Context-based sensorimotor routing is a hallmark of executive control. Pharmacological inactivations in rats have implicated the midbrain superior colliculus (SC) in this process. But what specific role is this, and what circuit mechanisms support it? Here we report a subset of rat SC neurons that instantiate a specific link between the representations of context and motor choice. Moreover, these neurons encode animals’ choice far earlier than other neurons in the SC or in the frontal cortex, suggesting that their neural dynamics lead choice computation. Optogenetic inactivations revealed that SC activity during context encoding is necessary for choice behavior, even while that choice behavior is robust to inactivations during choice formation. Searches for SC circuit models matching our experimental results identified key circuit predictions while revealing some a priori expected features as unnecessary. Our results reveal circuit mechanisms within the SC that implement response inhibition and context-based vector inversion during executive control.
This is a preview of subscription content, access via your institution
Open Access articles citing this article.
Nature Open Access 27 July 2022
Signal Transduction and Targeted Therapy Open Access 08 June 2022
Subscribe to Nature+
Get immediate online access to the entire Nature family of 50+ journals
Subscribe to Journal
Get full journal access for 1 year
only $8.25 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Get time limited or full article access on ReadCube.
All prices are NET prices.
Processed behavioral, electrophysiological, optogenetic and video data are publicly available on GitHub: https://github.com/Brody-Lab/Proanti. Raw data are archived at Princeton University and available from the corresponding author upon reasonable request. Modeling data are publicly available on GitHub: https://github.com/carlosbrody/superior_colliculus_mutual_inhibition.
All software used for behavioral training is available on the Brody lab website at http://brodylab.org/code/proanti-code. All custom data analysis and modeling codes are freely available on the corresponding GitHub repositories: https://github.com/Brody-Lab/Proanti (analysis) and https://github.com/carlosbrody/superior_colliculus_mutual_inhibition (modeling).
Miller, E. K. & Cohen, J. D. An integrative theory of prefrontal cortex function. Annu. Rev. Neurosci. 24, 167–202 (2001).
Hallett, P. E. Primary and secondary saccades to goals defined by instructions. Vis. Res. 18, 1279–1296 (1978).
Duan, C. A., Erlich, J. C. & Brody, C. D. Requirement of prefrontal and midbrain regions for rapid executive control of behavior in the rat. Neuron 86, 1491–1503 (2015).
Weiler, J. & Heath, M. Task-switching in oculomotor control: unidirectional switch-cost when alternating between pro- and antisaccades. Neurosci. Lett. 530, 150–154 (2012).
Munoz, D. P. & Everling, S. Look away: the anti-saccade task and the voluntary control of eye movement. Nat. Rev. Neurosci. 5, 218–228 (2004).
Everling, S. & Fischer, B. The antisaccade: a review of basic research and clinical studies. Neuropsychologia 36, 885–899 (1998).
Hutton, S. B. & Ettinger, U. The antisaccade task as a research tool in psychopathology: a critical review. Psychophysiology 43, 302–313 (2006).
Lo, C.-C. & Wang, X.-J. Conflict resolution as near-threshold decision-making: a spiking neural circuit model with two-stage competition for antisaccadic task. PLoS Comput. Biol. 12, e1005081 (2016).
Everling, S. & Johnston, K. Control of the superior colliculus by the lateral prefrontal cortex. Philos. Trans. R. Soc. Lond. B Biol. Sci. 368, 20130068 (2013).
Felsen, G. & Mainen, Z. F. Neural substrates of sensory-guided locomotor decisions in the rat superior colliculus. Neuron 60, 137–148 (2008).
Sparks, D. L. & Hartwich-Young, R. The deep layers of the superior colliculus. Rev. Oculomot. Res. 2, 213–255 (1989).
Wurtz, R. H. & Goldberg, M. E. Activity of superior colliculus in behaving monkey. 3. Cells discharging before eye movements. J. Neurophysiol. 35, 575–586 (1972).
Evans, D. A., Stempel, A. V., Vale, R. & Branco, T. Cognitive control of escape behaviour. Trends Cogn. Sci. 23, 334–348 (2019).
Gandhi, N. J. & Katnani, H. A. Motor functions of the superior colliculus. Annu. Rev. Neurosci. 34, 205–231 (2011).
Robinson, D. A. Eye movements evoked by collicular stimulation in the alert monkey. Vis. Res. 12, 1795–1808 (1972).
