Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Article
  • Published:

Long-term stability of single neuron activity in the motor system

Nature Neuroscience volume 25, pages 1664–1674 (2022)Cite this article

Subjects

Abstract

How an established behavior is retained and consistently produced by a nervous system in constant flux remains a mystery. One possible solution to ensure long-term stability in motor output is to fix the activity patterns of single neurons in the relevant circuits. Alternatively, activity in single cells could drift over time provided that the population dynamics are constrained to produce the same behavior. To arbitrate between these possibilities, we recorded single-unit activity in motor cortex and striatum continuously for several weeks as rats performed stereotyped motor behaviors—both learned and innate. We found long-term stability in single neuron activity patterns across both brain regions. A small amount of drift in neural activity, observed over weeks of recording, could be explained by concomitant changes in task-irrelevant aspects of the behavior. These results suggest that long-term stable behaviors are generated by single neuron activity patterns that are themselves highly stable.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: A paradigm for interrogating long-term neural and behavioral stability.
Fig. 2: Analyzing neural stability in a recurrent network model.
Fig. 3: Experimental recordings of behavior and neural activity.
Fig. 4: Single neuron activity in DLS and MC is stable over time.
Fig. 5: Long-term drift of task-specific movement patterns in the lever-pressing task.
Fig. 6: Neural activity is stable during an innate behavior (WDS).

Similar content being viewed by others

Data availability

All data have been used previously by Dhawale et al.43. See https://github.com/KrisJensen/stability_paper_code for instructions on how to download the subset of data used for this paper.

Code availability

The code used to train all models, perform all analyses, and generate all figures is available online: https://github.com/KrisJensen/stability_paper_code.

References

  1. Krakauer, J. W. & Shadmehr, R. Consolidation of motor memory. Trends Neurosci. 29, 58–64 (2006).

    CAS  PubMed  Google Scholar 

  2. Melnick, M. J. Effects of overlearning on the retention of a gross motor skill. Res. Q. Am. Assoc. Health, Phys. Educ. Recreat. 42, 60–69 (1971).

    CAS  Google Scholar 

  3. Park, S.-W. & Sternad, D. Robust retention of individual sensorimotor skill after self-guided practice. J. Neurophysiol. 113, 2635–2645 (2015).

    PubMed  PubMed Central  Google Scholar 

  4. Churchland, M. M. et al. Neural population dynamics during reaching. Nature 487, 51–56 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  5. Haith, A. M. & Krakauer, J. W. in Progress in Motor Control, Vol. 782 (eds Richardson, M. J. et al.) 1–21 (Springer, 2013).

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

    CAS  PubMed  PubMed Central  Google Scholar 

  7. Park, S.-W., Dijkstra, T. & Sternad, D. Learning to never forget—time scales and specificity of long-term memory of a motor skill. Front. Comput. Neurosci. 7, 111 (2013).

    PubMed  PubMed Central  Google Scholar 

  8. Fu, M., Yu, X., Lu, J. & Zuo, Y. Repetitive motor learning induces coordinated formation of clustered dendritic spines in vivo. Nature 483, 92–95 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  9. Holtmaat, A. & Svoboda, K. Experience-dependent structural synaptic plasticity in the mammalian brain. Nat. Rev. Neurosci. 10, 647–658 (2009).

    CAS  PubMed  Google Scholar 

  10. Xu, T. et al. Rapid formation and selective stabilization of synapses for enduring motor memories. Nature 462, 915–919 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  11. Yang, G., Pan, F. & Gan, W.-B. Stably maintained dendritic spines are associated with lifelong memories. Nature 462, 920–924 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  12. Clopath, C., Bonhoeffer, T., Hübener, M. & Rose, T. Variance and invariance of neuronal long-term representations. Philos. Trans. R. Soc. Lond. B Biol. Sci. 372, 20160161 (2017).

