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:

Learning binds new inputs into functional synaptic clusters via spinogenesis

Nature Neuroscience volume 25, pages 726–737 (2022)Cite this article

Subjects

Abstract

Learning induces the formation of new excitatory synapses in the form of dendritic spines, but their functional properties remain unknown. Here, using longitudinal in vivo two-photon imaging and correlated electron microscopy of dendritic spines in the motor cortex of mice during motor learning, we describe a framework for the formation, survival and resulting function of new, learning-related spines. Specifically, our data indicate that the formation of new spines during learning is guided by the potentiation of functionally clustered preexisting spines exhibiting task-related activity during earlier sessions of learning. We present evidence that this clustered potentiation induces the local outgrowth of multiple filopodia from the nearby dendrite, locally sampling the adjacent neuropil for potential axonal partners, likely via targeting preexisting presynaptic boutons. Successful connections are then selected for survival based on co-activity with nearby task-related spines, ensuring that the new spine preserves functional clustering. The resulting locally coherent activity of new spines signals the learned movement. Furthermore, we found that a majority of new spines synapse with axons previously unrepresented in these dendritic domains. Thus, learning involves the binding of new information streams into functional synaptic clusters to subserve learned behaviors.

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: Characterization of movement-related signals in dendritic spines in M1 during motor learning.
Fig. 2: iGluSnFR fluorescence signals are dependent on glutamate sensitivity.
Fig. 3: New spines form near enlarged, movement-related spines in the motor cortex during motor learning.
Fig. 4: Functional clusters of new spines and movement-related spines preferentially encode learned movements.
Fig. 5: Correlated light and electron microscopy reveals patterns of microstructures surrounding functional clusters.
Fig. 6: Most new spines represent novel connections on the dendritic segment.
Fig. 7: Proposed model of locally facilitated spinogenesis during motor learning.

Similar content being viewed by others

Data availability

Data are available upon request from the corresponding authors.

Code availability

Code used to analyze data and generate figures for this paper are available upon request from the corresponding authors.

References

  1. Makino, H., Hwang, E. J., Hedrick, N. G. & Komiyama, T. Circuit mechanisms of sensorimotor learning. Neuron https://doi.org/10.1016/j.neuron.2016.10.029 (2016).

  2. Hofer, S. B., Mrsic-Flogel, T. D., Bonhoeffer, T. & Hübener, M. Experience leaves a lasting structural trace in cortical circuits. Nature 457, 313–317 (2009).

    Article  CAS  PubMed  Google Scholar 

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

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. 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).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Iacaruso, M. F., Gasler, I. T. & Hofer, S. B. Synaptic organization of visual space in primary visual cortex. Nature https://doi.org/10.1038/nature23019 (2017).

  6. Wilson, D. E., Whitney, D. E., Scholl, B. & Fitzpatrick, D. Orientation selectivity and the functional clustering of synaptic inputs in primary visual cortex. Nat. Neurosci. 19, 1003–1009 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Scholl, B., Wilson, D. E. & Fitzpatrick, D. Local order within global disorder: synaptic architecture of visual space. Neuron https://doi.org/10.1016/j.neuron.2017.10.017 (2017).

  8. Kerlin, A. et al. Functional clustering of dendritic activity during decision-making. Elife https://doi.org/10.7554/eLife.46966 (2019).

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

    Article  CAS  PubMed  Google Scholar 

  10. 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 

  11. Marvin, J. S. et al. An optimized fluorescent probe for visualizing glutamate neurotransmission. Nat. Methods https://doi.org/10.1038/nmeth.2333 (2013).

  12. Marvin, J. S. et al. Stability, affinity, and chromatic variants of the glutamate sensor iGluSnFR. Nat. Methods https://doi.org/10.1038/s41592-018-0171-3 (2018).

  13. Hires, S. A., Zhu, Y. & Tsien, R. Y. Optical measurement of synaptic glutamate spillover and reuptake by linker optimized glutamate-sensitive fluorescent reporters. Proc. Natl. Acad. Sci. USA https://doi.org/10.1073/pnas.0712008105 (2008).

  14. Chen, T.-W. et al. Ultrasensitive fluorescent proteins for imaging neuronal activity. Nature 499, 295–300 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Kerlin, A. et al. Functional clustering of dendritic activity during decision-making. eLife https://doi.org/10.7554/eLife.46966 (2018).

  16. 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).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Takahashi, N. et al. Locally synchronized synaptic inputs. Science 335, 353–356 (2012).

    Article  CAS  PubMed  Google Scholar 

  18. De Roo, M., Klauser, P. & Muller, D. LTP promotes a selective long-term stabilization and clustering of dendritic spines. PLoS Biol. https://doi.org/10.1371/journal.pbio.0060219 (2008).

  19. Kastellakis, G., Cai, D. J., Mednick, S. C., Silva, A. J. & Poirazi, P. Synaptic clustering within dendrites: an emerging theory of memory formation. Prog. Neurobiol. 126, 19–35 (2015).

  20. Kleindienst, T., Winnubst, J., Roth-Alpermann, C., Bonhoeffer, T. & Lohmann, C. Activity-dependent clustering of functional synaptic inputs on developing hippocampal dendrites. Neuron https://doi.org/10.1016/j.neuron.2011.10.015 (2011).

  21. Makino, H. & Malinow, R. AMPA receptor incorporation into synapses during LTP: the role of lateral movement and exocytosis. Neuron 64, 381–390 (2009).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Makino, H. & Malinow, R. Compartmentalized versus global synaptic plasticity on dendrites controlled by experience. Neuron https://doi.org/10.1016/j.neuron.2011.09.036 (2011).

