Bidirectional synaptic plasticity rapidly modifies hippocampal representations

  1. Aaron D Milstein
  2. Yiding Li
  3. Katie C Bittner
  4. Christine Grienberger
  5. Ivan Soltesz
  6. Jeffrey C Magee
  7. Sandro Romani

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    This manuscript uses a combination of high-quality in vivo electrophysiology and modelling to demonstrate that Behavioural Time Scale Plasticity (BTSP) is bidirectional, and the amplitude and direction of this plasticity are dictated by the current weight of the inputs and not by the correlated activity of pairs of neurons. These findings challenge our current views on synaptic plasticity, which are primarily based on Hebb's concept. In addition, the network model used in this study demonstrates that this type of plasticity can rapidly reshape population activity to respond to environmental clues. This study will be of interest to the broad neuroscience audience and foster new ideas on biological and artificial learning.

    (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 and Reviewer #2 agreed to share their names with the authors.)

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Learning requires neural adaptations thought to be mediated by activity-dependent synaptic plasticity. A relatively non-standard form of synaptic plasticity driven by dendritic calcium spikes, or plateau potentials, has been reported to underlie place field formation in rodent hippocampal CA1 neurons. Here, we found that this behavioral timescale synaptic plasticity (BTSP) can also reshape existing place fields via bidirectional synaptic weight changes that depend on the temporal proximity of plateau potentials to pre-existing place fields. When evoked near an existing place field, plateau potentials induced less synaptic potentiation and more depression, suggesting BTSP might depend inversely on postsynaptic activation. However, manipulations of place cell membrane potential and computational modeling indicated that this anti-correlation actually results from a dependence on current synaptic weight such that weak inputs potentiate and strong inputs depress. A network model implementing this bidirectional synaptic learning rule suggested that BTSP enables population activity, rather than pairwise neuronal correlations, to drive neural adaptations to experience.

Article activity feed

  1. Version published to 10.7554/elife.73046 on eLife
  2. eLife

    Author Response:

    Reviewer #1:

    This paper uses intracellular patch-clamp recordings of hippocampal CA1 pyramidal cells in awake mice running on a cued treadmill to investigate whether dendritic plateau potentials, which can induce place fields in silent cells through Behavioral Time Scale Plasticity (BTSP), could also modify the spatial modulation of existing place cells. They report that plateau potentials can lead to the formation of a secondary place field by synaptic potentiation while reducing the primary place field by synaptic depression. As for place fields induced in silent cells, the spatial extend of this bi-directional plasticity depends on the speed of the animal during induction suggesting a fixed time course. Further analysis revealed that the sign and magnitude of Vm changes varied in a distance/time -dependent manner from the location/time of plateau induction such that Vm tended to increase at plateau location and to decrease away from the plateau in both directions adding a bidirectional property to previously described BTSP. The sign and magnitude of the plasticity also depends on the value of Vm at the time/location of plateau induction such that if Vm is more hyperpolarized than -55 mV the plasticity induces a depolarization at the plateau location and less hyperpolarization away from that location as observed in silent cells while if Vm is more depolarized than -55 mV the plasticity induces a smaller depolarization and more hyperpolarization further away. The authors then used Vm manipulations and computational modeling to show that the critical factor was not the absolute Vm level but instead the level of potentiation of activated synapses. Finally, the authors used a network model of CA1 to show that BTSP can account for over-representation at reward location within a familiar environment. Altogether, this work represents a nice combination of cutting-edge experimental work and modeling that shed new lights on cellular mechanisms allowing experience -dependent modifications of spatial maps in familiar environments.

    Major comments:

    1. The procedure for BTSP induction is described without much details in the text and method. According to the text from 1 to 8 laps/ stimulation were used to induce the secondary place field. Why is there such variability? I guess that sometimes one stimulation is not enough to induce the place field but this should be clearly stated in the introduction. Also how does one know if a new place field is induced? Sometimes the new and old place field strongly overlap (e.g. Fig. 1C, blue trace) and it must be difficult "by eye" to decide if a new place field was induced. For readers to get a better idea of the all process could you report the average number of laps/stimulations used to induce the secondary place field? Could you mention how many laps/stimulations were used for the example traces shown in Figure 1C. Was the success rate 100%? How many laps were recorded after induction and used to compute the average traces shown in Figure 1C? Could you please report those numbers in the text and the figure legend? Also, in case a primary place field was induced by BTSP how many laps/stimulations were used? Was it more difficult to induce a secondary place field compared to a primary one?

