科技科学

Targeted stimulation of motor cortex ensembles drives learned movements

Nature正文已收录本站

Data availability

The data used in this study are available at Zenodo (https://doi.org/10.5281/zenodo.22216381)48. Source data are provided with this paper.

Code availability

The code used in this study is available at Zenodo (https://doi.org/10.5281/zenodo.22216381)48.

References

  1. Peters, A. J., Chen, S. X. & Komiyama, T. Emergence of reproducible spatiotemporal activity during motor learning. Nature 510, 263–267 (2014).

    Article  ADS  CAS  PubMed  Google Scholar 

  2. Guo, L. et al. Dynamic rewiring of neural circuits in the motor cortex in mouse models of Parkinson’s disease. Nat. Neurosci. 18, 1299–1309 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  3. Kawai, R. et al. Motor cortex is required for learning but not for executing a motor skill. Neuron 86, 800–812 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Guo, J.-Z. et al. Cortex commands the performance of skilled movement. eLife 4, e10774 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

  5. Hwang, E. J. et al. Disengagement of motor cortex from movement control during long-term learning. Sci. Adv. 5, eaay0001 (2019).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  6. Makino, H., Hwang, E. J., Hedrick, N. G. & Komiyama, T. Circuit mechanisms of sensorimotor learning. Neuron 92, 705–721 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Peters, A. J., Liu, H. & Komiyama, T. Learning in the rodent motor cortex. Ann. Rev. Neurosci. 40, 77–97 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Kogan, E., Lu, J. & Zuo, Y. Cortical circuit dynamics underlying motor skill learning: from rodents to humans. Front. Mol. Neurosci. 16, 1292685 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

  9. Chen, S. X., Kim, A. N., Peters, A. J. & Komiyama, T. Subtype-specific plasticity of inhibitory circuits in motor cortex during motor learning. Nat. Neurosci. 18, 1109–1115 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Hedrick, N. G. et al. Learning binds new inputs into functional synaptic clusters via spinogenesis. Nat. Neurosci. 25, 726–737 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Hedrick, N. G., Wright, W. J. & Komiyama, T. Local and global predictors of synapse elimination during motor learning. Sci. Adv. 10, eadk0540 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Ren, C. et al. Global and subtype-specific modulation of cortical inhibitory neurons regulated by acetylcholine during motor learning. Neuron 110, 2334–2350.e8 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Makino, H. et al. Transformation of cortex-wide emergent properties during motor learning. Neuron 94, 880–890.e8 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Hwang, E. J., Dahlen, J. E., Mukundan, M. & Komiyama, T. Disengagement of motor cortex during long-term learning tracks the performance level of learned movements. J. Neurosci. 41, 7029–7047 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Marshel, J. H. et al. Cortical layer–specific critical dynamics triggering perception. Science 365, eaaw5202 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Packer, A. M., Russell, L. E., Dalgleish, H. W. P. & Häusser, M. Simultaneous all-optical manipulation and recording of neural circuit activity with cellular resolution in vivo. Nat. Methods 12, 140–146 (2015).

    Article  CAS  PubMed  Google Scholar 

  17. Adesnik, H. & Abdeladim, L. Probing neural codes with two-photon holographic optogenetics. Nat. Neurosci. 24, 1356–1366 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Russell, L. E. et al. All-optical interrogation of neural circuits in behaving mice. Nat. Protoc. 17, 1579–1620 https://doi.org/10.1038/s41596-022-00691-w (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Piantadosi Sean, C. et al. Holographic stimulation of opposing amygdala ensembles bidirectionally modulates valence-specific behavior via mutual inhibition. Neuron 112, 593–610.e5 https://doi.org/10.1016/j.neuron.2023.11.007 (2024).

    Article  CAS  PubMed  Google Scholar 

  20. Douglas, R. J. & Martin, K. A. C. Neuronal circuits of the neocortex. Annu. Rev. Neurosci. 27, 419–451 (2004).

    Article  CAS  PubMed  Google Scholar 

  21. Douglas, R. J. & Martin, K. A. C. Recurrent neuronal circuits in the neocortex. Curr. Biol. 17, R496–R500 (2007).

    Article  CAS  PubMed  Google Scholar 

  22. Harris, K. D. & Shepherd, G. M. G. The neocortical circuit: themes and variations. Nat. Neurosci. 18, 170–181 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Muñoz-Castañeda, R. et al. Cellular anatomy of the mouse primary motor cortex. Nature 598, 159–166 (2021).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

  24. Oldenburg, I. A. et al. The logic of recurrent circuits in the primary visual cortex. Nat. Neurosci. 27, 137–147 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Riehle, A. & Requin, J. The predictive value for performance speed of preparatory changes in neuronal activity of the monkey motor and premotor cortex. Behav. Brain Res. 53, 35–49 (1993).

