Cerebellar control of a unitary head direction sense
1. Summary
The head-direction (HD) system, a key neural circuit for navigation, consists of several anatomical structures containing neurons selective to the animal’s head direction. HD cells exhibit ubiquitous temporal coordination across brain regions, independently of the animal’s behavioral state or sensory inputs. Such temporal coordination mediates a single, stable, and persistent HD signal, which is essential for intact orientation. However, the mechanistic processes behind the temporal organization of HD cells are unknown. By manipulating the cerebellum, we identify pairs of HD cells recorded from two brain structures (anterodorsal thalamus and retrosplenial cortex) that lose their temporal coordination, specifically during the removal of the external sensory inputs. Further, we identify distinct cerebellar mechanisms that participate in the spatial stability of the HD signal depending on sensory signals. We show that while cerebellar protein phosphatase 2B-dependent mechanisms facilitate the anchoring of the HD signal on the external cues, the cerebellar protein kinase C-dependent mechanisms are required for the stability of the HD signal by self-motion cues. These results indicate that the cerebellum contributes to the preservation of a single and stable sense of direction.
2. Data description
Data were obtained from the anterodorsal thalamic nucleus (ADN) and/or the retrosplenial cortex (RSC) of 25 adult male mice (six L7-PKCI with seven control littermates; six L7-PP2B with six control littermates), aged from three to six months during the recordings. Mice were housed singly in a standard transparent polycarbonate cage (25x16x14cm; LxWxH) on a 12-hour light/dark cycle. All the recordings were performed during the light cycle. In L7-PKCI mice, the pseudo-substrate PKC inhibitor (PKCI) was expressed under the control of the pcp-2 (L7) gene promoter, ensuring a selective suppression of PKC-dependent mechanisms in Purkinje cells. In L7-PP2B mice, the selective deletion of PP2B in Purkinje cells was obtained using the Cre-loxP-system (flox/flox, cre/+).
Data is provided in two separate sets for separate purposes. Note that, in both datasets, not all of the cells are necessarily HD cells. The dataset contains a broader cell population involving all the HD cells which were detected using different methods. Therefore a cell may or may not be identified as an HD cell using different methods (i.e. GLM, spike time shuffling and using vector length or wwatsson U2 in one or multiple sessions).
3. Dataset1
3.1. General description of dataset
• This dataset contains unique single cells (no duplicate cells accross different days of recordings). This is the dataset used to generate all the comparisons for single-cell properties (including Figures 1, and 3B-H of the PNAS paper).
• Each .mat file in this dataset represents one set of recordings from one or multiple cells (rows) across six sessions (columns).
• For each .mat file, the data is provided as two “Strcut” type variables: “cInputs” contains raw data involving spiking timestamps for each cell and general tracking information (animal’s head direction and timestamps). “cOuputs” contains the analysis output (e.g. involving firing or tuning information) calculated from that raw data. Each cell within a Strcut represents a variable in M×N (Cell#×Session#) format, where the rows represent unique cells and the column represents the session number (all data recorded in six sessions; for description see figure S5A in the associated published paper). Therefore, note that one row represents the same cell across different sessions. If a cell is not recorded in a session, the values are assigned as NaN or the associated matrix is left empty.
3.2. Elements in .mat files
• cInputs.spktTot: Spike timestamps for recorded cells
• cInputs.spkdTot: Animal’s momentary head direction (in angle) at the time of spiking
• cInputs.postTot: Animal’s tracking timestamps (in seconds; note that neuralynx doesn’t sample the tracking data at a constant rate but very close to 25 Hz. Therefore actual timestamps are provided. When a constant sample rate is required, timestamp data can be interpolated).
• cInputs.posdTot: Ainmal’s momentary head direction (in angle) aligned with the tracking timestamps
• cOutputs.sessionName: The specific name for each set of recordings (information regarding mice, number, and date can be found here)
• cOutputs.tetNo: The tetrode number from which a cell is recorded. Each mouse is implanted with 2x Microdrive, each carrying a bundle of 4 or 8 tetrodes. For detail see the table below. In short, for mice numbers 107-108, all tetrodes are recorded from ADN. For mice numbers 122-161, tetrodes 1-4 are recorded from ADN, and tetrodes 5-8 from RSC. For mice numbers 174-177, tetrodes 1-8 are recorded from ADN, and tetrodes 9-16 are from RSC.
• cOutputs.ClsN: The cell number associated with automatic spike sorting which was preserved after manual clustering (Note that the same cell –columns- might be ascociated with different cluster number across different sessions).
• cOutputs.sesLimTime: All sessions are recorded for 10 min. However, if any session is disrupted, values of experiment length are shown here. An empty value means the session length is the normal 10-minute-long.
• cOutputs.isoDist: Spike-cluster isolation distance (quality of spike sorting for a given cluster representing a cell)
• cOutputs.meanRate: Overall mean firing rate in Hz (this value is simply the division of the number of spikes over the length of the session).
