Distinct representations of body and head motion are dynamically encoded by Purkinje cell populations in the macaque cerebellum

  1. Omid A Zobeiri
  2. Kathleen E Cullen

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    Zobeiri and Cullen address the important question of how the cerebellum transforms multiple streams of sensory information into an estimate of the motion of the body in the world. They find that Purkinje cells, the inhibitory principal neurons of the cerebellar cortex, have multimodal and highly diverse responses to vestibular and neck proprioceptive inputs. Notably, this information is combined in a way that is different than what is seen in downstream fastigial neurons, which reflect either head or body motion, but not both. The experiments are well executed, generating data that provide important and novel insights, but there are shortcomings in the model put forward to account for these results.

    (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 ability to accurately control our posture and perceive our spatial orientation during self-motion requires knowledge of the motion of both the head and body. However, while the vestibular sensors and nuclei directly encode head motion, no sensors directly encode body motion. Instead, the integration of vestibular and neck proprioceptive inputs is necessary to transform vestibular information into the body-centric reference frame required for postural control. The anterior vermis of the cerebellum is thought to play a key role in this transformation, yet how its Purkinje cells transform multiple streams of sensory information into an estimate of body motion remains unknown. Here, we recorded the activity of individual anterior vermis Purkinje cells in alert monkeys during passively applied whole-body, body-under-head, and head-on-body rotations. Most Purkinje cells dynamically encoded an intermediate representation of self-motion between head and body motion. Notably, Purkinje cells responded to both vestibular and neck proprioceptive stimulation with considerable heterogeneity in their response dynamics. Furthermore, their vestibular responses were tuned to head-on-body position. In contrast, targeted neurons in the deep cerebellar nuclei are known to unambiguously encode either head or body motion across conditions. Using a simple population model, we established that combining responses of~40-50 Purkinje cells could explain the responses of these deep cerebellar nuclei neurons across all self-motion conditions. We propose that the observed heterogeneity in Purkinje cell response dynamics underlies the cerebellum’s capacity to compute the dynamic representation of body motion required to ensure accurate postural control and perceptual stability in our daily lives.

Article activity feed

  1. Version published to 10.7554/elife.75018 on eLife
  2. eLife

    Author Response:

    Reviewer #2 (Public Review):

    The manuscript addresses an important question regarding sensory processing related to self-motion. The main experiment is clearly described and demonstrates that neurons display a diversity of responses from purely reflecting vestibular input (head-in-space motion) to predominantly body motion, and any combination between. Of particular interest, is that the response of the Purkinje cells are profoundly different than its downstream target, the fastigial neurons which signal only head-in-space or body motion. This substantive difference in neural representations between these two connected brain regions is surprising.

    The manuscript also provides a simple population model to show that fastigial responses could be generated from Purkinje cell activity, but only from combining at least 40 neurons. While the model provides some insight on the potential interaction between Purkinje cells and fastigial neurons, I think the model assumes no other input to the fastigial neurons. However, I would assume that there is likely a strong input from mossy fibers onto the fastigial neurons that also target the Purkinje cells. This mossy fiber input will certainly provide vestibular and neck proprioceptive input to the fastigial nucleus. Thus, the Purkinje cell input may be essential for countering the mossy fiber input leading to separate representations for head and body motion in the fastigial nucleus.

    We agree this is an important point. To address the reviewer’s concern, we performed additional modeling in order to consider the influence of mossy fiber inputs. Specifically, following the reviewer’s suggestion below, mossy fiber input was modeled using random patterns of vestibular and neck proprioceptive input. Prior studies have shown that the dynamics of vestibular nuclei neuron responses strongly resemble those of unimodal fastigial neurons in rhesus monkeys (i.e., they encode vestibular input and are insensitive to neck proprioceptive inputs, Roy & Cullen, 2001). In contrast, reticular formation neurons responses to such yaw head and/or neck rotations have not yet been described. We therefore simulated mossy fiber input first as a summation of vestibular and neck proprioceptive inputs, for which the gains and phases were randomly drawn from a distribution, comparable to that previously reported (Mitchell er al. 2017) in the vestibular nuclei (Fig. 7-figure supplement 3). We then further explored the effect of systematically altering this simulated mossy fiber input - relative to the reference distribution of mossy fiber inputs - by i) doubling the gain, ii) reducing the gain by half, iii) doubling the phase, and iv) reducing the phase by half (Fig. 7-figure supplement 4). Overall, we found that the addition of such simulated mossy fiber did not dramatically alter our estimate of the population Purkinje cell population size required to generate rFN neurons responses (~50 versus 40; Fig. 7-figure supplement 3&4).