    We have now included additional details in a new Supplementary Figure S1, and have added the following text to the Materials and Methods:

    P25, L664: “Plasticity was induced in vivo by injecting current (700 pA, 300 ms) intracellularly into recorded CA1 neurons to evoke dendritic plateau potentials at the same position on the circular treadmill for multiple consecutive laps. In most cases, plateaus were evoked on five consecutive laps (Figure S1D, left). However, during some experiments, large changes in the spatial Vm ramp depolarization could be observed to develop after as few as one plateau (consistent with the observation that plasticity could be induced by a single spontaneously-occurring plateau), and so fewer induction laps were used. In other experiments, plateaus were induced on more than five consecutive laps if place field expression remained weak after the first five trials (Figure S1D, left). The source of this variability across cells/animals is not yet clear, and requires future investigation. Overall, this procedure induced changes in spatial Vm ramp depolarization in 100% of cells in which it was attempted by three investigators. In some cells, the initial place field was first induced by this procedure, and then the procedure was repeated a second or third time in the same cell with plateaus induced at different locations. In those cases, there was no systematic difference in the number of plateaus required to induce the first place field compared to subsequent fields (Figure S1D, right).”

    We also now report in the legend of Figure 1 and in the Materials and Methods (P27, L716) that spatial Vm ramps before and after plasticity were computed by averaging across 10 laps.

    1. Along the same line the behavior of the mice is not extensively described. However, behavior and notably running speed seems to have a major impact on the spatial extend of the plasticity. Could you plot the speed profile of mice below voltage traces for example in Figure 1B, S1I and so on. Also could you show lap by lap speed profiles superimposed for one recording session to have an idea of the variability and overall stereotypy of behavior. It appears that the place fields can span the border between laps (i.e. start before and end after the reward zone). How is it possible if animal stops at the end of a lap to get reward? Usually these stops induce a state change with reduced theta oscillations, which is less favorable to place cell coding. Is it the case that some animals do not stop at reward location? Could you give more details here?

    We have now included additional details in a new Supplementary Figure S1, and have added the following text to the Materials and Methods:

    P26, L679: “Since the time window for plasticity induction by BTSP extends for seconds around each plateau, and plateaus were typically evoked on multiple consecutive laps, the changes in synaptic weights induced by BTSP depended on the run behavior of the animals across all induction laps. We showed in Figure 3D that the spatial width of place fields induced by BTSP varied with the average velocity of animals across all plasticity induction laps. Another factor that contributed to the spatial width of induced fields is the proximity of the evoked plateaus to the reward site, as animals tended to stop running briefly to lick near the fixed reward site. Variability across laps in either the run velocity or the duration of pauses could pose a challenge in trying to relate spatial changes in Vm ramp depolarization to the time delay to the plateau (see below). Figure S1 shows the full run trajectories of animals during all plasticity induction laps for the five example cells shown in Figure 1. While some variability across induction laps was observed, each animal tended to run consistently at similar velocities across laps.”

    Please also note that, on the circular treadmill, the place fields of presynaptic neurons in CA3 can “wrap around” the track (e.g. see presynaptic firing rates schematized in Figure 3J). In some cases, this meant that the same synapse that generated an eligibility trace and underwent plasticity before the animal stopped to lick for reward, was also activated and contributed to Vm ramp depolarization once the animal continued running, since the spatial positions were traversed contiguously. In the model, spatial CA3 inputs were considered to be silent during pauses in running.

    1. The rationale behind the analysis of delta Vm against time from plateau induction shown in Figure 2E, 3 and 4 and associated supplementary figures is not clear from the text and method sections. If I understood well this analysis uses the difference between average Vm of several laps after the induction laps minus the average Vm of several laps before the induction laps but then uses the speed of the animal during the induction laps to convert this delta Vm trace in the temporal domain. But this assumes a relatively constant behavior of the animal during induction. If induction is performed over 1 or 2 laps the chance of a constant speed are probably higher than if it is performed over say 7-8 laps. If the animal slows down consistently or even stops during induction laps 6-8 but runs fast during induction laps 1-5 how does one interpret the DeltaVm over time representation? Authors should report the number of laps used for induction for the traces illustrated in Fig. 2 and the time against position traces for all individual induction laps superimposed on top of the average in Fig. 2C and delta Vm against time traces for all individual induction laps superimposed on top of the average in Fig. 2E.

