Optogenetic Frequency Scrambling Of Hippocampal Theta Oscillations Dissociates Working Memory Retrieval From Hippocampal Spatiotemporal Codes Part 3

Optogenetic pacing of MS neurons leads to non-physiological theta synchrony

Surprisingly, 8 Hz optogenetic stimulations that are associated with the pacing of theta oscillations led to decreased performance when applied during either encoding or retrieval of episodic memory in the NPOR task, as well as spatial working memory retrieval in the DNMTS task. To further understand the effects of pacing theta on hippocampal physiology, we analyzed phase locking of theta oscillation before and during 8 Hz optogenetic stimulations (Supplementary Fig. 7a) and found that such stimulation led to increased phase synchrony across stimulation epochs (Supplementary Fig. 7b). Furthermore, we analyzed the cross-correlation of theta oscillations across dorsal CA1 (~1 mm distance between electrodes along the septotemporal axis (Supplementary Fig. 7c) and found that pacing theta rhythms at 8 Hz led to an increased cross correlation between those two electrodes (Supplementary Fig. 7d).

Genetics is the science that studies genetic inheritance and variation, and memory is an aspect of great concern. So, what does genetics have to do with memory? This article will explore issues related to memory from a genetic perspective.

First of all, we know that genetic genes are the basis for human physical and intellectual development, and whether these genes are superior will affect people's memory. The latest research shows that people's IQ and memory are affected by genetic factors. Research shows that genetic genes play a crucial role in intelligence and memory. Therefore, some people are born with stronger memories, while others need to work hard and train to improve their memory.

Secondly, environmental factors are also important factors in memory development. We not only need the support of genetic genes but also need to improve our memory through scientific training and good living habits. For example, we can insist on exercising every day, getting enough sleep, following meal-matching principles, etc., which can help improve our brain function and memory.

Finally, ordinary people, do not overemphasize the influence of genetic factors on memory development. Because genetic factors are not absolute, people's memory and intelligence can still improve themselves by improving their efforts and training. We should always maintain a positive and continuous learning attitude, and constantly explore and develop our memory potential. I believe that only in this way can we become smarter and have a greater sense of accomplishment. To sum up, the relationship between genetics and memory is very close, but we should not overemphasize genetic factors, but should constantly improve our memory and intelligence through self-improvement and training.

It can be seen that we need to improve our memory. Cistanche deserticola can significantly improve memory because Cistanche deserticola is a traditional Chinese medicinal material with many unique effects, one of which is to improve memory. The efficacy of minced meat comes from the various active ingredients it contains, including acid, polysaccharides, flavonoids, etc. These ingredients can promote brain health in various ways.

Click know 10 ways to improve memory

Discussion

Within the septohippocampal system, the exact causal relationships between (1) MS activity, (2) hippocampal oscillations, (3) hippocampal neuron activity, and (4) behavior, including memory, remain an active area of research. In particular, whether and how the MS supports encoding of place and time in the hippocampus, as well as its specific contribution to memory function, remain unclear. Here, we leveraged more recent techniques that allowed us to record >1000 neurons while performing optogenetic stimulation of MS neurons with a sub-second resolution to control theta rhythms and assess their role in hippocampal physiology and memory. Importantly, using an excitatory opsin and either scrambled or 8 Hz stimulations, we were able to consistently and robustly abolish or pace hippocampal theta, respectively. 

Compared to MS inhibition using either an inhibitory opsin or pharmacological compounds (e.g. muscimol), our approach enables a within-subject comparison of two opposed states (paced vs abolished theta) while maintaining activity levels in MS-PV cells. We combined a tone-cued linear track task that, together with information-theoretic approaches, allowed us to disentangle spatial and temporal hippocampal codes. This method alleviates the need for arbitrary thresholds, enabling a standardized approach to calcium imaging analysis. While a large portion of CA1 pyramidal neurons expressed a mixture of spatiotemporal codes, we focused our analyses on neurons tuned specifically to place, time, or distance traveled.

While spatial coding has been extensively investigated in CA1, temporal and distance codes have recently gained more interest. Temporal28,35–37 and distance37 codes have been extracted by clamping visuospatial cues or extracted analytically using generalized linear models in virtual reality paradigms26,27,38. Here, we propose an approach to disentangle spatial, temporal, and distance codes using an information theoretic approach which, together with our cuedalternation task, enables the analysis of real-world spatiotemporal and multiplexed codes in freely moving animals. A large amount of CA1 pyramidal neurons encode multiplexed information of location, distance, and time as previously reported26,27,38 in addition to self-motion signals such as acceleration, speed, and orientation60.