    Article  CAS  PubMed  Google Scholar 

  26. Churchland, M. M., Afshar, A. & Shenoy, K. V. A central source of movement variability. Neuron 52, 1085–1096 (2006).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Churchland, M. M., Cunningham, J. P., Kaufman, M. T., Ryu, S. I. & Shenoy, K. V. Cortical preparatory activity: representation of movement or first cog in a dynamical machine? Neuron 68, 387–400 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Shenoy, K. V., Sahani, M. & Churchland, M. M. Cortical control of arm movements: a dynamical systems perspective. Annu. Rev. Neurosci. 36, 337–359 (2013).

    Article  CAS  PubMed  Google Scholar 

  29. Afshar, A. et al. Single-trial neural correlates of arm movement preparation. Neuron 71, 555–564 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Churchland, M. M. & Shenoy, K. V. Preparatory activity and the expansive null-space. Nat. Rev. Neurosci. 25, 213–236 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Vyas, S., Golub, M. D., Sussillo, D. & Shenoy, K. V. Computation through neural population dynamics. Annu. Rev. Neurosci. 43, 249–275 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Sauerbrei, B. A. et al. Cortical pattern generation during dexterous movement is input-driven. Nature 577, 386–391 (2020).

    Article  ADS  CAS  PubMed  Google Scholar 

  33. Yuste, R., MacLean, J. N., Smith, J. & Lansner, A. The cortex as a central pattern generator. Nat. Rev. Neurosci. 6, 477–483 (2005).

    Article  CAS  PubMed  Google Scholar 

  34. Pandarinath, C. et al. Latent factors and dynamics in motor cortex and their application to brain–machine interfaces. J. Neurosci. 38, 9390–9401 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. Long, M. A., Jin, D. Z. & Fee, M. S. Support for a synaptic chain model of neuronal sequence generation. Nature 468, 394–399 (2010).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  36. Moll, F. W. et al. Thalamus drives vocal onsets in the zebra finch courtship song. Nature 616, 132–136 (2023).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  37. Ramot, A. et al. Motor learning refines thalamic influence on motor cortex. Nature 643, 725–734 https://doi.org/10.1038/s41586-025-08962-8 (2025).

    Article  ADS  CAS  PubMed  Google Scholar 

  38. Tanaka, Y. H. et al. Thalamocortical axonal activity in motor cortex exhibits layer-specific dynamics during motor learning. Neuron 100, 244–258.e12 (2018).

    Article  CAS  PubMed  Google Scholar 

  39. Dacre, J. et al. A cerebellar-thalamocortical pathway drives behavioral context-dependent movement initiation. Neuron 109, 2326–2338.e8 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Inagaki, H. K. et al. A midbrain-thalamus-cortex circuit reorganizes cortical dynamics to initiate movement. Cell 185, 1065–1081.e23 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Takahashi, N. et al. Thalamic input to motor cortex facilitates goal-directed action initiation. Curr. Biol. 31, 4148–4155.e4 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Laje, R. & Buonomano, D. V. Robust timing and motor patterns by taming chaos in recurrent neural networks. Nat. Neurosci. 16, 925–933 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Bachschmid-Romano, L., Hatsopoulos, N. G. & Brunel, N. Interplay between external inputs and recurrent dynamics during movement preparation and execution in a network model of motor cortex. eLife 12, e77690 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Hennequin, G., Vogels, T. P. & Gerstner, W. Optimal control of transient dynamics in balanced networks supports generation of complex movements. Neuron 82, 1394–1406 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Gillett, M., Pereira, U. & Brunel, N. Characteristics of sequential activity in networks with temporally asymmetric Hebbian learning. Proc. Natl Acad. Sci. USA 117, 29948–29958 (2020).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  46. Weiler, N., Wood, L., Yu, J., Solla, S. A. & Shepherd, G. M. G. Top-down laminar organization of the excitatory network in motor cortex. Nat. Neurosci. 11, 360–366 (2008).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Pachitariu, M. et al. Suite2p: beyond 10,000 neurons with standard two-photon microscopy. Preprint at bioRxiv https://doi.org/10.1101/061507 (2017).

