Multi-area activity in mouse motor cortex associated with one- and two-handed oromanual dexterity

Reviewer #1 (Public review):

Summary:

This study describes motor cortical activity patterns during food handling in mice, investigating whether the hand/s used is reflected in distinct neural activity. The experiments focus on forelimb M1 and M2 (fM1, fM2) and an oral-manual region LOM. The main findings are that fM1 and fM2 have largely similar relationships with forelimb control, and LOM neurons are more broadly tuned. These conclusions are reached using a variety of analyses spanning straightforward firing rate analyses, selectivity metrics, PCA, and GLM decoding methods to assess tuning generalizability. The study's significance is strengthened by including analyses of bimanual control, and in this sphere, there are descriptive data and analyses that aficionados of cortical control of dexterous behaviors will find useful. The use of unimanual control is useful as a point of comparison here, but less novel overall. There are a number of places where the descriptions of what is being analyzed, what is being concluded, and data reporting should be strengthened and clarified. Additionally, the study could be greatly improved by consolidating figures and the analyses shown, since many are redundant. Many of the analyses need clearer reporting of means and effect sizes in the text, rather than just statistical outcomes. Overall, at this juncture, the study presents analyses of a unique dataset that may seed future investigations of mechanisms of bimanual coordination.

Strengths:

There are relatively few studies that compare neural activity across bimanual and unimanual control. This study uses a naturalistic food handling task to explore neural relationships to forelimb kinematics under these conditions. The uniqueness of the task and analysis target is a strength of the study.

The authors remain fairly conservative and make few strong claims in the study, which may be warranted given the diversity of tuning profiles they observed.

Weaknesses:

There are a number of statistical tests that were accompanied by too little information to critically evaluate. Means and effect sizes needed to be better reported; some details of analyses were difficult to parse, making the strength of the conclusions difficult to evaluate.

Reviewer #2 (Public review):

Summary:

Barrett et al. examine how neural activity in the mouse motor cortex varies when a movement is performed with the ipsilateral or contralateral forelimb. First, they train animals to grasp and manipulate a pellet of food with either the left forepaw, the right forepaw, or both. Next, they measure activity in the primary and secondary forelimb motor areas (fl-M1 and fl-M2) and in the classical tongue-jaw area (tj-M1 / LOM). While responses in the forelimb areas are diverse, with some neurons preferring ipsilateral or bilateral movements, a plurality of cells prefer the contralateral limb. In LOM, by contrast, little limb selectivity is observed. At the neural population level, structure is preserved across conditions in LOM, but not in the forelimb areas. Finally, paw position can be decoded from activity in all three areas, and the LOM decoder generalized across limbs.

Strengths:

While previous studies in macaques have compared motor cortical activity during movement (and perturbation) of the contralateral and ipsilateral arms, no analogous work has been undertaken in rodents. This paper closes this knowledge gap by showing, for the first time, moderate-to-strong lateralization in the forelimb motor cortical areas of mice transporting grasped food pellets to the mouth, and a relative absence of lateralization in the classical tongue-jaw area. On the whole, I think this is a solid paper that reports novel observations of interest to the motor systems community.

Weaknesses:

The central question posed is whether cortical activity depends on the effector(s) used (ipsi forelimb, contra forelimb, or both). The corresponding hypotheses (Figure 1) are somewhat coarse-grained and are not mutually exclusive. One might expect to see condition-independent, limb-selective, and uni-/bimanual-selective signals in motor cortex (though their magnitudes could differ substantially), and to find these signals intermingled at the level of single neurons. The authors may wish to consider setting up a more focused question. For example, can bimanual responses be explained as a sum of the unimanual responses from the left and right limbs?

In the area usually identified as tongue-jaw motor cortex (here referred to as LOM), unit and population activity look quite similar for ipsilateral, contralateral, and bilateral forelimb reaches. The most parsimonious explanation is that the activity is related mostly to mouth and tongue movements, rather than limb movements. Systematic mapping studies with microstimulation in the rat (Neafsy et al., Brain Res. Rev. 1986) and optogenetic stimulation in the mouse (Mayrhofer et al., Neuron 2019) tend to support the idea that tjM1/LOM is specialized for control of the tongue and mouth. Thus, I'm not entirely convinced that it "encodes ingestion-related forelimb parameters necessary for oromanual coordination." The authors could say more about this issue: what specific limb-related parameters do they think are encoded, why would these parameters be effector-independent, what evidence for this encoding is presented here, and how can limb- and mouth-related components be distinguished? The problem could potentially be addressed experimentally, as well, by delivering food pellets directly to the mouth while preventing manipulation with the paws, but this experiment isn't strictly necessary.