We found that frequency scrambling and 8 Hz optogenetic stimulations drastically abolished or paced theta rhythms, respectively, and led to decreased overall activity in a subpopulation of CA1 pyramidal cells while not causing significant changes in place cell activity, similar to previous reports using either pharmacological inhibition36,46,47 or optogenetic9,48 pacing of MS or septal inputs to the hippocampus. As MS neurons are known to be the primary driver of hippocampal oscillations, we expected that MS stimulation would be associated with disrupted time or distance cell activity in conditions of reduced theta rhythms, but no changes occurred when theta was abolished or paced. Additionally, the pacing of hippocampal oscillations to 8 Hz did not lead to changes in the quality of spatial codes as reported previously9 and did not alter temporal codes as observed in our behavioral paradigm.

Although it was reported that time (but not place) cells could depend directly on the medial entorhinal cortex (MEC) inputs61, recent experimental evidence suggests that MEC lesions do not lead to any alterations in hippocampal time cell physiology62. Notably, a more recent investigation found that optogenetic pacing of the MS did not disrupt the spatial codes of grid cells in the entorhinal cortex63, further suggesting that MS activity is not directly involved in spatial codes within the hippocampal formation. Together with our results, this indicates that temporal codes could either be the result of MS-dependent computations within the entorhinal cortex63 (1) or be generated intrinsically within the hippocampus itself (2).

While early studies found that the MS played an essential role in maintaining time cells and supporting working memory36, recent experimental evidence shows that the exact contribution of time cells to working memory performance could be less than previously thought62. Since MS-PV neurons likely maintain significant activity levels during frequency scrambling optogenetic stimulation (in contrast to inhibition approaches), our results strongly support the role of precisely timed septal activity in supporting working memory, as hypothesized previously. Although we observed reduced memory performance when using optogenetic stimulation during retrieval, stimulation during the encoding phase of the DNMTS tasks was not associated with reduced memory in the DNMTS task. Pacing theta oscillations at 8 Hz during either the encoding or retrieval phase of the NOPR task led to impaired memory retrieval. Our electrophysiological analyses further reveal that such stimulation caused distant regions of dorsal CA1 to be entrained at the same phase, with non-physiological phase locking. While we did not directly investigate the causal relationship between phase changes and memory performance, spike timing, and phase precession were also found to be associated with altered working memory function despite leaving place coding intact64. Moreover, phase locking of pyramidal neurons to theta oscillations has previously been shown to be a predictor of memory performance65.

Notably, one limitation of our approach to establishing temporal and distance tuning properties is that it requires four times more sampling than a regular linear track which increases the chances of photobleaching when adding stimulation conditions to the experimental timeline. While we found that spatial codes remained unaffected by optogenetic stimulations within the same session, more recent imaging hardware that includes sensors with increased sensitivity could help prevent photobleaching and allow for longer recording sessions, thus sampling of temporal and distance cells along with stimulations. We also found that the pool of neurons that significantly and specifically encode only one variable is small in comparison to the number of conjunctive neurons, making sampling requirements for these highly specific neurons much higher.

We provide experimental evidence that manipulation of MS does not alter the locomotor speed and that, conversely, locomotor speed does dictate theta frequency. Even though the place and time codes were preserved during our MS stimulations, a portion (~6%) of cells were directly modulated and could account for memory impairments observed in our behavioral assays. PV interneurons in the hippocampus have previously been reported to be part of a microcircuit essential in regulating memory consolidation66,67, and our optogenetic manipulations could be associated with the silencing of a portion of these cells. Moreover, although we did not observe changes in frequency bands other than theta, we cannot exclude that CA1 pyramidal neuron spike timing could be drastically altered when scrambling theta oscillations, while 8 Hz stimulation was shown to not result in increased hippocampal activity18, which could explain the differential effects of those stimulation patterns on working memory performance. The memory impairments we observed were likely non-cholinergic: firstly, in our immunohistological experiments, we found virtually no expression of ChrimsonR in ChAT neurons of the MS. Secondly, our optogenetic stimulations were not associated with changes in hippocampal ripples, while previous reports indicate that stimulation of ChAT neurons is associated with reduced ripple frequency45,57. ChrimsonR transfection of MS VGLUT2 neurons is also unlikely since activation of these neurons is associated with a direct increase in locomotor activity 68, which we did not observe. 