    Another issue is the limited number of neurons recorded in the secondary experiment with only 12 bimodal neurons and 5 unimodal (although there appears to be only 4 neurons in Figure 5C). Such a small sample impacts the estimated tuning properties of Purkinje neurons in Figure 5D and the results from the population model. This needs to be clearly recognized.

    We have revised the RESULTS to clarify the numbers of Purkinje cells that were tested (13 bimodal and 4 unimodal Purkinje cells). For comparison, in our Brooks and Cullen study, tuning curves were computed for 10 bimodal and 12 unimodal rFN. We note that i) unimodal Purkinje cells make up a relatively small percentage of anterior vermis Purkinje cells and ii) similar to unimodal rFN, our small sample of unimodal 9 Purkinje cells did not demonstrate significant tuning. In contrast, all bimodal Purkinje cells in our sample demonstrated significant tuning. To simulate responses for the bimodal Purkinje cells that were not held long enough to test during gain-field paradigm (i.e., Fig 5), we generated tuning curves drawn from a normal distribution estimated from 13 bimodal Purkinje cells. We appreciate this was not clear in the original submission and have revised the METHODS section to clarify our approach. Overall, while we recognize that our sample size is small, we nevertheless found it interesting that including this our results from this protocol did not increase the estimated population size relative to that estimated using our other dynamic protocols.

    Reviewer #3 (Public Review):

    In this study, the authors characterize the simple spike discharges of Purkinje cells in the anterior vermis of the macaque during passive vestibular and neck proprioceptive stimulation. The activity of most Purkinje cells encoded both vestibular (whole-body rotation) and proprioceptive (body-under-head rotation) stimuli. Although the vestibular and proprioceptive responses were, on average, antagonistic in the preferred direction, consistent with a partial transformation from head to body coordinates, response properties for both modalities were highly variable across neurons. Most cells responded under combined vestibular and proprioceptive stimulation (head-on-body rotation), and these responses were well-approximated by the average of the responses to each modality individually. Vestibular responses exhibited gain-field-like tuning with changes in head-on-body position, though these changes were significantly smaller than the shifts observed for neurons downstream in the rostral fastigial nucleus. Finally, a weighted average of the responses of approximately 40 Purkinje cells provided a good fit to the responses of postsynaptic fastigial neurons.

    Overall, these results provide important and novel insights into the implementation of coordinate transformations by cerebellar circuitry. The experiments are well-designed, the data high quality, the analyses reasonable, and the conclusions justified by the data. The manuscript is clear and well-written, and will be of interest to a broad neuroscientific audience. I have no major concerns. I have a few minor suggestions for improving this manuscript, described below.

    1 - The authors may wish to discuss earlier work in the decerebrate cat by Denoth et al. (1979, Pflügers Archiv), which provided evidence that the responses of Purkinje cells in the anterior vermis to head-on-body tilt is relatively well-approximated by averaging the responses to neck and macular stimulation alone.

    We thank the reviewer for bringing this reference to our attention and have revised the INTRODUCTION and DISCUSSION to include the early work of Denoth et al.,1979.

    2 - To better convey the heterogeneity of responses across the sample of Purkinje cells, two additional supplemental figure panels might be useful: (1) the vestibular, proprioceptive, summed, and combined sensitivities in each direction (as in the Fig. 3C insets) for each individual neuron (perhaps as a series of subpanels), and (2) scatterplots of response phase for proprioceptive vs vestibular stimulation for bimodal neurons (with separate panels for preferred and non-preferred directions).

    We agree that this is a useful way to emphasize the heterogeneity of bimodal Purkinje cells responses and have added the requested response phase scatterplots for proprioceptive vs vestibular stimulation (Fig 2 - figure supplement 2C&D). We have also made a figure showing the summation model for each individual neuron. However, because our Purkinje cell population included 73 neurons, this figure includes a corresponding 73X2 =146 polar plots (i.e., two plot each cell, one for ipsi and contralateral motion). Given the immense size of this figure, we elected not to include this figure in the supplementary material in the revised manuscript.