    PubMed  PubMed Central  Google Scholar 

  13. Driscoll, L. N., Pettit, N. L., Minderer, M., Chettih, S. N. & Harvey, C. D. Dynamic reorganization of neuronal activity patterns in parietal cortex. Cell 170, 986–999.e16 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  14. Kargo, W. J. & Nitz, D. A. Improvements in the signal-to-noise ratio of motor cortex cells distinguish early versus late phases of motor skill learning. J. Neurosci. 24, 5560–5569 (2004).

    CAS  PubMed  PubMed Central  Google Scholar 

  15. Peters, A. J., Lee, J., Hedrick, N. G., O’Neil, K. & Komiyama, T. Reorganization of corticospinal output during motor learning. Nat. Neurosci. 20, 1133–1141 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  16. Schoonover, C. E., Ohashi, S. N., Axel, R. & Fink, A. J. P. Representational drift in primary olfactory cortex. Nature 594, 541–546 (2021).

    CAS  PubMed  Google Scholar 

  17. Rule, M. E., O’Leary, T. & Harvey, C. D. Causes and consequences of representational drift. Curr. Opin. Neurobiol. 58, 141–147 (2019).

    CAS  PubMed  PubMed Central  Google Scholar 

  18. Gallego, J. A., Perich, M. G., Miller, L. E. & Solla, S. A. Neural manifolds for the control of movement. Neuron 94, 978–984 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

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

    CAS  PubMed  PubMed Central  Google Scholar 

  20. Jensen, K., Stone, T.-C. & Hennequin, G. Scalable Bayesian GPFA with automatic relevance determination and discrete noise models. Adv. Neural Inf. Process. Syst. 34, 10613–10626 (2021).

    Google Scholar 

  21. Rokni, U., Richardson, A. G., Bizzi, E. & Seung, H. S. Motor learning with unstable neural representations. Neuron 54, 653–666 (2007).

    CAS  PubMed  Google Scholar 

  22. Qin, S. et al. Coordinated drift of receptive fields during noisy representation learning. Preprint at bioRxiv https://doi.org/10.1101/2021.08.30.458264 (2021).

  23. Duncker, L., Driscoll, L., Shenoy, K. V., Sahani, M. & Sussillo, D. In Proc. 34th Conference on Neural Information Processing Systems (eds Larochelle, H. et al.) 14387–14397 (NeurIPS, 2020).

  24. Kao, T.-C., Jensen, K., van de Ven, G., Bernacchia, A. & Hennequin, G. Natural continual learning: success is a journey, not (just) a destination. Adv. Neural Inf. Process. Syst. 34, 28067–28079 (2021).

    Google Scholar 

  25. Katlowitz, K. A., Picardo, M. A. & Long, M. A. Stable sequential activity underlying the maintenance of a precisely executed skilled behavior. Neuron 98, 1133–1140.e3 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

  26. Sizemore, M. & Perkel, D. J. Premotor synaptic plasticity limited to the critical period for song learning. Proc. Natl Acad. Sci. USA 108, 17492–17497 (2011).

    CAS  PubMed  PubMed Central  Google Scholar 

  27. Carmena, J. M., Lebedev, M. A., Henriquez, C. S. & Nicolelis, M. A. L. Stable ensemble performance with single-neuron variability during reaching movements in primates. J. Neurosci. 25, 10712–10716 (2005).

    CAS  PubMed  PubMed Central  Google Scholar 

  28. Chestek, C. A. et al. Single-neuron stability during repeated reaching in macaque premotor cortex. J. Neurosci. 27, 10742–10750 (2007).

    CAS  PubMed  PubMed Central  Google Scholar 

  29. Flint, R. D., Scheid, M. R., Wright, Z. A., Solla, S. A. & Slutzky, M. W. Long-term stability of motor cortical activity: implications for brain machine interfaces and optimal feedback control. J. Neurosci. 36, 3623–3632 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  30. Fraser, G. W. & Schwartz, A. B. Recording from the same neurons chronically in motor cortex. J. Neurophysiol. 107, 1970–1978 (2012).

