Spinal premotor interneurons controlling antagonistic muscles are spatially intermingled

  1. Remi Ronzano
  2. Sophie Skarlatou
  3. Bianca K Barriga
  4. B Anne Bannatyne
  5. Gardave Singh Bhumbra
  6. Joshua D Foster
  7. Jeffrey D Moore
  8. Camille Lancelin
  9. Amanda M Pocratsky
  10. Mustafa Görkem Özyurt
  11. Calvin Chad Smith
  12. Andrew J Todd
  13. David J Maxwell
  14. Andrew J Murray
  15. Samuel L Pfaff
  16. Robert M Brownstone
  17. Niccolò Zampieri
  18. Marco Beato

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    The experiments presented in this extensive study by Ronzano et al. are a tour-de-force investigating the spatial organization of premotor interneurons in the mouse spinal cord to re-examine the fundamental question of whether there is spatial segregation of interneurons with monosynaptic connections to motoneurons innervating functionally antagonistic (flexor and extensor) pairs of limb muscles. The authors' premotor circuit mapping experiments, involving four different collaborating laboratories applying an extensive set of complementary rabies virus-based trans-synaptic circuit tracing techniques, convincingly demonstrate complete spatial overlap among flexor and extensor premotor interneurons, contradicting previous mapping results that suggest spatial segregation. The present results revise our understanding of the spatial organization of spinal premotor circuits with fundamental implications for understanding spinal motor circuit function.

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

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Elaborate behaviours are produced by tightly controlled flexor-extensor motor neuron activation patterns. Motor neurons are regulated by a network of interneurons within the spinal cord, but the computational processes involved in motor control are not fully understood. The neuroanatomical arrangement of motor and premotor neurons into topographic patterns related to their controlled muscles is thought to facilitate how information is processed by spinal circuits. Rabies retrograde monosynaptic tracing has been used to label premotor interneurons innervating specific motor neuron pools, with previous studies reporting topographic mediolateral positional biases in flexor and extensor premotor interneurons. To more precisely define how premotor interneurons contacting specific motor pools are organized, we used multiple complementary viral-tracing approaches in mice to minimize systematic biases associated with each method. Contrary to expectations, we found that premotor interneurons contacting motor pools controlling flexion and extension of the ankle are highly intermingled rather than segregated into specific domains like motor neurons. Thus, premotor spinal neurons controlling different muscles process motor instructions in the absence of clear spatial patterns among the flexor-extensor circuit components.

Article activity feed

  1. Version published to 10.7554/elife.81976 on eLife
  2. Version published to 10.1101/2021.02.10.430608v5 on bioRxiv
  3. Version published to 10.1101/2021.02.10.430608v4 on bioRxiv
  4. eLife

    Author Response

    Reviewer #2 (Public Review):

    SIGNIFICANCE: Movement is based on the coordinated activation and deactivation of muscle groups that depend on the timing and strength of firing of the motoneurons connected to them. Motoneuron recruitment ultimately depends on the activity of local interneurons. By difference to other CNS regions, the interneurons in the spinal cord controlling motor output display a very high diversity in genetics, anatomy, localization, and electrophysiological properties. Making sense of the interneuronal circuits that modulate motor output to the different muscles of the body has revealed to be quite complex. One technique proposed over 10 years ago is the use of retrograde transsynaptic-monosynaptic tracing with modified rabies virus injected in single muscles to define premotor connections to individual motor pools controlling single muscles. Using this technique, the original authors suggested that interneurons controlling flexors and extensors occupied different locations in the spinal cord. This idea was an extension of pioneering work from the Jessell lab at Columbia University demonstrating that positional identity determined input connectivity of motoneurons, at least from Ia afferents. This principle, if extended to premotor spinal interneurons would simplify mechanisms by which extensor and flexor interneuron networks could be connected and controlled. In this paper, the authors combine data from four independent groups to show this principle might not be correct. In other words, interneurons connected to individual motor pools are highly intermingled in the spinal cord. This raises the bar for understanding both the intrinsic organization principles of interneuron microcircuits in the spinal cord (if any) and how they develop their specific connectivity.