  23. Hedrick, N. et al. Rho GTPase complementation underlies BDNF-dependent homo- and heterosynaptic plasticity. Nature 538, 104–108 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Kwon, H.-B. & Sabatini, B. L. Glutamate induces de novo growth of functional spines in developing cortex. Nature 474, 100–104 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Matsuzaki, M., Honkura, N., Ellis-Davies, G. C. R. & Kasai, H. Structural basis of long-term potentiation in single dendritic spines. Nature 429, 761–766 (2004).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Cichon, J. & Gan, W.-B. Branch-specific dendritic Ca2+ spikes cause persistent synaptic plasticity. Nature 520, 180–185 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Murakoshi, H., Wang, H. & Yasuda, R. Local, persistent activation of Rho GTPases during plasticity of single dendritic spines. Nature 472, 100–104 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Spacek, J. & Harris, K. M. Three-dimensional organization of smooth endoplasmic reticulum in hippocampal CA1 dendrites and dendritic spines of the immature and mature rat. J. Neurosci. https://doi.org/10.1523/jneurosci.17-01-00190.1997 (1997).

  29. Deller, T. et al. Synaptopodin-deficient mice lack a spine apparatus and show deficits in synaptic plasticity. Proc. Natl. Acad. Sci. USA https://doi.org/10.1073/pnas.1832384100 (2003).

  30. Hwang, E. J. et al. Disengagement of motor cortex from movement control during long-term learning. Sci. Adv. https://doi.org/10.1126/sciadv.aay0001 (2019).

  31. Conner, J. M., Culberson, A., Packowski, C., Chiba, A. A. & Tuszynski, M. H. Lesions of the basal forebrain cholinergic system impair task acquisition and abolish cortical plasticity associated with motor skill learning. Neuron 38, 819–829 (2003).

    Article  CAS  PubMed  Google Scholar 

  32. 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 

  33. Losonczy, A. & Magee, J. C. Integrative properties of radial oblique dendrites in hippocampal CA1 pyramidal neurons. Neuron 50, 291–307 (2006).

    Article  CAS  PubMed  Google Scholar 

  34. Niell, C. M., Meyer, M. P. & Smith, S. J. In vivo imaging of synapse formation on a growing dendritic arbor. Nat. Neurosci. https://doi.org/10.1038/nn1191 (2004).

  35. Ziv, N. E. & Smith, S. J. Evidence for a role of dendritic filopodia in synaptogenesis and spine formation. Neuron https://doi.org/10.1016/S0896-6273(00)80283-4 (1996).

  36. Toni, N. et al. Synapse formation on neurons born in the adult hippocampus. Nat. Neurosci. https://doi.org/10.1038/nn1908 (2007).

  37. Dailey, M. E. & Smith, S. J. The dynamics of dendritic structure in developing hippocampal slices. J. Neurosci. https://doi.org/10.1523/jneurosci.16-09-02983.1996 (1996).

  38. Lendvai, B., Stern, E. A., Chen, B. & Svoboda, K. Experience-dependent plasticity of dendritic spines in the developing rat barrel cortex in vivo. Nature 404, 876–881 (2000).

    Article  CAS  PubMed  Google Scholar 

  39. Zuo, Y., Lin, A., Chang, P. & Gan, W. -B. Development of long-term dendritic spine stability in diverse regions of cerebral cortex. Neuron 46, 181–189 (2005).

  40. Cruz-Martín, A., Crespo, M. & Portera-Cailliau, C. Delayed stabilization of dendritic spines in fragile X mice. J. Neurosci. 30, 7793–7803 (2010).

    Article  PubMed  PubMed Central  Google Scholar 

  41. Holtmaat, A. J. G. D. et al. Transient and persistent dendritic spines in the neocortex in vivo. Neuron 45, 279–291 (2005).

    Article  CAS  PubMed  Google Scholar 

  42. Frank, A. C. et al. Hotspots of dendritic spine turnover facilitate clustered spine addition and learning and memory. Nat. Commun. https://doi.org/10.1038/s41467-017-02751-2 (2018).

  43. Knott, G. W., Holtmaat, A., Wilbrecht, L., Welker, E. & Svoboda, K. Spine growth precedes synapse formation in the adult neocortex in vivo. Nat. Neurosci. 9, 1117–1124 (2006).

    Article  CAS  PubMed  Google Scholar 

  44. Yang, Y. et al. Selective synaptic remodeling of amygdalocortical connections associated with fear memory. Nat. Neurosci. https://doi.org/10.1038/nn.4370 (2016).

  45. Lee, K. J. et al. Motor skill training induces coordinated strengthening and weakening between neighboring synapses. J. Neurosci. https://doi.org/10.1523/JNEUROSCI.0848-12.2013 (2013).

  46. Toni, N., Buchs, P. A., Nikonenko, I., Bron, C. R. & Muller, D. LTP promotes formation of multiple spine synapses between a single axon terminal and a dendrite. Nature https://doi.org/10.1038/46574 (1999).

  47. Niculescu, D. et al. A BDNF-mediated push–pull plasticity mechanism for synaptic clustering. Cell Rep. https://doi.org/10.1016/j.celrep.2018.07.073 (2018).

  48. Bloss, E. B. et al. Single excitatory axons form clustered synapses onto CA1 pyramidal cell dendrites. Nat. Neurosci. https://doi.org/10.1038/s41593-018-0084-6 (2018).

  49. Chicurel, M. E. & Harris, K. M. Three‐dimensional analysis of the structure and composition of CA3 branched dendritic spines and their synaptic relationships with mossy fiber boutons in the rat hippocampus. J. Comp. Neurol. https://doi.org/10.1002/cne.903250204 (1992).

  50. Kasthuri, N. et al. Saturated reconstruction of a volume of nocortex. Cell https://doi.org/10.1016/j.cell.2015.06.054 (2015).

  51. Govindarajan, A., Kelleher, R. J. & Tonegawa, S. A clustered plasticity model of long-term memory engrams. Nat. Rev. Neurosci. 7, 575–583 (2006).

    Article  CAS  PubMed  Google Scholar 

  52. Poirazi, P. & Mel, B. W. Impact of active dendrites and structural plasticity on the memory capacity of neural tissue. Neuron 29, 779–796 (2001).

    Article  CAS  PubMed  Google Scholar 

  53. Tjia, M., Yu, X., Jammu, L. S., Lu, J. & Zuo, Y. Pyramidal neurons in different cortical layers exhibit distinct dynamics and plasticity of apical dendritic spines. Front. Neural Circuits 11, 43 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

  54. Mitani, A. & Komiyama, T. Real-time processing of two-photon calcium imaging data including lateral motion artifact correction. Front. Neuroinform. https://doi.org/10.3389/fninf.2018.00098 (2018).