    PubMed  Google Scholar 

  31. Ganguly, K. & Carmena, J. M. Emergence of a stable cortical map for neuroprosthetic control. PLoS Biol. 7, e1000153 (2009).

    PubMed  PubMed Central  Google Scholar 

  32. Liberti, W. A. et al. Unstable neurons underlie a stable learned behavior. Nat. Neurosci. 19, 1665–1671 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  33. Stevenson, I. H. et al. Statistical assessment of the stability of neural movement representations. J. Neurophysiol. 106, 764–774 (2011).

    PubMed  PubMed Central  Google Scholar 

  34. Sadeh, S. & Clopath, C. Contribution of behavioural variability to representational drift. eLife 11, e77907 (2022).

    PubMed  PubMed Central  Google Scholar 

  35. Willett, J. A. et al. The estrous cycle modulates rat caudate–putamen medium spiny neuron physiology. Eur. J. Neurosci. 52, 2737–2755 (2020).

    PubMed  Google Scholar 

  36. Miller, E. M., Shankar, M. U., Knutson, B. & McClure, S. M. Dissociating motivation from reward in human striatal activity. J. Cogn. Neurosci. 26, 1075–1084 (2014).

    PubMed  Google Scholar 

  37. Sheppard, P. A. S., Choleris, E. & Galea, L. A. M. Structural plasticity of the hippocampus in response to estrogens in female rodents. Mol. Brain 12, 22 (2019).

    PubMed  PubMed Central  Google Scholar 

  38. Gallego, J. A., Perich, M. G., Chowdhury, R. H., Solla, S. A. & Miller, L. E. Long-term stability of cortical population dynamics underlying consistent behavior. Nat. Neurosci. 23, 260–270 (2020).

    CAS  PubMed  PubMed Central  Google Scholar 

  39. Hunnicutt, B. J. et al. A comprehensive excitatory input map of the striatum reveals novel functional organization. eLife 5, e19103 (2016).

    PubMed  PubMed Central  Google Scholar 

  40. Musall, S., Kaufman, M. T., Juavinett, A. L., Gluf, S. & Churchland, A. K. Single-trial neural dynamics are dominated by richly varied movements. Nat. Neurosci. 22, 1677–1686 (2019).

    CAS  PubMed  PubMed Central  Google Scholar 

  41. Kingma, D. P. & Ba, J. Adam: a method for stochastic optimization. Preprint at https://arxiv.org/abs/1412.6980 (2014).

  42. Deitch, D., Rubin, A. & Ziv, Y. Representational drift in the mouse visual cortex. Curr. Biol. 31, 4327–4339.e6 (2021).

    CAS  PubMed  Google Scholar 

  43. Dhawale, A. K., Wolff, S. B. E., Ko, R. & Ölveczky, B. P. The basal ganglia control the detailed kinematics of learned motor skills. Nat. Neurosci. 24, 1256–1269 (2021).

    CAS  PubMed  Google Scholar 

  44. Poddar, R., Kawai, R. & Ölveczky, B. P. A fully automated high-throughput training system for rodents. PLoS One 8, e83171 (2013).

    PubMed  PubMed Central  Google Scholar 

  45. Dhawale, A. K. et al. Automated long-term recording and analysis of neural activity in behaving animals. eLife 6, e27702 (2017).

    PubMed  PubMed Central  Google Scholar 

  46. Kubota, Y. et al. Stable encoding of task structure coexists with flexible coding of task events in sensorimotor striatum. J. Neurophysiol. 102, 2142–2160 (2009).

    PubMed  PubMed Central  Google Scholar 

  47. Sheng, M., Lu, D., Shen, Z. & Poo, M. Emergence of stable striatal D1R and D2R neuronal ensembles with distinct firing sequence during motor learning. Proc Natl Acad. Sci. USA 116, 11038–11047 (2019).