    STRENGTHS AND WEAKNESSES: The authors propose that the conflicting conclusions occur because technical differences. The technique is based on complementation of rabies virus glycoprotein (G) in specific targeted motoneurons infected with a glycoprotein deficient rabies virus (RVdG). The way G and RVdG are delivered to specific motoneurons controlling one muscle differ. Originally this was accomplished by co-injecting RVdG and an AAV-G vectors in the same muscle simultaneously. However as previously published by a different group and now confirmed by the authors, this approach also infects muscle sensory afferents capable of transynaptically labeling populations of interneurons in the spinal cord anterogradely. This results in the labeling of mixed interneuron populations through their output to specific motor pools and/or their input from primary afferents of the same muscle. To avoid this problem the authors used transgenic approaches to induce expression of G in all motoneurons (not sensory neurons) and obtain muscle specifity by injecting RVdG in single muscles. Unfortunately, there is no single gene that selects only motoneurons for transgenic expression and tools for intersectional approaches were not available. Therefore, G is unavoidably expressed in some interneurons, in addition to motoneurons. These interneurons could be an additional source of transsynaptic jumps if they receive the RVdG from the motoneurons, raising the possibility that some labeling is the result of disynaptic, not monosynaptic, connections. The authors try to control for this possibility by comparing two different cre lines to direct G expression in motoneurons and each with different types of additional interneurons targeted. The results in both lines are similar raising confidence in the main conclusions. Moreover, the authors indicate that some motoneurons outside the intended pools are also labeled because motoneuron-to-motoneuron connections. In other words, the starter neurons for tracing monosynaptic connections are not as homogeneous or specific to a single motor pool as desired. This is acknowledged as a current limitation and is addressed in the discussion by proposing possible alternative approaches. Despite this weakness, the main conclusion of the study remains strong.

    A second technical issue raised by the authors is that of possible leakage during injection in the muscle. To reduce this possibility the authors reduced the volume injected compared to previous studies from 5 to 1 microliter and checked post-hoc the injection site for possible leakage (these are neonatal pups with muscles volumes of 2-3 microliters or less). In addition, they make a nice comparison injecting different titers of RVdG to demonstrate that variations in the number of infected motoneurons of one or two orders of magnitude does not alter the main conclusion on the topographic positioning of the interneurons connected to different motor pools. One weakness is that the exact numbers of motoneurons that start the tracing is impossible to evaluate and this prevents accurate comparisons across experiments. This is because cell death induced by the rabies virus is to be expected and only a variable subset of surviving neurons can be identified. Currently, this is an unavoidable characteristic of the technique. Nevertheless, there is a nice correlation between titer, surviving motoneuron numbers and interneurons labeled in number and location. The large number of replicates and their consistency further raises confidence in the authors claim of high specificity and replicability during injection, despite variable numbers of recovered motoneurons. The authors conclude that it is very important to check for the number and localization of starter motoneurons to confirm specificity after the injections. This reviewer totally agrees and is surprised this was not done in the experiment in which they try to replicate previous experiments by co-injecting in muscle AAV-G and RVdG.

    We agree with the reviewer that ideally the starter cells should have been identified in all the experiments. However, data were collected independently, at very different times in each of the labs involved, with different initial aims and there was no prior agreement on the details of injection and post-processing. The realization that we had similar experiments, performed with different techniques, led us to pool our observation together in order to give a picture of the distribution of premotor interneurons, the leitmotif of this paper, and a great effort has been devoted to ensure that all the cell counting procedures were uniform across labs. The lack of initial coordination is the reason why in some datasets the starter cells have not been quantified. Furthermore, in the previous version we had wrongly indicated that motor neurons analysed at Glasgow University were identified by ChAT expression. We have corrected this in the current version, since for those experiments motor neurons were only identified by location within any of the motor nuclei and size (diameter greater than 30 µm). On the other hand, since we have started comparing results, we have agreed on a uniform way of analysing and representing the data using the same normalization criteria. Therefore, while we cannot compare quantities like the ratio of secondary and primary infected cells for all the experiments (but we do it for the subset in which this is possible, see new Figure 4-Figure supplement 3 and comment number 3 below), the positional analysis has been done following the exact same criteria.

    One final problem with interpretation is that, for yet unknown reasons, the technique is dependent on the age of the animal and cannot be implemented in mature animals. Therefore, the connectivity revealed here is the one present during the first few days of life in the mouse. This is a period of significant synaptogenesis and synaptic selection and de-selection. The authors are encouraged to discuss further this limitation when interpreting interneuron connectivity in adult from studies in neonates.

    A very important point, see detailed answer to comment number 10 below.

    In summary, the authors have introduced new technical variations to trace premotor interneurons and challenge a major idea in the field, that is that interneuron connected to flexors and extensors occupy different positions in the spinal cord. The technique has still some weaknesses. 1) possibility of disynaptic jumps, 2) accurate identification of starter neurons, 3) restriction to neonates. However, the authors strengthen their conclusions considering alternatives and introducing a large number of controls (two cre lines, different titers, large number of animals analyzed, large numbers and consistency of replicates, independent counting in different labs... etc). This is an important and very useful study that suggests topographic localization is not a major organizing principle for interneuron connections with motor pools. It remains to be investigated then what are the organizational mechanisms that couple interneurons to functional distinct motor pools.