Johnston, K., Koval, M. J., Lomber, S. G. & Everling, S. Macaque dorsolateral prefrontal cortex does not suppress saccade-related activity in the superior colliculus. Cereb. Cortex 24, 1373–1388 (2014).
Basso, M. A. & May, P. J. Circuits for action and cognition: a view from the superior colliculus. Annu Rev. Vis. Sci. 3, 197–226 (2017).
Wolf, A. B. et al. An integrative role for the superior colliculus in selecting targets for movements. J. Neurophysiol. 114, 2118–2131 (2015).
Krauzlis, R. J., Lovejoy, L. P. & Zénon, A. Superior colliculus and visual spatial attention. Annu. Rev. Neurosci. 36, 165–182 (2013).
Crapse, T. B., Lau, H. & Basso, M. A. A role for the superior colliculus in decision criteria. Neuron 97, 181–194.e6 (2018).
Horwitz, G. D., Batista, A. P. & Newsome, W. T. Representation of an abstract perceptual decision in macaque superior colliculus. J. Neurophysiol. 91, 2281–2296 (2004).
Duan, C. A. et al. A cortico-collicular pathway for motor planning in a memory-dependent perceptual decision task. Nat. Commun. 12, 2727 (2021).
Schmitt, L. I. et al. Thalamic amplification of cortical connectivity sustains attentional control. Nature 545, 219–223 (2017).
Erlich, J. C., Bialek, M. & Brody, C. D. A cortical substrate for memory-guided orienting in the rat. Neuron 72, 330–343 (2011).
Kopec, C. D., Erlich, J. C., Brunton, B. W., Deisseroth, K. & Brody, C. D. Cortical and subcortical contributions to short-term memory for orienting movements. Neuron 88, 367–377 (2015).
Felsen, G. & Mainen, Z. F. Midbrain contributions to sensorimotor decision making. J. Neurophysiol. 108, 135–147 (2012).
Everling, S., Dorris, M. C., Klein, R. M. & Munoz, D. P. Role of primate superior colliculus in preparation and execution of anti-saccades and pro-saccades. J. Neurosci. 19, 2740–2754 (1999).
Rigotti, M. et al. The importance of mixed selectivity in complex cognitive tasks. Nature 497, 585–590 (2013).
Mante, V., Sussillo, D., Shenoy, K. V. & Newsome, W. T. Context-dependent computation by recurrent dynamics in prefrontal cortex. Nature 503, 78–84 (2013).
Pagan, M., Urban, L. S., Wohl, M. P. & Rust, N. C. Signals in inferotemporal and perirhinal cortex suggest an untangling of visual target information. Nat. Neurosci. 16, 1132–1139 (2013).
Hanks, T. D. et al. Distinct relationships of parietal and prefrontal cortices to evidence accumulation. Nature 520, 220–223 (2015).
Prinz, A. A., Bucher, D. & Marder, E. Similar network activity from disparate circuit parameters. Nat. Neurosci. 7, 1345–1352 (2004).
Fisher, D., Olasagasti, I., Tank, D. W., Aksay, E. R. F. & Goldman, M. S. A modeling framework for deriving the structural and functional architecture of a short-term memory microcircuit. Neuron 79, 987–1000 (2013).
Bittner, S. R. et al. Interrogating theoretical models of neural computation with deep inference. Preprint at bioRxiv https://doi.org/10.1101/837567 (2019).
Dean, P., Mitchell, I. J. & Redgrave, P. Contralateral head movements produced by microinjection of glutamate into superior colliculus of rats: evidence for mediation by multiple output pathways. Neuroscience 24, 491–500 (1988).
May, P. J. The mammalian superior colliculus: laminar structure and connections. Prog. Brain Res. 151, 321–378 (2006).
Machens, C. K., Romo, R. & Brody, C. D. Flexible control of mutual inhibition: a neural model of two-interval discrimination. Science 307, 1121–1124 (2005).
Cutsuridis, V., Smyrnis, N., Evdokimidis, I. & Perantonis, S. A neural model of decision-making by the superior colliculus in an antisaccade task. Neural Netw. 20, 690–704 (2007).
Wong, K.-F. & Wang, X.-J. A recurrent network mechanism of time integration in perceptual decisions. J. Neurosci. 26, 1314–1328 (2006).
Steinmetz, N. A., Zatka-Haas, P., Carandini, M. & Harris, K. D. Distributed coding of choice, action and engagement across the mouse brain. Nature 576, 266–273 (2019).
Uylings, H. B. M., Groenewegen, H. J. & Kolb, B. Do rats have a prefrontal cortex? Behav. Brain Res. 146, 3–17 (2003).