    CAS  PubMed  PubMed Central  Google Scholar 

  48. Insafutdinov, E., Pishchulin, L., Andres, B., Andriluka, M., Schiele, B. in Computer Vision – ECCV 2016, Vol. 9910 (eds Leibe, B. et al.) 34–50 (Springer, 2016).

  49. Wolff, S. B. E., Ko, R. & Ölveczky, B. P. Distinct roles for motor cortical and thalamic inputs to striatum during motor learning and execution.Sci. Adv. 8, eabk0231 (2022).

    CAS  PubMed  PubMed Central  Google Scholar 

  50. Kanwal, J. K. et al. Internal state: dynamic, interconnected communication loops distributed across body, brain, and time. Integr. Comp. Biol. 61, 867–886 (2021).

    PubMed  PubMed Central  Google Scholar 

  51. Chaisanguanthum, K. S., Shen, H. H. & Sabes, P. N. Motor variability arises from a slow random walk in neural state. J. Neurosci. 34, 12071–12080 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  52. Churchland, M. M. Using the precision of the primate to study the origins of movement variability. Neuroscience 296, 92–100 (2015).

    CAS  PubMed  Google Scholar 

  53. Stevenson, I. H. Omitted variable bias in GLMs of neural spiking activity. Neural Comput. 30, 3227–3258 (2018).

    PubMed  Google Scholar 

  54. Mehler, D. M. A. & Kording, K. P. The lure of misleading causal statements in functional connectivity research. Preprint at https://arxiv.org/abs/1812.03363 (2020).

  55. Pillow, J. W. et al. Spatio-temporal correlations and visual signalling in a complete neuronal population. Nature 454, 995–999 (2008).

    CAS  PubMed  PubMed Central  Google Scholar 

  56. Fletcher, A. & Harding, V. An examination of the ‘wet dog’ shake behaviour in rats produced by acute administration of sodium n-dipropylacetate. J. Pharm. Pharmacol. 33, 811–813 (1981).

    CAS  PubMed  Google Scholar 

  57. Marshall, J. D. et al. Continuous whole-body 3D kinematic recordings across the rodent behavioral repertoire. Neuron 109, 420–437.e8 (2021).

    CAS  PubMed  Google Scholar 

  58. Bedard, P. & Pycock, C. J. ‘Wet-dog’ shake behaviour in the rat: a possible quantitative model of central 5-hydroxytryptamine activity. Neuropharmacology 16, 663–670 (1977).

    CAS  PubMed  Google Scholar 

  59. Hatsopoulos, N. G. & Suminski, A. J. Sensing with the motor cortex. Neuron 72, 477–487 (2011).

    CAS  PubMed  PubMed Central  Google Scholar 

  60. Zagha, E. et al. The importance of accounting for movement when relating neuronal activity to sensory and cognitive processes. J. Neurosci. 42, 1375–1382 (2022).

    CAS  PubMed  PubMed Central  Google Scholar 

  61. Srinivasan, M. & Ruina, A. Computer optimization of a minimal biped model discovers walking and running. Nature 439, 72–75 (2006).

    CAS  PubMed  Google Scholar 

  62. Todorov, E. & Jordan, M. I. Optimal feedback control as a theory of motor coordination. Nat. Neurosci. 5, 1226–1235 (2002).

    CAS  PubMed  Google Scholar 

  63. Dhawale, A. K., Miyamoto, Y. R., Smith, M. A. & Ölveczky, B. P. Adaptive regulation of motor variability. Curr. Biol. 29, 3551–3562.e7 (2019).

    CAS  PubMed  PubMed Central  Google Scholar 

  64. Wu, H. G., Miyamoto, Y. R., Castro, L. N. G., Ölveczky, B. P. & Smith, M. A. Temporal structure of motor variability is dynamically regulated and predicts motor learning ability. Nat. Neurosci. 17, 312–321 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  65. Rule, M. E. & O’Leary, T. Self-healing codes: how stable neural populations can track continually reconfiguring neural representations. Proc. Natl Acad. Sci. USA 119, e2106692119 (2022).