Although the precise temporal arrangement of CA1 pyramidal spikes could explain, at least partially, the effects of MS stimulation on memory, alternative mechanisms should be taken into consideration. Notably, in addition to the hippocampus and the entorhinal cortex, MS PV neurons project to the retrosplenial cortex3 and could be responsible for some of the memory impairments observed here.

In summary, using calcium imaging, optogenetics, and electrophysiology, we found that theta rhythms can be paced or abolished using stimulation of the MS. Such stimulation impaired episodic as well as working memory retrieval. These effects were non-cholinergic and did not disrupt hippocampal ripple activity. Finally, while a small fraction of hippocampal neurons responded directly to optogenetic stimulation of the MS, place, time, and distance cells were not disrupted by manipulations of theta oscillations. Together these results suggest that while MS input to the hippocampus plays an essential role in memory, multiplexed codes in CA1 pyramidal neurons might not be the direct substrate for such functions.

Methods

Animals

All procedures were approved by the McGill University Animal Care Committee and the Canadian Council on Animal Care (protocol 2015- 7650). A total of n = 41 male (n = 20) and female (n = 21) 8–16 weeks old, B6;129P2 PV-Cre mice (Jackson Laboratory, RRID: IMSR_JAX:017320) were used in this study. n = 5 mice were used for combining optogenetics with calcium imaging; n = 3 mice were implanted for calcium imaging, optogenetics, and electrophysiological controls; n = 4 mice were used in electrophysiological studies; n = 29 mice were transfected and implanted for behavioral assays. Mice were housed individually on a 12-h light/dark cycle at 22 °C and 40% humidity with food and water ad libitum. All experiments were carried out during the light portion of the light/dark cycle.

Adeno-associated viral vectors

Adeno-associated virus AAV5.CamKII.GCaMP6f.WPRE.SV40 (Addgene # 100834, obtained from the University of Pennsylvania Vector Core) was used in all calcium imaging experiments. Adeno-associated virus (AAV) of serotype dj (hybrid capsid created from eight different AAV serotypes) AAVdj-hSyn-ChrimsonR-tdTomato were obtained from the Vector Core Facility at Oregon Health and Science University in Portland, Oregon. Although he is a housekeeping gene, we did not observe transfection in cholinergic cells (see “Results” and Fig. 2b–e). An eYFP construct without the ChrimsonR sequence was used as a control (termed YFP control in this manuscript).

Surgical procedures

Mice were anesthetized with isoflurane (5% induction, 0.5–1 % maintenance) and placed in a stereotaxic frame (David Kopf Instruments). Body temperature was maintained with a heating pad, eyes were hydrated with gel (Optixcare), and Carprofen (10 ml/kg) was administered subcutaneously. The skull was completely cleared of all connective tissue and small craniotomies were performed using a dental drill for subsequent injection or implant.

Viral injections. All viral injections were performed using a glass pipette connected to a Nanoject III (Drummond) injector. 500 nl of AAVdjhSyn-ChrimsonR-tdTomato (or eYFP control) was delivered into the MS at a rate of 1 nL/s, at the following coordinates based on reference mouse stereotaxic atlas69 (distance from Bregma in mm): anteroposterior (AP) 0.85, mediolateral (ML) 0, dorsoventral (DV) −4.50 using a 5° angle in the ML plane. After surgery, animals were monitored until recovery.

Fiber optic implant. Two weeks post-injection, mice were anesthetized for implantation following the same surgical procedure. A 200 μm diameter fiber optics with ceramic ferrule (Thorlabs) was implanted at the same coordinates. Implants were cemented in place using C&B-Metabond® (Patterson dental). Black nail polish was applied over the dental cement to block light emission during optogenetic stimulation.

Electrode implants. An array of 7 tungsten microelectrodes (~1 MΩ impedance) was lowered in dorsal CA1 spanning through stratum pyramidale (pyr), stratum radiatum (rad), and stratum lacunosum moleculare (lm). Screws placed in the bone above the frontal cortex and cerebellum served as ground and reference, respectively. Following electrode, ground, and reference placement, dental cement was applied to secure the implant permanently to the skull.

