Increased cortical plasticity leads to memory interference and enhanced hippocampal-cortical interactions

  1. Irene Navarro Lobato
  2. Adrian Aleman-Zapata
  3. Anumita Samanta
  4. Milan Bogers
  5. Shekhar Narayanan
  6. Abdelrahman Rayan
  7. Alejandra Alonso
  8. Jacqueline van der Meij
  9. Mehdi Khamassi
  10. Zafar U Khan
  11. Lisa Genzel

  • Curated by eLife

    eLife logo

    eLife assessment

    This important study reveals that slow plasticity in the neocortex is essential to prevent memory interference. The method of artificially increasing plasticity in the prefrontal cortex of rats during learning and its effect on sleep physiology, when memories are believed to be reprocessed, is solid. The work will be of interest to neuroscientists interested in learning and memory.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Our brain is continuously challenged by daily experiences. Thus, how to avoid systematic erasing of previously encoded memories? While it has been proposed that a dual-learning system with ‘slow’ learning in the cortex and ‘fast’ learning in the hippocampus could protect previous knowledge from interference, this has never been observed in the living organism. Here, we report that increasing plasticity via the viral-induced overexpression of RGS14414 in the prelimbic cortex leads to better one-trial memory, but that this comes at the price of increased interference in semantic-like memory. Indeed, electrophysiological recordings showed that this manipulation also resulted in shorter NonREM-sleep bouts, smaller delta-waves and decreased neuronal firing rates. In contrast, hippocampal-cortical interactions in form of theta coherence during wake and REM-sleep as well as oscillatory coupling during NonREM-sleep were enhanced. Thus, we provide the first experimental evidence for the long-standing and unproven fundamental idea that high thresholds for plasticity in the cortex protect preexisting memories and modulating these thresholds affects both memory encoding and consolidation mechanisms.

Article activity feed

  1. Version published to 10.7554/elife.84911 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Public Review):

    The authors use a newly developed object-space memory task comprising of a "Stable" version and "Overlapping" version where two objects are presented in two locations per trial in a square open field. Each version consists of 5 training trials of 5-min presentations of an object-space configuration, with both object locations staying constant across training trials in the Stable condition, and only one object location staying fixed in the Overlapping condition. Memory is tested in a test trial 24 hours later where the opposite configuration is presented - overlapping configuration presented for the Stable condition and stable configuration presented for the Overlapping condition - with the thesis that memory in this test trial for the Overlapping condition will depend on the accumulated memory of spatial patterns over the training trials, whereas memory for the test trial in the Stable condition can be due to episodic memory of last trial or accumulated memory. Memory is quantified using a Discrimination Index (DI), comparing the amount of time animals spend exploring the two object locations.

    Here, animals in other groups are also presented with an interference trial equivalent to the test trial, to test if the memory of the Overlapping condition can be disrupted. The behavioral data show that for RGS14 over-expressing animals, memory in the Overlapping condition is diminished compared to controls with no interference or controls where over-expression is inhibited, whereas memory in the Stable condition is enhanced. This is interpreted as interference in semantic-like memory formation, whereas one-shot episodic memory is improved. The authors speculate that increased cortical plasticity should lead to increased and larger delta waves according to the sleep homeostasis hypothesis, and observe that instead increased cortical plasticity leads to less non-REM sleep and smaller delta waves, with more prefrontal neurons with slower firing rates (presumably more plastic neurons). They further report increased hippocampal-cortical theta coherence during task and REM sleep, increased NonREM oscillatory coupling, and changes in hippocampal ripples in RGS14 over-expressing animals.

    While these results are interesting, there are several issues that need to be addressed, and the link between physiology and behavioral results is unclear.

    1. The behavioral results rely on the interpretation that the Overlapping condition corresponds to semantic-like memory and the Stable condition corresponds to episodic-like memory. While the dissociation in memory performance due to interference seen in these two conditions is intriguing, the Stable condition can correspond not just to the memory of the previous trial but also accumulated memory of a stable spatial pattern over the 5 testing trials, similar to accumulated memory of a changing spatial pattern in the Overlapping pattern.