    CAS  PubMed  PubMed Central  Google Scholar 

  66. Lütcke, H., Margolis, D. J. & Helmchen, F. Steady or changing? Long-term monitoring of neuronal population activity. Trends Neurosci. 36, 375–384 (2013).

    PubMed  Google Scholar 

  67. Huang, L. et al. Relationship between simultaneously recorded spiking activity and fluorescence signal in GCaMP6 transgenic mice. eLife 10, e51675 (2021).

    CAS  PubMed  PubMed Central  Google Scholar 

  68. Mante, V., Sussillo, D., Shenoy, K. V. & Newsome, W. T. Context-dependent computation by recurrent dynamics in prefrontal cortex. Nature 503, 78–84 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  69. Rule, M. E. et al. Stable task information from an unstable neural population. eLife 9, e51121 (2020).

    CAS  PubMed  PubMed Central  Google Scholar 

  70. Otchy, T. M. et al. Acute off-target effects of neural circuit manipulations. Nature 528, 358–363 (2015).

    CAS  PubMed  Google Scholar 

  71. Margolis, D. J. et al. Reorganization of cortical population activity imaged throughout long-term sensory deprivation. Nat. Neurosci. 15, 1539–1546 (2012).

    CAS  PubMed  Google Scholar 

  72. Hengen, K. B., Lambo, M. E., Van Hooser, S. D., Katz, D. B. & Turrigiano, G. G. Firing rate homeostasis in visual cortex of freely behaving rodents. Neuron 80, 335–342 (2013).

    CAS  PubMed  Google Scholar 

  73. Mrsic-Flogel, T. D. et al. Homeostatic regulation of eye-specific responses in visual cortex during ocular dominance plasticity. Neuron 54, 961–972 (2007).

    CAS  PubMed  Google Scholar 

  74. Golowasch, J., Casey, M., Abbott, L. F. & Marder, E. Network stability from activity-dependent regulation of neuronal conductances. Neural Comput. 11, 1079–1096 (1999).

    CAS  PubMed  Google Scholar 

  75. Marder, E. & Goaillard, J.-M. Variability, compensation and homeostasis in neuron and network function. Nat. Rev. Neurosci. 7, 563–574 (2006).

    CAS  PubMed  Google Scholar 

  76. Kirkpatrick, J. et al. Overcoming catastrophic forgetting in neural networks. Proc. Natl Acad. Sci. USA 114, 3521–3526 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  77. Kao, T.-C., Sadabadi, M. S. & Hennequin, G. Optimal anticipatory control as a theory of motor preparation: a thalamo-cortical circuit model. Neuron 109, 1567–1581.e12 (2021).

    CAS  PubMed  PubMed Central  Google Scholar 

  78. Rodriguez, A. & Laio, A. Clustering by fast search and find of density peaks. Science 344, 1492–1496 (2014).

    CAS  PubMed  Google Scholar 

  79. Williams, A. H. et al. Discovering precise temporal patterns in large-scale neural recordings through robust and interpretable time warping. Neuron 105, 246–259.e8 (2020).

    CAS  PubMed  Google Scholar 

  80. Keshtkaran, M. R. et al. A large-scale neural network training framework for generalized estimation of single-trial population dynamics. Preprint at bioRxiv https://doi.org/10.1101/2021.01.13.426570 (2021).

Download references

Acknowledgements

We are grateful to K. Hardcastle, C. Pehlevan, T.-C. Kao, G. Hennequin and M. Schimel for their feedback on the manuscript. This work was supported by a Gates Cambridge scholarship and Nordea-fonden (K.T.J.); a Helen Hay Whitney Foundation Postdoctoral Fellowship, the Zuckerman STEM Leadership Program postdoctoral fellowship, and the Weizmann Institute of Science - National Postdoctoral Award for Advancing Women in Science (N.K.H.); a Life Sciences Research Foundation and Charles A. Kings Foundation postdoctoral fellowship (A.K.D.); an EMBO postdoctoral fellowship ALTF1561-2013 and an HFSP postdoctoral fellowship LT 000514/2014 (S.B.E.W.); and National Institutes of Health grants R01-NS0993231 and R01-NS105349 (B.P.Ö.).