  55. Kremer, J. R., Mastronarde, D. N. & McIntosh, J. R. R. Computer visualization of three-dimensional image data using IMOD. J. Struct. Biol. 116, 71–76 (1996).

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

We thank K. O’Neil, Q. Chen, O. Arroyo, L. Hall, P. Yao and E. Hall for technical assistance; M. Banghart, B. Bloodgood and members of the laboratory of T.K., especially A. Ramot, for discussion; and E. Gjoni for assistance with EM analysis. This research was supported by grants from NIH (R01 NS091010, R21 NS112750, R21 NS109722, R01 DC014690, R01 EY025349 and P30 EY022589), Pew Charitable Trusts and David & Lucile Packard Foundation to T.K.; NIH (F32 NS103267 and K99 NS114175) to N.G.H.; U24 NS120055, R01 DA038896, R24 GM137200 and R01 GM082949 to M.E.; and U01 MH114829, R01 DA049787 and R01 NS121231 to B.K.L.

Author information

Authors and Affiliations

  1. Neurobiology Section, Division of Biological Sciences, University of California, San Diego, La Jolla, CA, USA

    Nathan G. Hedrick, Zhongmin Lu, Surbhi Singhi, Peter Nguyen, Yessenia Magaña, Sayyed Jilani, Byung Kook Lim & Takaki Komiyama

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

    Nathan G. Hedrick, Zhongmin Lu, Surbhi Singhi, Peter Nguyen, Yessenia Magaña, Sayyed Jilani, Byung Kook Lim & Takaki Komiyama

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

    Nathan G. Hedrick, Zhongmin Lu, Eric Bushong, Surbhi Singhi, Peter Nguyen, Yessenia Magaña, Sayyed Jilani, Mark Ellisman & Takaki Komiyama

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

    Nathan G. Hedrick, Zhongmin Lu, Surbhi Singhi, Peter Nguyen, Yessenia Magaña, Sayyed Jilani & Takaki Komiyama

  5. Center for Research in Biological Systems, University of California, San Diego, School of Medicine, La Jolla, CA, USA

    Eric Bushong & Mark Ellisman

  6. National Center for Microscopy and Imaging Research, University of California, San Diego, La Jolla, CA, USA

    Eric Bushong & Mark Ellisman

Authors
  1. Nathan G. Hedrick
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Zhongmin Lu
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Eric Bushong
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Surbhi Singhi
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Peter Nguyen
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Yessenia Magaña
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Sayyed Jilani
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Byung Kook Lim
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Mark Ellisman
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Takaki Komiyama
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

This work was conceived by N.G.H. and T.K. All in vivo experiments were performed by N.G.H. and Z.L. with assistance from Y.M. and S.J. and analyzed by N.G.H. and T.K. CLEM was performed by N.G.H, Z.L., E.B. and M.E., and analyzed by N.G.H., S.S. and P.N. B.K.L. provided viral vectors. The paper was written by N.G.H. and T.K. with inputs from all authors.

Corresponding authors

Correspondence to Nathan G. Hedrick or Takaki Komiyama.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Neuroscience thanks Carlos Portera-Cailliau and the other, anonymous, reviewer(s) 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 Experimental setup, learning metrics, and image quality assurance.

(a,b) Task performance improves over days of training. (a) The percentage of trials resulting in reward significantly increases over learning (p = 2e-31; Pearson’s correlation coefficient). Data points correspond to the mean fraction of successful trials ± SEM. **** p < 0.001. n = 53 mice. (b) The reaction time (time from cue onset to movement onset; black) as well as the time from cue to reward (red) and movement onset to reward significantly decrease over learning (p = 8e-35 for cue-to-reward; p = 7e-50 for cue-to-movement; p = 6e-4 for movement onset to reward, Pearson’s correlation coefficient). Data points correspond to mean time ± SEM. **** p < 0.001. n = 53 mice. (c) Schematic of imaging schedule. In each animal, three dendritic fields were imaged on each of the first three training days (Imaging Session 1, IS1), after which each field was imaged in 5-day intervals for two additional imaging sessions (IS2 and IS3). This yielded three imaging sessions for each field: early, middle, and late sessions. (d) Spine density does not significantly change over learning (1-way ANOVA, main effect of session: F = 0.94, d.f. = 2, p = 0.39). Individual dendrites are plotted as colored dots; black line corresponds to the median spine density ± bootstrapped 95% CI.n = 76 dendrites / 23 mice.(e) Spine event frequency is stable over learning for MRS (p = 0.46, rank-sum test), and decreases for nonMRSs (p = 1e-5, rank-sum test). Spine event frequencies are higher for MRSs in late sessions (early: p = 0.17, late: p = 7e-7). Bars represent median ± bootstrapped 95% CI. n = 898 early MRSs, 820 late MRSs, 869 early nonMRSs, 788 late nonMRS. Multiple comparisons were corrected using the FDR method. (f) Example lever position trace, showing movement periods in green shadings. (g) Example of frame-wise 2d correlation with the corresponding calculated reference image (that is, an iteratively aligned and average-projected version of the imaging field) associated with the behavioral window shown in (d). Prior to motion correction (red line), frame correlations with the reference image show with a decrease during movements. After motion correction, mean frame correlations are higher regardless of movements (blue line). (h) Example x-y pixel shifts used for motion correction of each frame in the behavioral window shown in (f,g), showing that extra correction during movements successfully compensated for movement artifacts. (i) Average projection images from the ~4 s behavioral window indicated in (g). Prior tomotion correction, images are blurry and individual structures are difficult to resolve because of misalignment across frames (top). After motion correction, images are sharp, and individual spines are visible (bottom). (j) Motion correction generates stable frames over movement. When aligned to movement onset, frame correlations with the reference image (as in (e)) decrease sharply during movements in the raw, pre-correction images (red line), but are stable during movements for the post-correction frames (blue line). Curves represent mean change in reference imaged correlation. Shaded portions correspond to SEM.

Extended Data Fig. 2 Workflow of in vivo imaging followed by correlated ex vivo electron microscopy.

