Feedback inhibition underlies new computational functions of cerebellar interneurons

  1. Hunter E Halverson
  2. Jinsook Kim
  3. Andrei Khilkevich
  4. Michael D Mauk
  5. George J Augustine

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    In this manuscript, the authors describe an inhibitory pathway from Purkinje cells in the cerebellum to a subset of molecular layer interneurons. The authors use in-vivo recordings to characterize these synaptic connections and probe their function during a delay conditioning task in vivo and using computer simulations. This is informative and an advance, but some claims regarding the function of this pathway need stronger substantiation. This is relevant to experimentalists and modelers interested in the cerebellum.

    (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 #2 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

The function of a feedback inhibitory circuit between cerebellar Purkinje cells and molecular layer interneurons (MLIs) was defined by combining optogenetics, neuronal activity recordings both in cerebellar slices and in vivo, and computational modeling. Purkinje cells inhibit a subset of MLIs in the inner third of the molecular layer. This inhibition is non-reciprocal, short-range (less than 200 μm) and is based on convergence of one to two Purkinje cells onto MLIs. During learning-related eyelid movements in vivo, the activity of a subset of MLIs progressively increases as Purkinje cell activity decreases, with Purkinje cells usually leading the MLIs. Computer simulations indicate that these relationships are best explained by the feedback circuit from Purkinje cells to MLIs and that this feedback circuit plays a central role in making cerebellar learning efficient.

Article activity feed

  1. Version published to 10.7554/elife.77603 on eLife
  2. eLife

    Author Response

    Reviewer 1# (Public Review):

    Purkinje cells (PCs) in the cerebellum extend axonal collaterals along the PC layer and within the molecular layer. Previous anatomical studies have shown the existence of these tracts and recently, the existence of functional synapses from PCs to PCs, molecular layer interneurons (MLIs), and other cell types was demonstrated by Witter et al., (Neuron, 2016) using optogenetics. In this manuscript, Halverson et al., first characterize the PC to MLI synapse properties in the slice using optogenetics and dual patch recordings. They then use computer simulations to predict the role of these connections in eyelid conditioning and test these predictions using in vivo recordings in rabbits. Authors claim that PCs fire before their target MLIs and that their activity is anticorrelated. They further suggest that the special class of MLIs receiving inhibitory input from PCs might serve to synchronize PCs during eyelid conditioning.

    Major comments:

    1. The manuscript is quite long with 9 main figure panels and 6 supplementary figures. The flow of the results is not smooth. While the first 4 figures are nicely done in terms of their results and organization, the same cannot be said about the rest of the figures.

    To address this concern, we have revised the Results section extensively. We believe that it is now much more accessible and better integrated.

    In fact, it would make sense to split the manuscript in two, one describing the synaptic properties and circuit mapping of the PC-PC-MLI circuit and the other describing their role in eyelid conditioning. As it stands, this manuscript is a tough read and difficult to get through.

    We acknowledge that our results, which were done in two different labs and employed a variety of different techniques, could have been split into two (or even more) separate papers. However, we believe that there is high value to our readers in providing a comprehensive study that integrates many different types of analyses to attack the same fundamental question. That is why we chose to organize the content in the way that we did and that is why we still prefer to keep the entire story together. However, we do agree with the reviewer’s point that the previous version was unwieldy and too challenging to understand. Therefore, we have invested a lot of effort to improve the readability of the revised version.

    Further, the authors have not connected the initial slice physiology with the later in vivo work to argue for their presence in the same paper. For example, the quantal content measurement, the short-term plasticity, the mobilization rate measurement, etc do not figure in the latter half of the manuscript at all. I strongly suggest carving figures 1-4 out into a separate manuscript.

    The slice work motivated the computational simulation and the in vivo recordings of MLI activity. While it is true that it is hard to correlate every aspect of the slice work (e.g. quantal content measurements, etc.) with the in vivo recordings, and vice versa, there are elements of each that have informed the other. As a result, consistent properties of the PC-to-PC-MLI circuit emerged. We have highlighted the cross-connections more in the revised manuscript, including the following passages:

    1. “Only a subset of MLIs (8.7%) showed clear inverse correlation with eyelid PCs and they were within approximately 120 µm of eyelid PCs. These connectivity rates and distances are comparable to our observations in cerebellar slices, where we found that approximately 5-6% of MLIs receive PC feedback inhibition (Figure 1b) that extends over 200 m or less (Figure 3).” (p.13, para 1)

    2. “The pattern of cross-correlation between connected PCs and PC-MLIs was qualitatively similar to that observed in slices…” (p.14, para 2).