    Yes! We completely agree on this. We do not claim the stable condition corresponds to episodic-like memory, instead we refer to it as simple memory, since it can be solved either way (one trial memory or cumulative memory). We now expanded this in the discussion to make it clearer.

    Here, it is puzzling that in the behavioral control with no interference (Figure 1D), memory in the Stable and Overlapping condition is unchanged in the test trial, with the DI statistically at 0 in the test trial. In the original description of the Object Space task by the authors in the referenced paper, the measure of memory was a Discrimination Index significantly higher than 0 in both the Stable and Overlapping conditions. This discrepancy needs to be reconciled. Is the DI for the interference trial shown in Fig. S1 significantly different than 0? No statistics or description is provided in the figure legend here.

    As mentioned above, we apologize that we oversimplified the description. The 24h interference trial would be what corresponds to the original test trial. We added a clarifying figure for comparison in S1 (bar graph in addition to the violin plot) and stats. Performance was for all groups and conditions above chance, replicating our previous results.

    1. The physiology experiments compare Home cage (HC) conditions to the Object Space task (OS) throughout the manuscript. While some differences are seen in the control and RGS14 over-expressing animals, there is no comparison of the Stable vs. Overlapping condition in the physiology experiments. This precludes making explicit links between physiological observations and behavioral effects.

    As also mentioned above, we have now added analysis exploring the detailed OS conditions. We would like to thank the reviewers for giving us the opportunity of doing so.

    1. The authors speculate that learning will result in larger and more delta waves as per the synaptic homeostasis hypothesis. It should be noted here that an alternative hypothesis is that there should also be a selective increase in synaptic plasticity for learning and consolidation. The authors do observe that control animals show more frequent and higher-amplitude delta waves, but rather than enhancing this process, RGS14 animals with increased plasticity show the opposite effect. How can this be reconciled and linked with the behavioral data in the Stable and Overlapping condition?

    In the context of the Object Space Task, we would expect all behavioural conditions (Stable and Overlapping) to induce synaptic changes since learning does occur also in the Stable condition (see also performance on 24h trial). Thus, especially homeostatic responses such as increase in delta amplitude, we would expect for all experiences independent if subtle statistical rules are presented or not. In contrast, detailed processing, extracting underlying regularities is rather proposed by the Sleep for Active Systems Consolidation Hypothesis to occur during hippocampal-cortical interactions in form of delta/ripple/spindle interactions (with different theories emphasising different types of interactions). As mentioned above, we now add a more specific analysis in this regards, where we can show that the two OS conditions that involve moving objects (where thus potentially statistical regularities can be extracted) show a higher percentage of ripples occurring after large slow oscillations in comparison to home cage or the simple learning condition Stable. In contrast, RGS14 already has higher participation in both control conditions, emphasising that in these animals all experiences are treated by the brain as significant learning condition, explaining the behavioural effect (increased interference due to better memory for the interference). Further, we expanded in the discussion how in RGS we sometimes see an enhancement of learning effects but sometimes see a more complex interaction of what we would expect from physiological learning.

    Similarly, there is an increase in slower-firing neurons in RGS14 over-expressing animals. Slower-firing neurons have been proposed to be more plastic in the hippocampus based on their participation in learned hippocampal sequences, but appropriate references or data are needed to support the assertion that slower-firing neurons in the prefrontal cortex are more plastic.

    As described above, we have expanded the discussion including other citations that also consider the cortex. We can show that our changes would be expected if one turns the cortex as plastic as the hippocampus.

    1. It is noted that changing cortical plasticity influences hippocampal-cortical coupling and hippocampal ripples, suggesting a cortical influence on hippocampal physiological patterns. It has been previously shown that disrupting prefrontal cortical activity does alter hippocampal ripples and hippocampal theta sequences (Schmidt et al., 2019; Schmidt and Redish, 2021). The current results should be discussed in this context.