Author information

Author notes

  1. Ashesh K. Dhawale

    Present address: Centre for Neuroscience, Indian Institute of Science, Bangalore, India

  2. Steffen B. E. Wolff

    Present address: Department of Pharmacology, University of Maryland School of Medicine, Baltimore, MD, USA

Authors and Affiliations

  1. Department of Organismic and Evolutionary Biology and Center for Brain Science, Harvard University, Cambridge, MA, USA

    Kristopher T. Jensen, Naama Kadmon Harpaz, Ashesh K. Dhawale, Steffen B. E. Wolff & Bence P. Ölveczky

  2. Computational and Biological Learning Lab, Department of Engineering, University of Cambridge, Cambridge, UK

    Kristopher T. Jensen

Authors
  1. Kristopher T. Jensen
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Naama Kadmon Harpaz
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Ashesh K. Dhawale
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Steffen B. E. Wolff
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Bence P. Ölveczky
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

B.P.Ö., K.T.J., A.K.D. and N.K.H. conceived the study. A.K.D. and S.B.E.W. collected the data. K.T.J. and N.K.H. analyzed the data. K.T.J., N.K.H. and B.P.Ö. interpreted the data and wrote the manuscript. All authors reviewed the manuscript.

Corresponding author

Correspondence to Bence P. Ölveczky.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Neuroscience thanks the anonymous reviewers for their contribution to the peer review of this work.

Additional information

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

Extended data

Extended Data Fig. 1 RNN parameter interpolation.

(a) Mean correlation between the initial conditions (left), recurrent weight matrices (center), and readout weight matrices (right) of the simulated RNNs as a function of time difference for the stable and drifting networks. Shading indicates standard deviation across 10 repetitions of training and interpolation.

Extended Data Fig. 2 Kinematics of all animals.

(a) Heatmaps showing the forelimb trajectories of each animal on every trial across all days. x-axes indicate time within trial and y-axes indicate trial number from first (top) to last (bottom). Each column corresponds to a single animal (first three: DLS, last three: MC). The rows illustrate the trajectories of the right forelimb parallel and perpendicular to the floor, followed by the left forelimb parallel and perpendicular to the floor. The second animal from the left corresponds to the example used in Figs. 3a, 5a and b. (b) Heatmaps showing the z-scored velocity of each animal on every trial across all days for the animals in a. The rows illustrate the velocity of the right forelimb parallel and perpendicular to the floor followed by the left forelimb parallel and perpendicular to the floor.

Extended Data Fig. 3 Similarity as a function of time difference for all neurons.

We computed the PETH correlation as a function of time difference for all neurons, taking the average across all pairs of days separated by the same time difference for each neuron. This figure shows the average similarity as a function of time difference for neurons recorded in DLS (left) or MC (right) during the lever-pressing task (a) and the wet-dog shake behavior (b). Upper panels indicate all neurons recorded for at least 3 days, lower panels indicate neurons which were recorded for at least 14 (a) or 10 (b) days and therefore included in Fig. 4c or 6c. Neurons were sorted by recording duration.

Extended Data Fig. 4 Stability as a function of time difference for different recording durations.

(a) We performed analyses as in Fig. 4c, plotting the neural similarity as a function of time difference for neurons recorded for at least N days, with N ranging from 5 to 17 (c.f. N = 14 in Fig. 4c). Error bars indicate standard error across units, and dashed lines indicate controls as in Fig. 4c. (b) As in a, now for the wet-dog shake behavior instead of the lever-pressing task (c.f. Fig. 6c).

Extended Data Fig. 5 Latent stability and neural decoding.