    3 - Can the authors provide additional information on the approximate location of the recorded neurons (lobule and zone or mediolateral position)? Is it possible that some project to the vestibular nuclei, rather than the rFN? This consideration seems especially relevant for the interpretation of the pooling analysis in Fig. 6, which seems to assume that Purkinje cells are sampled from a sagittal zone with overlapping projections in the rFN (or, at least, that the response properties of the sampled neurons are representative of the properties in a corticonuclear zone). Some additional discussion on this point would be helpful.

    The recorded neurons were located in the lobules II-V of the anterior vermis, ~0 to 2 mm from the midline. We now include this information in the revised METHODS. As noted by the reviewer, Purkinje cells in this region of the anterior vermis project to the vestibular nuclei as well as to the rFN (Voogd et al. 1991). Nevertheless, using comparable stimulation protocols, we have previously shown that the responses of vestibular nuclei neurons are comparable to those of unimodal rFN neurons (Brooks et al., 2015). Specifically, both vestibular nuclei and unimodal rFN neurons are insensitive to proprioceptive stimulation and demonstrated comparable responses to vestibular stimulation. Thus, our present modeling results regarding the population convergence required to account for unimodal rFN neurons can be directly applied to vestibular nuclei neurons. We have revised the DISCUSSION to consider this point.

    4 - When weighted averages of Purkinje cell responses are used to model rFN responses, my intuition would be that w_i is near zero for v-shaped and rectifying Purkinje cells. That is, the model would mostly ignore them, as data from both directions appear to be included. Is this the case? A more detailed description of the fitting procedure would also be helpful.

    To address the reviewers’ concerns regarding the Purkinje cell weights, we have added a new inset to Fig 7C. As can be seen, model weights are well distributed across different Purkinje cells. Further, to confirm that the distribution of the weights of Purkinje cells inputs are distributed over different classes of PCs we now illustrate the weight distributions for (a) linear vs. v-shaped vs. rectifying Purkinje cells, (b) bimodal vs. unimodal Purkinje cells, (c) Type I vs. Type II Purkinje cells and (d) Purkinje cells with agonistic vs. antagonistic vestibular and proprioceptive sensitivities. These results are shown in Figure 7-supplemental figures 1&2. Overall, we found that distribution of the weights was not biased towards linear cells, but rather were similarly distributed across all three groups. This was true for our modeling of both bimodal and unimodal rFN cells (compare Fig 7- figure supplement 1 vs. Fig 7- figure supplement 2). As can be seen in this Figure, we likewise found comparable results for the weights of Type I vs. Type II Purkinje cells, unimodal vs. bimodal Purkinje cells, and/or vestibular / proprioceptive agonist vs. antagonist bimodal neurons. Finally, as detailed above in our response to the reviewers’ consensus comments, we have also revised the METHODS section to provide a more detailed description of linear regression method.

    5 - Another potential interpretive issue in the averaging analysis concerns the presence of noise on single trials. The authors could briefly comment on whether more Purkinje cells might be needed to predict rFN responses on a single trial in real time.

    This is an interesting question; we have revised the DISCUSSION to consider this point.

  3. eLife

    Evaluation Summary:

    Zobeiri and Cullen address the important question of how the cerebellum transforms multiple streams of sensory information into an estimate of the motion of the body in the world. They find that Purkinje cells, the inhibitory principal neurons of the cerebellar cortex, have multimodal and highly diverse responses to vestibular and neck proprioceptive inputs. Notably, this information is combined in a way that is different than what is seen in downstream fastigial neurons, which reflect either head or body motion, but not both. The experiments are well executed, generating data that provide important and novel insights, but there are shortcomings in the model put forward to account for these results.

    (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 #1 (Public Review):

    In this manuscript, Zobeiri and colleagues investigate the activity of cerebellar Purkinje cells relative to various types of head and neck movements. They find that many Purkinje cells encode both vestibular and neck proprioceptive information. Notably, vestibular and proprioceptive information tend to be in the same direction, whereas in previous recordings this group found that in the fastigial cerebellar nucleus, the direct target of this region, vestibular and proprioceptive information tend to cancel each other out. The authors put forward a model that suggests this difference is explained by convergence of different Purkinje cells onto fastigial neurons.

    The authors are obviously experts in this field, and the analyses are performed at a highly sophisticated level. However, I found the manuscript quite difficult to read, and struggled at times to understand the significance of the results. What seems to be the most significant finding, that Purkinje cell activity can explain the encoding of neurons in the fastigial nucleus, is not well developed given how counterintuitive the result appears.