Because the corticospinal tract is strongly lateralized, cortical activity presumably has a smaller effect on ipsilateral than contralateral motor output. Somatosensory feedback should also be relatively lateralized for the forelimb areas. The authors could say a bit more about this issue and how it relates to their data and conclusions in the Discussion.

An important limitation of the behavioral task is that it involves only a single stereotyped movement for each limb, instead of multiple directions, speeds, or loads. This issue and its consequences for the analyses (especially those in Figures 7-10) and conclusions could be discussed.

Reviewer #3 (Public review):

Summary:

Barrett et al. compare the responses of different parts of the mouse primary and secondary motor cortex in the context of a task where the animals manipulate and eat food using either or both hands. They find that roughly half the activity is conserved when reaching with one hand vs. the other hand, or with both. Similarity of activity was somewhat higher in the "lateral oral and manual" (LOM) part of the motor cortex, consistent with notions of a more generalized oromanual function there.

Strengths:

This work aims at addressing two worthwhile questions in a mouse model of motor control: (1) what specializations do we have for controlling feeding movements, and (2) how are the arms and hands coordinated with one another? The authors develop a simple but innovative apparatus to block either hand during food handling, track the behavior at high temporal fidelity, and record a sizable neural dataset. The analyses come from numerous angles to take good advantage of the data, and succeed in showing multiple lines of evidence for greater invariance in LOM than in the forelimb parts of M1 and M2.

Weaknesses:

There are several limitations of the current study. Most importantly, the behavior presents an inherent challenge: there is only one type of movement for each of the three conditions (contra hand, ipsi hand, and bimanual). This is entirely reasonable from the perspective that this is the ethological behavior when feeding, but it limits what analyses are possible. In particular, it precludes disentangling the neural relationship with many correlated aspects of behavior, and limits identifying population-level features of the neural activity meaningfully. This means that there are a number of alternative possible sources of the neuron-level area differences found here, and the population-level features may not be reliable. Second, the behavior tracking was used at a relatively coarse level, and thus the relationships to various behavioral variables were left less distinguishable than they might have been. Finally, there may be an issue with the coordinates of what is being called forelimb M1 here, which may include some hindlimb M1.

Author response:

Reviewer #1 (Public review):

Summary:

This study describes motor cortical activity patterns during food handling in mice, investigating whether the hand/s used is reflected in distinct neural activity. The experiments focus on forelimb M1 and M2 (fM1, fM2) and an oral-manual region LOM. The main findings are that fM1 and fM2 have largely similar relationships with forelimb control, and LOM neurons are more broadly tuned. These conclusions are reached using a variety of analyses spanning straightforward firing rate analyses, selectivity metrics, PCA, and GLM decoding methods to assess tuning generalizability. The study's significance is strengthened by including analyses of bimanual control, and in this sphere, there are descriptive data and analyses that aficionados of cortical control of dexterous behaviors will find useful. The use of unimanual control is useful as a point of comparison here, but less novel overall. There are a number of places where the descriptions of what is being analyzed, what is being concluded, and data reporting should be strengthened and clarified. Additionally, the study could be greatly improved by consolidating figures and the analyses shown, since many are redundant. Many of the analyses need clearer reporting of means and effect sizes in the text, rather than just statistical outcomes. Overall, at this juncture, the study presents analyses of a unique dataset that may seed future investigations of mechanisms of bimanual coordination.

Strengths:

There are relatively few studies that compare neural activity across bimanual and unimanual control. This study uses a naturalistic food handling task to explore neural relationships to forelimb kinematics under these conditions. The uniqueness of the task and analysis target is a strength of the study.

The authors remain fairly conservative and make few strong claims in the study, which may be warranted given the diversity of tuning profiles they observed.

Weaknesses:

There are a number of statistical tests that were accompanied by too little information to critically evaluate. Means and effect sizes needed to be better reported; some details of analyses were difficult to parse, making the strength of the conclusions difficult to evaluate.

We will improve these aspects of the presentation and reporting of statistical analyses in the revised manuscript.

Reviewer #2 (Public review):

Summary:

Barrett et al. examine how neural activity in the mouse motor cortex varies when a movement is performed with the ipsilateral or contralateral forelimb. First, they train animals to grasp and manipulate a pellet of food with either the left forepaw, the right forepaw, or both. Next, they measure activity in the primary and secondary forelimb motor areas (fl-M1 and fl-M2) and in the classical tongue-jaw area (tj-M1 / LOM). While responses in the forelimb areas are diverse, with some neurons preferring ipsilateral or bilateral movements, a plurality of cells prefer the contralateral limb. In LOM, by contrast, little limb selectivity is observed. At the neural population level, structure is preserved across conditions in LOM, but not in the forelimb areas. Finally, paw position can be decoded from activity in all three areas, and the LOM decoder generalized across limbs.