    We would like to thank the reviewer for these suggestions, they are now incorporated in the manuscript.

    Reviewer #2 (Public Review):

    In this paper, the authors provide evidence to support the longstanding proposition that a dual-learning system/systems-level consolidation (hippocampus attains memories at a fast pace which are eventually transmitted to the slow-learning neocortex) allows rapid acquisition of new memories while protecting pre-existing memories. The authors leverage many techniques (behavior, pharmacology, electrophysiology, modelling) and report a host of behavioral and electrophysiological changes on induction of increased medial prefrontal cortex (mPFC) plasticity which are interesting and will be of significant interest to the broad readership.

    The experimental design and analyses are convincing (barring some instances which are discussed below). The following recommendations will bolster the strength/quality of the manuscript:

    1. Certain concerns regarding the interpretation and analysis of the behavioral data remain. The authors need to clarify if increased mPFC plasticity leads to only an increase in one-shot memory or 'also' interference of previous information. It seems that the behavioral results could also be explained by the more parsimonious explanation that one-shot memory is improved. Do the current controls tease apart these two scenarios?

    We agree we cannot disentangle if one memory is just stronger than the other or if its an overwriting effect. We added this now to the discussion. Of note, we do not think it actually would be possible to distinguish these two effects behaviourally in rodents, or at least we cannot think of a fitting study design that would enable the contrast.

    Additionally, the authors need to clarify why the 'no trial' and 'anisomycin' controls for the stable task perform at chance levels on exposure to a new object-place association on test day (Fig 1D).

    Violin plots are sometimes hard to see. Here simple bar plots where you can see that the animals are not at chance at the 72h test in the control conditions.

    Finally, further description of how the discrimination index (exploration time of novel-exploration time of familiar/sum of both) is recommended i.e., in the stable condition, which 'object' is chosen as 'novel' (as both are in the same locations) for computing the index (Fig 1). Do negative DI values imply a neophobia to novel objects (and thus are a form of memory; this is also crucial because the modelling results (Fig 1E) use both neophilia and neophobia while negative discrimination indexes are considered similar to 0 for interpreting the behavioral results, as stated on page 3, lines 84-86?

    We added this now to the methods (For Overlapping it is moved location – stable location, for Stable it is location-to-be-moved-at-test – stable location and for random which is assigned as moved and stable is random, and then for each divided by total time). We agree that neophilia/neophobia (especially changes in the distribution) can be an issue and have discussed it in detail in Schut et al NLM 2020 where we see difference in absolute beta values (thus controlling for philia/phobia differences). We also discuss there why it is difficult to control for this in the DI in more detail. In short, one could use absolute values but then it is difficult to determine what a group chance-level would look like. However, luckily here there is not issue since we did not observe difference in neophilic or phobic tendencies while running the experiments. Critically the interference trial (that can also function as simple test trial) confirms that as a group animals show positive DI and neophilia.

    1. The authors report lower firing rates in RGS14414 animals during the task in Fig 2F. It is indeed remarkable how large the reported differences are. The authors need to rule out any differences in the behavioral state of the animals in the two groups during the task, i.e., rest vs. active exploration/movement dynamics. Are only epochs during the task while the animals interact with the objects used for computing the firing rates (same epochs as Fig 1)? If not, doing so will provide a useful comparison with Fig 1. Additionally, although the authors make the case for slow firing rate neurons being important for plasticity (based on Grosmark and Buzsaki, 2016), it is crucial to note that the firing rate dynamic (slow vs. fast) in that study for the hippocampus is defined based on the whole recorded session (predominated by sleep), indeed the firing rates of the two groups (slow vs. fast/plastic vs. rigid) during the task/maze-running do not differ in that study. Therefore, the results here seem incongruent with the Grosmark and Buzsaki paper. Since this finding is central to the main claim of the authors, it either warrants further investigation or a re-interpretation of their results.