Implant for calcium imaging. We injected the AAV5.CamKII.GCaMP6f virus (200 nL at 1 nl s−1 ) in hippocampal CA1 using the following coordinates: anteroposterior (AP) − 1.86 mm from bregma, mediolateral (ML) 1.5 mm, dorsoventral (DV) 1.5 mm. Two weeks following the injection, mice were anesthetized with isoflurane and the skull was cleared. A <2 mm diameter craniotomy was performed in the skull above the injection site. An anchor screw was placed on the posterior plate above the cerebellum. The dura was removed, and the portion of the cortex above the injection site was aspirated using a vacuum pump until the corpus callosum was visible. These fiber bundles were then gently aspirated without applying pressure on the underlying hippocampus, and a 1.8 mm diameter gradient index (GRIN; Edmund Optics) lens was lowered at the following coordinates: AP − 1.86 mm from bregma, ML 1.5 mm, DV 1.2 mm. The GRIN lens was permanently attached to the skull using C&B-Metabond (Patterson Dental), and Kwik-Sil (World Precision Instruments) silicone adhesive was placed on the GRIN to protect it. Four weeks later, the silicone cap was removed and CA1 was imaged using a miniscope mounted with an aluminum base plate while mice were under light anesthesia (<0.5% isoflurane) to allow the visualization of cell activity. When a satisfying field of view was found (large neuronal assembly, visible landmarks), the base plate was cemented above the GRIN lens, the mini scope was removed, and a plastic cap was placed on the base plate to protect the GRIN lens.

Simultaneous calcium imaging and electrophysiological recordings. To control for the effects of GRIN implants on hippocampal theta as well as cross-talk between mini-scope and optogenetic excitation lights, we attached tungsten micro-electrodes to GRIN lenses. To this end, we placed GRIN lenses horizontally under a low-magnification microscope in a dust-free environment. A tungsten microelectrode was gently placed on the top edge of the GRIN lens using a micromanipulator. We used the known diameter of the GRIN lens as a reference unit to estimate the desired protrusion of the electrode (~50 µm, further assessed digitally after taking a microphotograph of the preparation) by our planned implantation coordinates. Small amounts (~50 µL) of superglue were deposited on the top edge of the GRIN lens with the electrode in place and left to dry for ~10 min, before applying the next layer of glue. Five thin layers were used to maintain the electrode attached to the GRIN lens. After implantation of these GRIN-electrode assemblies using the protocol described above, the protruding wires were gently bent and hidden under a protective cap (1–1.5 cm high) retrieved from the pear of a manual suction pipette as a replacement for Kwik-Sil.

In vivo behavioral procedures

Habituation. Mice were gently handled for ~5 min over one week, with progressive habituation to the plugging procedure (fiber optic, miniscope, and electrophysiological pre-amplified tethers). Animals were then water-scheduled (2 h access per day).

Miniscope recordings. Miniscopes (V3) were assembled using open-source instructions (miniscope.org). Imaging data were acquired using a CMOS imaging sensor (Aptina, MT9V032) and multiplexed through a lightweight coaxial cable. Data was acquired using a data acquisition (DAQ) box connected via a USB host controller (Cypress, CYUSB3013). Animal behavior was recorded using a consumer-grade webcam (Logitech) mounted above the environment. Calcium and behavioral data were recorded using miniscope.org, source DAQ custom acquisition software. The DAQ simultaneously acquired behavioral and cellular imaging streams at 30 Hz as uncompressed.avi files of 1000 frames for 15-minute recording sessions, along with corresponding timestamps to enable precise temporal alignment of behavioral and calcium imaging data.

In vivo electrophysiological recordings. Following 1 week of postsurgical recovery and one week of habituation to the tethering setup, LFP from implanted mice was recorded. All recorded signals from implanted electrodes were amplified by the tether pre-amplifier before being digitized at 22 kHz using a digital recording system (Neuralynx, USA). Recordings of each channel signal were saved along with video recordings and TTL signals from the Laser Diode for subsequent analysis.

