program Investissements d’Avenir launched by the French Government and implemented by the ANR, with the references, PER-SU (LRR) and ANR-10-LABX-BioPsy (LRR). The group of LRR is member of the Labex BioPsy and ENP Foundation. Labex are supported by French State funds managed by the ANR within the Investissements d'Avenir programme under reference ANR-11-IDEX-0004–02. We thank Roxanna Ureta for help with histology, Lilith Sommer for help with behavioral experiments, Gregory Sedes and Nadine Francis for help developing analysis codes, and Richard Apps for his insightful comments. We are grateful to Richard Hawkes and Izumi Sugihara for generously providing the aldolase C antibody. Finally, we thank all members of the CEZAME team for helpful discussions of the experiments and manuscript.

Ethics

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Ministère de l'Enseignement Supérieur et de la Recherche. All of the animals were handled according to approved institutional animal care agreement B75 0510. The protocol was approved by the Committee on the Ethics of Animal Experiments (expérimentation animale n°5) protocols #00895.02 and APAFIS#4315-2016042708195884v1. All surgery was performed under isoflurane gas anesthesia, and every effort was made to minimize suffering.

Senior and Reviewing Editor

  1. Richard B Ivry, University of California, Berkeley, United States

Reviewer

  1. Peter Strick, University of Pittsburgh

Publication history

  1. Received: September 10, 2018
  2. Accepted: May 30, 2019
  3. Version of Record published: June 17, 2019 (version 1)

Copyright

© 2019, Watson et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,115
    Page views
  • 198
    Downloads
  • 3
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)