    As mentioned in the main points, we now added the firing rate analysis (including new groups splits) for wake in the sleep box, NREM and REM separately. Each time the same results are obtained. Currently, we do not yet have the tracking and video synchronization set-up, therefore we cannot split the task for specific behaviours.

    However, we now also cite Buzsaki’s original log-normal brain review, where he first proposed the idea. There he also shows same effects as we do, in that the general firing rate distribution is the same for task and different sleep stages, just overall shifted. The analysis from Grosmark included more strigent subselection of neurons to be able to also argue that incorporation into run/replay-sequences could not have been biased by firing rate per se (instead of plasticity). However, the original proposition from Buzsaki does fit to our results. He further presents hippocampus vs cortex firing rates, which also confirm the idea (hippocampus more plastic and has slower firing rates). We included this figure above in the general comments. Further, we now expanded the discussion in this point.

    1. A concern remains as to how many of the electrophysiological changes they observe (firing rate differences, LFP differences including coupling, sleep state differences, Figs. 2-4) support their main hypothesis or are a by-product of injection of RGS14414 (for instance, one might argue that an increased 'capability' to learn new information/more plasticity might lead to more NREM sleep for consolidation, etc.). The authors need to carefully interpret all their data in light of their main hypothesis, which will substantially improve the quality/strength of the manuscript.

    We now expanded the discussion, included more structure and also include that we cannot disentangle if the cellular changes or sleep oscillation changes or an interaction of both is the cause of the result. Furthermore, we added that we cannot distinguish if the interference memory is stronger or actually overwrites the original training memory.

    Reviewer #3 (Public Review):

    The authors set out to test the idea that memories involve a fast process (for the acquisition of new information) and a slow process (where these memories are progressively transferred/integrated into more-long term storage). The former process involves the hippocampus and the latter the cerebral cortex. This 'dual-learning' system theoretically allows for new learning without causing interference in the consolidation of older memories. They test this idea by artificially increasing plasticity in the pre-limbic cortex and measuring changes in different learning/memory tasks. They also examined electrophysiological changes in sleep, as sleep is linked to memory formation and synaptic plasticity.

    The strengths of the study include a) meticulous analyses of a variety of electrophysiological measurements b) a combination of neurobiological and computational tools c) a largely comprehensive analysis of sleep-based changes. Some weaknesses include questions about the technique for increasing cortical plasticity (is this physiological?) and the absence of some additional experiments that would strengthen the conclusions. However, overall, the findings appear to support the general idea under examination.

    This study is likely to be very impactful as it provides some really new information about these important neural processes, as well as data that challenges popular ideas about sleep and synaptic plasticity.

    We would like to thank the reviewer for these positive comments. Answers to the weaknesses are presented below in the recommendations for the authors.

  3. eLife

    eLife assessment

    This important study reveals that slow plasticity in the neocortex is essential to prevent memory interference. The method of artificially increasing plasticity in the prefrontal cortex of rats during learning and its effect on sleep physiology, when memories are believed to be reprocessed, is solid. The work will be of interest to neuroscientists interested in learning and memory.

  4. eLife

    Reviewer #1 (Public Review):

    The authors use a newly developed object-space memory task comprising of a "Stable" version and "Overlapping" version where two objects are presented in two locations per trial in a square open field. Each version consists of 5 training trials of 5-min presentations of an object-space configuration, with both object locations staying constant across training trials in the Stable condition, and only one object location staying fixed in the Overlapping condition. Memory is tested in a test trial 24 hours later where the opposite configuration is presented - overlapping configuration presented for the Stable condition and stable configuration presented for the Overlapping condition - with the thesis that memory in this test trial for the Overlapping condition will depend on the accumulated memory of spatial patterns over the training trials, whereas memory for the test trial in the Stable condition can be due to episodic memory of last trial or accumulated memory. Memory is quantified using a Discrimination Index (DI), comparing the amount of time animals spend exploring the two object locations.

