Experience-driven rate modulation is reinstated during hippocampal replay

  1. Margot Tirole
  2. Marta Huelin Gorriz
  3. Masahiro Takigawa
  4. Lilia Kukovska
  5. Daniel Bendor

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    We all have had days where there were multiple distinct memorable experiences that we successfully remember as distinct. This paper for the first time focuses on the important question of whether the resting/sleeping hippocampus maintains a clear distinction between replays of different environments and finds that in fact, replays of different tracks are distinct in the sense that both the right sets of neurons are coactive AND their firing rates in replay reflect their firing rates during experiences.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Replay, the sequential reactivation within a neuronal ensemble, is a central hippocampal mechanism postulated to drive memory processing. While both rate and place representations are used by hippocampal place cells to encode behavioral episodes, replay has been largely defined by only the latter – based on the fidelity of sequential activity across neighboring place fields. Here, we show that dorsal CA1 place cells in rats can modulate their firing rate between replay events of two different contexts. This experience-dependent phenomenon mirrors the same pattern of rate modulation observed during behavior and can be used independently from place information within replay sequences to discriminate between contexts. Our results reveal the existence of two complementary neural representations available for memory processes.

Article activity feed

  1. Version published to 10.7554/elife.79031 on eLife
  2. eLife

    Evaluation Summary:

    We all have had days where there were multiple distinct memorable experiences that we successfully remember as distinct. This paper for the first time focuses on the important question of whether the resting/sleeping hippocampus maintains a clear distinction between replays of different environments and finds that in fact, replays of different tracks are distinct in the sense that both the right sets of neurons are coactive AND their firing rates in replay reflect their firing rates during experiences.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

  3. eLife

    Reviewer #1 (Public Review):

    This paper used two linear tracks of different cues as two contexts and tested the rate modulation of contexts during behavior and during replay events. They showed that not only sequential information, but rate information also are encoding information and that they are reinstated during replay events. This is super exciting! The data about how things change during sleep is also timely and important.

    My primary criticism of this paper is that it misses the opportunity to give some key details about the statistics of neural activity during 'ripples' rather than studying identified replay events. A secondary criticism is that they limit their analyses to neurons that have place fields in both environments. I think the activity of the other 3 categories of neurons (active in Track 1 only, active in Track 2 only, and not active in either track) are also of critical interest.

  4. eLife

    Reviewer #2 (Public Review):

    This study by Tirole et al. addresses to what extent differences in firing rate that occurs during the awake experience of two different tracks are replayed during SWRs.

    In principle, this is a topic broadly relevant to our understanding of the circuit-level mechanisms and neural coding of memory, because it can provide insight into the ways in which experience is transformed into memory traces, and in particular, whether an entire coding modality (firing rate patterns) is available for replay. However, I didn't have an easy time situating this study in the context of the existing literature. When I first read the title, I expected this work was going to address the question of if there is replay of rate-remapped experiences, which is still an understudied topic (but see Takahashi, 2015) and would be important to examine. But once I realized that the two experiences here are actually more like global remapping, it was less clear to me what is novel here.

    My best guess about what's novel is that even though on the one hand, many studies have shown a distinguishable replay of two (or more) distinct experiences, e.g. different mazes like in Karlsson et al. 2009, different arms of a T-maze in Gupta et al. 2010, the overlapping central stem element of different trajectories in various mazes (Takahashi, 2015 and work from the Jadhav lab). On the other hand, there have been extremely detailed examinations of the contributions of firing rate changes (as distinct from temporal order or synchrony) as in Farooq et al. 2019. But perhaps the authors think that the intersection of those two kinds of work has not been studied, that is, how much do firing rate changes specifically contribute to the replay of two distinct experiences? In any case, regardless of whether I understood that correctly or not, the authors need to be more explicit in the introduction and discussion in contextualizing their work. I also suspect that the current findings are a direct logical consequence of putting together these well-established previous results; this would not mean the current work isn't a useful advance, but it would moderate the novelty and general interest.