  48. Chen, Q. et al. Targeted stimulation of motor cortex ensembles drives learned movements. Zenodo https://doi.org/10.5281/zenodo.22216381 (2026).

Download references

Acknowledgements

We dedicate this manuscript to our friend and colleague, A.W., and acknowledge her surviving parents F. Wu and F. Qu. We thank B. Morales, A. Medina, D. Arakelyan and E. Hall for technical assistance and other Komiyama Lab members, especially J. Li and E. Gjoni, for their discussion. We also thank all our friends who have reached out to us to offer support during the difficult time.

Funding

This work was supported by the following funding sources: the NIH grants R01 NS125298, R01 NS091010, R01 DC018545 and R01 MH128746 (T.K.); the National Science Foundation grant 2024776 (T.K.); and the Simons Collaboration on the Global Brain Pilot Award (T.K.).

Author information

Author notes

  1. These authors contributed equally: Qiyu Chen, An Wu

  2. Deceased: An Wu

Authors and Affiliations

  1. Department of Neurobiology, University of California San Diego, La Jolla, CA, USA

    Qiyu Chen, An Wu, Bin Yu, Soyoung Chae, Zijing Tan, Assaf Ramot, Johnatan Aljadeff & Takaki Komiyama

  2. Center for Neural Circuits and Behavior, University of California San Diego, La Jolla, CA, USA

    Qiyu Chen, An Wu, Bin Yu, Soyoung Chae, Zijing Tan, Assaf Ramot, Johnatan Aljadeff & Takaki Komiyama

  3. Department of Neurosciences, University of California San Diego, La Jolla, CA, USA

    Qiyu Chen, An Wu, Bin Yu, Soyoung Chae, Zijing Tan, Assaf Ramot & Takaki Komiyama

  4. Halıcıoğlu Data Science Institute, University of California San Diego, La Jolla, CA, USA

    Takaki Komiyama

  5. Kavli Institute for Brain and Mind, University of California San Diego, La Jolla, CA, USA

    Takaki Komiyama

Authors

  1. Qiyu Chen
  2. An Wu
  3. Bin Yu
  4. Soyoung Chae
  5. Zijing Tan
  6. Assaf Ramot
  7. Johnatan Aljadeff
  8. Takaki Komiyama

Contributions

A.W. and T.K. conceptualized the study. A.W., Q.C., T.K. and J.A. devised the methodology. Q.C., A.W., B.Y., Z.T. and A.R. conducted the investigation. Q.C. helped with visualization. Q.C., A.W., and S.C. performed the formal analysis. Q.C. validated the study. A.W., Q.C. and T.K. helped with the resources. Q.C. and A.W. handled the software. Q.C. curated the data. T.K. helped with funding acquisition. T.K. was in charge of project administration. T.K. supervised the study. Q.C. and T.K. wrote the original draft. Q.C., T.K., J.A., B.Y., S.C., Z.T. and A.R. reviewed and edited the paper.

Corresponding author

Correspondence to Takaki Komiyama.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature thanks Kayvon Daie and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended Data Fig. 1 Additional information on experimental setup.

a. Example lever trajectories of rewarded trials in an expert session from one mouse. Grey, individual trials; black, mean. b. Reward rate, reaction time, cue onset to reward time, and movement stereotypy (median trial-to-trial correlation coefficients of lever movements for each session) at expert stage (N = 15 mice). Box and whisker plots: median and interquartile range. c. Left, example field of view of in vivo two-photon calcium imaging. Right, z-scored time series of estimated spikes from example neurons outlined in the image in the middle. This panel is representative of all 19 animals imaged in this study. d. Left, average population activity of all movement-active, movement-suppressed and non-movement-related neurons. Mean ± SEM across session-averages. Right, proportions of movement-active, movement-suppressed and non-movement-related neurons. N = 13874/25/15 (neurons/sessions/mice). e. Stimulation spatial accuracy. Left, schematic of experimental design. Stimulation locations were positioned at increasing distances from the target neuron while recording target neuron activity. Right, trial-averaged activity of target neurons at various centroid distances between target neuron and stimulation location. Box and whisker plots: median and interquartile range. N = 20/4/1 (targets/sessions/mice).