    The weaknesses summarized in the paragraph above are addressed in detail below in the answers to the specific comments.

    Reviewer #3 (Public Review):

    The manuscript by Ronzano et al presents a rigorous neuroanatomical study to convincingly demonstrate that there is no difference in the medio-lateral organization of flexor and extensor premotor interneurons. The study uses monosynaptic restricted transsynaptic tracing from ankle flexor and extensor muscles with several (4) strategies for delivery of the G protein complement and delta G Rabies virus, and additional (2) variations that consider titer and cre line. The authors went to great lengths here in attempt to replicate prior studies for which they had initial conflicting findings. Further, the experiments are performed in laboratories in four different locations. The variations on the Rabies and complement delivery, regardless of lab performing the experiment and analysis, all converge on the same conclusion. Aside from the primary conclusion, the paper can be used as a manual for anyone considering transsynaptic tracing as it details the benefits and caveats of each strategy with examples.

    The initial conflicting results put the onus on the authors to demonstrate where the divergence occurred. The authors took a highly comprehensive approach, which is a clear strength of the paper. All of the data is fully and transparently presented. Standardizations and differences between experiments run or analyzed in each lab are well laid out. Figure 1 and Table 2 provide a great summary of the techniques and their limitations. These are also well thought out and discussed within each section of results.

    The only thing missing is a likely explanation for the differences seen. Although the authors made several attempts to provide such explanation, the question remains - how did two groups who published independent studies using different strategies demonstrate flexor and extensor separation in the dorsal horn, when this study, using several strategies in multiple labs, show that the premotor neurons are in complete overlap? Additional small differences in methodologies could be identified which are not discussed and may provide potential explanations, but only for discrepancies in results of single techniques, not for all of the strategies used. The lack of reason for the discrepancy with prior studies despite the extensive efforts is unsatisfying, but, most importantly, the experiments were rigorously performed and the data support the conclusions presented.

    We thank the reviewer for the positive comments and we share the opinion that the discrepancy is unsatisfying. While we propose possible explanations, despite the extensive efforts, we could not provide a definite answer, but we hope that making our work public and all the data available, will trigger even more efforts from the rest of the community.

  5. eLife

    Evaluation Summary:

    The experiments presented in this extensive study by Ronzano et al. are a tour-de-force investigating the spatial organization of premotor interneurons in the mouse spinal cord to re-examine the fundamental question of whether there is spatial segregation of interneurons with monosynaptic connections to motoneurons innervating functionally antagonistic (flexor and extensor) pairs of limb muscles. The authors' premotor circuit mapping experiments, involving four different collaborating laboratories applying an extensive set of complementary rabies virus-based trans-synaptic circuit tracing techniques, convincingly demonstrate complete spatial overlap among flexor and extensor premotor interneurons, contradicting previous mapping results that suggest spatial segregation. The present results revise our understanding of the spatial organization of spinal premotor circuits with fundamental implications for understanding spinal motor circuit function.

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

  6. eLife

    Reviewer #1 (Public Review):

    The experiments presented in this extensive study by Ronzano et al. are a tour-de-force investigating the spatial organization of premotor interneurons in the mouse spinal cord to re-examine the fundamental question of whether there is spatial segregation of interneurons with monosynaptic connections to motoneurons innervating functionally antagonistic (flexor and extensor) pairs of limb muscles. Such segregation has been proposed from earlier studies utilizing strategies for retrograde trans-synaptic tracing of spinal premotoneurons with rabies virus (RabV) following muscle injection. This spatial organization has been suggested to provide an anatomical substrate for labeled line inputs from proprioceptive afferents to motor neurons with possibly organization advantages for motor control. The present premotor circuit mapping experiments, involving four different collaborating laboratories applying an extensive set of complementary RabV-based trans-synaptic circuit tracing techniques, convincingly demonstrate complete spatial overlap among flexor and extensor premotor interneurons, contradicting the previous mapping results that suggest spatial segregation. The present results revise our understanding of the spatial organization of spinal premotor circuits and provide an alternative view of the role of interneuron positioning in sensory input connectivity without specific spatial patterning of output connectivity to motoneurons, with fundamental implications for understanding motor circuit function.

    Strengths of these studies include:

    1. The investigators systematically tested and directly compared most of the available premotor circuit tracing strategies utilizing genetically modified mouse strains and viruses, as well as the previous approaches, with all tests replicating the spatial overlap of flexor and extensor premotor interneurons.