    Beyond this overall question of how the work relates to the extant literature, I have a suggested modification to the data analysis. I think that the quality of the data and the care taken in the analyses were very high in general, so I do not have any major concerns, and the conclusions are very thoroughly supported. However, I wonder if there is a way to simplify some of the analyses and make them a bit more straightforward to interpret. As the authors have realized, there is potential for a circularity in the analysis, in the sense that to compare firing rate differences for two tracks between Track and Replay, Replay events first need to be assigned to one or the other (decoded) Track. But then any firing rate differences may be contributing to the output of the decoder, rendering the analysis circular. I understand the authors use various methods like the firing-rate-insensitive method in Figure 2 to deal with this crucial issue. But wouldn't a simpler way be to leave out the cell whose firing rates are being analyzed out of the decoding step so that the labeling of Replay events is independent of that cell? This seems an intuitive and rigorous way to address the central question the authors have. Is there some reason why that isn't done?

  5. eLife

    Reviewer #3 (Public Review):

    In general, I find this to be an experimentally and analytically sound paper. The observation that rate information is preserved in hippocampal replay is hinted at in previous work, but to my knowledge, has not yet been explicitly quantified as the authors have done here. Thus, this work is novel and, in my opinion, an important contribution to our understanding of hippocampal network function.

    The large number of control analyses strongly support the core finding of this work. I feel that the authors have very convincingly demonstrated that rate information is represented along with spatial information in replay.

    While I can think of many suggestions to follow up on this work, I have no major concerns regarding the experiments, analyses, or interpretation of the manuscript.

  6. eLife

    Author Response

    Reviewer #1 (Public Review):

    My primary criticism of this paper is that it misses the opportunity to give some key details about the statistics of neural activity during 'ripples' rather than studying identified replay events. A secondary criticism is that they limit their analyses to neurons that have place fields in both environments. I think the activity of the other 3 categories of neurons (active in Track 1 only, active in Track 2 only, and not active in either track) are also of critical interest.

    We agree with the reviewer that it is important to demonstrate that the main observations are not due to a small subset of neurons or replay events. We have described above the inclusion of Figure 1- figure supplement 6, where the threshold for replay detection is made less stringent and the ratio of significant replay events/candidate replay events are now reported in the manuscript. To address the concern that the analysis is limited to neurons only with place fields on both tracks, we have added four more subpanels to Figure 1-figure supplement 6, where we perform our regression analysis on all spatially tuned (pyramidal) neurons (Figure 1-figure supplement 6E), neurons with only place fields on one track (track 1 and track 2 neurons will be in the upper right and lower left quadrant of plot respectively, Figure 1-figure supplement 6F), neurons with peak amplitude <1Hz on each tracks (Figure 1-figure supplement 6G) and finally, interneurons (Figure 1-figure supplement 6H). Consistent with our previous findings, we observe significant regressions for POST replay events for all spatially tuned neurons and neurons with place fields only one track. Conversely, neurons that were not active on either track and interneurons are not rate modulated by experience during replay.

    It is important to note that replay detection uses all spatially tuned cells, but the regression analysis is limited to cells active on both tracks in the main analysis. The reason for this is now explained in more detail in the revised manuscript (page 5):

    “It is important to note that a significant regression would be expected when analyzing neurons with a place field only on one track, as they are expected to participate in replay events of this track, while being silent during the replay of the other track. As such, our regression analysis only analyzed place cells active on both tracks and stable across the whole run (Figure 1-figure supplement 1B and see Methods).”

    Reviewer #2 (Public Review):

    This study by Tirole et al. addresses to what extent differences in firing rate that occurs during the awake experience of two different tracks are replayed during SWRs.

    In principle, this is a topic broadly relevant to our understanding of the circuit-level mechanisms and neural coding of memory, because it can provide insight into the ways in which experience is transformed into memory traces, and in particular, whether an entire coding modality (firing rate patterns) is available for replay. However, I didn't have an easy time situating this study in the context of the existing literature. When I first read the title, I expected this work was going to address the question of if there is replay of rate-remapped experiences, which is still an understudied topic (but see Takahashi, 2015) and would be important to examine. But once I realized that the two experiences here are actually more like global remapping, it was less clear to me what is novel here.