Seamans, J. K., Lapish, C. C. & Durstewitz, D. Comparing the prefrontal cortex of rats and primates: insights from electrophysiology. Neurotox. Res. 14, 249–262 (2008).
Everling, S. & DeSouza, J. F. X. Rule-dependent activity for prosaccades and antisaccades in the primate prefrontal cortex. J. Cogn. Neurosci. 17, 1483–1496 (2005).
Johnston, K. & Everling, S. Monkey dorsolateral prefrontal cortex sends task-selective signals directly to the superior colliculus. J. Neurosci. 26, 12471–12478 (2006).
Zénon, A. & Krauzlis, R. J. Attention deficits without cortical neuronal deficits. Nature 489, 434–437 (2012).
McPeek, R. M. & Keller, E. L. Deficits in saccade target selection after inactivation of superior colliculus. Nat. Neurosci. 7, 757–763 (2004).
Sahibzada, N., Dean, P. & Redgrave, P. Movements resembling orientation or avoidance elicited by electrical stimulation of the superior colliculus in rats. J. Neurosci. 6, 723–733 (1986).
Dean, P., Redgrave, P. & Westby, G. W. Event or emergency? Two response systems in the mammalian superior colliculus. Trends Neurosci. 12, 137–147 (1989).
Aschauer, D. F., Kreuz, S. & Rumpel, S. Analysis of transduction efficiency, tropism and axonal transport of AAV serotypes 1, 2, 5, 6, 8 and 9 in the mouse brain. PLoS ONE 8, e76310 (2013).
Franceschi, G. & Solomon, S. G. Visual response properties of neurons in the superficial layers of the superior colliculus of awake mouse. J. Physiol. 596, 6307–6332 (2018).
Pagan, M. & Rust, N. C. Dynamic target match signals in perirhinal cortex can be explained by instantaneous computations that act on dynamic input from inferotemporal cortex. J. Neurosci. 34, 11067–11084 (2014).
Tibshirani, R. J. & Efron, B. An introduction to the bootstrap. Monogr. Stat. Appl. Probab. 57, 1–436 (1993).
Goldman, M. S. Memory without feedback in a neural network. Neuron 93, 715 (2017).
Murphy, B. K. & Miller, K. D. Balanced amplification: a new mechanism of selective amplification of neural activity patterns. Neuron 61, 635–648 (2009).
Schaub, M. T., Billeh, Y. N., Anastassiou, C. A., Koch, C. & Barahona, M. Emergence of slow-switching assemblies in structured neuronal networks. PLoS Comput. Biol. 11, e1004196 (2015).
We thank K. Osorio and J. Teran for animal and laboratory support. This work was funded by the Howard Hughes Medical Institute. C.A.D. was supported by a Howard Hughes Medical Institute predoctoral fellowship. C.A.D. and M.P. are supported by a Simons Collaboration on the Global Brain postdoctoral fellowship.
The authors declare no competing interests.
Peer review information Nature Neuroscience thanks Gidon Felsen and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Asymmetries between Pro and Anti response time (RT), accuracy, and task switch cost in implanted rats. a, Normalized RT distributions of an example rat. Histograms of correct Pro (n = 3894 trials) and correct Anti (n = 3323) RTs are shown on top and error Pro (n = 1239 trials) and error Anti (n = 1161 trials) RTs are shown in the bottom. Each curve is normalized to have a total area of 1. Median RTs for Pro and Anti hits and errors are indicated by vertical bars; 95% confidence intervals across trials for each trial type are indicated by horizontal bars. b, RT summary of 16 individual rats (7 for SC and PFC neural recordings and 9 for optogenetic inactivation experiments). Left: median RTs for Anti hits and Pro hits for all rats (n = 16). ***P = 4 × 10−4, two-sided bootstrap test. Right: RT difference between Pro and Anti, hits and errors, averaged across all rats (n = 16). For each rat, the difference between median RTs of paired conditions was calculated. White bar shows the mean and s.e.m. across rats for Anti hit RTs minus Pro hit RTs, P = 4 × 10−4, two-sided bootstrap test. Green bar shows Pro hit RTs minus Pro error RTs, P = 4 × 10−4, two-sided bootstrap test. Orange bar shows Anti hit RTs minus Anti error RTs, P = 4 × 10−4, two-sided bootstrap test. c, Pro and Anti performance for individual rats (n = 16). Mean and s.e.m. of Pro and Anti performance are computed over sessions for each rat and plotted against each other. Average Pro (green) and Anti (orange) performance across rats was plotted in the upper left corner (n = 16). Pro versus Anti, P = 0.003, two-sided bootstrap test. d, Switch cost asymmetry. Left: percent correct as a function of trial number relative to a task block switch for one example rat. Each data point is the mean and s.e.m. across trials for Pro and Anti accuracy on three trials before and after the switch. Right: average accuracy switch cost for Pro trials (P = 4 × 10−4) and Anti trials (P = 4 × 10−4) across rats (n = 16). The cost of switching to Pro was larger than the cost of switching to Anti (P = 0.002), two-sided bootstrap tests.
