Altered basal ganglia output during self-restraint

  1. Bon-Mi Gu
  2. Joshua D Berke

  • Curated by eLife

    eLife logo

    eLife assessment

    This study builds upon previous results of the authors to study the neural computations within the basal ganglia that support behavioral proactive inhibition. Here, the authors identify features of neural activity in the SNr that correlate with proactive inhibition, including changes in firing rate and neural variability, and how both of these variables are influenced by an animal's outcome history. The analyses are rigorous and provide important insights into the neural dynamics in the basal ganglia that support proactive inhibition.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Suppressing actions is essential for flexible behavior. Multiple neural circuits involved in behavioral inhibition converge upon a key basal ganglia output nucleus, the substantia nigra pars reticulata (SNr). To examine how changes in basal ganglia output contribute to self-restraint, we recorded SNr neurons during a proactive behavioral inhibition task. Rats responded to Go! cues with rapid leftward or rightward movements, but also prepared to cancel one of these movement directions on trials when a Stop! cue might occur. This action restraint – visible as direction-selective slowing of reaction times – altered both rates and patterns of SNr spiking. Overall firing rate was elevated before the Go! cue, and this effect was driven by a subpopulation of direction-selective SNr neurons. In neural state space, this corresponded to a shift away from the restrained movement. SNr neurons also showed more variable inter-spike intervals during proactive inhibition. This corresponded to more variable state-space trajectories, which may slow reaction times via reduced preparation to move. These findings open new perspectives on how basal ganglia dynamics contribute to movement preparation and cognitive control.

Article activity feed

  1. Version published to 10.7554/elife.82143 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Public Review):

    Gu et al. examine how activity in the substantia nigra pars reticulata (SNr) contributes to proactive inhibition - the suppression of upcoming actions - by recording SNr activity in rats performing a task requiring them to be prepared to cancel a planned movement. This task was developed in a previous study by the same authors in which they examined how globus pallidus pars externa (GPe) activity depends on proactive inhibition (Gu et al., 2020), which motivated the present focus on SNr. The task is rich and the complementary analyses of how the neural activity relates to the behavior, at the level of individual neurons and populations, are appropriate and illuminating. Overall, this study is well done and has the potential to be a nice contribution to our understanding of how the SNr, and therefore the basal ganglia, mediate behavioral inhibition. Addressing a few questions, however, would improve the paper.

    We appreciate both the positive comments and constructive criticism.

    • It is not obvious why the presence or absence of proactive inhibition should be determined on a session-by-session basis. It seems quite possible that proactive inhibition is not an all-or-none phenomenon, and also that it might be exhibited to a greater or lesser extent across a session (e.g., due to changes in motivational drive). It would therefore strengthen the paper to better explain the rationale for comparing neural activity across entire sessions "with" and "without" proactive inhibition. Within-session variation in proactive inhibition could be quite advantageous, allowing for within-neuron comparisons. It is even possible that the differences in neural activity that the authors report here using session-by-session analysis are an underestimate of the true effect of proactive inhibition.

    It is true that some of our analyses compare whole sessions with- and without- overall behavioral evidence for proactive inhibition. But our primary results come from within-session comparisons of Maybe-Stop to No-Stop trials. For this purpose, the session-wide assessment of proactive inhibition is primarily a screen for which sessions to use for within-session analysis.

    It would be desirable if we could use behavior to determine the degree of proactive inhibition on each individual trial, and then compare this to neural measures. Unfortunately, this is not generally feasible in our experiments. Our key evidence for proactive inhibition is the prolongation of reaction times (RTs). However, RTs are famously highly variable over trials. This variability likely reflects a variety of factors, not simply proactive inhibition. For example, in our previous paper (Gu et al. 2020) we showed that dividing trials into slower and faster RTs did not reproduce the same neural differences as comparing Maybe-Stop to No-Stop trials.

    An alternative approach to investigating proactive inhibition is to focus on the increased restraint that typically follows over-hasty responses. We found that when rats fail to Stop, on the next trial the degree of SNr variability increases (Fig. 6). We have now expanded this analysis to include additional types of errors. We find that another form of over-hasty action, premature responses before the Go cue, are also followed on the next trial by increased SNr variability (Fig. 6- supp1). By contrast, other error types (wrong choices; failure to respond quickly enough) do not provoke greater variability. These additional within-session analyses provide convergent evidence for increased variability as an adaptive response to failures evoked by excessive haste.