    Here, animals in other groups are also presented with an interference trial equivalent to the test trial, to test if the memory of the Overlapping condition can be disrupted. The behavioral data show that for RGS14 over-expressing animals, memory in the Overlapping condition is diminished compared to controls with no interference or controls where over-expression is inhibited, whereas memory in the Stable condition is enhanced. This is interpreted as interference in semantic-like memory formation, whereas one-shot episodic memory is improved. The authors speculate that increased cortical plasticity should lead to increased and larger delta waves according to the sleep homeostasis hypothesis, and observe that instead increased cortical plasticity leads to less non-REM sleep and smaller delta waves, with more prefrontal neurons with slower firing rates (presumably more plastic neurons). They further report increased hippocampal-cortical theta coherence during task and REM sleep, increased NonREM oscillatory coupling, and changes in hippocampal ripples in RGS14 over-expressing animals.

    While these results are interesting, there are several issues that need to be addressed, and the link between physiology and behavioral results is unclear.

    1. The behavioral results rely on the interpretation that the Overlapping condition corresponds to semantic-like memory and the Stable condition corresponds to episodic-like memory. While the dissociation in memory performance due to interference seen in these two conditions is intriguing, the Stable condition can correspond not just to the memory of the previous trial but also accumulated memory of a stable spatial pattern over the 5 testing trials, similar to accumulated memory of a changing spatial pattern in the Overlapping pattern.

    Here, it is puzzling that in the behavioral control with no interference (Figure 1D), memory in the Stable and Overlapping condition is unchanged in the test trial, with the DI statistically at 0 in the test trial. In the original description of the Object Space task by the authors in the referenced paper, the measure of memory was a Discrimination Index significantly higher than 0 in both the Stable and Overlapping conditions. This discrepancy needs to be reconciled. Is the DI for the interference trial shown in Fig. S1 significantly different than 0? No statistics or description is provided in the figure legend here.

    1. The physiology experiments compare Home cage (HC) conditions to the Object Space task (OS) throughout the manuscript. While some differences are seen in the control and RGS14 over-expressing animals, there is no comparison of the Stable vs. Overlapping condition in the physiology experiments. This precludes making explicit links between physiological observations and behavioral effects.

    2. The authors speculate that learning will result in larger and more delta waves as per the synaptic homeostasis hypothesis. It should be noted here that an alternative hypothesis is that there should also be a selective increase in synaptic plasticity for learning and consolidation. The authors do observe that control animals show more frequent and higher-amplitude delta waves, but rather than enhancing this process, RGS14 animals with increased plasticity show the opposite effect. How can this be reconciled and linked with the behavioral data in the Stable and Overlapping condition? Similarly, there is an increase in slower-firing neurons in RGS14 over-expressing animals. Slower-firing neurons have been proposed to be more plastic in the hippocampus based on their participation in learned hippocampal sequences, but appropriate references or data are needed to support the assertion that slower-firing neurons in the prefrontal cortex are more plastic.

    3. It is noted that changing cortical plasticity influences hippocampal-cortical coupling and hippocampal ripples, suggesting a cortical influence on hippocampal physiological patterns. It has been previously shown that disrupting prefrontal cortical activity does alter hippocampal ripples and hippocampal theta sequences (Schmidt et al., 2019; Schmidt and Redish, 2021). The current results should be discussed in this context.

  5. eLife

    Reviewer #2 (Public Review):

    In this paper, the authors provide evidence to support the longstanding proposition that a dual-learning system/systems-level consolidation (hippocampus attains memories at a fast pace which are eventually transmitted to the slow-learning neocortex) allows rapid acquisition of new memories while protecting pre-existing memories. The authors leverage many techniques (behavior, pharmacology, electrophysiology, modelling) and report a host of behavioral and electrophysiological changes on induction of increased medial prefrontal cortex (mPFC) plasticity which are interesting and will be of significant interest to the broad readership.