    3. “This correspondence strengthens the conclusion that putative PC-MLI identified in vivo are equivalent to the PC-MLI identified in slices” (p. 15, para 1).

    4. “The need for relatively large changes in PC activity in vivo highlights the importance of the frequency-independent synaptic synaptic transmission at the PC-to-PC-MLI synapse illustrated in Figure 2.” (p. 16, para. 1)

    We have more closely harmonized the style of all figures, to subliminally emphasize the close connection between the slice and in vivo results.

    Above we have addressed the suggestion to split the paper into two. Instead of breaking up the paper, we worked hard to better integrate the two parts and make them easier to read as a whole.

    1. Authors conclude that eyelid PCs and eyelid PC-MLIs are inversely correlated and that PCs precede PC-MLIs during CRs and therefore could drive their activity. Both of these points are insufficiently justified by their analysis. First, it is not clear how eyelid PCs are identified – I’m assuming this is based on negative correlation with CRs just like positively correlated MLIs are assigned as eyelid PC-MLIs.

    We apologize for failing to mention that eyelid PCs were identified by the presence of US -evoked (eyelid stimulation) complex spikes. This criterion is completely independent of the responses of PCs during expression of eyelid CRs and also provides an in vivo tool for identifying the “eyelid” region of the cerebellar cortex, which should also be where eyelid PC-MLIs are located. To address this omission, we have now describe the method used to identify eyelid PCs in the Methods section (p. 31, para. 2) and the Results section (p. 11, para. 2).

    If this is how PCs and PC-MLIs are identified, then the inverse correlation between the two cell types results from this definition itself. And, their activity pattern during CRs, illustrated in many figure panels is hardly surprising.

    Yes of course this would be circular logic, but it is not at all what we did! Again, we apologize for the confusion.

    Second, to show that PCs fire ahead of PC-MLIs, the authors calculate the difference in fractional change in spike rate before and after the start of the CR (PC-MLI). Their reasoning is that if the bulk of firing rate change happened before the start of CR for PCs, but at the start or later for PC-MLIs, then this value will be positive, else it will be negative. The distribution of these values was shifted to the positive side leading them to conclude that PCs fire ahead of PC-MLIs. However, this is a huge logical jump. The sign of (PC-MLI) is dependent on the depth of modulation in each cell type as well and does not necessarily indicate relative timing. In any case, such caveats have not been ruled out in their analysis. This analysis to establish timing is unconvincing. Would it not be better to look at the timing of the spike modulation start directly rather than the round-about method they are using?

    We agree with the reviewer that PC and PC-MLI activities undergo complex time-dependent changes, particularly during CRs, which makes it challenging to have a single parameter that uniquely represents the differences in timing between the activity of the two cell types. In our revised manuscript, we have addressed this issue by creating a new section that is entirely devoted to analysis of the temporal relationship between PC and PC-MLI activity (pp. 14-17). In brief, here are the main lines of evidence that PCs fire prior to PC-MLIs, both in baseline conditions and during conditioned eyelid responses.

    We have provided 3 types of evidence that PCs fire prior to putative PC-MLIs during baseline activity:

    1. A spike-triggered average of PC and putative PC-MLI activity during baseline firing showed a modest decrease in PC-MLI firing rate in response to a PC action potential (Figure 8a; see also Figure 8-figure supplement 1a and 1b).

    2. A pause in PC activity caused a very substantial rise in activity in putative PC-MLIs (Figure 8c; see also Figure 8-figure supplement 1c).

    3. A burst of PC activity caused a decline in putative PC-MLI activity (Figure 8d; see also Figure 8-figure supplement 1d).