    • It is difficult to rule out alternative explanations for the observed differences in SNr activity. While the authors acknowledge this point in the 3rd paragraph of the discussion, they only discuss one potential alternative - reward expectation. Another difference between maybe-stop and no-stop trials is the likelihood that a particular target should be selected, which has also been shown to modulate SNr activity (Basso & Wurtz, 2002). As is often the case with complex behavioral tasks, there may be many other differences between trial types that may contribute to differences in neural activity. It would be helpful for the authors to more fully explain how their results relate to contextual modulation of SNr activity, and why the dependence of SNr activity on proactive inhibition may be a novel finding.

    We have expanded the Discussion to include additional alternative explanations.

    • A natural question arising from this study, as with most studies of neural recordings during behavior, is the causal nature of the neural activity. It would be non-trivial and beyond the scope of the current study to perform the sort of perturbations that could determine whether population variability causally relates to preparation to suppress actions. But it would be useful to discuss future experiments that might be able to test causality.

    We added in Discussion the possibility of using optogenetic manipulations of specific inputs to SNr, to help determine their distinct contributions to SNr firing patterns and proactive behavior.

    Reviewer #2 (Public Review):

    The authors have recorded the activity of neurons in the rat substancia nigra pars reticulata (SNr) while animals performed a version of a stop-signal task. The goal of this study is to investigate and describe the contribution of SNr in proactive inhibitory control. By examining single-cell responses as well as population activity, the authors show that increasing the probability of stop signal trials induces several changes in SNr responses. First, specific populations of SNr neurons increase their activity during proactive, direction-specific inhibition. At the population level, neurons are biased away from the side of the movement that has to be potentially inhibited. Second, during proactive inhibition, neuron activity is more variable, both at the single-cell and population levels. Finally, the authors show that animals' outcome history influences both firing rates and variability of neuron responses in the current trial. Especially, neural variability is increased following a failure to inhibit a movement.

    Strengths

    The manuscript provides an interesting and timely insight into the role of the basal ganglia output nucleus in movement initiation control. The paper is often clearly and concisely written (although see one issue related to this below). One of the main strengths of the work is to allow an interesting comparison with recent work by the same team, aimed at investigating the responses of another basal ganglia nucleus (GPe) in the same task, using similar analyses (this comparison is not extensively exploited in the discussion section though). Another potential strength is the use of different analysis scales. The authors investigated single-unit responses as well as population "trajectories" in the neural state space. This is an interesting option that could have been better motivated, given that the two approaches assume quite different brain operations.

    Thank you for the interest and careful comments.

    Weaknesses

    The analyses and results sometimes lack clarity and details. For instance, and unless I missed the information, it is not clearly stated whether "maybe-stop" trial analyses only include Go trials or if (failed) Stop trials are also considered. Moreover, quite complicated figures are often described very briefly in the main text. Methods are also often too succinctly described, and sometimes refer to a previous publication (Gu et al., 2020) that readers did not necessarily read.

    We have made a range of changes to make the analyses and their rationale more clear. This includes specifying that Maybe-Stop trials include both Go and Stop trials (and why). We have also added more details in both main text and Methods.

    There are some points that the authors might need to discuss more. Especially, a global picture of the role of the different basal ganglia nuclei during movement control would have been appreciated. Also, the authors monitored the activity of the rat basal ganglia output. We would have appreciated more information regarding the impact of this output activity on SNr target areas, as compared to their previous work that focused on GPe for instance. Another example concerns the observation that SNr activity is elevated during active inhibition regardless of the firing rate pattern before movement (increase or decrease). As noted by the authors themselves, this is inconsistent with the classical role assigned to the basal ganglia output nucleus (i.e. a decrease in activity promotes movement). Despite that this observation is of potential interest to readers working on the basal ganglia, it is not discussed.

    The revised Discussion includes a section on how altered basal ganglia output may affect targets to alter behavior.

  3. eLife

    eLife assessment

    This study builds upon previous results of the authors to study the neural computations within the basal ganglia that support behavioral proactive inhibition. Here, the authors identify features of neural activity in the SNr that correlate with proactive inhibition, including changes in firing rate and neural variability, and how both of these variables are influenced by an animal's outcome history. The analyses are rigorous and provide important insights into the neural dynamics in the basal ganglia that support proactive inhibition.