It has previously been reported that stable neural activity can be identified in a common latent space even when there is a turnover of recorded neurons43. As we show in Fig. 2e, this can be consistent with either stable or drifting single unit activity. While we have already shown a high degree of similarity for single neurons, here we investigate whether ‘aligning’ the neural activity between sessions can identify a common subspace with even higher similarity. These analyses require simultaneous recording of a large population of neurons, which in general was not the case in our dataset (c.f. Fig. 3c). Instead, we considered a single week of recording in a single animal with recordings from DLS (day 8-14 in Fig. 3c), where we simultaneously recorded 16 neurons firing at least 10 spikes during the task on each day. (a) We first computed the similarity as a function of time difference as the correlation between single neuron PETHs, averaged across neurons (black line). We then proceeded to align the neural activity on each pair of days using CCA and computed the similarity in the resulting aligned space as the average correlation across all dimensions. This CCA-aligned similarity was generally lower than the similarity averaged over individual neurons, suggesting that the neuron-aligned coordinate system is more stable than the CCA-aligned alternative (note that CCA performs a greedy alignment rather than finding the optimal alignment, which would provide an upper bound on the single neuron similarity). Shadings indicate standard error across all pairs of days with a given time difference. (b) We proceeded to consider population decoding of behavior from neural activity, using the same data as in a. We fitted a linear model to predict the trajectories of the left and right forelimbs from neural activity on each day using crossvalidated ridge regression, and we tested the models on data from all other days. Here, we plot the performance as a function of time difference, averaged across the vertical and horizontal dimensions and both forelimbs. Line and shading indicate mean and standard error across pairs of days with a given time difference. (c) We proceeded to compute stability indices for the data in b to see whether there was a significant negative trend. We bootstrapped the individual datapoints (before taking the mean) 10,000 times and estimated stability indices from each surrogate dataset. The distribution over the resulting stability indices was not significantly smaller than 0 (one-sided p = 0.48). (d) While the analysis in a suggests that the single neurons provide a good coordinate system for stable representations, it does not address the question of whether an aligned low-dimensional manifold can provide better decoding43. We therefore proceeded to train a population decoding model as in b, but where the decoder was trained on the top 10 PCs from a single day and tested on the top 10 PCs from every other day after alignment via CCA43 (blue dashed line). We found that decoding performance from this aligned latent space was almost identical to the decoding performance from raw neural activity (black line). This provides further evidence that the stable aligned dynamics identified in previous work are the result of stable single unit tuning curves. Shading indicates standard error across pairs of days with a given time difference. (e) Finally, we considered how the relationship between kinematics and neural activity changed over time at a single neuron level. We used the GLM discussed in Fig. 5e to predict neural activity from behavior. This GLM was trained on the first day of recording for each neuron and tested on each subsequent day. The figure shows the correlation between the predicted firing rate and true spike count as a function of time difference, averaged across all neurons which were recorded for at least a week and had a training correlation of at least 0.1. Blue indicates neurons recorded from DLS (n = 58 units), red from MC (n = 61 units), and shadings indicate standard errors across neurons. Dashed lines indicate the average correlation across neurons from hold-one-out crossvalidation on all trials from the first day of recording.

Extended Data Fig. 6 Exponential model fits and stability indices.

(a) Plots of PETH similarity against time difference for four example units (colors) together with exponential fits illustrating a range of different decay rates, same-day similarities, and durations of recording. Note that one of these example units (cyan) exhibits an apparent increase in stability over time due to the noisy nature of the data. Indeed, in a perfectly stable model (such as the stable RNN in Fig. 2e), neurons will be as likely to exhibit such an increase as they are to exhibit a decrease in similarity over time, leading to a median stability index of 0. Such noise is mitigated by increasing recording durations. (b) Distribution of the mean error of each model fit across the population of neurons recorded from MC (red) or DLS (blue). Vertical dashed lines indicate quartiles of the distributions. (c) Stability indices for all neurons recorded from DLS (left; blue) or MC (right; red) during the lever-pressing task. Solid lines indicate exponential fits as in Fig. 4e. As the time difference increases, the variance decreases (due to the increase in data), and the median stability index gradually increases (c.f. solid lines). (d) As in c, for the WDS behavior.