(a) Workflow of correlated light and electron microscopy (CLEM) to identify the in vivo- imaged dendrites for EM. From top left moving clockwise: in vivo images of dendrites imaged during an experiment are acquired a final time for reference of the most updated structural information. Vascular maps are then generated by 2p in vivo imaging of TRITC-Dextran (injected retro-orbitally on final day of in vivo imaging), capturing an ~200𝜇m z-stack in the area surrounding the dendrite of interest for detailed structural landmarks (panel 2 and white square in panel 3) and a low-zoom (usually 1 zoom/ ~1000𝜇m2, panel 3) to provide the larger structural context with respect to the vasculature. Arrows indicate the vasculature used for aligning the subsequent images. After the animal is perfused with fixative, thin sections are cut, and bright-field imaging is used to find the vasculature and approximate the location of the dendrite (white square in panel 4). Confocal images are then acquired using the slice containing the target vasculature so as to ensure the location of the target dendrite (white square in panel 5). The slice of interest is then processed for EM, thereby permanently rendering the tissue photo-inaccessible. The tissue is then exposed to iterative steps of X-Ray tomography and trimming (guided by software-based non-rigid alignment) so as to isolate the dendrite within a small volume amenable to EM. Shown is an alignment of the confocal stack with the x-ray tomogram. Colored arrowheads point to the dendrites of interest identified in subsequent images. Serial-section scanning-blockface EM is then performed on the final tissue block, and the confocal images are aligned to the output using the accumulated fiducial structures over previous imaging modalities. The final result (panel 8) is an aligned stack of confocal fluorescence and EM data, allowing a fluorescence overlay onto the EM stack. Using this alignment, the target dendrite and constituent spines are then reconstructed using the in vivo 2p image as reference for the domains of interest. These experiments were repeated independently in n = 4 mice with comparable results (shown below). (b) After reconstruction of the target dendrite(s) and spines (left), the synaptically connected axonal partner of each spine was traced to the full extent of the imaging volume or to the furthest point of confident identification (right). In this volume, 59 axons were successfully reconstructed to various extents. (c) Results of all remaining CLEM experiments from the current study. Left, in vivo 2-photon average projections, aligned EM data, and the corresponding reconstruction of the target dendrites are shown. Reconstruction images are angled so as to accentuate the correspondence between 2p and EM. Middle, zoomed-in visualizations of the reconstructions of target dendrites prior to reconstruction of synaptically connected axons. The number of axons reconstructed in each volume were: 137 (top), 74 (middle), and 60 (bottom), for a total of 330 axons (including the 59 axons from (b)). Note that the axonal partners of spines near the ends of dendrites were often not reconstructed due to their proximity to the edge of the sample volume.

Extended Data Fig. 3 Characterization of movement-related signals at single dendritic spines.

Schematic of MRS definition. The dot product of the ΔF/F0 trace with binarized movement traces defines the movement score, x, for a given spine. Movements are then shuffled in time 1000x (without breaking individual movements), and the corresponding shuffled scores calculated. The real and shuffled scores are then ranked together. Those spines whose movement score is above the 97.5th percentile are considered MRS. Example traces of movement onset-aligned activity for 12 individual MRSs (green, top) and 12 new spines (blue, bottom). Movement onset is indicated by black vertical dashed line. Activity onset timing is indicated with red triangles. Mean ± SEM. Explanation of event detection based on previously published9,16 methods. Left, full ΔF/F0 trace for an example spine. For each trace, noise is estimated based on the negative portion of the trace, reflected about the x-axis. Values above 1x the standard deviation of this trace are considered above the ‘baseline’ threshold (light blue), and values above 2x the standard deviation are considered above the ‘active’ threshold (red). Green rectangle indicates portion of the trace inspected in right. Upper right, zoomed-in section of the trace for easier visualization. Tan region indicates example region used to demonstrate end and start times of active periods. End times (red asterisk) are defined as when a smoothed version of the trace falls below the active threshold, and start times (green asterisk) are defined by moving backwards in time from each end point to the first point where the raw trace falls below the baseline threshold. All frames meeting these criteria are assigned a value of 1, and all other frames a value of 0. In this way, events are considered continuous unless they fall below the baseline threshold. The resulting binarized trace (bottom right) indicates when spines are active and inactive. Relationship of spine active periods to movements. Using the binarized activity described in (c), the relationship of spines’ activity can be shown as either the fraction of movements during which a given spine is active (left) or the fraction of activity for a given spine that occurs during movements (right). Spines are subdivided into MRSs and nonMRS, as described in (a). Bars correspond to the medians ± 95% CI. MRSs are active during a higher fraction of movements than nonMRSs in both early (p = 8e-52; rank-sum test) and late (p = 2e-70; rank-sum test) sessions, and while MRSs are active during a similar fraction of movements across sessions (p = 0.95; rank-sum test), nonMRSs show a significant decrease (p = 5e-7; rank-sum test). Similarly, a higher fraction of MRS activity occurs during movements than nonMRSs in both early (p = 1e-72; rank-sum test) and late (p = 4e-93; rank-sum test) sessions, though both MRSs (p = 9e-101; rank-sum test) and nonMRS (p = 4e-105; rank-sum test) show an increased fraction of activity occurring during movements by late sessions. All comparisons were two-sided. Multiple comparisons were corrected using the FDR method. **** p < 0.001. The co-activity rates of simulated spine pairs (Methods) correlate better with the geometric mean of the constituent spine pairs’ event frequencies than the arithmetic mean, demonstrating that the geometric mean is a more suitable normalization factor for co-activity rates. Overall statistics for arithmetic mean: r = 0.91, p < 0.0001; geometric mean: r = 0.95, p < 0.0001. The r-value was higher for the fit between geometric mean and co-activity rates in 1000/1000 shuffles.

Extended Data Fig. 4 Spine identification and CLEM-based corroboration.