    We have now included additional details in a new Supplementary Figure S1, and have added the following text to the Materials and Methods:

    P27, L732: “In order to relate spatial changes in Vm ramp depolarization to the time delay to a plateau (e.g. Figures 2E, 2F, 3A – 3F, 3I, 4B, 4C, 4E, 4F and 6E), we assigned to each spatial position the shortest time delay to plateau that occurred across multiple induction laps (Figure S1). This is a conservative estimate, as the shortest delay between presynaptic activity and postsynaptic plateau will generate the largest overlap between eligibility traces (ET) and instructive signals (IS), and will result in the largest changes in synaptic weight. While this method is imperfect and did discard variability in running behavior across laps, it enabled direct comparison of the time-course of BTSP across neurons. We also note that, to generate the modeling results shown in Figure 6, the full run trajectory of each animal during all induction laps, including pauses, was provided as input to the model (see details below). This resulted in good quantitative agreement between experimentally-recorded and modeled spatial Vm ramps (Figure 6D).”

    1. In Figure 3D it is unclear where the PCs within-field data comes from. The n = 26 suggests that this data includes all stimulation but in most cases induction was performed by stimulating outside place field location (as shown in Figure 1D) and induction is done by stimulating always in the same position (except for 2/24 cells were there was a third induction). Could you please specify?

    We acknowledge this it was confusing how data points were selected for inclusion in the category “PCs (within-field)” in Figures 3D and 4C. For each of the 26 inductions performed in cells with pre-existing place fields, cells were most depolarized at spatial bins within their place fields, but were also relatively hyperpolarized at positions outside their place fields. On this background, we induced plasticity by evoking plateau potentials at a fixed location, which was at a different distance from the initial place field in different cells, as highlighted in Figure 1D. In Figure 3D, we sought to determine if changes in Vm ramp were different at positions that were depolarized, compared to positions that were hyperpolarized. To do this, we pooled data from all 26 inductions, selecting only the spatial bins in each recording where the Vm ramp depolarization exceeded a threshold of -56 mV.

    We have now revised the text to clarify this point:

    P8, L184: “In Figures 3D – F, we examined this further by comparing data from initially hyperpolarized silent cells (black; n=29 inductions, see Figure S3 and Materials and Methods) to data from place cells (dark red; n=26 inductions). Place cells were on average more depolarized before plasticity than silent cells (Figure 3D), and more depression occurred in place cells compared to silent cells (Figure 3E). However, each place cell had both spatial positions where it was depolarized within its place field, and positions where it was hyperpolarized out-of-field. To determine if spatial positions that were initially depolarized were associated with larger depression, we grouped Vm ramp data from all place cells, considering only spatial bins where each cell was more depolarized than a threshold of -56 mV (light red traces labeled “PCs (within-field)” in Figures 3D – F). Indeed, more depression and less potentiation was induced in place cells at those spatial positions that were initially most depolarized (Figure 3E).”

    1. In the modeling experiment (Figure 6 A) it is unclear why the dV/dt trace show no change for synapses activated before the plateau (unlike what is illustrated in Figure 5C). In my understanding the eligibility trace of these synapses shown in green allow them to be potentiated by a certain amount that depends on the overlap of their eligibility trace with the instructive signal. Maybe to facilitate understanding authors could show the post-synaptic potentials before and after plateau induction in Figure 5A.

    The Reviewer pointed out that in Figure 5, changes in synaptic weight occur for inputs activated before a plateau, but this appeared to not occur in Figure 6A. This is not the case – in both Figures, an eligibility trace (ET) is generated at the time of presynaptic activity, but changes in synaptic weight do not occur until later when the plateau arrives and an instructive signal (IS) is generated. In the example shown in Figure 6A, nonzero changes in synaptic weight occur for all presynaptic inputs (each row in the bottom panel labeled dW/dt). However, these changes do not begin until after the plateau is initiated. We have revised the text to clarify this point:

    P20, L498: “Note that, at inputs activated before the onset time of the plateau, changes in synaptic weight (bottom row) do not begin until after plateau onset when the instructive signal IS and the signal overlap ET*IS are nonzero.”