Optogenetic stimulation. Laser stimulation was delivered through a fiber optic cord (200 μm diameter) using a laser diode fiber light source (Doric Lenses). Light intensity was calibrated and wavelength corrected using the Power Meter Bundle with the PM100D Console and S130C Slim Photodiode Sensorlight (Thorlabs). Every pacing stimulation (including 8 Hz) was performed using 5 ms square pulses. To apply scrambled light stimulation, we used an Arduino microcontroller to generate a white noise oscillation signal directly fed into the analog input of the laser diode driver. To perform randomly selected frequency stimulation, we used a standard 5 s ON, 5 s OFF protocol, but for each stimulation epoch, we used an Arduino microcontroller to randomly select a stimulation frequency in the theta band. When applying optogenetic stimulation during behavior, a loose piece of heat shrink tubing was fitted around the junction between the patch cord and the mouse ferrule implant to limit visible light emission. Light intensities are expressed as nominal power, as measured at the tip of the fiber optic implant–cord assembly (before surgical placement), and corrected for the appropriate wavelength.

Behavioral assays

Sequential tone linear track. Mice were water-scheduled and could access water only for 2 hours per day from 6–8 p.m. To disentangle spatial, temporal, and distance codes, we build a 134 cm long linear track using medium gray Lego® bricks, which allows easy modifications and implementations without the need to permanently modify the structure of the maze. Pyroelectric sensors were placed at each end of the linear track and were connected to an Arduino microcontroller. Each detection triggered a new tone in sequence, indicating the progress to reward delivery. The following tones were used and delivered using a piezo speaker: 1 s beeps at 2000 Hz, 250 ms beeps at 3000 Hz, and a continuous tone at 4000 Hz. When the last, continuous tone was triggered, a reward (10% sucrose in water) was delivered at the starting end of the linear track in the cap of a 15 mL Falcon tube. Changes in running direction before triggering the next tone were considered errors and did not trigger reward delivery. Performance was measured as the number of correct trials (with no error) divided by the number of total trials.

Novel object place recognition. On the first day, mice were allowed to freely explore a 45 × 45 cm dark gray open field that contained visual cues (large white horizontal and vertical gratings) on its walls for 10 min. On the second day, two identical objects (helping hand base) were presented for 10 min. On the third day, mice were allowed to explore the same open field while the location of one of the two objects was displaced. To control for potential spatial preference biases, both the initial, as well as the displaced position of objects, were randomized. Mice have attributed a random testing order that was kept identical throughout the three days of testing. Behavior was recorded with a video camera (Logitech) and analyzed offline. Behavioral analysis was performed blind to the genotype and treatment. Object explorations were defined as epochs where mice have their nose within 1 cm of an object. RI was computed as the following:

Automated delayed non-match to sample task. Mice were water scheduled and trained in a continuous T-maze to a delayed non-match to sample task. Briefly, each trial was divided into two phases: sample and test. In the sample phase, one arm was blocked and mice were forced to explore the opposite arm, where they received a 50 µL of 10% sucrose water reward. After completing the sample phase, mice were subjected to a delay (10 s) in the starting compartment. Then, during the test phase, both arms could be explored, but only the opposite (unexplored) arm was baited so that mice had to alternate locations between the sample and test phases. Mice were subjected to 10 trials (sample + test) per day, for 10 consecutive days, and the daily success rate was calculated as the number of correct trials divided by the total number of trials.

Post-mortem histological analyses

After completion of behavioral testing, mice were deeply anesthetized with a mixture of ketamine/xylazine/acepromazine (100, 16, 3 mg/kg, respectively, intraperitoneal injection) and perfused transcardially with 4% paraformaldehyde (PFA) in PBS. Brains were extracted and postfixed overnight in PFA at 4 °C and subsequently washed in PBS for an additional 24 h at 4 °C. Brains and sections were cryoprotected in a solution of 30% ethylene glycol, 30% glycerol, and 40% PBS until used. Each brain was then sectioned at 50 µm using a vibratome: every section was sequentially collected in 4 different 1.5 mL tubes to allow different analyses (electrode location, immunohistochemistry).