a, A light in the center port indicates that rats should nose poke there to initiate a trial and keep their noses there until the center light offset (‘‘fixation’’ period). During the first 1 s of the fixation period, a Pro or Anti sound is played to indicate the current task, followed by a 500-ms silent delay. The center light is then turned off, indicating that the animal is now free to withdraw from the center port, and the moment it withdraws, a left or right light is turned on to indicate the target location. The temporal gap between fixation offset (that is, end of the delay period) and target stimulus onset was controlled by animals and was thus variable on each trial (mean = 127 ms after fixation offset). Reaction Time (RT) is defined as the time from target onset until side poke. The 3 vertical lines correspond to the vertical lines in Fig. 1–3. b, Similar to a, for optogenetic sessions. To ensure that all sub-trial optogenetic inactivation conditions have the same laser duration (750 ms, green shade), rats were trained on a modified version of the behavior where the task cue period and the delay period both lasted 750 ms. Choice period inactivation started at the onset of visual target and lasted 750 ms, covering the time it took animals to form and execute the orienting choice into the side poke (690.8 ± 39.1 ms, mean ± s.e.m. across animals’ median RT in optogenetic inactivation sessions).
Extended Data Fig. 3 Individual PFC neurons encode task and choice variables during flexible sensorimotor routing.
a-c, Same as in Fig. 1c-e, for the PFC population (291 out of 331 total neurons).
Similar to Fig. 1c,e, separated by recordings from individual rats. Mean performance on Pro and Anti trials during each rat’s recording sessions are shown above each panel. Rat ‘J205’, ‘A117’, and ‘Z014’ had implants both in SC and in PFC. Rats with fewer than 20 neurons were excluded from this analysis.
Extended Data Fig. 5 Relationship between task context (Pro/Anti) and choice (Contra/Ipsi) d’ across the SC population.
a, For each SC neuron, the signed Pro/Anti d’ computed at the time of peak Pro/Anti selectivity was plotted against the signed Choice d’ computed at the time of peak Choice selectivity. No correlation is observed (Pearson’s correlation coefficient r = 0.06, n = 193, P = 0.3821, t-test). b, Correlation between Pro/Anti d’ and Choice d’ for the whole SC population computed at all time points. The two black dashed lines in the color bar indicate the correlation values that are not significantly different from 0 (P > 0.05). The correlation is significantly different than 0 only at times shortly after the appearance of the target stimulus. Positive correlation corresponds to either a Pro (d’ > 0) and Contra (d’ > 0) preference or an Anti (d’ < 0) and Ipsi (d’ < 0) preference. Bin size = 250 ms, centered (that is, it includes spikes from ± 125 ms relative to the plotted timepoint.
a, SC population decoding performance (mean ± s.d.) for linear classification of correct Pro versus Anti trials (task, red line), Go-Left versus Go-Right trials (choice, blue line), and Left-Light versus Right-Light trials (light stimulus, black line). Compared to the early and strong choice information in the SC population, linearly decodable information related to the light stimulus appeared later and weaker, suggesting that information being received by deep SC layer neurons about which side the Light is on is combined nonlinearly and very rapidly with context information, to produce early, linearly decodable information about choice. b, Matrix of light stimulus (left side light/ right side light) selectivity for the SC neural population, similar to Fig. 1e.