Strengths:

While previous studies in macaques have compared motor cortical activity during movement (and perturbation) of the contralateral and ipsilateral arms, no analogous work has been undertaken in rodents. This paper closes this knowledge gap by showing, for the first time, moderate-to-strong lateralization in the forelimb motor cortical areas of mice transporting grasped food pellets to the mouth, and a relative absence of lateralization in the classical tongue-jaw area. On the whole, I think this is a solid paper that reports novel observations of interest to the motor systems community.

Weaknesses:

The central question posed is whether cortical activity depends on the effector(s) used (ipsi forelimb, contra forelimb, or both). The corresponding hypotheses (Figure 1) are somewhat coarse-grained and are not mutually exclusive. One might expect to see condition-independent, limb-selective, and uni-/bimanual-selective signals in motor cortex (though their magnitudes could differ substantially), and to find these signals intermingled at the level of single neurons. The authors may wish to consider setting up a more focused question. For example, can bimanual responses be explained as a sum of the unimanual responses from the left and right limbs?

In the revised manuscript, we will clarify that the possibilities illustrated in figure 1 are not intended as mutually exclusive. We indeed find all of these signals intermingled. In terms of a more focused question amenable to hypothesis testing, this can be expressed as: for each dimension (laterality vs manuality) are the activity patterns in each area closer to those predicted by invariance or dependence, as compared to the other areas? The various statistical analyses in the paper all essentially boil down to testing this question. Broadly, the answer is yes: we see activity closer to the invariant prediction LOM, and activity closer to the dependent prediction in fl-M1 and fl-M2. Testing whether bimanual activity can be explained as a simple linear sum of the left and right unimanual might provide additional insight into this question, and this analysis will be presented in the revised manuscript.

In the area usually identified as tongue-jaw motor cortex (here referred to as LOM), unit and population activity look quite similar for ipsilateral, contralateral, and bilateral forelimb reaches. The most parsimonious explanation is that the activity is related mostly to mouth and tongue movements, rather than limb movements. Systematic mapping studies with microstimulation in the rat (Neafsy et al., Brain Res. Rev. 1986) and optogenetic stimulation in the mouse (Mayrhofer et al., Neuron 2019) tend to support the idea that tjM1/LOM is specialized for control of the tongue and mouth. Thus, I'm not entirely convinced that it "encodes ingestion-related forelimb parameters necessary for oromanual coordination." The authors could say more about this issue: what specific limb-related parameters do they think are encoded, why would these parameters be effector-independent, what evidence for this encoding is presented here, and how can limb- and mouth-related components be distinguished? The problem could potentially be addressed experimentally, as well, by delivering food pellets directly to the mouth while preventing manipulation with the paws, but this experiment isn't strictly necessary.

The issue of orofacial movement confounds is an important one that we made a point of addressing in the discussion. There are three main points that we believe cast doubt on this as the most likely explanation for the effector-invariant representation in LOM.

First, while we do not disagree that LOM has an important role in tongue and jaw control, there is plenty of evidence from mapping and behavioural studies (which we cite in the introduction) that it also plays a role in forelimb motor control as well.

Second, while we cannot observe all orofacial movements, we have previously shown that the jaw is less active when the hands and LOM are most active (Barrett et al., 2024). Conversely, LOM firing is much lower during chewing, when the tongue and jaw are very active.

Finally, the correlation between LOM firing and forelimb movements is not merely a coarse-grained one on the timescale of active manipulation vs passive holding phases. LOM firing closely tracks the position of the forelimb(s) on fast timescales and with near-zero lag, giving better decoding than from fl-M1 or fl-M2, as we have shown here and previously (Barrett et al., 2022). If we assume that LOM only encodes orofacial movements, then this result implies that orofacial movements correlate with forelimb position better than fl-M1 or fl-M2 firing correlates with forelimb position.

The revised manuscript will include an expanded discussion to clarify these and related points.

Because the corticospinal tract is strongly lateralized, cortical activity presumably has a smaller effect on ipsilateral than contralateral motor output. Somatosensory feedback should also be relatively lateralized for the forelimb areas. The authors could say a bit more about this issue and how it relates to their data and conclusions in the Discussion.

We will discuss this in the revised manuscript.

An important limitation of the behavioral task is that it involves only a single stereotyped movement for each limb, instead of multiple directions, speeds, or loads. This issue and its consequences for the analyses (especially those in Figures 7-10) and conclusions could be discussed.

This limitation applies to the analyses relating to the transport-to-mouth movement (Figures 3-7). The population correlation structure and decoding analyses (Figures 8-10) consider activity throughout the full duration of food handling, which involves a much greater variety of movements (Barrett et al., 2020). Indeed, this was a major motivation for including these analyses. The revised manuscript will clarify this point.

Reviewer #3 (Public review):

Summary:

Barrett et al. compare the responses of different parts of the mouse primary and secondary motor cortex in the context of a task where the animals manipulate and eat food using either or both hands. They find that roughly half the activity is conserved when reaching with one hand vs. the other hand, or with both. Similarity of activity was somewhat higher in the "lateral oral and manual" (LOM) part of the motor cortex, consistent with notions of a more generalized oromanual function there.

Strengths:

This work aims at addressing two worthwhile questions in a mouse model of motor control: (1) what specializations do we have for controlling feeding movements, and (2) how are the arms and hands coordinated with one another? The authors develop a simple but innovative apparatus to block either hand during food handling, track the behavior at high temporal fidelity, and record a sizable neural dataset. The analyses come from numerous angles to take good advantage of the data, and succeed in showing multiple lines of evidence for greater invariance in LOM than in the forelimb parts of M1 and M2.

Weaknesses:

There are several limitations of the current study. Most importantly, the behavior presents an inherent challenge: there is only one type of movement for each of the three conditions (contra hand, ipsi hand, and bimanual).

See our response to reviewer #2 above regarding the variety of movements. We agree that this a limitation for the unit-level and PCA analyses, but one that is alleviated by the population correlation and decoding analyses, which relate to complex ongoing movements.

This is entirely reasonable from the perspective that this is the ethological behavior when feeding, but it limits what analyses are possible. In particular, it precludes disentangling the neural relationship with many correlated aspects of behavior, and limits identifying population-level features of the neural activity meaningfully. This means that there are a number of alternative possible sources of the neuron-level area differences found here, and the population-level features may not be reliable.

Important behavioural confounds include orofacial movements, non-specific movement initiation signals, and arousal. Orofacial movements we have discussed above in our response to Reviewer 2. Movement initiation signals would likely be transient and well-timed to movement onset, hence this may explain some of the effector-independent activity in fl-M1 and fl-M2 (consistent with e.g. (Kaufman et al., 2016)). However, we do not believe this to be the case in LOM as its activity is delayed and sustained relative to movement initiation. Regarding arousal, the mouse is actively engaged in consuming the food even when the hands are stationary, so there is no a priori reason to believe that arousal varies rapidly during the behaviour. Consistent with this, measurements of noradrenergic activity in the locus coeruleus during consumption suggest that arousal varies on slow timescales, on the order of seconds to tens of seconds (Sciolino et al., 2022). Such slow variation in arousal would not explain the rapid but condition-invariant changes in firing in any of the cortical areas studied here. The updated manuscript will include more detailed discussion of these points.

Second, the behavior tracking was used at a relatively coarse level, and thus the relationships to various behavioral variables were left less distinguishable than they might have been.

Behavior tracking was performed with kilohertz temporal resolution and submillimeter spatial resolution.

Finally, there may be an issue with the coordinates of what is being called forelimb M1 here, which may include some hindlimb M1.

Our recording coordinates are based on the territory of corticospinal neurons retrogradely labeled from C6 spinal cord, medial to any layer 4 labelling, as reported in our previous study (Yamawaki et al., 2021). Thus we are confident in calling this area forelimb M1.

References:

Barrett, J. M., Martin, M. E., Gao, M., Druzinsky, R. E., Miri, A., & Shepherd, G. M. G. (2024). Hand-jaw coordination as mice handle food is organized around intrinsic structure-function relationships. The Journal of Neuroscience, 44(42), e0856242024.

Barrett, J. M., Martin, M. E., & Shepherd, G. M. G. (2022). Manipulation-specific cortical activity as mice handle food. Current Biology, 32(22), 4842-4853.e6.

Barrett, J. M., Tapies, M. G. R., & Shepherd, G. M. G. (2020). Manual dexterity of mice during food-handling involves the thumb and a set of fast basic movements. PLOS ONE, 15(1), e0226774.

Kaufman, M. T., Seely, J. S., Sussillo, D., Ryu, S. I., Shenoy, K. V., & Churchland, M. M. (2016). The Largest Response Component in the Motor Cortex Reflects Movement Timing but Not Movement Type. eNeuro, 3(4).

Sciolino, N. R., Hsiang, M., Mazzone, C. M., Wilson, L. R., Plummer, N. W., Amin, J., Smith, K. G., McGee, C. A., Fry, S. A., Yang, C. X., Powell, J. M., Bruchas, M. R., Kravitz, A. V., Cushman, J. D., Krashes, M. J., Cui, G., & Jensen, P. (2022). Natural locus coeruleus dynamics during feeding. Science Advances, 8(33), eabn9134.

Yamawaki, N., Raineri Tapies, M. G., Stults, A., Smith, G. A., & Shepherd, G. M. (2021). Circuit organization of the excitatory sensorimotor loop through hand/forelimb S1 and M1. eLife, 10, e66836.