  3. eLife

    Evaluation Summary:

    This manuscript uses a combination of high-quality in vivo electrophysiology and modelling to demonstrate that Behavioural Time Scale Plasticity (BTSP) is bidirectional, and the amplitude and direction of this plasticity are dictated by the current weight of the inputs and not by the correlated activity of pairs of neurons. These findings challenge our current views on synaptic plasticity, which are primarily based on Hebb's concept. In addition, the network model used in this study demonstrates that this type of plasticity can rapidly reshape population activity to respond to environmental clues. This study will be of interest to the broad neuroscience audience and foster new ideas on biological and artificial learning.

    (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 and Reviewer #2 agreed to share their names with the authors.)

  4. eLife

    Reviewer #1 (Public Review):

    This paper uses intracellular patch-clamp recordings of hippocampal CA1 pyramidal cells in awake mice running on a cued treadmill to investigate whether dendritic plateau potentials, which can induce place fields in silent cells through Behavioral Time Scale Plasticity (BTSP), could also modify the spatial modulation of existing place cells. They report that plateau potentials can lead to the formation of a secondary place field by synaptic potentiation while reducing the primary place field by synaptic depression. As for place fields induced in silent cells, the spatial extend of this bi-directional plasticity depends on the speed of the animal during induction suggesting a fixed time course. Further analysis revealed that the sign and magnitude of Vm changes varied in a distance/time -dependent manner from the location/time of plateau induction such that Vm tended to increase at plateau location and to decrease away from the plateau in both directions adding a bidirectional property to previously described BTSP. The sign and magnitude of the plasticity also depends on the value of Vm at the time/location of plateau induction such that if Vm is more hyperpolarized than -55 mV the plasticity induces a depolarization at the plateau location and less hyperpolarization away from that location as observed in silent cells while if Vm is more depolarized than -55 mV the plasticity induces a smaller depolarization and more hyperpolarization further away. The authors then used Vm manipulations and computational modeling to show that the critical factor was not the absolute Vm level but instead the level of potentiation of activated synapses. Finally, the authors used a network model of CA1 to show that BTSP can account for over-representation at reward location within a familiar environment. Altogether, this work represents a nice combination of cutting-edge experimental work and modeling that shed new lights on cellular mechanisms allowing experience -dependent modifications of spatial maps in familiar environments.

    Major comments:

    1. The procedure for BTSP induction is described without much details in the text and method. According to the text from 1 to 8 laps/ stimulation were used to induce the secondary place field. Why is there such variability? I guess that sometimes one stimulation is not enough to induce the place field but this should be clearly stated in the introduction. Also how does one know if a new place field is induced? Sometimes the new and old place field strongly overlap (e.g. Fig. 1C, blue trace) and it must be difficult "by eye" to decide if a new place field was induced. For readers to get a better idea of the all process could you report the average number of laps/stimulations used to induce the secondary place field? Could you mention how many laps/stimulations were used for the example traces shown in Figure 1C. Was the success rate 100%? How many laps were recorded after induction and used to compute the average traces shown in Figure 1C? Could you please report those numbers in the text and the figure legend? Also, in case a primary place field was induced by BTSP how many laps/stimulations were used? Was it more difficult to induce a secondary place field compared to a primary one?

    2. Along the same line the behavior of the mice is not extensively described. However, behavior and notably running speed seems to have a major impact on the spatial extend of the plasticity. Could you plot the speed profile of mice below voltage traces for example in Figure 1B, S1I and so on. Also could you show lap by lap speed profiles superimposed for one recording session to have an idea of the variability and overall stereotypy of behavior. It appears that the place fields can span the border between laps (i.e. start before and end after the reward zone). How is it possible if animal stops at the end of a lap to get reward? Usually these stops induce a state change with reduced theta oscillations, which is less favorable to place cell coding. Is it the case that some animals do not stop at reward location? Could you give more details here?