Extended Data Fig. 7 Results are not dependent on time-warping.

In this figure, we reproduce some of the key analyses of the paper after aligning trials by ‘trimming’ to a fixed duration rather than the time-warping used in the main text. (a) Neural similarity as a function of time difference for neurons recorded for at least 14 days in the lever-pressing task in either DLS (blue) or motor cortex (red). Note the similarity with Fig. 4c using time-warping. Lines and shading indicate mean and standard error across units. (b) Kinematic similarity in the lever-pressing task as a function of time difference across all animals. Solid line and shading indicate mean and standard error across animals after trimming. Dashed line indicates the mean after time-warping. Note that time-warping better aligns kinematics, which is the primary motivation for its use in the main text. (c) correlation between neural similarity and kinematic similarity on consecutive days (c.f. Fig. 5d). (d-f) As in a-c, now for the wet-dog shake behavior.

Extended Data Fig. 8 Exponential fits for different subsampled recording durations.

Black lines indicate the mean across units of the neural similarity as a function of time difference for units recorded for at least 14 days during the lever-pressing task (c.f. Fig. 4c). We fitted exponential models to the mean data, considering only data up to and including increasing time differences (legend). As the subsampled ‘recording duration’ increases, so does the stability index learned in the exponential model for both neurons recorded in DLS (a) and MC (b). If the observed increase in stability with recording duration is due to latent processes with autocorrelations on the order of days, we would expect the neural similarity to decrease to some saturating baseline value, γ. We therefore also fitted a model to the average similarity across neurons as a function of time difference, which assumes a decay to such a baseline (\(\rho = \beta e^{\alpha \,\delta t} + \gamma\); red dashed lines). This model yielded an asymptotic correlation of γ = 0.71 for DLS and γ = 0.58 for MC, suggesting a high degree of neural similarity at long timescales.

Extended Data Fig. 9 Behavioral drift and inter-press intervals.

(a) Correlations between mean velocity profiles plotted against time difference for all pairs of days in each animal. Top row: lever-pressing task; bottom row: wet dog shakes. Blue indicates animals with recordings from DLS, red from MC. (b) Distribution of correlations between time difference and behavioral similarity across all animals, generated by a bootstrap analysis of the data in a. All animals exhibit a significant negative correlation between behavioral similarity and time difference in both the lever-pressing task and wet dog shake behavior (p < 0.001; one-sided bootstrap test). (c) Inter-press interval (IPI) for each animal, convolved with a 200-trial Gaussian filter. Time is normalized from 0 to 1 for each animal (n = 9365 ± 6886 trials, mean ± std). Black horizontal line indicates 700 ms. (d) We computed the IPI autocorrelation as a function of trial number and normalized time by the average number of trials per day for each animal (colored lines). Black line and shading indicate mean and standard error across animals. Task performance is only correlated over short timescales of 0.5-1 days despite behavioral drift on timescales of weeks (c.f. panel a). This suggests that behavioral changes are predominantly along ‘task-null’ directions that do not affect performance.

Extended Data Fig. 10 Task-modulation of neurons in the lever-pressing task and wet-dog shake behavior.

(a) A PETH was computed across all trials for each neuron in 20 ms bins, and the time bin identified with the maximum deviation from the mean across all time bins. The corresponding z-score was computed, and the distribution of absolute values of these z-scores plotted across all DLS neurons for the lever-pressing task (black) and wet-dog shake behavior (blue). (b) As in a, now for neurons recorded from MC.

Supplementary information

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Jensen, K.T., Kadmon Harpaz, N., Dhawale, A.K. et al. Long-term stability of single neuron activity in the motor system. Nat Neurosci 25, 1664–1674 (2022). https://doi.org/10.1038/s41593-022-01194-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41593-022-01194-3

This article is cited by

Search

Advanced search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing