Optogenetic Frequency Scrambling Of Hippocampal Theta Oscillations Dissociates Working Memory Retrieval From Hippocampal Spatiotemporal Codes Part 1

The precise temporal coordination of activity in the brain is thought to be fundamental for memory function. Inhibitory neurons in the medial septum provide a prominent source of innervation to the hippocampus and play a major role in controlling hippocampal theta (~8 Hz) oscillations. While pharmacological inhibition of medial septal neurons is known to disrupt memory, the exact role of septal inhibitory neurons in regulating hippocampal representations and memory is not fully understood. 

The relationship between brain activity and memory is closely related. Our memory is closely tied to the workings of our brain, one of the main organs in our body that controls how we think, feel, and act. Therefore, if we want to have better memories, we need to use our brains more efficiently.

First, we need to understand how our brains work. There is a part of the brain called the hippocampus, which is our memory center. It is responsible for converting processed information into long-term memory and storing it in our brains. The factors that affect the function of the hippocampus are mainly the blood circulation of the brain and the quantity and quality of nerve cells.

Secondly, we should pay attention to keeping the brain active. By continually learning new things, the brain engages in information processing and memory recall more frequently, which promotes hippocampus health and promotes improved memory. Therefore, we can keep our brains active and improve our memory by reading books, learning languages, exercising memory, etc.

Finally, we must pay attention to maintaining good living habits. For example, getting enough sleep, eating a balanced diet, and exercising regularly can help promote a healthy brain and body, thereby improving our memory.

In short, brain activity and memory are closely related. We can promote the health and training of the brain in many aspects, thereby improving our memory and coping with various challenges in life and work. It can be seen that we need to improve our memory. Cistanche deserticola can significantly improve our memory because Cistanche deserticola is a traditional Chinese medicinal material that has 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 Short-term Memory how to improve

Here, we dissociate the role of theta rhythms in spatiotemporal coding and memory using an all-optical interrogation and recording approach. We find that optogenetic frequency scrambling stimulations abolish theta oscillations and modulate a portion of neurons in the hippocampus. Such stimulation decreased episodic and working memory retrieval while leaving hippocampal spatiotemporal codes intact. Our study suggests that theta rhythms play an essential role in memory but may not be necessary for hippocampal spatiotemporal codes.

The precise temporal coordination of neuronal activity is thought to be fundamental for memory encoding and retrieval. In particular, the medial septum (MS) has been proposed to act as the main zeitgeber to downstream structures and provides the largest subcortical inputs to the hippocampus1. The MS is a neurochemically heterogeneous structure composed of GABAergic2, mainly parvalbumin (PV)-positive neurons3, along with cholinergic4 and a smaller population of glutamatergic neurons5,6. 

MS PV cells project directly onto GABAergic interneurons in the hippocampus giving rise to feedforward inhibitory control of hippocampal pyramidal cells3. Additionally, PV interneurons within the hippocampus are essential in pacing theta (~8 Hz) rhythms7, and optogenetic stimulation of MS PV neurons directly8,9 or of their terminals10 in the hippocampus are associated with frequency-specific pacing of hippocampal oscillations, whereas inhibition of the MS in vivo has been associated with reduced theta oscillation power11–13.

While complete MS optogenetic inhibition has been associated with spatial memory impairments14, these effects could potentially be attributed to disruption of cholinergic functions15,16, which are known to be critical for memory. More recently, the activity of noncholinergic MS PV cells is necessary for memory encoding and retrieval17,18. Notably, the hippocampus is also a core structure for episodic19,20 and working21 memory. Since the MS is essential in generating and maintaining hippocampal theta rhythms, disruption of MS activity is likely to impact downstream hippocampal physiology and working memory22.
Although the exact physiological mechanisms of hippocampus-dependent memory are currently unknown, it has been proposed that hippocampal place cells23 that encode specific locations of a given context could support episodic memory24. In the hippocampal subfield CA1, spatial tuning depends on contextual sensory inputs25. Other variables such as time and distance can also be encoded during visually guided locomotion26,27 but also in the absence of sensory cues28, likely using internal information including self-motion29 (see Mehta30 and McNaughton31 for review). 

Representations of time and space can be represented conjunctively in hippocampal neurons, and such multiplexed spatiotemporal codes could be a candidate substrate for working memory 32–34. In addition to clamping distal visual cues using a treadmill35–37 or using virtual reality paradigms26,27,38, spatiotemporal codes have also been extracted analytically using generalized linear models that implement space, time, and distance37. However, such approaches have not been employed extensively on recordings of neuronal activity during free exploration.

Several studies suggest that hippocampal theta rhythms could underlie temporal codes since theta rhythms tightly orchestrate hippocampal activity. While time cells have also been reported in both CA1 and CA3 of rodents performing tasks that do not require working memory41, pharmacological inhibition of the MS results in specific disruption of time but not place cells and is associated with decreased working memory36. An important drawback of pharmacological approaches is that they do not distinguish the relative contribution of GABAergic versus cholinergic cells to memory function. Notably, inhibition of MS cholinergic activity was found to alter hippocampal spatial representations43 and decrease working memory performance44,45. Surprisingly, pharmacological inhibition of the MS was associated with reduced theta oscillation power but not place fields46, and this resilience of place cell activity during diminished theta was not due to experience-related plasticity mechanisms47. 

Previous attempts at inhibiting MS GABAergic interneurons specifically using optogenetics, were associated with only a partial reduction, but not complete disruption of theta signals13. In turn, optogenetic pacing of theta oscillation has only been associated with minor changes in place cell characteristics, including a slight shift in firing frequency9 and phase48. Additionally, while it is hypothesized that MS inputs could directly control hippocampal temporal codes, causal evidence is still lacking. To this day, the exact role of MS-PV neurons in orchestrating hippocampal spatiotemporal codes and memory remains unknown.

Here, we controlled MS PV activity using optogenetics to pace or abolish theta oscillations using a red-shifted excitatory opsin. We propose an approach to completely abolish hippocampal theta rhythms based on optogenetic frequency scrambling stimulations of MS neurons. Alternatively, pacing theta rhythms at their natural frequency in the same animals provides within-subject controls. We combined optogenetic control with calcium imaging of CA1 pyramidal cells in mice running on a linear track with sequential tones. 

In these conditions, we could separate place, time, and distance cells using an information theoretic approach. When performing optogenetic frequency scrambling of theta signals, both place and time representations were preserved and only a small subset of CA1 pyramidal cells was modulated by stimulation. We next found that ablation of theta oscillation was associated with impaired working memory retrieval suggesting that MS PV cells play a critical role in generating hippocampal theta oscillations that are necessary for memory retrieval but are not involved in spatiotemporal representations.

Results

CA1 pyramidal cells encode spatiotemporal information

To examine the spatiotemporal codes in large populations of CA1 principal cells, we injected a viral vector expressing GCaMP6fast under a CamKII promoter in the CA1 region of the hippocampus, implanted a GRIN lens above the injection site, and performed calcium imaging recordings of pyramidal neurons using open-source miniscopes49,50 (Fig. 1a, b; see Supplementary Fig. 1 for detailed histology). We extracted spatial footprints of neurons (Fig. 1c) and their corresponding calcium transients (Fig. 1d) using CNMFe51. To tease apart the spatial and temporal tuning properties of principal cells, we developed a task combining a linear track with three-tone cues triggered by motion sensors at both ends of the track. A new tone was instantly triggered at the end of each run, informing the mice of their progression toward reward delivery. 

Every fourth run was cued with a high-pitched, continuous tone that was associated with the delivery of a reward at the end of the linear track (Fig. 1e). The absolute location of each mouse, along with the time elapsed and distance traveled since the departure from the reward site were monitored (Fig. 1f). Using these variables and binarized neuronal activity, we computed probabilistic tuning curves (Fig. 1g) and derived mutual information (MI) between neuronal activity and location, time, as well as the distance for each recorded cell. In contrast to correlation-based analyses, MI does not assume linear, monotonic relationships between neuronal activity and behavioral variables but rather expresses the amount of uncertainty of one variable that can be explained by the other. 