    3. The rationale behind the analysis of delta Vm against time from plateau induction shown in Figure 2E, 3 and 4 and associated supplementary figures is not clear from the text and method sections. If I understood well this analysis uses the difference between average Vm of several laps after the induction laps minus the average Vm of several laps before the induction laps but then uses the speed of the animal during the induction laps to convert this delta Vm trace in the temporal domain. But this assumes a relatively constant behavior of the animal during induction. If induction is performed over 1 or 2 laps the chance of a constant speed are probably higher than if it is performed over say 7-8 laps. If the animal slows down consistently or even stops during induction laps 6-8 but runs fast during induction laps 1-5 how does one interpret the DeltaVm over time representation? Authors should report the number of laps used for induction for the traces illustrated in Fig. 2 and the time against position traces for all individual induction laps superimposed on top of the average in Fig. 2C and delta Vm against time traces for all individual induction laps superimposed on top of the average in Fig. 2E.

    4. In Figure 3D it is unclear where the PCs within-field data comes from. The n = 26 suggests that this data includes all stimulation but in most cases induction was performed by stimulating outside place field location (as shown in Figure 1D) and induction is done by stimulating always in the same position (except for 2/24 cells were there was a third induction). Could you please specify?

    5. In the modeling experiment (Figure 6 A) it is unclear why the dV/dt trace show no change for synapses activated before the plateau (unlike what is illustrated in Figure 5C). In my understanding the eligibility trace of these synapses shown in green allow them to be potentiated by a certain amount that depends on the overlap of their eligibility trace with the instructive signal. Maybe to facilitate understanding authors could show the post-synaptic potentials before and after plateau induction in Figure 5A.

  5. eLife

    Reviewer #2 (Public Review):

    What causes the synapses to give way in long-term potentiation (LTP) or long term depression (LTD)? Much work in the past 30 years has focused on the relative timing of pre- and post-synaptic spikes (STDP), post-synaptic voltage and the presence of neuromodulators as the most salient factors. But theory has not been able to lead to fast and accurate learning using only these factors, nor has experiment been able to show their causal role in producing representations. The discovery that relative timing between pre-synaptic activity and plateau potentials CA1 pyramidal cells regulates a fast and strong expression of LTD, which causes new representations in behaving animals, has offered a new hope.

    The original studies on behavioural time-scale plasticity (BTSP) left many crucial questions unanswered. Foremost is the absence of LTD. Are plateaus not engaging LTD at all or is LTD regulated by other factors altogether? Milstein et al. provide an answer to this question. LTD is engaged by plateaus, but only for synapses that have been potentiated, or synapses we expect have a strong weight. According to this point of view, previous attempt did not observed LTD because the protocol triggered only synapses with weak synaptic weights.

    Surprisingly, the dependence of synaptic plasticity on pre-post timing remains approximately symmetric (as in Bittner et al.); with LTD being engaged at the same time whether the post-synaptic plateau or the pre-synaptic firing came before one another. This enhances the paradigm shift offered by BTSP. First because it suggests a convergence of synaptic mechanisms: the effect of pre-synaptic firing on the trace left by the post-synaptic plateau is the same as the effect of post-synaptic plateau on the trace left by pre-synaptic firing. And second, because this asymmetry is firmly against Hebb's postulate. Pre-synaptic firing does not have to 'repeatedly or persistently take part in firing' the post synaptic neuron, a plateau generated by another pathway before the occurence of pre-synaptic firing is sufficient to induce both LTP and LTD.

    The work of Milstein et al. is also important because it contains a computational model, a hypothesis consistent with their experimental data and the molecular pathways involved in synaptic plasticity. They show that when the two opposing forces for LTP and LTD have distinct parameters (LTD being triggered by lower temporal coincidence but with a higher sensitivity and saturating earlier than LTP), they can recapitulate their experimental observations in vivo.

    I believe this is likely to become a seminal article in the field. The only weak point I see is I think the results from the modelling section could come out more clearly, particularly the crucial dependence of the model on having distinct functions for LTP and LTD.

  6. Version published to 10.1101/2020.02.04.934182v2 on bioRxiv
  7. Version published to 10.1101/2020.02.04.934182v1 on bioRxiv