Source data

Extended Data Fig. 2 Movement trial classification.

a. Histograms of the maximum lever speed within 300 ms of trial onset for all trials, voluntary trials, early-target stimulation trials, late-target stimulation trials, and omission trials, respectively. Vertical dashed lines indicate the thresholds used to define movement trials. No-movement trials used in Fig. 5 were defined with a stricter criterion (Methods). Voluntary: N = 1199/25/15; early: N = 328/15/11; late: N = 237/10/8; omission: N = 204/25/15 (trials/sessions/mice).

Source data

Extended Data Fig. 3 Example late-target stimulation.

a. Left, example two-photon calcium imaging of three planes in M1 L2/3 spanning 60 µm in depth, showing target neurons (blue) and non-target neurons (beige). Right, trial-averaged responses of late-onset target and non-target neurons in opto trials aligned to stimulation onset. Red vertical bar indicates stimulation timing.

Source data

Extended Data Fig. 4 Additional movement quantification.

a. Left, average lever speed over the 0.9-sec period starting from cue/stimulation onset in movement trials in each trial type (voluntary vs early: p = 1.81 × 10-8, voluntary vs late: p = 1.64 × 10-14, voluntary vs. omission: p = 3.81 × 10-3, early vs late: p = 0.430, early vs. omission: p = 0.679, late vs. omission: p = 0.442 for trial type, mixed-effects model, two-sided). Right, total lever displacement in each trial type (voluntary vs early: p = 3.12 × 10-8, voluntary vs late: p = 1.66 × 10-14, voluntary vs. omission: p = 4.33 × 10-3, early vs late: p = 0.403, early vs. omission: p = 0.717, late vs. omission: p = 0.430 for trial type, mixed-effects model, two-sided). Voluntary: N = 729/25/15; early: N = 184/15/11; late: N = 107/10/8; omission: N = 42/25/15 (trials/sessions/mice). b. Left, correlation with the learned movement in each trial type (voluntary vs early: p = 5.24 × 10−12, voluntary vs late: p = 1.23 × 10−12, voluntary vs omission: p = 7.41 × 10−11, early vs late: p = 2.45 × 10−4, early vs omission: p = 0.003, late vs omission: p = 0.900 for trial type, mixed-effects model, two-sided). Right, reaction time of movements in each trial type (voluntary vs early: p = 1.10 × 10−11, voluntary vs late: p = 1.55 × 10−30, voluntary vs omission: p = 0.571, early vs late: p = 1.98 × 10−6, early vs omission: p = 0.034, late vs omission: p = 2.94 × 10−8 for trial type, mixed-effects model, two-sided). Voluntary: N = 489/25/25; early: N = 184/15/11; late: N = 107/10/8; omission: N = 42/25/15 (trials/sessions/mice). For this figure, box-and-whisker plots: median and interquartile range, + indicates mean. Same analysis as in Fig. 2c,d, except that omission trials are included for comparison.

Source data

Extended Data Fig. 5 Non-target early-onset neuron recruitment by optogenetic stimulation.

a. Left, fraction of responsive non-target early-onset neurons per trial for early-, late-target stimulation movement trials and late-target stimulation no-movement trials (early move. vs. late move.: p = 1.95 × 10−3; early move. vs. late no-move.: p = 5.40 × 10−8; late move. vs. late no-move.: p = 0.015 for trial type, mixed-effects model, two-sided). Right, response amplitudes of the responsive non-target early-onset neurons for early-, late-target stimulation movement trials and late-stimulation no-movement trials (early move. vs. late move.: p = 0.875; early move. vs. late no-move.: p = 0.013; late move. vs. late no-move.: p = 1.87 × 10−3 for trial type, mixed-effects model, two-sided). Early move.: N = 188/15/11; late move.: N = 107/10/8; late no-move.: N = 64/10/8 (trials/sessions/mice). Box and whisker plots: median and interquartile range, + indicates mean. This figure uses the same analysis procedure as Fig. 2e, except that only non-target early-onset neurons are examined, rather than all non-target neurons.