The significance of MI values was tested using shuffled surrogates that underwent circular permutations (n = 1000) to preserve the temporal dynamics of calcium transients. Neurons that encoded exclusively one variable with an MI greater than shuffled surrogates 95% of the time (p ≤ 0.05) were labeled as either place-modulated (spatial), time-modulated (temporal), or distance-modulated (see Methods). For the following analyses, we focused on candidate cells that only encoded one significant variable (Fig. 1h). Importantly, time-modulated cells were not systematically active at particular locations, and place-modulated cells were not systematically active at a given time (Fig. 1i, j). 

While the majority of cells encoding a single variable were place-modulated, a large portion of neurons were conjunctive neurons that encoded more than one variable (17.98 ± 1.77%). In contrast, cells encoding place exclusively represented 9.02 ± 1.61% of the total recorded population, while 1.79 ± 0.68% selectively encoded distance and 1.27 ± 0.09% selectively encoded time (Fig. 1k; additional examples of neurons tuned to time, space, or distance along with their information content are shown in Supplementary Fig. 2).

Although our information-theoretic approach can disentangle overlapping variables by isolating cells that only significantly encode one variable, we further tested the relevance of each cell type in encoding spatiotemporal variables using a naive Bayesian classifier to decode location (Fig. 1l–n), the time elapsed (Fig. 1o–q), and distance traveled (Fig. 1r–t) on the linear track52. 

We estimated the current state of each mouse by computing the maximum a posteriori (MAP) value given neuronal activity and bootstrapped tuning curves computed using actual or circularly shuffled binarized activity (Fig. 1l, o, r; see Methods for detailed protocol). The quality of predictions was assessed using confusion matrices (Fig. 1m, p, s) and by computing the Euclidean distance between the predicted state and the actual state (Fig. 1n, q, t). Importantly, our Bayesian decoder yielded an average error of 16.58 cm, which was significantly lower than when decoding from shuffled surrogates (50.95 cm; paired t-test, t4 = 19.75, p ≤ 0.0001), and decoding using spatially modulated cells was significantly more accurate than when using non-spatially modulated cells (paired t-test, t4 = 34.54, p ≤ 0.0001; Fig. 1n). Similarly, the average decoding error for time elapsed was 6.45 s, which was significantly lower than error computed using shuffled surrogates (19.12 s; paired t-test,t4 = 18.01, p ≤ 0.0001). 

Decoding using time-modulated cells yielded significantly better accuracy compared to non-time-modulated cells (paired t-test, t4 = 3.163, p = 0.0341; Fig. 1q). Finally, the average distance error using our decoder was 61.40 cm, which was significantly lower than that of shuffled surrogates (189.6 cm; t-test, t4 = 28.79, p ≤ 0.0001). Decoding using distance-modulated cells yielded significantly lower errors compared to non-distance-modulated cells (t-test, t4 = 4.595, p = 0.0101; Fig. 1t).

Selective MS optogenetic control of theta oscillations

To examine the relative contribution of MS-generated theta signals to hippocampal spatiotemporal codes, we transfected the red-shifted excitatory opsin ChrimsonR in the MS (Fig. 2a). In contrast to inhibitory opsins, ChrimsonR allowed us to either scramble or pace theta signals within subjects. Additionally, ChrimsonR is more effective than the more widely used Channelrhodopsin-2 and provides the capability to combine optogenetics with calcium imaging53. 14.91 ± 3.57% of PV cells expressed ChrimsonR, which we found sufficient to exert pervasive control over hippocampal oscillations (n = 4 mice; Fig. 2b, c). In contrast, we found virtually no expression of ChrimsonR in ChAT cells (1.05 ± 1.052% of ChAT cells also expressed ChrimsonR; Fig. 2d, e).