Immunohistology. Using one tube of collected brain sections (25% sampling) for each analysis, sections were washed for 3 × 5 min in PBS to remove the cryoprotective solution. Sections were incubated overnight with PGT (0.45% Gelatin and 0.25% Triton in PBS) at 4 °C. Next, slices were incubated with primary antibodies: either 1:200 goat anti-choline-acetyl-transferase from Millipore or 1:500 mouse anti-PV monoclonal IgG1 from Sigma-Aldrich in PGT at room temperature for 48 h or 2 h respectively. Following washes of 10, 20, and 30 min, sections were then incubated with secondary antibodies [1:2000 donkey anti-goat coupled with Alexa 488 or 1:500 goat anti-mouse IgG1 coupled to Alexa 488 (Life Technologies)] in PGT for 45 min. Following 10, 20, and 30 min washes in PBS, sections were then mounted on glass slides and permanently coverslipped with Fluoromount mounting medium that contained DAPI. Only mice with the histologically confirmed placement of implants were included in this study. For GRIN lenses, the surface of the lens had to be <100 µm above stratum pyramidale, and GCaMP6f expression was validated using fluorescence microscopy. Electrophysiological implants had to include at least one microelectrode in CA1 stratum radiatum or stratum pyramidale. Finally, the tips of fiber optics had to be within 100 µm of the MS region, and proper construct expression was assessed using fluorescence microscopy.

Data analysis

Except for analyses of SWRs and the explicit impact of locomotion on physiological recordings, electrophysiological and calcium imaging analyses were performed only on periods of locomotion (>2 cm s−1 in the open field; >5 cm s−1 on the linear track).

Automatized tracking of behavior. To extract information about the position, velocity, and head direction of mice, we used DeepLabCut70,71. Briefly, we trained a model to detect mice’s ears, nose, body, and tail base. The head direction was estimated using the angle between either each ear or the nose and body, depending on measurement availability. Location data was interpolated to calcium imaging sampling frequency using linear interpolation. Velocity was extracted by computing Δd/Δt where d is the distance and t time and subsequently smoothed the result by applying a Gaussian filter with sigma = 33 ms to remove detection artifacts. Velocity signals were used to identify periods of locomotor activity and compute place, time, and distance modulation of activity specifically for those periods.

Calcium imaging analysis. Calcium imaging data analysis was performed using MATLAB 2020a and Python 3.8.4. Video recordings were analyzed using the Miniscope Analysis pipeline (https://github. com/etterguillaume/MiniscopeAnalysis). Briefly, rigid (rotation and translation) motion correction was applied using NoRMCorre72, and videos were spatially downsampled (3×) before concatenation. Calcium traces were extracted using CNMFe51 using the following parameters: gSig = 3 pixels (width of Gaussian kernel), gSiz = 20 pixels (approximate neuron diameter), background_model = ‘ring’, spatial_algorithm = ‘hals’, min_corr = 0.8 (minimum pixel correlation threshold), min_PNR = 8 (minimum peak-to-noise ratio threshold).

Raw calcium traces were filtered to remove high-frequency fluctuations and binarized: Briefly, neurons were considered active when normalized calcium signal amplitude exceeded two standard deviations, and the first-order derivative was above 0 (see ref. 52 for additional details on the methodology52). To extract neurons tuning to specific variables, location, time, and distance were binned (location: 3 cm bins; time: 1 s bins; distance: 3 cm bins). From binarized signals, we computed the marginal likelihood of cells being active P Að Þ that we use as a proxy for neuronal activity. More importantly, we then derive activity likelihood or the ‘probability of being active given the state of a variable’ P A j Si
using binned variables:

where M is the total number of possible behavioral states, and P(Si∩Aj) is the joint probability of the animal being in bin I concurrently with activity level j (0 or 1). To assess the significance of the obtained MI value, we then generated 1000 shuffled surrogates using random circular permutations. We chose circular permutations because they remove the temporal relationship between neuronal activity and behavior, while still preserving the temporal structure of calcium transients and thus lead to more conservative results (as opposed to complete randomization of every data point, which inflates the significance value of results). Because shuffled surrogates were not systematically normally distributed, we used a non-parametric approach where the p-value (pN) corresponds to the number of data points from the shuffled distribution that is greater than the actual data for each bin, divided by the number of permutations52,73.