    2. The authors utilized a mouse genetic strategy combining a Cre conditional allele expressing RabV glycoprotein G from the rosa locus with either the ChAT::Cre or Olig2::Cre mouse lines, which in contrast to previous RabV-based approaches, enables selective and potentially high levels of G expression in all motoneurons at the time of RabV muscle injection and likely robust transsynaptic transfer for premotor neuron labeling.

    3. The authors present a very useful instructive exposition of the currently available techniques for labeling premotor interneurons outlining experimental strategies and indicating advantages and disadvantages for interpretation of results by illustrating RabV trans-synaptic transfer pathways that could confound experimental results.

    4. The authors also used transgenic strategies in combination with their other approaches to differentiate inhibitory or putative excitatory premotor interneurons controlling the activity of flexor and extensor muscles and demonstrated from technically elegant spatial analyses that flexor and extensor premotor neurons were always spatially intermingled regardless of their neurotransmitter identity.

    5. The authors further confirmed the lack of spatial segregation by pooling together all the results obtained with the different circuit tracing methods.

    6. The authors thoroughly discuss the limitations of their mouse genetic strategy for circuit tracing including off-target G complementation in cells other than the targeted cholinergic motoneurons with the possibility of labeling disynaptic pathways via cholinergic spinal interneurons. Also considered is the problem in identifying the number of motoneurons with G complementation, which is a main determinant of reproducibility in RabV tracing experiments and a key parameter for comparing results from different circuit tracing approaches.

    7. Overall the experiments are rigorously performed with a design that reduces biases associated with the various RabV-based circuit tracing methods, and the data although very extensive with numerous data source files and supplemental illustrations, are clearly presented.

  7. eLife

    Reviewer #2 (Public Review):

    SIGNIFICANCE: Movement is based on the coordinated activation and deactivation of muscle groups that depend on the timing and strength of firing of the motoneurons connected to them. Motoneuron recruitment ultimately depends on the activity of local interneurons. By difference to other CNS regions, the interneurons in the spinal cord controlling motor output display a very high diversity in genetics, anatomy, localization, and electrophysiological properties. Making sense of the interneuronal circuits that modulate motor output to the different muscles of the body has revealed to be quite complex. One technique proposed over 10 years ago is the use of retrograde transsynaptic-monosynaptic tracing with modified rabies virus injected in single muscles to define premotor connections to individual motor pools controlling single muscles. Using this technique, the original authors suggested that interneurons controlling flexors and extensors occupied different locations in the spinal cord. This idea was an extension of pioneering work from the Jessell lab at Columbia University demonstrating that positional identity determined input connectivity of motoneurons, at least from Ia afferents. This principle, if extended to premotor spinal interneurons would simplify mechanisms by which extensor and flexor interneuron networks could be connected and controlled. In this paper, the authors combine data from four independent groups to show this principle might not be correct. In other words, interneurons connected to individual motor pools are highly intermingled in the spinal cord. This raises the bar for understanding both the intrinsic organization principles of interneuron microcircuits in the spinal cord (if any) and how they develop their specific connectivity.

    STRENGTHS AND WEAKNESSES: The authors propose that the conflicting conclusions occur because technical differences. The technique is based on complementation of rabies virus glycoprotein (G) in specific targeted motoneurons infected with a glycoprotein deficient rabies virus (RVdG). The way G and RVdG are delivered to specific motoneurons controlling one muscle differ. Originally this was accomplished by co-injecting RVdG and an AAV-G vectors in the same muscle simultaneously. However as previously published by a different group and now confirmed by the authors, this approach also infects muscle sensory afferents capable of transynaptically labeling populations of interneurons in the spinal cord anterogradely. This results in the labeling of mixed interneuron populations through their output to specific motor pools and/or their input from primary afferents of the same muscle. To avoid this problem the authors used transgenic approaches to induce expression of G in all motoneurons (not sensory neurons) and obtain muscle specifity by injecting RVdG in single muscles. Unfortunately, there is no single gene that selects only motoneurons for transgenic expression and tools for intersectional approaches were not available. Therefore, G is unavoidably expressed in some interneurons, in addition to motoneurons. These interneurons could be an additional source of transsynaptic jumps if they receive the RVdG from the motoneurons, raising the possibility that some labeling is the result of disynaptic, not monosynaptic, connections. The authors try to control for this possibility by comparing two different cre lines to direct G expression in motoneurons and each with different types of additional interneurons targeted. The results in both lines are similar raising confidence in the main conclusions. Moreover, the authors indicate that some motoneurons outside the intended pools are also labeled because motoneuron-to-motoneuron connections. In other words, the starter neurons for tracing monosynaptic connections are not as homogeneous or specific to a single motor pool as desired. This is acknowledged as a current limitation and is addressed in the discussion by proposing possible alternative approaches. Despite this weakness, the main conclusion of the study remains strong.