Example of ROIs. Top, in vivo image under consideration. Bottom, elliptical ROIs manually drawn in the initial analysis of this dendrite. Magenta ellipses correspond to ROIs that were successfully located in EM, while blue ellipses show those that were not confidently identified. Two example regions are indicated for closer inspection in (b). (b) Zoomed-in regions shown in (a) for closer inspection of ROI drawing. Top row: deconvolved in vivo images of the two regions shown in (a). Deconvolved images were used throughout this study to aid in spine identification. Second row: corresponding EM reconstructions showing the same spines imaged in vivo. Third row: reconstructions of the associated regions of dendrite, but with post-synaptic density (PSD) reconstructions shown in red. Fourth row: example EM micrographs demonstrating PSDs on a selection of spines (numbers correspond to spines in above image). Scale bars in the fourth row (green) correspond to 0.5 µm. (c) Summary of the fraction of ROIs identified in vivo that were successfully located in EM. The vast majority of structures drawn as spines in vivo were also found in EM. Of those structures that were not located, most were due to technical failures in EM acquisition (1 of the 16 of such failures was due to debris from a previous slice obfuscating the region of the target dendrite, and 15 were due to slices that were skipped, likely due to uneven cutting). We also identified one highly localized dendritic region that appeared damaged/blebbed despite a healthy appearance in vivo, preventing the assessment of two spines observed in vivo. The four remaining ROIs were not located due to unknown reasons, suggesting either mislabeling or elimination of these spines between the final in vivo imaging session and EM processing (see a potential example of this in the blue-encircled spine in (b). (d) Summary of the fraction of all spines identified in EM (within the dendrites captured in vivo) that were also labeled as ROIs in vivo. The majority of spines that were not identified in vivo were co-axial with either another (usually larger, see (e)) spine (red slice), or the dendritic branch (purple slice). A small subset of spines located in EM were obscured by another fluorescent structure (for example, a labeled axon) in vivo (blue slice). The remaining spines were not identified in vivo for unknown reasons, which might include the rapid formation of spines between the final imaging session and EM processing, or most of these spines being below the detectable size for in vivo imaging (see (e)). (e) Summary histograms of the volumes (as calculated in EM) of spines that were also captured in vivo (green) and spines that were not identified in vivo (gray). The vast majority of spines that were identified in EM but not labeled in vivo were small in size; 67% of ‘missed’ spines were within the 3rd percentile of the ‘captured’ population. The medians of the ‘missed’ and ‘captured’ groups are indicated with black and green arrows, respectively.

Extended Data Fig. 5 Analysis of no-task controls.

(a) Schematic of ‘no-task’ condition compared to the typical ‘learning’ condition. Unlike in the learning condition, the no-task condition administers water rewards at the end of each cue period, regardless of the movement of the lever. Mice were head-fixed in the experimental apparatus only during the 9 imaging sessions. (b) Heat map of lever press correlations across sessions of the ‘no-task’ experiment. Black portions correspond to days during which mice were not placed in the experimental apparatus. n = 14 mice. (c) Lever movement correlations, both within and across adjacent sessions, did not increase over time in the no-task controls. Significance of the relationship was evaluated with Pearson’s correlation coefficient, with p-values shown in figure text. Data points correspond to mean ± SEM correlation values for each session. n = 14 mice. (d) Activity event frequency of different spine types for the no-task condition. In both early and late sessions, MRSs and nonMRSs show comparable event frequencies (early: p = 0.07, late: p = 0.4, rank-sum test). MRSs and nonMRS slightly decrease (MRSs: p = 0.0035, nonMRSs: p = 9e-8) their activity event frequencies over the experiment. Multiple comparisons corrected using the FDR method. When pooled with the learning-group data, only late MRSs are significantly higher in the learning group than in the no-task control after correcting for multiple comparisons (p = 0.006, rank-sum test). Conversely, nonMRS showed higher event frequencies in both early and late sessions for the no-task group (early: p = 2e-4, late: p 0.002, rank-sum test). n = 433 early MRSs, 1270 early nonMRSs, 506 late MRSs, 918 late nonMRSs. Bars represent median ± 95% CI. Individual data points (spines) shown as black dots. (e) MRS density was lower in the no-task group (0.21/μm, 95% CI = [0.17, 0.25]) than the learning group ((0.34/μm, 95% CI = [0.29, 0.38]; p = 4e-4, rank-sum test). Bars represent median values ± 95% CI. Circles correspond to individual dendrites, color-coded by animal. Note that all imaged dendrites are shown, including those that show no new spine formation. n = 40 no-task group dendrites, 76 learning-group dendrites. (f) The new spine density on imaged dendrites was lower in the no-task group (0.02/μm, 95% CI = [0.010 0.027]) than the learning group (0.039/μm, 95% CI = [0.022, 0.054]) (p = 0.04, rank-sum test). Bars represent median values. Circles correspond to individual dendrites, color-coded by animal. Note that all imaged dendrites are shown, including those that show no new spine formation. n = 40 no-task control dendrites, 76 learning-group dendrites. (g) A higher fraction of new spines was transient in the control group vs. learning group (p = 0.02, Chi-squared test). (h) Functional clustering was intact in the no-task condition, but co-activity was generally less pronounced. Much like the learning condition, MRSs showed higher overall co-activity levels than nonMRSs (3-way ANOVA [MRS status × distance × session], main effect of MRS status: F = 976.58, d.f. = 1, p = 2e-211; post-hoc using least-significant difference: p < 0.0001 for MRS vs. nonMRS across all distance bins when collapsing across sessions or when preserving session identity). We observed a main effect of distance (F = 109.38, d.f. = 5, p = 5e-211), and the first distance bin was higher than all other bins (p = 1e-30, post-hoc using LSD), indicating that functional clustering is present. By comparing to the learning-group data (Fig. 1j), we find that overall co-activity values were higher in the learning group (4-way ANOVA [MRS status × distance × session × condition], main effect of condition, p = 6e-54; post-hoc using LSD). We also observe a significant MRS status × condition interaction (F = 106.27, d.f. = 1, p = 7e-25), with post-hoc analysis revealing that both MRSs (p = 2e-57) and nonMRS (p = 1e = 5) show higher co-activity rates in the learning condition than their no-task counterparts. n = 6376 early MRS pairs, 49894 early nonMRS pairs, 8088 late MRS pairs, 47274 late nonMRS pairs. Y-axis scale was set to match that of the learning group (Fig. 1j). Mean ± SEM. (i) Neither MRS density (r = 0.03, p = 0.64) nor nonMRS density (r = −0.10, p = 0.16, Pearson’s correlation coefficient) changed as a function of distance to the nearest new spine in the no-task controls. Mean ± SEM. (j) New spines formed during the no-task control are weakly functionally clustered with MRSs. Like in the learning condition, new spines formed in the no-task condition show slightly higher overall co-activity rates with MRSs vs. nonMRSs (2-way ANOVA [MRS status × distance], main effect of MRS status: F = 4.2, d.f. = 1, p = 0.04) and closer spine pairs generally had higher co-activity (main effect of distance: F = 2.3, d.f. = 5, p = 0.04). No significant interaction term was observed. When comparing to the learning condition data (Fig. 4c), we observed a significant main effect of MRS status and distance (3-way ANOVA [MRS status × distance × condition], MRS status: F = 22.85, d.f. = 1, p = 2e-6; distance: F = 14.87, d.f. = 5, p = 2e-14), but not condition (that is no-task vs. learning). However, we found a significant interaction between distance and condition (F = 5.66, d.f. = 5, p = 3e-5), suggesting that the spatial relationship of co-activity rates differs across the two conditions. Indeed, post-hoc comparisons show that only the learning condition shows significant functional clustering, such that the first distance bin is greater than all other distance bins (p < 0.0001 for learning condition at distance = 0–5um vs. all other bins; p > 0.05 for no-task condition at distance = 0–5um vs. all other bins). n = 789 new spine-MRS pairs, 2029 new spine-nonMRS pairs. Y-axis scale was set to match that of the learning group (Fig. 3c). Mean ± SEM. (k) Like in the learning condition, new spine-MRS pairs show activity that is enriched during movements, more so than new spine activity alone (p = 7e-5; sign-rank test) or MRS activity alone (p = 0.037; sign-rank test). Multiple comparisons were corrected for using FDR. n = 21 fields showing new spine-MRS co-activity / 25 total fields. Bars represent median ± 95% CI. (l) Movements containing new spine-MRS co-activity are not more correlated with the learned movement pattern (LMP) than movements lacking such activity (p = 0.43, sign-rank test). Data points correspond to the median correlation of movements occurring with (blue) or without (red) co-activity of new spine-MRS pairs in a given field. n = 18 fields showing new spine-MRS co-activity during at least 3 movements / 21 fields showing new spine-MRS co-activity / 25 total fields. (m) The functional separation of sustained new spine-MRS pairs from other pair types is conspicuously absent. New spine-MRS pairs, irrespective of survival, showed slightly higher overall co-activity rates (3-way ANOVA [MRS status × distance × survival], main effect of MRS: p = 0.001), and spine pairs were generally functionally clustered (main effect of distance, p = 0.02), but no interaction terms were significant. Notably, we observed a main effect of new spine survival (p = 0.01), but post-hoc inspection revealed that transient new spine-MRS co-activity was actually slightly higher than the other groups (p < 0.02 for all comparisons). Mean ± SEM.