    We have an additional 3 lines of evidence showing that PCs fire prior to putative PC-MLIs during CRs:

    1. Simultaneous recordings of the time course in changes in PC and putative PC-MLI activity during CRs indicate that PC activity usually declined prior to the activity of putative PC-MLIs. This is clearly visible in the examples shown in Figure 9c, as well as the averaged data shown in Figure 9-figure supplement 1a.

    2. We measured the delay between the time at which PC activity reached 50% of its minimum during the CS and the time at which the activity of putative PC-MLIs reached 50% of its maximum during single trials. Whenever CRs were observed, PCs reached their half-maximal response before putative PC-MLIs did (Figure 9─figure supplement 1b).

    3. We also measured the collective timing of changes in the activity of putative PC-MLIs and eyelid PCs during conditioning across all of our paired recordings. This was done by calculating a ratio representing the magnitude of changes in activity prior to CR onset, normalized to the peak amplitude of the change during the entire interval. The distribution of differences in the timing of changes in PC and PC-MLI activity has a mean that is greater than zero (Figure 10), indicating that eyelid PCs decreased their activity before putative PC-MLIs increased their activity in a majority of cases.

    We hope that these improvements have adequately addressed the reviewer’s concern.

    1. Many figure panels make the same point and appear redundant. For example, that PCs and PC-MLIs are inversely correlated with each other in vivo during CRs is shown in Figure 7, figure 8a, S2, S4, and S5. Of course, in each case, the data are sorted differently (according to ISI, CR initiation, cumulative distributions, etc.,) but surely, the point regarding inverse relationship can be conveyed more concisely?

    As mentioned in response to the reviewer’s previous comment, we have made significant changes to this part of the manuscript, including creating a section that addresses the temporal relationship between PC and PC-MLI activity. This has involved removing some of the analyses listed by the reviewer, adding some new analysis and relegating some previous figures to the supplementary materials. We believe that these changes allow the manuscript to efficiently clarify the relationship between PC and PC-MLI activity and highlight the value of each type of analysis that is included.

    1. Several details are missing in the methods section even though parts of it may have been published before. For instance, how are CRs calculated in the simulation? Methods state that 'The averaged and smoothed activity of the eight deep nucleus neurons was used to represent the output of the simulation and the predicted "eyelid response" of the simulation'. It is not clear what the nature of this transform is and if any calibration factors were used. How comparable are the simulated CRs in kinetics and amplitude to experimental CRs?

    In response to this comment, we have revised the Methods section to include much more detail about the simulation methods, including two schematic diagrams. The methods employed for the experimental work in the paper are already described in detail.

    The simulation can produce simulated CRs (smoothed histogram of nucleus activity) with kinematic variables that are comparable to experimental CRs. A detailed account of how this is accomplished is described in Medina and Mauk (2000) and are briefly summarized on p. 39, para.2: “The averaged and smoothed activity of the eight deep nucleus neurons was used to represent the output of the simulation and the predicted “eyelid response” of the simulation.” Although this approach is not intended to simulate the precise kinematics of an eyeblink, comparison of Figures 11a (simulation) and 6a (rabbit) show that there is a reasonable concordance. The real value of the simulation is in predicting the relative changes in eyelid responses that occur during conditioning.

  3. eLife

    Evaluation Summary:

    In this manuscript, the authors describe an inhibitory pathway from Purkinje cells in the cerebellum to a subset of molecular layer interneurons. The authors use in-vivo recordings to characterize these synaptic connections and probe their function during a delay conditioning task in vivo and using computer simulations. This is informative and an advance, but some claims regarding the function of this pathway need stronger substantiation. This is relevant to experimentalists and modelers interested in the cerebellum.

    (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 #2 agreed to share their name with the authors.)

  4. eLife

    Reviewer #2 (Public Review):

    The authors show that Purkinje cells (PCs) may inhibit a subset of nearby molecular layer interneurons (MLIs); this connection is never reciprocal, i.e. PC activity can regulate MLI activity and thus the modulation of other PCs, but is not triggering a delayed modulation of its own activity. Modeling and in vivo recordings demonstrate that this MLI recruitment - as well as plasticity at parallel fiber inputs onto these PCs - is required for the temporal relationships that enable spike pauses in PCs and related conditioned responses. The work is very informative and will advance the field. Minor concerns that should be addressed are whether the model allows for an assessment of how many MLIs need to be modulated to explain the pause in PC firing and what the cause for the early phase in the PC pause is before MLI spike rates are modulated. In addition, the model presented here includes bidirectional parallel fiber-PC plasticity, but it is not described what the specific roles of LTD and LTP might be, and whether indeed there is a role for LTP at all.