  5. eLife

    Reviewer #2 (Public Review):

    The manuscript addresses an important question regarding sensory processing related to self-motion. The main experiment is clearly described and demonstrates that neurons display a diversity of responses from purely reflecting vestibular input (head-in-space motion) to predominantly body motion, and any combination between. Of particular interest, is that the response of the Purkinje cells are profoundly different than its downstream target, the fastigial neurons which signal only head-in-space or body motion. This substantive difference in neural representations between these two connected brain regions is surprising.

    The manuscript also provides a simple population model to show that fastigial responses could be generated from Purkinje cell activity, but only from combining at least 40 neurons. While the model provides some insight on the potential interaction between Purkinje cells and fastigial neurons, I think the model assumes no other input to the fastigial neurons. However, I would assume that there is likely a strong input from mossy fibers onto the fastigial neurons that also target the Purkinje cells. This mossy fiber input will certainly provide vestibular and neck proprioceptive input to the fastigial nucleus. Thus, the Purkinje cell input may be essential for countering the mossy fiber input leading to separate representations for head and body motion in the fastigial nucleus.

    Another issue is the limited number of neurons recorded in the secondary experiment with only 12 bimodal neurons and 5 unimodal (although there appears to be only 4 neurons in Figure 5C). Such a small sample impacts the estimated tuning properties of Purkinje neurons in Figure 5D and the results from the population model. This needs to be clearly recognized.

  6. eLife

    Reviewer #3 (Public Review):

    In this study, the authors characterize the simple spike discharges of Purkinje cells in the anterior vermis of the macaque during passive vestibular and neck proprioceptive stimulation. The activity of most Purkinje cells encoded both vestibular (whole-body rotation) and proprioceptive (body-under-head rotation) stimuli. Although the vestibular and proprioceptive responses were, on average, antagonistic in the preferred direction, consistent with a partial transformation from head to body coordinates, response properties for both modalities were highly variable across neurons. Most cells responded under combined vestibular and proprioceptive stimulation (head-on-body rotation), and these responses were well-approximated by the average of the responses to each modality individually. Vestibular responses exhibited gain-field-like tuning with changes in head-on-body position, though these changes were significantly smaller than the shifts observed for neurons downstream in the rostral fastigial nucleus. Finally, a weighted average of the responses of approximately 40 Purkinje cells provided a good fit to the responses of postsynaptic fastigial neurons.

    Overall, these results provide important and novel insights into the implementation of coordinate transformations by cerebellar circuitry. The experiments are well-designed, the data high quality, the analyses reasonable, and the conclusions justified by the data. The manuscript is clear and well-written, and will be of interest to a broad neuroscientific audience. I have no major concerns. I have a few minor suggestions for improving this manuscript, described below.

    1 - The authors may wish to discuss earlier work in the decerebrate cat by Denoth et al. (1979, Pflügers Archiv), which provided evidence that the responses of Purkinje cells in the anterior vermis to head-on-body tilt is relatively well-approximated by averaging the responses to neck and macular stimulation alone.

    2 - To better convey the heterogeneity of responses across the sample of Purkinje cells, two additional supplemental figure panels might be useful: (1) the vestibular, proprioceptive, summed, and combined sensitivities in each direction (as in the Fig. 3C insets) for each individual neuron (perhaps as a series of subpanels), and (2) scatterplots of response phase for proprioceptive vs vestibular stimulation for bimodal neurons (with separate panels for preferred and non-preferred directions).

    3 - Can the authors provide additional information on the approximate location of the recorded neurons (lobule and zone or mediolateral position)? Is it possible that some project to the vestibular nuclei, rather than the rFN? This consideration seems especially relevant for the interpretation of the pooling analysis in Fig. 6, which seems to assume that Purkinje cells are sampled from a sagittal zone with overlapping projections in the rFN (or, at least, that the response properties of the sampled neurons are representative of the properties in a corticonuclear zone). Some additional discussion on this point would be helpful.

    4 - When weighted averages of Purkinje cell responses are used to model rFN responses, my intuition would be that w_i is near zero for v-shaped and rectifying Purkinje cells. That is, the model would mostly ignore them, as data from both directions appear to be included. Is this the case? A more detailed description of the fitting procedure would also be helpful.

    5 - Another potential interpretive issue in the averaging analysis concerns the presence of noise on single trials. The authors could briefly comment on whether more Purkinje cells might be needed to predict rFN responses on a single trial in real time.

  7. Version published to 10.1101/2021.10.25.465748v1 on bioRxiv