Extended Data Fig. 6 Additional spatial analyses of dendritic structural and functional features with respect to new spines.

(a) Overall spine density as a function of distance from new spines. While there is a trend towards higher spine densities closer to new spines, the effect is not significant (Pearson’s correlation coefficient; r = −0.10, p = 0.13). Data points correspond to the mean spine density ± SEM of all dendrites displaying new spine formation (n = 50 dendrites across 21 mice). (b) A large fraction of new spine-MRS co-activity occurs during movements regardless of the distance between the spines (ANOVA with post-hoc test using the least-significant difference; main effect of distance p = 0.11). Data points correspond to the mean ± SEM of all new spine-MRS pairs (n = 1658 pairs across 50 dendrites from 21 mice). (c) Movements coincident with new spine-MRS coactivity have similarly high correlations with the learned movement pattern regardless of the distance between the spines (ANOVA with post-hoc test using the least-significant difference; main effect of distance: p = 0.85). Data points correspond to the mean (± SEM) correlation of movements coincident with co-activity of a given new spine-MRS pair (n = 1658 pairs / 50 dendrites / 21 mice). (d) The fraction of movements coincident with co-activity of a particular new spine-MRS pair shows a significant decrease with increasing distance between the spines, illustrating that a higher number of movements is encoded by closer spine pairs (ANOVA with post-hoc test using the least-significant different; main effect of distance: p = 0.9e-05; p < 0.02 for post-hoc comparison of first distance bin vs. all other distance bins). Mean ± SEM. (n = 1658 pairs / 50 dendrites / 21 mice).

Extended Data Fig. 7 Supporting evidence of in vivo spine volume estimates using iGluSnFR.

Effect of different enlargement threshold cutoff values (from 1.1, light green, to 2, magenta) on the relationship between the probability of MRS enlargement and distance to the nearest new spine. Data points correspond to the mean probability of spine enlargement (± SEM) for MRSs in a given distance bin for each new spine imaged. n = 118 new spines, 697 MRSs present on 50 new-spine-containing dendrites across 21 mice. Summary of Pearson’s statistics for data shown in (a). The relationship is significant for threshold values up to 1.5-fold change in spine volume. Data points correspond to either the calculated Pearson’s correlation coefficient (black, left axis), or the corresponding p-value (red, right axis) calculated from statistical tests on the data groups shown in (a). (c,d) Relationship between estimated spine volume and spine event frequency in early (c) and late (d) sessions, revealing a lack of positive correlation, suggesting that our methods using iGluSnFR fluorescence do not overestimate the volume of highly active spines. Linear fit of data shown in red. Significance of the relationship was determined from Pearson’s correlation coefficient, with r and p-values shown in figure text. **** p < 0.001, ns: not significant. (e) Evaluation of the effect of removing active periods from image projections prior to calculation of spine volume. In 11 of the 23 ‘learning group’ animals (769 spines), spine volume was re-calculated after removal of all frames in which a given spine was considered ‘active’. ‘Activity-removed’ spine volume estimates correlate strongly and significantly with spine volume estimates from full-length time series projections. Unity line shown in dashed blue. Linear fit of data shown in red. Significance of the relationship was determined from Pearson’s correlation coefficient, with r and p-values shown in figure text. **** p < 0.001 (f) Example in vivo, EM, and reconstruction images illustrating the presence of spine apparati, a signature of mature and potentiated spines, in spines that showed enlargement in vivo during learning. Green arrows indicate spines of interest. Each spine of interest in the in vivo images is outlined for spine size comparisons (red = early; green = late), and the early and late outlines are overlaid in the bottom left corner of each late session. Red arrows in the EM images indicate spine apparati. Reconstructed spine apparati are shown as yellow structures within spines (magenta).