• cOutputs.bursting: Ratio of spikes detected as bursts (interspike intervals at 6ms) over the whole number of spikes.
• cOutputs.glmType: Type of cell detected by GLM framework (Generalized linear model; Matlab toolbox glmfit). The spiking activity of a cell was modeled as a discrete-time point process in the GLM framework assuming HD, AHV, or Speed component. Cells tagged as “H”, for HD, “A” for AHV, and ”S” for speed cells. conjunctive cells are tagged with these letters together, for example, “HS” is a cell detected as a conjunctive HD-by-speed cell. Note that a cell might be detected HD/speed/AHV in one session but not in another session.
• cOutputs.pShuffleVecLen: p-value for HD cells (null hypothesis: cell is not HD cell). An alternative method to GLM for the identification of HD cells, this p-value compares the value of HD vector length with chance level statistics. Here, the spike train of individual cells was time-shifted along the tracking samples by a variable randomly chosen between 20 s and the total trial length minus 20 s. The shifted values that were passing the length of the trial were wrapped to the beginning. Such a shuffling procedure decouples the spikes from the animal’s head direction while keeping the spike timing relations intact. For the shuffled spikes, the value of the animal’s momentary head direction was estimated from HD samples and the vector length value was measured. This procedure was repeated 100 times, and the distribution of the outcomes was generated. The 95th or 99th (or else) percentile value of this shuffled distribution can be used as the classification criterion.
• cOutputs.pShuffleWU2: Same procedure as “pShuffleVecLen” but instead of vector length, the value of Watson U2 was used as a criterion for measuring the p-value (statistical significance identifying HD cell).
• cOutputs.DR: HD distributive ratio (See details in the PNAS paper).
• cOutputs.meanAngle: Preferred firing direction (PFD), the angle of the resultant vector obtained from the head direction angles during the spiking of a cell.
• cOutputs.vecLen: Mean vector length, the magnitude of the resultant vector obtained from the head direction angles collected during the spiking of a cell.
• cOutputs.InfoBtsPrSpk: Shannon information for HD cell activity as bits/spike.
• cOutputs.stab5min: Intra-session directional stability (Pearson correlation) of HD tuning curve calculated from 5-minutes sub sessions.
• cOutputs.stab2min: Intra-session directional stability (average Pearson correlation) of HD tuning curve for 2-minutes sub-sessions.
• cOutputs.polTunCrv: A bin estimate of HD tuning curve in sixty bins of 6° (number of spikes in each head direction bin divided by the amount of time spent in that bin, smoothed by a gaussian kernel).
4. Dataset2
4.1. General description of dataset
• Contains unique cell pairs but not unique single cells (accordingly duplicate cells accross days of recordings may exist).
• This dataset contains two folders, one folder contains information about all single cells used in this dataset (folder: SingleCellsInfo) and the second folder contains the spike cross-correlation for simultaneously recorded cell pairs (folder: Xcorrs_CellPairs).
• The purpose of this dataset (compared to dataset1) was to maximize the number of cell pairs without having unique single cells across different trials of recordings from the same animal. For instance, if cells 1 and 2 are recorded on day 1, and cells 1 and 3 are recorded on day 2, in dataset 1, cell 1 is only used from day 1 ( or the day with the best spike-cluster quality for that cell). However, in dataset 2, cell 1 is represented twice as it provides two unique cell pairs: pair 1×2 and pair 1×3.
• .mat files in the folder “SingleCellsInfo” involve the same information as dataset1 for all the cells used in this dataset.
• .mat files in folder “Xcorrs_CellPairs” include the spike cross-correlation for all cell pairs in each mouse.
4.2. How to use the dataset
The information from “SingleCellsInfo” can be used to identify the HD cell pair. This information is processed in the sheet “batch” of the excel file (“Excel_Cell selection\ Paired_circ50_ALLids7.xlsx”) and HD cells, mouse genotype groupings, and the recorded brain structures are marked as well as the parameters that are used to select the cell pairs. These parameters include “idHDs” which is the id of HD cell pairs according to criteria chosen in excel-cell $DL$1-$DX$1 within both “C1_Basics” (information for the first cell in a pair) and “C2_Basics” (information for the second pair in a pair) are indexed based on their properties. Using these information, one can index the cells, like the example that is provided in the “Codes\ PairInfo_7121.mat” file. This mat file is called by “xCorr_Paired4.m” to create cross-correlation graphs for the indexed population. For instance, the current “PairInfo_7121.mat” file has indexed HD cells by assuming GLM as the classifying method, a “Distributive ratio - DR” criteria, and using a minimum isolation distance value of 10. One can simply change the “InputPath” in the code “xCorr_Paired4” and run this code to see a sample of results (which is used in the paper).
For more details, see the README.docx and/or the supporting information of the published paper online at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2214539120/-/DCSupplemental.*
3. Cite as
M Fallahnezhad, J Le Méro, X Zenelaj, J Vincent, C Rochefort, and L Rondi-Reig. (2023). Cerebellar control of a unitary head direction sense. PNAS.