Source data

Extended Data Fig. 6 General properties of target neurons do not explain their distinct stimulation outcomes.

a. Left, fraction of early- or late-onset targets located in different imaging depths (p = 0.153 for group × location, mixed-effects model, two-sided). Middle, fraction of early- or late-onset targets located along the anterior-posterior axis within the field of view (p = 0.667 for group × location, mixed-effects model, two-sided). Right, fraction of early- or late-onset targets located along the medial-lateral axis (p = 0.813 for group × location, mixed-effects model, two-sided). Mean ± SEM. Early: N = 15/11; late: N = 10/8 (sessions/mice). b. Left, the distribution of nearest neighbor distances between early-onset targets and a shuffled control (p = 0.930 for data type, mixed-effects model, two-sided). Middle, same as left for late-onset (p = 0.133 for data type, mixed-effects model, two-sided). Right, Comparison between early- and late-onset targets (p = 0.262 for target type, mixed-effects model, two-sided). Early: N = 285/15/11; shuffled (early): N = 285000/15/11; late: N = 188/10/8; shuffled (late): N = 188000/10/8 (neurons/sessions/mice). c. Left, the peak activity amplitude of early- vs. late-onset targets during voluntary movements (p = 0.351 for target type, mixed-effects model, two-sided). Right, the response consistency of early- vs. late-onset targets during voluntary movements (p = 0.516 for target type, mixed-effects model, two-sided). Response consistency was measured as the fraction of movements during which a given neuron was active. Early: N = 285/15/11; late: N = 188/10/8 (neurons/sessions/mice). For this figure, box and whisker plots: median and interquartile range, + indicates mean.

Source data

Extended Data Fig. 7 Population activity including no-movement trials, with direct comparison to omission trials.

This figure is analogous to Figs. 1h,i and 2e,f, with the difference that all trials were included for each trial type, including no-movement trials. a. Activity of movement-active, movement-suppressed, and non-movement-related neurons (separated by horizontal dashed lines) averaged across learned movement trials, opto trials, and omission trials separately, from an example session. Each row represents a non-target neuron. b. Average activity of non-target neurons during the 1 sec following cue/stimulation onset sorted by amplitudes in learned movements from the same example session (opto-induced: p = 2.26 × 10−32, R = 0.518; omission: p = 0.125, R = −0.072. Pearson’s correlation, two-sided). Mean ± SEM. c. Leftmost, average activity of non-target neurons during learned movements versus in early-target stimulation trials pooled across sessions (p = 1.59 × 10−84 for learned activity, mixed-effects model, two-sided; Pearson’s correlation, R = 0.202). Each dot represents the activity of one neuron averaged during the 1 sec following cue/stimulation onset, averaged across trials. Center-left, average activity of non-target neurons during learned movements versus in late-target stimulation trials pooled across sessions (p = 2.44 × 10−20 for learned activity, mixed-effects model, two-sided; Pearson’s correlation, R = 0.127). Center-right, average activity of non-target neurons during learned movements versus in omission trials pooled across sessions (p = 0.075 for learned activity, mixed-effects model, two-sided; Pearson’s correlation, R = −0.018). Rightmost, correlation coefficients between non-target neuron population activity in learned movements and early-, late-target stimulation, and omission movements calculated for individual sessions (early vs. late: p = 0.081, early vs. omission: p = 5.10 × 10−5, late vs. omission: p = 4.15 × 10−3; Wilcoxon rank-sum test, two-sided). Early: N = 8789/15/11; late: N = 5252/10/8; omissions: N = 14041/25/15 (neurons/sessions/mice). d. Left, fraction of responsive non-target neurons per trial for early-target and late-target stimulation trials (p = 1.85 × 10−4 for trial type, mixed-effects model, two-sided). Right, response amplitudes of the responsive non-target neurons for early-target and late-target stimulation trials (p = 0.160 for trial type, mixed-effects model, two-sided). Early: N = 328/15/11; late: N = 237/10/8 (trials/sessions/mice). For this figure, box and whisker plots: median and interquartile range, + indicates mean.

Source data

Extended Data Fig. 8 Activity-movement relationship analysis including target neurons.