To determine the modulation of cells by optogenetic stimulations, we used a similar approach but computed the MI between neuronal activity and the binarized stimulation signals (thus treated as a behavioral state). The same circular shuffling procedure was then used to extract statistical significance.

where P(S|A) is the posterior probability distribution of states given neuronal activity. Using only epochs with velocity >5 cm s−1, a training dataset was generated using 90% of the data. The remaining 10% of the data was used for testing. A decoding error was calculated using 50 bootstrapped surrogates and a pool of 160 cells using randomly chosen data points with replacement. Every neuron is assumed to be independent of each other, which in practice is not the case and leads to greater reconstruction errors, but decreases computational time. The population posterior probability was derived from the following equation:

Tracking cells over multiple days. Neurons were tracked over multiple days using CellReg74: https://github.com/zivlab/CellReg (v1.5.3). Briefly, spatial footprints were aligned using rigid alignment to correct for rotations and translations. After alignment, we considered candidate sets of cells to be the same neuron if their maximal distance was <12 µm, and used the modeled spatial correlation threshold (usually in the range 0.6–0.8) to determine the identity of cell pairs across days. Finally, we assessed the stability of the spatial representation using pairwise field correlation (Pearson correlation of tuning curves).

Electrophysiological analysis. Electrophysiological data analysis was performed using MATLAB 2020a using both the signal processing as well as the wavelet toolbox. Wavelet convolution was applied to LFP signals using complex Morlet wavelets (‘cmor1–1.5’ in MATLAB) when both time- and frequency-domain accuracy was required. Moving window Fourier convolution (2 s window in the theta band, 5 s window in the gamma band, 10 ms moving steps) was used when frequency domain accuracy was privileged over time-domain accuracy (e.g., to plot dominant frequency when pacing using optogenetics). Analysis of power spectral densities was performed when mice were running at 5 cm s−1 or above unless described otherwise.

Oscillation strength. OS was computed as the ratio of cumulative power spectral density around the peak oscillation frequency ±1 Hz to the cumulative band power in the theta (4–12 Hz) band. This metric becomes 1 when all power spectral density falls within the peak oscillation frequency (e.g. 8 Hz if stimulating at that frequency).

SWRs detection and analysis. To monitor SWRs, mice were allowed to freely explore an open field for 10 min while recording. Stimulations (scrambled or 8 Hz) were performed with a 5 s ON, 5 s OFF paradigm. Only periods of quiet restfulness were considered for analysis. To this end, we computed the z-scored ratio of theta/delta power after filtering for each frequency band, performing a Hilbert transform, and only considered periods where the resulting value was under 0. To detect SWRs, we filtered LFP signals in the 150–250 Hz frequency band and subsequently z-scored. Ripples were detected using the findpeaks function in MATLAB, with the following parameters: threshold = 4 sd, minpeakwidth = 0 s, minpeakdistance = 0.03 s.

References

1. Colgin, L. L. Rhythms of the hippocampal network. Nat. Rev. Neurosci. 17, 239–249 (2016). 

2. Tóth, K., Freund, T. F. & Miles, R. Disinhibition of rat hippocampal pyramidal cells by GABAergic afferents from the septum. J. Physiol. 500, 463–474 (1997). 

3. Unal, G., Joshi, A., Viney, T. J., Kis, V. & Somogyi, P. Synaptic targets of medial septal projections in the hippocampus and extrahippocampal cortices of the mouse. J. Neurosci. 35, 15812–15826 (2015). 

4. Simon, A. P., Poindessous-Jazat, F., Dutar, P., Epelbaum, J. & Bassant, M.-H. Firing properties of anatomically identified neurons in the medial septum of anesthetized and unanesthetized restrained rats. J. Neurosci. 26, 9038–9046 (2006). 

5. Scotty, F. et al. Distinct electrophysiological properties of glutamatergic, cholinergic and GABAergic rat septohippocampal neurons: novel implications for hippocampal rhythmicity. J. Physiol. 551, 927–943 (2003). 

6. Manseau, F., Danik, M. & Williams, S. A functional glutamatergic neuron network in the medial septum and diagonal band area. J. Physiol. 566, 865–884 (2005). 

7. Amilhon, B. et al. Parvalbumin interneurons of the hippocampus tune population activity at theta frequency. Neuron 86, 1277–1289 (2015). 

8. Etter, G. et al. Optogenetic gamma stimulation rescues memory impairments in an Alzheimer’s disease mouse model. Nat. Commun. 10, 1–11 (2019). 

9. Zutshi, I. et al. Hippocampal neural circuits respond to optogenetic pacing of theta frequencies by generating accelerated oscillation frequencies. Curr. Biol. 28, 1179–1188.e3 (2018). 

10. Bender, F. et al. Theta oscillations regulate the speed of locomotion via a hippocampus-to-lateral septum pathway. Nat. Commun. 6, 8521 (2015).

For more information:1950477648nn@gmail.com