a, Effect of full-trial and sub-trial inactivations of bilateral SC on Pro (green) and Anti (orange) error rate (mean and s.e.m.) compared to YFP controls (gray). Full-trial: n = 662, 615 for Pro and Anti inactivation trials; n = 362, 322 for Pro and Anti control trials. ***P = 4 × 10−4, two-sided permutation test. Task cue: n = 413, 401 for Pro and Anti inactivation trials; n = 290, 271 for Pro and Anti control trials. Delay: n = 562, 527 for Pro and Anti inactivation trials; n = 315, 260 for Pro and Anti control trials. ***P = 4 × 10−4; **P = 0.0012, two-sided permutation tests. Choice: n = 547, 506 for Pro and Anti inactivation trials; n = 319, 261 for Pro and Anti control trials. All paired statistics shown here are computed using a two-sided permutation test, shuffled 5000 times. b, Effect of full-trial and sub-trial inactivations of bilateral SC on response time (RT). For each behavioral session, a median RT on non-stimulated control trials is calculated and subtracted from the RTs on inactivation trials, and these normalized RT changes due to inactivation are plotted here. Each curve is normalized to have a total area of 1. Vertical bars show the median RT changes for correct Pro and Anti trials; s.e.m. across trials for each trial type are indicated by horizontal bars. A shift to the right indicates slowing due to inactivation and a shift to the left indicates speeding.
Extended Data Fig. 8 Variability across model solutions in dynamics and parameters, and common functional properties.
a, Distribution of choice preference (d’) for Pro and Anti model units from 373 individual model solutions during the choice period (Methods). Note that although most Anti units (red shading) were Ipsi-preferring, we also observed Anti/Contra-preferring units (red shaded counts to the right of zero), similar to the SC neural data (Fig. 4). In contrast, all Pro units (gray shading) were Contra-preferring. b, The dimensionality of parameters across model solutions, and of dynamics across model solutions (n = 373 solutions). Eight SVD dimensions are required to explain 90% of the variance in dynamics across model solutions. Ten PCA dimensions are required to explain 90% of the variance in parameters across model solutions. c, Variance explained by each dimension of PCA performed on each model solution’s dynamics. Full trial: PCA computed on all time points. Delay period only: PCA computed only during the delay period. Target period only: PCA computed only during the target period. Mean ± s.d across 373 model solutions. d, The connectivity matrix of each model solution was analyzed via the Schur Decomposition (Methods). All solutions (n = 373) contained one of each of the following functional modes: All, Side of brain, Task, and Diagonal. The percentage of solutions with positive eigenvalues for each mode is reported.
a, Schematic of the 6-node SC model, in which each hemisphere contains two Anti pools and one Pro pool. b, Format and results similar to Fig. 7e. Histogram of horizontal weights between the two Pro units (as illustrated by the insert cartoon) for all 36 six-node model solutions. Red arrow marks average value across solutions. Solutions do not require inhibitory weights between the two Pro/Contra pools. c, Format and results similar to Fig. 7c. Scatter plot of diagonal weights (from Anti units to the Pro unit on the opposite hemisphere) against vertical weights (from Anti units to the Pro unit on the same hemisphere), for all model solutions. Each dot represents the average weights from the two Anti units in a solution. Red line marks unity. d, Format and results similar to Extended Data Fig. 8a. Histogram of choice d’ for Pro and Anti nodes during the choice period (n = 36 model solutions). We observed both Anti/Ipsi and Anti-Contra-preferring units, with a majority of Anti/Ipsi units, as in the experimental data. e, Similar to Extended Data Fig. 8d. Percentage of model solutions with positive eigenvalues for each Schur mode type, based on Schur Decomposition analysis of the connectivity matrix. The solution networks (n = 36 solutions, red, mean ± 95% CI) are compared against 10,000 random networks (black) with the same symmetry and parameter value constraints. Dashed line indicates results from 10,000 random networks with the same symmetry, but not the parameter value constraints.
The distribution of parameter values across all solutions (n = 373) is plotted for each of the 16 free parameters (Methods). Vertical dashed line marks zero for reference. Red arrow marks average parameter value across solutions. The weight parameters determined the connectivity matrix between units. The noise parameter was the variance of white noise added to each unit on each time step. The Pro and Anti rule input weights determined the strength of the task context inputs to either the Pro or Anti units. The stimulus input determined the weight of the stimulus to either the Left or Right units. The Pro bias term was a constant input to only the Pro units. The target period input was a constant input to all nodes, only during the target period. The constant input was a bias term during all time points for all units. The opto strength was the fraction of each node’s output that was transmitted to the other nodes during inactivations; a strength of 1 is no inactivation, a strength of 0 is complete inactivation.
About this article
Cite this article
Duan, C.A., Pagan, M., Piet, A.T. et al. Collicular circuits for flexible sensorimotor routing. Nat Neurosci 24, 1110–1120 (2021). https://doi.org/10.1038/s41593-021-00865-x
This article is cited by
Signal Transduction and Targeted Therapy (2022)