    A second technical issue raised by the authors is that of possible leakage during injection in the muscle. To reduce this possibility the authors reduced the volume injected compared to previous studies from 5 to 1 microliter and checked post-hoc the injection site for possible leakage (these are neonatal pups with muscles volumes of 2-3 microliters or less). In addition, they make a nice comparison injecting different titers of RVdG to demonstrate that variations in the number of infected motoneurons of one or two orders of magnitude does not alter the main conclusion on the topographic positioning of the interneurons connected to different motor pools. One weakness is that the exact numbers of motoneurons that start the tracing is impossible to evaluate and this prevents accurate comparisons across experiments. This is because cell death induced by the rabies virus is to be expected and only a variable subset of surviving neurons can be identified. Currently, this is an unavoidable characteristic of the technique. Nevertheless, there is a nice correlation between titer, surviving motoneuron numbers and interneurons labeled in number and location. The large number of replicates and their consistency further raises confidence in the authors claim of high specificity and replicability during injection, despite variable numbers of recovered motoneurons. The authors conclude that it is very important to check for the number and localization of starter motoneurons to confirm specificity after the injections. This reviewer totally agrees and is surprised this was not done in the experiment in which they try to replicate previous experiments by co-injecting in muscle AAV-G and RVdG.

    One final problem with interpretation is that, for yet unknown reasons, the technique is dependent on the age of the animal and cannot be implemented in mature animals. Therefore, the connectivity revealed here is the one present during the first few days of life in the mouse. This is a period of significant synaptogenesis and synaptic selection and de-selection. The authors are encouraged to discuss further this limitation when interpreting interneuron connectivity in adult from studies in neonates.

    In summary, the authors have introduced new technical variations to trace premotor interneurons and challenge a major idea in the field, that is that interneuron connected to flexors and extensors occupy different positions in the spinal cord. The technique has still some weaknesses. 1) possibility of disynaptic jumps, 2) accurate identification of starter neurons, 3) restriction to neonates. However, the authors strengthen their conclusions considering alternatives and introducing a large number of controls (two cre lines, different titers, large number of animals analyzed, large numbers and consistency of replicates, independent counting in different labs... etc). This is an important and very useful study that suggests topographic localization is not a major organizing principle for interneuron connections with motor pools. It remains to be investigated then what are the organizational mechanisms that couple interneurons to functional distinct motor pools.

  8. eLife

    Reviewer #3 (Public Review):

    The manuscript by Ronzano et al presents a rigorous neuroanatomical study to convincingly demonstrate that there is no difference in the medio-lateral organization of flexor and extensor premotor interneurons. The study uses monosynaptic restricted transsynaptic tracing from ankle flexor and extensor muscles with several (4) strategies for delivery of the G protein complement and delta G Rabies virus, and additional (2) variations that consider titer and cre line. The authors went to great lengths here in attempt to replicate prior studies for which they had initial conflicting findings. Further, the experiments are performed in laboratories in four different locations. The variations on the Rabies and complement delivery, regardless of lab performing the experiment and analysis, all converge on the same conclusion. Aside from the primary conclusion, the paper can be used as a manual for anyone considering transsynaptic tracing as it details the benefits and caveats of each strategy with examples.

    The initial conflicting results put the onus on the authors to demonstrate where the divergence occurred. The authors took a highly comprehensive approach, which is a clear strength of the paper. All of the data is fully and transparently presented. Standardizations and differences between experiments run or analyzed in each lab are well laid out. Figure 1 and Table 2 provide a great summary of the techniques and their limitations. These are also well thought out and discussed within each section of results.

    The only thing missing is a likely explanation for the differences seen. Although the authors made several attempts to provide such explanation, the question remains - how did two groups who published independent studies using different strategies demonstrate flexor and extensor separation in the dorsal horn, when this study, using several strategies in multiple labs, show that the premotor neurons are in complete overlap? Additional small differences in methodologies could be identified which are not discussed and may provide potential explanations, but only for discrepancies in results of single techniques, not for all of the strategies used. The lack of reason for the discrepancy with prior studies despite the extensive efforts is unsatisfying, but, most importantly, the experiments were rigorously performed and the data support the conclusions presented.

  9. Version published to 10.1101/2021.02.10.430608v3 on bioRxiv
  10. Version published to 10.1101/2021.02.10.430608v2 on bioRxiv
  11. Version published to 10.1101/2021.02.10.430608v1 on bioRxiv