Extended Data Fig. 8 Cases of axon-sharing between spines.

Two additional example cases of axon sharing between spines on imaged dendrites. The left-most column shows reconstructions of the dendrites and its spines along with the axon being shared between two spines. Each synapse occurring with the dendrite of interest is circled, and corresponding EM images illustrating synaptic contact are shown on the right (further indicated with red arrows). The top example shows a double connection between two pre-existing spines immediately after a dendritic branch point. The bottom example shows a double connection formed by a new spine and a nearby pre-existing spine.

Extended Data Fig. 9 Limiting analysis to rewarded movement-related spines (rMRSs) yields reproduces main findings.

Pre-existing rMRSs show strong functional clustering. Data points correspond to mean ± SEM. n = 443 early rMRSs / 2052 condendritic rMRS-rMRS pairs; 1472 early nonrMRS-nonrMRS pairs; 571 late rMRSs / 3278 late nonrMRS-nonrMRS pairs. (b) rMRS density nearby (≤10μm) new spines is higher than expected by chance, as estimated by randomizing the location of new spines. Statistical significance was determined by a one-sided comparison of the median of each shuffle across all new spines to the real data median; the resulting fraction of shuffles that are consistent with the null hypothesis (that is that there is not a higher nearby MRS density than expected by chance) corresponds to the p value; p < 0.001. Inset, median nearby rMRS density of all new spines compared to the median chance estimate. Error bars correspond to bootstrapped 95% CI. n = 118 new spines across 50 dendrites. (c) rMRS density decreases as a function of distance from the nearest new spines, whereas nonrMRS density does not. MRS density: Spearman’s r = −0.16, p = 0.01; nonMRS density: Spearman’s r = 0.03, p = 0.67). Data points correspond to mean ± SEM. n = 118 new spines across 50 dendrites. (d) rMRS enlargement probability decreases as a function of distance from the nearest new spine. Spearman’s r = −0.11, p = 0.048). Data points correspond to mean ± SEM. n = 118 new spines; 443 rMRSs present on 50 new spine-containing dendrites. (e) New spines are functionally clustered with rMRSs in a manner similar to MRSs. Data points correspond to mean ± SEM. n = 1091 new spine-rMRS pairs; 2774 new spine-nonrMRS pairs / 118 new spines across 50 dendrites / 21 mice. (f) The fraction of activity of new spine-rMRS pairs occurring during movements trends higher than either new spine activity alone (p = 1e-6) or rMRS activity alone (p = 0.1; sign-rank rest correcting for multiple comparisons using the FDR method). n = 34 fields. Bars correspond to the median fraction of cases for each group present in each of the 34 fields imaged. Error bars represent bootstrapped 95% CI. (g) Rewarded movements associated with new spine-rMRS co-activity are not more correlated with the learned movement pattern (LMP) than rewarded movements lacking such activity (p = 0.14). Thus, it appears that preferential occurrence of co-activity during rewarded movements lead to the similarity of movements to LMP shown in Fig. 4g. (h) The survival of new spines can be predicted by co-activity with nearby rMRSs. New spine-rMRS pairs showed higher overall co-activity rates than other spine pairs (3-way ANOVA [rMRS status × distance × survival], main effect of rMRS status: p = 3e-11), and co-activity values were significantly impacted by distance (main effect of distance, p = 0.0006). We observed a main effect of new spine survival (p = 0.007), and trends for interaction terms distance × survival (p = 0.055) and rMRS × distance × survival (p = 0.058). Given the trend in the three-way interaction term, we performed post-hoc comparisons of the four groups across each distance bin. Sustained new spine-rMRS co-activity at close distances (within 5μm) was significantly higher in 17 of the 18 possible comparisons against the other three groups (p values range from 7e-12 to 0.02 for the 17 significant comparisons, p = 0.47 for the final comparison), and was higher than values within the same group at longer distances (>15, p < 0.008). Conversely, transient new spines show co-activity that is not higher at closer distances (transient new spine-MRS pairs’ co-activity within 5μm is not different than any other distance bin (p values range from 0.27 to 0.73); transient new spine-nonMRS pairs’ show co-activity that is significantly higher (p = 0.02) than one other bin and significantly lower (p = 0.046) than another bin), suggesting that no significant functional clustering is present between transient new spines and other spines. Further, neither transient spine group showed a significant correlation with distance (transient new spines-MRS co-activity vs. distance, r = 0.06, p = 0.39; transient new spines-nonMRS co-activity vs. distance: r = −0.07, p = 0.14), while sustained new spine co-activity is significantly negatively correlated with distance (sustained new spine-MRS co-activity vs. distance: r = −0.18, p = 0.0001; sustained new spine-nonMRS co-activity vs. distance: r = −0.14, p = 6e-5). Thus, surviving new spines’ co-activity with rMRSs is higher and more spatially selective than transient new spines. n = 28 transient new spines, 230 transient new spine-rMRS pairs, 669 transient new spine-nonrMRS pairs; 57 sustained middle-session new spines, 530 sustained new spine-rMRS pairs, 1259 sustained new spine-nonrMRS pairs. Mean ± SEM.

Extended Data Fig. 10 Main findings are robust against changes in the threshold for MRS definition.