We then implanted mice with fiber optics over the MS. We performed 638 nm laser stimulation while recording local field potentials (LFP) in CA1 (Fig. 2f). We found that MS scrambled and 8 Hz optogenetic stimulation could disrupt or pace theta oscillations, respectively (Fig. 2g). While baseline natural theta displays some frequency variability in the 4–12 Hz frequency band, 8 Hz stimulations led to consistent and stable hippocampal oscillations at that frequency. In contrast, scrambled stimulations consistently abolished theta rhythms (Fig. 2h). 

We found that oscillation strength (OS) in the theta band (see Methods) was significantly decreased by scrambled stimulations (0.45 ± 0.01) compared to baseline epochs (0.67 ± 0.01, p ≤ 0.0001) and were not significantly different from OS of the white noise control signal (0.50 ± 0.01, p = 0.99). On the other hand, 8 Hz stimulations increased theta power significantly (0.82 ± 0.01, p ≤ 0.0001; n = 59 epochs; Fig. 2i) compared to scrambled stimulations. We also found a significant interaction between our stimulation patterns and the LFP frequency band (F10 = 6.467, p ≤ 0.0001). In particular, scrambled frequency stimulation significantly decreased theta power (0.341 ± 0.06 portion of baseline theta band power; p = 0.0394, pairwise t-test), while 8 Hz stimulations significantly increased theta power (3.302 ± 0.76 portion of baseline theta band power, p = 0.0004, pairwise t-test; Fig. 2j) leaving other frequency bands unaltered.

While our calcium imaging and electrophysiological analyses only included periods of locomotion (see Methods), we also found that we were able to reliably abolish (Supplementary Fig. 3a, b) or pace (Supplementary Fig. 3a, c) theta oscillations regardless of locomotor state (including periods of restfulness). While natural theta OS is correlated to locomotor speed (Pearson R2 = 0.059, p = 0.001; n = 179 independent epochs; Supplementary Fig. 3d), abolishing theta led to a loss of such correlation (Pearson R2 = 0.008, p = 0.223; n = 179 independent epochs; Supplementary Fig. 3e), as did 8 Hz stimulations (Pearson R2 = 0.0008, p = 0.714; n = 177 independent epochs; Supplementary Fig. 3f) suggesting that locomotor states did not override the effects of optogenetic stimulations on theta oscillations.

Hippocampal sharp-wave ripples (SWRs) play an essential role in memory consolidation54–56 and stimulation of MS cholinergic neurons has been associated with reduced ripple activity57 and impaired working memory45. Although we found virtually no expression of ChrimsonR in MS cholinergic neurons, it was essential to measure the impact of our MS optogenetic stimulation on ripple physiology. To this end, we recorded CA1-LFP and performed 5s ON, and 5s OFF scrambled optogenetic stimulation in freely behaving mice exploring an open field (Supplementary Fig. 4a). We measured attributes of ripple events before and during scrambled stimulation, and found no changes in power (unpaired, two-tailed t-test, t6 = 0.076, p = 0.941; Supplementary Fig. 4b, left panel), frequency of occurrence (unpaired, two-tailed t-test,t6 = −1.688, p = 0.142; Supplementary Fig. 4c, left panel), or width (unpaired, two-tailed t-test, t6 = 0.124, p = 0.905; Supplementary Fig. 4d, left panel). 

Similarly, applying 8 Hz optogenetic stimulation had no discernable effects on ripple power (unpaired, two-tailed t-test, t6 = 0.378, p = 0.718; Supplementary Fig. 4b, right panel), frequency (unpaired, two-tailed t-test, t6 = −1.643, p = 0.151; Supplementary Fig. 4c, right panel), and width (unpaired, two-tailed t-test, t6 = −0.138, p = 0.894; Supplementary Fig. 4d, right panel). Together with our histological results and a previous report that optogenetic stimulation of MS cholinergic neurons reduces the occurrence of SWRs57, we find that these optogenetic stimulations do not affect cholinergic inputs to the hippocampus.

For more information:1950477648nn@gmail.com