  5. eLife

    Reviewer 1# (Public Review):

    Purkinje cells (PCs) in the cerebellum extend axonal collaterals along the PC layer and within the molecular layer. Previous anatomical studies have shown the existence of these tracts and recently, the existence of functional synapses from PCs to PCs, molecular layer interneurons (MLIs), and other cell types was demonstrated by Witter et al., (Neuron, 2016) using optogenetics. In this manuscript, Halverson et al., first characterize the PC to MLI synapse properties in the slice using optogenetics and dual patch recordings. They then use computer simulations to predict the role of these connections in eyelid conditioning and test these predictions using in vivo recordings in rabbits. Authors claim that PCs fire before their target MLIs and that their activity is anticorrelated. They further suggest that the special class of MLIs receiving inhibitory input from PCs might serve to synchronize PCs during eyelid conditioning.

    Major comments:

    1. The manuscript is quite long with 9 main figure panels and 6 supplementary figures. The flow of the results is not smooth. While the first 4 figures are nicely done in terms of their results and organization, the same cannot be said about the rest of the figures. In fact, it would make sense to split the manuscript in two, one describing the synaptic properties and circuit mapping of the PC-PC-MLI circuit and the other describing their role in eyelid conditioning. As it stands, this manuscript is a tough read and difficult to get through. Further, the authors have not connected the initial slice physiology with the later in vivo work to argue for their presence in the same paper. For example, the quantal content measurement, the short-term plasticity, the mobilization rate measurement, etc do not figure in the latter half of the manuscript at all. I strongly suggest carving figures 1-4 out into a separate manuscript.

    2. Authors conclude that eyelid PCs and eyelid PC-MLIs are inversely correlated and that PCs precede PC-MLIs during CRs and therefore could drive their activity. Both of these points are insufficiently justified by their analysis. First, it is not clear how eyelid PCs are identified - I'm assuming this is based on negative correlation with CRs just like positively correlated MLIs are assigned as eyelid PC-MLIs. If this is how PCs and PC-MLIs are identified, then the inverse correlation between the two cell types results from this definition itself. And, their activity pattern during CRs, illustrated in many figure panels is hardly surprising.

    Second, to show that PCs fire ahead of PC-MLIs, the authors calculate the difference in fractional change in spike rate before and after the start of the CR (PC-MLI). Their reasoning is that if the bulk of firing rate change happened before the start of CR for PCs, but at the start or later for PC-MLIs, then this value will be positive, else it will be negative. The distribution of these values was shifted to the positive side leading them to conclude that PCs fire ahead of PC-MLIs. However, this is a huge logical jump. The sign of (PC-MLI) is dependent on the depth of modulation in each cell type as well and does not necessarily indicate relative timing. In any case, such caveats have not been ruled out in their analysis. This analysis to establish timing is unconvincing. Would it not be better to look at the timing of the spike modulation start directly rather than the round-about method they are using?

    3. Many figure panels make the same point and appear redundant. For example, that PCs and PC-MLIs are inversely correlated with each other in vivo during CRs is shown in Figure 7, figure 8a, S2, S4, and S5. Of course, in each case, the data are sorted differently (according to ISI, CR initiation, cumulative distributions, etc.,) but surely, the point regarding inverse relationship can be conveyed more concisely?

    4. Several details are missing in the methods section even though parts of it may have been published before. For instance, how are CRs calculated in the simulation? Methods state that 'The averaged and smoothed activity of the eight deep nucleus neurons was used to represent the output of the simulation and the predicted "eyelid response" of the simulation'. It is not clear what the nature of this transform is and if any calibration factors were used. How comparable are the simulated CRs in kinetics and amplitude to experimental CRs?

  6. Version published to 10.1101/2022.03.03.482855v1 on bioRxiv