    My best guess about what's novel is that even though on the one hand, many studies have shown a distinguishable replay of two (or more) distinct experiences, e.g. different mazes like in Karlsson et al. 2009, different arms of a T-maze in Gupta et al. 2010, the overlapping central stem element of different trajectories in various mazes (Takahashi, 2015 and work from the Jadhav lab). On the other hand, there have been extremely detailed examinations of the contributions of firing rate changes (as distinct from temporal order or synchrony) as in Farooq et al. 2019. But perhaps the authors think that the intersection of those two kinds of work has not been studied, that is, how much do firing rate changes specifically contribute to the replay of two distinct experiences? In any case, regardless of whether I understood that correctly or not, the authors need to be more explicit in the introduction and discussion in contextualizing their work. I also suspect that the current findings are a direct logical consequence of putting together these well-established previous results; this would not mean the current work isn't a useful advance, but it would moderate the novelty and general interest.

    Beyond this overall question of how the work relates to the extant literature, I have a suggested modification to the data analysis. I think that the quality of the data and the care taken in the analyses were very high in general, so I do not have any major concerns, and the conclusions are very thoroughly supported. However, I wonder if there is a way to simplify some of the analyses and make them a bit more straightforward to interpret. As the authors have realized, there is potential for a circularity in the analysis, in the sense that to compare firing rate differences for two tracks between Track and Replay, Replay events first need to be assigned to one or the other (decoded) Track. But then any firing rate differences may be contributing to the output of the decoder, rendering the analysis circular. I understand the authors use various methods like the firing-rate-insensitive method in Figure 2 to deal with this crucial issue. But wouldn't a simpler way be to leave out the cell whose firing rates are being analyzed out of the decoding step so that the labeling of Replay events is independent of that cell? This seems an intuitive and rigorous way to address the central question the authors have. Is there some reason why that isn't done?

    We thank the reviewer for this feedback, and agree it is important to emphasize the novel contributions of the manuscript (as we see it), and clarify this further if needed. The reviewer is correct that there are several studies that have looked at rate remapping during reactivation. We have cited some of these, but have now updated our citations in the intro and discussion based on the comments here. While we have avoided directly criticizing a particular study in our earlier draft of the manuscript, these previous studies are affected generally by several issues:

    1. replay detection methods were sensitive to rate modulation, creating a circular argument for the existence of rate modulation in replay. [Our study thoroughly addresses this with several controls].
    2. the analysis of reactivations rather than replay, which lacks the statistical rigor of sequence detection [we have focused on replay using a strict threshold for significance]
    3. Replay/reactivations are analyzed for a single environment, making it difficult to distinguish between rate modulation and changes in the overall excitability levels of neurons maintained over behavior and sleep. [our studies uses two tracks to avoid this potential issue].
    4. When multiple contexts were decoded, neurons that only fired in one context were not removed from the analysis, artificially “inflating” any observed rate modulation. [we have circumvented this issue by only analyzing neurons with place fields in both environments]

    The suggestion to repeat the analysis and leave one neuron out for replay detection is excellent, however this was avoided due to the required processing time- to run our complete analysis takes more than a week, and repeating this for each possible “leave-one-out” combination would take significantly longer (this has to be done independently for each neuron). We used multiple controls (track rate shuffle, replay rate shuffle, rank order correlation- figure 2, figure 2—figure supplement 2) to eliminate any possibility that a neuron’s firing rate could influence replay detection. Specifically, for rank-order correlation based replay detection, each burst of spikes is only treated as a single event (median of spike times in the burst), which directly circumvents the problem of firing rate biasing replay event selection.

  11. Version published to 10.1101/2021.07.15.452506v3 on bioRxiv
  12. Version published to 10.1101/2021.07.15.452506v2 on bioRxiv
  13. Version published to 10.1101/2021.07.15.452506v1 on bioRxiv