    The experimental design and analyses are convincing (barring some instances which are discussed below). The following recommendations will bolster the strength/quality of the manuscript:

    1. Certain concerns regarding the interpretation and analysis of the behavioral data remain. The authors need to clarify if increased mPFC plasticity leads to only an increase in one-shot memory or 'also' interference of previous information. It seems that the behavioral results could also be explained by the more parsimonious explanation that one-shot memory is improved. Do the current controls tease apart these two scenarios? Additionally, the authors need to clarify why the 'no trial' and 'anisomycin' controls for the stable task perform at chance levels on exposure to a new object-place association on test day (Fig 1D). Finally, further description of how the discrimination index (exploration time of novel-exploration time of familiar/sum of both) is recommended i.e., in the stable condition, which 'object' is chosen as 'novel' (as both are in the same locations) for computing the index (Fig 1). Do negative DI values imply a neophobia to novel objects (and thus are a form of memory; this is also crucial because the modelling results (Fig 1E) use both neophilia and neophobia while negative discrimination indexes are considered similar to 0 for interpreting the behavioral results, as stated on page 3, lines 84-86?

    2. The authors report lower firing rates in RGS14414 animals during the task in Fig 2F. It is indeed remarkable how large the reported differences are. The authors need to rule out any differences in the behavioral state of the animals in the two groups during the task, i.e., rest vs. active exploration/movement dynamics. Are only epochs during the task while the animals interact with the objects used for computing the firing rates (same epochs as Fig 1)? If not, doing so will provide a useful comparison with Fig 1. Additionally, although the authors make the case for slow firing rate neurons being important for plasticity (based on Grosmark and Buzsaki, 2016), it is crucial to note that the firing rate dynamic (slow vs. fast) in that study for the hippocampus is defined based on the whole recorded session (predominated by sleep), indeed the firing rates of the two groups (slow vs. fast/plastic vs. rigid) during the task/maze-running do not differ in that study. Therefore, the results here seem incongruent with the Grosmark and Buzsaki paper. Since this finding is central to the main claim of the authors, it either warrants further investigation or a re-interpretation of their results.

    3. A concern remains as to how many of the electrophysiological changes they observe (firing rate differences, LFP differences including coupling, sleep state differences, Figs. 2-4) support their main hypothesis or are a by-product of injection of RGS14414 (for instance, one might argue that an increased 'capability' to learn new information/more plasticity might lead to more NREM sleep for consolidation, etc.). The authors need to carefully interpret all their data in light of their main hypothesis, which will substantially improve the quality/strength of the manuscript.

  6. eLife

    Reviewer #3 (Public Review):

    The authors set out to test the idea that memories involve a fast process (for the acquisition of new information) and a slow process (where these memories are progressively transferred/integrated into more-long term storage). The former process involves the hippocampus and the latter the cerebral cortex. This 'dual-learning' system theoretically allows for new learning without causing interference in the consolidation of older memories. They test this idea by artificially increasing plasticity in the pre-limbic cortex and measuring changes in different learning/memory tasks. They also examined electrophysiological changes in sleep, as sleep is linked to memory formation and synaptic plasticity.

    The strengths of the study include a) meticulous analyses of a variety of electrophysiological measurements b) a combination of neurobiological and computational tools c) a largely comprehensive analysis of sleep-based changes. Some weaknesses include questions about the technique for increasing cortical plasticity (is this physiological?) and the absence of some additional experiments that would strengthen the conclusions. However, overall, the findings appear to support the general idea under examination.

    This study is likely to be very impactful as it provides some really new information about these important neural processes, as well as data that challenges popular ideas about sleep and synaptic plasticity.

  7. Version published to 10.1101/2022.11.21.517356v1 on bioRxiv