Shifting the threshold for defining MRSs to being greater than either the 95th percentile of shuffles or the 99.5th percentile of shuffles has negligible effects on the main findings. (a) Functional clustering of pre-existing spines is similar when using both 95th and 99.5th percentiles to define MRSs. 95th percentile cutoff: MRSs are still more co-active than nonMRS (3-way ANOVA, main effect of MRS status: F = 1092.04, d.f. = 1, p = 2e-233), and spines show distance-dependent co-activity rates (main effect of distance: F = 132.07, d.f. = 6, p = 3e-138), such that closer pairs are more co-active (Pearson’s correlation coefficient, r values range from 0.14–0.22 and p < 0.0001 for all 8 separate groups shown). More conservative cutoff results in higher co-activity rates of MRS pairs. n = 687 early MRSs / 4318 early MRS-MRS pairs / 1080 early nonMRSs / 13621 early nonMRS-nonMRS pairs / 685 late MRSs / 4687 late MRS-MRS; 1230 late nonMRS / 12318 late nonMRS-nonMRS pairs at the 99.5th percentile; n = 954 early MRSs / 7719 early MRS-MRS pairs / 813 early nonMRSs / 8534 early nonMRS-nonMRS pairs / 868 late MRSs / 7313 late MRS-MRS pairs / 788 late nonMRSs / 8967 late nonMRS-nonMRS pairs. (b) The higher density of nearby MRSs is robust to the different MRS thresholds. Main plots show cumulative probability distributions of the nearby MRS density for real new spines (light blue) and chance, estimated by shuffling new spine locations 1000 times for each new spine (250 individual shuffles shown in gray, median of all shuffles shown in black). Statistical significance was determined by comparing the median of each shuffle across all new spines to the real data median; the resulting fraction of shuffles that are consistent with the null hypothesis (that is that there is not a higher nearby MRS density than expected by chance) corresponds to the p value. 95th percentile MRS density: p = 0.003. 99.5th percentile MRS density: p = 0.012. n = 118 new spines across 50 dendrites. (c) The distance-dependent decrease in MRS density is similar across different thresholds. 95th percentile: MRSs: Spearman’s rank coefficient, r = −0.12, p = 0.045, nonMRS: r = 0.07, p = 0.212. MRSs show a significantly different slope than nonMRS (p < 0.001). 99.5th percentile: MRSs: r = −0.11, p = 0.052. n = 1390 new spine-MRS pairs / 2475 new spine-nonMRS pairs at 99.5th percentile; 1731 new spine-MRS pairs / 2131 new spine-nonMRS pairs at 95th percentile. (d) Volume increase probability with respect to new spine location is similar across different MRS thresholds. 95th percentile cutoff: r = −0.07, p = 0.062 (Pearson’s correlation coefficient. 99.5th percentile cutoff: r = −0.−8, p = 0.06 (Pearson’s correlation coefficient). n = 1390 new spine-MRS pairs / 2475 new spine-nonMRS pairs at 99.5th percentile; 1734 new spine-MRS pairs / 2131 new spine-nonMRS pairs at 95th percentile. (e) New spine functional clustering is nearly identical across different thresholds. 95th percentile MRSs: New spine-MRS pairs are more co-active than new spine-nonMRS pairs (2-way ANOVA, main effect of MRS status: F = 18.71, d.f. = 1, p = 2e-5). New spine co-activity generally shows a distance-dependent change in co-activity rates (main effect of distance: F = 20.97, d.f. = 6, p = 1e-20), an effect that did not depend on MRS status, (MRS status by distance interaction: F = 1.5, d.f. = 5, p = 0.19). Both groups showed higher co-activity at the 0–5μm bin than all other bins (post-hoc using LSD: p < 0.05 for all). n = 1390 new spine-MRS pairs / 2475 new spine-nonMRS pairs at 99.5th percentile; 1734 new spine-MRS pairs / 2131 new spine-nonMRS pairs at 95th percentile. (f) The functional separation of transient and sustained new spines with respect to their co-activity with nearby spines is comparable for both thresholds. 95th percentile cutoff: as in other conditions, new spine-MRS pairs are more co-active than new spine-nonMRS pairs (3-way ANOVA, main effect of MRS status: F = 24.91, d.f. = 1, p = 7e-7), and pairs are differentially co-active at closer distances (main effect of distance: F = 5.77, d.f. = 5, p = 3e-5). Only surviving new spines show a significant negative correlation with distance (surviving new spine-MRS vs. distance, r = −0.16, p = 2e-5 [Spearman’s: r = −0.13, p < 0.001]; surviving new spine-nonMRS vs distance, r = −0.13, p = 0.003 [Spearman’s r = 0- 0.13, p = 0.003]; transient new spine-MRS co-activity vs distance, r = −0.07, p = 0.2 [Spearman’s r = −0.09, p = 0.13], transient new spine-nonMRS co-activity vs. distance: r = −0.04, p = 0.54 [Spearman’s r = −0.06, p = 0.38], Pearson’s correlation coefficient). n = 339 transient new spine-MRS pairs / 560 transient new spine-nonMRS pairs / 618 sustained new spine-MRS pairs / 1171 sustained new spine-nonMRS pairs at 99.5th percentile; 451 transient new spine-MRS pairs / 448 transient new spine-nonMRS pairs / 833 sustained new spine-MRS pairs / 956 sustained new spine-nonMRS pairs at 95th percentile. (g) The fraction of activity occurring during movements is highly similar for different MRS thresholds. The fraction of new spine-MRS co-activity events occurring during movements is higher than the fraction of new spine-only (95th: p = 1e-6; 99.5th: p = 7e-5) and MRS-only (95th: p = 0.003; 99.5th: p = 2e-4, rank-sum test) events. n = 34 fields. Median ± 95% CI. (h) The correlation of movements coinciding with new spine-MRS co-activity with the learned movement pattern (LMP) is robust to MRS threshold. Both thresholds yield MRSs that, when co-active with new spines, signal movements more similar to the LMP than movements lacking such activity (95th: p = 0.002, 99.5th: p = 0.0009, sign-rank test). n = 34 fields. Individual data points correspond to the median correlation of all movements concurrent with co-activity (blue) or lacking such activity (red) of each imaged field. Overall median of each group shown in color-coded line.

Supplementary information

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Hedrick, N.G., Lu, Z., Bushong, E. et al. Learning binds new inputs into functional synaptic clusters via spinogenesis. Nat Neurosci 25, 726–737 (2022). https://doi.org/10.1038/s41593-022-01086-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41593-022-01086-6

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