  4. eLife

    Reviewer #1 (Public Review):

    Gu et al. examine how activity in the substantia nigra pars reticulata (SNr) contributes to proactive inhibition - the suppression of upcoming actions - by recording SNr activity in rats performing a task requiring them to be prepared to cancel a planned movement. This task was developed in a previous study by the same authors in which they examined how globus pallidus pars externa (GPe) activity depends on proactive inhibition (Gu et al., 2020), which motivated the present focus on SNr. The task is rich and the complementary analyses of how the neural activity relates to the behavior, at the level of individual neurons and populations, are appropriate and illuminating. Overall, this study is well done and has the potential to be a nice contribution to our understanding of how the SNr, and therefore the basal ganglia, mediate behavioral inhibition. Addressing a few questions, however, would improve the paper.

    - It is not obvious why the presence or absence of proactive inhibition should be determined on a session-by-session basis. It seems quite possible that proactive inhibition is not an all-or-none phenomenon, and also that it might be exhibited to a greater or lesser extent across a session (e.g., due to changes in motivational drive). It would therefore strengthen the paper to better explain the rationale for comparing neural activity across entire sessions "with" and "without" proactive inhibition. Within-session variation in proactive inhibition could be quite advantageous, allowing for within-neuron comparisons. It is even possible that the differences in neural activity that the authors report here using session-by-session analysis are an underestimate of the true effect of proactive inhibition.

    - It is difficult to rule out alternative explanations for the observed differences in SNr activity. While the authors acknowledge this point in the 3rd paragraph of the discussion, they only discuss one potential alternative - reward expectation. Another difference between maybe-stop and no-stop trials is the likelihood that a particular target should be selected, which has also been shown to modulate SNr activity (Basso & Wurtz, 2002). As is often the case with complex behavioral tasks, there may be many other differences between trial types that may contribute to differences in neural activity. It would be helpful for the authors to more fully explain how their results relate to contextual modulation of SNr activity, and why the dependence of SNr activity on proactive inhibition may be a novel finding.

    - A natural question arising from this study, as with most studies of neural recordings during behavior, is the causal nature of the neural activity. It would be non-trivial and beyond the scope of the current study to perform the sort of perturbations that could determine whether population variability causally relates to preparation to suppress actions. But it would be useful to discuss future experiments that might be able to test causality.

  5. eLife

    Reviewer #2 (Public Review):

    The authors have recorded the activity of neurons in the rat substancia nigra pars reticulata (SNr) while animals performed a version of a stop-signal task. The goal of this study is to investigate and describe the contribution of SNr in proactive inhibitory control. By examining single-cell responses as well as population activity, the authors show that increasing the probability of stop signal trials induces several changes in SNr responses. First, specific populations of SNr neurons increase their activity during proactive, direction-specific inhibition. At the population level, neurons are biased away from the side of the movement that has to be potentially inhibited. Second, during proactive inhibition, neuron activity is more variable, both at the single-cell and population levels. Finally, the authors show that animals' outcome history influences both firing rates and variability of neuron responses in the current trial. Especially, neural variability is increased following a failure to inhibit a movement.

    Strengths
    The manuscript provides an interesting and timely insight into the role of the basal ganglia output nucleus in movement initiation control. The paper is often clearly and concisely written (although see one issue related to this below). One of the main strengths of the work is to allow an interesting comparison with recent work by the same team, aimed at investigating the responses of another basal ganglia nucleus (GPe) in the same task, using similar analyses (this comparison is not extensively exploited in the discussion section though). Another potential strength is the use of different analysis scales. The authors investigated single-unit responses as well as population "trajectories" in the neural state space. This is an interesting option that could have been better motivated, given that the two approaches assume quite different brain operations.

    Weaknesses
    The analyses and results sometimes lack clarity and details. For instance, and unless I missed the information, it is not clearly stated whether "maybe-stop" trial analyses only include Go trials or if (failed) Stop trials are also considered. Moreover, quite complicated figures are often described very briefly in the main text. Methods are also often too succinctly described, and sometimes refer to a previous publication (Gu et al., 2020) that readers did not necessarily read.
    There are some points that the authors might need to discuss more. Especially, a global picture of the role of the different basal ganglia nuclei during movement control would have been appreciated. Also, the authors monitored the activity of the rat basal ganglia output. We would have appreciated more information regarding the impact of this output activity on SNr target areas, as compared to their previous work that focused on GPe for instance. Another example concerns the observation that SNr activity is elevated during active inhibition regardless of the firing rate pattern before movement (increase or decrease). As noted by the authors themselves, this is inconsistent with the classical role assigned to the basal ganglia output nucleus (i.e. a decrease in activity promotes movement). Despite that this observation is of potential interest to readers working on the basal ganglia, it is not discussed.

  6. Version published to 10.1101/2022.04.23.489276v2 on bioRxiv
  7. Version published to 10.1101/2022.04.23.489276v1 on bioRxiv