These are equivalent to Figs. 3a,b and 4a,b except that the analysis includes target neuron activity. a. Left, post-cue population activity distance to the learned activity negatively correlated with correlation with the learned movement in voluntary trials (p = 6.84 × 10−11 for activity distance, mixed-effects model; Pearson’s correlation, R = −0.357). Each dot represents an individual trial. Right, post-stimulation population activity distance to the learned activity negatively correlated with correlation with the learned movement in opto trials (p = 1.79 × 10−4 for activity distance, mixed-effects model; Pearson’s correlation, R = −0.271). Post-cue: N = 313/15/11; post-stimulation: N = 184/15/11 (pairs/sessions/mice). b. Left, pre-cue population activity distance to the learned activity negatively correlated with correlation with the learned movement in voluntary trials (p = 6.06 × 10−5 for activity distance, mixed-effects model; Pearson’s correlation, R = −0.224). Each dot represents an individual trial. Right, pre-stimulation population activity distance to the learned activity negatively correlated with correlation with the learned movement in opto trials (p = 0.017 for activity distance, mixed-effects model; Pearson’s correlation, R = −0.174). Pre-cue: N = 313/15/11; pre-stimulation: N = 184/15/11 (pairs/sessions/mice).

Source data

Extended Data Fig. 9 Initial state population activity predicts movement variability and movement probability.

In Fig. 4, population activity distance to learned activity in the pre-cue/stimulation period was smaller in voluntary trials compared to opto trials. This was because the population activity was measured in the principal component space defined by all voluntary trials. The current figure panels a-c are equivalent to Fig. 4 except that PCA was performed using a random subset (50 %) of only the learned movement trials (“learned activity”). a. Population activity distance to the learned activity before the trial start negatively correlated with correlation with the learned movement in opto trials (left, p = 0.016, mixed-effects model, activity distance; Pearson’s correlation, R = −0.177), and in voluntary trials (p = 2.54 × 10−6 for activity distance, mixed-effects model; Pearson’s correlation, R = −0.262). Each dot represents an individual trial. Opto: N = 184/15/11; voluntary: N = 313/15/11 (pairs/sessions/mice). b. Population activity distance to the learned activity before the trial start showed a correlation between pre- and post-stimulation periods in opto trials (left, p = 1.27 × 10−6 for pre-stimulation activity distance, mixed-effects model; Pearson’s correlation, R = 0.834) and between pre- and post-cue periods in voluntary trials (right, p = 3.51 × 10−4 for pre-cue activity distance, mixed-effects model; Pearson’s correlation, R = 0.850). Opto: N = 184/15/11; voluntary: N = 313/15/11 (pairs/sessions/mice). c. Population activity distance before the trial start negatively correlated with correlation of subsequent movements between voluntary trials (left, p = 9.40 × 10−5 for activity distance, mixed-effects model; Pearson’s correlation, R = −0.091), between opto trials (middle, p = 0.003 for activity distance, mixed-effects model; Pearson’s correlation, R = −0.103), and between voluntary and opto trials (right, p = 1.07 × 10−4 for activity distance, mixed-effects model; Pearson’s correlation, R = −0.083). Voluntary: N = 3413/15/11; opto: N = 1173/15/11; voluntary-opto: N = 4131/15/11 (pairs/sessions/mice). d. A Lasso logistic decoder trained on pre-cue population activity of voluntary trials was applied to opto and omission trials. Lines show separately fit mixed-effects curves relating initial state (decoder output) to movement probability. Initial state predicts movement probability in both opto and omission trials (Opto: p = 2.20 × 10−5, omission: p = 0.011 for initial state, mixed-effects model). However, the two curves significantly differ from each other (χ²(2) = 15.83, p = 3.65 × 10−4, likelihood ratio test). Dots and error bars show the mean ± SEM of the actual movement probability and initial state within quartile bins of initial state. Opto: N = 94/6/5; omission: N = 34/6/5 (trials/sessions/mice).

Source data

Extended Data Fig. 10 Population activity in non-target neurons in no-movement trials.

a. Average population activity of non-target neurons in early-target stimulation movement trials (left) and no-movement trials (right). Mean ± SEM pooled across sessions. b. Same as a but for late-target stimulation.

Supplementary information

Source data

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Chen, Q., Wu, A., Yu, B. et al. Targeted stimulation of motor cortex ensembles drives learned movements. Nature (2026). https://doi.org/10.1038/s41586-026-11174-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • Version of record:

  • DOI: https://doi.org/10.1038/s41586-026-11174-3